article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the dicke model@xcite describes the interaction of @xmath1 two - level systems with bosonic field modes . in the thermodynamic limit , and for a single bosonic mode
, it undergoes a quantum phase transition ( qpt ) from a normal to a superradiant phase when the atom - field coupling strength exceeds a critical value@xcite .
the model has been experimentally realised by an bose - einstein - condensate trapped in an optical cavity@xcite .
dissipation due to photon emission has furthermore been modeled , e.g. by heisenberg - langevin equations @xcite , the keldysh approach @xcite or by use of hartree - fock - bogoliubov theory @xcite . in this paper , we study the full photon counting statistics of the driven dissipative single - mode dicke model by including a counting field @xmath2 in the master equation , which we analytically solve in the @xmath0-representation in the thermodynamic limit @xmath3 .
we obtain analytic expressions for time - dependent cumulants of the photon counting statistics , as well as the asymptotic cumulant generating function ( cgf ) .
the cgf is shown to consist of two parts : a macroscopic component , which is poissonian in nature and has a characteristic rate proportional to the mean occupation of the cavity mode ( order parameter of the system ) ; on top of this comes a contribution describing fluctuations about the mean behaviour , which has a non - trivial form and is divergent around the quantum phase transition .
in addition , we identify the three phases ( normal , superradiant , intermittent ) and the corresponding critical coupling parameters @xmath4 that characterize the dissipative phase transition . the structure of this paper is as follows : in sec .
[ ch : model ] we introduce the model and obtain the three critical coupling parameters , in sec .
[ ch : mseq_p_represent_counting ] we solve the master equation with a counting field @xmath2 , in sec .
[ ch : results ] we give an analytical expression for the cgf and in sec .
[ ch : discussion ] we discuss our results and connect them to other works .
the dicke hamiltonian ( @xmath5 ) @xmath6 describes our isolated system . here
, @xmath7 are the collective atomic angular momentum operators that describe an ensemble of @xmath1 two level atoms with a level splitting @xmath8 ; @xmath9 is the length of pseudo - spin with value @xmath10 for bosonic realisation @xcite , @xmath11 and @xmath12 are the ladder operators for the optical mode with an energy @xmath13 , and @xmath14 is the coupling strength between the optical mode and the atoms . as in refs .
@xcite we interpret eq . as an effective model in a frame rotating at the frequency of an external driving laser , @xmath15 .
we describe dissipation in this system with the master equation in lindblad - form @xcite @xmath16 \notag \\ & \quad - \frac{\gamma}{2 }
\biggl { [ } a_1^\dag a_1 \rho(t ) - 2 a_1 \rho(t ) a_1^\dag + \rho(t ) a_1^\dag a_1 \biggr { ] } \,,\end{aligned}\ ] ] where @xmath17 is the density matrix of the atom - cavity system and @xmath18 is the rate of photon loss from the cavity . , the master equation is obtained from a cavity - bath coupling hamiltonian @xmath19 , in a frame rotating with driving frequency @xmath15 .
this also means , that @xmath14 is a function of @xmath15 . ]
+ to make analytical progress , we use the holstein - primakoff - transformation @xcite , which allows one to describe @xmath1-two - level - atoms by a single bosonic mode .
this is done by the transformation @xmath20 where @xmath21 and @xmath22 are the ladder operators of the introduced atomic mode . in this representation
, the dicke - hamiltonian reads @xmath23 the isolated dicke - hamiltonian @xmath24 in the thermodynamic limit @xmath3 is known to have two phases : normal and superradiant @xcite . in the normal phase and for @xmath3
only @xmath24 becomes effectively@xcite , @xmath25 which has the form of two simply - coupled harmonic oscillators .
the master equation for this phase may be obtained by simply replacing @xmath24 in eq . with this form . + in contrast , the superradiant phase is characterised by the macroscopic occupation of the @xmath11 and @xmath21 modes . in order to describe this phase
, we insert a mean - field - ansatz @xmath26 into both @xmath24 and the dissipator term of the master equation , eq . .
let us start with @xmath24 , where we insert eq . into eq . .
we then expand the square roots and neglect terms with powers of @xmath1 in the denominator @xcite .
assuming @xmath27 we obtain in the thermodynamic limit @xmath28 with @xmath29 with @xmath30 making the displacements in the dissipator , the complete master equation then reads @xmath31 \notag\\ & - i \left[h_{s}^{(2)},\rho(t)\right ] - \frac{\gamma}{2 } \left(c^\dag c \rho(t ) + \rho(t ) c^\dag c - 2 c \rho(t ) c^\dag\right)\,.\end{aligned}\ ] ] now , we first determine the macroscopic occupation @xmath32 in the superradiant phase using the master equation . in eq .
we have already separated the terms that are linear in the new operators @xmath33 .
these terms are proportional to the square root of the particle number @xmath1 , so they diverge in the thermodynamic limit .
we determine the macroscopic occupation , such that all parts @xmath34 of the master equation vanish .
that means the first commutator has to be zero which is fulfilled if @xmath35 next , we insert the @xmath36 from eq . ,
factor out the operators @xmath37 and get an expression of the form @xmath38 . to achieve the identity we set all braces to zero and
arrive at four equations .
the last two of them are equal , because we have assumed @xmath27 .
these equations can be solved ( see appendix [ cha : msgl_determin_macroskopic_parameter ] ) to yield @xmath39 with@xcite @xmath40 which differs from the isolated case by the inclusion of the loss - rate @xmath18 . for @xmath41 , @xmath42 and @xmath43 are non - zero .
the coupling @xmath44 is our first of three critical points .
it is the coupling above which the macroscopic excitation exists .
there is also one trivial solution , @xmath45 , that recovers the normal phase .
some insight into the behaviour of the system can be obtained by just considering the non - dissipative part , i.e. the hamiltonians @xmath46 and @xmath47 with displacements chosen as in the foregoing .
we study here the eigenvalues of these two hamiltonians and in the next section we will rewrite the master equation , eq . , in corresponding diagonal bases
. both hamiltonians can be diagonalised by a transformation of the following form @xcite @xmath48 where @xmath49 represents the old basis and @xmath50 , the new basis , in which the system is diagonal . in both cases ,
the diagonalising matrix has the same structure but different internal parameters , which are listed in appendix eq . - .
after diagonalisation , @xmath46 or @xmath47 assume the form @xmath51 with eigenenergies @xmath52 the forms of which are given in eqs .
, . in the normal phase ,
one of the eigenenergies @xmath53 of @xmath46 has a zero at @xmath54 with @xmath55 the vanishing eigenenergy marks the end of the normal phase @xcite .
coming from the superradiant phase , the same eigenenergy vanishes , but in this case at a critical value of @xmath56 : @xmath57 we recognize that @xmath58 and the eigenenergies in the superradiant phase also depend on the bath - coupling - parameter @xmath18 .
furthermore @xmath59 , so we obtain a gap , where the eigenenergies become complex and our analysis breaks down , see the discussion below . in fig .
[ fig1:eigenenergien_h_eff_aus_nagy_msc_mit_gap ] , the excitation energy of the closed system ( case @xmath60 ) and open system ( @xmath61 ) are plotted .
( solid lines ) . also plotted are the energies for the equilibrium ( @xmath62 ) case ( dotted lines ) .
for @xmath63 we have a normal phase , and for @xmath64 a superradiant phase . at @xmath54 and @xmath65 the @xmath66 energy vanishes . for @xmath62 , @xmath67 and @xmath58 coincide , but out of equilibrium , a gap appears .
parameters : @xmath68 ( case @xmath60 : @xmath69 ) . ]
in the following , our aim is to count how many photons are lost to the environment , and to determine the properties of the photon distribution function . to this end
, we introduce a counting field @xcite @xmath2 into the master equation @xmath70 \notag \\ & & \!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! \!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! - \frac{\gamma}{2 } \biggl { [ } a_1^\dag a_1 \rho(\chi , t ) - 2 e^{i\chi } a_1 \rho(\chi , t ) a_1^\dag + \rho(\chi , t ) a_1^\dag a_1 \biggr { ] } \!\!. \end{aligned}\ ] ] to make our calculations analytically tractable , we make a rotating wave approximation ( rwa ) .
we transform the master equation , eq .
, to a diagonal basis @xmath50 of @xmath46 in the normal phase or of @xmath47 in the superradiant phase using the relations @xmath71 where the coefficients are defined in appendix [ cha : diagonalisation_parameter ] .
next , we transform into the interaction picture and neglect all fast rotating terms such as @xmath72 for @xmath73 .
we obtain the following rwa - master equation for @xmath74 @xmath75 this master equation describes two uncoupled harmonic oscillators in interaction with two independent thermal baths at different temperatures @xcite .
we see that each exchange of quanta with these baths , either into or out , is associated with the emission a of ` physical ' photon ( i.e. one with annihilation operator @xmath11 ) . the final term in eq .
( [ eq1:msgl2_rwa_energie_nicht_null ] ) is only nonzero in the superradiant phase and represents a poissonian process with rate proportional to a macroscopic excitation @xmath76 .
now we transform the master equation into the @xmath0-representation using the following ansatz @xmath77 where @xmath78 is the eigenstate of the annihilating operator @xmath50 with eigenvalue @xmath79 . using @xmath80 and integration by parts ,
we get a partial differential equation for the quasi distribution @xmath81 , @xmath82 equation can be solved as @xmath83 if @xmath84 is fulfilled . since @xmath85 is a bi - linear operator , the solution of eq .
can be written in the following form @xmath86 substituting into @xmath87 we obtain seven coupled first order differential equations for the functions @xmath88 , see appendix eq . .
taking a displaced 2-dimensional gaussian distribution with a standard deviation @xmath89 as an initial condition , @xmath90 we can solve this system analytically . the steady state solution for @xmath91
can also be calculated and is listed in appendix eq . .
with knowledge of the @xmath0-representation of the density matrix with the counting field @xmath2 , we can now calculate the mean occupation of the original modes and the cumulant generating function for the photon counting statistics . as a final remark
, the rwa would normally not be valid exactly near the points @xmath92 , because here one of the eigenenergies @xmath93 vanishes .
terms like @xmath94 can no longer be neglected on the basis that they are quickly rotating
. however , we repeated our calculations ( @xmath95 ) in the singular coupling limit of the master equation where terms like @xmath96 approach unity and found that , for the parameters studied , no significant qualitative differences with the rwa method arose .
we first calculate the occupation number of the photonic and atomic modes . in our @xmath0-representation
, an operator average is given by @xmath97 which we use to calculate @xmath98 and @xmath99 .
we have to use the relation between the old and the diagonal basis , because the @xmath0-representation has been evaluated in the diagonal basis .
we also have to do the calculation in the interaction picture and use the rwa , as we have done in the calculation of the @xmath0-function . in both phases
the time - dependent average of the optical and atomic modes has the same structure of the form @xmath100 from which we can read off the characteristic relaxation times @xmath101 where @xmath102 is the rotation angle of the decoupling , eqs . , . as a result
, the two modes develop at two different time scales . in the following ,
we look at the long - time solution . in the normal phase , using the long - time solution of eq .
we get ( see appendix [ cha : avarage_and_p ] ) @xmath103 from which we recognize the expected divergence at @xmath54 , eq . . in the superradiant phase
, we have a finite macroscopic occupation , which is much greater that the fluctuations around it .
we obtain @xmath104 fig .
[ fig1:msgl2_rwa_npsr_stat_besetz_v3 ] shows the occupation number of the field and the atomic modes as a function of @xmath14 in the steady state and gives results that are approximately equivalent to those of nagy and co - workers @xcite .
we see that the fluctuations diverge around the phase transition , but there is again the undefined area for @xmath105 $ ] .
this area has been also seen in the energy plot . in the normal phase
the occupation of both modes decrease with smaller coupling parameter @xmath14 .
decreasing @xmath14 weakens the coupling between the optical and the field mode . for @xmath106 the atomic mode
decouples completely and only the field mode becomes damped . in the superradiant phase ,
the occupation of both modes first decreases with increasing @xmath14 but then the occupation of the field mode increases .
the increasing @xmath14 reduces the coupling @xmath107 , eq .
, which explains the decreasing .
on the other hand , the @xmath108-term in the hamiltonian then becomes dominant .
for @xmath109 the @xmath110-mode becomes approximately decoupled but squeezed because of the presence of the @xmath108-term , see eq . , which explains the increase of occupation .
that means , only the @xmath111-mode is then damped by the bath .
the macroscopic occupation @xmath32 exists only in the superradiant phase , where it is much greater than the fluctuation , and vanishes for @xmath112 .
.,title="fig : " ] + we now derive the cumulant generating function ( cgf ) @xmath113 of the photon counting statistics@xcite , defined as @xmath114 the @xmath115-th cumulant can be obtained by @xmath116 in the long - time limit , the cgf can be calculated from our @xmath0-function including a counting field @xmath2 , eq . as follows @xmath117 where @xmath118 is defined in appendix eq . and terms @xmath119
have being neglected . inserting the quantity @xmath120 from eq .
, we obtain the cgf for both phases , @xmath121 the cgf thus consists of two parts .
the first is the cgf of a poissonian process with rate @xmath122 . in the superradiant phase
, @xmath123 is macroscopic ( proportional to the number of atoms ) and this contribution dominates .
the k - th cumulants from this macroscopic contribution are simply @xmath124 . on top of this contribution , which is absent in the normal phase ( where @xmath125 ) , comes a further contribution to the cgf which arises from fluctuations about the mean - field displacements .
the cumulants due to this contribution are given by @xmath126 .
with this result we are now able to calculate all cumulants in the long - time limit .
for example , the first cumulant reads @xmath127 where @xmath128 we see that @xmath129 is proportional to the mean occupation of the optical mode that directly couples to the environment .
this fits with the experiment use of the mean macroscopic flux @xmath130 as a measure of the order parameter of the superradiant phase transition @xcite . as must be , the cumulant generating function is linear in time .
we mention that the source and deriving the master equation using @xmath24 and @xmath131 again gives a master equation for @xmath17 similar to eq .
but without let the heating terms @xmath132 . ] of the photon current is the driving frequency @xmath15 .
is macroscopic , @xmath133 is the corresponding fluctuation , eq . .
for @xmath134 the cumulants starting by the third are growing .
we can see also a tiny gap around @xmath44 corresponding to the undefined area @xmath135 of our theory .
parameters : @xmath136 . ]
but with fano factors not cumulants .
[ fig1:msgl2_rwa_fano_gegen_l ] ] in fig .
[ fig1:msgl2_rwa_kumsteigung_fano_gegen_l ] we have plotted the first five cumulants divided by time as a function of the coupling parameter @xmath14 .
we see all cumulants diverging near the phase transition @xmath137 , where one eigenvalue of the hamiltonian becomes zero .
going away from the critical region , our system emits less to the environment and therefore @xmath138 decreases . for
@xmath109 , @xmath139 become constant , however , higher cumulants such as @xmath140 grow .
however , in the superradiant phase all fluctuations can be neglected because of the existence of a macroscopic poisson distribution @xmath141 ( note the scaling @xmath142 in fig .
[ fig1:msgl2_rwa_kumsteigung_fano_gegen_l ] ) , which is much greater than the fluctuation .
we see again the undefined area around @xmath44 .
[ fig1:msgl2_rwa_fano_gegen_l ] shows the fluctuation part of the fano factors @xmath143 .
if we use the complete solution of eq . , we can calculate the cumulants as a function of time @xmath144 .
the evolution of the first cumulants is shown in fig .
[ fig1:kumulant_time_development ] .
we see that there are at least two time scales .
this behaviour is due to the two timescales @xmath145 and @xmath146 , eq .
, of the both modes .
as we can see in the same picture , @xmath98 develops in an analogous fashion . .
after some characteristic time scale , @xmath98 approximately reaches the steady state value , and on the same time scale the cumulants change their slope .
parameters : @xmath147 . ]
we have derived the cumulant generating function for a driven single - mode dicke hamiltonian coupled to a bath .
the combination of the thermodynamic limit and the use of the @xmath0 representation allows us to obtain exact expressions not only for the asymptotic cgf and cumulants , but also those at finite times . the cgf and photon - counting cumulants consist of two parts : a macroscopic ( order @xmath148 ) contribution and fluctuations around it ( order @xmath149 ) .
although the fluctuation component would be masked by the macroscopic contribution , this latter is only non - zero in the superradiant phase , such that the fluctuation - component could be experimentally accessed in the normal phase .
we have identified three critical values @xmath4 for the coupling constant @xmath14 . the value @xmath150 , eq .
, marks the phase transition , the area @xmath41 has a non - vanishing macroscopic mode occupation @xmath151 , eq . .
both the value of @xmath150 and of @xmath151 coincide with results in the literature @xcite .
the expression for @xmath67 and @xmath58 also agree with previous calculations based an equations of motions @xcite , however , only up to the first order in @xmath18 ( which is a small correction anyway ) .
for example , our result + \gamma a_1 \rho a_1^\dag$ ] with non - hermitian @xmath152 .
one of its eigenvalues has a zero at @xmath153 .
its real part for @xmath154 @xmath155 corresponds to , but only up to a factor 2 in the @xmath156 term . ] for @xmath157,@xmath158 , eq .
differs from that of dimer@xcite et al . by a factor 2 in the term of @xmath159 . the mean photon occupation number calculated for the normal phase by ztop@xcite et al .
coincides with our result up to the first order in @xmath18 .
we ascribe these small difference as to the different diagonalisation procedures for the effective hamiltonians . as in previous works@xcite ,
the status of the region @xmath160 , around the critical point making the normal - superradiant transition remains an open issue . in our approach ,
the effective hamiltonian is not stable in that region .
this could indicate a limitation of the simple mean field approach close to @xmath44 in the dissipation model in a way to fulfill @xmath161 and applying the rwa to the eq . ,
all the critical values @xmath162 coincide . in this case
the critical coupling for the phase transition is independent of the bath coupling @xmath18 ] .
we thank i. lesanovsky and s. genway for useful discussions .
the authors gratefully acknowledge financial support from the daad and dfg grants br @xmath164 , @xmath165 , sfb @xmath166 , and grk @xmath167 .
in eq . , the parameters for the diagonalisation are @xmath168 in the normal phase , @xmath169 .
the eigenenergies are @xmath170 the rotation angle for the decoupling is @xmath171 in the superradiant phase , the eigenenergies of @xmath172 are @xmath173 the rotation angle fulfills @xmath174
the three equations for the macroscopic occupation parameters @xmath175 , eq . , are @xmath176 insertion of , in produces @xmath177 using these values of @xmath178 , we insert them into the parameters of @xmath172 , eq . , @xmath179 finally , the set of coupled differential equations for the @xmath0-representation , eq . , reads ( functions depends on @xmath144 and @xmath2 ) @xmath180 the solution of this system in the long time limit is @xmath181 first , we find a steady state solutions of the last six equations , where time derivatives of @xmath182 @xmath183 are zero .
then we insert them into the differential equation for @xmath184 , eq . , integrate , drop terms that are not dominant in case of @xmath185 and
obtain the equation for @xmath186 .
in order to obtain @xmath98 , we have to use the relation between the old and diagonal basis , because the p - representation has been evaluated in the diagonal basis . @xmath187
24ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1103/physrev.93.99 [ * * , ( ) ] link:\doibase 10.1103/physreva.7.831 [ * * , ( ) ] link:\doibase 10.1016/0003 - 4916(73)90039 - 0 [ * * , ( ) ] link:\doibase 10.1103/physreve.67.066203 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.90.044101 [ * * , ( ) ] @noop * * ( ) link:\doibase 10.1103/physrevlett.104.130401 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.043637 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.043001 [ * * , ( ) ] @noop * * , ( ) @noop ( ) link:\doibase 10.1103/physreva.86.013641 [ * * , ( ) ] link:\doibase 10.1103/physreve.86.012101 [ * * , ( ) ] link:\doibase 10.1103/physrev.58.1098 [ * * , ( ) ] link:\doibase 10.1103/physreva.85.013817 [ * * , ( ) ] @noop _ _ , vol .
( , ) @noop _ _ ( , ) link:\doibase 10.1103/physrevb.67.085316 [ * * , ( ) ] link:\doibase 10.1007/s00340 - 009 - 3436 - 9 [ * * , ( ) ] link:\doibase 10.1103/physreva.75.013804 [ * * , ( ) ] | we consider a driven single mode dicke - hamiltonian coupled to a dissipative zero - temperature bath .
we derive the cumulant generating function for emitted photons of this quantum - critical system by using a @xmath0-representation of the master equation in the thermodynamic limit .
this cumulant generating function is shown to consist of two parts : a macroscopic component , which is poissonian in nature with characteristic rate proportional to the order parameter of the system ; and a part describing fluctuations which is non - trivial in form and divergent around the quantum phase transition . |
ever since eggen , lynden - bell & sandage ( 1962 , hereafter els ) interpreted the correlations between metallicity and kinematics of 221 stars in the solar neighbourhood as evidence in favour of a monolithic collapse formation model of the milky way , the mode of assembly of the milky way has been one of the main goals of observational surveys of the chemistry , age , and kinematics of stars in the galaxy .
soon after the publication of els , it became clear that some data provided challenges to the els interpretation , paving the way to more complex formation scenarios for the milky way .
for example , the properties of the stellar halo , as traced by the globular cluster system , deviate from the trends seen in the sample analyzed by els ( searle & zinn 1978 , hereafter sz ) and have long been viewed as consistent with a formation process that involved the merger of a number of smaller subunits rather than the gradual collapse envisioned by els .
the vigorous observational progress that followed the els and sz papers ( recently reviewed by freeman & bland - hawthorn 2002 ) have led to the development of a milky way formation scenario that borrows from the basic tenets of both the els and sz hypotheses . in this
, the stellar spheroid is built through a number of early mergers , whereas the stellar disk is regarded as the outcome of the smooth , dissipative deposition ( and transformation into stars ) of gas cooling more or less continuously off the intergalactic medium .
although the disk is built smoothly in this scenario , it is not stationary , and may evolve as a result of internal inhomogeneities such as spiral arms and molecular clouds , as well as by the effect of minor mergers with external satellite galaxies .
indeed , the presence of two distinct components in the galactic disk ( the thin and thick disks , gilmore & reid 1983 ) is widely ascribed to the dynamical `` heating '' of an early thin disk by a satellite roughly @xmath2 gyr ago ( quinn & goodman 1986 ; quinn , hernquist & fullagar 1993 ; robin et al .
1996 ; wyse 2004 ) .
axis is chosen to coincide with the rotation axis of the disk component .
the orbital plane is initially seen approximately edge - on in the @xmath3-@xmath4 projection and is inclined @xmath5 relative to the plane of the disk . as the orbit of the satellite decays and approaches the disk , it is rapidly brought into the plane of the disk before disruption .
[ figs : orbev ] ] various versions of this scenario have long guided the interpretation of newer datasets which , thanks to dedicated observing campaigns ; to the availability of accurate distance measurements from hipparcos ; as well as to exquisite spectroscopy possible from 8 - 10 m telescopes , have expanded to include hundreds of stars with reliable three - dimensional kinematics and detailed chemical abundance information .
most of these stars , however , are in the immediate vicinity of the sun , and therefore the analysis suffers from the subtle and inevitable biases that the sun s location in the galaxy brings about
. the bias may be modest if the stellar distribution function is a smooth function of galactic phase - space coordinates ( may & binney 1986 ) , but it may imply severe limitations if a large number of coherent dynamical groups relicts from past accretion events populate the galaxy , as these may well be under- or over - represented in the vicinity of the sun . the relatively quiescent formation scenario of the galaxy described above the galactic disk is after all the dominant stellar component of the galaxy has encouraged the interpretation of the abundance patterns of nearby stars in terms of global collapse models where enrichment proceeds according to the general scenario outlined by tinsley ( 1979 ) .
their success in reproducing the correlations between kinematics and chemistry of solar neighbourhood stars notwithstanding , such models have traditionally been less successful at accounting for the sizable scatter around the mean trends shown in the data , as well as by the presence of abundance oddities in stars with otherwise unremarkable kinematics ( carney et al .
1997 ; king 1997 ; hanson et al . 1998
; fulbright 2002 ) . .
top panel shows the evolution of the distance to the center of the main galaxy , @xmath6 , the distance to the rotation axis of the disk , @xmath7 , and the vertical distance to the plane of the disk , @xmath4 , as a function of time .
second panel shows the evolution of the instantaneous inclination of the orbital plane of the satellite relative to the plane of the disk .
note that the satellite is brought onto an orbit nearly coplanar with the disk before disruption .
third panel shows the evolution of the circularity of the satellite s orbit , defined as the ratio between the @xmath4-component of its angular momentum and that of a circular orbit of the same binding energy .
note that the orbit becomes nearly circular before disruption .
bottom panel shows the evolution of the self - bound stellar mass of the satellite with time until disruption.[figs : orbpar ] ] from a cosmological perspective , a slow and quiescent buildup of the galaxy is at odds with the mode of assembly envisioned in currently popular hierarchical models of structure formation such as the @xmath8cdm scenario ( bahcall et al .
1999 ) , where merging is expected to have contributed significantly to the various populations of the galaxy ( see , e.g. , steinmetz & navarro 2002 ) .
in particular , the presence of an old , extended stellar disk component ( such as that hypothesized to be the progenitor of today s thick disk ) is awkward to accommodate in such models , since the dynamical fragility of stars in disk - like orbits seems to preempt an active merging history .
indeed , the presence of old stars in the disk components of the galaxy is often used to argue that the galaxy has not been disturbed by mergers throughout most of its life ( see , e.g. , wyse 2004 ) ; clearly an unusual assembly history in a universe where structure grows hierarchically .
the precise age of the galactic disk components is still somewhat controversial , given that ages of individual stars are notoriously difficult to measure accurately , but the thin disk of the galaxy is known to contain a fair fraction of metal - poor ( and presumably quite old ) stars .
for example , roughly @xmath9 of stars in the dynamically - unbiased catalog of metal - deficient ( [ fe / h ] @xmath10 ) nearby stars compiled by beers et al .
( 2000 , hereafter b00 ) have thin disk - like angular momenta
. numerical simulations of hierarchical galaxy formation have suggested a way to reconcile the presence of old , metal - poor disk stars with the hectic merging activity envisioned in hierarchical formation scenarios .
abadi et al . ( 2003a , b ) report , for example , that @xmath11 of dynamically - selected thin - disk stars in their cosmological simulation of the formation of a disk galaxy are actually older than the epoch of the galaxy s last major merger .
essentially all ( @xmath12 ) of these old stars are brought in at late times by satellite galaxies whose orbits are circularized and dragged into the disk by dynamical friction before disruption .
the core of satellites that disrupt after their orbits have been nearly circularized may thus contribute stars to the thin disk component upon disruption .
in contrast , satellites that are disrupted before their orbits are substantially eroded by dynamical friction ( or whose orbits are roughly polar to the disk ) will contribute most of their stars to the spheroidal component .
more massive satellites are likely to see their orbits more severely affected by dynamical friction before disruption ( they are typically denser and thus more resilient to disruption ) , and are therefore likely to shed their stars onto more rotationally supported orbits than those of less massive companions more prone to disruption .
this suggests that broad correlations between metallicity and rotation may arise without the need for progressive in - situ enrichment of gas cooling and settling into the disk . indeed , most metal poor stars in the solar neighbourhood
may actually be the overlapping debris of numerous satellites disrupted in the past .
in such scenario , the trend for rotational support to increase with metallicity would just reflect the mass - metallicity relation of their progenitor galaxies , coupled with the more efficient orbital decay of massive satellites in the potential of the galaxy .
if this scenario is correct , it should be possible to identify at least some of the relicts of such accretion events as coherent groups in phase space .
a number of such kinematically - distinct associations have been recognized in the stellar halo ; for example , the sagittarius dwarf ( ibata , gilmore & irwin 1994 ) ; the retrograde rotation of halo stars in the direction of the north galactic pole ( majewski 1992 ) ; and the identification of substructure in the outer halo by helmi et al .
( 1999 ) .
accretion onto the disk has also been recently recognized .
one example is provided by the `` monoceros ring '' of stars in the outer disk of the galaxy , which has been interpreted as material recently stripped from a dwarf galaxy that may still remain hidden in the disk ( yanny et al .
2003 ; newberg et al .
2002 ; helmi et al .
2003 ; ibata et al .
2003 ; rocha - pinto et al .
2003 ; crane et al . 2003 ; martin et al .
further examples are provided by the `` arcturus stream '' ; a group of stars at the apocenter of their eccentric , disk confined orbits that has been interpreted as debris from the disruption of a satellite in the plane of the galactic disk ( eggen 1971 ; navarro , helmi & freeman 2004 ) ; as well as by the discovery of an unexpected , slowly - rotating component above and below the plane of the disk ( gilmore , wyse & norris 2002 ) .
we examine here the possibility that other solar - neighbourhood stars may have originated from material stripped from disrupted satellites . as with the analysis of the `` arcturus group '' , we shall focus on stars with orbits confined to relatively small excursions from the plane of the disk ( but not necessarily on circular orbits ) and use numerical simulations to motivate and guide the search for common orbital characteristics in catalogs of nearby metal - deficient stars .
we describe briefly the numerical simulations and present our main results in section [ sec : numexp ] .
we use these to motivate the analysis of samples of nearby stars with accurate kinematics and chemical abundance measurements in section [ sec : omcen ] .
we conclude with a brief summary and discussion of our main conclusions in section [ sec : conc ] .
we analyze the accretion and disruption of a satellite galaxy in the simulation of the formation of a disk galaxy presented by abadi et al .
( 2003a , b ) .
the simulation follows self - consistently the evolution in a @xmath8cdm universe of a small region surrounding a target galaxy , excised from a large periodic box and resimulated at higher resolution preserving the tidal fields from the whole box .
the simulation includes the gravitational effects of dark matter , gas and stars , and follows the hydrodynamical evolution of the gaseous component using the smooth particle hydrodynamics ( sph ) technique .
dense , cold gas is allowed to turn into stars at rates consistent with the empirical schmidt - like law of kennicutt ( 1998 ) .
the energetic feedback of evolving stars is included as a heating term on the surrounding gas , but its effectiveness in curtailing star formation is low .
the transformation of gas into stars thus tracks closely the rate at which gas cools and condenses at the center of dark matter halos .
details of the simulation are given in abadi et al .
( 2003a , b ) .
as discussed by these authors , at @xmath13 the main galaxy resembles more an early - type spiral than the milky way .
although the luminosity of the disk component ( @xmath14 ) and the circular speed at @xmath15 kpc ( @xmath16 km s@xmath17 ) are comparable to those of our galaxy we emphasize that this is _ not _ a model of the milky way , and that our main goal is to understand _ qualitatively _ rather than quantitatively the nature of debris from tidally disrupted satellites .
we choose to illustrate the analysis with the same satellite used by helmi et al .
( 2003 ) in their discussion of the various formation scenarios of the `` monoceros ring '' .
this is a @xmath18 satellite which is accreted and disrupted at @xmath19 by the main galaxy .
the satellite turns around at @xmath20 from a radius of @xmath21 ( physical ) kpc , and first approaches the galaxy at @xmath22 , when it reaches a pericentric radius of @xmath23 ( physical ) kpc .
its highly eccentric orbit is inclined by @xmath5 relative to that of the disk of the galaxy .
these orbital parameters evolve quickly , as dynamical friction breaks the satellite s orbit and steadily reduces its apocentric radius .
[ figs : orbev ] shows the trajectory of the satellite in three orthogonal projections , chosen so that the @xmath4 axis coincides with the rotation axis of the disk of the main galaxy , which is located at the origin of the coordinate system .
this trajectory follows the center of mass of the self - bound stellar core of the satellite , computed by iteratively removing escapers using the procedure outlined in detail by hayashi et al .
( 2003 ) . for short
, we shall use `` satellite '' to denote this self - bound group of stars that track the surviving core of the satellite .
the orbit of the satellite decays steadily , gradually becoming more circular and more closely aligned with the plane of the disk .
this is shown in fig .
[ figs : orbpar ] , where we show the evolution of the satellite s distance to the galaxy ( top panel ) ; the inclination of the orbit relative to the disk ( second panel ) ; the orbital circularity is defined as the ratio between the orbital angular momentum and that of a circular orbit with the same binding energy , @xmath24 . ] ( third panel ) ; and the self - bound stellar mass of the satellite ( bottom panel ) . in the last few orbits
the satellite s core decays to an almost circular orbit coplanar with the disk .
although the orbital energy ( as measured by the satellite s apocenter ) declines gradually throughout the evolution , the inclination and circularity of the satellite change noticeably as the satellite approaches the disk of the main galaxy , which extends out to @xmath25-@xmath26 kpc at this redshift .
the fast decay into the plane seen after @xmath27 is accompanied by rapid circularization of the orbit of the remaining self - bound core ( fig .
[ figs : orbpar ] ) , and is consistent with the results of previous work on satellite decay in disk - dominated or flattened systems ( see , e.g. , quinn & goodman 1986 ; pearrubia , kroupa & boily 2002 , and references therein ) . the stellar component of the satellite is disrupted mainly during the three last pericentric passages , between @xmath28 and @xmath29 .
the self - bound mass of the satellite core drops to zero after @xmath19 , and the satellite debris quickly phase - mixes into a torus - shaped configuration , as shown in fig .
[ figs : posxyz ] .
the well - mixed , relaxed configuration of the debris at late times hides the presence of several well - defined , dynamically coherent groups of well - defined ( specific ) binding energy ( @xmath24 ) and angular momentum ( @xmath30 ) .
this is shown in fig .
[ figs : histos ] , where several distinct groups of particles are clearly apparent in the @xmath24 and @xmath30 histograms after disruption .
the `` peaks '' in these histograms may be traced to particles lost during each of the last few three pericentric passages prior to full disruption .
this is illustrated by the dashed curve in fig .
[ figs : histos ] , which correspond to stars lost at @xmath31 .
the dot - dashed curve corresponds to stars lost during the last two pericentric passages , at @xmath32 and @xmath33 , respectively . in general ,
the satellite sheds two groups of particles at each pericentric passage , corresponding roughly to the `` inner '' and `` outer '' tidal arms seen in the @xmath34 panel of fig .
[ figs : posxyz ] , so we expect several distinct groups of particles in fig . [ figs : histos ] . because of overlaps between groups , as well as the intrinsic spread in the @xmath24 and @xmath30 distributions due to the finite velocity dispersion of the satellite , roughly three prominent groups are clearly identifiable by their distinct energy and angular momentum after disruption .
the evolution of these groups in position - velocity space is shown in figs .
[ figs : vphi ] and [ figs : vr ] , which show the rotation ( @xmath35 ) and radial ( @xmath36 ) velocity as a function of their distance ( @xmath7 ) to the rotation axis of the disk .
the solid curve in fig .
[ figs : vphi ] shows the circular velocity on the plane of the disk of the main galaxy . dashed hyperbolae in this figure are lines of constant specific angular momentum which indicate the location of three prominent `` peaks '' in the @xmath30 distribution shown in fig .
[ figs : histos ] .
these groupings are also seen in the @xmath36-@xmath7 plane , as shown in fig .
[ figs : vr ] .
solid line curves in the @xmath37 panel of this figure indicate the average trajectory of particles with various energies and angular momenta .
the innermost and outermost loops trace orbits with ( @xmath24/km s@xmath38 , @xmath30/km s@xmath17 kpc ) of order ( @xmath39 , @xmath40 ) and ( @xmath41 , @xmath40 ) , corresponding roughly to two of the peaks in the distribution of stripped stars seen in figure [ figs : histos ] . these groupings in @xmath24-@xmath30 space are expected to be long lived , provided that the galactic potential remains relatively quiescent afterwards . as discussed by abadi et al .
( 2003a , b ) , the disruption of satellites on coplanar orbits provides a mechanism for adding stars of various ages and metallicities to the disk components of a galaxy .
these authors concentrate their analysis on the debris that contributes to the thin disk , and argue that accretion events are possibly responsible for the majority of old thin - disk stars in the solar neighbourhood .
satellites like the one we analyze here do not contribute substantially to the thin disk , as few particles in the debris of this satellite end up on nearly circular orbits .
however , most of the debris is confined to a plane roughly coincident with that of the disk of the galaxy , as shown in fig .
[ figs : posxyz ] , and might therefore be considered part of a thick disk component . ) at various radii . the solid curve is the histogram measured in the range ( @xmath42 , @xmath43 ) kpc , the dashed curve corresponds to the range ( @xmath44 , @xmath45 ) kpc , and the dashed curve to ( @xmath2 , @xmath26 ) kpc .
these locations are chosen to bracket the apocentric radii corresponding to the three most prominent `` peaks '' in the distribution of specific energies shown in fig .
[ figs : histos ] , and are intended to illustrate the broad , symmetric , `` double - humped '' velocity distribution measured by observers in the disk at radii intermediate between the characteristic apocentric radii of the debris .
[ figs : histvr ] ] to an observer in the disk , the local properties of the satellite debris will depend significantly on the position of the observer relative to the pericenter and apocenter of each of the groups mentioned above
. local detection of these groupings as recognizable features against the background of galactic stars would be easier near apocenter , since stars tend to crowd there during their orbits .
a possible debris detection strategy would thus be to search samples of nearby stars for groups with : ( i ) modest vertical motions , ( ii ) negligible net galactocentric radial velocity , and ( iii ) common angular momentum , and to probe them for corroborating traits suggestive of a common origin .
one example of such approach is provided by the `` arcturus group '' reported by navarro et al .
these authors argue that the significant excess of disk stars with angular momentum similar to arcturus seen in the b00 catalog , together with other , more indirect evidence ( such as the tight correlation in the [ @xmath1/fe ] versus [ fe / h ] plane , and the discovery of an unexpected population with similar rotation speed above and below the galactic plane by gilmore et al .
( 2002 ) ) , constitute persuasive evidence for the extra - galactic origin of the bright star arcturus and of other members of its namesake group . on the other hand , for observers located _ between _ the apocenter and pericenter of a well - mixed group the debris would show locally as groups of stars with a rather broad , symmetric ( and at times `` double - humped '' ) distribution of galactocentric radial velocities .
this is illustrated in fig .
[ figs : histvr ] , where we show the @xmath36 histogram of stars chosen at various radii across the disk after disruption ( @xmath46 ) .
these radii are chosen to lie _ between _ the apocenters of the most prominent groups identified in fig .
[ figs : histos ] .
the relatively symmetric , double - peaked distribution of galactocentric radial velocities is a telltale sign of a past accretion event that left a substantial fraction of stars on orbits more energetic than the observer s .
we shall use this result below to motivate the search for coherent dynamical structures in samples of nearby stars that may have originated in the disruption of former satellites of the milky way .
given the clear trend for the metallicity of galaxies in the local group to increase with luminosity ( van den bergh 1999 ) , signatures of the accretion and disruption of dwarf galaxies are expected to be more prevalent in samples of metal - poor stars . b00
provide a catalog of metal - deficient stars in the vicinity of the sun , compiled with the specific aim of minimizing kinematic selection biases .
this is the largest sample of metal - poor ( [ fe / h ] @xmath47 ) stars with available fe abundances , distances , and radial velocities .
proper motions are also available for many of these stars , making it a good sample to search for substructure ( see , e.g. , chiba & beers 2000 ; brook et al .
2003 ) .
the top dot - dashed histogram in fig .
[ figs : vphihist ] shows the distribution of galactocentric rotation speeds for all stars in the b00 sample .
the full sample is dominated by the halo and the canonical `` thick disk '' component .
a decomposition of the @xmath48 distribution into three gaussians representing the `` canonical '' components of the galaxy is shown in fig .
[ figs : vphihist ] . from left to right , the halo , thick disk , and thin disk components are assumed to have ( @xmath49 , @xmath50 ) equal to ( @xmath51,@xmath52 ) , ( @xmath53,@xmath54 ) and ( @xmath55,@xmath56 ) km s@xmath17 , and are found to contribute @xmath57 , @xmath58 , and @xmath9 of the sample , respectively .
the halo distribution is normalized to match the number of counter - rotating stars , whereas the relative contributions of the thick and thin disks are chosen to match the @xmath48 distribution of the remaining co - rotating stars .
the fit obtained is shown by the thick solid curve in fig .
[ figs : vphihist ] .
most random realizations are unable to account for the `` excess '' of stars labelled as the `` arcturus group '' at @xmath59 km s@xmath17 ; only @xmath60 out of @xmath61 realizations give a number of stars comparable to what is observed ( navarro et al .
2004 ) . of similar significance
is the excess of stars on slightly retrograde orbits ( @xmath62 km s@xmath63 km s@xmath17 ) labelled as the `` @xmath0cen group '' in fig .
[ figs : vphihist ] .
stars in this group have specific angular momenta similar to that of the globular cluster @xmath0cen , the most luminous and unusual of the milky way globular clusters . although the typical metallicity of stars in @xmath0cen is low ( its distribution peaks at [ fe / h ] @xmath64 ; see smith 2004 for a review ) , there is convincing evidence for a long and complex star formation and enrichment history in the structure of the giant branch of cluster stars ( norris , freeman & mighell 1996 ; lee et al .
1999 ) , and in the unusual s - process element abundances , which point to the retention of elements fabricated in agb stars ( norris & da costa 1995 ) .
@xmath47 ) stars of beers et al .
( 2000 ) ( dashed histogram ) , decomposed into three canonical galactic ( gaussian ) components : halo , thick disk , and thin disk , as labelled .
the solid curve denotes the sum of all three components .
note the excess of stars lagging the local standard of rest ( lsr ) rotation by @xmath65 km s@xmath17 , coincident with the arcturus group , as well as that for mildly retrograde orbits labelled as the @xmath0cen group .
see text for full details .
[ figs : vphihist ] ] these properties strongly suggest that the cluster did not form on its present orbit , which is rather bound ( apocenter @xmath66 kpc ) , confined to the disk ( @xmath67 kpc ) , and mildly retrograde ( dinescu , girard & van altena 1999 ) .
this orbit implies frequent passages through the disk , and may only be reconciled with significant gas retention and repeated episodes of star formation if @xmath0cen formed elsewhere and its orbit has decayed substantially over time .
efficient decay requires more mass than presently attached to the cluster , implying that @xmath0cen was likely much more massive in the past .
indeed , it has long been hypothesized that it was the nucleus of a rather massive and dense dwarf accreted and disrupted in the tidal field of the milky way ( freeman 1993 ) .
although the apocenter of @xmath0cen s orbit lies at present within the solar circle , a substantial fraction of the stars of its parent galaxy may have been shed into orbits of higher energy that now intersect the solar neighbourhood .
this has motivated several authors to search for debris from @xmath0cen s parent galaxy in catalogs of nearby metal - poor stars .
dinescu ( 2002 ) , for example , argues that the retrograde signature noted in fig .
[ figs : vphihist ] for stars with @xmath68 km s@xmath17 is most prevalent in samples with metallicities restricted to @xmath69 [ fe / h ] @xmath70 ; a range that overlaps the metallicity distribution of @xmath0cen stars .
tidal debris tends to settle into a structure more vertically extended than the thin disk of the galaxy , and therefore it may be detectable _ in situ _ above and below the plane of the galaxy .
mizutani , chiba & sakamoto ( 2003 ; see also chiba & mizutani 2004 ) have recently argued that this may explain the presence of a high - velocity ( @xmath71 km s@xmath17 ) peak in the line - of - sight velocities of stars in the direction against galactic rotation recently presented by gilmore et al .
( 2002 ) .
stars have accurate kinematics and distances , as well as reliable element - to - element abundance ratios . the excess of stars corresponding to the arcturus and @xmath0cen groups noted in the beers et al .
( 2000 ) catalog ( fig .
[ figs : vphihist ] ) is also clearly seen here in this independent compilation .
[ figs : jzgrat ] ] our analysis is complementary to that of these authors , but focuses instead on the galactocentric radial velocity distribution of disk - confined stars with angular momenta similar to @xmath0cen s , as well as on the detailed abundance patterns of such stars . ) versus radial ( @xmath72 ) velocities of metal - poor stars in the gratton et al .
( 2003 ) compilation ( open circles ) .
filled circles are used to highlight @xmath0cen group candidates .
these are stars with angular momenta similar to @xmath0cen s ( starred symbol ) , and with relatively low vertical velocity ( see conditions as labelled in the figure ) .
the @xmath72 distribution of the @xmath73 candidate stars ( dashed histogram , multiplied by 4 for clarity , scale on right axis ) differs markedly from that of the rest of the stars in compilation ( solid histogram ) .
only 3 candidate stars have @xmath74 km s@xmath17 , whereas @xmath75 would be expected from the overall distribution .
[ figs : uw ] ] gratton et al .
( 2003 , hereafter gcclb ) have compiled element - to - element abundance ratio measurements for about @xmath76 stars in the solar neighbourhood with accurate kinematical data , bringing together their own measurements with literature data from the work of nissen & schuster ( 1997 ) , fulbright ( 2000 ) , and prochaska et al .
the angular momentum distribution of these stars is presented in fig .
[ figs : jzgrat ] .
the `` excess '' of stars in the b00 catalog corresponding to the @xmath0cen and arcturus groups noted in fig .
[ figs : vphihist ] are also clearly seen in this compilation , underscoring the robustness of these kinematic features to various selection techniques . following the motivation outlined in section [ sec : numexp ] , we analyze the galactocentric radial ( @xmath72 ) and vertical ( @xmath77 ) velocity , @xmath78 , and @xmath77 to denote radial , tangential and vertical velocities in galactic coordinates , measured relative to the lsr .
@xmath72 is positive outwards , @xmath78 in the direction of the sun s rotation , and @xmath77 towards the north galactic pole .
we assume that the sun s ( @xmath72,@xmath78,@xmath77 ) velocity is ( @xmath79,@xmath80,@xmath81 ) km s@xmath17 . ]
distribution of @xmath0cen group stars in fig .
[ figs : uw ] .
open circles show the vertical and radial velocities of all stars in the gcclb compilation , whereas the filled circles highlight those with angular momenta in the `` peak '' associated with @xmath0cen in fig .
[ figs : jzgrat ] , i.e. , @xmath30 between @xmath82 and @xmath83 km s@xmath17 kpc .
we choose to focus on those stars which , like @xmath0cen , have their orbits roughly confined to the disk , and restrict the candidate list further to stars with @xmath84 km s@xmath17 in order to weed out unrelated halo stars .
thirteen stars are selected by these two criteria as candidate members of the group , and their @xmath72 distribution is shown by the dashed histogram in fig . [ figs : uw ] ( for clarity , the number of stars in each bin is multiplied by four relative to the scale on the right axis ) .
the radial velocity distribution of these stars is in obvious contrast with those of all stars in the gcclb compilation , which is well described by a gaussian of zero mean and dispersion @xmath85 km s@xmath17 ( solid histogram in fig .
[ figs : uw ] ) , as expected for stars selected preferentially near the apocenter of their orbits .
the @xmath0cen group candidates , on the other hand , span a large symmetric range in @xmath72 , from @xmath86 to @xmath83 km s@xmath17 , with no obvious central tendency ( i.e. , few stars near apocenter ) and a velocity dispersion exceeding @xmath87 km s@xmath17 . indeed ,
only three out of thirteen candidates have @xmath74 km s@xmath17 , whereas one would have expected three times as many had their @xmath72 distribution been similar to that of the whole ensemble . with the obvious caveat that the number of candidates is rather small ,
one is tempted to interpret the broad , roughly symmetric @xmath72 distribution as evidence that some of the candidate stars are part of a dynamically coherent group shed by @xmath0cen s dwarf prior to disruption .
the apocenter of this group is outside the solar circle , and therefore group stars are only seen moving towards apocenter with high positive @xmath72 velocity or coming from the apocenter with high negative @xmath72 .
support for the common origin of this group comes from analyzing the abundances of candidate stars .
[ figs : fehafe ] shows the abundance of @xmath1 elements and iron for stars in the gcclb sample . as is customary ,
the @xmath1-element abundance is expressed as the logarithm of its abundance relative to fe in solar units , [ @xmath1/fe ] , and the iron abundance is given as a logarithmic measure of its ratio to h also in solar units , [ fe / h ] .
open circles correspond to all stars in the gcclb compilation , whereas the filled circles indicate the @xmath0cen group candidates , as in fig .
[ figs : uw ] .
remarkably , the @xmath0cen group candidates are not distributed at random amongst the sample , but rather trace a well - defined narrow track in this plane , spanning a wide range in iron abundance ( @xmath88 [ fe / h ] @xmath89 ) which is consistent with the spread in metallicity measured for individual stars in @xmath0cen ( suntzeff & kraft 1996 ) .
this well defined track in the [ @xmath1/fe ] versus [ fe / h ] plane is roughly consistent with simple one - zone self - enrichment models ( see , e.g. , matteucci & francois 1989 ) , but where pollution by the iron - rich ejecta of type ia supernovae becomes important at lower metallicity than for `` typical '' stars in the halo , where [ @xmath1/fe ] drops precipitously for [ fe / h ] @xmath90 .
intriguingly , two candidate stars appear to deviate from this trend , and are highlighted by starred symbols in fig . [
figs : fehafe ] .
this is consistent with the above interpretation , as these two stars are dynamically quite distinct from the rest ; although they share the angular momentum of @xmath0cen , their low @xmath72 velocity ( @xmath91 km s@xmath17 ) sets them apart from the rest as unrelated stars on much more tightly bound orbits . the chemical coherence highlighted by the filled symbols in fig .
[ figs : fehafe ] is thus manifest only when stars of coherent kinematics ( i.e. similar @xmath24 and @xmath30 ) are considered part of the group .
overall , this pattern is consistent with that expected for a satellite system that self - enriched to a metallicity of order one - tenth of the sun on a time scale somewhat longer than envisioned for the bulk of metal - poor stars in the vicinity of the sun .
we conclude that the chemical and dynamical coherence of stars in the @xmath0cen group is highly suggestive of a common origin , most likely a single dwarf system that was dragged and disrupted into the disk of the galaxy early during its assembly history .
cen group .
note that these stars delineate a well defined sequence in the abundance ratios that may be interpreted as resulting from a protracted episode of star formation where a system self - enriched to a metallicity of order one - tenth of solar .
two candidate stars deviate from the main trend , and are highlighted with starred symbols . these two stars , however , are kinematically distinct from the rest .
although they share the angular momentum of @xmath0cen , they are at the apocenter of their orbits ( @xmath91 km s@xmath17 ) in sharp contrast with the rest of @xmath0cen group candidates , which are moving through the solar neighbourhood to or from their apocenter at an average radial velocity of @xmath92 km s@xmath17 .
these stars are unlikely to be related to @xmath0cen s progenitor .
[ figs : fehafe ] ]
we use numerical simulations to analyze the dynamical properties of tidal debris stripped from a satellite on a highly eccentric orbit roughly coplanar with the disk of the main galaxy upon disruption . because of the small pericentric radius of the orbit , the stellar component of the satellite is disrupted by the tidal shocks that accompany the last few pericentric passages , splitting the bulk of stars into coherent groups of common energy and angular momentum on eccentric orbits roughly confined to the disk .
observers located between the pericenter and apocenter of such groups would see the debris as stars of common angular momentum but with a wide , symmetric ( and at times `` double - peaked '' ) distribution of galactocentric radial velocities .
a group of stars with these characteristics stands out in catalogs of nearby metal - poor stars and share the peculiar kinematics of the unusual globular cluster @xmath0cen .
these stars orbit the galaxy on retrograde orbits but confined to relatively small vertical excursions from the plane of the disk , and show extremely well - correlated element - to - element abundance ratios .
they span a range in [ fe / h ] comparable to that of individual stars in @xmath0cen , and follow closely the [ @xmath1/fe ] versus [ fe / h ] correlation expected for a simple closed - box self - enrichment model .
the dynamical and chemical coherence of stars in this group are highly suggestive of a common origin , which we ascribe to the dwarf galaxy that brought @xmath0cen into the galactic disk .
the @xmath0cen group thus shares some of the properties of the `` arcturus group '' ; a group of stars of moderate velocity dispersion and singular metal abundance ( eggen 1971 ) , that navarro et al .
( 2004 ) have recently claimed to be part of the debris from another past accretion event .
however , the @xmath0cen and arcturus groups also differ in a number of properties . for example , although arcturus group members have negligible net radial motions and are thus near the apocenter of their orbits , @xmath0cen group stars are passing by the solar neighbourhood at fairly high speeds to or from apocentric radii far outside the solar circle .
in addition , most arcturus group members have highly enhanced [ @xmath1/fe ] ratios relative to solar , and are in this regard similar to the majority of metal - poor stars in the solar neighbourhood ( see figure 3 of navarro et al .
on the other hand , the [ @xmath1/fe ] ratio of @xmath0cen group members approaches the solar value at relatively low metallicity ( [ @xmath1/fe ] @xmath93 at [ fe / h ] @xmath94 ; see fig .
[ figs : fehafe ] ) .
thus , several stars in this group have unusually low [ @xmath1/fe ] enhancement relative to typical halo stars of comparable metallicity .
this is important , as it suggests an `` extra - galactic '' resolution for the puzzle surrounding stars with anomalously low [ @xmath1/fe ] in the solar neighbourhood : namely , all such stars may come from disrupted dwarfs ( carney et al . 1997 ; king 1997 ; hanson et al .
1998 ; fulbright 2002 ) .
indeed , low [ @xmath1/fe ] ratios are consistent with those of stars in the dwarf satellites of the milky way ( draco ; sculptor ; and sagittarius ; see , e.g. , shetrone et al .
2001 , 2003 ; bonifacio et al .
2004 ; venn et al .
2004 ) .
the prominence of the arcturus and @xmath0cen groups in samples of metal - deficient stars ( the two combined make up to @xmath95 of the stars in the gcclb sample ) suggests that accretion events have probably contributed a significant number of the metal - deficient stars in the solar neighborhood .
these two examples demonstrate that at least some , and perhaps many , old stars on orbits confined to the disk did not form _ in situ _ , but were brought into the galaxy by accretion events .
it is clearly important to search exhaustively for signatures of past accretion events in the solar neighbourhood in order to quantify properly the contribution of tidal debris to the inventory of metal - poor stars in the galaxy , as this may seriously affect the validity and applicability of galactic formation and evolution models .
for example , els - inspired interpretations of the correlation between metallicity and rotational support for nearby metal - poor stars in terms of a global collapse rest on the assumption that their orbits have not evolved drastically over time @xmath96 ( see also bekki & chiba 2000 for further discussion ) .
our comment , on the other hand , refers to all stars in the beers et al .
( 2000 ) sample , i.e. , those with [ fe / h ] @xmath47 . ] . on the other hand ,
if most metal - poor stars in the solar neighbourhood have been brought into the galaxy by accretion events , a drastically different scenario may be required to accommodate this correlation .
one possibility is that this may just reflect the increasing efficiency of dynamical friction in circularizing the orbits of more massive satellites the most effective contributors of stars relatively rich in metals .
is it possible that the whole trend seen in fig .
[ figs : fehafe ] is caused by the overlap of stars stripped from numerous satellites , each of which may have self - enriched roughly as a closed box prior to accretion ?
it would be premature to argue that the data is fully consistent with such scenario , but nor is it obviously ruled out by the data . for example , the scarcity of metal - poor stars with low [ @xmath1/fe ] may be explained by arguing that they originate in the smallest and most metal - poor of all accreted satellites ( i.e. , the likes of sculptor , draco , or the progenitor of @xmath0cen ) .
overall , few stars are locked up in such small satellites , and therefore it is not surprising that they contribute a relatively small fraction of all the metal - deficient stars in the solar neighbourhood . in this interpretation , the main trend in the [ @xmath1/fe ] versus [ fe / h ] plane ( fig .
[ figs : fehafe ] ) is dominated by the enrichment process of the few massive dwarfs whose debris dominates the metal - poor star counts in the solar neighbourhood .
one example would be the progenitor of the arcturus group , which probably self - enriched rapidly to [ fe / h]@xmath97 before disruption , and therefore contributed mostly stars highly enhanced in [ @xmath1/fe ] .
a number of critical enquiries should be pursued before one may elevate this speculation into a working hypothesis to guide the interpretation of future datasets .
these enquiries will soon become possible , as the completion of large - scale surveys such as those planned by the rave collaboration ( steinmetz et al .
2003 ) or by gaia ( perryman et al . 2001 ) promise to provide a comprehensive appraisal of the importance of kinematical substructure in the dynamics of the galaxy .
detailed abundance studies of stars selected from such surveys to belong dynamically to either the arcturus or @xmath0cen groups may corroborate the `` closed - box '' evolutionary sequence outlined for the progenitors of these stellar groups .
( we note , however , that this may only apply to the low - mass end of accreted dwarfs , as large satellites like , say , the lmc , may themselves have been built through the assembly of smaller subunits . ) the discovery of further dynamically and chemically coherent groups associated with stars of anomalous chemistry in the solar neighbourhood would be especially helpful to confirm the extra - galactic origin of such stars .
finally , ongoing studies of the kinematics of thick disks in external galaxies ( see , e.g. , dalcanton , yoachim & bernstein 2004 ) hold the promise of extending these studies beyond the unique environment of the milky way and to unravel the cosmological role of accretion events in the formation of the various galactic components .
am acknowledges support from the comit mixto eso - chile and from the facultad de ciencias fsicas y matemticas of the universidad de chile .
jfn is supported by canada s nserc , the canadian foundation for innovation , and the alexander von humboldt foundation .
smith v.v . , 2004 , in mcwilliam a. , rauch m. , eds . , carnegie observatories astrophysics series , vol .
4 : origin and evolution of the elements , cambridge university press , cambridge , p. 186
( astro - ph 0310654 ) wyse r. f. g. , 2004 , in livio m. , ed . , proc .
2003 may symposium space telescope science institute , the local group as an astrophysical laboratory , cambridge university press , cambridge , in press ( astro - ph/0402636 ) | we use numerical simulations to investigate the orbital characteristics of tidal debris from satellites whose orbits are dragged into the plane of galactic disks by dynamical friction before disruption .
we find that these satellites may deposit a significant fraction of their stars into the disk components of a galaxy , and use our results to motivate the search for accretion relicts in samples of metal - poor disk stars in the vicinity of the sun .
satellites disrupted on very eccentric orbits coplanar with the disk are expected to shed stars in `` trails '' of distinct orbital energy and angular momentum during each pericentric passage .
to an observer located between the pericenter and apocenter of such orbits , these trails would show as distinct groupings of stars with low vertical velocity and a broad , symmetric , often double - peaked distribution of galactocentric radial velocities .
one group of stars with these characteristics stands out in available compilations of nearby metal - poor stars .
these stars have specific angular momenta similar to that of the globular cluster @xmath0cen , long hypothesized to be the nucleus of a dwarf galaxy disrupted by the milky way tidal field .
in addition to their kindred kinematics , stars in the @xmath0cen group share distinct chemical abundance characteristics , and trace a well - defined track in the [ @xmath1/fe ] versus [ fe / h ] plane , consistent with simple closed - box enrichment models and a protracted star formation history . the dynamical and chemical coherence of this group suggests that it consists of stars that once belonged to the dwarf that brought @xmath0cen into the galaxy .
the presence of this and other `` tidal relicts '' in the solar neighbourhood suggest an extra - galactic origin for the presence of nearby stars with odd kinematics and chemistry , and implies that accounting for stars contributed by distinct satellite galaxies may be crucial to the success of models of galactic chemical enrichment .
[ firstpage ] galaxy : disk galaxy : formation galaxy : kinematics and dynamics galaxy : structure |
@xmath0-frieze patterns have been considered for the first time in @xcite with a slightly more special definition . meanwhile , many papers have treated different types of @xmath0-patterns .
let us shortly recapitulate some definitions and results .
let @xmath9 , @xmath10 .
an _ @xmath0-frieze pattern _ ( or _
_ @xmath0-frieze__for short ) is an array @xmath11 of the form @xmath12 where there are @xmath13 zeros on the left and right in each row , @xmath14 are numbers in some field , and such that every ( complete ) adjacent @xmath4 submatrix has determinant @xmath5
. it may be convenient to extend the pattern to the right and to the left by repeating the rows ; in this case one has to multiply every second repetition by @xmath15 : @xmath16 we call @xmath17 the _ height _ of @xmath11 .
we call the frieze @xmath11 * _ integral _ if all @xmath14 are in @xmath18 , * _ non - zero _ if @xmath19 for all @xmath20 , * _ positive _ if all @xmath14 are positive real numbers , * _ periodic _ if there exists an @xmath21 such that @xmath22 for all @xmath23 and @xmath24 , * _ _ specialized__if any adjacent @xmath8 submatrix not containing too many zeros from the border has nonzero determinant , * _ tame _ if all adjacent @xmath7 submatrices have determinant @xmath25 , * _ wild _ if it is not tame
. sometimes it will be useful to view the extended @xmath11 as a matrix @xmath26 , where @xmath27 and thus @xmath28 . 1 .
conway - coxeter friezes are exactly the integral positive @xmath29-friezes .
they are all tame , specialized , and periodic .
2 . extending the array @xmath30 periodically ( in all directions ) we obtain a tame integral positive @xmath1-frieze .
it has been proved several times in the literature that tame friezes are periodic , see for example @xcite .
the main reason is ( compare @xcite ) : let @xmath32 be a tame @xmath0-friezeand denote @xmath33 then the matrices @xmath34 are all equal for fixed @xmath35 and @xmath36 , and they are of the form @xmath37 where the last column has alternating signs and @xmath38 if the frieze is positive .
since @xmath11 is tame , it is periodic with period @xmath39 . thus the sequence of @xmath40 is periodic , and @xmath41 .
further , the complete frieze pattern is uniquely determined by this sequence of @xmath40 , thus by a sequence of @xmath39 tuples of the form @xmath42 .
an interesting question concerning tame friezes is : there are only finitely many tame integral positive @xmath1-friezes of height @xmath43 .
more precisely , the number of such friezes could be @xmath44 .
notice that there are only finitely many tame integral positive @xmath29-friezes ( see for example @xcite ) , that there are only finitely many tame integral positive @xmath1-friezes of height less than @xmath43 ( @xmath45 , @xmath46 , @xmath47 of heights @xmath48 resp . ) , but that there are infinitely many tame integral positive @xmath1-friezes of height greater than @xmath43 ( see @xcite ) .
notice also that the examples in section [ wfp ] show that there are infinitely many wild integral positive @xmath0-friezes of a fixed height .
the ( apparently ) most useful identity in the context of @xmath0-friezesis sylvester s identity : if we denote @xmath49 the determinant of the @xmath50 adjacent submatrix at position @xmath51 in the array @xmath52 ( i.e. @xmath53 ) , then @xmath54 for @xmath55 ( assume @xmath56 ) .
[ cornerpos ] if @xmath57 are positive , then @xmath58 is positive as well .
specialized@xmath0-friezesare tame by equation ( [ impeq ] ) for @xmath59 : @xmath60 if @xmath11 is a specialized@xmath0-frieze , then it is uniquely determined by any sequence of @xmath13 successive rows .
if we choose @xmath61 , @xmath62 , @xmath63 to be transcendent and algebraically independent , then @xmath11 is specialized .
replacing every entry @xmath64 in @xmath11 by @xmath65 we obtain the _ dual _ frieze @xmath66 .
if @xmath11 is tame , then @xmath66 is a transposed , translated copy of @xmath11 ( see @xcite or @xcite ) . thus : a tame frieze is specializedif and only if it is nonzero .
but is there a tame frieze which is not specialized ?
yes : extending the array @xmath67 periodically ( in all directions ) we obtain a tame but not specializedintegral @xmath1-frieze .
the following lemma ( together with the fact that tame friezes are periodic ) generalizes the main result of @xcite . if @xmath11 is an @xmath0-friezesuch that @xmath68 then @xmath11 is specialized(and thus tame and periodic ) .
it suffices to prove ( see ( i ) and ( ii ) below ) that ( [ lemass ] ) holds for the next sequence of @xmath13 rows of the pattern : induction then gives @xmath69 for @xmath70 , @xmath71 , and @xmath72 .
but then ( [ lemass ] ) also holds for the frieze given by @xmath73 .
the same argument thus shows that @xmath69 for @xmath70 , @xmath71 , and all @xmath36 .
in particular , the case @xmath74 implies that @xmath11 is specialized .
\(i ) let @xmath75 .
then @xmath76 , @xmath77 , and @xmath78 are positive by ( [ lemass ] ) ( where @xmath79 if @xmath59 ) , @xmath80 , and @xmath81 because it is on the border of the frieze .
thus by corollary [ cornerpos ] for @xmath82 , @xmath83 , @xmath75 , we get @xmath84 .
induction over @xmath85 yields @xmath86 for @xmath87 as well ( notice that we always have @xmath81 ) .
\(ii ) but then we can move to the next entries to the right , @xmath88 , @xmath89 : as before , @xmath90 , @xmath91 , and @xmath92 are positive by ( [ lemass ] ) , and @xmath93 by the induction in ( i ) .
hence by an induction over @xmath35 and for each @xmath35 an induction over @xmath85 ( like in ( i ) ) we obtain @xmath94 for all @xmath95 and @xmath96 .
are there wild @xmath0-friezes ? yes : but what about non - periodic wild friezes ?
the key difference between tame and wild friezes is : if @xmath11 is a wild @xmath0-frieze , then there exist @xmath13 successive rows which do not uniquely determine the following row by the @xmath0-condition .
this happens if and only if some adjacent @xmath8 submatrix within these rows has determinant zero .
define a directed graph @xmath98 which has the set of @xmath99-tuples of vectors in @xmath100 as vertices and such that two tuples @xmath101 and @xmath102 are connected by an edge @xmath103 if @xmath104 and if @xmath105 satisfy the @xmath0-condition when embedded into a frieze . in this example , @xmath106 .
we choose @xmath107 vertices in the graph @xmath108 : let @xmath109 @xmath110 @xmath111 @xmath112 @xmath113 @xmath114 now putting the pieces @xmath115 together , where successive pieces overlap with one row and according to the following subgraph of @xmath108 , wild friezes really are wild . figure [ fig_wildgraphs ] consists of two further examples of subgraphs of @xmath98 .
every closed path corresponds to a periodic integral positive @xmath1-frieze .
as above , we may construct infinitely many non periodic friezes from these pictures .
( of course , we have omitted a list of the corresponding vertices . )
we now display a @xmath1-frieze which is integral , positive , not periodic , and with unbounded entries .
let @xmath119 , and denote by @xmath120 the array displayed in figure [ q0q1 ] where the entries @xmath121 are given in figure [ entri ] .
see figure [ q0q1 ] for examples .
one can check that concatenating @xmath137 produces an integral , positive @xmath1-frieze in which the entries increase without limit .
the integrality comes from the fact that the above formulas are expressions for integral recursions . for example
, @xmath138 satisfies @xmath139 we computed this example with the following technique .
start with an arbitrary vertex in @xmath98 and successively construct adjacent vertices ( with entries within a certain bound ) .
this will produce a graph with loops and many branches .
some of them turn out to be leaves or `` dead ends '' ( at least we find no continuation within the bounds ) . now , after removing some dead ends , we look for a longest path in our constructed subgraph .
then we look for a pattern among the vertices on the path . | in this note , among other things , we show : 1 . there are periodic wild @xmath0-frieze patterns whose entries are positive integers .
2 . there are non - periodic @xmath0-frieze patterns whose entries are positive integers .
3 . there is an @xmath1-frieze pattern whose entries are positive integers and with infinitely many different entries . in this note , among other things ,
we show : there are periodic wild slk - frieze patterns whose entries are positive integers .
there are non - periodic slk - frieze patterns whose entries are positive integers .
there is an sl3-frieze pattern whose entries are positive integers and with infinitely many different entries .
conway - coxeter frieze patterns are arrays of positive integers bounded by diagonals of ones such that every adjacent @xmath2 subdeterminant is one .
they were introduced in @xcite and classified for the first time in @xcite .
more than @xmath3 years later , they appeared in a natural way in several other contexts , for instance in the fields of cluster algebras and of nichols algebras . among the generalizations of frieze patterns that have been introduced since their first appearance ( see @xcite ) are the @xmath0-frieze patterns which we will discuss in more detail in this note .
in the literature , @xmath0-frieze patterns satisfy an @xmath0-condition ( adjacent @xmath4 subdeterminants are @xmath5 ) , usually they consist of integers , and sometimes the entries are required to be positive .
for example , the array @xmath6 is an @xmath1-frieze pattern . in the case of tame friezes ,
i.e. when all adjacent @xmath7-subdeterminants are zero , these three conditions appear to be very natural ; on the one hand they imply that the entries of the frieze are specializations of variables of a cluster algebra , on the other hand , positivity has a nice geometric interpretation .
integrality could turn out to be natural because of some yet unknown representation theoretical motivation ( as for example given by the nichols algebras in the case of conway - coxeter friezes , see @xcite , @xcite ) .
however , in the wild ( not tame ) case , the examples presented in this note suggest that the usual conditions on the entries do not seem to be the natural ones .
in fact , the determinants of the adjacent @xmath8 submatrices in an @xmath0-frieze pattern will probably play a more important role than the entries of the frieze themselves . and
we believe that to assume that these determinants are nonzero ( the frieze is then tame ) is more natural than to assume positivity of the entries .
we exhibit some facts on tame friezes in the second section . in the last section we provide many examples of wild @xmath0-friezes illustrating that it is probably hopeless to classify all wild friezes ( even when the entries are positive integers ) , although to my knowledge , no wild friezes with positive integral entries were known before . *
acknowledgement : * i would like to thank c. bessenrodt , t. holm , p. jrgensen , and s. morier - genoud for calling my attention to some of the questions addressed here and for many further valuable comments . |
after many years of preparations and constructions , the large hadron collider at cern will start its operation soon .
one of the most important questions the lhc should provide an answer to is related to the nature of the electroweak gauge symmetry breaking . in the standard model ( sm ) it
is achieved by adding a doublet of higgs fields and arranging the parameters such that the symmetry is spontaneously broken .
although the sm is extremely successful in describing the experimental data , there are many arguments suggesting that it can not be the ultimate theory , the sm higgs sector is unnatural can not explain why the electroweak scale is so small with respect to the planck scale ; the sm can not account for the matter - antimatter asymmetry of the universe , nor for the dark matter in fact , the above shortcomings point towards a new physics at the tev scale . among many propositions for solving part of these problems in the physics area beyond the standard model , supersymmetry ( susy )
is generally considered most elegant and respected . not invented or designed to
, it can accommodate or explain some of the outstanding problems of the sm : stabilizing the gap between electroweak and planck scale , unifying the gauge couplings , inducing radiative electroweak symmetry breaking .
it provides a candidate for dark matter ( dm ) , offers new ideas on the matter - antimatter asymmetry , while its unique mathematical structure can provide a link to physics at the gut / planck scale and , in local form , paving the path to gravity . in the simplest n=1 supersymmetric extension of the sm ,
each sm particle is paired with a sparticle that differs in spin by half a unit . since
none of the sparticles has been seen so far , the supersymmetric lagrangian must be supplemented by susy breaking terms that keep unseen superpartners out of experimental reach while retaining renormalizability and maintaining perturbatively stable hierarchy of scales .
experimental constraints , mainly from flavor and higgs physics , limit the allowed parameter space and play an increasingly restrictive role in building models of susy breaking . however , the successes of supersymmetry do not rely on its simplest realization .
in fact , non - minimal realizations may ameliorate the flavor problem .
for example , dirac gauginos ( in contrast to majorana in the mssm ) forbid some of the couplings and often lead to additional suppression of contributions from loops with gauginos in flavor - changing processes .
such scenarios can be based on d - term supersymmetry breaking models @xcite or continuous r - symmetries @xcite .
a dirac gaugino requires additional fermionic degrees of freedom .
they can be provided by adding a chiral super - multiplet in the adjoint representation of the gauge group @xcite . here
we will present elements of the phenomenology of the scalar partner of the dirac gluino ( sgluon ) , as worked out in a recent paper @xcite [ and confronted with ref .
in the mssm gluinos are majorana fields with two degrees of freedom to match gluons in the color - octet vector super - multiplet . to provide the two additional degrees of freedom for dirac fields , the usual n=1 gluon / gluino vector super - multiplet @xmath0 may be paired with an additional n=1 color - octet chiral super - multiplet @xmath1 of extra gluinos and scalar @xmath2 fields to a vector hyper - multiplet of n=2 supersymmetry .
[ similarly , the electroweak sector , not to be discussed here , is supplemented by additional su(2)@xmath3 and u(1)@xmath4 super - multiplets . ] the n=2 mirror ( s)fermions are assumed to be very heavy in order to avoid chirality problems .
this hybrid model expands to n=2 only in the gaugino sector @xcite .
standard mssm @xmath5 and new gluinos @xmath6 couple minimally to the gluon field @xmath7 + \overline{\tilde{g } ' } \gamma^{\mu } [ { g}_{\mu } , { \tilde{g } ' } ] ) \ , , \label{eq : gluino}\ ] ] where @xmath8 denotes the qcd coupling , the fields being color - octet matrices ( @xmath9 with the gell - mann matrices @xmath10 ) ; @xmath5 and @xmath6 are two 4-component majorana spinor fields .
quark and squark fields interact only with the standard gluino , @xmath11 since only their hyper - multiplet partners ( assumed to be heavy ) couple to @xmath6 , as required by n=2 supersymmetry .
soft supersymmetry breaking generates masses for the gluino fields @xmath5 and @xmath6 , @xmath12 diagonal terms are induced by the individual majorana mass parameters @xmath13 and @xmath14 while an off - diagonal term corresponds to the dirac mass .
diagonalization gives rise to two majorana mass eigenstates , @xmath15 and @xmath16 , with masses @xmath17 and @xmath18 .
there are two limiting cases of interest : in the limit @xmath19 the standard mssm gluino is recovered ; in the limit of vanishing majorana mass parameters @xmath13 and @xmath14 with @xmath20 , the mixing is maximal and the two majorana gluino states are paired to a dirac state , @xmath21 dirac gluinos are characteristically different from majorana gluinos ; for detailed discussion we refer to @xcite . here
we present one example from @xcite in fig .
[ fig : d / m ] , where the partonic cross sections for different - flavor squark production @xmath22 mediated by the gluino t - channel exchange are plotted as a function of a dirac / majorana control parameter @xmath23 , assuming partonic center - of - mass energy @xmath24 gev , @xmath25 gev and @xmath26 gev .
here @xmath27 , @xmath28 with @xmath29 kept fixed along @xmath30 $ ] .
the parameter @xmath23 allows for a continuous transition from @xmath31 , where the mssm limit is reached with one majorana gluino ( the second being infinitely heavy ) , to @xmath32 that corresponds to two degenerate majorana fields combined to a dirac gluino .
for equal @xmath33 and @xmath34 masses , @xmath35 are non - zero in the majorana limit but vanish in the dirac , while @xmath36 reaches the same value in both limits .
the new gluinos are accompanied by a color - octet complex scalar field @xmath2 . in the simplest realization , in parallel to the degenerate majorana gluinos combined to the dirac gluino , we assume that the real and imaginary components of the scalar field @xmath2 are degenerate with the ( sgluon ) mass denoted by @xmath37 @xcite . at tree level , apart from the @xmath38 and @xmath39 couplings as required by the gauge symmetry , the sgluons couple to the gluino pair via the yukawa - type interaction @xmath40 where @xmath41 are the su(3)@xmath42 structure constants . when supersymmetry is broken spontaneously
, the dirac gluino mass generates a scalar coupling between @xmath2 and squark pair @xcite @xmath43 where @xmath44 and @xmath45 squarks contribute with opposite signs , as demanded by the general form of the super - qcd @xmath46-terms .
although sgluons are @xmath45-parity even , the above couplings imply that tree - level single production of @xmath2 s in gluon - gluon or quark - antiquark collisions is not possible .
only @xmath2 pairs can be produced in gluon collisions as well as in @xmath47 annihilation . since sgluons couple to squarks and gluinos , loop - induced couplings to the sm fields are of interest .
however , even at the one - loop level gluino loops do not contribute to the @xmath48 coupling as a consequence of bose symmetry since the coupling is even in momentum space but odd , @xmath49 , in color space . ) , while gluons always couple to the same pair , eq .
( [ eq : gluino ] ) . ] ( 300,60)(0,70 ) ( 0,95)[r]@xmath50 ( 15,95)[c ] @xmath2 ( 100,95)[c ] @xmath51 ( 25,95)(50,95)3 ( 50,95)(90,115)3 ( 90,75)(50,95)3 ( 90,115)(90,75)3 ( 125,115)[l ] @xmath52 ( 125,75)[l ] @xmath52 ( 90,115)(120,115)25 ( 90,75)(120,75)25 ( 150,95)[r ] @xmath2 ( 200,100)[c ] @xmath51 ( 155,95)(180,95)3 ( 200,95)(20,180,0)3 ( 200,95)(20,0,180)3 ( 220,95)(250,115)24 ( 220,95)(250,75)24 ( 255,115)[l ] @xmath52 ( 255,75)[l ] @xmath52 ( 300,60)(0,70 ) ( 0,95)[r]@xmath53 ( 15,95)[r ] @xmath2 ( 70,115)[c ] @xmath54 ( 95,95)[l ] @xmath51 ( 25,95)(50,95)3 ( 50,95)(90,115 ) ( 90,75)(50,95 ) ( 90,75)(90,115)3 ( 125,115)[l ] @xmath55 ( 125,75)[l ] @xmath56 ( 90,115)(120,115 ) ( 120,75)(90,75 ) ( 150,95)[r ] @xmath2 ( 200,115)[c ] @xmath51 ( 225,98)[l ] @xmath54 ( 155,95)(180,95)3 ( 180,95)(220,115)3 ( 220,75)(180,95)3 ( 220,75)(220,115 ) ( 220,115)(250,115 ) ( 250,75)(220,75 ) ( 255,115)[l ] @xmath55 ( 255,75)[l ] @xmath56 on the other hand , triangle diagrams involving squark / gluino lines ( fig .
[ fig : loops ] ) generate @xmath57 and @xmath58 couplings .
the interaction lagrangian however , eq . ( [ eq : sqq ] )
, implies that all @xmath44- and @xmath45-squark contributions to the couplings come with opposite signs so that they cancel each other for mass degenerate squarks .
in addition , the quark - antiquark coupling is suppressed by the quark mass as evident from general chirality rules .
sgluons have a large color charge and thus might be copiously produced at the lhc .
first we discuss their decay modes , and then production processes at the lhc .
sgluons will decay into different channels that include gluinos , squarks , gluons and quarks . at tree level
the @xmath2 particles can decay to a pair of dirac gluinos @xmath59 or into a pair of squarks , @xmath60 & = & \frac{3 \alpha_s m_{\sigma}}{4 } \beta_{\tilde{g}}\ , ( 1+\beta^2_{\tilde{g}})\ , , \\ \gamma [ \sigma \to \tilde{q}_a \tilde{q}_a^ * ] & = & \frac { \alpha_s |m_3^d|^2 } { 4 m_\sigma } \beta_{\tilde{q}_a}\ , , \label{eq : tree2}\end{aligned}\ ] ] where @xmath61 are the velocities of @xmath62 ( @xmath63 ) . if @xmath64 , one or both of these sparticles can be virtual .
squarks and gluinos will further cascade to sm particles and lsp .
loop - induced couplings will mediate decays into gluon or quark - antiquark pairs @xmath65|^2\,,\\
\gamma(\sigma \rightarrow q \bar q ) & = & \frac{9 \alpha_s^3 } { 128 \pi^2 } \frac{|m_3^d|^2 m_q^2}{m_\sigma}\ , \beta_q [ ( m^2_\sigma-4 m_q^2 ) |{\cal i}_s|^2 + m^2_\sigma \ , |{\cal i}_p|^2 ] \ , .
\label{loop - qq}\end{aligned}\ ] ] in eq.([loop - gg ] ) , @xmath66 and @xmath67 is the standard function from a squark circulating in the loop @xcite . in eq.([loop - qq ] ) , the effective scalar ( @xmath68 ) and pseudoscalar ( @xmath69 ) couplings take the form ( @xmath70 ) @xmath71 \,,\end{aligned}\ ] ] where @xmath72 , @xmath73 , @xmath74 , @xmath75 , and the squark / gluino denominators are ( @xmath76 ) + @xmath77 , + @xmath78 . + if @xmath79 and @xmath34 of a given flavor mix , the subscripts @xmath80 in equations above have to be replaced by @xmath81 labeling the squark mass eigenstates , and the contribution from this flavor is suppressed by the factor @xmath82 .
note that , because of the chirality structure , decays to light quarks are suppressed by the quark mass and that both loop - induced decays are absent if @xmath44 and @xmath45 squarks are degenerate the ordering between the tree - level and loop - induced decay modes depends on the values of various soft breaking parameters .
so long as gluino - pair decay channels are shut kinematically , even for small @xmath44-@xmath45 squark mass splitting the sgluon decays into two gluons , and to a @xmath83 pair if kinematically allowed , always dominate over tree - level off - shell four body decays @xmath84 and @xmath85 . increasing the gluino mass increases the @xmath86 coupling . as a result , the partial width into two gluons ( due to pure squark loops ) increases , while the @xmath83 partial width ( due to mixed squark gluino loops ) decreases since the increase of the @xmath87 couplings is over compensated by the gluino mass dependence of the propagators
of course , the tree
level , two
body decays of eqs .
( [ eq : tree1],[eq : tree2 ] ) will dominate if they are kinematically allowed .
well above all thresholds the partial width into gluinos always dominates : it grows @xmath88 while the partial width into squarks asymptotically scales like @xmath89 since the supersymmetry breaking @xmath90 coupling has mass dimension 1 , while the supersymmetric @xmath91 coupling is dimensionless .
the above qualitative features can be seen in fig .
[ fig : br ] , where the branching ratios for @xmath2 decays are plotted for two different squark / gluino mass hierarchies .
moderate mass splitting between the @xmath44 and @xmath45 squarks of the five light flavors , and somewhat greater for soft breaking @xmath92 masses have been assumed : @xmath93 , the off diagonal element of the squared @xmath92 mass matrix @xmath94 ; and the gluino is a pure dirac state , i.e. @xmath95 . sgluons can be produced singly in gluon - gluon collisions via squark loops ; the production via @xmath96 annihilation is negligible for light incoming quarks .
the partonic cross section is given by @xmath97 = \frac{\pi^2 } { m^3_\sigma } \gamma(\sigma \to gg)\,,\ ] ] where the partial width for @xmath98 decays is given in eq.@xmath99([loop - gg ] ) .
the resulting cross section for single @xmath2 production at the lhc is shown by the blue curves in fig.@xmath99[fig : lhcsigma ] ( with the lo cteq6l parton densities @xcite ) .
the solid curve has been calculated for the parameter set of the right frame of fig.@xmath99[fig : br ] , while the dashed one is for the benchmark point sps1a@xmath100 @xcite . since sps1a@xmath100 has a somewhat smaller gluino mass ( interpreted here as a dirac mass ) it generally leads to smaller cross sections for single @xmath2 production .
the signatures for single @xmath2 production are potentially very exciting .
however , the 2-gluon decay channel must be discriminated from the large sm background . on the other hand
, large production cross section in the gluon fusion would imply diminished decay rates to other channels , which in addition do not allow a direct reconstruction of @xmath37 .
detailed experimental simulations are needed to see if the single @xmath2 production can be detected as a resonance above the sm and mssm backgrounds .
sgluons can also be pair - produced in @xmath96 and @xmath101 processes , @xmath102 & = & \frac{4 \pi \alpha_s^2}{9s } \,\beta^3_\sigma \ , , \\
\sigma [ gg \to \sigma\sigma^{\ast } ] & = & \frac{15 \pi \alpha_s^2\beta_\sigma}{8s } [ 1 + \frac{34}{5}\ , \frac{m_\sigma^2}{s}-\frac{24}{5 } ( 1-\frac{m_\sigma^2}{s})\frac{m_\sigma^2}{s}\ , l_\sigma ] \,,\end{aligned}\ ] ] where @xmath103 is the invariant parton - parton energy , @xmath104 and @xmath105 are the mass and center - of - mass velocity of the @xmath2 particle , and @xmath106 .
the cross section for @xmath2-pair production at lhc , @xmath107 , is shown by the solid red curve in fig.@xmath99[fig : lhcsigma ] for the @xmath2-mass range between 500 gev and 2 tev . with values from several picobarn downwards ,
a sizable @xmath108 event rate can be generated .
as expected , due to large color charge of the sgluon , the @xmath108 cross section exceeds stop or sbottom - pair production ( red dashed line ) , mediated by a set of topologically equivalent feynman diagrams , by more than an order of magnitude . in the scenario of the right frame of fig .
[ fig : br ] , the single @xmath2 cross section can exceed the @xmath2-pair production cross section for @xmath109 tev ; taking @xmath110 , as in the left frame of fig .
[ fig : br ] , lead to a very small single @xmath2 production cross section . in general
one can not simultaneously have a large @xmath111 and a large @xmath112 .
with the exception of @xmath113 decays , all the @xmath2 decay modes give rise to signatures that should be easily detectable .
most spectacular signature results from @xmath114 decay , each @xmath2 decaying into at least four hard jets and two invisible neutralinos as lsp s .
@xmath2-pair production then generates final states with a minimum of eight jets with high sphericity and four lsp s . in such events the transverse momenta of the hard jets produced and the vector sum of the momenta of the four @xmath115 in the final state , which determines the measured missing transverse momentum @xmath116 , are markedly different from the corresponding mssm gluino or squark production with the same mass configurations @xcite .
other interesting final states are four - top states @xmath117 , which can dominate if @xmath118 and @xmath44-@xmath45 mixing is significant in the stop sector , and @xmath119 if @xmath120 .
these channels also lead to four lsps in the final state , plus a large number of hard jets . on the other hand , the @xmath121 final state , which can be the dominant mode if the two body decays into squarks and gluinos
are kinematically excluded , might allow the direct kinematic reconstruction of @xmath37 .
observation in addition @xmath122 final state might indicate a substantial mixing in the up - type squark sector .
models with dirac gauginos offer an interesting alternative to the mssm scenario . embedded into theories of extended supersymmetries
, they predict the presence of scalar particles in the adjoint representation of the gauge groups .
the color - octet scalars , sgluons , ( if kinematically accessible ) can be copiously produced at the lhc .
their signatures are distinctly different from the usual mssm topologies . depending on the masses of the particles involved , either multi - jet final states with high sphericity and large missing transverse momentum are predicted , or four top quarks should be observed in @xmath123 production . if the mass splitting between @xmath44 and @xmath45 squarks is not too small , loop induced single @xmath2 production may also have a sizable cross section . though this channel suffers in general from large backgrounds
, identifying the @xmath2 particle as a resonance in 2-gluon final states would truly be an exciting experimental observation .
supported in parts by the korea research foundation grant ( krf-2008 - 521-c00069 ) , the german bmbf grant ( 05ht6pda ) , the eu research networks `` universenet '' ( mrtn - ct-2006 - 035863 ) , `` forcesuniverse '' ( mrtn - ct-2004 - 005104 ) , `` the quest for unification '' ( mrtn - ct-2004 - 503369 ) , and the ec programme``particle physics and cosmology : the interface '' ( mtkd ct2005029466 ) .
jk is grateful to the theory division for the hospitality extended to him at cern , pmz to the insitute of theoretical physics e at rwth aachen .
100 p. j. fox , a. e. nelson and n. weiner , jhep * 0208 * ( 2002 ) 035 [ arxiv : hep - ph/0206096 ] .
y. nomura , d. poland and b. tweedie , nucl .
b * 745 * , 29 ( 2006 ) [ arxiv : hep - ph/0509243 ] .
l. j. hall and l. randall , nucl .
b * 352 * , 289 ( 1991 ) ; g. d. kribs , e. poppitz and n. weiner , arxiv:0712.2039 [ hep - ph ] .
p. fayet , nucl .
b * 113 * ( 1976 ) 135 ; l. lvarez - gaum and s. f. hassan , fortsch .
* 45 * ( 1997 ) 159 [ arxiv : hep - th/9701069 ] .
s. y. choi , m. drees , j. kalinowski , j. m. kim , e. popenda and p. m. zerwas , arxiv:0812.3586 [ hep - ph ] .
t. plehn and t. m. p. tait , arxiv:0810.3919 [ hep - ph ] and corrected version [ _ in preparation _ ] .
choi , m. drees , a. freitas and p.m. zerwas , phy . rev .
d * 78 * ( 2008 ) 095007 [ arxiv:0808.2410 [ hep - ph ] ] .
j. pumplin , d. r. stump , j. huston , h. l. lai , p. m. nadolsky and w. k. tung , jhep * 0207 * ( 2002 ) 012 [ arxiv : hep - ph/0201195 ] .
j. a. aguilar - saavedra _ et al .
_ , eur .
j. c * 46 * ( 2006 ) 43 [ arxiv : hep - ph/0511344 ] . | elements of the phenomenology of color - octet scalars ( sgluons ) , as predicted in the hybrid n=1/n=2 supersymmetric model , are discussed in the light of forthcoming experiments at the cern large hadron collider . |
in this paper we obtain new continuity results for tangent cones along interiors of limit geodesics in gromov - hausdorff limits of manifolds with lower ricci curvature bounds .
our main technical result is the following [ thm - main ] for any @xmath4 and @xmath5 , there exist @xmath6 , @xmath7 and @xmath8 such that the following holds : + suppose that @xmath9 is a complete @xmath1-dimensional riemannian manifold with @xmath10 and let @xmath11\to m$ ] be a unit speed minimizing geodesic .
then for any @xmath12 with @xmath13 and any @xmath14 there exist subsets @xmath15 ( @xmath16 ) with @xmath17 and a @xmath18-bilipschitz onto map @xmath19 , that is , @xmath20 is bijective and @xmath21 for any @xmath22 with @xmath23
. let @xmath24 ( @xmath16 ) be the renormalized volume measures at @xmath25 .
it s then obvious that under the assumptions of the theorem we have @xmath26 for some universal @xmath27 .
let @xmath28 where @xmath29 . by passing to a subsequence
we can assume that the renormalized volume measures @xmath30 on @xmath31 converge to a measure @xmath32 on @xmath0 @xcite .
for a point @xmath33 let @xmath34 be a tangent cone at @xmath35 corresponding to some @xmath36 .
again , up to passing to a subsequence we can assume that the renormalized measures @xmath37 converge to a renormalized measure @xmath38 on @xmath39 ( note that @xmath40 .
given @xmath41 we will call tangent cones @xmath42 @xmath16 together with the limit measures _ same scale _ if they come from the same rescaling sequence @xmath36 . using precompactenss and a standard arzela - ascoli type argument theorem [ thm - main ] easily yields [ cor - main ] for any @xmath4 and @xmath5
, there exist @xmath7 and @xmath8 such that the following holds : let @xmath43 where @xmath29 .
let @xmath11\to x$ ] be a unit speed geodesic which is a limit of geodesics in @xmath44 .
let @xmath32 be a renormalized limit volume measure on @xmath0 .
then for any @xmath12 with @xmath13 there exist subsets @xmath45 ( @xmath16 ) in the unit ball around the origin @xmath46 in the same scale tangent cones @xmath47 ( @xmath16 ) such that @xmath48 and there exists a map @xmath49 satisfying 1 .
@xmath50 is @xmath51-bilipschitz onto ; 2 .
@xmath52 in particular , @xmath53 ( @xmath54 ) is absolutely continuous with respect to @xmath55 ( @xmath56 ) . in @xcite colding and naber
show that under the assumptions of corollary [ cor - main ] same scale tangent cones along @xmath2 vary hlder continuously in @xmath57 .
corollary [ cor - main ] implies that hlder continuity of tangent cones also holds in measure - metric sense with respect to the renormalized limit volume measures on the tangent cones .
this does not follow from the results of @xcite which do not address measured continuity .
since same scale tangent cones do not need to exists for all @xmath57 for any given scaling sequence , we state the hlder continuity quanitatively using sturm distance @xmath58 which metrizes the measured gromov - hausdorff topology on the class of spaces in question ( * ? ? ?
* lemma 3.7 ) .
[ mm - cont-1 ] there exist @xmath59 such that the following holds .
let @xmath43 where @xmath29 .
let @xmath11\to x$ ] be a unit speed geodesic which is a limit of geodesics in @xmath44 .
then for any @xmath12 with @xmath13 we have that @xmath60 where @xmath61 are same scale tangent cones and @xmath62 is the unit ball around the vertex in @xmath63 .
note that bishop - gromov volume comparison implies that in corollary [ cor - main ] the set @xmath64 is @xmath65 dense in @xmath66 for @xmath16 and hence same scale tangent cones @xmath3 are hlder continuos in the pointed gromov - hausdorff topology .
of course , this is already known by @xcite .
let @xmath0 be a limit of @xmath1-manifolds with ricci curvature bounded below . recall that a point @xmath67 is called @xmath68-regular if _
every _ tangent cone @xmath69 is isometric to @xmath70 .
the collection of all @xmath68-regular points is denoted by @xmath71 .
( when the space @xmath0 in question is clear we will sometimes simply write @xmath72 ) .
the set of regular points of @xmath0 is the union @xmath73 the set of singular points @xmath74 is the complement of the set of regular points .
it was proved in @xcite that @xmath75 with respect to any renormalized limit volume measure @xmath32 on @xmath0 . moreover , by ( * ? ? ?
* theorem 4.15 ) , @xmath76 and @xmath32 is absolutely continuous on @xmath71 with respect to the @xmath68-dimensional hausdorff measure .
in particular , @xmath77 it was further shown in ( * ? ? ?
* theorem 1.18 ) that there exists unique integer @xmath68 such that @xmath78 altogether this implies that there exists unique integer @xmath68 such that @xmath79 moreover , it can be shown ( theorem [ thm - dim ] below ) that this @xmath68 is equal to the largest integer @xmath80 for which @xmath81 is non - empty .
following colding and naber we will call this @xmath68 _ the dimension of @xmath0 _ and denote it by @xmath82 .
( note that it is not known to be equal to the hausdorff dimension of @xmath0 in the collapsed case ) .
corollary [ cor - main ] immediately implies [ thm - main2 ] under the assumptions of corollary [ cor - main ] the dimension of same scale tangent cones @xmath83 is constant for @xmath84 . for @xmath85 sufficiently close to @xmath86 let @xmath87 ( @xmath16 ) , @xmath49 be provided by corollary [ cor - main ]
let @xmath88 .
suppose @xmath89 , say @xmath90 . by using and corollary [ cor - main ] ( ii ) we can assume that @xmath91 . by above this
means that @xmath92 . since @xmath50 is lipschitz we have @xmath93 .
since @xmath94 is absolutely continuous with respect to the @xmath95-dim hausdorff measure on @xmath96 and @xmath97 this implies that @xmath98 .
this is a contradiction since @xmath99 .
note that a cusp `` can exist in the limit space of manifolds with lower ricci curvature bound , for example , a horn ( * ? ? ?
* example 8.77 ) .
theorem [ thm - main2 ] indicates that a ' ' cusp " can not occur in the interior of limit geodesics .
in particular , it provides a new way to rule out the trumpet ( * ? ? ?
* example 5.5 ) and its generalizations ( * ? ? ?
* example 1.15 ) .
moreover , it shows that the following example can not arise as a gromov - hausdorff limit of manifolds with lower ricci bound , even through the tangent cones are hlder ( in fact , lipschitz ) continuous along the interior of geodesics .
this example can not be ruled out by previously known results .
let @xmath100 . ] then @xmath101 for @xmath102 and @xmath103 .
let @xmath0 be the double of @xmath104 along its boundary . then all points not on the @xmath35-axis are in @xmath105 and along the @xmath35-axis we have that for @xmath102 , @xmath106 ( double of @xmath107 ( i.e. it s a cone @xmath108 ) degenerating to @xmath109 .
so @xmath110 but @xmath111 for @xmath102 .
lastly , any segment of the geodesic @xmath112 is unique shortest between its end points and hence it s a limit geodesic if @xmath104 is a limit of manifolds with @xmath113 .
hence theorem [ thm - main2 ] is applicable to @xmath2 and therefore @xmath0 is not a limit of @xmath1-manifolds with @xmath113 .
note that one can further smooth out the metric on @xmath0 along @xmath114axis@xmath115 to obtain a space @xmath116 with similar properties but which in addition is a smooth riemannian manifold away from the @xmath35-axis .
in particular @xmath116 is non - branching .
next we want to mention several semicontinuity results about the colding - naber dimension which further suggest that this notion is a natural one .
let @xmath117 be the space of pointed gromov - hausdorff limits of manifolds with @xmath118 .
recall the following notions from @xcite let @xmath119 . * @xmath120 such that some tangent cone @xmath39 splits off isometrically as @xmath121 .
* @xmath122 such that every tangent cone @xmath39 splits off isometrically as @xmath121 . *
@xmath123 such that there exist @xmath124 , @xmath104 and @xmath125 such that @xmath126 . by
* lemma 2.5 ) there exists @xmath127 such that if @xmath128 for some @xmath129 then @xmath130 for all sufficiently small @xmath131 .
suppose @xmath132 , @xmath133 and @xmath134 .
then @xmath135 which obviously implies that @xmath136 for all large @xmath137 as well . by
above this implies that @xmath138 for all large @xmath137 .
this together with yields the following result of honda proved in ( * ? ? ?
* prop 3.78 ) using very different tools .
* prop 3.78 ) [ thm - dim - semicont ] let @xmath139 and @xmath140 .
. then @xmath142 .
in other words , the dimension function is lower semicontinuous on @xmath117 with respect to the gromov - hausdorff topology .
this theorem , applied to the convergence @xmath143 for @xmath144 , immediately gives the following result which also directly follows from ( * ? ? ?
* prop 3.1 ) and .
[ thm - dim ] let @xmath119 .
then @xmath82 is equal to the largest @xmath68 for which @xmath145 . another immediate consequence of theorem
[ thm - dim - semicont ] is the following ( * ? ? ?
* prop 3.78 ) [ cor - dim - tcone ] let @xmath119 .
then for any @xmath33 and any tangent cone @xmath39 it holds that @xmath146 it is obvious from that for any @xmath119 we have @xmath147 .
however , as was mentioned earlier , the following natural question remains open .
let @xmath119 .
is it true that @xmath148 ?
let @xmath149\to m^n$ ] be a unit speed shortest geodesic in an @xmath1-manifold with @xmath150 .
in @xcite colding and naber constructed a parabolic approximation @xmath151 to @xmath152 given as the solution of the heat equation with initial conditions given by @xmath152 , appropriately cut off near the end points of @xmath2 and outside a large ball containing @xmath2 .
they showed that @xmath153 provides a good approximation to @xmath154 on an @xmath131-neighborhood of @xmath155}$ ] .
in particular , they showed that @xmath156 for all @xmath157 .
they used this to show that for any @xmath158 most points in @xmath159 remain @xmath131-close to @xmath2 under the reverse gradient flow of @xmath160 for a definite time @xmath161 . in section
[ sec : grad - flow ] we show that the same holds true for the reverse gradient flow @xmath162 of @xmath153 .
next , the standard weak type 1 - 1 inequality for maximum function applied to the inequality ( [ cn.hess.int ] ) implies that @xmath163 as well .
this implies that for every @xmath35 in a subset @xmath164 in @xmath159 of almost full measure the integral @xmath165 is small ( see estimate ( [ e.max.small ] ) ) .
using a small modification of a lemma from @xcite this implies that for any such point @xmath35 and _ any _
@xmath166 most points in @xmath167 remain @xmath168-close to @xmath169 for all @xmath170 under the flow @xmath162 .
this then easily implies that @xmath162 is bilipschitz on @xmath164 using bishop - gromov volume comparison and triangle inequality .
we are very grateful to aaron naber for helpful conversations and to shouhei honda for bringing to our attention results of @xcite and @xcite .
we are also very greateful to the referee for pointing out that our results imply corollary [ mm - cont-1 ] .
in this section we will list most of the technical tools needed for the proof of theorem [ thm - main ] . throughout the rest of the paper , unless indicated otherwise , we will assume that all manifolds @xmath171 involved are @xmath1-dimensional complete riemannian satisfying @xmath172 we will need the following result of cheeger and colding : ( * ? ? ?
* theorem 2.11)[seg - ineq ] given @xmath1 and @xmath173 there exists @xmath174 such that the following holds .
let @xmath175 be a nonnegative measurable function .
then for any @xmath176 and @xmath177 it holds @xmath178 where @xmath179 denotes a minimal geodesic from @xmath180 to @xmath181 .
let @xmath182\to m$ ] be a minimizing unit speed geodesic with @xmath183 where @xmath184 . to simplify notations and exposition from now on
we will assume that @xmath185 .
let @xmath186 , and let @xmath187 be the excess function .
the following result is a direct consequence of ( * ? ? ?
* theorem 2.8 ) and , as was observed in @xcite , using the fact that @xmath188 it immediately implies the abresch - gromoll estimate @xcite .
* theorem 2.8 ) [ thm - gen - ab - gr ] there exist @xmath189 , @xmath190 such that for any @xmath191 , @xmath192 it holds @xmath193 fix @xmath194 and let @xmath195 be parabolic approximations to @xmath196 constructed in @xcite .
they are given by the solutions to the heat equations @xmath197 for appropriately constructed cutoff function @xmath198 .
we will need the following properties of @xmath199 established in @xcite .
* lemma 2.10 ) [ lem : lap - comp ] there exists @xmath189 such that @xmath200 ( * ? ? ?
* theorem 2.19 ) [ t : mainregthm1 ] there exist @xmath201 such that for all @xmath202 there exists @xmath203 $ ] such that the following properties are satisfied 1 .
@xmath204 for any @xmath205 with @xmath206 [ e : emain1 ] 2 .
@xmath207.[e : emain2 ] 3 . @xmath208 .
[ e : emain3 ] 4 . @xmath209 .
[ e : emain4 ] we will need the following lemma ( cf .
* lemma 3.4 ) ) .
[ lem : dist - growth ] let @xmath0 be a smooth vector field on @xmath210 and let @xmath211 be smooth curves .
let @xmath212 .
then @xmath213 where @xmath214\to m$ ] is a shortest geodesic from @xmath215 to @xmath216 .
here @xmath217 means the norm of the full covariant derivative of @xmath0 i.e. norm of the map @xmath218 .
in particular , if @xmath219 is smooth and @xmath220 , then @xmath221 the lemma easily follows from the first variation formula for distance functions and the triangle inequality .
let @xmath222 be a nonnegative function .
consider the maximum function @xmath223 for @xmath224 $ ] .
we ll set @xmath225 .
the following lemma is well - known @xcite .
[ lem : mf - est][lem : weak11 ] suppose @xmath9 has @xmath226 and let @xmath222 be a nonnegative function
. then the following holds . 1 . if @xmath227 with @xmath228 then @xmath229 is finite almost everywhere .
2 . if @xmath230 then @xmath231 for any @xmath232 .
3 . if @xmath227 with @xmath233 then @xmath234 and @xmath235 .
this lemma easily generalizes to functions defined on subsets as follows : [ cor : mx ] let @xmath236 and @xmath237 be measurable .
let @xmath238 be measurable such that @xmath239 where @xmath240 . here
@xmath241 denotes the @xmath242-neighborhood of @xmath243 .
then @xmath244 let @xmath245 .
obviously , @xmath246 for any @xmath247 .
the result follows by applying lemma [ lem : mf - est ] ( iii ) to @xmath248 .
let @xmath162 be the _ reverse _ gradient flow of @xmath249 ( i.e. the gradient flow of @xmath250 ) and let @xmath251 be the reverse gradient flow of @xmath160 .
we first want to show that for most points @xmath252 we have that @xmath253 for all @xmath254 and @xmath255 $ ] for some uniform @xmath256 .
note that this ( and more ) is already known for @xmath251 by @xcite .
following colding - naber we use the following [ defn : a - b ] for @xmath257 define the set @xmath258 .
similarly , we define @xmath259 .
an important technical tool used to prove the main results of @xcite is the following ( * ? ? ?
* proposition 3.6)[prop - dist - grad ] there exist @xmath260 and @xmath261 such that if @xmath254 and @xmath262 then @xmath263 as in theorem [ t : mainregthm1 ] we have @xmath264 unlike colding - naber we prefer to work with the gradient flow of the parabolic approximation @xmath265 rather than the gradient flows of @xmath196 , because the gradient flow of @xmath265 provides better distance distortion estimates since in that case the two terms outside the integral in lemma [ lem : dist - growth ] vanish and the resulting inequality scales better in the estimates involving maximum function ( see lemma [ lem : dtr ] below ) .
therefore , our first order of business is to establish the following lemma which says that proposition [ prop - dist - grad ] holds for the gradient flow of @xmath266 as well : [ lem : distort1 ] there exists @xmath267 and @xmath268 such that if @xmath269 and @xmath270 then @xmath271 we have @xmath272 and @xmath273 the proof of proposition [ prop - dist - grad ] uses bootstrapping in @xmath274 starting with infinitesimally small ( depending on @xmath210 ! ) @xmath131 ( cf .
lemma [ lem : dtr ] below ) for which the claim easily follows from bochner s formula applied to @xmath160 along @xmath2 .
we do nt utilize bootstarpping in @xmath131 and instead use that the result has already been established for the gradient flow of @xmath275 .
of course , we only need to prove the second inequality as the first one holds by proposition [ prop - dist - grad ] for some @xmath276 . by possibly making @xmath277 smaller
we can ensure that it satisfies theorem [ t : mainregthm1 ] .
let @xmath278 be small ( how small it will be chosen later ) .
let @xmath279 we wish to show that @xmath280 contains @xmath281 $ ] for some uniform @xmath282 .
obviously @xmath280 is open in @xmath281 $ ] so it s enough to show that it s also closed . to establish this it s enough to show that if @xmath283 and @xmath284 then @xmath285 .
for any @xmath286 we define @xmath287 to be the characteristic function of the set @xmath288 . the same argument as in @xcite shows that @xmath289 indeed , we have @xmath290 where the last inequality follows from the fact that @xmath291 by lemma [ lem : lap - comp ] and hence the jacobian of @xmath162 satisfies @xmath292 similar inequality holds for @xmath251 by bishop - gromov volume comparison .
since @xmath293 , @xmath294 by definition , by the segment inequality ( theorem [ seg - ineq ] ) we have @xmath295 \int_{b_{5r}(\gamma(t - s))}|\operatorname{hess}_h| \notag\\ & \leq c(n,\delta)\,r\,\operatorname{vol}(b_{5r}(\gamma(t - s)))\int_{b_{5r}(\gamma(t - s))}|\operatorname{hess}_h| \notag\\ & = c(n,\delta)\,r\,\operatorname{vol}(b_{5r}(\gamma(t - s)))^2\fint_{b_{5r}(\gamma(t - s))}|\operatorname{hess}_h| \notag\\ & \le c(n,\delta)\,r\,\operatorname{vol}(b_{r}(\gamma(t - s)))^2\fint_{b_{5r}(\gamma(t - s))}|\operatorname{hess}_h|\ , , \notag\end{aligned}\ ] ] where the last inequality follows by bishop - gromov .
thus , @xmath296 dividing by @xmath297 we get . by (
* cor 3.7 ) we have that @xmath298 for some universal @xmath299 and therefore @xmath300 let @xmath301 then by and theorem [ t : mainregthm1 ] we have that @xmath302 let @xmath303 and for @xmath304 let us define @xmath305 here @xmath306 is small and chosen first .
then @xmath307 is chosen later depending on @xmath308 . by (
* , equations ( 115 ) ( 117 ) ) we can assume that @xmath309 if @xmath310 for some universal @xmath311 .
therefore , by construction we have that @xmath312 and hence @xmath313 we choose @xmath308 so that @xmath314 in and .
let @xmath315 ( this intersection is non - empty for small @xmath316 by bishop - gromov ) and let @xmath317 .
we will fix @xmath318 satisfying the above conditions from now on .
we claim that then @xmath319 .
indeed @xmath320 for all @xmath321 since @xmath322 .
so by the triangle inequality it s enough to show that @xmath323 for any @xmath324 .
let @xmath325 .
this set is obviously open and connected in @xmath326 $ ] .
we claim that @xmath327 $ ] .
let @xmath328 .
note that for any @xmath329 we have that @xmath330 .
therefore , by and for any @xmath329 we have @xmath331 if @xmath332 is chosen small enough . next ,
recall that by ( * ? ? ?
* lemma 2.20(3 ) ) we have that for any @xmath252 , @xmath333 further , by theorem [ thm - gen - ab - gr ] we know that @xmath193 therefore , without losing generality by making the sets @xmath334 slightly smaller we can assume that for any @xmath304 we have @xmath335 thus , for all @xmath304 we have @xmath336 if @xmath337 is small enough .
therefore , by lemma [ lem : dist - growth ] and using and we get that @xmath338 by the triangle inequality , @xmath339 by continuity the same holds for @xmath340 and hence @xmath341 .
thus @xmath342 is both open and closed in @xmath326 $ ] and therefore @xmath327 $ ] . unwinding this
further we see that this means that @xmath343 . therefore , by @xmath344 when @xmath308 was chosen small enough so that @xmath345 .
hence @xmath285 .
therefore , @xmath280 is both open and closed and @xmath346 .
the proof of lemma [ lem : distort1 ] shows that @xmath347 for appropriately chosen @xmath307 depending on @xmath308 .
moreover , the proof shows that @xmath307 can be chosen to be of the form @xmath348 for some @xmath311 .
in view of this means that the conclusion can be strengthened as follows ( cf .
@xcite ) [ lem : distort2 ] for every @xmath349 and @xmath350 that there exists @xmath351 such that the set @xmath352 satisfies @xmath353 moreover , @xmath354 can be chosen to be of the form @xmath355 for some @xmath311 .
when @xmath308 is sufficiently small this means that most points in @xmath159 remain close to the geodesic @xmath2 under the flow @xmath162 .
also , provided @xmath345 using bishop - gromov , the above lemma , and give the following [ lem - comp - av1 ] let @xmath356 be a nonnegative measurable function .
then @xmath357 for any @xmath170 as in lemma [ lem : distort2 ] .
it is obvious that all results of this section concerning the flow of @xmath358 are also true for the flow of @xmath359 .
the goal of this section is to prove the following equivalent version of theorem [ thm - main ] suppose @xmath9 is complete with @xmath10 and let @xmath11\to m^n$ ] be a unit speed minimizing geodesic
. then for any @xmath12 with @xmath13 and any @xmath14 there are subsets @xmath15 ( @xmath16 ) such that @xmath362 and there exists a @xmath363-bilipschitz map @xmath19 for some @xmath364 .
corollary [ cor : mx ] essentially means that all estimates involving integrals of @xmath366 from the previous section remain true for @xmath367 .
in particular , for @xmath277 as in theorem [ t : mainregthm1 ] and any @xmath368 we have @xmath369 by corollary [ cor : mx ] this implies that @xmath370 where @xmath371 .
it is clear that all results from the previous section work for this @xmath265 as well as @xmath372 .
therefore , everywhere in the previous section where we used we could have used instead .
indeed , we have for any @xmath373 , @xmath374 [ lem : dtr ] given @xmath232 , there exists ( explicit ) @xmath381 such that the following holds .
suppose @xmath9 has @xmath236 and @xmath382 is a vector field with compact support , which depends on time but is piecewise constant in time .
let @xmath383 be the integral curve of @xmath382 with @xmath384 and assume that @xmath385 on @xmath386 for all @xmath387 $ ] .
let @xmath388 be the flow of @xmath382 .
define the distortion function @xmath389 of the flow on scale @xmath131 by the formula @xmath390 put @xmath391 .
then for any @xmath392 we have @xmath393 and there exists @xmath394 such that @xmath395 and @xmath396 . in @xcite lemma [ lem : dtr ]
is stated for divergence free vector fields .
we want to apply it to @xmath401 which is not divergence free .
however , recall that by lemma [ lem : lap - comp ] it does satisfy @xmath402 and hence the jacobian of its flow map satisfies @xmath403 pointwise .
the proof given in @xcite goes through verbatim with a straightforward change in one place using instead of the flow of @xmath0 being volume - preserving .
we include the proof here for reader s convenience .
notice that all estimates are trivial if @xmath404 .
therefore it suffices to prove the statement with a universal constant @xmath405 for all @xmath406 .
we put @xmath407 and determine @xmath408 in the process .
we proceed by induction on the size of @xmath131 .
suppose the result holds for some @xmath411 .
it suffices to prove that it then holds for @xmath131 . by induction assumption
we know that for any @xmath57 there exists @xmath412 such that for any @xmath413 $ ] we have @xmath414 and @xmath415 where we used @xmath416 in the inequality .
this easily implies that @xmath417 are comparable for all @xmath57 .
more precisely , for any @xmath418 $ ] we have that @xmath419 with a computable universal @xmath420 . put @xmath421 @xmath422 as @xmath423 is monotonously increasing , we deduce that if @xmath424 , then @xmath425 .
since @xmath426 and @xmath162 satisfies @xmath427 it follows @xmath428 where we used that @xmath429 .
we would like to point out that using @xmath430 instead of @xmath431 ( which is true for flows of harmonic maps ) in the above inequality is the only place where the proof of lemma [ lem : dtr ] differs from the proof of ( * ? ? ?
* lemma 3.7 ) . ) if @xmath215 is not in the cut locus of @xmath216 and @xmath432\rightarrow m$ ] is a minimal geodesic between @xmath215 and @xmath216 , then by lemma [ lem : dist - growth ] @xmath433 combining the last two inequalities with the segment inequality we deduce note that the choice of the constant @xmath435 can be made explicit and _ independent _ of the induction assumption .
we deduce @xmath436 and thus the subset @xmath437 satisfies @xmath438 it is elementary to check that @xmath439$\,.}\end{aligned}\ ] ] then arguing as before we estimate that @xmath440 using @xmath441 and the volume estimate ( [ vol est1 ] ) this gives @xmath442 this completes the induction step with @xmath443 and @xmath444 . in order to remove the restriction @xmath445 one
can just increase @xmath405 by the factor @xmath446 , as indicated at the beginning .
since @xmath452 in we have @xmath457 and therefore by the definition of @xmath458 for all @xmath459 it holds @xmath460 this means that we can apply corollary [ cor - good - pts ] at all such points with @xmath461 and @xmath462 .
let @xmath463 be the constant provided by corollary [ cor - good - pts ] .
let @xmath464 .
let @xmath465 and set @xmath466 . using corollary [ cor - good - pts ] , by bishop - gromov
we can be assured that @xmath467 provided @xmath468 and @xmath469 is chosen small enough .
pick any @xmath470 .
likewise , using bishop - gromov we can find @xmath471 and @xmath472 .
therefore , by corollary [ cor - good - pts ] we have @xmath473 and @xmath474 summing up the above inequalities and using the triangle inequality we get @xmath475 this shows that @xmath162 is @xmath476-lipschitz on @xmath458 .
next , let us show that it s bilipschitz .
as before , using bishop - gromov , we can find @xmath477 and @xmath478 .
this implies that @xmath479 and @xmath480 by the triangle inequality this yields @xmath481 which finally proves part of lemma [ lem - bilip ] .
therefore , by bishop - gromov we can find @xmath483 .
since by construction we have that @xmath484 , by lemma [ lem : distort2 ] we have that @xmath485 . by part , for any @xmath486 we also have that @xmath487 . applying the triangle inequality
we get that @xmath488 . this yields and finishes the proof of lemma [ lem - bilip ] and hence of theorem [ thm - main1 ] . | we show that if @xmath0 is a limit of @xmath1-dimensional riemannian manifolds with ricci curvature bounded below and @xmath2 is a limit geodesic in @xmath0 then along the interior of @xmath2 same scale measure metric tangent cones @xmath3 are hlder continuous with respect to measured gromov - hausdorff topology and have the same dimension in the sense of colding - naber .
plus3pt minus3pt plus3pt minus3pt |
so far , we have shown how the susy / non - susy duality proposed by aganagic et al . can have a corresponding type iia brane picture .
apart from conventional ones , where anti - branes are wrapped on a cy by hand , the setup here involves changing the orientation of local two - cycles through a varying background ns - flux . this dose work because @xmath10-field gives a khler moduli @xmath11 of arbitrary sign to shrinking two - cycles and hence controls their flops . on type iia side ,
we interpret this background as two crossing ns5-branes where @xmath12-branes appear naturally for flipped orientations .
consequently , simultaneous presence of d4- and @xmath12-branes soon suggests a way to realize various kinds of susy / non - susy vacua via adjusting parameters the ns - flux contains .
moreover , curved ns5-branes on @xmath13-plane with @xmath1 susy correspond to what has been known as the @xmath0 seiberg - witten theory .
one can further add fi parameters to partially break @xmath1 down to @xmath2 . resorting to type iia brane pictures
, we see this process is clearly visualized in figure [ partial susy ] .
the final @xmath2 vacuum is arrived at once the tree - level superpotential @xmath14 takes the form of @xmath15 .
we also considered m - theory lift . without @xmath16 perturbation ,
the m - theory curve itself is either a degenerated seiberg - witten curve on @xmath13-plane or a loop equation of dv matrix model on @xmath17-plane . though adding higher @xmath16 terms has no effect on the planar loop equation , we find that @xmath2 effective superpotential which involves @xmath18-cycles on @xmath13-plane gets modified
. in particular , it seems that the above partial susy breaking process can be described by deforming one given m - theory curve in order to incorporate quantum effects .
it is thus of interest to compare this observation with field theory results found in @xcite .
we leave these problems to future works .
we are grateful to k. hashimoto for helpful discussions .
ko would like to thank t. higaki and k. maruyoshi for useful comments .
ko is supported in part by grant - in - aid for scientific research ( no.19740120 ) from the ministry of education , culture , sports , science and technology .
tst is supported in part by the postdoctoral program at riken .
l. hollands , j. marsano , k. papadodimas and m. shigemori , _ nonsupersymmetric flux vacua and perturbed n=2 systems , " _ arxiv:0804.4006 [ hep - th ] h. ooguri and y. ookouchi , _ `` meta - stable supersymmetry breaking vacua on intersecting branes , '' _ phys
. lett .
b * 641 * , 323 ( 2006 ) [ arxiv : hep - th/0607183 ] a. giveon and d. kutasov , _ brane dynamics and gauge theory , " _ rev .
phys . * 71 * 983 ( 1999 ) [ arxiv : hep - th/9802067 ] a. hanany and e. witten , _ type iib superstrings , bps monopoles , and three - dimensional gauge dynamics , " _ nucl .
b * 492 * 152 ( 1997 ) [ arxiv : hep - th/9611230 ] r. dijkgraaf and c. vafa , nucl .
b * 644 * 3 ( 2002 ) [ arxiv : hep - th/0206255 ] ; nucl .
b * 644 * 21 ( 2002 ) [ arxiv : hep - th/0207106 ] ; [ arxiv : hep - th/0208048 ] f. cachazo , k. intriligator , and c. vafa , _
a large n duality via a geometric transition , " _ nucl .
b * 603 * 3 ( 2001 ) [ arxiv : hep - th/0103067 ] i. antoniadis , h. partouche and t. r. taylor , phys .
b * 372 * , 83 ( 1996 ) [ arxiv : hep - th/9512006 ] ; i. antoniadis and t. r. taylor , fortsch .
phys . *
44 * , 487 ( 1996 ) [ arxiv : hep - th/9604062 ] s. ferrara , l. girardello and m. porrati , phys . lett .
b * 376 * , 275 ( 1996 ) [ arxiv : hep - th/9512180 ] h. partouche and b. pioline , nucl .
* 56b * , 322 ( 1997 ) [ arxiv : hep - th/9702115 ] k. fujiwara , h. itoyama and m. sakaguchi , prog .
phys . * 113 * , 429 ( 2005 ) [ arxiv : hep - th/0409060 ] ; nucl .
b * 723 * , 33 ( 2005 ) [ arxiv : hep - th/0503113 ] ; k. fujiwara , nucl . phys .
b * 770 * , 145 ( 2007 ) [ arxiv : hep - th/0609039 ] | we study the susy / non - susy duality proposed by aganagic et al . from type iia string and m - theory perspectives . we find that our brane configuration generalizes the so - called @xmath0 seiberg - witten theory on the one hand , and provides a way to realize non - susy vacua by intersecting ns5-branes on the other hand .
we also argue how the partial susy breaking from @xmath1 down to @xmath2 can be clearly visualized through the brane picture .
[ cols= " < " , ] finally , we comment on the partial susy breaking from @xmath1 to @xmath2 through the above brane picture .
the partial susy breaking is discussed in @xcite for abelian gauge group .
the non - abelian generalization is well investigated and established by authors of @xcite . according to these early works ,
@xmath1 theory is firstly perturbed by introducing a general prepotential of the form and its susy is broken down to @xmath2 thereof upon adding fayet - iliopoulos ( fi ) parameters . as mentioned throughout this paper ,
the prepotential @xmath3 describes how ns5-branes get deformed in @xmath4-space , see figure [ partial susy ] ( a ) .
three fi parameters correspond to the relative position of two ns5-branes in @xmath5-space .
henceforth , turning on fi parameters means that two ns5-branes are separated from each other in @xmath5-space .
in addition , a new direction by which the bare coupling constant is measured should be defined due to the presence of fi parameters . in figure [ partial susy ] ( a ) , initial positions of d4-branes are not fixed .
but if d4-branes still remain at their initial positions , they become non - parallel when the curved ns5-brane is pulled along @xmath6 , see figure [ partial susy ] ( b ) .
in other words , susy can no longer be maintained for an off - shell choice of @xmath7 vev .
to recover susy , d4-branes should be re - distributed appropriately at critical loci , i.e. @xmath8 ( on - shell condition ) .
now , since susy is recovered to @xmath2 , we can as well recognize the tree - level superpotential as @xmath9 with coefficients rescaled suitably , see figure [ partial susy ] ( c ) . to this end , the partial susy breaking mechanism can thus be understood pictorially from the _ extended _ brane configuration .
the above argument is valid only for the classical ( uv ) theory . in order to extend this picture to the full quantum theory
, we need to replace the ns5/d4 system with a mqcd curve . pulling out
one ns5-brane then corresponds to deforming the curve .
the projective information of the mqcd curve contains both the _ extended _ sw curve and loop equation ( or , alternatively , generalized konishi anomaly equation ) of @xmath2 theory .
therefore , it is interesting to see how these aspects transform according to the partial susy breaking deformation of the curve . |
the purpose of this paper is to derive the continued fraction form of generating function of some new @xmath0-analogs of eulerian numbers @xmath1 introduced by lauren williams @xcite building on work of alexander postnikov @xcite and to link them to generalized patterns .
these numbers are related to the number of alignments and weak exceedances of permutations and come from the enumeration of totally positive grassmann cells .
we refer to @xcite and @xcite for details .
we start by some definitions .
let @xmath2 be a permutation of @xmath3 $ ] .
then let * @xmath4 ; @xmath5 .
* @xmath6 ; @xmath7 . * @xmath8 ; @xmath9 . for example
let @xmath10 then @xmath11 , @xmath12 , and @xmath13 .
+ we can also define these parameters using the permutation diagram . if @xmath14 is a permutation of @xmath3 $ ] .
we draw a line and put the numbers from 1 to @xmath15 and we draw an edge from @xmath16 to @xmath17 above the line if @xmath18 and under the line otherwise .
then @xmath19 is the number of pairs of nested edges above ( resp . under ) the line and @xmath20
is the number of pairs of edges such that one is above the line and the other under and such that their support do not intersect . ]
the permutation diagram of @xmath10 is on figure [ perm ] .
the pairs of edges contributing to @xmath21 are @xmath22 , @xmath23 and @xmath24 .
the pair of edges contributing to @xmath25 is @xmath26 .
the pairs of edges contributing to @xmath20 are @xmath27 and @xmath28 .
the number of alignements of a permutation @xmath14 is equal to @xmath29 this definition looks a bit different but is equivalent to the definition in @xcite .
the number of weak exceedances of a permutation @xmath14 is the cardinal of the set @xmath30 . in @xcite
the following result was proved : the number of permutations of @xmath3 $ ] with @xmath31 weak exceedances and @xmath32 alignments is the coefficient of @xmath33 $ ] in @xmath34_q^n q^{k(i-1)}\left ( { n\choose i}q^{k - i}+{n\choose i-1}\right).\ ] ] where @xmath35_q=1+q+\ldots + q^{n-1}$ ] .
these numbers have the property that if @xmath36 they are the eulerian numbers , if @xmath37 they are the narayana numbers and if @xmath38 they are the binomial coefficients .
see @xcite for details .
lauren williams proved in @xcite that @xmath39 is equal to @xmath40x + [ i]xy)}.\ ] ] here we exhibit the generating function @xmath41 in a continued fraction form .
@xmath42 with @xmath43_q+[n]_q$ ] and @xmath44_q^2 $ ] and @xmath3_q=1+q+\ldots + q^{n-1}$ ] .
[ big ] to prove this theorem we introduce the notion of crossings .
let * @xmath45 ; @xmath46 .
* @xmath47 , @xmath48 .
we define the number of crossings of a permutation @xmath14 to be @xmath49 .
we can again define those parameters using the permutation diagram .
@xmath50 is the number of pairs of edges above the line that intersect and @xmath51 is the number of pairs of edges under the line that cross .
for @xmath10 we have @xmath52 and @xmath53 the pairs of edges contributing to @xmath54 are @xmath55 and @xmath56 .
the pair of edges contributing to @xmath57 is @xmath58 . + the proof of the theorem is direct with the following two propositions .
the coefficient @xmath59 $ ] in @xmath60 is the number of permutations of @xmath3 $ ] with @xmath31 weak exceedances with @xmath32 crossings .
[ main1 ] for any permutation with @xmath31 weak exceedances the number of crossings plus the number of alignments is @xmath61 [ main ] we prove these two propositions in section 2 .
the proof of the first one uses a bijection of foata and zeilberger between permutations and weighted bicolored motzkin paths @xcite .
the proof of the second one uses basic properties of permutations .
then in section 3 we use a bijection due to franon and viennot @xcite and the bijection of foata and zeilberger and use results from @xcite to link with generalized patterns .
one result is : there is a one - to - one correspondance between permutations with @xmath31 weak exceedances and @xmath62 alignments and permutations with @xmath63 descents and @xmath32 occurrences of the generalised pattern @xmath64 .
[ pattern ] the link between alignments and generalized patterns was conjectured by steingrimsson and williams @xcite . in section 4
we generalize to the case of decorated permutations . in section 5
we show how these numbers appear naturally in the stationary distribution of the asep model .
we use a bijection of foata and zeilberger @xcite between permutations and weighted bicolored motzkin paths . we could also use the bijection of biane @xcite .
we refer to @xcite for a compact definition of these bijections .
a bicolored motzkin path of length @xmath15 is a sequence @xmath65 such that @xmath66 for @xmath67 such that if @xmath68 then @xmath69 for @xmath67 and @xmath70 . to any permutation
@xmath14 we associate a pair @xmath71 made of a bicolored motzkin path @xmath72 and a weight @xmath73 .
the path is created using the following rules : * @xmath74 if @xmath75 and @xmath76 * @xmath77 if @xmath75 and @xmath78 * @xmath79 if @xmath80 and @xmath76 * @xmath81 if @xmath80 and @xmath78 the weight is created using the following rules : * @xmath82 if @xmath83 . *
@xmath84 if @xmath85 .
for example @xmath86 gives the path @xmath87 and the weight @xmath88 .
+ we now need if @xmath18 then @xmath89 and if @xmath80 then @xmath90 * proof . *
it is easy to prove by induction that @xmath91 using the definitions given in section 1 , if @xmath18 then @xmath92 and if @xmath80 then @xmath93 this implies the result .
let @xmath94 be the set of pairs @xmath71 obtained from permutations of @xmath3 $ ] .
using the machinery developed in @xcite , we get directly that if : @xmath95 then @xmath96 with @xmath43_{p , q}+[n]_{p , q}$ ] and @xmath44_{p , q}^2 $ ] . +
this gives a generalization of proposition [ main1 ] .
we define the number of nestings of a permutation @xmath14 to be @xmath97 .
the coefficient of @xmath98 $ ] in @xmath99 is the number of permutations @xmath14 of @xmath3 $ ] with @xmath31 weak exceedances , @xmath32 crossings and @xmath100 nestings .
note that this bijection implies that the number of permutations with @xmath31 weak exceedances and @xmath32 crossings and @xmath100 nestings is equal to the number of permutations with @xmath31 weak exceedances and @xmath32 nestings and @xmath100 crossings .
similar results are known for set partitions and matchings @xcite .
we suppose that @xmath14 is a permutation of @xmath3 $ ] with @xmath31 weak exceedances . for any @xmath16 with @xmath101 , we first define : * @xmath102 * @xmath103 note that @xmath104 and that for @xmath80 , @xmath105 + therefore @xmath106 for @xmath18 , let @xmath107 it is easy to see that : @xmath108 therefore @xmath109 where @xmath110 it is easy to see that for @xmath80 , @xmath111 and therefore that @xmath112 as they are pairwise disjoint then @xmath113 combining equations ( [ eq1 ] ) and ( [ eq2 ] ) we get that @xmath114 this concludes the proof of proposition [ main ] .
continued fractions like the one presented in equation ( [ best ] ) were studied combinatorially in @xcite .
a descent in a permutation is an index @xmath16 such that @xmath115 .
an ascent in a permutation is an index @xmath16 such that @xmath116 .
the pattern 31 - 2 ( resp .
2 - 31 , 13 - 2 ) occurs in @xmath14 if there exist @xmath117 such that @xmath118 ( resp .
@xmath119 , @xmath120 ) .
+ theorem 10 in @xcite associated with theorem 22 in @xcite tells us that : the coefficient of @xmath121 $ ] in @xmath122 is the number of permutations @xmath14 of @xmath3 $ ] with @xmath123 descents , @xmath32 occurences of the patterns @xmath124 and @xmath100 occurences of the pattern @xmath125 .
this can also be proved bijectively thanks to a bijection of franon and viennot @xcite .
we present now that bijection .
see also @xcite . given a permutation @xmath126 , we set @xmath127 and @xmath128 .
. then @xmath16 is * a valley if @xmath130 * a double ascent if @xmath131 * a double descent if @xmath132 * a peak if @xmath133 * the beginning ( resp .
end ) of a descent if @xmath134 ( resp .
@xmath135 ) * the beginning ( resp .
end ) of a ascent if @xmath136 ( resp .
@xmath137 ) .
if @xmath129 then we define @xmath138 ( resp .
@xmath139 ) to be the number of indices @xmath140 ( resp .
@xmath141 ) such that @xmath142 .
+ to any permutation , we associate a pair @xmath71 made of a bicolored motzkin path @xmath72 and a weight @xmath73 . the path is created using the following rules : * @xmath74 if @xmath16 is a valley * @xmath77 if @xmath16 is a double ascent * @xmath79 if @xmath16 is a double descent * @xmath81 if @xmath16 is a peak the weight is created using the following rules : * @xmath143 if @xmath83 . *
@xmath144 if @xmath85 .
for example @xmath145 gives the path @xmath87 and the weight @xmath146 .
+ now we prove the following lemma for any @xmath16 @xmath147 * proof . *
we prove this lemma by induction . if @xmath16 is equal to 1 then @xmath16 is the beginning of an ascent and @xmath148 . if @xmath149 then @xmath150 where @xmath151 is zero , one or minus one .
it is easy to see that @xmath151 is one if @xmath152 is the end of a descent and @xmath16 is the beginning of an ascent , @xmath151 is minus one if @xmath152 is the end of an ascent and @xmath16 is the beginning of a descent and 0 otherwise . that gives exactly the lemma .
+ let @xmath94 be the set of pairs @xmath71 obtained from permutations of @xmath3 $ ] . using the machinery developed in @xcite
, we get directly that @xmath153 is @xmath99 defined in equation ( [ best ] ) .
+ combining the bijection of franon and viennot and the inverse of bijection of foata and zeilberger @xcite we get that the number of permutations of @xmath3 $ ] with @xmath31 descents , @xmath32 occurences of the patterns @xmath124 and @xmath100 occurences of the pattern @xmath125 is equal to number of permutations @xmath14 of @xmath3 $ ] with @xmath123 weak exceedances and @xmath32 crossings and @xmath100 nestings .
we also propose the direct mapping . starting from a permutation @xmath14 with @xmath31 descents , @xmath32 occurences of the patterns @xmath124 and @xmath100 occurences of the pattern @xmath125
, we form a permutation @xmath154 with @xmath31 descents , @xmath32 occurences of the patterns @xmath124 and @xmath100 occurences of the pattern @xmath125 .
we first form two two - rowed arrays @xmath155 and @xmath156 .
the first line of @xmath155 contains all the entries of @xmath14 that are the beginning of a descent .
they are sorted in increasing order .
the second line of @xmath155 contains all the entries of @xmath14 that are the end of a descent .
they are sorted such that if @xmath16 is in that row then @xmath139 in @xmath14 is the number of entries of the row to the right of @xmath16 smaller than @xmath16 .
the second line of @xmath155 contains all the entries of @xmath14 that are the end of a descent sorted in increasing order .
+ the first line of @xmath156 contains all the entries of @xmath14 that are the beginning of an ascent and that are sorted in increasing order .
the second line of @xmath156 contains all the entries of @xmath14 that are not the end of a descent .
they are sorted such that if @xmath16 is in that row then @xmath139 in @xmath14 is the number of entries of the row to the left of @xmath16 greater than @xmath16 .
we create the permutation @xmath154 which is the union of @xmath155 and @xmath156 .
+ for example if @xmath157 then the @xmath125 sequence is @xmath158 then @xmath159 and @xmath160 then @xmath161 .
+ we conclude this section by proving proposition [ pattern ] . given a permutation @xmath2 .
let @xmath162 . if @xmath14 has @xmath63 ascents ( or @xmath123 descents ) and @xmath32 occurences of the pattern @xmath124 , then @xmath163 has @xmath63 descents and @xmath32 occurences of the pattern @xmath164 . with the previous proposition this gives proposition [ pattern ] .
we can also derive the generating function of @xmath165 these were introduced in @xcite .
the following corollary is an easy consequences of theorem [ big ] .
let @xmath166 .
@xmath167 with @xmath168_q$ ] and @xmath169_q^2 $ ] .
lauren williams @xcite proved that @xmath170x+[i]xy)}\ ] ] it would be interesting to have a direct proof of the identity of the formal series form and the continued fraction form of these generating functions @xcite .
+ we can also interpret these results combinatorially . in @xcite the coefficient of @xmath171 $ ] in @xmath172 is interpreted in terms of decorated permutations with @xmath31 weak exceedances and @xmath32 alignments .
let us define these notions .
decorated permutations are permutations where the fixed points are bicolored @xcite .
we color these fixed points by colors @xmath173 .
we say that @xmath174 if @xmath75 or @xmath175 and colored with color @xmath176 .
we say that @xmath177 if @xmath80 or @xmath175 and colored with color @xmath178 . for a decorated permutation @xmath14 and @xmath16 , let * @xmath179 * @xmath180 * @xmath181 * @xmath182 * @xmath47 as for permutations we define @xmath183 . with these notions
, we can again define the number of alignments ( resp .
nestings , crossings ) of a decorated permutation @xmath14 as @xmath184 ( resp .
@xmath97 , @xmath185 ) . the number of weak exceedances ( resp .
descedances ) of a decorated permutation is the cardinality of the set @xmath186 ( resp .
@xmath187 ) . let @xmath188 with @xmath168_{p , q}$ ] and @xmath169_{p , q}^2 $ ] .
a direct generalization of the bijection la foata - zeilberger on decorated permutations gives : the coefficient of @xmath98 $ ] in @xmath189 is the number of decorated permutations @xmath14 of @xmath3 $ ] with @xmath31 weak exceedances , @xmath32 is the sum of the crossings and the descedances and @xmath100 nestings .
we can also make a direct link with the alignments : for any decorated permutation @xmath14 with @xmath31 weak exceedances @xmath190 * proof . *
the proof is omitted as it follows exactly the same steps as the proof of proposition [ main ] . + we leave as an open question to derive the bijection la franon - viennot on decorated permutations to interpret the @xmath172 in terms of descents and generalized patterns .
the asep model @xcite consists of black particles entering a row of @xmath15 cells , each of which is occupied by a black particle or vacant .
a particle may enter the system from the left hand side , hop to the right or to the left and leave the system from the right hand side , with the constraint that a cell contains at most one particle .
we will say that the empty cells are filled with white particles @xmath191 .
a basic configuration is a row of @xmath15 cells , each containing either a black @xmath192 or a white @xmath191 particle .
let @xmath193 be the set of basic configurations of @xmath15 particles .
we write these configurations as though they are words of length @xmath15 in the language @xmath194 .
the asep defines a markov chain @xmath195 defined on @xmath193 with the transition probabilities @xmath196 , @xmath197 , and @xmath0 .
the probability @xmath198 , of finding the system in state @xmath199 at time @xmath200 given that the system is in state @xmath201 at time @xmath202 is defined by : * if @xmath203 and @xmath204 then @xmath205 * if @xmath206 and @xmath207 then @xmath208 * if @xmath209 and @xmath210 then @xmath211 * otherwise @xmath212 for @xmath213 and @xmath214 .
see an example for @xmath215 in figure [ chainp ] . for @xmath215 . ]
this markov chain has a unique stationary distribution @xcite .
the case @xmath37 was studied combinatorially by enrica duchi and gilles schaeffer @xcite . in @xcite
another combinatorial approach was taken to treat the general case .
@xcite let @xmath216 be the set of bicolored motzkin paths of length @xmath15 .
the weight of the path in @xmath217 is 1 .
the weight of any path @xmath218 denoted by @xmath219 is the product of the weights of its steps .
the weight of a step , @xmath220 , starting at height @xmath221 is given by : @xmath222_q\\ \mbox{if } p_i = \bar{e } & \mbox { then } & w(p_i ) = [ h]_q+q^h/\alpha \\ \mbox{if } p_i = e & \mbox { then } & w(p_i ) = [ h]_q+q^h/\beta \\
\mbox{if } p_i = s & \mbox { then } & w(p_i ) = [ h]_q + q^{h}/(\alpha\beta)-q^{h-1}(1/\alpha-1)(1/\beta-1 ) .
\end{aligned}\ ] ] given a path @xmath218 , @xmath223 is the basic configuration such that each @xmath224 and @xmath225 step is changed to @xmath191 and each @xmath226 and @xmath227 step is changed to @xmath192 .
let @xmath228 and @xmath229 @xcite at the steady state , the probability that the chain is in the basic configuration @xmath201 is @xmath230 we can use these results and observations from the previous sections to get : if @xmath231 , at the steady state , the probability that the chain is in a basic configuration with @xmath31 particles is @xmath232 before proving that theorem , we need a lemma there is a weight preserving bijection between @xmath233 the set of bicolored motzkin paths of length @xmath15 where the weight of any step starting at height @xmath221 is @xmath234_q$ ] and where @xmath31 is the number of steps @xmath227 plus the number of steps @xmath226 and @xmath235 the set of motzkin paths of length @xmath236 where the weight of any step starting at height @xmath221 is @xmath234_q$ ] if the step is @xmath227 or @xmath226 and @xmath237_q$ ] otherwise and where @xmath238 is the number of steps @xmath227 plus the number of steps @xmath226 .
* this is a general version of the classical bijection between bicolored motzkin path of length @xmath15 and bicolored motzkin path of length @xmath236 where one of the horizontal steps can not appear at height 0 .
the proof consists of transforming each step @xmath227 ( resp .
@xmath226 , @xmath224 , @xmath225 ) weighted by @xmath239 by two steps @xmath240 ( resp .
@xmath241 , @xmath242 , @xmath243 ) each of them weighted by @xmath244 .
then we add a @xmath227 step at the beginning and add a step @xmath225 at the end both with weight @xmath245 and apply the reverse map . + * proof of the theorem .
* we want to compute for @xmath231 @xmath246 now we use the previous lemma and get @xmath247 using @xcite and the definition of the weight of the steps of @xmath248 , we conclude that : @xmath249 and therefore that @xmath250 .
several open problems naturally arise : * can we define generalized patterns for decorated permutations ?
* can we generalize these @xmath0-eulerian numbers to understand the asep when @xmath251 or @xmath252 ?
* can we extend the definition of @xmath31-crossing and @xmath31-nesting that were defined for matchings and set partitions @xcite ?
99 p. biane , permutations suivant le type dexcdance et le nombre dinversions et interprtation combinatoire dune fraction continue de heine .
european j. combin . 14
( 1993 ) , no .
4 , 277284 .
r. brak , s. corteel , j. essam and a. rechnitzer , a combinatorial derivation of the pasep equations , fpsac2005 , to appear .
w. y. c. chen , e. y. p. deng , r. r. x. du , r. stanley and c. h. yan , crossings and nestings of matchings and partitions , submitted .
a. claesson , t. mansour , counting occurrences of a pattern of type @xmath253 or @xmath254 in permutations . adv . in appl .
( 2002 ) , no .
2 , 293310 .
r. clarke , e. steingrmsson , and j. zeng , new euler - mahonian statistics on permutations and words . adv . in appl .
18 ( 1997 ) , no .
3 , 237270 .
b. derrida , m.r .
evans , v. hakim , v. pasquier , _ exact solution of a 1d asymmetric exclusion model using a matrix formulation _
a26 , 1493 - 1517 ( 1993 ) .
e. duchi , g. schaeffer , _ a combinatorial approach to jumping particles _
, jal of comb theory a , to appear ` arxiv math.co/0407519 ` p. flajolet , combinatorial aspects of continued fractions , disc . math . 41 ( 1982 ) , 145 - 153 .
d. foata and d. zeilberger , denert s permutation statistic is indeed euler - mahonian .
83 ( 1990 ) , no .
1 , 3159 . j. franon and x. viennot , permutations selon les pics , creux , doubles descentes , doubles montes , nombres deuler .
nombres de genochi , disc .
( 1979 ) , 21 - 35 .
m. klazar , on identities concerning the numbers of crossings and nestings of two edges in matchings , preprint .
c. macdonald j. gibbs and a. pipkin , biopolymers , 6 1 ( 1968 ) .
a. postnikov , webs in totally positive grassman cells , preprint .
x. viennot , une thorie combinatoire des polynmes orthogonaux , notes de cours , uqam , montral 1988 .
williams , private communication .
williams , enumeration of totally positive grassmann cells . adv .
( 2005 ) , no .
2 , 319342 | we derive the continued fraction form of the generating function of some new @xmath0-analogs of the eulerian numbers @xmath1 introduced by lauren williams building on work of alexander postnikov .
they are related to the number of alignments and weak exceedances of permutations .
we show how these numbers are related to crossing and generalized patterns of permutations we generalize to the case of decorated permutations .
finally we show how these numbers appear naturally in the stationary distribution of the asep model . |
the general problem studied here concerns conformal riemannian metrics of constant positive curvature with conic singularities on a compact riemann surface @xmath3 . for the history of the problem and its relevance to mathematics and physics we refer to @xcite .
the goal is to classify such metrics up to isometry , in particular _ to determine for a given riemann surface , how many such metrics exist with prescribed singularities and prescribed angles at each singularity . _
analytically the problem consists in the study of the liouville equation @xmath4 on the punctured riemann surface , with prescribed singularities at the punctures .
if @xmath5 is the length element of the metric in a local conformal coordinate , then @xmath6 , and equation ( [ liouville ] ) expresses the fact that the curvature equals @xmath0 .
the behavior at a singularity @xmath7 is @xmath8 , where @xmath9 , @xmath10 is the local coordinate which is @xmath11 at @xmath12 , and @xmath13 is the angle at the singularity . here and in what follows we measure the angles at the singularities in _ half - turns _ , for example `` integer angle '' means an integer multiple of @xmath14 radians .
an important special case is that of _ symmetric _ metrics , that is metrics which are invariant with respect to an anti - conformal involution which leaves all singularities fixed . in the case of the sphere , such metrics are in one - to - one correspondence with spherical polygons @xcite .
so the problem in this case consists in classification and enumerating spherical polygons of prescribed conformal type with prescribed angles at the corners .
every surface of constant curvature @xmath0 is locally isometric to a piece of the standard sphere .
this isometry is analytic in conformal coordinates , so it admits an analytic continuation along every path not passing through the singularities , and we obtain a multi - valued analytic function @xmath15 where @xmath16 are the singularities , and @xmath17 is the riemann sphere .
this function @xmath18 is called the _ developing map _ , and the length element of the metric is expressed in the form @xmath19 monodromy of the developing map consists of linear - fractional transformations ( rotations of the sphere ) , so the schwarzian derivative @xmath20:=\frac{f'''}{f'}-\frac{3}{2}\left(\frac{f''}{f'}\right)^2\ ] ] is single valued on @xmath21 .
the behavior of @xmath18 near the conic singularities @xmath16 implies that @xmath22 $ ] has double poles at these points , with principal parts related to the angles ( see below ) . in the case of the sphere , @xmath22 $ ] is a rational function @xmath23 , and we obtain a schwarz differential equation @xmath22=r$ ] .
general solution of this schwarz equation is a ratio of two linearly independent solutions of the linear differential equation with regular singular points @xmath24 in the case of the sphere , @xmath23 in this equation is a rational function with poles of order @xmath25 at the singularities , and the coefficients at @xmath26 are determined by the angles .
the exponent differences at the singular points are the angles . after a normalization @xmath27 free parameters
remain which are called _
accessory parameters_. in terms of this equation , our general problem is equivalent to the following : _ for given singular points and the exponent differences , how many values of accessory parameters exist , so that the monodromy group of ( [ ode ] ) is conjugate to a subgroup of @xmath28 .
_ in the case of three singularities , there is no accessory parameter , the equation ( [ ode ] ) is hypergeometric , and the problem of classification of the metrics was completely solved for this case in @xcite .
the case of four singularities corresponds to heun s equation which has one accessory parameter .
the assumption that the angles are half - integer leads to the special case of the heun equation , which is equivalent to the so - called darboux
treibich verdier equation @xcite , and which is easier to study because its general solution is meromorphic , and can be expressed in terms of abelian integrals .
for @xmath29 , the existing results show that the general problem described above is very difficult , see , for example , the recent study of metrics with one singularity of integer angle on a torus @xcite , or classification of such symmetric metrics with four angles , at least one of them integer , on the sphere @xcite . in this paper
we study new special cases .
we restrict ourselves to the cases of four singularities with integer angles on tori and half - integer angles on the sphere . in the first part of this paper
we establish a relation between the metrics on the sphere and the metrics on tori , and restate the results obtained in @xcite . in the second part we study the simplest family which is not covered by the existing results , metrics on `` rectangular '' tori with angles @xmath30 . by the results of the first part ,
this is equivalent to the study of symmetric metrics on the sphere with angles @xmath31 .
we thank walter bergweiler and chang - shou lin for useful discussions of the problems considered here .
we consider a metric of curvature @xmath0 on a torus @xmath32 with four conic singularities at half - periods , @xmath33 , where @xmath34 and the metric has an integer angle at each singularity .
let the angles be @xmath35 .
then the developing map @xmath36 must satisfy the schwarz equation @xmath37=r$ ] , where @xmath23 is an elliptic function of the form @xmath38 with @xmath39 and @xmath40 , as is explained below .
the general solution is a ratio of two linearly independent solutions of the equation @xmath41 the exponents at the singularities of ( [ 1 ] ) are determined from the characteristic equation @xmath42 whose solutions are @xmath43 , so the difference of the exponents is @xmath44 , and these are our prescribed angles of the metric on @xmath45 .
monodromy of the developing map can not contain a parabolic transformation , so there should be no logarithms in the local solutions of ( [ 1 ] ) near the singularities , and equation ( [ 1 ] ) must have trivial monodromy at @xmath46 .
in other words , all singularities must be _
this gives three polynomial conditions on the parameters @xmath47 , see , for example @xcite .
in addition to this , we have conditions on the accessory parameter @xmath48 which ensure that monodromy of ( [ 1 ] ) is unitarizable , that is conjugated to a subgroup of @xmath49 .
in particular , these conditions of absence of logarithms are always satisfied if all @xmath50 , in which case ( [ 1 ] ) becomes the darboux treibich verdier equation @xmath51 for this equation an explicit formula for the general solution of ( [ 1 ] ) can be written @xcite .
the monodromy coming from the fundamental group of the torus must be conjugate to an abelian subgroup of @xmath28 .
there are only two kinds of such groups @xcite : the klein viergroup @xmath52 and a cyclic group ( finite or infinite ) . by a post - composition of @xmath36 with an element of @xmath28 , which does not change the metric
, we can achieve that @xmath36 satisfies one of the following : @xmath53 or @xmath54 we call two metrics _ equivalent _ if their developing maps are related by a linear - fractional transformation .
thus in the case ( [ b ] ) each metric is a member of a one - parametric family of equivalent metrics with developing maps @xmath55 , @xmath56 .
* theorem 1 .
* _ suppose that @xmath57 in ( [ 1 ] ) .
then in case ( [ a ] ) the metric is even , that is invariant under @xmath58 , while in case ( [ b ] ) each equivalence class contains an even metric .
_ this generalizes results from @xcite , sections 2.1 and 5.1 . _
_ it is sufficient to show that @xmath59 with @xmath60 . that @xmath36 satisfies ( [ 2 ] ) with _
some _ @xmath61 is clear because the schwarzian @xmath37 $ ] is even .
it is also clear from ( [ 2 ] ) that @xmath62 so @xmath63 is either an elliptic transformation with multiplier @xmath64 or the identity .
if @xmath65 , the proof is finished , so assume that the multiplier of @xmath63 is @xmath64 .
suppose now that @xmath36 is a meromorphic function , satisfying ( [ 2 ] ) , and @xmath66 where @xmath3 is a linear - fractional transformation .
we claim that in this case the equation @xmath67 must hold .
indeed , ( [ s ] ) implies @xmath68 and also @xmath69 using ( [ 2 ] ) , ( [ ss ] ) and ( [ ins ] ) we obtain @xmath70 which implies ( [ com ] ) .
suppose that ( [ a ] ) holds .
both equations in ( [ a ] ) are of the form ( [ s ] ) with @xmath71 , thus @xmath72 . using the first equation ( [ a ] )
in which @xmath73 , we conclude that @xmath74 or @xmath75 , @xmath76 . in the first case ( [ idem ] )
implies @xmath77 , so @xmath60 . in the second case ,
we use the second equation in ( [ a ] ) which implies that @xmath63 commutes with @xmath78 , and conclude again that @xmath77 and @xmath60 .
suppose now that ( [ b ] ) holds , and @xmath79
. then ( [ com ] ) gives @xmath80 substituting @xmath81 to ( [ kwa ] ) , after simple computation we obtain @xmath82 and @xmath83 if @xmath84 , then @xmath85 so @xmath86 and @xmath87 . using ( [ idem ] )
we conclude that @xmath88 . if @xmath89 , then @xmath90 , so @xmath91 and @xmath75 , and after multiplication of @xmath36 by a constant we obtain @xmath60 .
finally , if we have ( [ b ] ) with both multiplies @xmath92 , then @xmath36 is an elliptic function , with @xmath93 , where @xmath63 is an elliptic transformation satisfying ( [ 2 ] ) .
let @xmath94 be such that @xmath95 then the metric with the developing map @xmath96 is even .
this completes the proof of theorem 1 .
every even metric on the torus is a pull - back of a metric from the sphere via the @xmath97 function .
thus we obtain a metric on the sphere , which has @xmath98 singularities at the critical values of @xmath97 that is at @xmath99 , and the angles at these singularities are @xmath100 .
we call @xmath18 the developing map on the sphere , so that @xmath101 .
now we consider four cases .
* _ all @xmath102 are odd . _
then the monodromy of @xmath18 is generated by four non - trivial involutions .
* proposition 1 .
* _ in case 1 , @xmath36 satisfies ( [ b ] ) .
_ this generalizes a result ( * ? ? ?
* thm . 0.3 ) .
_ proving this by contradiction , assume that we have ( [ a ] ) .
then the monodromy group of @xmath36 is the klein viergroup @xmath103 .
let @xmath104 be conformal involutions of @xmath105 with fixed points @xmath46 , @xmath106 their images in the monodromy group are non - trivial elements of @xmath107 , because all @xmath102 are odd .
as @xmath107 contains only three non - trivial elements , two of the distinct involutions @xmath108 and @xmath109 have the same image , so the image of @xmath110 in the monodromy group must be the identity . but @xmath110 is a translation by some @xmath111 , so we obtain a contradiction with ( [ a ] ) , which proves the proposition . in the remaining cases ,
some @xmath102 are even , thus the metric on the sphere has at least one integer angle .
all these cases were studied in detail and completely classified in @xcite .
we state the results .
* _ exactly one of the @xmath102 is even .
_ let @xmath112 be the even angle .
then it follows from ( * ? ? ?
* proposition 4.2 ) that there are at most @xmath113 distinct metrics , and for generic torus there are exactly @xmath113 of them .
the accessory parameters @xmath48 are found from the algebraic equation that expresses the absence of logarithms in the solution of the heun equation .
* case 3 . *
_ exactly two of the @xmath102 are even .
_ let @xmath114 , @xmath115 and assume that @xmath116 then according to ( * ? ? ?
4.1 ) there are two subcases : subcase 3a .
@xmath117 is even , @xmath118 , and the number of classes of metrics is at most @xmath119 with equality for generic tori , subcase 3b .
@xmath117 is odd , @xmath120 , and the number of classes of metrics is at most @xmath121 with equality for generic tori .
if three of the @xmath102 are even then all four must be even , because the product of local monodromies is the identity , and we obtain the last case : * case 4 .
* _ all @xmath102 are even . _ in this case , @xmath18 is a rational function with four critical points of orders @xmath122 .
such functions were classified in @xcite .
so in all cases 2,3,4 , we have a complete classification of metrics in question . in the next sections we consider a special family which belongs to case 1 .
here we begin to study spherical quadrilaterals whose angles are odd multiples of @xmath123 .
we call them _ spherical rectangles_. metrics on the sphere corresponding to spherical rectangles are symmetric with respect to a circle , and without loss of generality we may assume that this circle is the real line .
singularities lie on the real line .
consider the ramified covering of degree @xmath25 from a torus to the sphere , ramified over these singularities .
it is a composition of an element of @xmath124 with the weierstrass @xmath97 function whose invariants @xmath125 are real .
such weierstrass functions are defined on `` rectangular '' tori , which means that @xmath126 .
the pull back of the metric on the torus has singularities with integer angles at half - periods @xmath46 . in each quadrilateral
we mark one corner .
two quadrilaterals are considered equal if there is an orientation preserving isometry sending one to another and sending the marked corner to the marked corner .
we usually choose the upper half - lane @xmath127 as a conformal model of a spherical polygon .
then the corners @xmath16 become points on the real line , and the metric has the length element @xmath128 , @xmath6 where @xmath129 satisfies ( [ liouville ] ) .
then the spherical polygon is a pair @xmath130 , where @xmath131 has conic singularities at @xmath16 and the sides @xmath132 are geodesic with respect to the metric @xmath131 .
developing map of a quadrilateral satisfies the schwarz equation associated with a heun equation , as explained in the introduction ( see also @xcite ) .
our first result is * theorem 2 . *
_ let @xmath133 be the developing map of a spherical quadrilateral whose angles are odd multiples of @xmath123 .
then there are two opposite sides whose @xmath18-images are contained in the same circle . _
thus there are two types of such spherical quadrilaterals : in the first type one of the two sides whose images belong to the same circle ends at the marked corner , and in the second type such a side begins at the marked corner . to state our main result we define the conformal modulus .
let us map conformally our quadrilateral onto the rectangle @xmath134 , so that the marked corner is mapped to @xmath11 .
then @xmath135 is called the modulus .
* theorem 3 . * _ there are two continuous families of spherical quadrilaterals with angles @xmath136 , and angle @xmath137 at the marked corner .
one family consists of quadrilaterals of the first type , and the modulus in this family varies from @xmath11 to @xmath138 .
the second family consists of quadrilaterals of the second type , and the modulus varies from @xmath139 to @xmath140 .
each family contains exactly one quadrilateral of the given modulus in the described range .
_ so there is a `` prohibited interval '' @xmath141 $ ] for the moduli of spherical quadrilaterals with angles @xmath136 .
theorem 3 is proved in the next section .
_ proof of theorem 2 .
_ let us enumerate the great circles @xmath142 which contain the images of sides according to positive orientation of the boundary , so that @xmath143 is a residue modulo @xmath98 .
we may assume that @xmath144 and @xmath145 are real and imaginary axes , then @xmath146 must cross the real axis perpendicularly at the points @xmath147 , and @xmath148 must cross the imaginary axis perpendicularly at the points @xmath149 . without loss of generality
@xmath150 and @xmath151 .
if the @xmath98 circles are in general position ( that is there are no triple intersections ) , then both @xmath152 and @xmath153 are finite , positive , and there is a crossing point @xmath154 of @xmath146 and @xmath148 .
the centers of @xmath146 and @xmath148 are at @xmath155 and @xmath156 .
their radii are @xmath157 and @xmath158 .
then we obtain from the right triangle @xmath159 that the square of the distance between the centers is @xmath160 on the other hand the angle between the radii of @xmath146 and @xmath148 at @xmath154 is right , so the same distance is @xmath161 a contradiction .
therefore the images of two opposite sides must lie on the same circle .
the non - generic cases are easy and we leave them to the reader .
this proves theorem 2 .
let our three circles be the real line , the line @xmath162 through the origin making angle @xmath163 with positive ray , and the unit circle .
the @xmath18-images of the two opposite sides lie on the unit circle and the images of the other two sides on @xmath164 and @xmath162 , respectively .
we have one continuous parameter @xmath165 , and the modulus of the quadrilateral depends on the angles and on @xmath163 .
the group generated by reflections in the @xmath18-images of the sides has a subgroup of index @xmath25 consisting of conformal maps ; this subgroup is the monodromy group of our developing map @xmath166 .
the monodromy group is generated by @xmath167 elliptic involutions with fixed points @xmath168 , @xmath169 and @xmath170 , so it consists of the maps of the form @xmath171 where @xmath172 is any integer .
when @xmath163 is rational , this group is finite , it is a dihedral group , and the developing map is algebraic . for this case
, there is a theorem of klein ( see @xcite ) which says that the heun equation is a pullback of a hypergeometric equation with finite monodromy group .
this means that the developing map factors as @xmath173 , where @xmath174 is a solution of the hypergeometric equation ( in our case @xmath175 is a conformal map of the triangle with angles @xmath176 onto the upper half - plane , and @xmath177 is a belyi rational function which can be computed algebraically ) .
our monodromy group ( [ monodromy ] ) belongs to the class of groups considered by van vleck which consist of transformations of the form @xmath178 van vleck @xcite proved that fuchsian differential equations with such projective monodromy groups can be characterized by the property that they have two linearly independent solutions whose product is a polynomial .
this remarkable property was first discovered by hermite @xcite for the case of lam equation , and generalized by darboux @xcite to equation ( [ dtv ] ) .
it permits to solve the equation in a closed form .
if @xmath179 is this polynomial , then two linearly independent solutions can be represented in the form @xmath180 hermite found the method of determining @xmath179 in terms of @xmath181 and the accessory parameter .
equation on the accessory parameter comes from the condition that the monodromy group has the special form ( [ monodromy ] ) . in section 4
we perform all computations explicitly in a simple case .
to pull back our metric on a torus , we model our spherical quadrilateral on a flat rectangle , instead of the half - plane . consider the rectangle @xmath182 , where @xmath183 and @xmath184 are positive and @xmath185 . now
the developing map @xmath18 is defined on this rectangle and maps it to the standard sphere . without loss of generality
one can assume that @xmath186 and @xmath187 , but sometimes we will use @xmath183 and @xmath188
. then the developing map extends by reflection to the whole plane .
we assume without loss of generality that the vertical sides of the rectangle are mapped by @xmath18 into the unit circle , and the side @xmath189 to the real line .
then the developing map is real on the real line and satisfies @xmath190 which is a special case of ( [ b ] ) . to this one
can add @xmath191 and similar properties for conformal involutions with fixed points at @xmath46 . if the angles @xmath192 of our spherical rectangle are @xmath193 ,
then the map @xmath18 is ramified at the corners and is @xmath194-to-@xmath0 near @xmath46 , that is @xmath195 has a zero of order @xmath196 at @xmath46 and no other zeros . in ( [ periods ] ) , @xmath163 and the parameter @xmath197
are connected by some relation which also depends on the @xmath102 .
thus we have the problem : _ for given periods and given integers @xmath102 , how many functions @xmath18 exist ( up to proportionality ) that satisfy ( [ periods ] ) and have prescribed critical point pattern : critical points of order @xmath196 at @xmath111 and no other critical points . _ the schwarzian derivative of @xmath18 is doubly periodic with periods @xmath198 , and has double poles at @xmath111 , where we must have apparent singularities and the exponent difference must be @xmath102 .
so our developing map is the ratio of two solutions of the darboux
treibich verdier equation ( [ dtv ] ) .
thus we have an equivalent problem : _ for given periods and @xmath199 , how many real values of @xmath48 exist , so that a ratio @xmath200 of some linearly independent solutions of ( [ 1 ] ) has properties ( [ periods ] ) with some real @xmath163 . _ ( parameter @xmath48 must be real because the schwarzian of @xmath18 is real on the real line , and @xmath201 are real on the real line for the lattices with real @xmath183 and imaginary @xmath188 . )
in this section we prove theorem 3 . we consider a spherical rectangle whose angles are odd multiples of @xmath123 ( measured in half - turns , so that @xmath123 means @xmath1 radians ) .
we take the angles @xmath136 . * lemma 2 . * _ there are two families of spherical rectangles with angles @xmath31 , each depending on parameter @xmath165 .
rectangles of these families are isometric to subsets of the sphere whose oriented boundaries are shown in fig . 1 . _ _ proof . _
the area of a rectangle with angles as in lemma 2 is @xmath123 of the area of the sphere ; this follows from the gauss bonnet formula .
we can normalize the developing map so that the image of the boundary is contained in the union of the real line , the line @xmath162 as in the previous section , and the unit circle .
the image of any side can not cover more than @xmath123 of the great circle to which this image belongs , otherwise this image would contain either a half - plane or the exterior or the interior of the unit disk . since the area of the whole rectangle is @xmath123 of the area of the sphere , it would follow that the rectangle is isometric to a hemisphere , which is of course impossible . from theorem 1 we know that some pair of opposite sides is mapped by the developing map into the same circle . without loss of generality
we assume that this circle is the unit circle .
let @xmath202 be these two sides .
they are oriented in the usual way , so that the rectangle is on the left of each side .
the angle at the beginning of @xmath203 is the same as that at the beginning of @xmath204 , and it is either @xmath123 or @xmath137 .
so we have ( at least ) two types of such rectangles .
consider a rectangle of the first type .
let the sides be @xmath205 in the order of positive orientation .
without loss of generality , the image of @xmath206 is on the real line , oriented in the increasing direction .
this image must be either @xmath207 or its complement .
without loss of generality , let it be the complement .
then the image of @xmath203 must be a clockwise arc of the unit circle from @xmath64 to some point @xmath208 in the upper half - plane .
tracing the rest of the boundary we arrive at a picture in the top of fig . 1
. rectangles of type 2 are considered in the similar way .
this proves the lemma .
an alternative way to understand spherical quadrilaterals is through their _ nets _ @xcite . the developing map @xmath18 maps the sides of a quadrilateral @xmath209 to three transversal great circles on the sphere .
the _ net _
@xmath210 of @xmath209 is the preimage in @xmath209 of these three circles . for each circle @xmath211 , its preimage in @xmath209
is called @xmath211-_net _ , denoted @xmath212 . an _ arc _ of the net @xmath212 is a connected component of @xmath213 .
the net @xmath210 defines a triangulation of @xmath209 , each face of it being mapped by @xmath18 one - to - one onto one of the triangles into which the three circles partition the sphere .
* lemma 3 . *
_ let @xmath211 be the circle to which two opposite sides @xmath203 and @xmath204 of a quadrilateral @xmath209 with angles @xmath214 are mapped .
then there is an arc @xmath215 of @xmath212 with the ends at the two corners of @xmath209 having angles @xmath137 . _ _ proof .
_ since the area of @xmath209 is @xmath123 of the area of the sphere , and each corner having angle @xmath137 has three faces of @xmath210 adjacent to it , with the total area greater than @xmath216 of the area of the sphere , there should be a face @xmath36 of @xmath210 adjacent to both of these corners .
thus these corners are connected by one of the sides of @xmath36 which must be an arc of @xmath212 because @xmath211 is the only circle to which the images of both of these corners belong
. * corollary . *
_ each quadrilateral @xmath209 with angles @xmath31 is a union of two isometric halfs of hemispheres , each of them bounded by the arcs of two great circles intersecting at a right angle . _
* the statement of lemma 3 is a special case of the following statement , which will be proved elsewhere : _ let @xmath209 be a spherical @xmath217-gon such that all its sides are mapped to three transversal great circles by the developing map , and all its corners are mapped to some intersection points of those circles . then there is ( in general , non - unique ) triangulation of @xmath209 by @xmath27 disjoint arcs of its net , each arc having both ends at the corners of @xmath209 .
_ we recall that our quadrilateral has marked corner , and two quadrilaterals are considered equal if there is an isometry between them which preserves the orientation and sends the marked corner to the marked corner .
we choose the marked corner so that the angle at it is @xmath137 , and the two families are distinguished by the property that the side ending at the marked corner in the first family has the image on the same circle as its opposite side , while in the second family the side that ends at the marked corner and its opposite side have images on the different circles .
note that the choice of one of the two corners with the angle @xmath137 as marked is not important , as our families have isometries @xmath218 and @xmath219 , respectively , exchanging the two corners .
let us map our quadrilateral on a rectangle with two sides @xmath220 $ ] and @xmath221 $ ] , where @xmath11 is the marked corner .
then @xmath135 is the _ modulus _ of the quadrilateral .
now we return to the proof of theorem 3 .
our two families are shown in fig . 1 .
the marked corner is circled .
consider the first family .
taking log we obtain the region in fig . 2 .
.,width=384 ] now according to schwarz and christoffel , logarithm of the developing map defined in the upper half - plane is @xmath222 this should be compared with van vleck s general formula ( [ vv ] ) .
the residue at @xmath223 equals : @xmath224 then the residues at @xmath206 are @xmath225 by the residue theorem .
this gives the condition @xmath226 this is the `` bethe - ansatz equation '' for this case .
thus @xmath227 and @xmath228 therefore our integral is @xmath229 where the branch of the square root in the upper half - plane is chosen so that the integrand is positive on @xmath230 .
here @xmath223 plays the role of the accessory parameter , and is simply related to it .
the condition is that @xmath231 because the part of the integral in ( [ 3 ] ) over @xmath232 $ ] is imaginary . from this condition
@xmath233 must be found . to regularize our integral on @xmath207 .
write it as @xmath234 now @xmath235 and @xmath236 the equation for @xmath223 is @xmath237 .
we would like to know that @xmath238 is uniquely determined by @xmath239 .
this would mean that there is at most one quadrilateral with prescribed modulus in each family .
* proposition 1 .
* _ there exists @xmath240 with the following property . for
every @xmath241 there exists unique @xmath242 such that @xmath237 .
moreover , @xmath243 as @xmath244 and @xmath245 as @xmath246 .
_ here @xmath247 is a known ( increasing ) function of @xmath239 @xcite see also ( [ k ] ) below .
_ proof of proposition 1_. to prove uniqueness of @xmath223 it is sufficient to show that after multiplication of @xmath36 by some function of @xmath239 and @xmath223 which is non - zero in the considered range , the result is strictly decreasing in @xmath223 for each @xmath239 .
first we consider the function @xmath248 then we do partial fraction decomposition @xmath249 the second partial fraction @xmath250 is strictly decreasing in @xmath238 , for each @xmath251 , so it remains to prove that @xmath252 is decreasing . to do this we deform the contour of integration ( see fig . 3 )
we consider the simple closed curve @xmath215 on the riemann surface of the integrand which goes once around the segment @xmath253 $ ] in the negative direction ( clockwise ) .
then @xmath254 then we deform the contour so that the part in the upper half - plane consists of the interval @xmath255 , @xmath256 oriented from @xmath23 to @xmath239 , the interval @xmath257 oriented from @xmath258 to @xmath259 and a large half - circle @xmath260 .
the integral over the half - circle is @xmath261 times residue at infinity , thus it is pure imaginary . on the interval @xmath255 the integral has the form @xmath262 where @xmath263 and it is decreasing as a function of @xmath223 . on the interval @xmath257 the integral has similar form , and also decreasing as a function of @xmath223 .
this proves that for every @xmath239 there exists at most one @xmath223 such that @xmath237 .
let this value be @xmath264 . when @xmath265 , we have @xmath266 and @xmath264 is increasing .
so we have some interval @xmath267 on which @xmath223 is increasing .
when @xmath239 passes @xmath0 , @xmath223 passes @xmath11 , and we obtain the second family .
when @xmath268 , the quadrilateral degenerates must have the shape as in fig .
4 . for @xmath268.,width=384
] the conformal map is @xmath269 we want to keep @xmath270 but the residue at @xmath223 now must be @xmath271 , and the residue at @xmath206 must be @xmath272 .
that is @xmath273 notice that the equation for @xmath206 is exactly the same as before , thus @xmath227 , and that the expression under the root is now negative , and we obtain @xmath274 as desired .
the condition now is that @xmath275 and this must define @xmath276 .
the integral is now @xmath277 the only difference in comparison with the previous integral is that factor @xmath278 is replaced by @xmath279 under the square root , and the integral must be equal to @xmath280 .
this integral does not require desingularization .
consider the rescaled degenerate configuration , fig .
it is a quadrilateral but flat , not spherical .
.,width=384 ] the map of the upper half - plane on this region is given by @xmath281 which is the same that we obtain if we put @xmath282 in the integral in @xmath283 .
the condition would be now that @xmath284 but this diverges at one end .
von koppenfels @xcite integrates by parts to obtain his equation for critical @xmath239 .
the modulus of the degenerate region was computed by von koppenfels ( * ? ? ?
* ( 5.3.30 ) ) by solving the equation @xmath285 the solution is approximately @xmath286 the relation between our @xmath239 and @xmath287 is the following @xmath288 where @xmath289 this gives the critical value @xmath290 .
thus as @xmath239 varies from @xmath0 to @xmath291 , @xmath223 varies from @xmath11 to @xmath0 , and @xmath247 ( the modulus varies from @xmath11 to @xmath292 .
when @xmath265 , @xmath293 ; after this point is passed , we obtain the second family , with modulus @xmath294 near infinity . this second family is symmetric to the first , so there is a correspondence @xmath295 as seen from the pictures , and using the domain extension principle we have @xmath296 .
thus for every @xmath135 we have at most one quadrilateral with angles @xmath136 .
it is of the first type when @xmath297 and of the second type when @xmath298 .
this completes the proof of theorem 3 .
paper @xcite contains new proofs of some results from @xcite .
it is noticed in ( * ? ? ?
* remark 2 ) that equation ( [ k ] ) and its solution @xmath299 in ( [ kappap ] ) is a famous constant discussed in finch @xcite . to conform to finch s notation ,
let @xmath300 and @xmath301 there was a constant known for some time as the `` one - ninth '' constant in approximation theory , until gonchar and rakhmanov proved that it equals @xmath302 .
earlier the same constant @xmath302 implicitly appeared in euler s work on elastics .
halphen @xcite mentions @xmath223 as the unique solution of the equation @xmath303 and computes this solution to six digits .
but the most remarkable coincidence is the appearance of the related constants in @xcite , where there is a constant @xmath304 this constant @xmath305 appears in @xcite in the study of metrics of curvature @xmath0 with one conic singularity on real tori , but the real tori in @xcite are _ different _ from the real tori studied here !
we recall that there are two types of real tori : in the first type , which we called rectangular in this paper , the parameter @xmath306 is pure imaginary .
the second type , which is studied in @xcite can be called `` rhombic '' , and can be characterized by the property @xmath307 .
1 n. i. akhiezer , elements of the theory of elliptic functions , ams , providence , ri , 1990 . f. baldassari and b. dwork , on second order differential equations with algebraic solutions , amer . j. math . ,
101 ( 1979 ) 4276 .
w. bergweiler and a. eremenko , green s function and anti - holomorphic dynamics on a torus , arxiv:1507.01704(v2 ) .
ching - li chai , chang - shou lin and chin - lung wang , mean field equations , hyperelliptic curves and modular forms : i , arxiv:1502.03297 .
ching - li chai , chang - shou lin and chin - lung wang , mean field equations , hyperelliptic curves and modular forms : ii , arxiv1502.03295 .
g. darboux , sur une quation linaire , c. r. acad .
paris , t. 94 ( 1882 ) 16451648 .
a. eremenko , metrics of positive curvature with conic singularities on the sphere , proc .
( 2004 ) , 33493355 a. eremenko and a. gabrielov , rational functions with real critical points and the b. and m. shapiro conjecture in real enumerative geometry , ann .
, 155 ( 2002 ) 105129 .
a. eremenko and a. gabrielov , elementary proof of the b. and m. shapiro conjecture for rational functions , in the book : notions of positivity and the geometry of polynomials , trends in mathematics , springer , basel , 2011 , p. 167178 .
a. eremenko , a. gabrielov , m. shapiro and a. vainshtein , rational functions and real schubert calculus , proc .
ams , 134 ( 2006 ) , no .
4 , 949957 . a. eremenko , a. gabrielov and v. tarasov , metrics with conic singularities and spherical polygons , arxiv:1405.1738 a. eremenko , a. gabrielov and v. tarasov , metrics with four conic singularities and spherical quadrilaterals , arxiv:1409.1529 a. eremenko , a. gabrielov and v. tarasov , spherical quadrilaterals with three non - integer angles , arxiv:1504.02928 .
s. finch , mathematical constants , cambridge up , cambridge , 2003 .
s. fujimori , y. kawakami , m. kokubu , w. rossman , m. umehara and k. yamada , cmc-1 trinoids in hyperbolic 3-space and metrics of constant curvature one with conical singularities on the 2-sphere , proc .
japan acad . , 87 ( 2011 ) , 144149 . f. gesztesy and weikard , on picard potentials , differential integral equations 8 ( 1995 ) , no . 6 , 14531476 .
lisa goldberg , catalan numbers and branched coverings by the riemann sphere , adv .
85 ( 1991 ) , no . 2 , 129.144 .
h . halphen , trait des fonctions elliptiques et de leurs applications , premiere partie , paris , gauthier - villars , 1886 .
hermite , sur lquation de lam , extrait de feuilles authographies du course danalyse de lcole polytechnique , @xmath311 division , 1872 - 73 , @xmath312 leon .
oeuvres , t. iii , p. 118122 .
paris , gauthier - villars , 1912 .
f. klein , vorlesungen ber die hypergeometrische funktion , reprint of the 1933 original .
springer - verlag , berlin - new york , 1981 .
f. klein , vorlesungen ber das insider und die aufsnug der clonking von fften grade , birkhuser verlag , basel , 1993 .
w. von koppenfels , konforme abbildung besonderer kreisbogenvierecke , j. fr die reine und angew . math .
, 181 ( 1939 ) 83124 .
w. von koppenfels , und f. stallmann , praxis der konformen abbildung , springer , berlin , 1959 .
chang - shou lin and chin - lung wang , elliptic functions green functions and the mean field equations on tori , ann . math . , 172 ( 2010 ) 911954 .
f. luo and g. tian , liouville equation and spherical convex polytopes , proc .
( 1992 ) , no . 4 , 1119.1129 . g. mondello and d. panov , spherical metrics with conical singularities on a 2-sphere : angle constraints , arxiv:1505.01994 .
e. mukhin , v. tarasov and a. varchenko , the b. and m. shapiro conjecture in real algebraic geometry and the bethe ansatz , ann . of math .
( 2 ) 170 ( 2009 ) , no .
2 , 863881 . i. scherbak , rational functions with prescribed critical points , geom . funct .
12 ( 2002 ) , no . 6 , 13651380 .
g. tarantello , analytical , geometrical and topological aspects of a class of mean field equations on surfaces , discrete contin .
( 2010 ) , no .
3 , 931973 . m. troyanov , metrics of constant curvature on a sphere with two conical singularities , differential geometry ( pescola , 1988 ) , 29606 , lecture notes in math . , 1410 , springer , berlin , 1989 . m. troyanov , prescribing curvature on compact surfaces with conical singularities , trans .
soc . , 324 ( 1991 ) 793821 .
e. van vleck , a determination of the number of real and imaginary roots of the hypergeometric series , trans .
soc . , 3 ( 1902 ) 110131 .
a. veselov , on darboux treibich verdier potentials , lett .
96 ( 2011 ) 209216 . | we discuss conformal metrics of curvature @xmath0 on tori and on the sphere , with four conic singularities whose angles are multiples of @xmath1 . besides some general results we study in detail the family of such symmetric metrics on the sphere , with angles @xmath2 .
msc 2010 : 34m03 , 34m05 , 30c20 , 35j91 , 33e05 .
keywords : heun equation , darboux - treibich verdier potential , metrics with conic singularities , positive curvature , elliptic integrals . |
recent detections of afterglow emission in the x - ray , optical and radio has revolutionized gamma - ray burst ( grb ) astronomy ( for a recent review , see @xcite ) .
the afterglow emission appears to be reasonably well - described by the fireball model @xcite , in which a relativistically expanding shell of hot gas encounters an external quiescent medium .
a blast wave is created , sweeping up the ambient material , and accelerating relativistic electrons which then emit synchrotron radiation @xcite . in this model ,
the spectral flux of the afterglow is a broken power - law function of frequency which evolves with time in a way that depends on the physical parameters of the fireball and its surrounding medium , such as the total energy output of the grb source , the collimation angle of the outflow , the ambient gas density , the fraction of the post shock energy which is converted into magnetic fields and accelerated electrons , and the energy distribution of the accelerated electrons @xcite .
so far , observations of the afterglow spectrum over a wide range of frequencies and times was the primary method used to place constraints on these parameters ( e.g.@xcite ) .
crude constraints on the size of the emitting region at late times were derived based on considerations involving synchrotron self - absorption @xcite and radio scintillations due to the intervening interstellar medium of the milky - way galaxy @xcite . the afterglow image is predicted to appear as a thin ring on the sky @xcite at frequencies near or above the peak of the synchrotron emission @xcite , but more like a uniform disk around or below the synchrotron self - absorption frequency @xcite . in between breaks in the power - law spectrum of an afterglow ,
the contrast and width of this ring evolve self - similarly , i.e. the brightness profile maintains its shape as a function of radius when the latter is normalized by the expanding circular boundary of the image @xcite .
the self - similar brightness profile of the image changes only across spectral breaks @xcite .
the outer radius of the image expands radially at an apparent superluminal speed of @xmath7 , where @xmath8 is the lorentz factor of the emitting shock .
the image occupies an angle of @xmath9 relative to the center of the explosion due to relativistic beaming , and hence has a radius @xmath10 , where @xmath11 is the shock radius .
geometric time delay implies that the observed time , @xmath12 , and so the radius scales as @xmath13 . for an ambient density profile @xmath14
, one gets @xmath15 . therefore , @xmath16 with @xmath17 $ ] . for a uniform ambient medium with @xmath18 @xcite , @xmath19 , while for a wind profile with @xmath20 one gets @xmath21 .
the predicted differences between the afterglow images in these two cases are typically small @xcite .. ] by resolving the image of a grb afterglow at different frequencies , one can obtain precious , new constraints on the fireball model .
as originally pointed out by @xcite , after about one day , the angular size of a grb ring is typically a micro - arcsecond ( @xmath22as ) , i.e. of the same order as the angular einstein ring radius of a solar mass lens at cosmological distances , @xmath23 where @xmath24 is the lens mass , and @xmath25 , and @xmath26 , @xmath27 and @xmath28 are the angular diameter distances between the observer - source , observer - lens , and lens - source , respectively . this fortuitous coincidence , along with the fact that the afterglow image expands superluminally , means that lensing by an intervening solar - mass star will produce a detectable deviation in the afterglow light curve with a duration suitable for intense monitoring . as the afterglow image expands , different parts of it sweep by the lens and get amplified .
the differential magnification of the source image can be used to recover the radial structure of the afterglow image through a series of relative flux measurements @xcite .
the probability that a source at redshift @xmath29 will have a projected angular distance less than the einstein radius of any intervening star ( the optical depth ) is @xmath30 , where @xmath31 is the cosmological density of stars in units of the critical density .
the value of @xmath31 is constrained to be @xmath32 based of the known populations of luminous @xcite and faint @xcite stars and @xmath33 based on big - bang nucleosynthesis .
hence , @xmath34 @xcite , and roughly one out of a hundred afterglows should be strongly magnified ( although all events may show weak magnification signals due to a star located at ten einstein radii from their line - of - sight ; see @xcite ) .
intense monitoring of lensing events could provide invaluable information about the nature of grb afterglows .
recently , @xcite interpreted the unusual optical afterglow light curve of grb 000301c @xcite as being microlensed by an intervening @xmath35 star .
they found that the optical - infrared light curve was well - fit by a model in which the emission originated from a ring of fractional width @xmath5 , in agreement with earlier theoretical predictions .
@xcite argued that , if one adopts realistic surface brightness profiles for the afterglow image , the optical / infrared afterglow of grb 000301c can not be explained by microlensing .
@xcite showed , however , that microlensing can reproduce the observed light curve , provided that the source is significantly limb - brightened , namely that @xmath36 of the flux arises from the outer 25% of the area of the afterglow .
this requirement is met by the surface brightness profiles expected for emission frequencies above the cooling break frequency @xcite .
unfortunately , the light curve coverage in radio frequencies was too sparse to uniquely confirm the microlensing interpretation through , e.g. comparison of the inferred surface brightness profiles across spectral breaks .
the idea of using microlensing to study the images of gamma - ray burst afterglows is similar in spirit to the idea of using microlensing to resolve the atmospheres of stars in the magellenic clouds and the galactic bulge @xcite ; for a recent review , see @xcite . for background stars ,
the source radius stays constant and the resolution of the source occurs because the caustic structure of the lens sweeps over the face of the star due to the relative source - lens motion .
however , for grb afterglows the relative source - lens motion is negligible , and the resolution occurs as the image of the source itself is expanding and sweeping by the lens .
microlensing has been used to measure limb - darkening of the stars in the small magellenic cloud and the galactic bulge @xcite and also to measure the spatial variations of spectral lines across the face of two stars @xcite . @xcite
considered the signal - to - noise requirements for resolving stellar atmospheres with microlensing .
our goal here is to apply similar considerations to grb afterglows . in this paper
, we examine the observational constraints that can be obtained on afterglow images through a concerted monitoring effort of the grb community in future microlensing events . specifically , we determine the expected statistical errors in the recovered intensity profile as a function of various parameters and input assumptions , assuming that the afterglow and microlensing event parameters are known .
this significance of this work is twofold : first , we show that the radial intensity profile of afterglows can be determined with fairly good accuracy ( @xmath37 ) for reasonable expense of observational resources , and second , we present the effects of various assumptions on the error in the recovered intensity profile , in order to test the robustness of the conclusions and to provide guidelines for observers . in [ sec : grbs ] we present the formalism for microlensing of grb afterglows .
we describe the method for determining the expected errors in the recovered intensity profile in [ sec : errors ] , and apply it to a fiducial case in [ sec : scaling ] . in [ sec : inputs ] , we determine the effects of various input assumptions on the errors . finally , we summarize the implications of our primary results and discuss the effects of some of our ( unavoidably ) simplified assumptions in [ sec : conclusions ] . for concreteness , we will primarily focus on the scaling laws of a spherical fireball in a uniform medium , but derive a general scaling relation for the recovered errors in the relative intensity profile that is applicable for any power - law external density profile , @xmath38 . the two currently popular models , @xmath18 ( uniform medium ) and @xmath20 ( stellar wind ) yield qualitatively similar conclusions .
we do not consider jetted outflows , although a collimated outflow of opening angle @xmath39 would behave as if it is part of a spherical fireball at early time as long as @xmath40 @xcite .
we also assume that the afterglow flux exhibits a single power - law slope , which is appropriate if the observed frequency does not cross one of the several breaks in the afterglow spectrum ( see @xcite ) .
note that we do not assume any specific , model - dependent form for the intensity profile , i.e. we do not evaluate the errors on the coefficients of some parametric form for the profile .
rather , we evaluate the minimum attainable errors via direct inversion of the afterglow light curve .
for a spherical fireball , the observer sees a circular source image with an outer radius of angular size @xmath41 that evolves as , @xmath42 where @xmath43 is the angular size of the afterglow after @xmath44 , @xmath45 is the observed time in days , and @xmath46/[2(4-k)]$ ] for an external medium with density profile @xmath14 .
the angular size @xmath47 depends on the energy of the burst , the density of the ambient medium , and the redshift of the burst . for typical parameters ,
@xmath43 is of order @xmath48 .
as long as an observed frequency , @xmath49 , does not cross any of the time - dependent spectral break frequencies , the afterglow flux at that frequency evolves as a power - law of time ( e.g. , sari et al .
1999 ) @xmath50 where @xmath51 is the flux after @xmath44 .
@xcite provide a comprehensive study of the predicted image profile in the different spectral regimes .
now assume that an intervening compact object lies at an angular distance @xmath52 from the line - of - sight to the center of the afterglow .
we may then normalize the angular radius of the afterglow image in units of the einstein radius of this object ( eq . [ eqn : thetae ] ) , @xmath53 where @xmath54 since both @xmath43 and @xmath55 are @xmath56 for typical parameters , @xmath57 is of order unity .
the value of @xmath57 depends very weakly on the grb energy output and the ambient density normalization , but is more sensitive to the lens mass and the source and lens redshifts ( for @xmath18 , see eqs . 2 & 3 in garnavich et al .
2000 ) .
the magnification of an extended source is given by , @xmath58 where @xmath59 is the point - source magnification at vector position @xmath60 relative to the lens , and the integral is over the area of the source .
for lensing of a uniform circular source of radius @xmath61 by a single point mass with no external shear , the magnification obtains the value @xmath62 $ ] , where @xcite , @xmath63 .
\label{eqn : aus}\ ] ] here @xmath64 is the step function and @xmath65 is the angular separation ( impact parameter ) between the source center and the lens in units of the einstein ring radius . for a uniform source ( as applicable below the synchrotron self - absorption frequency ; see granot et al . 1999b and granot & loeb 2001 ) , the magnification history by a point - mass lens is completely specified by two parameters : @xmath66 and @xmath57 .
figure [ fig : fig2 ] shows the magnification as a function of time in days for @xmath67 and @xmath68 , and two density profiles , uniform ( @xmath69 ) and @xmath70 ( @xmath71 ) . at early times
when the source radius is small , the magnification is roughly fixed at its point - source value @xmath72 .
the peak magnification occurs at a time @xmath0 when @xmath73 , i.e. @xmath74 . the total flux from the afterglow is then by , @xmath75 where we have allowed for flux @xmath76 from any unresolved sources not being lensed , e.g. the host galaxy of the grb .
thus , in the simplest scenario the flux at a given frequency is a function of five parameters : @xmath77 , and @xmath66 . in order to illustrate how microlensing effectively resolves the image of the grb afterglow , we plot in the fraction of @xmath78 contributed by five equal area annuli as a function of time assuming @xmath79 , and a uniform source . for @xmath80 ,
the annuli contribute roughly equal flux , and the source is not resolved .
however , beginning at @xmath81 different annuli obtain different weights and the lens differentially magnifies the source .
the annulus that contributes most of the flux at a given time has a radius of @xmath82 in units of @xmath55 , where @xmath83 is the radius of that annulus at @xmath84 . defining the fractional radius @xmath85 , @xmath86 from inspection of and ,
it is clear that the the light curve need only be monitored from @xmath0 until @xmath87 to resolve the outer @xmath88 of the radial profile of the afterglow image .
consider an azimuthally symmetric afterglow divided into @xmath89 annuli , with annulus @xmath90 centered at radius @xmath91 .
we assume a self - similar behavior for which both the fractional radii @xmath92 and the mean relative intensity @xmath93 in each annulus @xmath90 are constants in time .
the total flux from the afterglow is simply the sum of the fluxes from each annulus .
the weight of each annulus is in turn the area of the image of each annulus of width @xmath94 , @xmath95 , \label{eqn : areai}\ ] ] times the mean intensity of that annulus . converting from an integral to a finite sum
, the total flux is simply , @xmath96 the coefficients one wishes to determine are the @xmath97 relative intensities @xmath98 .
these can be determined through inversion of the observed flux @xmath99 if @xmath100 , and @xmath101 are known .
the functions @xmath102 and @xmath101 in turn depend on the parameters @xmath103 , and @xmath57 .
schematically , the inversion can be done as follows : @xmath104 and @xmath76 can be determined from the late - time ( @xmath105 ) behavior of the afterglow light curve , when the magnification due to microlensing is negligible .
the unlensed flux of the afterglow could then be extrapolated back to @xmath84 , to determine @xmath51 .
the offset between the lensed and unlensed flux at early times @xmath106 then gives @xmath66 .
finally , a measurement of @xmath66 combined with @xmath0 provides @xmath57 .
therefore , all the parameters necessary to invert can ideally be determined without reference to the surface brightness distribution of the source image . in practice , of course , a global fit to all @xmath107 parameters must be done simultaneously , and this will introduce correlations between the parameters @xmath98 and @xmath108 , in turn inflating the errors on the measured values of @xmath98 . while @xmath78 is linearly dependent on @xmath98 ( see ) , this is not true for the parameters @xmath108 .
the nonlinear dependences of these parameters on the observed quantity @xmath78 make the expected errors on these parameters and their covariances with the parameters of interest , @xmath98 , extremely difficult and time consuming to calculate .
furthermore , the errors will depend sensitively on the exact coverage and accuracy of the afterglow photometry .
for an observed microlensing event , a full and careful determination of the errors on the inferred values of @xmath98 must account for their covariances with the other fit parameters .
however , including these effects here is beyond the scope of the paper .
therefore , when computing the error bars on @xmath98 , we will for simplicity assume that @xmath108 are perfectly known . we stress that the errors we derive should thus be regarded as lower limits to the actual errors .
now consider a series of flux measurements @xmath109 that are made at times @xmath110 with errors @xmath111 , and that are fit to .
the parameters of the fit are @xmath112 , and the variances in these parameters are @xmath113 , where @xmath114 is the covariance matrix , @xmath115 and @xmath116 .
the fractional errors are then @xmath117 . we assume poisson - noise limited precision , so that @xmath118 .
therefore , for a given set of assumptions about the microlensing event ( @xmath119 ) , the unlensed flux of the afterglow ( @xmath120 ) and the observational setup ( duration of observations , telescope diameter , etc . ) , the covariance matrix @xmath121 can be formed and the expected errors @xmath122 determined .
we now apply the formalism in [ sec : errors ] to derive an approximate scaling relation for @xmath122 . rather than adopt specific ( and therefore model dependent ) forms for the afterglow flux @xmath51 as a function of frequency and a specific observational setup
, we will simply assume that the instrument being used collects @xmath123 photons per second from the unlensed afterglow at @xmath84 . in [ sec : discussion
] , we will evaluate values of @xmath123 and thus the errors expected from a specific example of afterglow emission and observational setup . for our fiducial scenario , we assume a uniform source ( @xmath124 ) , a negligible background flux ( @xmath125 ) , and that observations are made continuously from the burst until very late times @xmath105 .
we also assume equal - area annuli , so that all bins have equal weight for a uniform source .
we then calculate @xmath122 for a range of values of @xmath104 , @xmath57 , @xmath66 , and @xmath97 .
numerically , we find the following approximate scaling relation for the fractional errors in the recovered intensity profile , @xmath126 equation ( [ eqn : scale ] ) is the primary result of this paper .
we find that it predicts the expected errors to an accuracy of @xmath127 for most combinations of parameters .
note that , and indeed the majority of the quantitative results in this paper , rely on four assumptions about the afterglow , namely : ( i ) the unlensed flux of the afterglow is a power - law function of time , @xmath128 ; ( ii ) the radius of the afterglow image scales as @xmath16 ; ( iii ) the image profile evolves self - similarly ; and ( iv ) the afterglow image is circularly symmetric .
these assumptions should hold at least approximately for some restricted part of most afterglow light curves .
we discuss these assumptions more thoroughly in
[ sec : conclusions ] .
the above equations can be used to determine the expected errors on the recovered intensity from observations performed with various instruments , apertures , and photon frequencies of afterglow light curves of arbitrary fluxes and power - law indices . because we have assumed that the observational precision is limited only by photon statistics , that there are no correlations with the other parameters ( see the discussion in [ sec : errors ] ) , and that the measurements are continuous , equation ( [ eqn : scale ] )
provides the minimum attainable error , at least for uniform sources and isolated lenses with no external shear . in the next section
, we will relax some of the assumptions leading to this result in order to evaluate their effect on the expected errors .
for the two cases of a uniform medium ( @xmath69 ) and a stellar wind medium ( @xmath71 ) , the exponents in are quite similar . therefore
, for the sake of simplicity , we will consider only the uniform medium case with @xmath69 , in the discussion that follows .
several of the terms in can be derived analytically , simply by noting that , according to photon statistics , the errors should scale as @xmath129 , where @xmath130 is the number of photons collected .
for example , the scaling with @xmath57 can be derived as follows : the time of the peak of the light curve scales as @xmath131 .
the unlensed flux at this time is @xmath132 .
the magnification structure is independent of @xmath57 ; however , all times are scaled by @xmath133 , so the time over which photons can be collected scales as this factor .
thus the total error should scale as @xmath134 , as found numerically .
the scaling with the impact parameter @xmath66 can be derived in a similar fashion , under the assumption that the peak magnification scales as @xmath135 . however , this is only strictly valid for @xmath136
. therefore , the scaling @xmath137 in is only approximate , and breaks down for @xmath138 .
numerically , we find that for @xmath139 , predicts errors that are too small by @xmath140 for @xmath141 ( the outer 70% of the area of the image ) .
the exponent for the scaling with @xmath97 is @xmath142 , similar to that found by @xcite for resolving the images of galactic stars by microlensing .
naively , one might expect the prefactor in to be considerably smaller : for @xmath143 and @xmath68 the total number of photons collected is of order @xmath144 ( with only a logarithmic dependence on the integration time during which the source is resolved ) .
divided over @xmath145 bins , one might expect the error per bin to be @xmath146 , about an order of magnitude smaller than the @xmath147 predicted by for a radius @xmath148 .
however , the errors increase with @xmath97 faster than one would naively expect , and so the error in each bin is closer to @xmath149 .
the most severe simplification made in deriving is the assumption of continuous measurement from the grb trigger until long after the peak time @xmath0 .
since the optical depth for microlensing is @xmath150 , at least @xmath151 afterglows must be monitored to detect one that has impact parameter @xmath152 .
if it was truly necessary to monitor all of these afterglows continuously to recover the intensity profile accurately , this would represent an overwhelming observational burden .
we therefore evaluate the errors expected under various assumptions about the starting time and the duration of the observations .
we scale the starting time , @xmath153 , and ending time , @xmath154 , in terms of the peak time through the relation @xmath155 , so that for given @xmath153 and @xmath154 , a fixed range of fractional radii , namely @xmath156 through @xmath157 , are probed regardless of the values of @xmath57 and @xmath66 ( see ) .
the top panel of shows the expected errors @xmath122 normalized to the fiducial errors in equation ( [ eqn : scale ] ) , for @xmath158 , corresponding to complete coverage of the microlensing event , and also for @xmath159 ( starting at the peak of the event ) and various values of @xmath154 .
note that here and throughout we plot @xmath122 as a function of @xmath160 , because we assume bins of equal area ( and hence equal weight for a uniform source ; see eq .
[ eqn : areai ] ) .
clearly observations before the peak of the event provide little information on the intensity profile .
moreover , the afterglow need only be monitored until @xmath161 to resolve the outer two thirds of the source radius .
the bottom panel of shows the expected errors when @xmath153 is varied .
delaying observations even somewhat past the peak of the event will result in seriously degraded information .
we conclude that measurements from @xmath0 until @xmath162 will provide almost as much information as continuous measurements from the grb trigger until late times .
since for typical parameters @xmath163 , the afterglow need only be aggressively monitored for about six days .
this suggests the following detection strategy for maximizing the number of lensed grbs for which the intensity profile can be recovered .
since the sampling necessary to determine that the grb is being microlensed is not as dense as that needed to recover the intensity profile , a large number of bursts can be monitored relatively infrequently from a global network of small telescopes . with real - time reduction , this network should be able to `` alert '' microlensing events before or near the peak time .
larger telescopes could then be used to densely sample the microlensed afterglow for the @xmath164 days necessary to resolve the intensity profile .
this is similar to the way galactic microlensing observations are coordinated @xcite . in this section ,
we explore the effects of realistic intensity profiles on the resultant errors . the optical image of the afterglow is expected to be limb - brightened , because light from the edge of the afterglow suffers the longest time delay , and thus was emitted at earlier times when the fireball was brighter ( waxman 1997c ; sari 1998 ; panaitescu & meszaros 1998 ; granot , piran , & sari 1999a , b ) . the contrast between the center and edge of the afterglow , as well as the sharpness of the cut - off in intensity at the outer edge of the afterglow , depends on the location of the observed frequency relative to the various spectral breaks in the broken power - law spectrum of afterglows ( granot et al .
1999a , b ; granot & loeb 2001 ) .
the exact shape of the relative intensity profile will affect the shape of the light curve , as well as the resultant errors , with the errors expected to be larger near the center ( where the relative intensity is smaller ) than those calculated assuming a uniform source .
previous studies of microlensing @xcite have parameterized the intensity profile as a uniform inner disk bounded by a uniform outer ring , with two parameters specifying the width @xmath165 and contrast @xmath166 of the ring .
of course , realistic intensity profiles do not exhibit such discontinuous features .
the precise calculation of these profiles for the different spectral regimes of grb afterglows is presented elsewhere ( granot & loeb 2001 ) . here , we adopt a parameterized form for the relative intensity profile that captures the qualitative aspects of realistic profiles , @xmath167(1-x^n ) , \label{eqn : iprof}\ ] ] where @xmath168 and @xmath169 are free parameters that define the shape of the profile , and @xmath170 is the normalization such that @xmath171 .
the index @xmath169 defines the sharpness of the cut - off , and is roughly analogous to the width of the ring @xmath165 .
the coefficient @xmath168 is related to the intensity at @xmath172 , and is roughly analogous to the contrast @xmath166 .
note that for @xmath173 and @xmath174 , this form reduces to a uniform source .
the top panel of shows @xmath175 for a relatively gradual cutoff , @xmath176 , and various values of @xmath168 .
the bottom panel of shows the fractional errors @xmath177 relative to the fiducial errors for a uniform source in equation ( [ eqn : scale ] ) , given the image profiles shown in the top panel .
note that we normalize the variances @xmath178 to the mean intensity @xmath170 rather than the intensity in that bin in order to avoid the fractional errors @xmath179 blowing up when @xmath180 for large @xmath168 .
in general , the resultant errors are a factor @xmath181 worse than than the uniform source case . in the most extreme example
where @xmath182 at the center of the image , the fractional errors can become quite large , but only for the innermost bin .
for the outer half of the area of the image , the errors can be smaller than the uniform source case .
thus , in this case , measurements will result in a robust upper limit to the intensity at the center of the image , with a clear measurement for the outer part of the image .
therefore the ring - like structure of the image will be accurately recovered .
we have also calculated , but do not show , the errors for the case of a sharp outer cutoff , @xmath183 .
the results for this case are qualitatively and quantitatively similar to the results for @xmath176 . we conclude that realistic intensity profiles will not result in significantly inflated errors relative to the uniform source estimate ( ) , at least for the majority of the source area .
we now consider more complicated models for the lens .
the majority of lenses are likely to reside in galaxies , and a significant fraction of these may be in binary systems . therefore , the assumption of an isolated single lens with no external shear used in the previous sections will not be valid .
we therefore consider two other types of lenses : a single lens with external shear @xcite , and a binary lens @xcite .
the former case is appropriate for a lens perturbed by either a wide binary companion or the potential of its host galaxy .
the latter is appropriate for binary - lenses with separations @xmath184 or smaller .
for a lens residing in a region of high optical depth , i.e. near the center of a galaxy , the magnification structure will not be well approximated as either an isolated lens with external shear or a binary lens . in this case
one may only be able to draw statistical inferences about the surface brightness profile of the source , as it will be difficult to reconstruct the magnification structure of the lens from the light curve alone .
we will not consider this regime here .
we compute the fractional errors in the recovered intensity profiles in the same manner as before , except that we must replace @xmath185 in with the uniform - source magnification appropriate for the given lens model .
we compute this magnification using the inverse ray - shooting method ( see , e.g. , @xcite ) as follows .
the generalized lens equation for point masses and external shear can be expressed in complex coordinates on the plane of the sky as @xcite , @xmath186 where @xmath187 is the complex source position , @xmath188 is the complex image position , @xmath189 is the shear , @xmath190 is the fractional mass of lens component @xmath191 , and @xmath192 is the position of this component . all distances are expressed in units of the angular einstein ring of the combined mass of the system . for a single lens , @xmath193 , @xmath194 , and @xmath195 . for a binary - lens with no external shear , @xmath196 , @xmath197 ,
the mass ratio is @xmath198 , the dimensionless projected separation is @xmath199 , and we choose the origin to be the midpoint of the binary .
we sample the image plane uniformly and densely , solving for the source position using the appropriate form of .
these positions are then binned in the source plane .
the magnification of each bin is then just the surface density of rays in the image plane divided by the surface density of rays in the source plane .
since this procedure conserves flux , the resulting magnification map can then be convolved with a source of arbitrary size to compute the magnification of the extended source .
we calculate magnification maps for @xmath200 and @xmath201 , and one binary - lens configuration with @xmath202 and @xmath203 . for the single lens with external shear ,
@xmath204 parameters are needed to specify the light curve , namely : the @xmath89 relative intensities @xmath205 , plus @xmath206 , the shear @xmath189 , and two parameters that specify the position of the grb with respect to the lens , which we denote as @xmath207 . for the binary lens ,
we assume that @xmath196 ; however two additional parameters are needed to describe the topology of the lens , namely @xmath208 and @xmath209 , and so there are a total of @xmath210 parameters .
when calculating the errors on the parameters @xmath205 , we will again assume that all other parameters are perfectly known .
here this assumption is somewhat less justified than the single lens case , not only because there are more parameters , but also because the effects of these parameters are in general not directly related to features in the light curve , and are likely to be subject to degeneracies or correlations , which will inflate the errors in the recovered @xmath205 .
such denegeracies or correlations are known to inflate the errors in limb - darkening measurements of stars using galactic binary - lens events , although not greatly so for well - measured light curves @xcite . in the galactic binary - lens case ,
determination of all parameter combinations that correspond to solutions to the observed light curve , and therefore identification of degeneracies and correlations between fit parameters , is a formidable task @xcite .
this is likely to also be true for grb afterglows .
a comprehensive study of parameter space is beyond the scope of this paper , and so we will simply assume that the binary lens parameters are known .
this assumption facilitates comparison with the isolated single - lens case , which is the main purpose of our discussion here .
we stress that the errors we derive under this assumption are almost certainly underestimates . in both the external shear and binary - lens case
, we will assume a uniform afterglow image ( see mao & loeb 2000 for other examples ) .
in we show the magnification light curves for the case of a single lens with external shear along the @xmath211-axis of magnitudes @xmath212 , and @xmath201 , both for a source centered on the x - axis of the lens , @xmath213 , and a source centered on the y - axis of the lens @xmath214 .
note that the caustics are most highly elongated along the x - axis , and thus for a given @xmath189 , light curves where the source is centered on the x - axis should show more dramatic deviations from the @xmath196 case than for sources centered on the y - axis ( mao & loeb 2000 ) . shows the resultant errors relative to the fiducial error estimate in equation([eqn : scale ] ) . in general , sources centered on the x - axis show more dramatic variations in the expected errors than sources centered on the y - axis . in all cases
the fractional errors are within a factor of @xmath215 of the @xmath196 expectation .
thus , at least for modest values of @xmath216 , we conclude that an external shear will not result in significantly inflated statistical errors , assuming that the covariances with the afterglow and microlensing parameters are small . in
, we show the magnification as a function of time for the binary lens with @xmath202 and @xmath203 , and a source centered at @xmath217 .
we also show the fraction of total flux contributed by five equal area bins as a function of time . for times
@xmath218 , the light curves resemble that for a single lens ( compare with ) , as the annuli become comparable to or larger than the caustic features .
the bottom panel of shows the resultant errors relative to the fiducial single - lens case .
again , the statistical errors differ by less than a factor of @xmath215 from the fiducial estimate in equation ( [ eqn : scale ] ) . observed afterglows often show evidence for a flattening of the light curve to a constant flux at late times .
this is usually interpreted as due to unresolved flux from the host galaxy of the grb . even if the flux from the host galaxy is small
, the sky background will eventually dominate over that of the afterglow .
in other words , it is inevitable that @xmath219 at late times . for ground - based infrared observations ,
the background flux is likely to be larger than the afterglow flux at all relevant times .
a finite background flux will increase the errors on the recovered parameters @xmath205 , and preferentially so for smaller radii , since the information on smaller radii comes at later times , when the background is more significant .
we explore this effect by introducing a background flux which amounts to some fraction of @xmath220 for the afterglow .
shows the expected fractional errors on the recovered intensity profile relative to the fiducial value in equation ( [ eqn : scale ] ) , assuming a power - law flux index of @xmath143 for the afterglow , for relative background fluxes of @xmath221 , and @xmath222 . at optical - infrared frequencies
, this corresponds to a background of magnitude @xmath223 relative to the ( unlensed ) afterglow at @xmath44 .
as expected , the errors generally increase as the background flux increases , especially for smaller radii . however ,
unless the background flux is higher than @xmath220 , the expected errors are @xmath181 times greater than the fiducial estimate . in other words , for optical frequencies , the background flux should not compromise the measurement of the intensity profile .
however , for , e.g. , ground - based observations in the near - infrared , where the sky background can be many magnitudes brighter than the afterglow flux at 1 day , the accuracy with which the intensity profile can be measured will considerably larger than the fiducial estimate ( ) . thus spaced - based near - infrared observations would improve the accuracy of the recovered intensity profile . as discussed in
[ sec : scaling ] , the results we have presented so far are quite general .
we now estimate the errors expected under realistic observational conditions and a realistic model .
to do so , we make estimates of the various terms in by adopting the parameters for the observed afterglow of grb 000301c .
@xcite fit the optical and infrared photometry of grb 000301c to double power - law flux decline magnified by the simple microlensing model of @xcite .
they find best - fit parameters of @xmath224 and @xmath225 for the power - law flux indices , similar to the values found originally by @xcite without including the microlensing modification . for simplicity ,
we ignore the break , and simply adopt a single power - law of slope @xmath226 .
we adopt the best - fit microlensing parameters for the entire dataset of @xmath227 and @xmath228 , somewhat smaller than that found by @xcite . for illustrative purposes
, we will simply adopt the @xcite value of @xmath228 . ] . finally , we extrapolate by eye the unlensed flux of the afterglow back to @xmath229 for the @xmath230 and @xmath231-bands .
table 1 shows the demagnified , extrapolated magnitudes at @xmath84 of the afterglow of grb 000301c , along with the assumed sky brightness , and @xmath232 , the number of photons per second per square meter of collecting telescope area @xmath233 .
note that , for @xmath227 and @xmath228 , the peak of the microlensing event ( for a uniform source ) occurs at @xmath234 .
we calculate the expected errors @xmath179 assuming a 4-m telescope , a seeing of @xmath235 , a uniform source , and observations between the peak , @xmath236 and @xmath237 .
we adopt a uniform source in order to provide a model - independent estimate of the fractional errors expected for realistic observing conditions .
image profiles in the optical and infrared are unlikely to be well - represented by a uniform source @xcite , however this will not affect the errors in the recovered profile by more than a factor of @xmath238 ( see [ sec : iprofs ] ) .
the results are shown in . for observations in the optical ,
the relative intensity profile can be recovered accurately , @xmath239 . for observations in the infrared , where the sky background is dominant , the expected errors are @xmath5 .
although we do not specifically consider other wavelengths here , we note that radio afterglows peak later and last longer than afterglows at shorter wavelengths ( @xcite and references therein ) .
this , combined with the lower energy of the photons , implies that the photon flux at a given time is larger in the radio than in the optical . at sufficiently high radio frequencies where the effect of scintillations is sub - dominant
, radio observations may provide more accurate measurements of the afterglow image .
since the afterglow image is expected to be different at radio frequencies @xcite , we advocate performing measurements over as wide a range of frequencies as possible . given that the fraction of afterglows that exhibit microlensing events with impact parameter @xmath152 scales as @xmath240 , it is interesting to consider what kind of errors one would expect for the specific example of grb 000301c considered above , but with larger impact parameters .
we therefore repeat the calculation above with the same parameters , except we now vary the impact parameter . specifically , we consider @xmath241 , and @xmath242 , as would be expected in @xmath243 , and @xmath244 of all grb afterglows . in
we show the results for the @xmath245-band ( where the background is the smallest ) , assuming that observations are taken from peak , @xmath246 until @xmath247 . for @xmath248 ,
the fractional errors in the recovered intensity profile are @xmath37 for the majority of the area of the image .
thus , it should be possible to recover the intensity profile to an accuracy of @xmath140 for @xmath5 of all afterglows . to test the robustness of this conclusion
, we also show in the results for the @xmath249 case , but considering a band with larger background flux ( @xmath250-band ) , and also a non - zero external shear ( @xmath251 with source on the shear axis ) .
we find that the error is not greatly inflated , @xmath252 .
of course , these results assume that the afterglow can be monitored for many months after the peak of the microlensing event .
practical aspects aside , such monitoring may not be possible even in principle in those cases where the emission is due to a highly collimated jet .
such afterglows will eventually exhibit breaks in the light curve and changes in the image structure when the lorentz factor of the fireball falls below the inverse of the opening angle of the jet . in this case , the errors on the annuli centered at smaller radii would be inflated ( see [ sec : dobs ] ) .
additional deviations from our results may occur as the fireball decelerates to non - relativistic speeds @xcite .
microlensing offers a unique method for probing the dynamics of grb fireballs . as the grb afterglow image expands , it is differentially magnified by the lens .
different annuli of the image are resolved as they cross the lens position at different times .
thus the light curve of the afterglow during the microlensing event can be inverted to obtain the radial intensity profile @xmath175 , where @xmath253 is the fractional radius of the circular image .
we developed the formalism necessary to calculate the expected fractional errors on the recovered intensity profile @xmath179 .
assuming continuous observations , uniform sources , single lenses , and zero background flux , we find that the expected errors follow a general scaling relation ( eq .
[ eqn : scale ] ) in terms of the afterglow flux , afterglow radius in units of the einstein radius , the angular separation between the afterglow and the lens , and the number of resolution elements in the recovered intensity profile .
we have tested the accuracy of this relation by , in turn , relaxing the assumptions of continuous observations , uniform sources , single lenses , and zero background flux .
we find that , for reasonable choices of the relevant parameters , none of these refinements changes the expected errors by more than a factor of @xmath238 .
therefore , the fiducial scaling relation is fairly robust .
notably , we have found that observations starting at the peak @xmath0 of the microlensing event and ending at @xmath254 result in almost the same accuracy as measurements of the entire afterglow .
therefore , one can monitor frequently _ only _ those afterglows that are significantly microlensed by `` alerting '' to these events before they reach their peak . finally , we calculated the errors expected for observations in the near - infrared and optical regimes of a specific example .
adopting the observed and inferred parameters for the ( possibly ) microlensed optical / infrared afterglow of grb 000301c , we find that the relative intensity profile can typically be measured with a resolution of 10 and accuracy of @xmath3 in the optical and @xmath4 in the infrared using 4m - class telescopes .
fluctuations in the light curve due to inhomogeneities in the ambient gas distribution are typically small and should not compromise this precision @xcite . in the course of this study , we have had to make a number of simplifying assumptions . these include : ( i ) the unlensed flux of the afterglow is a power - law function of time , ( ii ) the radius of the afterglow image scales as a power - law of time @xmath16 , ( iii ) the afterglow image is circularly symmetric , ( iv ) the intensity profile is self similar , i.e. @xmath175 is independent of time , ( v ) the lens resides in a region of relatively low optical depth , and ( vi ) the covariances between the surface brightness profile parameters and the afterglow and microlensing parameters are small .
the first assumption is a generic prediction of the simplest afterglow models , provided that the observed frequency does not cross one of the several spectral breaks , and that @xmath9 is smaller than the collimation angle of the ejecta . when a break in the light curve occurs either due to the crossing of a spectral break or jet effects
the intensity profile is also likely to change , thus also violating assumption ( iv ) .
when the break is due to jet effects , or when the fireball becomes non - relativistic , then the index @xmath255 of the power - law form for the image radius will change .
furthermore , breaks in the light curve will produce additional uncertainties in the microlensing and afterglow parameters by introducing additional correlations .
for example , in the case of a single power - law flux decline , the impact parameter @xmath66 can be found by extrapolating the late - time light curve ( when the afterglow is not lensed ) to very early times , when the afterglow is lensed by an amount that depends only on @xmath66 .
clearly , this procedure becomes more uncertain when the light curve exhibits a break .
the assumption of circular symmetry may be crude at best , especially for collimated outflows .
for single lenses , it is clear that the information gathered about the surface brightness profile is inherently one dimensional , and therefore departures from circular symmetry will be difficult to detect .
therefore , the surface brightness profile inferred assuming circular symmetry will be biased . in principle
, binary lenses or single lenses with external shear would provide more information about departures from circular symmetry than a simple single lens ; in practice , however , the additional freedom introduce by the more complex lens models may make extracting this information difficult .
our assumption of small optical depth is likely to hold in the outskirts of galaxies . for sight
lines passing near the center of an intervening galaxy , the magnification structure of the lens is unlikely to be well represented by any of the three lenses we considered : an isolated single lens , a single lens with external shear , or a binary lens . in the high - optical depth regime
, it may be difficult to reconstruct the relative intensity profile directly , due to the complexity of the magnification structure .
finally , the assumption that the covariances between the intensity profile parameters and the afterglow and microlensing parameters are small is unlikely to hold in practice .
this means that the errors we calculate are effectively lower limits to the true errors .
the effects of these various assumptions on the resultant errors on the recovered intensity profile are interesting and important topics for future study .
our basic results are encouraging .
errors of @xmath256 would likely provide very stringent constraints on afterglow models , and even errors of @xmath257 would be interesting ( see fig . 2 in granot & loeb 2001 ) .
errors of the latter magnitude can be obtained from events with larger impact parameters . for the specific afterglow parameters adopted above
, we find errors of @xmath258 in the @xmath245-band for impact parameters in units of @xmath55 of @xmath259 and @xmath260 .
since the number of expected events scales as @xmath240 , this implies that interesting results might be obtained for a significant sample of afterglows .
the forthcoming swift satellite could provide hundreds of afterglow targets per year .
a network of 1 m class telescopes similar to that used in gravitational microlensing searches of the local group @xcite would provide an ideal method for `` alerting '' the community to _ microlensed _ afterglows , which could in turn be intensively monitored to provide information about the structure of the afterglow image .
although this represents a considerable expenditure of resources , the information gained would be invaluable , and furthermore can not be currently acquired by any other method .
we thank andrei gruzinov for helpful discussions . this work was supported in part ( for sg ) by nasa through 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 nas5 - 26555 , and ( for al ) by grants from the israel - us bsf ( bsf-9800343 ) , nsf ( ast-9900877 ) , and nasa ( nag5 - 7039 ) . @xmath261 + & magnitude & @xmath262 & @xmath263 + @xmath245 & 20.1 & 22.8 & 43 + @xmath264 & 20.6 & 22.5 & 77 + @xmath265 & 20.1 & 21.5 & 78 + @xmath250 & 19.7 & 20.8 & 130 + @xmath266 & 19.3 & 19.3 & 128 + @xmath267 & 18.3 & 15.9 & 194 + @xmath231 & 16.8 & 13.7 & 540 + | microlensing of a gamma - ray burst afterglow by an intervening star can be used to infer the radial structure of the afterglow image .
near the peak of the microlensing event , the outer edge of the image is more highly magnified than its central region , whereas the situation is reversed at later times due to the rapid radial expansion of the image on the sky .
thus , the microlensed afterglow light curve can be inverted to recover the self - similar radial intensity profile of the afterglow image .
we calculate the expected errors in the recovered intensity profile as a function of the number of resolution elements , under the assumption that the afterglow and microlensing event parameters are known . for a point - mass lens and uniform source
, we derive a simple scaling relation between these parameters and the resultant errors .
we find that the afterglow need not be monitored for its entire duration ; rather , observations from the peak magnification time @xmath0 of the microlensing event until @xmath1 are sufficient to resolve the majority of the afterglow image .
thus , microlensing events can be alerted by relatively infrequent observations of afterglows and then monitored intensively , without significant loss of information about the afterglow intensity profile .
the relative intensity profile of @xmath2 of all afterglows can be measured with 10 resolution elements to an accuracy of @xmath3 in the optical and @xmath4 in the infrared , using 4m - class telescopes .
weak microlensing events with large impact parameters are more common ; we estimate that for @xmath5 of afterglows the image profile may be inverted to a fractional accuracy @xmath6 through frequent optical observations .
we also calculate the effects of external shear due to the host galaxy or a binary companion , and contamination by background light from the host galaxy .
# 1#23.6pt 0(i / i)_0 |
suppose that @xmath0 is a complete and connected riemannian manifold of dimension @xmath1 , that @xmath2 is a closed embedded submanifold of @xmath0 of dimension @xmath3 and that @xmath4 is a brownian motion on @xmath0 starting at @xmath5 with explosion time @xmath6 .
the main objective of this paper is to study the distance between @xmath7 and @xmath2 for each @xmath8 .
this is not something which has previously been considered in the literature , the closest reference being the study of mean exit times given by gray , karp and pinksy in @xcite , and our main results are theorems [ th : estimatefinal ] , [ th : expenbm ] and [ th : concenthm ] .
we denote by @xmath9 the distance function and assume that there exist constants @xmath10 and @xmath11 such that the lyapunov - like condition @xmath12 holds off the cut locus of @xmath2 .
geometric conditions under which such an inequality arises , which allow for unbounded curvature , are given by theorem [ th : estimatefinal ] ( see corollary [ cor : maincor ] ) , which is derived from the classical heintze - karcher comparison theorem @xcite .
under such assumptions we deduce a variety of probabilistic estimates , which are presented in section [ se : section3 ] .
we do so first using a logarithmic sobolev inequality .
such inequalities were originally studied by gross in @xcite and we refer to the article @xcite for the special case of the heat kernel measure .
we then prove more general estimates using the it - tanaka formula of section [ se : section2 ] , which is derived from the formula of barden and le @xcite and which reduces to the formula of cranston , kendall and march @xcite in the one - point case .
the basic method is similar to that of hu in @xcite , who studied ( uniform ) exponential integrability for diffusions in @xmath13 for @xmath14 functions satisfying another lyapunov - like condition . indeed ,
several of our results could just as well be obtained for such functions , but we choose to focus on the distance function since in this case we have a geometric interpretation for the laplacian inequality .
section [ se : section2 ] also includes a characterization of the local time of brownian motion on a hypersurface and a couple of examples .
the probabilistic estimates of section [ se : section3 ] include theorem [ th : secradmomthm ] and its generalization theorem [ th : evenradmomthm ] , which provide upper bounds on the even moments of the distance @xmath15 for each @xmath8 .
using properties of laguerre polynomials we use these estimates to deduce theorems [ th : inequality2 ] and [ th : expenbm ] , which provide upper bounds on the moment generating functions of the distance @xmath15 and squared distance @xmath16 for each @xmath8 , respectively .
the latter estimate improves and generalizes a theorem of stroock . using theorem
[ th : expenbm ] and markov s inequality we then deduce theorem [ th : concenthm ] , which provides an concentration inequality for tubular neighbourhoods .
note that in this paper we do not assume the existence of gaussian upper bounds for the heat kernel ; the constants appearing in our estimates are all explicit .
this paper has been presented in such a way that it should be possible for the reader to read section [ se : section3 ] either before or after sections [ se : section1 ] and [ se : section2 ] .
finally , the author wishes to thank his supervisor xue - mei li for supporting this piece of work with numerous insightful suggestions .
he also wishes to thank david elworthy and an anonymous referee for providing additional comments and criticism .
we begin by deriving an inequality for the laplacian of the distance function .
this object can be written in terms of the jacobian determinant of the normal exponential map , so we begin with a comparison theorem and inequalities for this object .
the main references here are @xcite and @xcite .
suppose that @xmath0 is a complete and connected riemannian manifold of dimension @xmath1 with riemannian volume measure @xmath17 and that @xmath2 is a closed embedded submanifold of @xmath0 of dimension @xmath18 with induced measure @xmath19 and normal bundle @xmath20 .
fix @xmath21 and choose @xmath22 where @xmath23 denotes the first focal time along the geodesic @xmath24 with @xmath25 and @xmath26 .
if @xmath27 with @xmath28 a two - dimensional subspace of @xmath29 then denote by @xmath30 the sectional curvature of @xmath28 and let @xmath31 \end{aligned } \right\}.\ ] ] as regards the extrinsic geometry of the submanifold , denote by @xmath32 the shape operator associated to @xmath33 and by @xmath34 the eigenvalues of @xmath32 .
these eigenvalues are called the _ principle curvatures of @xmath2 with respect to @xmath33 _ and their arithmetic mean , denoted by @xmath35 , is called the _
mean curvature of @xmath2 with respect to @xmath33_. let @xmath36 denote the jacobian determinant of the normal exponential map @xmath37 ( which is simply the exponential map of @xmath0 restricted to @xmath38 ) and for @xmath39 define , for comparison , functions @xmath40 , @xmath41 , @xmath42 and @xmath43 by @xmath44 @xmath45 g_{\kappa}(t)&:= \frac{d}{dt}\log ( s_\kappa(t)/t)\\[0.2 cm ] f^\lambda_\kappa(t)&:= \frac{d}{dt}\log ( c_{\kappa } ( t ) + \lambda s_{\kappa } ( t ) ) .
\end{split}\ ] ] if @xmath46 is _ any _ constant such that @xmath47 then , as was proved by heintze and karcher in @xcite using a comparison theorem for jacobi fields , and which was generalized somewhat by kasue in @xcite , there is the inequality @xmath48 for all @xmath49 . furthermore
, if @xmath50 satisfies @xmath51 and @xmath52 is _ any _ constant such that @xmath53 then for @xmath54 there is the inequality @xmath55 for all @xmath49 and for @xmath56 there is the inequality @xmath57 for all @xmath49 .
heintze and karcher s method implies that the right - hand sides of inequalities , and are finite for all @xmath49 .
note that the ` empty sum is zero ' convention is used to cover the case @xmath54 in inequality .
we can use the heintze - karcher inequalities to deduce secondary estimates .
[ prop : propos ] for @xmath58 we have @xmath59 by gronwall s inequality and the fact that @xmath60 , this differential inequality implies an upper bound on @xmath61 . by a change of variables , upper bounds on @xmath61
imply upper bounds on the volumes of tubular neighbourhoods and the areas of the boundaries . in this way we can obtain estimates on these objects which are more explicit than those found in @xcite .
we will not present these calculations here ; they can be found in @xcite . to prove the proposition we will use two preliminary lemmas .
[ lem : case1 ] if @xmath62 then for @xmath49 we have @xmath63 setting @xmath64 we have @xmath65 and @xmath66 and from inequality it follows that @xmath67 for all @xmath49 .
the result follows by considering the cases @xmath68 and @xmath69 separately .
[ lem : case3 ] if @xmath70 then for @xmath49 we have @xmath71 fix @xmath72 and @xmath11 .
note that @xmath73 , @xmath74 and that by taylor s theorem the derivative of this function is strictly positive for positive @xmath75 .
therefore @xmath76 for @xmath77 and @xmath78 .
note that for each @xmath79 we have @xmath80 so @xmath81 is increasing on @xmath82 $ ] if and only if @xmath83 . if @xmath84 then @xmath85 is nonincreasing and @xmath86 .
conversely if @xmath83 then @xmath87 so @xmath88 is defined on @xmath89 and @xmath90 the lemma then follows from inequality by setting @xmath91 .
we can now prove propostion [ prop : propos ] . by lemmas [ lem : case1 ] and [ lem : case3 ]
it follows that @xmath92 & \leq & \sum_{i=1}^{n}\lambda_i(\xi ) \mathbb{1}_{\{{\underline{{\kappa}}_\xi(t_1 ) } \geq 0 \ } } + ( { m}-{n}-1)\sqrt{-{\underline{{\kappa}}_\xi(t_1 ) } } \mathbb{1}_{\{{\underline{{\kappa}}_\xi(t_1 ) } < 0\ } } \nonumber \\ & & + \sum_{i=1}^{{n } } \left ( \sqrt{-\underline{{\kappa}}_\xi(t_1)}\mathbb{1}_{\ { \vert \lambda_i(\xi ) \vert < \sqrt{-\underline{{\kappa}}_\xi(t_1)}\ } } + \lambda_i(\xi ) \mathbb{1}_{\{\vert \lambda_i(\xi ) \vert \geq \sqrt{-{\underline{{\kappa}}_\xi(t_1 ) } } \ } } \right ) \mathbb{1}_{\{{\underline{{\kappa}}_\xi(t_1 ) } < 0\ } } \nonumber \\[0.2 cm ] & \leq & \sum_{i=1}^{n}\vert\lambda_i(\xi)\vert \mathbb{1}_{\{{\underline{{\kappa}}_\xi(t_1 ) } \geq 0 \ } } + ( { m}-{n}-1)\sqrt{\vert{\underline{{\kappa}}_\xi(t_1 ) \wedge 0}\vert } \nonumber \\ & & + { n}\sqrt{\vert{\underline{{\kappa}}_\xi(t_1 ) \wedge 0}\vert } + \sum_{i=1}^{{n } } \vert\lambda_i(\xi)\vert \mathbb{1}_{\{{\underline{{\kappa}}_\xi(t_1 ) } < 0\ } } \nonumber \\[0.2 cm ] & = & ( { m}-1)\sqrt{\vert{\underline{{\kappa}}_\xi(t_1 ) \wedge 0}\vert } + \sum_{i=1}^{n}\vert \lambda_i(\xi)\vert \nonumber\end{aligned}\ ] ] as required .
note that the factor @xmath93 is reasonable since an orthonormal basis of a tangent space @xmath94 gives rise to precisely @xmath93 orthogonal planes containing the radial direction @xmath95 .
denote by @xmath96 the largest domain in @xmath38 whose fibres are star - like and such that @xmath97 is a diffeomorphism onto its image .
then that image is @xmath98 , where @xmath99 denotes the cut locus of @xmath2 .
recall that @xmath99 is a closed subset of @xmath0 with @xmath17-measure zero . with @xmath100
defined by @xmath101 the vector field @xmath102 will denote differentiation in the radial direction , which is defined off the union of @xmath2 and @xmath99 to be the gradient of @xmath9 and which is set equal to zero on that union .
if as in @xcite we define a function @xmath103 by @xmath104 we then have the following corollary of proposition [ prop : propos ] , in which @xmath105 denotes the distance to the cut locus along @xmath24 .
[ cor : corone ] suppose that there is a function @xmath106 such that for each @xmath107 and @xmath108 we have @xmath109 .
furthermore , suppose that the principal curvatures of @xmath2 are bounded in modulus by a constant @xmath110 .
then there is the estimate @xmath111 on @xmath98 . for each @xmath112 and @xmath108
we see by proposition [ prop : propos ] that @xmath113 since for each @xmath114 there exists a unique @xmath115 such that @xmath116 , the result follows for such @xmath117 by setting @xmath118 . for @xmath119 the radial derivative
is set equal to zero in which case the result is trivial .
furthermore , following from remarks made at the end of subsection [ ss : hkest ] , if there is a function @xmath120 such that for each @xmath107 and @xmath108 we have @xmath121 then for @xmath54 there is the estimate @xmath122 on @xmath123 and for @xmath56 with @xmath124 for each @xmath107 there is the estimate @xmath125 on @xmath123 .
thus we arrive at the main results of this section , which are the following theorem and its corollary .
[ th : estimatefinal ] suppose that @xmath0 is a complete and connected riemannian manifold of dimension @xmath1 and that @xmath2 is a closed embedded submanifold of @xmath0 of dimension @xmath18 .
denote by @xmath99 the cut locus of @xmath2 and @xmath9 the distance to @xmath2 and suppose that there exist constants @xmath126 such that one of the following conditions is satisfied on @xmath127 : ( c1 ) : : the sectional curvatures of planes containing the radial direction are bounded below by @xmath128 and there exists a constant @xmath129 such that the principal curvatures of @xmath2 are bounded in modulus by @xmath130 ; ( c2 ) : : @xmath54 and the ricci curvature in the radial direction is bounded below by @xmath131 ; ( c3 ) : : @xmath56 and the ricci curvature in the radial direction is bounded below by @xmath131 and there exists a constant @xmath110 such that the mean curvature of @xmath2 is bounded in modulus by @xmath130 .
then for @xmath132 , defined by , we have the inequality @xmath133 where @xmath102 denotes differentiation in the radial direction .
note that if @xmath54 then the mean curvature is not relevant and if @xmath134 then the sectional curvatures are not relevant but that the above estimates still make sense in either of these cases .
recall that we are primarily interested in the laplacian of the distance function .
if @xmath135 denotes the laplace - beltrami operator on @xmath0 then , as shown in @xcite , there is the formula @xmath136 on @xmath123 .
this yields the following corollary .
[ cor : maincor ] under the conditions of theorem [ th : estimatefinal ] we have @xmath137 on @xmath98 .
for the particular case in which @xmath2 is a point @xmath117 , it was proved by yau in @xcite that if the ricci curvature is bounded below by a constant @xmath138 then the laplacian of the distance function @xmath139 is bounded above by @xmath140 plus a constant depending on @xmath138 .
yau then used analytic techniques in @xcite to prove that this bound implies the stochastic completeness of @xmath0 .
a relaxation of yau s condition which allows the curvature to grow like a negative quadratic in the distance function is essentially optimal from the point of view of curvature and nonexplosion ; this is why we did not feel it necessary to present theorem [ th : estimatefinal ] in terms of a general growth function , although one certainly could .
we will return to this matter in subsection [ ss : exittimes ] . if in yau s example we set @xmath141 then inequality and taylor expansions imply @xmath142 for all @xmath8 if @xmath143 or for @xmath144 if @xmath145 , which yields the simple estimate @xmath146 on @xmath147 .
this has the advantage of taking into account the effect of positive curvature and in turn yields the laplacian estimate @xmath148 on @xmath147 , which is different to yau s bound .
note that inequalities and actually hold on the whole of @xmath0 in the sense of distributions .
in this section we show how the distance function relates to the local time of brownian motion on a hypersurface .
since the boundaries of regular domains are included as examples , this could yield applications related to the study of reflected brownian motion .
the main references here are @xcite , @xcite and @xcite .
the articles @xcite and @xcite approach geometric local time in the general context of continuous semimartingales from the point of view of tanaka s formula while @xcite approaches the topic for the special case of brownian motion using markov process theory .
suppose that @xmath149 is a brownian motion on @xmath0 defined up to an explosion time @xmath150 and that @xmath151 is a horizontal lift of @xmath149 to the orthonormal frame bundle with antidevelopment @xmath152 on @xmath13 .
in @xcite it was proved , using the theory of viscosity solutions , that the cut locus of @xmath2 is given by the disjoint union of two subsets @xmath153 and @xmath154 where the connected components of @xmath153 , of which there are at most countably many , are smooth two - sided @xmath93-dimensional submanifolds of @xmath0 and where @xmath154 is a closed subset of @xmath0 of hausdorff dimension at most @xmath155 ( and therefore polar for @xmath149 by @xcite ) .
also @xmath156 has @xmath17-measure
zero so it follows that @xmath157 where @xmath158 is a standard one - dimensional brownian motion , by lvy s characterization and the fact that @xmath151 consists of isometries .
furthermore , points belonging to @xmath153 can be connected to @xmath2 by precisely _ two _ length - minimizing geodesic segments , both of which are non - focal . using these observations
, it follows from ( * ? ? ?
* theorem 1 ) that @xmath159 is a continuous semimartingale .
in particular , if @xmath160 is a stopping time with @xmath161 then there exist continuous adapted nondecreasing and nonnegative processes @xmath162 and @xmath163 , whose associated random measures are singular with respect to lebesgue measure and supported when @xmath149 takes values in @xmath2 and @xmath153 , respectively , such that @xmath164 for all @xmath8 , almost surely , where @xmath165 here @xmath166 is _ any _ unit normal vector field on @xmath153 and the gteaux derivatives @xmath167 are defined for @xmath168 and @xmath169 by @xmath170 and @xmath171 . a detailed explanation of precisely how formula is derived from ( * ? ? ?
* theorem 1 ) can be found in the author s doctoral thesis .
note that the integral appearing in formula ( [ eq : rnformulaforbm ] ) is well - defined since the set of times when @xmath172 has lebesgue measure zero .
the process @xmath163 is given by the local time of @xmath173 at zero so long as the latter makes sense , while for the one - point case the process @xmath174 coincides with the geometric local time introduced in @xcite . the process @xmath162 , which we will refer to as _ the local time of @xmath149 on @xmath2 _ , satisfies @xmath175 where @xmath176 denotes the ( symmetric ) local time of the continuous semimartingale @xmath159 at zero .
it follows that if @xmath177 then the local time of @xmath149 on @xmath2 vanishes .
in fact , if @xmath149 is nonexplosive with @xmath56 then , since @xmath159 is a continuous semimartingale with @xmath178_s = ds$ ] , it follows that @xmath179 for all @xmath8 , almost surely .
note that the brownian motion considered here can be replaced with a brownian motion with locally bounded and measurable drift .
in this subsection we consider an application of a different approach , based on the theory of markov processes .
although it is not always necessary to do so , we will assume that @xmath0 is compact .
we will also assume that @xmath2 is a closed embedded hypersurface and that @xmath4 is a brownian motion on @xmath0 starting at @xmath180 .
the convexity based argument of @xcite , which was applied to the one point case , can be adapted to our situation and implies that with respect to the invariant measure @xmath17 the local time @xmath181 corresponds in the sense of @xcite to the induced measure @xmath19 . by @xcite
this implies the following theorem , in which @xmath182 denotes the transition density function for brownian motion .
[ th : revuzimpthm ] suppose that @xmath0 is compact , that @xmath2 is a closed embedded hypersurface and that @xmath4 is a brownian motion on @xmath0 starting at @xmath180 .
then @xmath183= \int_0^t \int_n p_s^m(x , y)d\operatorname{vol}_n(y)\,ds\ ] ] for all @xmath184 . in a subsequent article
we will calculate , estimate and provide an asymptotic relation for the rate of change @xmath185 $ ] . by a standard change of variables
, it follows that the expected value of the occupation times appearing inside the limit on the right - hand side of converge to the right - hand side of as @xmath186 .
if one could justify exchanging this limit with the expectation , one would obtain a different proof of theorem [ th : revuzimpthm ] .
the following corollary follows directly from theorem [ th : revuzimpthm ] and basic ergodicity properties of brownian motion .
[ cor : largetime ] suppose that @xmath0 is compact , that @xmath2 is a closed embedded hypersurface and that @xmath149 is a brownian motion on @xmath0 .
then @xmath187 = \frac{\operatorname{vol}_n(n)}{\operatorname{vol}_m(m)}.\ ] ] suppose @xmath188 ( i.e. the unit circle with the standard metric ) and let @xmath4 be a brownian motion starting at @xmath189 . by formula
it follows that @xmath190 for @xmath191 , where @xmath158 is a standard one - dimensional brownian motion .
but @xmath192 and @xmath193 is antipodal to @xmath180 , which is a distance @xmath194 away from @xmath180 , so as @xmath195 is supported on @xmath196 we deduce that @xmath197 for @xmath191 . now
@xmath198 as @xmath199 so @xmath200 = \int_{\mathbb{s}^1 } \frac{r^2_x(y)}{2\pi } d \operatorname{vol}_{\mathbb{s}^1}(y ) = \int_{-\pi}^{\pi}\frac{v^2}{2\pi}dv = \frac{\pi^2}{3}.\ ] ] thus by equations and it follows that @xmath201\right)\ ] ] which implies for large times @xmath75 the approximation @xmath202 = \frac{t}{2 \pi}-\frac{\pi}{6}+o(1).\ ] ] corollary [ cor : largetime ] and the previous example concern the behaviour of @xmath203 $ ] for large times , in the compact case . in the next example
we fix the time , instead considering the effect of expanding size of the submanifold , using balls in euclidean space as the example . for @xmath204 denote by @xmath205 the boundary of the open ball in @xmath13 of radius @xmath206 centred at the origin . if @xmath149 is a brownian motion on @xmath13 starting at the origin then @xmath207 is a markov process .
it follows from general theory , as in @xcite , that formula ( [ eq : revuzimp ] ) holds with @xmath208 and @xmath209 and therefore @xmath210 = \frac{\gamma\left(\frac{{m}}{2}-1,\frac{r^2}{2t}\right)}{\gamma\left(\frac{{m}}{2}\right)}\ ] ] where @xmath211 is the upper incomplete gamma function . in this setting the process @xmath212 corresponds to the local time of
an bessel process of dimension @xmath1 , started at the origin , at the value @xmath206 .
in particular , for the case @xmath213 we obtain @xmath214 = \gamma\left(0,\frac{r^2}{2t}\right).\ ] ] by differentiating the exponential of the right - hand side of equation we can then deduce the curious relation @xmath215\right ) = \gamma\ ] ] where @xmath216 denotes the euler - mascheroni constant .
in this section we combine the geometric inequalities of the first section with the it - tanaka formula of the second section and deduce probabilistic estimates for the radial moments of brownian motion with respect to a submanifold .
it is fairly standard practice to deduce exponential integrability from a logarithmic - sobolev inequality . in this subsection ,
we show how this can be done in a restricted version of the situation in which we are otherwise interested . for this , denote by @xmath217 the heat semigroup on @xmath0 ( acting on some suitable space of functions ) .
[ th : logsobest ] suppose that @xmath0 is a complete and connected riemannian manifold of dimension @xmath1 and that @xmath2 is a closed embedded submanifold of @xmath0 of dimension @xmath3 .
assume that there exist constants @xmath218 such that @xmath219 with at least one of the three conditions * ( c1 ) * , * ( c2 ) * or * ( c3 ) * of theorem [ th : estimatefinal ] satisfied with @xmath220
. then @xmath221\ ] ] for all @xmath222 and @xmath223\ ] ] for all @xmath224 , where @xmath225 let @xmath4 a brownian motion starting at @xmath226 , let @xmath227 be an exhaustion of @xmath0 by regular domains and denote by @xmath228 the first exit time of @xmath4 from @xmath229 . using it s formula and formula , corollary [ cor : maincor ] and jensen s inequality , we see that @xmath230 \leq r_n^2 ( x ) + ( { m}-{n } ) t + ( { n}\lambda + ( { m}-1)c_1 ) \int_{0}^{t } \mathbb{e}\left [ r_n^2(x_{s\wedge \tau_{d_i}}(x ) \right]^{\frac{1}{2 } } ds\ ] ] for all @xmath8 .
since the term @xmath231 is increasing in @xmath75 , bihari s inequality @xcite , which is a nonlinear integral form of gronwall s inequality , implies @xmath230 \leq \left(\left(r^2_n ( x)+({m}-{n } ) t \right)^{\frac{1}{2 } } + ( { n}\lambda + ( { m}-1)c_1 ) t/2 \right)^2\ ] ] for all @xmath8 , from which it follows that @xmath232 for all @xmath191 , by fatou s lemma .
now , bakry and ledoux discovered ( see @xcite or @xcite ) that condition implies the heat kernel logarithmic sobolev inequality @xmath233 for all @xmath234 and @xmath235 . by a slight generalization of the classical argument of herbst ( see @xcite ) it follows that for lipschitz @xmath236 with @xmath237 and @xmath238 we have @xmath239\ ] ] for all @xmath191 while it was proved in @xcite ( see also @xcite ) that the log - sobolev inequality implies @xmath240\ ] ] for all @xmath224 .
since @xmath9 is lipschitz with @xmath241 , inequality follows from by the estimate while inequality is proved similarly , by applying jensen s inequality to and using the estimate . to obtain exponential integrability for the heat kernel under relaxed curvature assumption we will use a different approach , which is developed in the next subsection . while the estimates and are , roughly speaking , the best we have under the conditions of theorem [ th : logsobest ] , the estimate does not reduce to the correct expression in @xmath13 , as we will see , and our later estimates will take into account positive curvature whereas the estimates of this subsection do not .
thus the later estimates are preferable from the point of view of geometric comparison . on the other hand , the estimates given by ( * ? ? ?
* theorem 8.62 ) , which concern the case @xmath242 , suggest that the ` double exponential ' feature of the estimates and ( which is the inevitable result of using herbst s argument and bakry and ledoux s log - sobolev constant , as opposed to being a consequence of our moment estimates ) is not actually necessary .
suppose now that @xmath4 is a brownian motion on @xmath0 with locally bounded and measurable drift @xmath243 starting from @xmath226 , defined upto an explosion time @xmath6 , and that @xmath2 is a closed embedded submanifold of @xmath0 of dimension @xmath244 .
we will assume for the majority of this section that there exist constants @xmath10 and @xmath11 such that the inequality @xmath245 holds off the cut locus . unless otherwise stated , future references to the validity of this inequality will refer to it on the domain @xmath123 . if @xmath243 satisfies a linear growth condition in @xmath9 then geometric conditions under which such an inequality arises are given by theorem [ th : estimatefinal ] ( see corollary [ cor : maincor ] ) , the content of which the reader might like to briefly review . in particular , there are various situations in which one can choose @xmath246 .
alternatively , if @xmath2 is a point and the ricci curvature is bounded below by a constant @xmath138 then inequality holds with @xmath247 and @xmath248 , as stated by inequality .
of course , if @xmath2 is an affine linear subspace of @xmath13 then inequality holds as an _ equality _ with @xmath249 and @xmath250 .
note that inequality does not imply nonexplosion of @xmath4 .
this is clear by considering the products of stochastically incomplete manifolds with ones which are not .
we therefore use localization arguments to deal with the possibility of explosion .
let @xmath253 be an exhaustion of @xmath0 by regular domains and denote by @xmath228 the first exit time of @xmath4 from @xmath229 .
note that @xmath254 and that this sequence of stopping times announces the explosion time @xmath6 .
then , by formula , it follows that @xmath255 for all @xmath8 , almost surely .
since the domains @xmath229 are of compact closure the it integral in is a true martingale and it follows that @xmath256 = r_n^2(x ) & - 2\,\mathbb{e}\left [ \int_0^{t\wedge \tau_{d_i } } r_n(x_s(x))d\mathbb{l}^{\operatorname{cut}(n)}_s(x(x))\right ] \\ & + \int_{0}^{t } \mathbb{e}\left[\mathbb{1}_{\lbrace s < \tau_{d_i}\rbrace}\left ( \frac{1}{2}\triangle + b \right ) r_n^2 ( x_s(x ) ) \right ] ds \end{split}\ ] ] for all @xmath8 , where exchanging the order of integrals in the last term is easily justified by the use of the stopping time and the assumptions of the theorem . before applying gronwall s inequality we should be careful , since we are allowing the coefficient @xmath257 to be negative
for this , note that @xmath258 = \mathbb{e}\left[\mathbb{1}_{\lbrace t < \tau_{d_i}\rbrace } r_n^2(x_t(x))\right]+\mathbb{e}\left[\mathbb{1}_{\lbrace t \geq \tau_{d_i}\rbrace } r_n^2(x_{\tau_{d_i}}(x))\right]\ ] ] and that the two functions @xmath259 , \quad t\mapsto \mathbb{e}\left[\mathbb{1}_{\lbrace t \geq \tau_{d_i}\rbrace}r_n^2(x_{\tau_{d_i}}(x))\right]\ ] ] are nondecreasing . if we define a function @xmath260 by @xmath261\ ] ] then @xmath260 is differentiable , since the boundaries of the @xmath229 are smooth , and it follows from and that we have the differential inequality @xmath262 for all @xmath8 . now applying gronwall s inequality to yields @xmath263 \leq r_n^2 ( x)e^{\lambda t } + \nu\left(\frac{e^{\lambda t}-1}{\lambda}\right)\ ] ] for all @xmath184 ,
from which the result follows by the monotone convergence theorem .
if one wishes to include on the right - hand of inequality a term that is linear in @xmath9 , as in the estimate , or simply a continuous function of @xmath9 , as in the estimates , and , with suitable integrability properties , then one can do so and use a nonlinear version of gronwall s inequality , such as bihari s inequality , to obtain an estimate on the left - hand side of .
we will refer the object on the left - hand side of inequality as _ the second radial moment of @xmath4 with respect to @xmath2_. to find an inequality for _ the first radial moment of @xmath4 with respect to @xmath2 _ one can simply use jensen s inequality .
note that @xmath264 and this provides the sense in which theorem [ th : secradmomthm ] and similar statements should be interpreted if @xmath265 . that a ricci curvature lower bound implies stochastic completeness
was originally proved by yau in @xcite , as mentioned earlier .
this was extended by ichihara in @xcite and hsu in @xcite to allow the ricci curvature to grow in the negative direction in a certain way ( like , for example , a negative quadratic in the distance function ) .
thus the following theorem is well - known in the one point case ( in terms of which it can be proved ) .
using theorem [ th : secradmomthm ] our proof is short , so we may as well include it so as to keep the presentation of this article reasonably self - contained .
so suppose that @xmath204 , let @xmath266 and denote by @xmath267 the first exit time of @xmath4 from the tubular neighbourhood @xmath268 . by following the proof of theorem [ th : secradmomthm ] , with the stopping times
@xmath228 replaced by @xmath270 , one deduces @xmath271 for all @xmath184 .
this crude exit time estimate implies that @xmath4 is nonexplosive since the compactness of @xmath2 implies that the stopping times @xmath270 announce the explosion time @xmath6 .
recall that if @xmath149 is a real - valued gaussian random variable with mean @xmath272 and variance @xmath273 then for @xmath274 one has the formula @xmath275 = \left(2\sigma^2\right)^{p}{p } !
l^{-\frac{1}{2}}_{p}\left(-\frac{\mu^2}{2\sigma^2}\right)\ ] ] where @xmath276 are the _ laguerre polynomials _ , defined by the formula @xmath277 for @xmath278 and @xmath279 ( for the properties of laguerre polynomials used in this article , see @xcite ) .
in particular , if @xmath4 is a standard brownian motion on @xmath280 starting from @xmath281 then @xmath282= \left(2t\right)^{p}{p } ! l^{-\frac{1}{2}}_{p}\left(-\frac{\vert x\vert^2}{2t}\right)\ ] ] for all @xmath184 . with this in mind
we prove the following theorem , a special case of which is theorem [ th : secradmomthm ] , which will be used in the next section to obtain exponential estimates .
theorem [ th : secradmomthm ] was stated separately because it constitutes the base case in an induction argument .
[ th : evenradmomthm ] suppose that there exist constants @xmath283 and @xmath11 such that inequality holds and let @xmath274
. then @xmath284 \leq \left(2r(t)e^{\lambda t}\right)^{p}{p } !
l^{\frac{\nu}{2}-1}_{p}\left(-\frac{r_n^2(x)}{2r(t)}\right)\ ] ] for all @xmath8 , where @xmath252 . by the assumption of the theorem
we see that on @xmath123 and for @xmath274 we have @xmath285 and by formula we have @xmath286 for all @xmath8 , almost surely , where the stopping times @xmath228 are defined as in the proof of theorem [ th : secradmomthm ] .
it follows that if we define functions @xmath287 by @xmath288\ ] ] then , arguing as we did in the proof of theorem [ th : secradmomthm ] , there is the differential inequality @xmath289 for all @xmath8 . applying gronwall s inequality yields @xmath290 for all @xmath191 and @xmath274 .
the next step in the proof is to use induction on @xmath291 to show that @xmath292 for all @xmath191 and @xmath274 .
inequality covers the base case @xmath293 . if we hypothesise that the inequality holds for some @xmath294 then by inequality we have @xmath295 for all @xmath8 , where @xmath296 .
using @xmath297 and properties of the gamma function it is straightforward to deduce inequality from inequality which completes the inductive argument . since @xmath10 we can then apply the relation @xmath298 which can be proved using leibniz s formula , to see that @xmath299 and so by inequality it follows that @xmath300 for @xmath184 and @xmath301 .
the result follows from this by the monotone convergence theorem .
we will refer the object on the left - hand side of inequality as _ the @xmath302-th radial moment of @xmath4 with respect to @xmath2_. one can deduce an estimate for _ the @xmath303-th radial moment of @xmath4 with respect to @xmath2 _ by jensen s inequality .
we conclude this subsection with an example where the even radial moments can be calculated explicitly .
[ ex : hyp3ex ] denote by @xmath304 the @xmath305-dimensional hyperbolic space with constant sectional curvatures equal to @xmath72 and suppose that @xmath4 is a brownian motion on @xmath304 starting at @xmath180
. then the densities of @xmath4 are given by the deterministic heat kernel formula @xmath306\ ] ] for @xmath307 and @xmath235 where @xmath308 by a change of variables it follows that for each @xmath274 we have @xmath309 = ( 2t)^{p}\frac{\gamma\left(\frac{3}{2}+{p}\right)}{\gamma\left(\frac{3}{2}\right)}\mathstrut_1 f_1 \left(\frac{3}{2}+{p},\frac{3}{2},-\frac{\kappa t}{2}\right)\ ] ] for all @xmath184 , where @xmath310 is the confluent hypergeometric function of the first kind , so in particular @xmath311 = 3t-\kappa t^2\ ] ] for all @xmath184 .
this ties in with the fact , proved by liao and zheng in @xcite , that on general @xmath0 if @xmath4 is a brownian motion starting at @xmath226 and if @xmath312 is the first exit time of @xmath4 from the geodesic ball @xmath313 then @xmath314= { m}t-\frac{1}{6}\operatorname{scal}(x)t^2+o(t^2)\ ] ] as @xmath315 , where @xmath316 might depend upon @xmath317 and where @xmath318 denotes the scalar curvature at @xmath180 ( since on @xmath304 the scalar curvature is constant and equal to @xmath319 ) .
recall that @xmath323 now , since @xmath324 it follows that @xmath325 and that @xmath326 for all @xmath327 . for such @xmath328
there is the bound @xmath329 and since the largest binomial coefficient is ` the middle one ' it follows that @xmath330 and therefore we have the somewhat crude bound @xmath331 for @xmath327 . substituting these bounds into equation yields @xmath332 from which the lemma follows since @xmath333 .
[ th : inequality2 ] suppose that there exist constants @xmath334 and @xmath11 such that inequality holds . then @xmath335\leq 1 + \left ( 1+\mathbf{r}(t,\theta , x)^{-\frac{1}{2}}\right)\left(\mathstrut_1 f_1\left(\frac{\nu}{2},\frac{1}{2 } , \mathbf{r}(t,\theta , x)\right)-1\right)\ ] ] for all @xmath336 , where @xmath337 with @xmath252 and where @xmath310 is the confluent hypergeometric function of the first kind . with the stopping times @xmath228 as in the proof of theorem [ th : secradmomthm ] , for @xmath274 with @xmath291
even we see by inequality that @xmath338 \leq \left(2r(t)e^{\lambda t}\right)^{\frac{{p}}{2 } } \gamma\left(\frac{{p}}{2}+1\right ) l^{\frac{\nu}{2}-1}_{\frac{{p}}{2 } } \left(-\frac{r_n^2(x)}{2r(t)}\right)\ ] ] and so , by jensen s inequality , if @xmath291 is odd then @xmath338 \leq \left(2r(t)e^{\lambda t}\right)^{\frac{{p}}{2 } } \left(\gamma\left(\frac{{p}+1}{2}+1\right ) l^{\frac{\nu}{2}-1}_{\frac{{p}+1}{2 } } \left(-\frac{r_n^2(x)}{2r(t)}\right)\right)^{\frac{{p}}{{p}+1}}.\ ] ] it follows from this and lemma [ lem : techlemma ] , since @xmath339 , that @xmath340\\ & \leq 1 + \sum_{{p}=1}^{\infty } \frac{\left(2\theta^2r(t)e^{\lambda t}\right)^{p}}{(2{p})!}{p}!l^{\frac{\nu}{2}-1}_{p}\left(-\frac{r_n^2(x)}{2r(t)}\right)\\ & \quad\quad+ \sum_{{p}=1}^{\infty } \frac{\left(2\theta^2r(t)e^{\lambda t}\right)^{\frac{2{p}-1}{2}}}{(2{p}-1)!}\left({p}!l^{\frac{\nu}{2}-1}_{p}\left(-\frac{r_n^2(x)}{2r(t)}\right)\right)^{\frac{2{p}-1}{2{p}}}\\ & \leq 1 + \sum_{{p}=1}^{\infty } \frac{\left(2\theta^2r(t)e^{\lambda t}\right)^{p}}{(2{p } ) ! } \left(12\left(1+\frac{r_n^2\left(x\right)}{2r(t)}\right)\right)^{p}\frac{\gamma\left(\frac{\nu}{2}+{p}\right)}{\gamma\left(\frac{\nu}{2}\right)}\\ & \quad\quad+ \sum_{{p}=1}^{\infty } \frac{\left(2\theta^2r(t)e^{\lambda t}\right)^{\frac{2{p}-1}{2}}}{(2{p}-1)!}\left ( \left(12\left(1+\frac{r_n^2\left(x\right)}{2r(t)}\right)\right)^{p}\frac{\gamma\left(\frac{\nu}{2}+{p}\right)}{\gamma\left(\frac{\nu}{2}\right)}\right)^{\frac{2{p}-1}{2{p}}}\\ & = \quad\quad \sum_{{p}=0}^{\infty } \frac{\left(24\theta^2\left(r(t)+\frac{r_n^2\left(x\right)}{2}\right)e^{\lambda t}\right)^{p}}{(2{p } ) ! } \frac{\gamma\left(\frac{\nu}{2}+{p}\right)}{\gamma\left(\frac{\nu}{2}\right)}\\ & \quad\quad+ \sum_{{p}=1}^{\infty } \frac{\left(24\theta^2\left(r(t)+\frac{r_n^2\left(x\right)}{2}\right)e^{\lambda t}\right)^{\frac{2{p}-1}{2}}}{(2{p}-1)!}\left(\frac{\gamma\left(\frac{\nu}{2}+{p}\right)}{\gamma\left(\frac{\nu}{2}\right)}\right)^{\frac{2{p}-1}{2{p}}}. \end{split}\ ] ] now using @xmath341 , @xmath342 and @xmath343 , together with the relation @xmath344 which can be seen directly from the definition of @xmath310 as a generalized hypergeometric series , we deduce that @xmath345&\leq \sum_{{p}=0}^\infty \frac{(4\mathbf{r}(t,\theta , x))^p}{(2{p})!}\frac{\gamma\left(\frac{\nu}{2}+{p}\right)}{\gamma \left(\frac{\nu}{2}\right)}\\ & \quad+ \mathbf{r}(t,\theta , x)^{-\frac{1}{2}}\sum_{{p}=1}^\infty \frac{(4\mathbf{r}(t,\theta , x))^p}{(2{p})!}\frac{\gamma\left(\frac{\nu}{2}+{p}\right)}{\gamma \left(\frac{\nu}{2}\right)}\\ & = 1 + \left ( 1+\mathbf{r}(t,\theta , x)^{-\frac{1}{2}}\right)\left(\mathstrut_1 f_1\left(\frac{\nu}{2},\frac{1}{2 } , \mathbf{r}(t,\theta , x)\right)-1\right ) \end{split}\ ] ] and the theorem follows by monotone convergence .
the right - hand side of is a continuous function of @xmath75 , @xmath346 and @xmath180 and since the function @xmath310 satisfies @xmath347 and @xmath348 it follows that if @xmath349 then the right - hand side of inequality converges to the limit of the left - hand side as @xmath350 .
furthermore , for the values of @xmath351 considered in the theorem the right - hand side of grows exponentially with @xmath352 ( in particular @xmath353).the theorem shows that under the assumptions of the theorem there is no positive time at which the left - hand side of is infinite . for @xmath354
the laguerre polynomials also satisfy the identity @xmath355 which is proved in @xcite .
it follows from this identity and equation that for a real - valued gaussian random variable @xmath149 with mean @xmath272 and variance @xmath273 we have for @xmath356 that @xmath357 = \left(1-\theta \sigma^2\right)^{-\frac{1}{2}}\exp\left[\frac{\theta\vert \mu \vert ^2 } { 2(1-\theta \sigma^2)}\right]\ ] ] so long as @xmath358 ( and there is a well - known generalization of this formula for gaussian measures on hilbert spaces ) .
in particular , if @xmath4 is a standard brownian motion on @xmath280 starting from @xmath281 then for @xmath184 it follows that @xmath359 = \left(1-\theta t\right)^{-\frac{1}{2}}\exp\left[\frac{\theta \vert x \vert^2 } { 2(1-\theta t)}\right]\ ] ] so long as @xmath360 . with this in mind we state the following theorem .
[ th : expenbm ] suppose there exists constants @xmath283 and @xmath11 such that inequality holds . then @xmath361 \leq \left(1-\theta r(t ) e^{\lambda t}\right)^{-\frac{\nu}{2}}\exp\left[\frac{\theta r_n^2(x)e^{\lambda t}}{2(1-\theta r(t ) e^{\lambda t})}\right]\ ] ] for all @xmath362 such that @xmath363 , where @xmath252 . using inequality and equation we see that @xmath364 = & \text { } \sum_{{p}=0}^{\infty } \frac{\theta^{p}}{2^{p}{p } ! } f_{x , i,2{p}}(t ) \\ \leq&\text { } \sum_{{p}=0}^{\infty } \left(\theta r(t)e^{\lambda t}\right)^{p}l^{\frac{\nu}{2}-1}_{p}\left(-\frac{r_n^2(x)}{2r(t)}\right ) \\ = & \text { } \left(1-\theta r(t ) e^{\lambda t}\right)^{-\frac{\nu}{2}}\exp\left[\frac{\theta r_n^2(x)e^{\lambda t}}{2(1-\theta r(t ) e^{\lambda t})}\right ] \nonumber \end{split}\ ] ] where we safely switched the order of integration using the stopping time .
the result follows by the monotone convergence theorem .
theorem [ th : expenbm ] improves upon the estimate given by the second part of a similar theorem due to stroock ( see ( * ? ? ?
* theorem 5.40 ) ) , since the latter concerns only the one point case , does not take into account positive curvature or the possibility of drift and does not reduce to the correct expression in flat euclidean space . for a qualitative picture of the behaviour of the above estimates , fix @xmath226 and consider the case in which @xmath4 is a brownian motion starting at @xmath180 .
denote by @xmath138 the infimum of the ricci curvature of @xmath0 and assume that @xmath365 .
then inequality implies that the assumptions of theorems [ th : inequality2 ] and [ th : expenbm ] hold when @xmath366 with @xmath247 and @xmath367 . for these parameters
we plot in figure [ fig : figh ] the right - hand sides of the inequalities and as functions of time for the three cases @xmath368 with @xmath369 and @xmath370 .
note that if @xmath371 then the left - hand sides of these inequalities are bounded , by myer s theorem .
the sharpness of our estimates allows for the following comparison inequality , for which we note that if @xmath54 then @xmath2 is vacuously both totally geodesic and minimal , since in this case it would be an at most countable collection of isolated points .
( i ) : : @xmath18 , the sectional curvature of planes containing the radial direction is non - negative and @xmath2 is totally geodesic ; ( ii ) : : @xmath372 , the ricci curvature in the radial direction is non - negative and @xmath2 is minimal . if @xmath373 denotes a brownian motion on @xmath374 starting at @xmath375 with @xmath376 then @xmath377 \leq \mathbb{e } \left[e^{\frac{\theta}{2 } \|y_t(y)\|^2_{\mathbb{r}^{{m}-{n}}}}\right]\ ] ] for all @xmath362 such that @xmath378 . to find a comparison theorem which takes into account negative curvature seems harder .
we can , however , perform an explicit calculation for the following special case , which compares favourably with our best estimate .
suppose that @xmath4 is a brownian motion on @xmath304 starting at @xmath180 .
then , using formulae and , one can show that @xmath379 = \left(1-\theta t\right)^{-\frac{3}{2}}\exp\left[-\frac{\theta \kappa t^2}{2(1-\theta t)}\right]\ ] ] for all @xmath235 and @xmath356 such that @xmath360 .
note that the explosion time of the right - hand side of formula is independent of @xmath380 .
if @xmath4 is a brownian motion on @xmath381 starting at @xmath180 then it is easy to see that @xmath382 for all @xmath235 .
note that the right - hand side of the asymptotic relation does not depend on the dimension @xmath1 .
returning to the setting of example [ ex : hyp3ex ] for a final time , we find another situation where there is a relation of the type .
a heat kernel comparison argument would suggest that a relation of this type should hold in general for a brownian motion @xmath4 on @xmath0 but as an inequality , so long as the ricci curvature is bounded below by a constant .
indeed , it follows from ( * ? ? ?
* theorem 8.62 ) that in this case there is the asymptotic estimate @xmath384 } r_x(x_t(x ) ) \geq r\bigg\rbrace \leq -\frac{1}{2t}.\ ] ] for the general setting considered in this article , we have the following theorem .
the proof follows a standard argument .
in particular , for @xmath356 and @xmath204 it follows from markov s inequality and theorem [ th : expenbm ] that @xmath386 \\ & \leq \left(1-\theta r(t ) e^{\lambda t}\right)^{-\frac{\nu}{2 } } \exp \left[\frac{\theta r_n^2(x ) e^{\lambda t}}{2(1-\theta r(t ) e^{\lambda t})}-\frac{\theta r^2}{2}\right ] \end{split}\ ] ] so long as @xmath363 .
if @xmath235 then choosing @xmath387 shows that for any @xmath388 and @xmath204 we have the estimate @xmath389\ ] ] from which the theorem follows since @xmath390 can be chosen arbitrarily close to @xmath391 after taking the limit . while theorem [ th : concenthm ] is trivial if @xmath0 is compact , the concentration inequality is still valid in that setting .
in fact , for @xmath204 suppose that @xmath392 and @xmath393 are constants such that the inequality @xmath394 holds on the tubular neighbourhood @xmath268 ( such constants always exist if @xmath2 is compact , by corollary [ cor : corone ] ) .
assuming that @xmath4 is non - explosive ( which would be the case if @xmath2 is compact , by theorem [ th : nonexp ] ) then the methods of this chapter can also be used to estimate quantities involving the process @xmath4 stopped on the boundary of the tubular neighbourhood .
we will not include such calculations here , to avoid extensive repetition , but doing so actually yields the exit time estimate @xmath395 } r_n(x_s(x ) ) \geq r \bigg\rbrace\leq ( 1-\delta)^{-\frac{\nu}{2}}\exp\left[\frac{r_n^2(x)\delta}{2r(t)(1-\delta)}-\frac{\delta r^2}{2r(t)e^{\lambda t}}\right]\ ] ] for all @xmath235 and @xmath396 , which improves inequality for the @xmath393 case .
the following two propositions and their corollaries constitute simple applications of theorems [ th : inequality2 ] and [ th : expenbm ] and provide bounds on the operator norm of certain feynman - kac semigroups , when acting on suitable banach spaces of functions .
[ prop : fkestpre2 ] suppose there exists constants @xmath334 and @xmath393 such that inequality holds . then @xmath397\leq 1 + \left ( 1+\mathbf{r}(t,\theta t , x)^{-\frac{1}{2}}\right)\left(\mathstrut_1 f_1\left(\frac{\nu}{2},\frac{1}{2 } , \mathbf{r}(t,\theta t , x)\right)-1\right)\ ] ] for all @xmath362 , where @xmath398 is defined by .
using the stopping times @xmath228 to safely exchange the order of integrals , we see by jensen s inequality that @xmath399 \leq \frac{1}{t}\int_0^t \mathbb{e } \left[\mathbf{1}_{\{s < \tau_{d_i}\ } } e^{t\theta r_n(x_s(x))}\right]ds.\ ] ] the result follows from this by the monotone convergence theorem and theorem [ th : inequality2 ] , since the right - hand side of inequality is nondecreasing in @xmath75 ( which is evident by the right - hand side of inequality and the fact that @xmath352 is nondecreasing in @xmath75 ) .
[ cor : vest2 ] suppose there exists constants @xmath334 and @xmath393 such that inequality holds and that @xmath400 is a measurable function on @xmath0 such that @xmath401 for some constant @xmath402 . then @xmath403 \leq e^{ct}\left(1 + \left ( 1+\mathbf{r}(t , ct , x)^{-\frac{1}{2}}\right)\left(\mathstrut_1 f_1\left(\frac{\nu}{2},\frac{1}{2 } , \mathbf{r}(t , ct , x)\right)-1\right)\right)\ ] ] for all @xmath8 , where @xmath398 is defined by .
[ prop : fkestpre ] suppose there exists constants @xmath283 and @xmath404 such that inequality holds .
then @xmath405\leq \left(1-\theta tr(t ) e^{\lambda t}\right)^{-\frac{\nu}{2}}\exp\left[\frac{\theta r_n^2(x ) te^{\lambda t}}{2(1-\theta
tr(t ) e^{\lambda t})}\right]\ ] ] for all @xmath362 such that @xmath406 .
[ cor : vest ] suppose there exists constants @xmath283 and @xmath11 such that inequality holds and that @xmath400 is a measurable function on @xmath0 such that @xmath407 for some constant @xmath402
. then @xmath408 \leq \left(1-c tr(t ) e^{\lambda t}\right)^{-\frac{\nu}{2}}\exp\left[ct+\frac{c r_n^2(x ) te^{\lambda
t}}{2(1-c tr(t ) e^{\lambda t})}\right]\ ] ] for all @xmath8 such that @xmath409 .
applications of the results and methods presented in this article have been explored in @xcite and will feature in a subsequent article .
in particular , we will show how one can use the jacobian inequality given by theorem [ th : estimatefinal ] and moment estimates for a certain elementary bridge process , related to brownian motion , to deduce lower bounds and an asymptotic relation for the _ integral of the heat kernel over a submanifold_. this object appears naturally in the study of _ submanifold bridge processes _ , for which the lower bounds imply a gradient estimate sufficient to prove a semimartingale property .
this is a new area of study which could lead to future developments related to the geometries of path or loop spaces .
shigeki aida , takao masuda , and ichir shigekawa , _ logarithmic sobolev inequalities and exponential integrability _ , j. funct .
126 * ( 1994 ) , no . 1 , 83101 .
mrstripmrextract1305064 ( 95m:60111 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip
] s. aida and d. stroock , _ moment estimates derived from poincar and logarithmic sobolev inequalities _ , math .
* 1 * ( 1994 ) , no . 1 , 7586 . mrstripmrextract1258492 ( 95f:60086 )
! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] dominique bakry and michel ledoux , _ a logarithmic sobolev form of the li - yau parabolic inequality _ ,
rev . mat .
* 22 * ( 2006 ) , no . 2 , 683702 .
mrstripmrextract2294794 ( 2008m:58051 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ]
dennis barden and huiling le , _ some consequences of the nature of the distance function on the cut locus in a riemannian manifold _ , j. london math .
( 2 ) * 56 * ( 1997 ) , no . 2 , 369383 . mrstripmrextract1489143 ( 99b:53059 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] michael cranston , wilfrid s. kendall , and peter march , _ the radial part of brownian motion .
ii . its life and times on the cut locus _ , probab . theory related fields * 96 * ( 1993 ) , no . 3 , 353368 .
mrstripmrextract1231929 ( 94g:60154 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] bruce k. driver and yaozhong hu , _ on heat kernel logarithmic sobolev inequalities _ , stochastic analysis and applications ( powys , 1995 ) , world sci . publ . ,
river edge , nj , 1996 , pp .
mrstripmrextract1453132 ( 98h:58183 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] pat fitzsimmons , jim pitman , and marc yor , _ markovian bridges : construction , palm interpretation , and splicing _ , seminar on stochastic processes , 1992 ( seattle , wa , 1992 ) , progr .
33 , birkhuser boston , boston , ma , 1993 , pp .
mrstripmrextract1278079 ( 95i:60070 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ]
alfred gray , _ tubes _ , second ed . , progress in mathematics , vol .
221 , birkhuser verlag , basel , 2004 , with a preface by vicente miquel .
mrstripmrextract2024928 ( 2004j:53001 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ]
alfred gray , leon karp , and mark a. pinsky , _ the mean exit time from a tube in a riemannian manifold _
, probability theory and harmonic analysis ( cleveland , ohio , 1983 ) , monogr .
textbooks pure appl . math .
98 , dekker , new york , 1986 , pp .
mrstripmrextract830234 ( 88a:58208 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] ernst heintze and hermann karcher , _ a general comparison theorem with applications to volume estimates for submanifolds _ , ann .
sci . cole norm .
( 4 ) * 11 * ( 1978 ) , no . 4 , 451470 .
mrstripmrextract533065 ( 80i:53026 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] pei hsu , _ heat semigroup on a complete riemannian manifold _
, ann . probab .
* 17 * ( 1989 ) , no . 3 , 12481254 .
mrstripmrextract1009455 ( 90j:58158 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] yaozhong hu , _ exponential integrability of diffusion processes _ , advances in stochastic inequalities ( atlanta , ga , 1997 ) , contemp . math . , vol .
234 , amer .
soc . , providence , ri , 1999 , pp .
mrstripmrextract1694764 ( 2000i:60089 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] kanji ichihara , _ curvature , geodesics and the brownian motion on a riemannian manifold .
explosion properties _ , nagoya math . j.
* 87 * ( 1982 ) , 115125 .
mrstripmrextract676590 ( 84m:58166b ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] atsushi kasue , _ a laplacian comparison theorem and function theoretic properties of a complete riemannian manifold _ ,
japan . j. math .
( n.s . ) * 8 * ( 1982 ) , no . 2 , 309341 .
mrstripmrextract722530 ( 85h:53031 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] hui ling le and dennis barden , _
it correction terms for the radial parts of semimartingales on manifolds _ , probab .
theory related fields * 101 * ( 1995 ) , no . 1 , 133146 . mrstripmrextract1314177 ( 96a:58211 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip
] n. n. lebedev , _ special functions and their applications _ , dover publications , inc . ,
new york , 1972 , revised edition , translated from the russian and edited by richard a. silverman , unabridged and corrected republication .
mrstripmrextract0350075 ( 50 # 2568 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] michel ledoux , _ concentration of measure and logarithmic sobolev inequalities _ , sminaire de probabilits , xxxiii , lecture notes in math .
1709 , springer , berlin , 1999 , pp .
mrstripmrextract1767995 ( 2002j:60002 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip
] m. liao and w. a. zheng , _ radial part of brownian motion on a riemannian manifold _
, ann . probab .
* 23 * ( 1995 ) , no . 1 , 173177 . mrstripmrextract1330766 ( 96j:58182 )
! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] carlo mantegazza and andrea carlo mennucci , _ hamilton - jacobi equations and distance functions on riemannian manifolds _ , appl .
* 47 * ( 2003 ) , no . 1 , 125 .
mrstripmrextract1941909 ( 2003h:49049 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ]
d. revuz , _ mesures associes aux fonctionnelles additives de markov .
i _ , trans .
* 148 * ( 1970 ) , 501531 .
mrstripmrextract0279890 ( 43 # 5611 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] takashi sakai , _ riemannian geometry _ ,
translations of mathematical monographs , vol .
149 , american mathematical society , providence , ri , 1996 , translated from the 1992 japanese original by the author .
mrstripmrextract1390760 ( 97f:53001 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] daniel w. stroock , _ an introduction to the analysis of paths on a riemannian manifold _
, mathematical surveys and monographs , vol .
74 , american mathematical society , providence , ri , 2000 .
mrstripmrextract1715265 ( 2001m:60187 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip
] s. j. taylor and n. a. watson , _ a hausdorff measure classification of polar sets for the heat equation _ , math .
cambridge philos .
* 97 * ( 1985 ) , no . 2 , 325344 .
mrstripmrextract771826 ( 86m:35077 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] shing tung yau , _ some function - theoretic properties of complete riemannian manifold and their applications to geometry _ , indiana univ .
j. * 25 * ( 1976 ) , no . 7 , 659670 . mrstripmrextract0417452 ( 54 # 5502 ) !
http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] shing tung yau , _ on the heat kernel of a complete riemannian manifold _ , j. math .
pures appl .
( 9 ) * 57 * ( 1978 ) , no . 2 , 191201 .
mrstripmrextract505904 ( 81b:58041 ) ! http://www.ams.org/mathscinet-getitem?mr=\@mrstrip[mr-mrstrip ] | we present a study of the distance between a brownian motion and a submanifold of a complete riemannian manifold .
we include a variety of results , including an inequality for the laplacian of the distance function derived from a jacobian comparison theorem , a characterization of local time on a hypersurface which includes a formula for the mean local time , an exit time estimate for tubular neighbourhoods and a concentration inequality .
we derive the concentration inequality using moment estimates to obtain an exponential bound , which holds under fairly general assumptions and which is sufficiently sharp to imply a comparison theorem .
we provide numerous examples throughout .
further applications will feature in a subsequent article , where we see how the main results and methods presented here can be applied to certain study objects which appear naturally in the theory of _ submanifold bridge processes_. + * keywords : * brownian motion ; local time ; concentration ; submanifold ; tube ; distance * ams msc 2010 : * 58j65 ; 53b21 ; 60j55 |
the d detector collected a data sample corresponding to an integrated luminosity of approximately 100 pb@xmath6 during the 19921996 tevatron run at @xmath8 gev .
the detector consists of three major components : a non - magnetic tracking system including a transition radiation detector in the central region , a liquid - argon calorimeter , and a toroidal magnetic muon spectrometer .
a detailed description of the d detector can be found in ref .
the tevatron @xmath9 collider is ideal for studying high @xmath10 phenomena .
it is a natural place to search for new physics beyond the standard model . in this talk ,
the results of the searches for supersymmetry and dirac monopoles are presented .
supersymmetry @xcite is a generalization of space - time symmetry .
it introduces for every standard model particle a supersymmetric partner that differs in spin by @xmath11 .
@xmath12-parity conservation requires that supersymmetric particles be produced in pairs and that the lightest supersymmetric particle ( lsp ) be stable .
the minimal supersymmetric standard model ( mssm ) is the simplest supersymmetric model . in the mssm , the gaugino - higgsino sector ( excluding gluinos )
is described by four parameters : @xmath13 , @xmath14 , @xmath15 , and @xmath16 , where @xmath13 and @xmath14 are the @xmath17 and @xmath18 gaugino mass parameters , @xmath15 is the higgsino mass parameter , and @xmath16 is the ratio of the vacuum expectation values of the two higgs doublets .
the gaugino - higgsino mixing gives in four neutral mass eigenstates ( neutralinos @xmath19 ) and two charged mass eigenstates ( charginos @xmath20 ) whose masses and couplings are fixed by the above four parameters .
supersymmetric models @xcite predicting photon production have been proposed as possible explanations of a recent event @xcite reported by the cdf collaboration . within the mssm
, it has been shown that the radiative decay of @xmath21 dominates in some regions of parameter space @xcite .
assuming @xmath22 is the lsp , the production of @xmath23 , either directly or indirectly from decays of other supersymmetry particles , will result in events with two high transverse energy ( @xmath24 ) photons and large missing transverse momentum ( @xmath25 ) and/or in events ( @xmath26 ) with one high @xmath24 photon , multile jets and large @xmath25 .
moreover , the @xmath0 events are also expected in supersymmetric models with a light gravitino ( @xmath27 ) being the lsp . in this case
, @xmath22 ( assumed to be the lightest superpartner of a standard model particle ) is unstable and decays into a photon plus a gravitino ( @xmath28 ) .
we present searches for new physics in the channels @xmath29 and @xmath30 at the fermilab tevatron .
the @xmath1 events with fewer than two jets are not considered here , due to large backgrounds from @xmath31jets production .
we interpret our results of the @xmath0 analysis in terms of @xmath32 , @xmath33 , and @xmath34 production and of the @xmath2 analysis in terms of squark ( @xmath35 ) and gluino ( @xmath36 ) production in supersymmetric models with a dominant @xmath7 decay .
the interpretation of the results of the @xmath0 analysis in models with a light @xmath27 can be found in ref .
@xcite .
the trigger used requires one electromagnetic ( em ) cluster with transverse energy @xmath37 gev , one jet with @xmath38 gev , and @xmath25@xmath39 gev ( @xmath25@xmath40 gev for about 10% of the data taken early in the tevatron run ) .
the jets in the trigger include non - leading em clusters .
photons are identified via a two - step process : the selection of isolated em energy clusters and the rejection of electrons .
the em clusters are selected from calorimeter energy clusters by requiring ( i ) at least 95% of the energy to be deposited in the em section of the calorimeter , ( ii ) the transverse and longitudinal shower profiles to be consistent with those expected for an em shower , and ( iii ) the energy in an annular isolation cone with radius 0.2 to 0.4 around the cluster in @xmath41 space to be less than 10% of the cluster energy , where @xmath42 and @xmath43 are the pseudorapidity and azimuthal angle .
electrons are removed by rejecting em clusters which have either a reconstructed track or a large number of tracking chamber hits in a road between the calorimeter cluster and the event vertex .
@xmath25 is determined from the energy deposition in the calorimeter for @xmath44 . to be selected as @xmath0 candidates
, events are first required to have two identified photons , one with @xmath45 gev and the other with @xmath46 gev , each with pseudorapidity @xmath47 or @xmath48 , the regions with good photon identification .
we denote the 28 events passing these photon requirements as the @xmath3 sample .
we then require @xmath25@xmath49 gev with at least one reconstructed vertex in the event to ensure good measurement of @xmath25 .
no requirement on jets is made .
two events satisfy all requirements .
the data used in this analysis correspond to an integrated luminosity of @xmath50 pb@xmath6 .
the principal backgrounds are multijet , direct photon , @xmath51 , @xmath52 , @xmath53 , and @xmath54 events from standard model processes with misidentified photons and/or mismeasured @xmath25 .
the background due to @xmath25 mismeasurement is estimated using events with two em - like clusters which satisfy looser em cluster requirements than those discussed above , and for which at least one of the two fails the em shower profile consistency requirement ( ii ) above .
in addition , these events must pass the photon kinematic requirements . by normalizing the number of events with @xmath25 @xmath55 gev in the qcd sample to that in the @xmath3 sample ,
we obtain a background of @xmath56 events due to @xmath25 mismeasurement for @xmath25 @xmath49 gev .
other backgrounds are due to events with genuine @xmath25such as those from @xmath31`@xmath57 ' ( where ` @xmath57 ' can be a real or a fake photon ) , @xmath54 , and @xmath58 production .
these events would fake @xmath0 events if the electrons were misidentified as photons .
we estimate their contribution using a sample of @xmath59`@xmath57 ' events passing the kinematic requirements , including that on @xmath25 . taking into account the probability ( @xmath60 , determined from @xmath61 data ) that an electron is misidentified as a photon
, we estimate a background of @xmath62 events . adding the two background contributions together yields @xmath63 events . no excess of events is observed .
= 3.0 in = 3.0 in the null results are used to set upper cross section limits on the pair production of scalar electron ( @xmath64 ) , scalar neutrino ( @xmath65 ) , and @xmath23 with @xmath66 , @xmath67 , and @xmath68 .
@xmath69 events are generated using the isajet program @xcite and are processed through the detector and trigger simulation and the reconstruction program .
the @xmath64 mass is set to 100 gev / c@xmath70 while the @xmath65 , @xmath23 , and @xmath22 masses are varied between 70100 gev / c@xmath70 , 5090 gev / c@xmath70 , and 3080 gev / c@xmath70 respectively .
the efficiency is found to be insensitive to the @xmath64 and @xmath65 masses and is largely determined by the mass difference @xmath71 as shown in fig .
[ fig : ggmet](a ) . for a given value of @xmath71 ,
the efficiencies for the three processes are the same within the errors , independent of the values of @xmath72 , @xmath73 , and @xmath74 . the 95% confidence level ( c.l . )
upper limits on the cross section as a function of @xmath71 is shown in fig .
[ fig : ggmet](b ) .
the limit is about 300 fb for large values of @xmath71 .
these results improve over those we published earlier @xcite , mainly due to the improved photon identification .
however , we note that the expected cross sections for @xmath75 processes are small , well below the experimental upper limits . for example
, the theoretical cross section for @xmath76 production is about 30 fb for @xmath77 gev / c@xmath70 .
to be selected as @xmath2 candidates , events are first required to have at least one identified photon with @xmath79 gev and pseudorapidity @xmath80 or @xmath81 , and two or more jets with @xmath82 gev and @xmath83 .
we denote events passing these requirements as the @xmath84 sample .
the @xmath25 distribution of the @xmath84 events is shown in fig .
[ fig : gmet_fig12](a ) . after requiring @xmath85 gev
, 378 events remain in the sample .
the data used in this analysis correspond to an integrated luminosity of @xmath86 pb@xmath6 .
the principal backgrounds are multijet , direct photon , @xmath52 , and @xmath87 events from standard model processes with jets or electrons misidentified as photons and/or mismeasured @xmath25 .
following the same procedure as in the @xmath0 analysis , the background due to @xmath25 mismeasurement is estimated to be @xmath88 events .
@xmath89 events with @xmath90 would fake @xmath2 events if the electrons were misidentified as photons .
this contribution is estimated to be @xmath91 using a sample of @xmath92 events passing all the kinematic requirements with the electron satisfying those imposed on the photon .
another background due to @xmath93 and @xmath94 production is found to be negligible .
the total background ( @xmath95 ) agrees well with the number of observed events .
the @xmath96 ( defined as the scalar sum of the @xmath24 of all jets with @xmath97 gev and @xmath98 ) distribution is shown in fig .
[ fig : gmet_fig12](b ) for both @xmath2 and background samples .
the background distribution reproduces the observed @xmath2 distribution well .
= 3.0 in = 3.0 in with the basic selection discussed above , the @xmath2 sample is dominated by the backgrounds . to optimize the selection criteria for supersymmetry
, we simulate squark and gluino production ( either in pair or in association with charginos / neutralinos ) using the pythia program @xcite .
the mssm parameter values are set to @xmath99 gev , @xmath100 , and @xmath101 gev .
this set of parameter values gives @xmath102 gev / c@xmath70 , @xmath103 gev / c@xmath70 , and @xmath68 .
sleptons ( @xmath104 ) and stop ( @xmath105 ) are assumed to be heavy .
events with @xmath23 in the final state are selected and are run through a detector simulation program , a trigger simulator , and the same trigger requirements , reconstruction , and analysis as the data .
the @xmath25 and @xmath96 distributions for @xmath106 gev / c@xmath70 events after the basic selection are shown in figs .
[ fig : gmet_fig12 ] .
the distributions expected from supersymmetry are considerably harder than those of the background events .
to increase sensitivity to supersymmetry , we introduce an @xmath96 cut and maximize the @xmath107 ratio by varying @xmath25 and @xmath96 requirements . here
@xmath108 is the efficiency for @xmath109 gev / c@xmath70 mc events and @xmath110 is the error on the estimated number of background events .
the optimized cuts are found to be @xmath111 gev and @xmath112 gev . with these additional cuts , five @xmath2 events
are observed while @xmath113 events are expected from the background processes .
the efficiency for supersymmetry signal varies from a few percent for low mass @xmath114 to approximately 25% for high mass @xmath114 .
mc studies show that the efficiency varies by 4% for different choices of @xmath13 , @xmath14 , @xmath16 , and @xmath15 which are consistent with @xmath68 and @xmath115 gev / c@xmath70 suggested in ref .
experimentally , the mass requirement is needed to ensure that photons from @xmath23 decays are reasonably energetic and are detected with good efficiency .
the variation is assigned as a systematic error in the efficiency .
the total fractional systematic error on the efficiency is 9%
. with five events observed and @xmath113 events expected from backgrounds , we observe no excess of events .
we compute 95% c.l .
upper limits on @xmath116 using a bayesian approach with a flat prior distribution for the signal cross section . the resulting upper limit as a function of squark / gluino mass
is displayed in fig .
[ fig : gmet_fig3 ] , for the case where @xmath117 , along with the theoretical cross sections , as calculated using the cteq3l parton distribution function ( p.d.f . ) @xcite .
the hatched band represents the range of the theoretical cross sections obtained by varying the supersymmetry parameter values with the constraints @xmath68 and @xmath118 .
the intersection of our limit curve with the lower edge of the theory band is at @xmath119 pb , leading to a lower limit for equal mass squarks and gluinos of 311 gev / c@xmath70 .
= 3.0 in the effect of light sleptons on the squark and gluino decays is studied by varying the slepton mass ( @xmath120 ) in mc .
the fraction of events containing at least one @xmath23 increases as the @xmath120 is decreased .
when @xmath120 is varied from 500 gev / c@xmath70 to 80 gev / c@xmath70 , the fraction increases about 25% for @xmath121 gev / c@xmath70 with equal gluino mass .
sleptons with mass below 80 gev / c@xmath70 have already been excluded @xcite .
the increased @xmath23 production increases the mass limit by approximately 10 gev / c@xmath70 . a light stop ( @xmath122 ) will also modify the squark and gluino decays and therefore affect the @xmath23 production .
we investigate this effect by setting @xmath123 gev / c@xmath70 which approximately corresponds to the current @xmath122 lower mass limit @xcite . a 15% reduction in @xmath23 production cross section
is observed .
this reduction lowers the limit for equal mass squarks and gluinos by about 6 gev / c@xmath70 . following the procedure above
, we obtain a low mass limit for gluinos ( squarks ) to be 233 gev / c@xmath70 ( 219 gev / c@xmath70 ) when squarks ( gluinos ) are heavy .
again , these limits vary approximately 10 gev / c@xmath70 if @xmath122 and/or sleptons are light .
one of the open questions of particle physics is the existence of dirac monopoles @xcite , hypothetical carriers of the magnetic charge proposed by p.m. dirac to symmetrize maxwell equations and explain the quantization of electric charge .
if such magnetic monopoles exist , then the elementary magnetic and electric charges ( @xmath124 and @xmath59 ) must be quantized according to the following formula : @xmath125 where @xmath126 is an unknown integer .
= 2.0 in dirac monopoles are expected to couple to photons with a coupling constant @xmath127 which is at least three orders of magnitude larger than the corresponding photon coupling to the electric charge ( @xmath128 ) .
therefore such monopoles could give rise to photon - photon rescattering via the box diagram shown in fig .
[ fig : mpole_diagram ] @xcite .
the contribution of this diagram for pointlike monopoles to diphoton production at hadron colliders was recently calculated @xcite and shown to be significant even for monopole masses comparable to the collider beam energy . since the virtuality ( @xmath129 ) of most incoming photons in the process of fig .
[ fig : mpole_diagram ] is small @xcite , the interacting partons scatter at very small angles and therefore escape the detector through the beam pipe .
thus , a signature for monopoles at hadron colliders is the production of a pair of isolated photons with high transverse energies .
this process gives a unique opportunity to find evidence for dirac monopoles or to set limits on the monopole mass .
previous monopole searches can be found in ref .
@xcite .
despite numerous studies , qed with pointlike monopoles is still not a complete theory . for example , it is not clear whether such a theory can be constructed to be renormalizable to all orders @xcite
. also , arguments exist ( see , e.g. @xcite ) that dirac monopoles must occupy a spatial volume of radius @xmath130 , where @xmath131 is the monopole mass , to accommodate the self energy implied by the large coupling .
in such a theory , it is possible that hard interactions of a monopole with photons would be weakened substantially by the effects of a monopole form factor .
the data used in this analysis represent an integrated luminosity of @xmath132 pb@xmath6 using a trigger which required the presence of an electromagnetic object with transverse energy @xmath24 above 40 gev .
this trigger did not require the presence of an inelastic collision , and therefore can be used to select low @xmath129 events typical of the process in fig .
[ fig : mpole_diagram ] .
the following offline selection criteria are : ( i ) at least two photons with @xmath133 gev and pseudorapidity @xmath134 ; ( ii ) missing transverse energy in the event @xmath25@xmath135 gev ; and ( iii ) no jets with @xmath136 gev and @xmath137 .
the jet veto requirement is used to select the low @xmath129 process in fig .
[ fig : mpole_diagram ] .
the trigger is @xmath138% efficient for this off - line selection .
photons are selected from the identified em clusters by requiring no tracks pointing toward the cluster from any of the event vertices .
the overall efficiency for photon identification is @xmath139 per photon .
the total efficiency for diphoton is @xmath140 .
this includes the efficiency of the @xmath25 veto @xmath141% as well as the identification efficiency for a pair of photons .
the above selection criteria define our base sample which contains 90 candidate events .
the main backgrounds to photon scattering through a monopole loop are due to : ( i ) diagrams similar to fig .
[ fig : mpole_diagram ] with other particles in the loop ; ( ii ) qcd production of dijets ( @xmath142 ) and direct photons ( @xmath143 ) ( with jets misidentified as photons due to fragmentation into a leading @xmath144 or @xmath42 decaying into a pair of spatially close photons , reconstructed as one em cluster ) , or direct diphotons ( @xmath3 ) ; and ( iii ) drell - yan dielectron production with electrons misidentified as photons due to tracking inefficiency .
background ( i ) is dominated by a virtual @xmath145-loop and has been shown to be negligible @xcite .
the other two background contributions are estimated from the data .
the qcd background is determined using the @xmath143 event sample collected with a single photon trigger , with the jet passing the same fiducial and kinematic cuts as the photon . by applying a jet - faking probability , we find the qcd background to be @xmath146 events .
direct photon and diphoton backgrounds are also included in this estimate . the drell - yan background is calculated from a sample of dielectron events passing the same fiducial and kinematic cuts as the signal sample .
multijet contamination of this sample is negligible .
the probability for a dielectron pair to be misidentified as a diphoton pair is found to be @xmath147 by comparing the number of events in the @xmath148 peak in the @xmath149 and @xmath3 samples passing loose kinematic cuts .
the drell - yan background in the base sample is @xmath150 events .
the overall background in the base sample is @xmath151 ( syst ) events , in good agreement with the 90 observed candidates . to optimize the sensitivity of this search to the monopole loop contribution
we apply a cut on the scalar sum of the transverse energies of all the photons in the event : @xmath152 .
we vary the @xmath153 cut threshold ( @xmath154 ) in 10 gev steps to achieve an expected background of 0.4 events .
such an optimization is based on the fact that for this expected background one has a 67% probability of observing no candidate events in the data in the absence of a signal . in such a case
@xcite , the limits on the signal do not depend on the exact background value or its uncertainties .
the agreement between the observed number of events and the predicted background as a function of @xmath155 is illustrated in fig .
[ fig : mpole_bck ] .
the @xmath156 gev cut corresponds to a background of @xmath157 events .
we set an upper limit for the production cross section of two or more photons with @xmath158 gev and @xmath159 : @xmath160 at the 95% c.l .
= 2.0 in since the data are consistent with the background hypothesis , we can set limits on the production of pointlike dirac monopoles .
we calculate the acceptance for the monopole signal using a fast mc program that generates diphoton events from a monopole loop according to the calculated differential cross section @xmath161 @xcite with a subsequent parametric simulation of the d detector . the overall acceptance for the monopole signal
is found to be @xmath162 .
the acceptance does not depend on the monopole mass for masses above the typical photon energy ( @xmath163 gev ) @xcite .
the total cross section for heavy monopole production at the tevatron is given by @xcite : @xmath164}\right)^8~{\rm fb,}\ ] ] where @xmath165 is a spin dependent factor @xcite : p = 0.085 , 1.39 , and 159 for monopole spin of 0 , 1/2 , and 1 , respectively .
the estimated error on this cross section due to choice of p.d.f . and to higher order qed effects is 30% @xcite .
additional uncertainties are associated with the @xmath166 subprocess in fig .
[ fig : mpole_diagram ] and with unitarity considerations .
the coupling constant @xmath167 is replaced with an effective coupling @xcite obtained by multiplying @xmath167 by a factor @xmath168 , where @xmath169 is the photon energy , typically 300 gev at the tevatron .
both unitarity and perturbation theory assumptions are satisfied when this factor is @xmath170 1 @xcite .
= 3.0 in comparing the lower bound of the theoretical cross section corrected for acceptance with the cross section limit set by this measurement , we obtain the following lower limits on the pointlike dirac monopole mass ( see fig .
[ fig : mpole_limits ] ) : @xmath171 we note that the effective coupling exceeds 1 and unitarity is violated close to the experimental bound . for values @xmath172 , the cross section will grow more slowly , approaching the usual @xmath173 behavior of a qed process @xcite which satisfies unitarity .
also , for lower monopole masses the effective parameter of the perturbation theory used in the calculations @xcite becomes too large , and therefore one would expect a non - negligible contribution of the higher order diagrams with four , six , etc .
photons in the final state .
the latter effect is , however , largely compensated by our analysis cut on the sum of the photon transverse energies ; if part of the signal cross section is due to the higher order diagrams , the above limits are unaffected .
in summary , we have performed searches for supersymmetry by searching for @xmath0 and @xmath2 events and for heavy pointlike dirac monopoles by searching for pairs of photons with high transverse energies .
our data agree with the expected background production . within the framework of the mssm with the choices of the parameter values consistent with @xmath68 and @xmath115 gev / c@xmath70 , we obtain a 95% c.l .
lower mass limit of 311 gev / c@xmath70 for equal mass squarks and gluinos .
using theoretical calculations @xcite we set 95% c.l .
lower limits on the dirac monopole mass for minimum magnetic charge @xmath174 in the range 610 to 1580 gev , depending on the monopole spin .
s. dimopoulos , s. thomas , and j.d .
wells , phys .
d * 54 * , 3283 ( 1996 ) ; s. dimopoulos
lett . * 76 * , 3494 ( 1996 ) ; k.s . babu , c. kolda , and f. wilczek , phys .
rev . lett . *
77 * , 3070 ( 1996 ) ; j.l .
lopez , d.v .
nanopoulos , and a. zichichi , phys .
lett . * 77 * , 5168 ( 1996 ) ; s. ambrosanio _ et al .
d * 54 * , 5395 ( 1996 ) ; h. baer _ et al .
d * 55 * , 4463 ( 1997 ) ; j. ellis , j.l .
lopez , and d.v .
nanopoulos , phys .
b * 394 * , 354 ( 1997 ) .
w. heisenberg and h. euler , z. phys .
* 38 * , 714 ( 1936 ) ; v. constantini , b. de tollis , and g. pistoni , nuovo cim .
* 2a * , 733 ( 1971 ) ; m. baillagreon , g. belanger , and f. boudjema , phys . rev .
d * 51 * , 4712 ( 1995 ) ; m. baillagreon , f. boudjema , e. chopin , and v. lafage , z. phys .
c * 67 * , 431 ( 1996 ) . | we have searched for diphoton events ( @xmath0 ) with large missing transverse momentum , @xmath1 events ( @xmath2 ) with two or more jets , and diphoton events ( @xmath3 ) with high transverse energies in @xmath4 collisions at @xmath5 tev using approximately 100 pb@xmath6 of data collected with the d detector at the fermilab tevatron in 19921996 . no excess of events beyond the expected backgrounds is observed .
the null results are interpreted in supersymmetric models with a dominant @xmath7 decay and in terms of dirac pointlike monopole production .
* new new - phenomena results from d * jianming qian + + + |
the nature and occupancy of the single - particle orbits for a nucleus can be determine from the spectroscopic factors ( sf ) . experimentally the sf can be measured with single - particle transfer reactions .
these reactions are of two types , first one is stripping in which one nucleon is stripped from the incoming projectile , while the second one is pick - up reaction in which one nucleon is picked up by the projectile . the examples of neuton transfer pick - up recations are like @xmath6 , @xmath7 and ( @xmath3he,@xmath8 ) , while stripping reactions are like @xmath9 , @xmath10 and ( @xmath8 , @xmath3he ) @xcite .
the sf is defined by a matrix element between initial state and final state corresponding to entrance channel and exit channels , respectively .
it is possible to describe the capture or emission of single - nucleons in stellar burning processes by calculating the nuclear matrix elements for a single - nucleon spectroscopic factors in the nuclear structure calculations .
the study of sf in different region of nuclear chart are reported in refs .
survey of excited state neutron spectroscopic factors for @xmath11 nuclei are reported by tsang _
et al _ @xcite .
in this work they extracted 565 neutron spectroscopic factors for @xmath12 and @xmath13 shell nuclei by analyzing @xmath9 angular distributions , they compared the experimental results with shell - model . to study different excited states and their spectroscopic factors for @xmath4 mg , earlier many experimental results were reported in refs .
recently , the spectroscopic factors for @xmath4 mg is reported in ref .
@xcite using @xmath2al(d,@xmath3he)@xmath4 mg reaction .
the structure of @xmath2al using different reaction channels were reported in refs .
the experimental results for sf of 14 excited states for @xmath2al using @xmath2al(d,@xmath5)@xmath4al reaction is reported in ref .
the study of @xmath4al and @xmath4 mg are important for astrophysics point of view .
the massive stars throughout the galaxy dominate in the production of @xmath4al @xcite , and it decays by @xmath14 to @xmath4 mg . in the present work our aim is to study recently available experimental spectroscopic factors data for @xmath2al(d,@xmath3he)@xmath4 mg @xcite and @xmath2al(d,@xmath5)@xmath4al @xcite in the frame work of nuclear shell - model .
we performed the shell - model calculation for spectroscopic factor using the two most suitable effective interactions usda and usdb for sd shell nuclei using shell - model code nushellx @xcite .
the hamiltonians , usda and usdb are based on a renormalized @xmath15 matrix by fitting two - body matrix elements with experimental data for binding energies and excitation energies for the @xmath12 shell nuclei @xcite .
the usda and usdb interactions are fitted by varying 30 and 56 linear combinations of one- and two - body matrix elements , respectively .
the rms deviations of 170 kev ( usda ) and 130 kev ( usdb ) were obtained between experimental and theoretical energies . in the present work before calculating the sf , first we examine the wavefunctions of concerned nuclei .
the comparison between the calculated and experimental levels for @xmath4 mg and @xmath4al reported in figs . 1 - 2 . the results of usdb interaction is much better than usda interaction .
in the case of @xmath4al we calculated only the g.s.(@xmath16 ) . with the comparison of theoretical and experimental spectroscopic factors it is possible to check the predictive power of effective interaction used in the shell - model calculations .
we can define the spectroscopic factor in terms of the reduced matrix elements of @xmath17 by : @xmath18 where the @xmath19 factor is by convention associated with heavier mass @xmath20 . here , @xmath21 indices distinguish the various basis states with the same @xmath22 value .
( color online ) comparison between calculated and experimental @xcite data for @xmath4mg.,width=415,height=377 ] ( color online ) comparison between calculated and experimental [ nndc ] data for @xmath4al.,width=415,height=377 ] .extracted values of @xmath23 for different excited states of @xmath4 mg from the reaction @xmath2al(d,@xmath3he ) at 25 mev .
we have taken experimental data from ref .
here @xmath24 , @xmath25 are @xmath26 and @xmath27 orbitals , respectively . [ cols=">,>,>,>,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] [ table2 ] experimentally the states of @xmath4 mg @xcite were studied by assuming pick - up from @xmath27 orbital only , and also few states by assuming configuration mixing of two lower orbital of @xmath12 shell - @xmath27 and @xmath26 single particle orbitals . in the table
i , we compared the experimental @xmath23 values with shell - model results for usda and usdb effective interactions .
as in the experiment the @xmath23 values were very large for @xmath25 ( @xmath27 ) transfer as compared with @xmath24 ( @xmath26 ) transfer for the first three states .
the shell - model results are larger for first two states , thus assigning larger single particle charateristics to these states . in the present work we have also predicted @xmath23 values for states up to @xmath28 6 mev ,
although experimental data is not available . in the fig
3 we have also shown variation of @xmath23 for experiment and theory . in this
we have also plotted @xmath29 values for theory and experiment .
the states of @xmath4al @xcite were experimentally studied by assuming pick - up from @xmath27 and @xmath26 single particle orbitals , while @xmath30 state at 3507 kev by pick - up from @xmath31 orbital . in the table
ii , we compared experimental @xmath23 values with shell - model results for usda and usdb effective interactions .
the experimentally observed values for spectroscopic factors up to @xmath28 4.7 mev reported in ref .
@xcite . in the present work we interpreted these measured sfs in term of shell - model calculations .
the experimental @xmath23 values for @xmath27 states given in the table ii .
the sf results with shell - model is slightly higher than experimental value for @xmath32 state .
for some states the sfs are very small this is because in these cases the wavefunctions are very fragmented .
there are two states @xmath33 at 1759 kev and @xmath34 at 4705 kev , with very small calculated spectroscopic factors .
this is because large cancellations of contributions from different components of the wave functions . in the fig .
4 we plotted @xmath23 for experiment and theory . in this
we have also plotted @xmath29 values for theory and experiment . the theory and experiment showing same trends .
motivated with recent experimental results of spectroscopic factors for @xmath2al(d,@xmath3he)@xmath4 mg and @xmath2al(d,@xmath5)@xmath4al reactions . in the present work we performed shell - model calculations for spectroscopic factors with new hamiltonians for @xmath12 shell usda and usdb .
the predictions from shell - model are found to be in a good agreement with the experimental data .
the sfs provide an independent test for the usda and usdb interactions used in the present shell - model calculations .
the present spectroscopic factor results for @xmath4 mg and @xmath4al are important for nucleosynthesis .
brussaard and p.w.m .
glaudemans , shell - model applications in nuclear spectroscopy"- ( north - holland publishing company , 1977 ) .
m.h . macfarlane and j.b .
french , rev .
* 32 * , 567 ( 1960 ) .
schiffer @xmath35 , phys .
lett . * 100 * , 112501 ( 2008 ) .
schiffer @xmath35 , phys .
lett . * 108 * , 022501 ( 2012 ) .
brown and w.a .
richter , phys .
c * 74 * , 034315 ( 2006 ) .
tsang , j. lee , and w.g .
lynch , phys .
lett . * 95 * , 222501 ( 2005 ) .
morfouace @xmath35 , phys .
b * 751 * , 306 ( 2015 ) .
j. lee , m.b .
tsang , and w.g .
lynch , phys .
c * 75 * , 064320 ( 2007 ) .
tsang @xmath35 , phys .
* 102 * , 062501 ( 2009 ) .
j. vernotte @xmath35 , phys .
c * 48 * , 205 ( 1993 ) . b. h. wildenthal @xmath36 , phys . rev . * 175 * , 1431 ( 1968 ) .
g. j. wagner @xmath35 , nucl .
a * 125 * , 80(1969 ) .
m. arditi @xmath35 , nucl . phys .
a * 165 * , 129(1969 ) .
j. j. kraushaar @xmath35 , phys .
c * 34 * , 1530 ( 1986 ) .
m. yassue @xmath35 , phys .
c * 42 * , 1279 ( 1990 ) .
m. burlein @xmath35 , phys .
c * 29 * , 2013 ( 1984 ) . v. srivastava @xmath37 , phys .
c * 93 * , 044601 ( 2016 ) .
k. a. chipps @xmath35 , phys .
c * 86 * , 014329 ( 2012 ) .
n. takahashi @xmath35 , phys .
c * 23 * , 1305 ( 1981 ) .
j. kroon et @xmath35 , nucl .
a * 204 * , 609 ( 1973 ) .
d. l. show @xmath35 , nucl .
a * 263 * , 293 ( 1977 ) .
r. r. betts @xmath35 , phys .
c * 8 * , 670 ( 1972 ) .
j. nurzynski @xmath35 , nucl .
a * 107 * , 581 ( 1968 ) .
n. a. vlasov , s. p.
kalinin , a. a. ogloblin , and v. i. chuev , sov .
jetp * 37 * , 844 ( 1960 ) .
r. diehl @xmath35 , nature ( london ) * 439 * , 45 ( 2006 ) .
v. srivastava @xmath35 , phys .
c * 91 * , 054611 ( 2015 ) .
b. a. brown , w. d. m. rae , e. mcdonald , and m. horoi , nushellx@msu .
brown and w.a .
richter , phys .
c * 74 * , 0343151 ( 2006 ) .
brown and w.a .
richter , phys .
c * 80 * , 034301 ( 2009 ) . | in the present work we calculated spectroscopic factor strengths for recently measured experimental data for @xmath0 and @xmath1 pick - up reactions @xmath2al(d,@xmath3he)@xmath4 mg [ phys .
rev .
c * 93 * , 044601 ( 2016 ) ] and @xmath2al ( d,@xmath5)@xmath4al [ phys . rev .
c * 91 * , 054611 ( 2015 ) ] within the framework of the shell - model .
we performed calculations with usda and usdb effective interactions , results are in a good agreement with the experimental data . |
in this section , we describe our hubbard model hamiltonian . we evaluate each individual term of eq .
( 1 ) in the main text using the 7d basis set defined by @xmath95 where the notation @xmath96 ( or @xmath97 ) indicates that both electrons are in the same dot , labelled @xmath98 , and @xmath99 indicates an empty dot .
the creation - annihilation operator combinations , @xmath100 , are then readily evaluated , as are the particle number operators @xmath101 , for dots @xmath102 , and spins @xmath103 .
we then obtain the following expressions for the individual terms in the hubbard hamiltonian , eq .
( 1 ) in the main text .
the tunnel coupling term is given by @xmath104 the onsite coulomb repulsion term describes the double - occupation energy cost for a single dot .
it is given by @xmath105 the detuning energies are given by @xmath106 since the basis states all belong to the same @xmath107 spin manifold , they all have the same zeeman energy , which we ignore here .
the local overhauser field energies due to nuclear fluctuations are given by @xmath108 where we define @xmath109 and @xmath110 to be the differences in local magnetic fields in the @xmath111 direction .
as explained in the main text , we only consider longitudinal ( @xmath111 ) components of the overhauser fields , defined by @xmath16 , as consistent with @xcite .
equation ( 3 ) of the main text then represents a set of 49 real coupled differential equations .
we solve these equations numerically and check that the trace - preserving condition is satisfied for the final density matrix , when the calculation is complete .
in this section , we estimate the effective exchange interactions @xmath112 that generate rotations .
we can use the results to provide initial estimates for the evolution periods for gate operations ; we use these estimates to optimize the gates , as discussed in the main text .
we now reduce the full @xmath69 hamiltonian , @xmath113 , to an effective @xmath80 hamiltonian for the logical qubit states .
we consider the ideal case with no nuclear fields , so @xmath114 and there is no coupling between the qubit states @xmath54 and the leakage state @xmath30 .
a schrieffer - wolff transformation @xcite to order @xmath115 in the small parameter @xmath116 yields the well - known heisenberg hamiltonian @xmath117 for the @xmath118 subspace of @xmath39 charge states . here ,
@xmath119 $ ] . in the absence of any coupling to the leakage state
, we can immediately project @xmath120 onto the @xmath80 logical qubit subspace , yielding @xmath121 where @xmath122 and @xmath123 are pauli matrices . here
, we may drop the constant term , giving eq .
( 2 ) in the main text . we can also obtain @xmath120 by directly performing a schrieffer - wolff transformation of @xmath124 onto the @xmath80 subspace , yielding @xmath125 where the denominator is given by @xmath126 by comparing eqs .
( [ eq : heff0 ] ) and ( [ eq : heff ] ) , we can identify the individual exchange interactions : @xmath127 diagonalizing eq .
( [ eq : heff ] ) , we obtain the energy splitting @xmath128 since @xmath64 is an even function in the variables @xmath24 and @xmath25 , we immediately find that @xmath129 when @xmath130 , establishing this setting as a detuning sweet spot . at the sweet spot ,
we find that @xmath131 corresponding to a @xmath1-rotation . indeed
, we see that @xmath1-rotations are achieved when either @xmath61 or @xmath62 . from eq .
( [ eq : heff ] ) , we see that rotations around the axis @xmath90 , used in the three - step @xmath87 protocol described in the main text , are defined by the line @xmath132 which correctly predicts the line of highest fidelities in fig .
2(b ) of the main text .
quantum process tomography ( qpt ) provides a means of characterizing quantum gates by comparing the ideal outcomes of gate operations with their actual outcomes . here , we follow the qpt recipe given in .
we solve the master equation , eq .
( 3 ) in the main text , for a specified pulse sequence for a given gate operation . for each simulation , the detuning parameters and the local magnetic fields
are held constant . using the simulation results
, we calculate the final fidelity , as outlined below . in the following section ,
we describe our method for performing statistical averages of those fidelities , taking into account the fluctuations of the detuning parameters and the random magnetic fields .
we now summarize the qpt method .
we consider a gate operation @xmath133 acting on an initial state described by the density matrix @xmath70 . here
, @xmath133 represents the final density matrix , and has no relation to the detuning parameter .
the @xmath134 operation can be expressed in terms of operation elements @xmath135 , such that @xmath136 the operation elements can be decomposed with respect to an orthogonal basis set of operators @xmath137 for the state space , such that @xmath138 where @xmath139 are complex numbers .
if @xmath70 describes a single qubit , then each @xmath137 is a @xmath80 matrix .
a convenient choice is the basis set @xmath140 where @xmath141 are pauli matrices .
we then have @xmath142 where the process matrix @xmath143 is defined as @xmath144 the process matrix can be fully characterized by initializing the system into linearly independent basis elements for the density matrix .
a convenient choice of initial states is @xmath29 , @xmath145 , @xmath146 , and @xmath147 .
we then perform appropriate linear combinations of gate operations on the initial states , as described in .
once the process matrix has been reconstructed , the process fidelity for a single - qubit rotation is given by @xcite @xmath148 + 1\right ) , \ ] ] where @xmath149 represents the ideal process matrix .
in the previous section , we described the calculation of qpt fidelities for individual simulations .
each simulation is performed for a constant value of the detuning parameters and the local nuclear fields . however , these parameters are all quasistatic , and we should perform an average over these quantities , as described in the main text , to describe the inhomogeneous broadening .
there are two different fluctuation axes for the detuning parameters ( @xmath24 and @xmath25 ) and two different axes for the nuclear fields ( @xmath150 and @xmath151 ) . while it is not computationally feasible to perform accurate , simultaneous averages over four different fluctuation axes , it is possible to perform simultaneous averages over two axes at a time .
we choose to perform simultaneous averages over the detuning fluctuations and the random overhauser fields separately , to distinguish the effects of charge and nuclear noise .
these calculations are computationally intensive .
we first consider the quasistatic random overhauser fields .
when @xmath152 , we only need to consider the longitudinal components of @xmath153 @xcite .
we model the probability distributions of these random fields as @xmath154 where @xmath155 is the standard deviation of the random fields .
the master equation is solved over a grid @xmath156 of size @xmath157 , with @xmath158 , while keeping @xmath24 and @xmath25 fixed .
the noise - averaged fidelity is then given by @xmath159 in our simulations , we choose @xmath150 and @xmath151 in the range ( -12 mt,+12 mt ) , and @xmath155=4 mt , as consistent with @xcite . similarly , we consider fluctuations of the detuning parameters keeping the local magnetic fields fixed .
we model the fluctuation probability distribution as @xmath160 over a grid @xmath161 of @xmath157 points , with @xmath162 .
the noise - averaged fidelity is then given by @xmath163 in our simulations , we choose @xmath164 and @xmath165 in the range ( -15 @xmath166ev,+15 @xmath166ev ) , and @xmath167=5 @xmath166ev , as consistent with @xcite .
figures 2(a ) and ( c ) of the main text show comparisons of nuclear and charge noise averaging results , while the insets show comparisons of charge noise averaging for three different values of @xmath14 . in this section , we extend these results by plotting nuclear noise - averaged results for three different values of @xmath14 , as shown in fig . s1 . as before
, we find that @xmath87 rotations have fidelities that are approximately 20 times worse than @xmath83 rotations , with optimal values that improve when @xmath14 is larger .
in previous sections , particularly in fig . 2 of the main text and fig .
s1 of the supplemental materials , we compared the effects of random nuclear fields and detuning fluctuations .
when we simulated detuning fluctuations , we adopted a fixed , characteristic magnetic field difference between the two dots . to complete this story , we perform the same simulation here , setting the static overhauser fields to zero .
this can be viewed as the ideal case for perfect , isotopically purified @xmath35si devices , whereas the previous simulations corresponded to gaas .
the results of our @xmath35si simulation are shown in fig .
s2 , assuming only detuning fluctuations .
we also show the equivalent gaas simulation for comparison , with the same detuning fluctuations but setting @xmath168 mt , as in fig . 2 of the main text .
we see that using @xmath35si is highly beneficial in two ways .
first , it yields an improvement in the maximum fidelity .
second , it lowers the optimal tunnel coupling , and therefore the gate speed , to a range that may be more convenient from a technological perspective . at even lower gate speeds ,
fast charge noise eventually degrades the gate fidelity . at higher tunnel couplings , where the effects of quasistatic charge noise and leakage to doubly - occupied charge states dominate the fidelity ,
the presence of a magnetic field difference is irrelevant .
once again , we find the optimal fidelity for @xmath1-rotations is approximately one order of magnitude better than for @xmath2-rotations , due to the presence of the sweet spot . | optimal working points or sweet spots " have arisen as an important tool for mitigating charge noise in quantum dot logical spin qubits .
a great challenge in quantum computation is to perform prescribed operations with very small error rates .
logical qubits are important for achieving this @xcite , since they are fundamental for quantum error correction @xcite .
moreover , logical qubits can have symmetries that give rise to so - called sweet spots , at which the effects of noise are suppressed @xcite .
several logical spin qubits have been proposed for quantum dot architectures @xcite . here , we consider the exchange - only logical qubit @xcite , formed of three electrons in a triple dot @xcite , as illustrated in fig .
[ fig : sebh](d ) .
this qubit has the advantage that it has the potential to be very fast , since all operations can be implemented without spatially varying magnetic fields .
the effects of charge noise can never be fully suppressed , even near a sweet spot @xcite . in this paper , we quantify the effect of sweet spots on gate fidelities by performing theoretical simulations of pulsed gate operations in an exchange - only qubit .
the sweet spot in this device occurs at the symmetry point shown in fig .
[ fig : sebh](b ) , where the detuning parameters @xmath0 , and the charge - induced fluctuations of the detuning @xcite are suppressed , to leading order .
( charge noise in the tunnel coupling @xcite is not suppressed at this point , but is not thought to be a dominant noise source @xcite . ) as consistent with recent experiments @xcite , @xmath1-rotations are performed at the sweet spot , while @xmath2-rotations are obtained by pulsing away from this point . in principle , the different rotations can be turned on and off independently . however , in practice it may be necessary to turn on the exchange interactions and magnetic field at all times , to suppress leakage into the non - logical sector of the hilbert space @xcite .
the exchange - only qubit provides an ideal platform for assessing the effect of sweet spots , since all gate operations are generated by the same physical process ( the exchange interaction @xcite ) .
the only difference between @xmath2 and @xmath1-rotations is their proximity to the sweet spot .
the fidelities of these operations can therefore be used to quantify the effectiveness of the sweet spot for mitigating charge noise .
this is in contrast with logical qubits where the different rotation axes correspond to different physical processes ( e.g. , exchange vs. magnetic couplings in singlet - triplet qubits @xcite ) . in ref .
@xcite , we provided a detailed account of magnetic noise from the overhauser fields of nuclear spins on the decoherence of an exchange - only qubit . here , we simulate realistic gate operations including quasistatic random overhauser fields @xcite and charge noise @xcite . in certain regimes
we find that the main limit on the gate fidelities arises from the overhauser fields , as consistent with experimental observations @xcite .
however , when the gates are properly optimized , we predict that charge noise should determine the upper bound on gate fidelities .
after optimization , we find that gate fidelities at the sweet spot are typically 20 times better than away from the sweet spot . _ theoretical model._ we model the coherent evolution of the exchange - only qubit using a 3-electron , 3-site hubbard model with the hamiltonian @xcite @xmath3 where the labels @xmath4 correspond to dot locations , @xmath5 , and @xmath6 refer to individual spin @xmath7 eigenstates , @xmath8 and @xmath9 are electron creation and annihilation operators , and @xmath10 is the electron number operator . the first term in eq .
( [ eq : hamiltonian ] ) describes the tunneling , with tunnel couplings @xmath11 .
we assume a symmetric , linear triple dot geometry , as shown in fig .
[ fig : sebh](d ) , with @xmath12 and @xmath13 .
the second term describes the onsite coulomb repulsion , with energies @xmath14 that are the same at every site .
the third term describes the local electrostatic potentials @xmath15 .
the fourth term describes the zeeman energy due to a uniform external magnetic field @xmath16 , with the land @xmath17-factor and bohr magneton @xmath18 .
the fifth term describes the local variations of the zeeman energy due to overhauser field fluctuations @xmath19 . here , we take @xmath20 because the lateral components of @xmath19 generate couplings between @xmath21 manifolds that are highly suppressed in the regime of large zeeman splittings , which we consider below .
we also ignore coulomb interactions between electrons in different dots .
the detuning @xmath22 is defined in analogy with experiments @xcite , and corresponds to the energy difference between the ( 2,0,1 ) and ( 1,0,2 ) charge configurations . for a triple
dot , there is also a second , independent detuning parameter @xcite , which we define here as @xmath23 . in experimental systems , the detunings @xmath24 and @xmath25 are controlled by voltages , including @xmath26 and @xmath27 , which are applied to the top - gates . a typical charge stability diagram is shown in fig .
[ fig : sebh](a ) for a fixed value of @xmath25 . .
( b ) energy level diagram of the @xmath28 manifold as a function of the detuning parameter @xmath24 . in the central region ,
the low energy states @xmath29-@xmath30 are in the ( 1,1,1 ) charge configuration , while the high energy states are doubly occupied .
( c ) bloch sphere representation of the logical qubit , with the rotation axes corresponding to @xmath31 ( left ) and @xmath32 ( right ) .
( d ) hubbard model of a triple quantum dot containing three electrons . ] the hilbert space associated with eq .
( [ eq : hamiltonian ] ) is large . for gaas - based devices ,
most leakage channels can be suppressed by enforcing sizeable energy splittings @xcite . as consistent with recent experiments @xcite
, we therefore consider the energy hierarchy @xmath33 , where @xmath34 is the exchange interaction generated by the tunnel couplings .
( in @xcite , we briefly consider @xmath35si - based devices , which do not require such an energy heirarchy , due to the absence of nuclear spins . ) since @xmath36 is large , the energy spectrum splits into manifolds of constant total spin @xmath21 .
our simulations focus on the seven states in the @xmath28 manifold , where the two qubit states are defined in the decoherence free subspace with @xmath37 @xcite . for a basis
set , we consider the seven eigenstates of eq .
( [ eq : hamiltonian ] ) when @xmath38 , consisting of three singly - occupied @xmath39 states , @xmath40 , @xmath41 , and @xmath42 , and four doubly - occupied states , @xmath43 , @xmath44 , @xmath45 , and @xmath46 . here
, the subscript denotes the dot index , @xmath47 , @xmath48 , and @xmath49 are the singlet and triplet states of two spins , and @xmath50 represents a dot with no electrons . @xmath51 and @xmath52 are the logical qubit states , @xmath53 is the main leakage state , and the doubly occupied states mediate the exchange interaction .
the gate simulations described below include the full set of seven basis states , in order to address questions of leakage and decoherence .
however it is instructive to consider the effective hamiltonian in the @xmath54 logical subspace @xcite , @xmath55 where @xmath56 and @xmath57 are exchange interactions .
the latter may be tuned independently as a function of the control parameters @xmath24 and @xmath25 , yielding a continuous set of rotations in the @xmath58-@xmath59 plane of the bloch sphere .
for example , we could independently set @xmath56 or @xmath57 to zero , yielding the pair of rotation axes shown in fig .
[ fig : sebh](c ) . from eq .
( [ eq : h2 ] ) , we see that @xmath1-rotations are obtained when @xmath60 . in @xcite ,
we show that this requirement is met when either @xmath61 or @xmath62 .
we also show that the special combination @xmath0 corresponds to a detuning sweet spot , because @xmath63 , where @xmath64 is the energy splitting between the qubit states .
since always - on exchange interactions are needed to prevent leakage , and since @xmath65 , eq .
( [ eq : h2 ] ) suggests that we can not achieve pure @xmath2-rotations .
we overcome this problem by implementing a three - step pulse sequence @xcite .
this procedure requires moving away from the sweet spot , with consequences for the decoherence and gate fidelity .
finally , we note that a complete set of single - qubit operations must include initialization and readout .
the latter are accomplished in experiments by adiabatically tuning the device to the @xmath66 or @xmath67 charge configurations in the far - detuned regime of fig .
[ fig : sebh](b ) @xcite . in our simulations
, we do not investigate readout and initialization ; we consider only the unitary gate operations .
moreover , we assume instantaneous ( diabatic ) pulses and do not investigate pulse imperfections . we consider only the errors caused by charge and nuclear noise sources , and by leakage outside the logical qubit hilbert space .
_ gate simulations._ we simulate the dynamics of the logical qubit gate operations by solving the master equation @xmath68-d[\rho(t)]\ ] ] for the @xmath69 density matrix , @xmath70 . the first term on the right - hand side of eq .
( [ eq : master ] ) describes the unitary evolution , while the second term describes the decoherence .
we consider dephasing from charge noise and random overhauser fields .
the nuclear fluctuations occur at frequencies much lower than the relevant electronic time scales @xcite ; we take them to be quasistatic with a gaussian distribution of width @xmath71 mt , as appropriate for gaas @xcite .
we consider two types of fast charge noise .
the virtually occupied states @xmath72-@xmath73 mediate exchange interactions , but they also contribute to double occupation dephasing errors of the form @xcite @xmath74 $ ] .
we also consider direct dephasing @xmath75 of the singly occupied states @xmath29-@xmath30 , with rates that depend on the derivative of the energy splitting @xmath76 between eigenstates @xmath77 and @xmath78 with respect to the detuning @xcite .
we assume that contributions from the individual detuning parameters contribute in quadrature , with the dephasing rates @xmath79^{1/2}$ ] . here
, the factor of 2 reflects the relative magnitudes of the @xmath24 and @xmath25 terms in the effective @xmath80 hamiltonian for the logical qubit states @xcite .
the resulting dephasing matrix is given by @xmath81 = d_u + d_\varepsilon = \\
& \hspace{-0.05 in } \begin{pmatrix } 0 & \gamma_{01}\rho_{01 } & \gamma_{02}\rho_{02 } & \gamma\rho_{03 } & \gamma\rho_{04 } & \gamma\rho_{05 } & \gamma\rho_{06 } \\ \gamma_{01}\rho_{01}^ * & 0 & \gamma_{12}\rho_{12 } & \gamma\rho_{13 } & \gamma\rho_{14 } & \gamma\rho_{15 } & \gamma\rho_{16}\\ \gamma_{02}\rho_{02}^ * & \gamma_{12}\rho_{12}^ * & 0 & \gamma\rho_{23 } & \gamma\rho_{24 } & \gamma\rho_{25 } & \gamma\rho_{26}\\ \gamma\rho_{03}^ * & \gamma\rho_{13}^ * & \gamma\rho_{23}^ * & 0 & 4\gamma\rho_{34 } & 3\gamma\rho_{35 } & \gamma\rho_{36}\\ \gamma\rho_{04}^ * & \gamma\rho_{14}^ * & \gamma\rho_{24}^ * & 4\gamma\rho_{34}^ * & 0 & \gamma\rho_{45 } & 3\gamma\rho_{46}\\ \gamma\rho_{05}^ * & \gamma\rho_{15}^ * & \gamma\rho_{25}^ * & 3\gamma\rho_{35}^ * & \gamma\rho_{45}^ * & 0 & 4\gamma\rho_{56}\\ \gamma\rho_{06}^ * & \gamma\rho_{16}^ * & \gamma\rho_{26}^ * & \gamma\rho_{36}^ * & 3\gamma\rho_{46}^ * & 4\gamma\rho_{56}^ * & 0 \end{pmatrix } . \nonumber\end{aligned}\ ] ] we treat the slow fluctuations of the detuning and overhauser fields by numerically solving the 49 coupled real differential equations in eq .
( [ eq : master ] ) for a fixed noise realization @xcite .
we then repeat the calculations for 625 realizations of overhauser field fluctuations and 961 realizations of detuning fluctuations , and perform the appropriate gaussian averages .
the simulations are performed on the open science grid at the university of wisconsin - madison @xcite .
the results reported here represent @xmath82 compute years . _ gate optimization._ we begin by considering @xmath83 rotations of the logical qubit .
as described above , these operations are performed at the sweet spot @xmath0 .
fluctuations of the detuning and the overhauser fields give rise to errors within the qubit subspace as well as leakage .
we monitor these effects by performing quantum process tomography ( qpt ) @xcite , beginning the simulations in four different initial states , and comparing the final results to the ideal final states for a fixed value of the tunnel coupling @xmath84 . in this procedure ,
the evolution period @xmath85 is treated as a variable .
the optimal value of @xmath85 is chosen by maximizing the fidelity @xmath86 obtained from qpt , with results shown in fig .
2(a ) . for small @xmath84 ,
the rotations are slow , and the fidelity is strongly suppressed by the quasistatic random overhauser fields . for large @xmath84 , the rotations are fast , and the fidelity is determined by a combination of charge noise and leakage . since the leakage process is coherent , the projection of the full density matrix onto the logical qubit subspace undergoes oscillations , as seen in the lower inset of fig .
2(a ) .
these oscillations are severe for large tunnel couplings , causing a deterioration of the fidelity as seen in the main figure .
we also investigate @xmath87 rotations of the logical qubit . as noted above
, it is not possible to perform a direct rotation around @xmath88 ; accurate rotations require multi - pulse gate sequences . here
, we consider a three - step procedure @xcite that can be visualized as shown in the lower inset of fig .
2(c ) .
the sequence consists of ( i ) a @xmath89-rotation around the @xmath90 axis on the bloch sphere , ( ii ) a @xmath83 rotation , and ( iii ) a final @xmath89-rotation around the @xmath90 axis . for steps ( i ) and ( iii ) , the values of @xmath24 and @xmath25 that determine the axis tilt
are not known _ a priori _ ; we find them by performing fidelity simulations for the desired gate operations in the absence of detuning and nuclear fluctuations .
the results are shown in fig .
2(b ) for a fixed value of @xmath84 , this calibration procedure is then repeated for other values of @xmath84 .
the three - step protocol is then optimized , step by step , by performing simulations to determine the evolution period @xmath85 that maximizes the fidelity of each step .
the final fidelities of the three - step @xmath87 protocol are shown in fig .
2(c ) .
we observe results similar to those in fig .
2(a ) .
however , the effects of leakage and charge noise are more severe because steps ( i ) and ( iii ) are not performed at sweet spots .
the suppression of the fidelity due to leakage is most obvious at large @xmath84 .
the lower inset shows a typical evolution projected onto the logical qubit bloch sphere . _ results and discussion._ the @xmath2 and @xmath1-rotation protocols used in fig .
2 are different . however , by comparing fidelities obtained using qpt , we can compare the final results effectively .
we observe that maximal fidelities ( or minimal infidelities , @xmath91 ) occur over a range of moderate to large tunnel couplings , @xmath92 - 20 ghz , that depends on the hubbard repulsion parameter @xmath14 .
our results also depend on the local field gradients @xmath93 , which determine the leakage rate .
the values of @xmath93 considered here are typical for gaas triple dots .
the optimal fidelities in fig . 2
occur on a plateau , whose value is largely determined by the detuning noise .
this is not the same conclusion reached in @xcite , where fidelity limits were attributed to nuclear noise .
we speculate that those experiments were performed at lower @xmath84 , below the plateau , where nuclear noise predominates .
we emphasize that larger @xmath84 should be used to achieve maximal fidelities .
our most important results are obtained by comparing the maximal fidelities of @xmath87 and @xmath83 rotations .
we find that @xmath2-rotations have maximum fidelities @xmath9420 times worse than @xmath1-rotations , which can be directly attributed to the fact that @xmath2-rotations occur away from the sweet spot , while @xmath1-rotations occur at the sweet spot .
the degradation of @xmath2-rotations is most noticeable for large @xmath84 , where the fidelity is dominated by charge noise .
measurements of the quantum dot hybrid qubit show a similar degradation of coherence away from a sweet spot @xcite .
this suggests that ac gating techniques could yield better fidelities than the dc pulsing techniques studied here , because the detuning is always centered at the sweet spot @xcite .
indeed , recent experiments on the exchange - only qubit have employed such a strategy @xcite . on the other hand , ac methods tend to produce somewhat slower gates , for which nuclear noise could be a problem .
finally , we note that our analysis has focused on gaas quantum dot devices , where nuclear noise is known to be important . for si - based devices , especially isotopically purified @xmath35si
, the nuclear noise can be very small . as a result
, si devices should yield better fidelities for dc pulsed gates , especially in the low-@xmath84 regime .
( see @xcite . ) at higher @xmath84 , where nuclear noise is not predominant , si and gaas exchange - only qubits should have similar fidelities .
we thank z. shi , x. wu , k. rudinger , j. gamble , and c. wong for helpful discussions .
we also thank the hep , condor , and chtc groups at uw - madison for computational support .
this work was supported in part by aro ( w911nf0910393 ) , nsf ( phy-1104660 and phy-1104672 ) , uw - madison ( 150 486700 4 ) , and by the united states department of defense .
the views and conclusions contained in this paper are those of the authors and should not be interpreted as representing the official policies , either expressed or implied , of the us government .
* appendix * in these appendices , we provide details about the calculations and simulations discussed in the main text .
section a describes the hamiltonian for the @xmath28 spin manifold .
section b provides analytical estimates for the exchange interactions in certain operating regimes of interest .
section c provides details of the quantum process tomography methods .
section d describes our statistical averaging procedure for treating quasistatic charge and nuclear noise .
section e describes some additional results for simulations with averages over overhauser field gradients .
section f describes results with no overhauser field gradients , consistent with pure , isotopically purified @xmath35si . |
let @xmath7 be a smooth hypersurface and @xmath8 a smooth , compactly supported surface measure on @xmath9 an averaging operator @xmath10 over @xmath4 is given by @xmath11 where @xmath12 is a complex valued function on @xmath13 in this euclidean setting , the averaging problem is to determine the optimal range of exponents @xmath14 such that @xmath15 where @xmath16 denotes the space of schwartz functions .
when @xmath4 is the unit sphere @xmath17 this problem is closely related to regularity estimates of the solutions to the wave equation at time @xmath18 , and it was studied by r.s .
strichartz @xcite .
it is well known that @xmath19 averaging results can be obtained by the decay estimates of the fourier transform of the surface measure @xmath8 on @xmath4 .
for instance , if @xmath20 for some @xmath21 , then the averaging inequality ( [ ai ] ) holds whenever @xmath22 where @xmath23 denotes the exponent conjugate to @xmath24 ( see ) .
thus , if @xmath25 and @xmath26 then the averaging estimate ( [ ai ] ) holds . since @xmath27 and @xmath28 estimates are clearly possible , we see from the interpolation theorem that if @xmath29 then @xmath19
estimates hold whenever @xmath30 lies in the triangle @xmath31 with vertices @xmath32 and @xmath33 moreover , it is well known that @xmath19 estimates are impossible if @xmath34 lies outside of the triangle @xmath35 such analogous phenomena were also observed in the finite field setting ( see , for example , ) .
+ on the other hand , if the optimal fourier decay estimate of the surface measure @xmath8 is given by @xmath36 then it is in general hard to prove sharp averaging results and some technical arguments are required to deal with the case ( see @xcite ) . in the finite field case ,
this was also pointed out by the authors in @xcite . + as an analogue of euclidean averaging problems , carbery , stones , and wright @xcite initially introduced and studied the averaging problem in finite fields and the author with shen has recently investigated the averaging problem over homogeneous varieties . in this introduction , we shortly review notation and definitions for averaging problems in the finite field setting and readers are referred to @xcite for further information and motivation on the averaging problem .
let @xmath37 be a @xmath38-dimensional vector space over a finite field @xmath2 with @xmath3 elements .
we endow the space @xmath37 with a normalized counting measure @xmath39 .
let @xmath40 be an algebraic variety .
then a normalized surface measure @xmath41 supported on @xmath4 can be defined by the relation @xmath42 where @xmath43 denotes the cardinality of @xmath4 ( see @xcite ) .
an averaging operator @xmath10 can be defined by @xmath44 where both @xmath12 and @xmath45 are functions on @xmath46 in this setting the averaging problem over @xmath4 is to determine @xmath47 such that @xmath48 where the constant @xmath49 is independent of functions @xmath12 and @xmath3 , the cardinality of the underlying finite field @xmath50 let @xmath51 we denote by @xmath52 to indicate that the averaging inequality ( [ apq ] ) holds .
the main purpose of this paper is to obtain the complete @xmath19 estimates of the averaging operators over varieties determined by nondegenerate quadratic form over @xmath2 .
let @xmath53 $ ] be a nondegenerate quadratic form .
define a variety @xmath54 we shall name this kind of varieties as a nondegenerate quadratic surface in @xmath55 since @xmath1 is a nondegenerate quadratic form , it can be transformed into a diagonal form @xmath56 with @xmath57 by means of a linear substitution ( see @xcite ) .
therefore , we may assume that any nondegenerate quadratic surface can be written by the form @xmath58 where @xmath59 the necessary conditions for the averaging estimates over @xmath60 were given in @xcite .
in fact , @xmath61 only if @xmath30 lies in the convex hull of @xmath62 and @xmath63 it is known from @xcite that this necessary conditions for @xmath64 are sufficient conditions if the dimension @xmath65 is odd .
on the other hand , it was observed in @xcite that if @xmath5 is even and @xmath60 contains a subspace @xmath66 with @xmath67 then @xmath64 only if @xmath30 lies in the convex hull of @xmath68 in this paper we show that ( [ imnec ] ) is also the sufficient conditions for @xmath64 in the specific case when the variety @xmath60 contains @xmath6-dimensional subspace with @xmath5 even
. see figure [ fig1 ] . in the case
that @xmath5 is even and @xmath60 contains a @xmath6-dimensional subspace , scaledwidth=80.0% ] [ mainthm ] let @xmath8 be the normalized surface measure on the nondegenerate quadratic surface @xmath69 as defined in ( [ defv ] ) .
suppose that @xmath5 is an even integer and @xmath60 contains a @xmath6-dimensional subspace .
then @xmath64 if and only if @xmath30 lies in the convex hull of @xmath70 if the dimension @xmath5 is even , then the diagonal entries @xmath71 can be properly chosen so that @xmath60 contains a @xmath6-dimensional subspace @xmath72 such an example is essentially the following one ( see theorem 4.5.1 of @xcite ) : @xmath73 and @xmath74 [ remgo ] from the observation ( [ imnec ] ) , we only need to prove the if " part of theorem [ mainthm ] . since @xmath75 is the normalized counting measure on @xmath37 , it follows from young s inequality that @xmath61 for @xmath76 thus , by duality and the interpolation theorem , it will be enough to prove that @xmath77 the authors in @xcite showed that this inequality holds for all characteristic functions on subsets of @xmath55 here , we improve upon their work by obtaining the strong - type estimate . in the remaining parts of this paper , we focus on providing the detail proof of theorem [ mainthm ] . in section 2 ,
we review the fourier analysis in finite fields and prove key lemmas which are essential in proving our main theorem .
the proof of theorem [ mainthm ] for even dimensions @xmath78 will be completed in section 3 .
namely , when @xmath78 is any even integer , the inequality ( [ rema ] ) will be proved in section 3 . in the final section ,
we finish the proof of theorem [ mainthm ] by proving the inequality ( [ rema ] ) for @xmath79
in this section we drive key lemmas which play a crucial role in proving theorem [ mainthm ] .
we begin by reviewing the discrete fourier analysis and readers can be reffered to @xcite for more information on it .
let @xmath2 be a finite field with @xmath3 elements . throughout this paper , we assume that @xmath3 is a power of odd prime so that the characteristic of @xmath2 is greater than two .
we denote by @xmath80 a nontrivial additive character of @xmath50 recall that the orthogonality relation of the canonical additive character @xmath80 says that @xmath81 where @xmath37 denotes the @xmath38-dimensional vector space over @xmath2 and @xmath82 is the usual dot - product notation .
denote by @xmath83 the vector space over @xmath2 , endowed with the normalized counting measure @xmath84 its dual space will be denoted by @xmath85 and we endow it with a counting measure @xmath86 if @xmath87 then the fourier transform of the function @xmath12 is defined on @xmath85 : @xmath88 we also recall the plancherel theorem : @xmath89 throughout this paper , we identify the set @xmath90 with the characteristic function on the set @xmath91 we denote by @xmath92 the inverse fourier transform of the normalized surface measure @xmath8 on @xmath60 in ( [ defv ] ) . recall that @xmath93 let @xmath94 denote the quadratic character of @xmath95 for each @xmath96 the gauss sum @xmath97 is defined by @xmath98 the absolute value of the gauss sum is given as follows ( see @xcite ) : @xmath99 it turns out that the inverse fourier transform of @xmath8 can be written in terms of the gauss sum .
the following was given in lemma 4.1 of @xcite .
[ fdecay ] let @xmath60 be the variety in @xmath37 as defined in ( [ defv ] ) , and let @xmath8 be the normalized surface measure on @xmath100 if @xmath101 is even , then we have @xmath102 here and throughout this paper we denote by @xmath94 the quadratic character of @xmath103 and we write @xmath104 for the gauss sum @xmath105 lemma [ fdecay ] yields the following corollary .
[ decay]let @xmath60 be the variety in @xmath37 as defined in ( [ defv ] ) and let @xmath8 be the normalized surface measure on @xmath100 if @xmath5 is even , then we have @xmath106 and @xmath107 by the definition of @xmath108 , it is clear that @xmath109 comparison with lemma [ fdecay ] shows that @xmath110 since @xmath111 , it follows that @xmath112 for @xmath5 even .
this proves ( [ sizes ] ) .
the inequality ( [ evendecay ] ) follows immediately from lemma [ fdecay ] because @xmath111 and @xmath113 for @xmath5 even .
it is clear from ( [ evendecay ] ) that if @xmath5 is even and @xmath60 is any nondegenerate quadratic surface in @xmath114 then @xmath115 recall that @xmath8 is the normalized surface measure on the nondegenerate quadratic surface @xmath60 . in the finite field setting , the bochner - riesz kernel @xmath116 is a function on @xmath85 and it satisfies that @xmath117 recall that @xmath118 denotes the counting measure on @xmath55 notice that @xmath119 if @xmath120 , and @xmath121 otherwise . also observe that @xmath122 here , the last equality follows because @xmath123 is defined on the vector space with the counting measure @xmath124 and its fourier transform @xmath125 is defined on the dual space with the normalized counting measure @xmath126 more precisely , if @xmath127 then @xmath128 our main lemma is as follows .
[ mainlem ] suppose that @xmath78 is even .
then , for every @xmath129 we have @xmath130 where @xmath116 is the bochner - riesz kernel . on the other hand , for every @xmath131
it follows that @xmath132 using the interpolation theorem , it suffices to prove that the following two inequalities hold for all d@xmath133 even : @xmath134 and @xmath135 the estimate ( [ samil ] ) can be obtained by applying young s inequality .
in fact , we see that @xmath136 since @xmath137 and @xmath138 the inequality ( [ samil ] ) is established . to prove the inequality ( [ sami ] ) , first use the plancherel theorem .
it follows that @xmath139 now , we recall that @xmath75 is the normalized counting measure but @xmath118 is the counting measure .
thus , the expression above is given by @xmath140 where the first line and the second line follow from the definition of @xmath116 and the inequality ( [ evendecay ] ) in corollary [ decay ] , respectively .
applying the plancherel theorem , it is clear that @xmath141 in order to obtain a good upper bound of @xmath142 , we shall conduct two different estimates on @xmath143 first , the plancherel theorem yields @xmath144 on the other hand , it follows that @xmath145 now , let @xmath146 which is also a nondegenerate quadratic surface with @xmath147 then the expression above can be written by @xmath148 now , we see from ( [ ko ] ) that if @xmath149 , then @xmath150 thus , we obtain that @xmath151 combining this with the inequality ( [ ei1 ] ) gives @xmath152 in conjunction with the inequality ( [ eii ] ) , this shows that @xmath153 since @xmath154 for @xmath155 , it also follows that @xmath156 a direct computation shows that this implies the inequality ( [ sami ] ) .
we complete the proof of lemma [ mainlem ] .
in this section we provide the complete proof of theorem [ mainthm ] in the case that @xmath78 is even .
the proof for @xmath157 shall be independently given in the following section .
the main reason is as follows .
lemma [ mainlem ] shall be used to prove theorem [ mainthm ] .
if @xmath78 is even , then we have seen that the inequality ( [ six ] ) of lemma [ mainlem ] follows by interpolating ( [ samil ] ) and ( [ sami ] ) . however ,
if @xmath38 is four , then such an interpolation is too meaningless to assert that ( [ six ] ) holds for @xmath79 as an alternative approach , the inequality ( [ four ] ) of lemma [ mainlem ] shall be applied to complete the proof for @xmath79 in this case we need more delicate estimates .
+ now we start proving theorem [ mainthm ] for @xmath78 even .
as mentioned in remark [ remgo ] , it is enough to prove the following statement .
[ res1 ] let @xmath60 be the variety in @xmath37 as defined in ( [ defv ] ) .
if @xmath78 is even , then we have @xmath158 where @xmath159
let @xmath160 and @xmath161 we aim to prove that for every complex - valued function @xmath12 on @xmath114 @xmath162 as in @xcite we proceed with the proof by decomposing the function @xmath12 to which the operator is applied into level sets .
without loss of generality , we may assume that @xmath12 is a nonnegative real - valued function and @xmath163 therefore , we may also assume that @xmath164 where @xmath165 are disjoint subsets of @xmath55 it follows from these assumptions that @xmath166 and hence for every @xmath167 @xmath168 recall that @xmath169 where @xmath116 is the bochner - riesz kernel .
it follows that @xmath170 since @xmath171 and @xmath75 is the normalized counting measure on @xmath114 it is clear from young s inequality that @xmath172 therefore , it suffices to prove the following inequality @xmath173 since we have assumed that @xmath174 , we see that @xmath175 also observe that @xmath176 from these observations , our task is to show that @xmath177 using ( [ simp ] ) , we see that @xmath178 where the last line follows by the symmetry of @xmath179 and @xmath180 and the inequality ( [ samil ] ) . now , for each @xmath181 we consider the following three sets : @xmath182 @xmath183 and @xmath184 since @xmath185 it is clear from ( [ six ] ) in lemma [ mainlem ] that our goal is to prove the following three inequalities : @xmath186 @xmath187 @xmath188 first , we prove that the inequality ( [ g1 ] ) holds .
since @xmath189 a direct computation shows that @xmath190 now recall from ( [ as2 ] ) that @xmath191 for all @xmath192 therefore , it follows that @xmath193 since @xmath194 and the sum over @xmath195 is a geometric series , we see that @xmath196 thus , the inequality ( [ g1 ] ) is established as follows : @xmath197 where we used the simple observation that @xmath198 and then the assumption @xmath199 + second , we prove that the inequality ( [ g2 ] ) holds .
let @xmath200 since @xmath201 we see that @xmath202 write @xmath203 as follows : @xmath204 since @xmath205 we notice from ( [ as2 ] ) that @xmath206 by the definition of the set @xmath207 , we also see that @xmath208 for all @xmath209 then , we have @xmath210 notice that @xmath211 and the geometric series over @xmath195 converges to @xmath212 because @xmath213 for @xmath214 from this observation and ( [ as1 ] ) , the inequality ( [ g2 ] ) follows because we have @xmath215 finally , we show that the inequality ( [ g3 ] ) holds .
as in the proof of the inequality ( [ g2 ] ) , we let @xmath216 for @xmath214 the value @xmath217 is written by @xmath218 notice from ( [ as2 ] ) that @xmath219 for all @xmath220 by the definition of @xmath221 , it is easy to notice that @xmath222 for @xmath223 it therefore follows that @xmath224 where we used the facts that @xmath225 @xmath226 for @xmath201 and @xmath227 and then the assumption ( [ as1 ] ) for the last equality .
thus , the inequality ( [ g3 ] ) holds and the proof of theorem [ res1 ] is complete .
as observed in remark [ remgo ] , it amounts to showing the following statement
. we will proceed by the similar ways as in the previous section .
however , the proof of the theorem will be based on ( [ four ] ) , rather than ( [ six ] ) in lemma [ mainlem ] .
we begin by recalling from ( [ four ] ) and ( [ sami ] ) that @xmath230 and @xmath231 we must show that for all complex - valued functions @xmath12 on @xmath232 @xmath233 as noticed in the previous section , it suffices to prove this inequality under the following assumptions : @xmath234 where @xmath165 are disjoint subsets of @xmath235 from these assumptions , it is clear that @xmath236 this clearly implies that @xmath237 according to ( [ same ] ) , it is enough to prove that @xmath238 since @xmath239 it is enough to show that @xmath240 by the symmetry of @xmath179 and @xmath180 and the h " older inequality , our task is to prove @xmath241 main steps to prove this inequality are summarized as follows . by considering the sizes of @xmath242 and @xmath243 , we first decompose @xmath244 as nine parts . next , using the estimates ( [ 6sam]),([sam ] ) , ( [ size ] ) , ( [ one ] ) , and a convergence property of a geometric series , we show that each part of them is @xmath245 which completes the proof of theorem [ res2 ] .
for the sake of completeness , we shall give full details . to do this ,
let us define the following @xmath246 sets : for @xmath247 @xmath248 @xmath249 @xmath250 @xmath251 @xmath252 @xmath253 @xmath254 @xmath255 @xmath256 a. carbery , b. stones , and j. wright , _ averages in vector spaces over finite fields , _ math .
( 2008 ) , * 144 * , 13 , 1327 .
h. iwaniec and e. kowalski , _ analytic number theory , _ colloquium publications , 53 ( 2004 ) .
a. iosevich and e. sawyer , _ sharp @xmath19 estimates for a class of averaging operators , _
fourier , grenoble , * 46 * , 5 ( 1996 ) , 13591384 .
d. koh and c. shen , _ harmonic analysis related to homogeneous varieties in three dimensional vector spaces over finite fields _ , canad .
64 ( 2012 ) , 1036 - 1057 .
d. koh and c. shen , _ extension and averaging operators for finite fields , _ proc .
edinb . math .
soc . , to appear .
w. littman , _
@xmath275 estimates for singular integral operators , _ proc .
pure math .
, * 23*(1973 ) , 479481 .
a. lewko and m. lewko , _ endpoint restriction estimates for the paraboloid over finite fields _ , proc .
( 2012 ) , 2013 - 2028 .
r. lidl and h. niederreiter , _ finite fields , _ cambridge university press , ( 1997 ) .
g. mockenhaupt , and t. tao , _ restriction and kakeya phenomena for finite fields _ , duke math . j. 121(2004 ) , no .
1 , 3574 . w. scharlau , _ quadratic forms _ ,
queen s papers on pure and applied math .
22 , queen s univ . ,
kingston , ontario , 1969 .
e.m.stein , _ harmonic analysis , _ princeton university press , 1993 .
strichartz , _ convolutions with kernels having singularities on a sphere , _ trans .
* 148 * ( 1970 ) , 461471 . | we study mapping properties of the averaging operator related to the variety @xmath0 where @xmath1 is a nondegenerate quadratic polynomial over a finite field @xmath2 with @xmath3 elements .
this paper is devoted to eliminating the logarithmic bound appearing in the paper @xcite . as a consequence ,
we settle down the averaging problems over the quadratic surfaces @xmath4 in the case when the dimensions @xmath5 are even and @xmath4 contains a @xmath6-dimensional subspace . |
in the hierarchical collapse scenario for structure formation in the universe , the number density of collapsed objects as a function of mass and cosmic time is a sensitive probe of cosmology .
the galaxy clusters that occupy the high - mass tail of this population provide a powerful and relatively clean tool for cosmology , since their growth is predominantly determined by linear gravitational processes . in the past , the local population of galaxy clusters has been used to jointly constrain the average matter density of the universe and the amplitude of perturbations in the density field ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
pushing observations to higher redshift breaks the degeneracy between those two parameters ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and allows properties of dark energy to be probed as well ( e.g. * ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
investigations of this type require sky surveys with well understood selection functions to find clusters , as well as a relation linking cluster mass with an observable .
a successful solution to the former requirement has been to identify clusters by the x - ray emission produced by hot intracluster gas , notably using data from the _ rosat _ all - sky survey ( rass ; * ? ? ?
the _ rosat _
brightest cluster sample ( bcs ; * ? ? ?
* ; * ? ? ?
* ) and _ rosat_-eso flux limited x - ray sample ( reflex ; * ? ? ?
* ) together cover approximately two - thirds of the sky out to redshift @xmath13 and contain more than 750 clusters .
the massive cluster survey ( macs ; * ? ? ? * ; * ? ? ?
* ) which at this writing contains 126 clusters and covers 55 per cent of the sky extends these data to @xmath14
. the most straightforward mass observable relation to complement these x - ray flux - limited surveys is the mass x - ray luminosity relation . for sufficiently massive ( hot ) objects at the relevant redshifts ,
the conversion from x - ray flux to luminosity is approximately independent of temperature , in which case the luminosities can be estimated directly from the survey flux and the selection function is identical to the requirement of detection . in a flux - complete survey further restricted to high luminosities ,
every cluster should thus be usable in the analysis , without the need for additional observations other than those required to calibrate the mass luminosity relation .
a disadvantage is that there is a large scatter in cluster luminosities at fixed mass ; however , sufficient data allow this scatter to be quantified empirically .
alternative approaches use cluster temperature @xcite , gas fraction @xcite or @xmath15 parameter @xcite to achieve tighter mass observable relations at the expense of reducing the size of the samples available for analysis .
the need to quantify the selection function in terms of both x - ray flux and a second observable additionally complicates these efforts . in this paper
, we use the observed x - ray luminosity function to investigate two cosmological scenarios , assuming a spatially flat metric in both cases : the first includes dark energy in the form of a cosmological constant ( @xmath16cdm ) ; the second has dark energy with a constant equation - of - state parameter , @xmath2 ( @xmath2cdm ) . in the latter case
, we account for the evolution of density perturbations in the dark - energy fluid , assuming that the dark - energy sound speed is equal to the speed of light . for each model ,
our results are in good agreement with findings from independent cosmological data sets , notably type ia supernovae ( snia ) , the cosmic microwave background ( cmb ) , the x - ray gas mass fraction of galaxy clusters ( @xmath17 ) , and measurements of cosmic shear .
the theoretical background for this work is reviewed in section [ sec : theory ] .
section [ sec : observations ] details the data used to constrain the mass luminosity relation and the cluster samples used to measure the x - ray luminosity function .
the analysis procedure is described in section [ sec : analysis ] and the cosmological results are presented in section [ sec : results ] .
the importance of various systematic effects is discussed in section [ sec : discussion ] . unless otherwise noted ,
masses and luminosities quoted in this paper or shown in figures are computed with respect to a spatially flat @xmath16cdm reference cosmology with hubble constant @xmath18 and @xmath19 .
luminosities and fluxes refer specifically to the 0.12.4@xmath20 energy band in the source and observer rest frames , respectively .
we will consistently use the notation @xmath21 to denote to the true luminosity of a cluster and @xmath22 to denote the luminosity inferred from observation .
we will also write , for example , @xmath0 to refer to the present day matter density in units of the critical density , whereas @xmath23 is the same quantity at redshift @xmath24 .
the variance of the linearly evolved density field , smoothed by a spherical top - hat window of comoving radius @xmath25 , enclosing mass @xmath26 , is @xmath27 here @xmath28 is the mean comoving matter density of the universe , @xmath29 is the linear power spectrum at redshift zero , @xmath30 is the fourier transform of the window function , and @xmath31 is the growth factor of linear perturbations at scales of @xmath32 , normalized to unity at @xmath33 .
we evaluate the transfer function , @xmath34 , and power spectrum , @xmath35 , using the camb package of @xcite .
* hereafter j01 ) and @xcite have shown that the predicted mass function of galaxy clusters of mass @xmath36 at redshift @xmath24 can be written in terms of a `` universal '' function of @xmath37 , @xmath38 which can be fit by a simple form , @xmath39 for cosmological - constant models .
it has since been verified that this fitting function is approximately universal ( within @xmath40 per cent ) among models with constant @xmath41 and some evolving-@xmath2 models @xcite .
we adopt the values @xmath42 , @xmath43 and @xmath44 from j01 , determined using a spherical overdensity group finder at an overdensity of 324 with respect to the mean matter density ( @xmath45 ) .
the number density per unit mass of galaxy clusters of mass @xmath36 at redshift @xmath24 is then @xmath46 following @xcite , we describe the relationship between mass and x - ray luminosity for massive clusters as self - similar ( e.g. * ? ? ?
* ) , modified by an additional redshift - dependent factor @xmath47 where @xmath48 and @xmath49 is the cluster mass defined at an overdensity of @xmath50 with respect to the critical density ( _ i.e. _ where the edge of the cluster is defined such that the mean density enclosed is a fixed factor @xmath50 times the critical density ) .
our model also includes a log - normal scatter of width @xmath51 in luminosity for a given mass .
there is good theoretical reason to expect that this scatter may evolve with redshift , since some of the variability in cluster luminosities can be attributed to the development of cool cores as well as the influence of merger events @xcite .
since current data are insufficient to directly detect evolution in the mass luminosity scatter ( @xcite and references therein ) , we allow for a generic , linear evolution in the scatter : @xmath52 the form of the self - similar evolution appearing in equation [ eq : mlpowlaw ] is appropriate for constraining the mass luminosity relation at fixed overdensity with respect to the critical density .
we have chosen to constrain the relation at an overdensity of @xmath53 in order to match our data ( section [ sec : rbdata ] ) , and because the simulations of @xcite indicate that the virial relations have lower scatter at this overdensity compared with higher overdensities .
the fact that the mass function and mass luminosity relation use different definitions of cluster mass requires us to convert masses from one overdensity to another during the analysis @xcite ; details of carrying out this conversion , assuming a spherically symmetric @xcite density profile , are reviewed in @xcite .
equations [ eq : mfcn_def][eq : mlsevol ] , combined with a suitable stochastic model linking the true luminosity of a cluster with its observed flux , provide the means to predict the galaxy cluster luminosity function based on a set of model parameters ( section [ sec : analysis ] ) .
we have determined the mass luminosity relation of galaxy clusters at low redshift using the sample of ( * ? ? ? * hereafter rb02 ) , shown in figure [ fig : ml_x1_y ] . in these data ,
the luminosities were measured from a combination of pointed _ rosat _ pspc and rass data .
the masses within @xmath54 ( @xmath55 ) were determined by fitting the surface brightness profile with a @xmath56-model @xcite and applying the hydrostatic equation , assuming that the intracluster gas is isothermal . here
@xmath54 is the radius within which the mean density is 500 times the critical density , @xmath57 ( not the universal mean matter density , @xmath28 ) . ) .
the quantities @xmath58 and @xmath59 are defined in equation [ eq : mlmodel2 ] and relate to log - mass and log - luminosity , respectively . as noted in the text , objects with @xmath60 or
@xmath61 are excluded from the analysis .
all masses and luminosities are computed for our reference cosmology . ]
the assumption of hydrostatic equilibrium introduces a bias into the derived masses due to the presence of non - thermal support and asphericity in the observed clusters .
this bias has been observed in simulations @xcite and in comparisons of masses derived from x - ray and weak gravitational lensing observations of the same targets @xcite , although the number of objects in these studies remains low .
@xcite find a bias of @xmath62 per cent at @xmath54 from simulations of 5 clusters .
@xcite find a smaller bias , @xmath63 per cent , from 16 simulated clusters at @xmath33 .
note that there are differences in the methodology of the two groups ; rasia et al .
use an isothermal @xmath56-model to determine the x - ray mass from mock images , while nagai et al . use a more sophisticated x - ray analysis and
additionally consider the observational bias in the determination of @xmath54 .
@xcite have compared masses determined from x - rays and lensing for 18 clusters and find a bias of @xmath64 per cent at @xmath54 , assuming the lensing masses are accurate .
we adopt the nagai et al .
values , @xmath63 per cent , as the nominal values of the mean bias , @xmath65 , and scatter in bias between clusters , @xmath66 . in order to account for the spread among the various results for @xmath65 and the small number of simulated clusters used to estimate @xmath66 , we assign a 20 per cent systematic uncertainty to both of these quantities ( see below ) .
we fit the log - linear model ( see equation [ eq : mlpowlaw ] ) @xmath67 where @xmath68 the full rb02 data set contains some small groups which can be influential in the estimation of the slope , @xmath56 .
since we only require the mass luminosity relation to be calibrated approximately one decade below the luminosity cut used for the x - ray luminosity function data ( section [ sec : xlf ] ) , we fit only the data with @xmath69 in our reference cosmology .
this choice mitigates the possibility of obtaining biased results if slope of the mass luminosity relation is different for massive clusters compared with smaller groups .
the process of fitting the model in equation [ eq : mlmodel ] is complicated by the presence of malmquist bias .
close to the flux limit for selection , any x - ray selected sample will preferentially include the most luminous sources for a given mass .
this results in a steepening of the derived mass luminosity relation and a bias in the inferred intrinsic scatter in luminosity for a given mass .
( the inferred intrinsic dispersion may be artificially reduced or increased , depending on the distribution of data with respect to the flux limit . ) the use of the extended sample of rb02 , rather than only their flux - limited hiflugcs sample , partially mitigates this effect by softening the flux limit .
furthermore , we eliminate the 6 clusters at redshifts @xmath70 which all lie close to the flux limit ; as figure [ fig : ml_z_x1 ] shows , the extent in luminosity of the remaining 86 clusters is roughly a decade at all redshifts . as our estimate ( see below ) of the intrinsic scatter in @xmath59 is roughly 0.12 ( _ i.e. _ 0.12 decades in luminosity ) , and the poisson measurement error is much smaller , we expect the effects of malmquist bias on our estimation of the power - law slope and intrinsic dispersion to be minimal . is defined in equation [ eq : mlmodel2 ] and is computed in our reference cosmology .
the data from rb02 which are excluded from the analysis ( see text ) , are shown as blue triangles .
the remaining clusters have an extent of roughly one decade in luminosity at all redshifts . ] a second consequence of malmquist bias is a strong apparent , but not necessarily physical , correlation between luminosity and redshift due to the fact that the flux limit corresponds to higher luminosities at higher redshifts .
( this is evident in figure [ fig : ml_z_x1 ] . ) within redshift 0.11 , we expect this false signal to be much greater than any real evolution ; we therefore fix @xmath71 when fitting the model .
although methods exist for performing linear regression on data with bivariate heteroscedastic errors and intrinsic scatter ( e.g. * ? ? ?
* ) , such methods do not provide a simple goodness - of - fit measure that can be associated with arbitrary values of the fit parameters .
in contrast , the @xmath72 statistic is an easily calculated measure of goodness - of - fit , but is biased by the presence of intrinsic scatter .
we compromise by using a modified @xmath72 statistic which accounts for intrinsic scatter by introducing an additional dispersion term , @xmath73 where @xmath74 is the measurement error on @xmath75 and @xmath76 is the additional dispersion .
this statistic is defined in terms of @xmath58 given @xmath59 because the luminosity measurement errors are negligible in comparison to the mass measurement errors ( figure [ fig : ml_x1_y ] ) .
the value of @xmath76 is chosen iteratively , adjusting it such that the best fit , found by minimizing @xmath77 using least - squares methods , has @xmath77 equal to the median of the chi - square distribution with @xmath78 ( 86 data points minus 2 free parameters ) degrees of freedom ( _ i.e. _ @xmath79 ) .
the @xmath77 statistic thus does not measure absolute goodness - of - fit , but goodness relative to the best fit .
note that in the cosmological analysis we use the fixed value @xmath80 from our reference cosmology ( see section [ sec : mllikelihood ] for details ) .
a common log - normal intrinsic scatter in luminosity for a given mass can be estimated from the scatter in the data about the best fit , with measurement errors subtracted in quadrature .
since these data are predominantly at very low redshift , we use this procedure to estimate the @xmath33 value of the intrinsic dispersion ( see equation [ eq : mlsevol ] ) , @xmath81,\ ] ] where @xmath82 and @xmath83 are the parameters describing the best fit . under assumptions of normal measurement errors and intrinsic scatter , the quantity @xmath84 is drawn from a chi - square distribution with @xmath85 degrees of freedom .
this likelihood for the measured value @xmath86 can then be used to constrain the model parameter @xmath87 . and
@xmath56 from the rb02 data ( solid lines ) . also shown is the best fit and 68.3 per cent confidence interval ( square with error bars ) on @xmath88 with fixed @xmath56 obtained from the @xcite data ( section [ sec : lensing ] ) .
the cosmology is fixed at our reference .
the dashed lines are the confidence regions that would have been obtained from the rb02 data if the masses were not corrected for bias due to the assumption of hydrostatic equilibrium .
the difference in the precision of the @xmath88 constraints between the corrected and uncorrected rb02 results comes from our marginalization over the uncertainty in the size of the correction . ]
the scatter in mass bias between clusters , @xmath66 , is propagated into the mass error bars , @xmath89 , and thus impacts directly the estimated intrinsic scatter , @xmath86 ( equation [ eq : mls_estimation ] ) .
this results in a degeneracy between @xmath66 and @xmath87 ; however , @xmath66 is not coupled to the other fit parameters , and so we can take its systematic uncertainty into account simply by increasing the width of the likelihood function for @xmath87 .
we therefore replace the chi - square likelihood discussed above by the gamma distribution ( of which the chi - square distribution is a special case ) . fixing the location of the distribution s peak and augmenting its variance by a factor of @xmath90 ,
the shape and scale parameters of the gamma distribution , @xmath91 and @xmath92 , can be determined by solving the system of equations @xmath93 in practice , the value @xmath94 approximately corresponds to the desired 20 per cent uncertainty in @xmath66 .
finally , we account for a 20 per cent systematic uncertainty in the mean bias , @xmath65 , by straightforwardly marginalizing over an appropriate prior . in detail , we parametrize the mean bias through @xmath95 , since this quantity is added to the log - masses , @xmath58 .
confidence regions for @xmath88 and @xmath56 obtained with cosmological parameters fixed at our reference and using uniform priors on @xmath88 , @xmath56 and @xmath87 are displayed in figure [ fig : dahle_rb_conf ] ( solid lines ) .
the best - fitting values and marginal 68.3 per cent confidence intervals are @xmath96 , @xmath97 and @xmath98 .
a quantile - quantile plot of the residuals in @xmath59 for the best fit ( figure [ fig : rb_qqplot ] ) confirms that the distribution of luminosities for a given mass is reasonably approximated by the log - normal distribution .
shows each residual @xmath99 against the quantity @xmath100 , where @xmath101 is the standard normal cdf and @xmath102 is the empirical cdf of the residuals .
if @xmath102 is approximately normal , the points will be distributed roughly along a line .
significant curvature would indicate a poor match in distribution . ] of the rb02 data for the best fitting relation to the normal distribution .
the straightness of the distribution indicates that the scatter is well approximated as normal in @xmath59 ( log - normal in luminosity ) .
( see footnote @xmath103 . )
inset : the histogram of @xmath59 residuals is directly compared with the normal distribution . ] in order to verify the appropriateness of the hydrostatic bias correction applied to the rb02 masses , we compare the results of section [ sec : rbdata ] with the data of @xcite , for which masses were measured using weak gravitational lensing .
these masses , with associated luminosities taken from the bcs and extended bcs ( ebcs ) catalogs , are shown in figure [ fig : dahle_x1_y ] ( black points ) .
in contrast to the rb02 data , the malmquist bias due to the ( e)bcs flux limit is clearly evident ; in particular , the least massive clusters in this data set are very likely to represent the upper tail of the distribution of luminosities for given mass . although the simple methods described above are unsuitable for fitting the mass luminosity relation when the data are subject to significant malmquist bias , the intersection of these two data sets near the highest observed masses indicates qualitatively that they are compatible . ) .
the effect of the ( e)bcs flux limit on the dahle data ( @xmath104 ) is clearly evident . at lower masses ( @xmath58 ) , the data set is progressively more likely to contain only the most luminous clusters for that mass , resulting in an apparent steepening of the relation . ]
we can quantify this observation by using the @xcite data to fit for the normalization , @xmath88 , by minimizing @xmath77 , while leaving the slope , @xmath56 , fixed at the best - fitting value from section [ sec : rbdata ] .
the resulting one - dimensional 68.3 per cent confidence interval is @xmath105 ( figure [ fig : dahle_rb_conf ] ) , in agreement with the corrected rb02 data .
in contrast , the lensing constraint is incompatible at the 68 per cent confidence level with the confidence regions that would have been obtained from the x - ray data without applying the bias correction ( dashed lines in figure [ fig : dahle_rb_conf ] ) .
we use three flux - limited surveys in our analysis : the bcs @xcite and reflex sample @xcite at low redshifts ( @xmath106 ) , and the macs @xcite at @xmath107 .
( we restricted the reflex sample to the southern hemisphere so that its coverage on the sky would not overlap the bcs . ) as discussed in section [ sec : xlflikelihood ] , a proper accounting of the intrinsic scatter in the mass luminosity relation involves convolving over all possible masses when evaluating the likelihood of a set of cosmological parameters ; it is therefore necessary for the mass luminosity relation to be well calibrated at luminosities significantly ( at least an order of magnitude ) below those allowed in the surveys .
we thus also restrict the analysis to clusters with large inferred luminosities @xmath108 .
the completeness of the reflex sample was investigated by @xcite and @xcite , and is thought to be @xmath109 per cent at a flux limit of @xmath110 .
the bcs completeness as a function of flux is quantified in @xcite ; we use a flux limit of @xmath111 where the bcs completeness is 90 per cent .
most of the incompleteness in the bcs is due to the inefficient extended - source detection of the rass ii algorithm , which is most severe at very low redshifts .
one consequence of our high luminosity cut is that the sample contains mostly higher redshift objects ( @xmath112 ) and not the large number of low redshift , low luminosity objects that would be included in a strictly flux - limited sample .
the reported completeness of the bcs is thus a significant underestimate for the sub - sample of very luminous clusters used in our analysis .
we have repeated the analysis of the bcs data alone ( see section [ sec : lcdmresults ] ) using a higher flux limit ( @xmath113 ) where the reported bcs incompleteness is negligible , and obtain a similar result .
we conclude that significant incompleteness is not present in the bcs , given our selection criteria .
the similar numbers of clusters ( 78 and 80 , respectively ) in the bcs and reflex samples satisfying the bcs flux limit and our luminosity cut also support this conclusion .
unlike the bcs and reflex , for which extended - source fluxes were measured using the voronoi tessellation and percolation @xcite and growth curve analysis @xcite algorithms , respectively , reported macs fluxes are measured within a fixed angular size aperture ( 5 arcmin in most cases ) . a redshift - dependent correction for missing flux , described in @xcite ,
is required when converting this aperture flux to total flux , both when determining luminosities for the detected clusters themselves and when computing the aperture flux - dependent survey sky coverage @xcite .
we adopt an aperture flux limit of @xmath114 , for which the completeness is @xmath109 per cent @xcite . using the flux and redshift limits and
the luminosity cut discussed above , the bcs , reflex and macs samples respectively contribute 78 , 130 and 34 clusters to our sample .
the luminosity redshift distribution of these data is displayed in figure [ fig : zlfig ] .
is indicated by the dashed line .
error bars are not shown .
note that even though no clusters above our adopted flux limit were found at redshifts @xmath115 , the lack of any such detections is significant in the analysis . ] in order to adequately take the effects of eddington bias into account when predicting the number of clusters above our flux and luminosity thresholds for a given cosmology and mass luminosity relation , we must be able to assign a probability to the luminosity observed from a cluster , given its true luminosity .
ultimately , the distribution is related to a poisson distribution in the number of photons detected ; however , the nontrivial conversion from photon count rate to unabsorbed flux and the variation in exposure times over the sky make the solution from first principles computationally difficult .
instead , we simplify the problem by assuming that the distribution @xmath116 , where @xmath22 and @xmath21 are the observed and true luminosities , is normal with mean @xmath21 . the variance is a function of the observed flux , which we estimate empirically for each sample by fitting the flux errors versus flux to a power - law model .
we find power - law slopes of 0.52 for the bcs and reflex samples and 0.56 for macs , consistent with approximately poisson scaling .
the primary purpose of this paper is to present an analysis based only on the x - ray luminosity function ( xlf ) data described above , along with the priors described in section [ sec : analysis ] .
however , we also show results obtained from other cosmological data for comparison purposes , along with results of a combined analysis , in section [ sec : results ] . specifically , the additional data consist of observations of galaxy cluster gas mass fractions ( @xmath17 ) , the cosmic microwave background ( cmb ) , and type ia supernovae ( snia ) .
the snia and @xmath17 results shown are identical to those in @xcite .
the @xmath17 data are reported in that work and the snia results are derived from the compilation of @xcite , which includes data from the essence survey ( 60 targets ; * ? ? ?
* ; * ? ? ?
* ) , the snls first year data ( 57 targets ; * ? ? ?
* ) , 45 nearby supernovae @xcite and the 30 high - redshift supernovae discovered by the hubble space telescope and reported by @xcite for which a ` gold ' rating was awarded ( 192 snia in total ) .
our analysis of the cosmic microwave background anisotropies uses three - year wilkinson microwave anisotropy probe ( wmap ) data , including marginalization over a plausible range in the amplitude of the sunyaev - zeldovich signal in galaxy clusters ( @xmath117 ) @xcite .
we use the october 2006 version of the wmap likelihood code . when analysing the cmb data , as with the xlf analysis , we include the effects of density perturbations in the dark - energy fluid . for our standard analysis
, we fix the dark - energy sound speed in units of the speed of light , @xmath118 , to 1.0 .
( some discussion of the influence of perturbations can be found in section [ sec : soundspeed ] . ) in addition , no prior on the scalar spectral index , @xmath119 , is used in the cmb analysis , since wmap places very strong constraints on this parameter ( @xmath120 ; @xcite ) ; as noted in section [ sec : priors ] , this degree of variation in @xmath119 produces negligible change in the xlf results .
we parametrize the full model fitted to the x - ray luminosity function data as @xmath121 , @xmath122 , @xmath123 , @xmath1 , @xmath119 , @xmath2 , @xmath124 , @xmath88 , @xmath56 , @xmath125 , @xmath126 , @xmath127 , @xmath128 , where @xmath129 and @xmath130 are the baryon and cold dark matter densities ( @xmath131 ) , @xmath124 is the normalization of the jenkins mass function ( equation [ eq : jmf ] ) , and the last 6 parameters describe the mass luminosity relation and its calibration .
in addition to the assumption of spatial flatness , we adopt the gaussian priors @xmath132 @xcite and @xmath133 @xcite from the hubble key project and big bang nucleosynthesis studies , respectively .
these latter priors are necessary because the likelihood depends very weakly on @xmath134 and @xmath135 , which enter only through the transfer function .
we additionally marginalize over a 20 per cent uncertainty in the normalization , @xmath124 , of the j01 mass function to account for the residuals of the fitting formula to their simulations over the mass range of interest and the expected variation among cosmologies .
we must also place a prior on the mass luminosity evolution parameters @xmath125 and @xmath136 , which are not constrained by our analysis of the rb02 data .
for the standard set of allowances used in this paper , we adopt the uniform priors @xmath137 , which corresponds to a limit of approximately 20 per cent evolution in the normalization of the mass luminosity relation within redshift 0.7 , and @xmath138 .
since the results are insensitive to the spectral index within a reasonable range ( section [ sec : priors ] ) , we fix @xmath139 in accordance with @xcite for the standard analysis .
the dark - energy equation of state was bounded by a uniform prior , @xmath140 . for the remaining model parameters , uniform priors on the physically allowed domains
were used .
these priors are summarized in table [ tab : priors ] ( labeled `` standard '' priors ) .
the sensitivity of our results to the choice of priors is analyzed in section [ sec : priors ] .
.priors used in the analysis .
all parameters not listed were assigned uniform priors on their physically allowed domains .
@xmath141 indicates the normal distribution with mean @xmath142 and variance @xmath143 , and @xmath144 indicates the uniform distribution with endpoints @xmath145 and @xmath146 .
[ tab : priors ] [ cols="<,^,<",options="header " , ] in order to assess the goodness of fit for the xlf data , we compare the number of clusters in each sample with the number predicted by the best - fitting set of parameters in the markov chains .
since the chains were generated at finite temperature , these do not necessarily correspond to the posterior modes , but they should be nearby . using the xlf data only , the best - fitting @xmath16cdm parameters predict 73 , 127 and 37 clusters in bcs , reflex and macs , respectively , while the best fitting @xmath2cdm parameters predict 76 , 130 and 35 .
both are in good agreement with the true samples , which respectively contain 78 , 130 and 34 clusters above flux and luminosity thresholds of section [ sec : xlf ] .
this agreement for the @xmath16cdm case is displayed graphically as a function of the flux limit in figure [ fig : goodness ] .
( the numbers quoted above correspond to the low - flux ends of the lines in the figure . )
the equivalent figure for @xmath2cdm looks similar .
above detect flux @xmath147 as a function of @xmath147 ( @xmath148 distribution ) for bcs ( dotted line , above / right ) , reflex ( solid line , above / right ) and macs ( solid line , below / left ) . shaded areas indicate the predictions and 68 per cent confidence regions of the best - fitting @xmath16cdm model .
due to the similar sky areas covered , the bcs and reflex predictions are essentially identical at @xmath149 . ] and @xmath1 for a constant-@xmath2 model using standard priors ( solid , black lines ) and weak priors on @xmath134 and @xmath122 ( dashed , blue lines ) . weakening these priors results in a slight expansion of the confidence region along the @xmath0-@xmath1 degeneracy axis .
the priors have no affect on the determination of @xmath2 . ]
the standard priors used in our analysis are conservative ; however , we have also performed analyses using the even more conservative priors labeled as `` weak '' in table [ tab : priors ] .
these weak - prior results effectively demonstrate the degeneracies between parameters of interest ( @xmath0 , @xmath1 and @xmath2 ) and each nuisance parameter , given the current level of statistical and systematic uncertainty .
two of these weak priors , on @xmath136 and @xmath150 , did not result in significant changes to any of the cosmological constraints .
the results of the other tests are summarized below .
one - dimensional , marginalized results for each of these analyses are also presented in table [ tab : constraints ] .
we consider @xmath134 and @xmath122 together , since both influence the results only through the shape of the power spectrum .
the results obtained by doubling the width of the gaussian priors on these parameters ( blue , dashed contours ) is compared with our standard - prior constraint in figure [ fig : priors_hwb_s8 ] .
the effect of weakening these priors is primarily to expand the constraints along the @xmath0-@xmath1 degeneracy curve .
the constraint on @xmath2 is not affected .
we next consider the effect of marginalizing over @xmath119 using a gaussian prior of width 0.05 , significantly larger than the constraint obtained from wmap @xcite .
like @xmath134 and @xmath122 , the effect of freeing this parameter is to weaken the constraints along the @xmath0-@xmath1 degeneracy , although the effect is relatively small , even with this wide prior ( left panel of figure [ fig : fwcdm_compare ] ) .
the weak prior on @xmath124 doubles the uncertainty on the j01 normalization to 40 per cent .
the result is a decrease in constraining power on @xmath0 and @xmath2 , but not significantly on @xmath1 .
the expansion of the dark - energy constraints can be seen in figure [ fig : priors_a_w ] . and @xmath2 for a constant-@xmath2 model using standard priors ( solid , black lines ) and a weak prior on the normalization of the mass function , @xmath124 ( dashed , blue lines ) .
the constraints on @xmath0 and @xmath2 are weakened , although @xmath1 is not greatly affected ( table [ tab : constraints ] ) . ] since @xmath125 describes only the redshift evolution of the mass luminosity relation , doubling the size of its uniform prior has no effect on the determination of the redshift - zero quantities @xmath0 and @xmath1 , as can be seen in table [ tab : constraints ] .
however , there is a degeneracy with @xmath2 , resulting in a poorer constraint ( table [ tab : constraints ] , prior 5 ) .
the presence of active galactic nuclei ( agn ) and other point - like x - ray emitters is a potential concern for the accuracy of the mass luminosity relation .
there is no possibility of subtracting these point sources from the rass data , since the number of photons in a cluster detection is typically too small .
however , they are subtracted from the rb02 luminosities , to the extent that _ rosat _ can resolve them .
in addition , there is the possibility of an increase in the density of agn over the redshift range of our data ( e.g. @xcite and references therein ) .
there may therefore be an inequivalence between luminosities inferred from the rass data and those used in the calibration of the mass luminosity relation . in our work studying macs clusters ( in preparation ) , we take advantage of the _ chandra _ x - ray observatory s high spatial resolution to efficiently identify point sources and determine their contribution to the total cluster luminosity .
we find that this contribution is at the few per cent level much smaller than the typical survey flux uncertainty and conclude that point source contamination is not an issue for the current study .
the bias in the measured masses of the mass luminosity data is completely degenerate with the mass function normalization , @xmath124 , which in turn affects inferences made on the other parameters .
consequently , it is important to include in the analysis the mean correction to masses motivated by departures from hydrostatic equilibrium , @xmath65 , and the scatter in this bias among clusters , @xmath66 , as well as the uncertainty in both quantities .
figure [ fig : massbias ] shows that the constraints obtained by fixing the mean bias and bias scatter at their nominal values ( blue , dashed contours ) are spuriously tighter than those of our standard analysis ( black , solid contours ) by @xmath151 per cent on @xmath0 and @xmath1 and @xmath152 per cent on @xmath2 . also shown is the result of neglecting the bias correction entirely ( red , solid contours ) , which results in lower values of @xmath1 compared with the nominal correction .
comparing the results for priors 7 and 8 in table [ tab : constraints ] with our standard - prior results , it is clear that the uncertainty in @xmath66 has a greater effect than the uncertainty in @xmath65 .
this is encouraging , since the former is due primarily to the very small number of simulations used to calibrate the observational mass bias . straightforwardly increasing
the number of simulated clusters should result in significant improvement to our constraints .
the mass distribution of clusters available to a flux - limited survey is influenced by the degree of scatter in flux for a given mass . as described in section [ sec : xlflikelihood ] , this effect can be decomposed into convolutions due to intrinsic scatter in the mass luminosity relation and measurement error in the survey flux determinations .
failure to account for either of these sources of scatter when evaluating the number of detectable clusters predicted by a set of model parameters can significantly bias the result .
the magnitude of the effect is demonstrated in figure [ fig : mlfscat ] , which shows the results for a @xmath2cdm cosmology obtained by ignoring either of these scatters individually . in each case , the results are biased towards lower @xmath0 and higher @xmath1 , along the degeneracy between the two parameters .
the constraints on cosmological parameters , including @xmath2 , are also spuriously tight in both cases .
( the joint @xmath0-@xmath2 constraints are not shown , but closely resemble the tighter results in figure [ fig : massbias ] . )
this results from the fact that any scatter in luminosity compounds the systematic effect of uncertainty in @xmath65 . in the case where intrinsic mass luminosity dispersion is ignored ,
the contributions due to uncertainty in @xmath87 , @xmath136 and @xmath66 also vanish from the error budget . in the analysis
thus far , we have accounted for dark - energy density perturbations assuming that the sound speed is equal to the speed of light ( @xmath153 ) .
table [ tab : constraints ] ( prior 12 ) shows that marginalizing over such a constant sound speed ( @xmath154 ) has no effect on our results ( but note that dark - energy models generally predict a time - varying sound speed ) .
@xcite , @xcite and @xcite showed that including such perturbations greatly reduces the ability of wmap data to discriminate among models with @xmath155 ( figure [ fig : fwcdm_compare ] , _ cf .
_ figures 15 and 16 of spergel et al . ) .
in contrast , the effect of dark - energy perturbations on the xlf constraints is less dramatic ; table [ tab : constraints ] ( prior 13 ) shows that an analysis with no perturbations results in somewhat larger constraints on @xmath2 , opposite in character from the wmap analysis .
we have presented new constraints on cosmological - constant ( @xmath16cdm ) and constant-@xmath2 ( @xmath2cdm ) dark - energy models using the observed x - ray luminosity function of the largest , most x - ray luminous galaxy clusters out to redshift 0.7 , in combination with standard priors on @xmath134 and @xmath122 . at 68.3 per cent confidence , we find @xmath4 and @xmath5 for a @xmath16cdm model , and @xmath6 , @xmath7 and @xmath8 for a @xmath2cdm model .
these results include marginalization over uncertainties in the theoretical mass function , non - self - similar evolution in the mass luminosity relation , and the bias in mass estimates from x - ray observations .
our results constitute the first precise determination of the dark - energy equation of state using measurements of the growth of cosmic structure observed in galaxy clusters , and provide additional support for the cosmological - constant model .
these results build upon , and are largely in agreement with , a number of earlier galaxy cluster studies ( see section [ sec : introduction ] ) .
our constraints on cosmological parameters are consistent with independent findings from studies of type ia supernovae , anisotropies in the cmb , the x - ray gas mass fraction of galaxy clusters , galaxy redshift surveys and leading cosmic shear surveys .
the agreement between the results from these independent techniques is reassuring , and motivates a combined analysis of the data in order to investigate more complex models of dark energy . combining our data with cmb , snia and @xmath17 data ,
we find @xmath10 , @xmath11 and @xmath12 . the results for @xmath1 presented here are somewhat higher than those from previous work based on the bcs and reflex data due to our correction to the masses used to constrain the mass luminosity relation .
the magnitude of this discrepancy underscores the need for an improved understanding of the observational biases resulting from asphericity , projection effects and hydrostatic disequilibrium .
more advanced and comprehensive simulations , calibrated by gravitational lensing studies , show considerable promise in this area .
broad - band , high spectral resolution x - ray data from , for example , the new x - ray telescope ( next ) and constellation - x , will allow precise measurements of gas velocities and non - thermal emission components in clusters , providing a more comprehensive understanding of the relevant gas physics . in the short term , simply increasing the number of simulated clusters used to calibrate the observational bias would reduce the uncertainty in the scatter in bias among clusters . since the uncertainty in this scatter is the dominant source of systematic uncertainty in @xmath1 , this approach provides a relatively easy means of improving the precision of our results . with regard to dark energy , the systematics due to the theoretical mass function and the redshift evolution of the mass observable relation are also important .
the former can be addressed with a large suite of cosmological simulations , allowing a more precise parametrization of the cosmological dependence of the mass function .
the latter necessitates a more rigorous study of galaxy cluster virial relations and their evolution , carefully accounting for selection effects .
such improved analysis techniques will be required if future , high - redshift x - ray ( e.g. spectrum - rg / erosita ) or sunyaev - zeldovich surveys are to be used to their full potential .
we thank adrian jenkins for providing computer code to evaluate the mass function of dark matter halos and gil holder , alexey vikhlinin and jeremy tinker for helpful discussions .
we also thank glenn morris and stuart marshall for computer support .
calculations for this work were carried out using the kipac xoc and orange compute clusters at the stanford linear accelerator center ( slac ) and the slac unix compute farm .
we acknowledge support from the national aeronautics and space administration ( nasa ) through nasa ltsa grant nag5 - 8253 , and though chandra award numbers dd5 - 6031x , go2 - 3168x , go2 - 3157x , go3 - 4164x , go3 - 4157x and go5 - 6133 , issued by the chandra x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of nasa under contract nas8 - 03060 . this work was supported in part by the u.s .
department of energy under contract number de - ac02 - 76sf00515 . am was additionally supported in part by a william r. and sara hart kimball stanford graduate fellowship . while this work was in revision , the 5-year wmap results were released ( @xcite and references therein ) .
the new wmap constraints are also in excellent agreement with our results .
since the difference between the 3-year and 5-year wmap results is only marginal , we have not repeated our combined analysis with the newer cmb data in this work . | we present constraints on the mean matter density , @xmath0 , the normalization of the density fluctuation power spectrum , @xmath1 , and the dark - energy equation - of - state parameter , @xmath2 , obtained from measurements of the x - ray luminosity function of the largest known galaxy clusters at redshifts @xmath3 , as compiled in the massive cluster survey ( macs ) and the local bcs and reflex galaxy cluster samples .
our analysis employs an observed mass luminosity relation , calibrated by hydrodynamical simulations , including corrections for non - thermal pressure support and accounting for the presence of intrinsic scatter .
conservative allowances for all known systematic uncertainties are included , as are standard priors on the hubble constant and mean baryon density .
we find @xmath4 and @xmath5 for a spatially flat , cosmological - constant model , and @xmath6 , @xmath7 and @xmath8 for a flat , constant-@xmath2 model ( marginalized 68 per cent confidence intervals ) .
our findings constitute the first determination of the dark - energy equation of state from measurements of the growth of cosmic structure in galaxy clusters , and the consistency of our result with @xmath9 lends additional support to the cosmological - constant model .
future work improving our understanding of redshift evolution and observational biases affecting the mass x - ray luminosity relation have the potential to significantly tighten these constraints .
our results are consistent with those from recent analyses of type ia supernovae , cosmic microwave background anisotropies , the x - ray gas mass fraction of relaxed galaxy clusters , baryon acoustic oscillations and cosmic shear . combining the new x - ray luminosity function data with current supernova ,
cosmic microwave background and cluster gas fraction data yields the improved constraints @xmath10 , @xmath11 and @xmath12 .
[ firstpage ] cosmological parameters large - scale structure of universe x - rays : galaxies : clusters . |
recently universality of lepton flavor in the decay of @xmath15 mesons was questioned by the lhcb group@xcite and conjectures based on the qcd with additional currents were given in @xcite .
in @xcite , violation of the flavor symmetry due to admixture of @xmath11 quark contribution to the @xmath9 quark contribution was studied , and we want to extend the theory to the decay of @xmath15 mesons . decay branching ratios of mesons to lepton pairs or a pion and lepton pairs are good testing ground of the standard model ( sm ) .
comparison of experimental decay branching ratios of @xmath18 , @xmath19 , @xmath20 and @xmath21 with that of sm were discussed in @xcite . in the sm the strength of transitions of a quark to different flavor states
is defined by the cabibbo - kobayashi - maskawa ( ckm ) matrices .
for example , the branching ratio of @xmath22 depends on the probability of the coupling of an @xmath12 quark + emitted vector particle to a @xmath23 quark + absorbing vector particle which depends on @xmath24 , and the branching ratio of @xmath25 depends on the probability of the coupling of a @xmath26 quark + emitted vector particle to a @xmath23 quark + absorbing vector particle which depends on @xmath27
. comparisons of theoretical meson decay matrix elements and experiments were done in various processes , and recently a summary of lattice simulation results was given by the flag group @xcite .
but the lattice can not give consistent branching ratios of @xmath25 and @xmath28 mesons together .
violation of @xmath29 symmetry or time reversal symmetry in the decay of @xmath15 mesons was studied in the difference of @xmath30 and @xmath29 violation was observed in @xmath31 decay but not in @xmath28 decay .
qualitative differences of @xmath32 and @xmath33 were also observed . since complete non leading order corrections in @xmath34 and @xmath35 transitions are not available@xcite , we apply cartan s octonion theory to solve the problem of branching ratios of @xmath15 mesons .
in @xcite we studied cartan s supersymmetry@xcite and studied coupling of fermions and vector particles which transform under groups @xmath36 and @xmath37 .
cartan considered only electromagnetic interactions , but extension to weak interaction is possible@xcite .
we applied the cartan s supersymmetric model to @xmath38 and found modification of coupling occurs in a certain channel as compared to the sm@xcite . in the section 2 ,
we summarize the status of meson decay branching ratios , and in the section 3 we explain the method of analyzing the meson decay using cartan s supersymmetry .
discussion and conclusion are given in section 4 .
when one uses results of @xmath39 meson decay to @xmath40 or @xmath41 states to analyses of @xmath25 meson or @xmath28 decay to @xmath42 or @xmath41 states , one encounters problems . from the universality of leptons ,
the strength of @xmath45 and that of @xmath46 are expected to be the same , but experimentally @xmath47 is observed by the lhcb group@xcite .
the inclusive @xmath48 decay modes were studied by several authors and an extension of randall and sundrum model@xcite to @xmath49 transition was tried in @xcite . in the custodially protected randall sundrum model ( rs@xmath50 ) ,
additional @xmath51 current in dark sector was proposed to solve @xmath52 anomalies @xcite .
we do not assume this current , but adopt the vertex with @xmath53 factor for the weak interaction , and consider two triangle diagrams in the coupling .
( left ) and @xmath54 ( right).,width=188 ] ( left ) and @xmath54 ( right).,width=188 ] the @xmath48 vertex requires renormalization of penguin diagrams .
( left ) and @xmath54 ( right).,width=188 ] ( left ) and @xmath54 ( right).,width=188 ] the lepton pairs can be @xmath16 which transform mainly to @xmath55 or @xmath56 which transform mainly to @xmath57 .
the ckm model implies higher order effect of coupling of @xmath55 to @xmath58 quark is dominated by the loop of @xmath9 quarks , but difficulty of the ckm model to explain the decay of @xmath59 transitions implies contamination of the loop of @xmath60 quarks in addition to @xmath61 quarks , since @xmath12 and @xmath11 quark belong to the same ckm sector .
( left ) , and @xmath62 ( right).,width=188 ] ( left ) , and @xmath62 ( right).,width=188 ] the decay of @xmath64 ( @xmath65 and @xmath66 ) in the standard model contains two @xmath67 verices and the intermediate state is @xmath57 or @xmath68 , as shown in fig.[bbarsd ] .
( left ) , and @xmath63 ( right).,width=226 ] ( left ) , and @xmath63 ( right).,width=226 ] they defined the lagrangian for @xmath69 as @xmath70 and averaged time - integrated branching ratios proportional to @xmath71 where @xmath72 is the lifetime of the heavior mass eigenstate of @xmath15@xcite , @xmath73 and @xmath74 are ckm matrix elements , was calculated for @xmath75 , and @xmath76 where @xmath77 is the average of the lifetime of the lighter mass and heavior mass eigenstate of @xmath15 , @xmath78 and @xmath79 are ckm matrix elements , was calculated for @xmath80 . in the sm ,
the decay branching ratio of @xmath25 and @xmath28 yields direct information on @xmath81 and @xmath82 , but uncertainty in the ckm matrix element @xmath82 was not seriously taken into account .
blanke et al.@xcite presented the review of @xmath83 , @xmath49 and @xmath84 . according to their work experimental world average branching ratio @xmath85 is slightly smaller than the theoretical value @xmath86@xcite . in the case of @xmath87
, there are larger experimental uncertainty@xcite , however the experimental world average branching ratio @xmath88 is much larger than the theoretical value @xmath89@xcite .
a two loop calculation of branching ratio including the diagram of the penguin diagram was done by @xcite for the @xmath90 .
if a neutrino is a majorana neutrino and @xmath91 and @xmath92 are indistinguishable , one can consider the decay through the box diagram@xcite .
but there is no clear experimental results that suggests @xmath93 .
if neutrino is massive , it is difficult to identify a neutrino and an antineutrino .
we expect the decay of @xmath90 and @xmath94 are not dominated by fig.[bbarsd ] , we study @xmath11 quark contribution is important . in the case of @xmath90
we consider the @xmath95 triangle diagram with a penguin loop studied as @xcite and the @xmath96 triangle diagram .
we do not assume that the penguin loop is given by a gluon but by a @xmath57 boson .
a gluon has unphysical components and it can be replaced by a gauge potential and ghost fields@xcite .
since the @xmath12 quark and the @xmath11 quark are in the same ckm sector , we expect in the @xmath95 triangle , some contribution of @xmath97 quark to the usual @xmath98 quark contribution . since a @xmath11 quark is heavior than a @xmath12 quark , the contribution is less important than @xmath63 or @xmath33 .
in @xcite we analyzed @xmath99 decay process using cartan s supersymmetry and compared the results with sm using ckm matrix elements .
we perform the similar analysis for @xmath39 meson and @xmath15 meson decays into lepton pairs . in the previous analysis of @xmath99 ,
a vector particle was emitted from a quark or an anti - quark in a meson and transformed to @xmath100 , but in the present case we consider processes in which a vector particle transform to @xmath41 .
the coupling of the vector particle which was the source of @xmath100 was pseudo scalar type @xmath101 the agreement of branching ratios of meson decays in theoretical models and the sm is not sufficient .
one could imagine new physics , in which heavy mesons @xmath102 instead of @xmath103 contribute@xcite .
however , it may be appropriate to study supersymmetry of quarks or leptons and vector particles more in detail .
in the sm , the simplest amplitudes of @xmath104 contains two @xmath67 vertices , since electromagnetic decay is impossible . in the cartan s supersymmetric model , @xmath105 and @xmath57 bosons
are replaced by vector particles @xmath106 and @xmath107 , and quarks and leptons are replaced by dirac spinors @xmath108 and @xmath109 .
incorporation of @xmath110 quarks allows decay amplitudes containing only one @xmath67 vertex from a part of @xmath111 .
simple diagrams of @xmath113 containing the vertex corresponding to that of @xmath114 are shown in fig.[sbardlla ] . in our diagram of @xmath83 shown in the left side of fig.[sbardlla ] , the symbol @xmath115 corresponds to @xmath57 boson with the momentum direction @xmath115 , the symbol @xmath116 corresponds to @xmath117 with the momentum direction @xmath116 , and the symbol @xmath107 corresponds to @xmath118 with the momentum direction @xmath107 .
the vector particle @xmath116 couple with @xmath119 of @xmath120 and create @xmath121 of anti - quark @xmath110 and the vector particle @xmath107 is emitted from @xmath122 of @xmath123 and create the @xmath119 . in order to make coupling of a quark to @xmath117 and @xmath118 bosons ,
it is necessary to include a @xmath67 vertex which couples @xmath110 and @xmath124 or @xmath123 and @xmath120 .
a quark emitted as @xmath125 can be absorbed as @xmath126 at the @xmath67 vertex .
quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] we extend the cartan s supersymmetric model to allow a spinor @xmath129 and an antispinor @xmath130 annihilate to a scalar particle @xmath131 . including the vertex @xmath132 vertices
, we can consider penguin diagrams of @xcite , which are shown in the right hand side of fig.[sbardlla ] . in the diagrams there are triangle diagrams of @xmath133 boson @xmath134 ) and a loop of @xmath57 boson @xmath135 . due to this loop , coupling of a quark to the @xmath57 boson(@xmath107 ) becomes possible , without including a @xmath67 vertex .
they appear as a correction from weak interactions . in our model , we allow transformation of @xmath136 to @xmath137 during propagations , and since @xmath138 consists of @xmath139 , and @xmath140 quark and @xmath141 quark are in the same sector of ckm model , the @xmath142 on the triangle is assumed to be @xmath143 .
in other words , we take into account the correction to the @xmath9 quark loop dominance in the sm . in the calculation of the right hand side of fig.[sbardlla ] , we fix the @xmath26 quark and the @xmath140 quark , both large components and exchange a vector particle between @xmath26 and @xmath140 . in the first diagram , the @xmath26 quark @xmath144 becomes after emitting a vector particle @xmath145 becomes a @xmath146 quark or a @xmath141 quark @xmath147 , and after emitting a particle @xmath148 corresponding to a @xmath57 boson becomes a quark @xmath149 and after emitting @xmath107 , becomes an anti - quark @xmath150 .
the @xmath140 quark @xmath121 emits a @xmath116 and becomes a quark @xmath151 .
the @xmath152 state of @xmath153 has an overlap with the vector particle @xmath154 and absorbs the previously emitted @xmath155 .
the @xmath57boson @xmath107 becomes an @xmath156 pair . by admitting an overlap of @xmath157 and @xmath158 state ,
@xmath159 transition becomes possible .
the contribution of @xmath11 or @xmath141 quark is expected to be not so important since @xmath160 and virtual transition to @xmath9 whose mass @xmath161 is not expected to be affected . in the case of @xmath25
we replace the @xmath140 quark to @xmath23 quark . in the case of @xmath28
we replace the @xmath26 quark in @xmath25 to an @xmath12 quark and @xmath162 quark to @xmath163 quark , since a @xmath11 quark is in the same ckm sector as an @xmath12 quark , and @xmath164 , we need to incorporate the effect of a @xmath11 quark .
the quark diagrams of @xmath63 processes are given in fig.[bbardlla ] and that of @xmath33 processes are given in fig.[bbarslla ] .
quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] quark diagrams containing @xmath127 ( @xmath128 ) vertices with different @xmath67 positions ( left ) , and @xmath95 triangle diagrams with different penguin loop positions ( right).,width=226 ] the qualitative difference of experimental and theoretical branching ratios of @xmath165 and @xmath166 may originate from the difference of quarks propagating on the triangle of two loop diagrams between @xmath12 and @xmath23 and @xmath26 and @xmath23 . in the case of @xmath28 , since the @xmath12 quark belongs to the same sector as the @xmath11 quark , after emitting a vector particle @xmath167 , it can be transformed to an antiquark @xmath141 and emits a vector particle @xmath107 which changes to @xmath41 , returns to a quark @xmath11 which emits a vector particle @xmath168 .
the @xmath11 quark absorbs the emitted vector particles @xmath167 and @xmath107 and transforms to a @xmath23 quark . in the case of @xmath25 ,
the @xmath26 quark is transformed to @xmath146 quark after emitting a vector particle @xmath169 as shown in the right - hand part of diagrams .
the product @xmath170 of ckm matrices gives the strength , and the @xmath146 quark emits an @xmath168 and transforms to a @xmath23 quark after absorbing the emitted @xmath107 on the @xmath67 vertex . the relative intensity of the left - hand side diagrams and the right - hand side diagrams of figs.[sbardlla ] , [ bbardlla ] , and fig.[bbarslla ] are not evident .
we expect the left - hand side dominates , and reflects the ckm matrix elements , however in @xmath22 the right - hand side diagram containing a @xmath11 quark propagation gives a slight suppression to the left - hand side dominant term , and in @xmath171 the right - hand side diagram containing a @xmath9 quark propagation is expected to give an enhancement to the dominant term .
description of dirac fermions is not unique .
the sub algebra @xmath172 with a basis @xmath173 where @xmath174 satisfy @xmath175 is equivalent to pauli algebra , defined by @xmath176 and @xmath177 ckm matrices were derived based on this algebra@xcite in the @xmath178 dimensional linear space @xmath179 over a field @xmath180 and exterior algebra @xmath181 , octonion appears by defining @xmath182 , and choosing @xmath183 as the basis of the field@xcite .
the commutation relations of @xmath174 are not same as those of cartan s @xmath184
. the rules of multiplication of @xmath174 follow clifford algebra@xcite .
lounesto@xcite pointed out that a product of clifford numbers @xmath185 and @xmath186 in @xmath178 dimensional linear space @xmath179 over a field @xmath180 defined by chevalley@xcite is @xmath187 where @xmath188 is the antisymmetric product , and @xmath189 is the contraction which depends on @xmath180 .
the time(space ) component of @xmath194 and that of @xmath195 satisfy the commutation relation @xmath196 similarly , those of @xmath197 and @xmath198 satisfy the commutation relation @xmath199 due to this difference of commutation relation of the space components and time components of @xmath123 and @xmath110 , we can define unique octonion in @xmath200 space .
we applied cartan s supersymmetry to analyze the decay branching ratios of @xmath39 and @xmath15 mesons to @xmath40 and @xmath41 . in the @xmath201 process , we assigned vector particles @xmath136 to @xmath57 boson of polarization @xmath136 and by allowing transition of @xmath120 and @xmath124 during the propagation of @xmath202 , we could make @xmath203 composed of large components @xmath123 or @xmath110
. the suppression of the @xmath204 can be explained , if one takes into account that the @xmath11 quark and @xmath12 quark are in the same sector , and correction to the ckm mechanism is large in @xmath204 than in @xmath205 .
when @xmath12 or @xmath11 quarks are not present in the meson system , the vertex of vector particles @xmath206(@xmath207 ) and the triangular loop of top quarks and one vector particle @xmath208 defines the decay branching ratios .
when @xmath12 or @xmath11 quarks are in the meson , the loop of @xmath209 modifies the @xmath208 contribution . in the case of @xmath205 , since @xmath26 quark is not in the sector of @xmath11 , only the loop of @xmath208 contribute , and if it adds the contribution of the one loop of @xmath206(@xmath207 ) , we can explain the enhancement of @xmath205 . in the processes of @xmath204 via exchange of majorana neutrino@xcite ,
one allows a transition from @xmath91 to @xmath92 . in our model ,
@xmath210 and their charge conjugates make sectors , and a neutrino in the sector of @xmath211 can transform to @xmath212 or @xmath213 in the notation of @xcite , which can be interpreted as @xmath214 in different triality sectors , which would not be detected by vector particles or leptons in the sector of the initial @xmath215 , since for detection of a particle @xmath216 in the sector @xmath217 , corresponding necessary condition @xmath218 also need to be transferred @xcite from different sectors , and we may be allowed to ignore the @xmath219 transition .
if our electromagnetic detector can detect electromagnetic field transformed by @xmath220 but can not detect electromagnetic field transformed by other four transformations , and there are six lepton sectors @xmath221 which can not be detected by our detectors , the presence of dark matter can be explained . we observed that starting from the meson wave function given by the large component of a quark and the large component of an anti - quark , qualitative features of the strength of the decay branching ratios can be obtained from cartan s supersymmetry .
cartan s octonion satisfies a triality automorphism that is supersymmetric , which octonions of @xcite do nt satisfy . in our system of fermions and vector fields with interaction @xmath222 , the operator @xmath67 yields coupling between a large component of a quark and a large component of an antiquark which enhances certain @xmath15 meson decay processes .
it makes the branching ratio of @xmath15 meson decays different from the ckm model .
if in the univese there are world which are transformed by @xmath223 and @xmath224 , and our electromagnetic detector can detect the world transformed by @xmath225 and @xmath37 , and the uncertainty principle applys not only in our world but also whole universe , we can understand the presence of dark matter .
departure from the standard model of @xmath226 and @xmath227 is due to the large correction from the @xmath11 quark in the decay of @xmath28 meson .
we ignored @xmath228 and @xmath229 etc , whose information will serve for establishing the model of @xmath15 mesons based on cartan s supersymmetry . 99 r. aaij et al .
( lhcb collaboration ) , measurement of form - factor independent observables in the decay @xmath230 , , 19,191801 ( 2013 ) .
arxiv:1308,1707[hep - ex ] . c. bobeth
, g. hiller and d. van dyk , general analysis of @xmath231 decays at low recoil , , 034016 ( 2013 ) .
c.bobeth et al . , @xmath232 in the standard model and reduced theoretical uncertainty , , 101801(2014 ) .
r. alonso , b. grinstein and j.m .
camalich , @xmath233 gauge invariance and the shape of new physics in rare @xmath15 decays , , 241802 ( 2014 ) .
d. a. sierra , f. staub and a. vicente , shedding light on the @xmath52 anomalies with a dark sector , arxiv : 1503.0607 v1(2015 ) .
s. furui , cartan s supersymmetry and weak and electromagnetic interactions , few body syst . * 56 * , 703(2015 ) , doi 10.1007/s00601 - 015 - 1016 - 6 , arxiv:1502.04524 [ hep - ph ] .
g. buchalla and a.j .
buras , the rare decays of @xmath234 and @xmath235 beyond leading logarithms , , 106(1994 ) .
m. blanke , a. j. buras , d.guadagnoli and c. tarantino , minimal flavour violation waiting for precise measurements of @xmath236 , @xmath237 and @xmath238 , arxiv : hep - ph/0604057v5 .
j. brod , m. gorbahn and e. stamou , two - loop electroweak corrections for the @xmath228 decays , , 034030 ( 2011 ) ; arxiv:1009.0947v2 [ hep - ph ] . m. gorbahn and u. haisch , charm - quark contribution to @xmath90 at next - to - next - to - leading order , arxiv hep - ph/0605203 v3 .
t. inami and c.s .
lim , , 297 ( 1981 ) .
becchi and g. ridolfi , _ an introduction to relativistic processes and the standard model of electroweak interactions _
springer ( 2014 ) .
buras , j. girrbach - noe , c. niehoff and d.m .
straub , @xmath239 decays in the standard model and beyond , arxiv:1409.4557 c. bobeth , m. gorbahn , t. hermann , m. misiak , e. stamou and m. steinhauser , @xmath64 in the standard model with reduced theoretical uncertainty , , 101801(2014 ) .
v. khachatryan et al .
( cms and lhcb collaborations ) , observation of the rare @xmath204 decay from the combined analysis of cms and lhcb data , arxiv:1411.4413 .
buras , d. buttazzo , j. girrbach - noe and r. knegjens , @xmath20 and @xmath240 in the standard model : status and perspectives , arxiv:1503.02693 .. y. amhis et al .
( heavy flavour averaging group ) , arxiv:1207.1158 , updates at http://www.slac.stanford.edu/xorg/hfag .
m. blanke , a.j .
buras and s. recksiegel , quark flavour observables in the littlest higgs model with t - parity after lhc run 1 , arxiv:1507.06316 v1 [ hep - ph ] .
f.beayjean , ch .
bobeth and s. jahn , constraints on tensor and scalar couplings from @xmath241 and @xmath242 , arxiv.1508.01526 v1[hep - ph ] .
s. aoki et al .
( flavour lattice averaging group ) , review of lattice results concerning low energy particle physics , arxiv:1310.8555 .
s. furui , fermion flavors in quaternion basis and infrared qcd , few body syst . * 52 * , 171(2012 ) ; .cartan , _ the theory of spinors _ , dover pub .
( 1966 ) p.118 .
becchhi and g. ridolfi , _ an introduction to relativistic processes and the standard model of electroweak interactions _ , 2nd edition , springer ( 2014 ) .
s. furui , cartan s supersymmetry and the violation of @xmath29 symmetry , arxiv:1505.05830 [ physics.gen-ph ] .
s. furui , cartan s supersymmetry and the decay of @xmath243 gev@xmath244 to @xmath245 and @xmath246 , pos , contribution to _ flavor physics @xmath247 violation _
, nagoya ( 2015 ) .
l. randall and r. sundrum , a large mass hierarchy from a small extra dimension , , 3370(1999 ) ; arxiv : hep - ph/9905221 .
l. randall and r. sundrum , an alternative to compactification , , 4690(1999 ) ; arxiv : hep - th/9906094 .
p. biancofiore , p. colangelo and f. de fazio , rare semileptonic @xmath248 decays in @xmath249 model , arxiv:1403.2944v2 ( 2014 ) .
p. biancofiore , p. colangelo , f. de fazio and e. scrimieri , exclusive @xmath250 induced transitions in @xmath249 model , arxiv:1408.5614v1 ( 2014 ) .
m. bitbol , the concept of measurement and time symmetry in quantum mechanics , philosophy of science * 55 * , 349 ( 1988 ) .
g. aragn - camarasa et al . , clifford algebra with mathematica , j. comput .
appl . math .
t. dray and c.a .
manogue , finding octonionic eigenvectors using mathematica , computer physics communications ( 1998 ) t. dray and c.a .
manogue , the octonionic eigenvalue problem , advances in applied clifford algebras * 8 * 341 ( 1998 ) .
i. r. porteous , clifford algebra and the classical groups .
, cambridge university press ( 1995 ) .
p. lounesto , clifford algebras and spinors , 2nd ed . ,
cambridge university press ( 2001 ) c. chevalley , theory of lie groups , overseas publications ltd . | the experimental decay branching ratios of mesons like @xmath0 and @xmath1 ( @xmath2 or @xmath3 ) do not agree completely with the standard model ( sm ) .
cartan s supersymmetry predicts relation of the coupling of vector particles @xmath4 , ( @xmath5 ) to dirac spinors of large components @xmath6 and small components @xmath7 . in the decay of @xmath8
, the cabibbo - kobayashi - maskawa(ckm ) model suggests that the contribution of @xmath9 quark dominates , while in the decay of @xmath10 , contribution of @xmath11 quark in the intermediate state is expected to be large , since @xmath12 and @xmath11 quark belong to the same ckm sector .
the relative strength of the @xmath9 quark and @xmath11 quark contribution in cartan s supersymmetric model has more freedom than that of the sm .
together with the problem of enhancement of @xmath13 in high @xmath14 region , we can understand the problem of branching ratios of @xmath15 decay into @xmath16 and @xmath17 , if the nature follows cartan s supersymmetry . |
the most important results in the main text and these supplemental materials are based on numerical methods .
therefore , the results can be considered exact within the domain of validity of the equations we used , without relying on analytic approximations .
the two - component bec can be described by the mean - field gross - pitaevskii equation ( gpe ) @xcite .
however , the typical analytical treatment , the thomas - fermi approximation ( tfa ) , @xcite which ignores the kinetic energy term , is not reliable in our case . it is known that tfa can not be used in the case of high density @xcite , which is the case we are studying .
instead , we numerically solve the gpe : @xmath143\psi_{i}\label{eq : gpe}\ ] ] where @xmath144 is the kronecker delta which can not be ignored if @xmath145 is of order one ; @xmath46 and @xmath146 are the single mode wavefunction and the number of particles of the @xmath47-th bec component respectively .
the normalization is @xmath147 and the density is given by @xmath148 .
the trapping potential is @xmath43 , with trapping strength @xmath149 , and the interaction strength is @xmath45 , with scattering length @xmath49 between component @xmath47 and @xmath150 .
our target is to find the ground state energy and wavefunction , which can be done by using the imaginary time method @xcite .
first , we use a wick rotation @xmath151 on eq .
( [ eq : gpe ] ) to obtain the corresponding diffusion equation , which is then reduced to two coupled 1d non - linear diffusion equations with the assumption of spherical symmetry .
finally , we let the system relax to the ground state with the fourth order runge - kutta method in time and finite difference method in space .
after finding the ground state wavefunction , we can use it to calculate the mean - field energy functional : @xmath152 = \sum_{i = a , b}n_{i}\int d^{3}r\left(\frac{\hbar^{2}}{2m}|\nabla\psi_{i}|^{2}+v_{i}|\psi_{i}|^{2}+\frac{1}{2}(n_{i}-1)u_{ii}|\psi_{i}|^{4}\right)+n_{a}n_{b}\int d^{3}ru_{ab}|\psi_{a}|^{2}|\psi_{b}|^{2}\label{eq : energy - functional}\ ] ] which depends on the spatial mode @xmath65 and the number of particles @xmath145 . note that the spatial modes @xmath65 depend implicitly on @xmath145 through eq .
( [ eq : gpe ] ) . in our scheme , the focus is the ground state energy @xmath34 as a function of @xmath153 and @xmath154 because the total number of particles @xmath155 in the two - component bec is fixed . after solving a set of becs with different small component in the range
@xmath156 $ ] , we fit the results up to fourth order to get the expansion coefficients @xmath157 fig .
[ cap : properties ] shows how the most relevant properties of the ground state of the two - component bec change with @xmath26 . for the scheme described in the main text ,
the interesting regime is when component b is located at the center of the trap .
this can be achieved with a slightly higher trapping for @xmath26 in this weakly phase separated regime as described in fig .
[ cap : properties ] with cat size @xmath71 .
note that in the case of equal trapping @xmath158 , the small component b will locate outside of component a because of the effective repulsion in this regime . as shown in fig .
[ cap : properties]a , the width of @xmath159 is close to the width of a gaussian at around @xmath160 , while at higher @xmath26 , the width is larger than the corresponding gaussian because of the self - repulsion with other atoms in the same component b. the same effects can be observed for the real density @xmath161 at the center ( fig . [
cap : properties]b ) , which is lower than the corresponding gaussian density with @xmath26 . when @xmath162 , the component b has higher density than the main component a. this allows us to ignore most effects of the component a , including the quantum depletion .
[ cap : properties]c - e shows the expansion coefficients @xmath124 .
note that both @xmath63 and @xmath67 have zero - crossing points . with zero kerr coefficient , @xmath91 ,
the system may be used to store spin states for a long time .
also , the zero third order , @xmath163 , suggests a regime to create good small spin cat states .
[ cap : properties]f shows the effects of the third order term on the shift of the `` best '' cat time @xmath122 , see definition below . qualitatively , the change in @xmath63 with respect to @xmath26 can be understood as follow .
the contributions to the kerr nonlinearity come from intra - species ( aa , bb ) and inter - species ( ab ) interactions , which have opposite sign to each other .
when the trapping is weak and identical for both components , the kerr nonlinearity is close to zero . also , for the phase separated regime , the component b is staying in the outer region .
when the trapping frequency @xmath26 for the b component is increased , the b component moves to the center and the overlap between a and b increases at first , which leads to an increase in the inter - species interaction term , resulting in a larger and negative kerr nonlinearity .
for very strong trapping of the b component , the overlap between a and b decreases again whereas the intra - species interaction for the b component increases strongly , leading to a large positive kerr nonlinearity .
this explains the crossover from negative to positive kerr nonlinearity as shown in fig .
in contrast , for the non - phase separated regime , the b component always stays inside the a component , and there is no crossover as discussed in section iv ( see fig .
numerical density distribution for both components ( a and b ) and its approximation with cat size @xmath71 and trapping strength @xmath164 .
@xmath165 and @xmath166 are the unperturbed wavefunction used by the first order perturbation calculation .
@xmath167 is another approximation .
see text for details .
other parameters used are the same as in fig .
[ cap : properties ] .
, scaledwidth=70.0% ] the numerically obtained spatial density distribution @xmath168 is shown in fig .
[ cap : wavefunc - demo ] .
the approximate solution of a harmonic oscillator ground state @xmath169 for b is good . if we follow a thomas - fermi approach similar to the one used in the previous paper @xcite by dropping the kinetic energy term in eq .
( [ eq : gpe ] ) , we will get @xmath170 . as expected , this approximation is not good and gives a negative density as shown in fig .
[ cap : wavefunc - demo ] .
in contrast , the tfa solution for a single component bec @xmath171 gives a fair approximation for a , given by @xcite : @xmath172 therefore we perform first order perturbation theory with the following splitting for the gpe : @xmath173 where @xmath174 is the unperturbed hamiltonian and the perturbation is given by @xmath175 . note that @xmath176 is used .
the solutions of @xmath174 are given by eq .
( [ eq : gpe - unpert - sol ] ) . to calculate the energy analytically
, we expand the ground state energy @xmath177 as the taylor series : @xmath178 note that the chemical potentials ( energy change with respect to the number of particles ) are given by @xmath179 .
since the main component @xmath21 in the scheme is much larger than the small component @xmath22 , or @xmath180 , the expansion can be carried out around the point @xmath181 : @xmath182\label{eq : eta2-def}\end{aligned}\ ] ] note that the @xmath183 here denote the exact chemical potentials from the gpe , which can be approximated by the unperturbed @xmath184 plus the perturbed chemical potential @xmath185 : @xmath186 using the unperturbed solutions eq .
( [ eq : gpe - unpert - sol ] ) , the chemical potential can be calculated as : @xmath187 where @xmath188 is the characteristic length of a gaussian .
note that the perturbation involves an integration whose range is chosen to be the whole space for simplicity , which is justified by the fact that component b is much narrower than component a when @xmath189 ( see fig .
[ cap : wavefunc - demo ] ) . substituting these results back into @xmath190 in eq .
( [ eq : eta1-def ] ) , we have : @xmath191 the third and fourth terms on the right hand side are the effective interaction between the main bec and the component @xmath22 .
the last term is the repulsion between the particles in component @xmath22 .
the fourth term is small when @xmath189 and can be ignored .
this result gives a very good approximation as demonstrated in fig .
[ cap : properties]c .
similarly , differentiating the chemical potential yields the second order term @xmath192 in eq .
( [ eq : eta2-def ] ) : @xmath193 \label{eq : eta2-approx}\end{aligned}\ ] ] the first three derivatives are smaller than the last term when @xmath189 and @xmath194 .
therefore , at high @xmath0 , the last term dominates @xmath195 , yielding @xmath196 as shown in fig .
[ cap : properties]b , eq .
( [ eq : eta2-approx - simple ] ) gives an order of magnitude estimation of @xmath195 .
note that we can also get the dominant term as calculated above by assuming component a to have a constant density distribution @xmath197 , at @xmath130 .
a better approximation should take into account the change in density @xmath198 as shown in fig .
[ cap : wavefunc - demo ] .
the maximum of the q function , @xmath199 , as a function of the relative time @xmath200 .
the `` best '' real cat time @xmath127 is defined as the time in which there are the two highest peaks in the @xmath1-function .
the leftmost peak corresponds to the initial coherent spin state ( css ) , with a value @xmath201 .
the peak at @xmath202 corresponds to the spin cat state ( cat ) .
the time at which the cat state is observed is shifted with respect to the ideal case @xmath203 because of the higher - order nonlinearities .
the highest peak at @xmath204 corresponds to the css at the revival time .
note that all fitting orders are included for determining @xmath127 .
the cat state at @xmath122 and css at @xmath205 are plotted in figs .
3a and 3b in the main text .
the other parameters used are the same as in fig .
[ cap : properties ] .
, scaledwidth=70.0% ] cat states can be distorted by higher order nonlinearities . thus we need to find out to what extent the cat states are distorted and whether the distortion is tolerable .
another practical problem is to figure out the optimal time to observe a cat state in real experiments .
we define the `` best '' cat time @xmath122 as the time with the two highest peaks in q function .
this definition is based on the feature of the cat states that two separated peaks in the q function should be distinguished clearly .
this method is illustrated by fig .
[ cap : qmax - tauc ] with the highest peak value @xmath206 plotted over time .
it is clear that the peak for the cat state is located near @xmath203 as expected . in practice , we search around the nearby region , say @xmath207 $ ] , for the highest peak .
the resulting @xmath122 corresponds to the best cat time .
we further manually check that there are indeed only two opposite peaks in phase space . the resulting shift in the cat time
is plotted in fig .
[ cap : properties]d .
note that the @xmath122 depends on @xmath77 , see also fig .
[ cap : properties]f . in the scheme ,
the output light is of the form @xmath208 with the initial condition @xmath209 and @xmath210 .
hence , the q - function without loss is @xmath211 where the phase space is defined by @xmath212 .
this equation is numerically evaluated to obtain the q - function for given @xmath124 , which are obtained by fitting the solutions of the gpe eq .
( [ eq : gpe ] ) . a few more figures corresponding to fig .
3a in main text are plotted in fig .
[ cap : q - function - nbar ] for different cat sizes @xmath77 .
one can see that the higher order effects ( @xmath213 ) are weak for small @xmath77 , but significant for larger @xmath77 .
the scheme should also work in the non - phase separated regime @xmath214 .
[ cap : na - varying - aab ] shows the coefficients @xmath63 , @xmath67 , @xmath68 for different values of the inter - species scattering length @xmath215 , with @xmath216 and @xmath217 .
the plots suggest that the kerr effect is also strong in the non - phase separated regime , but the higher order terms might limit the resulting cat size @xmath77 .
the main qualitative difference is that there are no zero - crossing points for @xmath124 in the non - phase separated regime .
these results further suggest that the weakly phase separated regime is advantageous because the higher - order terms are very small around @xmath218 .
cat time @xmath5 and one - atom loss time @xmath70 for rubidium with @xmath219 , @xmath79 and non - zero two - body loss rate @xmath220 .
it is clear that the time to lose one atom through the two - body loss within the same component , @xmath221 , dominates at high @xmath0 , which has a similar scaling as the cat time @xmath222 .
this limits the maximum @xmath77 to around 10 atoms .
parameters : rubidium atoms @xmath223rb with scattering length @xmath224 , @xmath225 , @xmath226 , where @xmath227 is the bohr radius .
the atom loss rates are @xmath228/s , @xmath229 , @xmath230/s , @xmath231m@xmath232/s , @xmath233m@xmath234/s @xcite , and @xmath235hz .
, scaledwidth=57.0% ] comparison of the numerical and approximation results for all loss processes corresponding to fig .
2a in the main text .
the solid curves show the numerical solutions of the time to lose one atom @xmath73 through an @xmath50-body process with particle combination @xmath74 , while the approximations are shown as dotted or dashed curves with the same color . the time to lose one atom through all processes
is calculated by @xmath236 .
note that the approximations here are essentially lower bounds for the numerical solutions , as shown in this figure .
see text for details . , scaledwidth=49.0% ] the atom loss rate for component @xmath47 is given by @xcite : @xmath237 where @xmath238 , @xmath220 , @xmath239 are the one , two and three particle collision loss rates .
note that the density @xmath240 in the equation also depends on the numbers of particles @xmath146 which decrease over time .
as discussed in the main text , the individual times to lose one particle through an @xmath50-body process with particle combination @xmath74 are defined as @xmath241 , @xmath242 and @xmath243 , where @xmath244 and @xmath245 are used as an approximation .
these time scales can be evaluated using numerical integration for the ground state density distribution obtained from solving eq .
( [ eq : gpe ] ) . here
, we further show the results for rubidium atoms with non - zero two - body loss rate in fig .
[ cap : rb - loss ] .
the loss due to two - body effects is significantly larger than that due to three - body effects in this case , which limits the maximum cat size to @xmath219 atoms , as compared to a few hundred atoms for the sodium example used in the main text .
note that this is only one possible choice of states for rubidium .
large cats may still be possible if appropriate internal states and other conditions can be found such that two - body loss is suppressed . for cat state creation
, the maximum loss of atoms in the component @xmath22 can not be larger than one atom .
therefore , we are trying to give a conservative estimation . since @xmath246 in the tfa
, we can use the maximum @xmath247 for the main bec , and the gaussian @xmath248 for component b in eq .
( [ eq : gpe - unpert - sol ] ) : @xmath249 the estimations for three body loss are shown in fig . [ cap : loss - approx ] , which suggests they are good lower bounds for @xmath250 and the time to lose one atom through all loss channels @xmath236 .
the estimation is better at small @xmath26 , since the density of component a is not repelled away so that the approximation @xmath251 is good .
the readout loss from spin states to light is treated using the beam splitter model with a given loss rate @xmath252 .
the state passing through the beam splitter is @xmath253 with @xmath254 , so the reduced density matrix @xmath255 is @xmath256 hence , the resulting q function with loss @xmath257 and initial coherent state @xmath258 can be written as : @xmath259 the term inside the big bracket is the bivariate poisson distribution so this summation is upper bounded by 1 . therefore the resulting @xmath260 is confined to the annulus @xmath261 .
hence , the effect of photon loss is to move the peak of the q - function toward the origin , as shown in fig . 3c and 3d in the main text .
this form of the q - function can be evaluated fairly efficiently with time complexity of order @xmath262 , which allows us to evaluate it for cat sizes of order a few hundred atoms .
since all @xmath66 depend on the total atom number @xmath9 , the statistical fluctuations in @xmath9 can cause dephasing ( equivalent to angular spreading in phase space for the q - function studied here ) , which can wash out all observable features of cat states ( consider for example the @xmath9-dependent rotation @xmath263 caused by @xmath264 ) .
the dephasing is small if the derivatives of the coefficients with respect to @xmath9 , @xmath265 , are small .
these quantities are plotted in fig .
[ cap : n - uncertainty]a and [ cap : n - uncertainty]b .
note that the constancy of @xmath266 in fig .
( [ cap : n - uncertainty]a ) can be understood from eq .
( [ eq : eta1-approx ] ) because @xmath267 is independent of @xmath26 .
also , the dephasing is linear in time , hence , a short cat time @xmath8 can significantly reduce the dephasing effects
. moreover the rotation generated by @xmath264 can be canceled by the opposite rotation generated by @xmath268 , as we derive below . first considering the expansion of @xmath269 around @xmath77
the relevant terms become @xmath270 on the right hand side , the first term gives a global phase which can be neglected .
the second term leads to a rotation in phase space .
writing @xmath271 and expanding the coefficients around @xmath272 one has @xmath273 where @xmath274 .
note that @xmath275 is the fluctuation in @xmath276 due to the uncertainty @xmath140 . substituting these back into the second term in eq .
( [ eq : etas_expand_n ] ) yields the dephasing term @xmath277 .
this dephasing term is the source of a @xmath140 dependent rotation in the @xmath278-plane , which is eliminated when the condition @xmath279 is satisfied , see fig .
[ cap : n - uncertainty]c .
in particular , we want to find out the maximum allowable @xmath140 that still preserves an observable spin cat state at the cat time @xmath5 .
therefore , we define @xmath280 , and the condition @xmath281 should be satisfied , or @xmath282 the higher order terms @xmath276 can also be included , yielding @xmath283 numerically , we find @xmath284 by taking the numerical derivative of @xmath66 .
the results in fig .
[ cap : n - uncertainty]d show that there is a large range of allowable uncertainty in atom number @xmath140 if @xmath0 is high enough .
note that this range is an estimation since only the first order approximation of @xmath285 is used .
in contrast , the accuracy requirement @xmath286 at low trapping @xmath26 is even higher than the high resolution of counting cold atoms of 1 in 1200 in a recent experiment @xcite .
the continuous loss of atoms from the bec can be described by the master equation : @xmath287+\sum_{i}(\hat{r}_{i}\rho\hat{r}_{i}^{\dagger}-\frac{1}{2}\hat{r}_{i}^{\dagger}\hat{r}_{i}\rho-\frac{1}{2}\rho\hat{r}_{i}^{\dagger}\hat{r}_{i})\ ] ] where @xmath288 is the density operator , @xmath289 is the hamiltonian of the system , @xmath290 is the lindblad operator for the loss channel in question , and the summation is over the different loss channels . in the main text ,
the 2-body loss is assumed to be zero and 3-body loss is much lower than the 1-body loss ( see fig .
3 ) in the regime considered .
therefore , only 1-body loss will be considered below to simplify the calculation .
the loss atoms from the large component causes a change in @xmath9 .
the resulting effects are similar to the fluctuations in @xmath9 considered in the previous section . here
we therefore focus on the small component . in this case
one can effectively describe the system by the density operator @xmath291 , with the initial state @xmath292 .
only one lindblad operator @xmath293 is needed .
the simplified master equation is : @xmath287+l_{1}(\hat{a}\rho\hat{a}^{\dagger}-\frac{1}{2}\hat{a}^{\dagger}\hat{a}\rho-\frac{1}{2}\rho\hat{a}^{\dagger}\hat{a})\ ] ] with the hamiltonian given by eq .
( [ eq : etas ] ) . the method used to simulate the system is the quantum jump method @xcite .
the results are shown in fig .
[ cap : q - loss - standard ] . for the standard kerr effect without higher - order terms
, it can be observed that the creation of spin cat state still results in two clear peaks in the q - function even when 2.5 atoms are lost on average .
in fact , the cat state is still visible even for an average loss of 5 atoms .
however , for the detection scheme , the system is required to evolve for @xmath294 , which limits the loss rate to @xmath295 as shown in the figure . for the parameters used in fig .
2b in the main text ( including higher - order nonlinearities ) , the average number of atoms lost is only 0.68 and the effect of the loss is small , see fig .
[ cap : q - loss - bec - real - standard ] . the main effect of
the loss is a fairly uniform background ring in the q function .
@xcite utilizes a similar collision induced cross - kerr nonlinearity in bec to implement photon - photon gates , while the current scheme uses a self - kerr nonlinearity to create spin cat states . the kerr effect in the previous scheme
is enhanced by increasing both scattering length ( through a feshbach resonance ) and the trapping frequency for both components .
however , the feshbach resonance induced atom loss can be very large @xcite , which will limit the maximum cat size .
not relying on a feshbach resonance also makes it possible to use the magnetic field to further eliminate atom loss .
also , both trapping frequencies should not be increased at the same time because it will result in high atom loss through the collision with the main bec .
instead , we suggest here to increase only the trapping frequency of small component .
this results in a similarly strong kerr effect , but with lower atom loss .
moreover , the treatment in the previous paper , which used the quantized mean - field gpe with tfa and first order perturbation theory , does not allow the study of higher order nonlinearities or atom number fluctuations .
our present approach allows us to study both of these effects , and we show that they can be significant .
the assumption of equal trapping frequencies also limits the previous treatment to the non - phase separated regime , which limits the choice of regimes with low atom loss , such as the sodium atom example used here .
furthermore , the density of the stored component in the previous scheme is much smaller ( about four orders of magnitude ) than the main component .
this raises the concern of other possible dominant effects on the same scale , such as quantum depletion @xcite .
as we have shown , these problem can be minimized in the current scheme by using a high enough @xmath26 so that the small component is located at the center with high density .
note that eq .
( [ eq : eta2-approx - simple ] ) , which is obtained as a limiting case for high @xmath26 here , is basically equivalent to the results of the treatment in ref .
@xcite .
11 b. d. esry , c. h. greene , j. burke , and j. l. bohn , phys .
lett . * 78 * , 3594 ( 1997 ) .
l.p . pitaevskii and s. stringari , _ bose - einstein condensation _ ( oxford university press , new york , 2003 ) .
pethick and h. smith , _ bose - einstein condensation in dilute gases _
( cambridge university press , london , 2008 ) .
ho and v. b. shenoy , phys . rev .
lett . * 77 * , 3276 ( 1996 ) .
h. pu and n. p. bigelow , phys .
80 * , 1130 ( 1998 ) .
m. l. chiofalo , s. succi , and m. p. tosi , phys .
e * 62 * , 7438 ( 2000 ) .
r. zhang , s. r. garner , and l. v. hau , phys .
lett . * 103 * , 233602 ( 2009 ) .
d. m. stamper - kurn , m. r. andrews , a. p. chikkatur , s. inouye , h .- j .
miesner , j. stenger , and w. ketterle , phys .
lett . * 80 * , 2027 ( 1998 ) .
a. rispe , b. he , and c. simon , phys .
lett . * 107 * , 043601 ( 2011 ) . c. chin , r. grimm , p. julienne and e. tiesinga , rev
phys . * 82 * 1225 ( 2010 ) .
y. li , y. castin , and a. sinatra , phys .
* 100 * , 210401 ( 2008 ) .
y. li , p. treutlein , j. reichel , and a. sinatra , the european physical journal b * 68 * , 365 ( 2009 ) .
d. b. hume , i. stroescu , m. joos , w. muessel , h. strobel , and m. k. oberthaler , phys .
lett . * 111 * , 253001 ( 2013 ) .
trail , k. almutairi , d.l .
feder , and b.c .
sanders , submitted to the arxiv simultaneously with the present paper . b. m. garraway and p. l. knight , pra**272 * * , 1266 ( 1994 ) .
m. b. plenio and p. l. knight , rev .
* 70 * , 101 ( 1998 ) . | we propose a method to create `` spin cat states '' , i.e. macroscopic superpositions of coherent spin states , in bose - einstein condensates using the kerr nonlinearity due to atomic collisions . based on a detailed study of atom loss ,
we conclude that cat sizes of hundreds of atoms should be realistic .
the existence of the spin cat states can be demonstrated by optical readout .
our analysis also includes the effects of higher - order nonlinearities , atom number fluctuations , and limited readout efficiency .
great efforts are currently made in many areas to bring quantum effects such as superposition and entanglement to the macroscopic level @xcite . a particularly dramatic class of macroscopic superposition states
are so - called cat states , i.e. superpositions of coherent states where the distance between the two components in phase space can be much greater than their individual size @xcite .
for example , the recent experiment of @xcite created a cat state of over one hundred microwave photons in a waveguide cavity coupled to a superconducting qubit .
it was essential for the success of the latter experiment that the loss in that system is extremely small , since even the loss of a single particle from a cat state of this size will lead to almost complete decoherence .
here we show that it should be possible to create cat states involving the spins of hundreds of atoms in another system where particle losses can be greatly suppressed , namely , bose - einstein condensates ( becs ) , where the spins correspond to different hyperfine states .
we use the kerr nonlinearity due to atomic collisions , which also played a key role in recent demonstrations of atomic spin squeezing @xcite .
in contrast to previous proposals @xcite we do not make use of josephson couplings to create the cat state , but rely purely on the kerr nonlinearity in the spirit of the well - known optical proposal of ref .
@xcite .
our approach is inspired by the experiment of ref .
@xcite , which stored light in a bec for over a second . ref .
@xcite proposed to use collision - based interactions in this system to implement photon - photon gates , see also ref .
@xcite . here
we apply a similar approach to the creation and optical detection of spin cat states . because of the great sensitivity of these states
, this requires a careful analysis of atom loss .
our theoretical treatment goes beyond that of ref .
@xcite , which was based on the thomas - fermi approximation ( tfa ) .
our new approach allows us to study several key imperfections in addition to loss , including higher - order nonlinearities , atom number fluctuations , and inefficient readout , and we conclude that their effects should be manageable .
( color online ) spin cat state creation ( a)-(d ) and detection ( e)-(g ) . in ( a)-(f )
the radially symmetric photons and spherically symmetric becs are represented by spatial density distributions .
( a ) a coherent light pulse is sent into the bec .
( b ) the light state is absorbed in the bec ( see inset ) , creating a css .
the shape of the input pulse is chosen such that the two - component bec is in its ground state after the absorption .
( c ) the trapping frequency @xmath0 for the small component is increased adiabatically .
the density of the small component now exceeds that of the large component at the center .
( d ) the collision - induced kerr nonlinearity drives the system into a spin cat state ( cat ) .
( e ) the trapping frequency is adiabatically reduced to its initial value .
( f ) the spin state is reconverted into light , whose husimi @xmath1 function @xcite is determined via homodyne detection @xcite .
( g ) expected shape of the @xmath2 function in phase space .
the coherent state at @xmath3 gives a single peak , while the cat state at @xmath4 yields two peaks .
further evolution for another interval @xmath5 returns the output light to a coherent state , yielding a single peak at @xmath6 .
this would not be possible if the two peaks at @xmath7 corresponded to an incoherent mixture , thus proving the existence of a coherent superposition of csss in the bec at @xmath8 @xcite .
, scaledwidth=49.0% ] our scheme is illustrated in fig .
[ cap : scheme - illustration ] .
the setup is similar to the experiment of ref .
@xcite ; see also ref . @xcite . in particular
, the light is converted into atomic coherences using a control beam ( slow and stopped light ) @xcite .
we start with a ground state bec with @xmath9 atoms in internal states @xmath10 . to create a spin state , a coherent light pulse , @xmath11 with mean
photon number @xmath12 and @xmath13 , is sent into the bec ( see fig . [
cap : scheme - illustration]a ) .
the light is absorbed by the bec and some atoms are converted into internal states @xmath14 as : @xmath15 where the fock state @xmath16 represents @xmath17 and @xmath18 excitations of wavefunctions @xmath19 and @xmath20 in the @xmath21 and @xmath22 components respectively .
note that @xmath23 is an excellent approximation of a css @xcite @xmath24 in the limit of @xmath25 which is the case in this scheme .
the described absorption process should prepare the two - component bec in its motional ground state to avoid the complication of unnecessary dynamics such as oscillations .
this can be achieved by matching the shape of the input pulse to the ground state of the effective trapping potential for the small component @xcite , provided that the effective trap is not too steep .
once the light has been absorbed , the trapping frequency @xmath26 is then increased adiabatically independently of @xmath27 , which can be achieved by combining optical and magnetic trapping . in the regime @xmath28 , a narrow wavefunction @xmath20 is formed at the center and its density can exceed the large component a , see fig .
[ cap : scheme - illustration]c .
this results in strong self - interaction and hence a large kerr nonlinearity . on the other hand ,
keeping @xmath27 low reduces the unwanted effects due to collision loss involving the large component .
the spin state will now evolve with time according to @xmath29 with @xmath30 . if the energy takes the kerr nonlinear form @xmath31 , then a spin cat state @xmath32 is formed at the time @xmath33 in full analogy with the proposal of ref .
@xcite .
the problem is thus reduced to the computation of the ground state energy @xmath34 .
the energy of the system can be calculated by the following mean - field energy functional @xmath35 $ ] : @xmath36 where @xmath37 , @xmath38 , @xmath39 and @xmath40 are the kinetic energy , potential energy , intra- and inter - component interaction energy respectively , given by @xmath41 , @xmath42 with spherically symmetric trapping @xmath43 , and @xmath44 with interaction strength @xmath45 . here
, @xmath46 are single particle wavefunctions for @xmath47-th component with normalization @xmath48 , @xmath49 are the scattering lengths , and @xmath50 is the atom mass .
the corresponding dynamic equation governing the system evolution is the gross - pitaevskii equation ( gpe ) @xcite . with the restriction @xmath51 and @xmath52 of spin states creation in eq .
( [ eq : light - to - spin - state ] ) , the nonlinearity in @xmath53 can be obtained by the expansion of the energy @xmath54 around @xmath55 as : @xmath56 where @xmath57 generates a global phase and @xmath58 with chemical potential @xmath59 ( @xmath60 ) is the energy to remove one atom from @xmath10 and add one atom to @xmath14 .
@xmath61 generates a simple rotation in phase space @xmath62 , which can be eliminated by a frame rotation .
the term @xmath63 is the kerr nonlinearity .
we obtain these coefficients by fitting the total energy @xmath34 with @xmath64 $ ] up to fourth orders in eq .
( [ eq : energy - expansion ] ) , where the numerical ground state @xmath65 of gpe used in eq .
( [ eq : total - energy ] ) is found by the imaginary time method @xcite .
this numerical approach is better than ref .
@xcite because we can avoid the problems associated with the tfa of high densities @xcite .
also , the high density for the small component at the center limits the negative effect of quantum fluctuations in the large component @xcite .
the latter are less important than the classical fluctuations in @xmath66 due to uncertainty in @xmath9 , whose effects will be discussed below .
moreover , we can now study the effects of higher - order nonlinearities ( in particular @xmath67 and @xmath68 ) .
( color online ) ( a ) the time to create a spin cat state @xmath69 ( thick red curve ) versus the time to lose one atom @xmath70 in component @xmath22 ( thick black curve ) as a function of the trapping frequency for the small component @xmath26 .
the size of the cat @xmath71 in this example .
one sees that @xmath72 is possible for sufficiently large @xmath26 .
the plot also shows the main individual loss channels contributing to the calculation of @xmath70 , where @xmath73 is the individual time of losing one atom through @xmath50-body collision with particle combinations @xmath74 .
it furthermore shows analytic approximations for @xmath75 ( red dotted curve ) and @xmath76 ( blue dashed curve ) , see text .
( b ) achievable cat size @xmath77 as a function of @xmath26 .
the shaded region corresponds to @xmath78 for a condensate size @xmath79 as in ( a ) .
the cat size can be increased somewhat by reducing @xmath9 ( dashed line ) .
we also show that there is a region where @xmath80 so that loss should really be negligible .
the green circles correspond to @xmath81 ( from left to right ) .
the star corresponds to the values used in fig . 3 , and the corresponding density distributions
are shown in fig .
1(d ) .
both plots are for @xmath82na with spin states @xmath83 , @xmath84 , scattering lengths @xmath85 , @xmath86 @xcite , loss coefficients @xmath87 , @xmath88 , @xmath89 @xcite , and a trapping frequency @xmath90 for the large component . ,
scaledwidth=49.0% ] fig .
[ cap : na - loss - phase ] shows our results for the spin cat creation time @xmath33 and achievable cat size @xmath77 , taking into account the effects of atom loss .
it is clear that the cat time @xmath5 decreases significantly as the trapping strength @xmath26 increases .
note that the kerr effect disappears ( @xmath91 ) around @xmath92hz , which may be used for long term storage . as mentioned above
, the reason for the strong kerr effect for large @xmath26 is that strong trapping potential forces @xmath20 into a highly localized gaussian @xmath93 .
the radius of @xmath20 is of the order of the characteristic length @xmath94 , and the density @xmath95 is peaked at the center @xmath96 which can be much higher than @xmath97 in our regime , see fig .
[ cap : scheme - illustration]d .
therefore , the system can be effectively described by @xmath98 , and the second order term is approximately @xmath99 , which is consistent with the first order perturbation theory in the supplemental materials @xcite .
the phase between the two components of the spin cat state is flipped by losing just one atom ( see the supplemental materials @xcite for more details on the effects of atom loss ) .
this means that @xmath5 must be smaller than the time to lose one atom @xmath70 , which depends on the density and thus @xmath77 . in our scheme ,
the loss of atoms in component @xmath21 will not affect the cat states directly , so we focus on the loss of component @xmath22 only , which can be estimated by the following loss rate equation @xcite : @xmath100 where @xmath101 is the approximate time to lose one atom through all possible loss channels if @xmath102 .
the loss rates @xmath103 correspond to the loss through @xmath50-body collisions involving particles in component b , where @xmath104 is due to collisions with the background gas , @xmath105 is due to spin exchange collisions , and @xmath106 is due to three - body recombination @xcite .
it is known that the two - particle loss can be eliminated by certain choices of internal states and control methods such as applying a microwave field in @xcite , or a specific magnetic field as in ref .
@xcite .
the latter example motivates our choice of parameters in fig .
[ cap : na - loss - phase ] . fig .
[ cap : na - loss - phase]a shows the time to lose one atom through different channels : @xmath107 for one - body loss and @xmath108 for three - body loss with different combination of collisions .
it can be observed that the high @xmath0 regime is dominated by the loss of @xmath109 , which corresponds to @xmath110 . for even larger values of @xmath26 than those shown in the figure , the three - body loss time @xmath110 becomes shorter than @xmath8 .
the small @xmath0 regime is dominated by the effect of @xmath111 .
see the supplemental materials @xcite for an approximate analytical treatment of atom loss .
the desirable region for experiments is @xmath78 which also depends on @xmath77 .
therefore , we can draw a @xmath77-@xmath26 phase diagram , which shows the achievable cat size as the shaded area in fig .
[ cap : na - loss - phase]b .
( color online ) optical demonstration of the spin cat state in the presence of various imperfections for the parameter values corresponding to the star in fig .
2(b ) ( @xmath71 and @xmath112hz ) .
the spin state is reconverted into light and the husimi phase space distribution function @xmath2 is determined via homodyne tomography .
( a ) includes the effects of the higher - order nonlinearities @xmath67 and @xmath68 .
two far separated peaks corresponding to the cat state are clearly visible at @xmath113 , and one peak corresponding to the revived coherent state at @xmath114 .
the shift of the cat creation time due to the higher - order terms is @xmath115 .
( b ) furthermore includes @xmath116 photon retrieval loss , which moves the peaks towards the origin , and @xmath117 uncertainty in the total atom number , which spreads the peaks in the angular direction.,scaledwidth=49.0% ] we now discuss how the existence of the spin cat states can be demonstrated via optical readout , see also fig .
1(e ) to 1(g ) .
our detection scheme is based on a revival argument and hence involves measurements at different times @xcite ( see also the related experiment of ref .
@xcite ) . in all cases
the readout process starts by reducing the trapping frequency adiabatically to its initial value .
then the spin state @xmath118 is reconverted into a state of light @xmath119 , followed by homodyne detection on the output light . using optical homodyne tomography @xcite
, we can reconstruct the husimi q - function @xcite @xmath120 with the density matrix @xmath121 .
the q - function allows us to visualize the resulting spin states of bec as a function of time .
higher - order nonlinearities distort the cat state and shift the cat creation time from @xmath8 for a pure kerr nonlinearity to a different observed value @xmath122 .
fig .
[ cap : q - results](a ) shows @xmath123 for @xmath112hz including up to fourth - order nonlinear terms @xmath124 .
two peaks at @xmath125 can be identified clearly . at the revival time @xmath126 ,
a single peak is recovered , which proves the existence of spin cat states in the bec at @xmath127 , as described in fig .
[ cap : scheme - illustration](g ) .
note that the definition of @xmath127 used is the time at which the q - function shows the two highest peaks .
in general , @xmath128 and hence @xmath129 for @xmath130 since @xmath20 is less localized than @xmath131 due to the repulsive self - interaction .
for the weakly phase separated regime ( @xmath132 ) used in fig .
[ cap : na - loss - phase ] , the effective compression from component a on @xmath20 can have the reverse effect .
this gives @xmath133 and thus nearly perfect cat states at @xmath134 hz .
further higher - order effects are shown in the supplemental materials @xcite . in current experiments
the light storage and retrieval process involves significant photon loss , e.g. about 90% loss in ref .
@xcite .
its main effect is to move the peaks towards the origin , see fig . 3b and supplemental materials @xcite .
one important requirement for achieving high absorption and emission efficiency is high optical depth .
for the example of fig .
[ cap : na - loss - phase ] , the optical depth can be estimated as @xmath135 with @xmath136 , wavelength @xmath137 nm and the bec radius @xmath138 m .
this is in principle sufficient to achieve an overall efficiency close to 1 @xcite .
another important experimental imperfection is the fact that the total atom number @xmath9 can not be precisely controlled from shot to shot .
this leads to fluctuations in the nonlinear coefficients @xmath124 .
the most important negative effect of these fluctuations is dephasing , i.e. angular spreading of the peaks in fig . 3 in phase space @xcite .
the magnitude of the angular spread at the time @xmath69 of the cat state creation can be estimated as @xmath139 , where @xmath140 is the uncertainty in @xmath9 , as discussed in more detail in the supplemental materials @xcite .
we find that the sensitivity of our scheme to atom number fluctuations is minimized for @xmath141hz .
fig .
3b shows that a 5% uncertainty in @xmath9 can be tolerated for @xmath142 hz ( even when occurring in combination with 90 % photon loss ) .
two key ingredients for the success of the present scheme are the use of a high trapping frequency for the small component and the achievement of very low loss .
the high trapping frequency enhances the strength of the kerr nonlinearity , making it possible to create cat states without relying on a feshbach resonance as proposed in ref .
@xcite .
this makes it possible to avoid the substantial atom loss typically associated with these resonances @xcite , and also allows one to use the magnetic field to eliminate two - body loss , which is critical .
for example , the loss rates for the choice of rubidium internal states discussed in ref .
@xcite would only allow cat sizes of order ten atoms , see the supplemental materials @xcite .
the high trapping frequency also helped us to suppress the unwanted effects of higher - order nonlinearities and atom number fluctuations .
if the readout efficiency could be increased significantly , then the present scheme could also be used to create optical cat states . besides their fundamental interest , both spin cat states and optical cat states are attractive in the context of quantum metrology @xcite .
we thank l. hau , c. trail , d. feder , k. almutairi and b. sanders for useful discussions .
this work was supported by aitf and nserc .
11 c. monroe , d. m. meekhof , b. e. king , and d. j. wineland , science * 272 * , 1131 ( 1996 ) . m. brune , e. hagley , j. dreyer , x. maitre , a. maali , c. wunderlich , j. m. raimond , and s. haroche , phys . rev
. lett . * 77 * , 4887 ( 1996 ) .
d. leibfried , e. knill , s. seidelin , j. britton , r. b. blakestad , j. chiaverini , d. b. hume , w. m. itano , j. d. jost , c. langer , r. ozeri , r. reichle and d. j. wineland , nature * 438 * , 639 ( 2005 ) . b. lcke , m. scherer , j. kruse , l. pezz , f. deuretzbacher , p. hyllus , o. topic , j. peise , w. ertmer , j. arlt , l. santos , a. smerzi and c. klempt , science * 334 * , 773 ( 2011 ) . m. arndt , o. nairz , j. vos - andreae , c. keller , g. van der zouw , and a. zeilinger , nature * 401 * , 680 ( 1999 ) . m. arndt and k. hornberger , nature physics 10 , * 271 * , ( 2014 ) . j. r. friedman , v. patel , w. chen , s. k. tolpygo , and j. e. lukens , nature * 406 * , 43 ( 2000 ) .
b. julsgaard , a. kozhekin , and e. s. polzik , nature * 413 * , 400 ( 2001 ) .
j. estve , c. gross , a. weller , s. giovanazzi , and m. k. oberthaler , nature * 455 * , 1216 ( 2008 ) . c. gross , t. zibold , e. nicklas , j. estve , and m. k. oberthaler , nature * 464 * , 1165 ( 2010 ) .
m. f. riedel , p. bhi , y. li , t. w. hnsch , a. sinatra , and p. treutlein , nature * 464 * , 1170 ( 2010 ) .
a. d. oconnell , m. hofheinz , m. ansmann , r. c. bialczak , m. lenander , e. lucero , m. neeley , d. sank , h. wang , m. weides , et al . ,
nature * 464 * , 697 ( 2010 ) .
a. i. lvovsky , r. ghobadi , a. chandra , a. s. prasad , and c. simon , nature physics * 9 * , 541 ( 2013 ) .
n. bruno , a. martin , p. sekatski , n. sangouard , r. t. thew , and n. gisin , nature physics * 9 * , 545 ( 2013 ) . t. a. palomaki , j. d. teufel , r. w. simmonds , and k. w. lehnert , science * 342 * , 710 ( 2013 ) .
b. vlastakis , g. kirchmair , z. leghtas , s. e. nigg , l. frunzio , s. m. girvin , m. mirrahimi , m. h. devoret , and r. j. schoelkopf , science * 342 * , 607 ( 2013 ) . j. i. cirac , m. lewenstein , k. mlmer , and p. zoller , phys . rev .
a * 57 * , 1208 ( 1998 ) .
d. gordon and c. m. savage , phys .
rev .
a * 59 * , 4623 ( 1999 ) .
b. yurke and d. stoler , phys .
rev .
lett . * 57 * , 13 ( 1986 ) .
r. zhang , s. r. garner , and l. v. hau , phys .
rev .
lett . * 103 * , 233602 ( 2009 ) .
a. rispe , b. he , and c. simon , phys .
rev .
lett . * 107 * , 043601 ( 2011 ) . c. vo , s. riedl , s. baur , g. rempe , and s. drr , phys .
rev .
lett . * 109 * , 263602 ( 2012 ) . c. gerry and p. knight ,
_ introductory quantum optics _ ( cambridge university press , london , 2005 ) .
a.i .
lvovsky and m.g .
raymer , rev .
mod . phys . * 81 * , 299 ( 2009 ) .
d. a. r. dalvit , j. dziarmaga , and w. h. zurek , phys .
rev .
a * 62 * , 013607 ( 2000 ) .
m. lettner , m. mcke , s. riedl , c. vo , c. hahn , s. baur , j. bochmann , s. ritter , s. drr , and g. rempe , phys .
rev .
lett . * 106 * , 210503 ( 2011 ) .
l. v. hau , s. e. harris , z. dutton , and c. h. behroozi , nature * 397 * , 594 ( 1999 ) .
m. fleischhauer and m.d .
lukin , phys .
rev .
lett .
84 , 5094 ( 2000 ) .
d.f .
phillips , a. fleischhauer , a. mair , r.l .
walsworth , and m.d .
lukin , phys .
rev .
lett .
86 , 783 ( 2001 ) . c. liu , z. dutton , c. h. behroozi , and l. v. hau ,
nature * 409 * , 490 ( 2001 ) . z. dutton and l. v. hau , phys .
rev .
a * 70 * , 053831 ( 2004 ) .
n. s. ginsberg , s. r. garner , and l. v. hau , nature * 445 * , 623 ( 2007 ) . j.m .
radcliffe , j. phys .
a : gen . phys . * 4 * , 313 ( 1971 ) .
b. d. esry , c. h. greene , j. burke , and j. l. bohn , phys .
rev .
lett . * 78 * , 3594 ( 1997 ) .
l.p .
pitaevskii and s. stringari , _ bose - einstein condensation _
( oxford university press , new york , 2003 ) .
c.j .
pethick and h. smith , _ bose - einstein condensation in dilute gases _
( cambridge university press , london , 2008 ) .
m. l. chiofalo , s. succi , and m. p. tosi , phys .
rev .
e * 62 * , 7438 ( 2000 ) .
h. pu and n. p. bigelow , phys .
rev .
lett . * 80 * , 1130 ( 1998 ) . c. m. trail , k. almutairi , d. l. feder , and b. c. sanders , phys .
rev . a * 89 * , 063823 ( 2014 ) .
see the supplemental materials for more details on our results and methods .
d. m. stamper - kurn , m. r. andrews , a. p. chikkatur , s. inouye , h .- j .
miesner , j. stenger , and w. ketterle , phys .
rev .
lett . * 80 * , 2027 ( 1998 ) . c. chin , r. grimm , p. julienne and e. tiesinga , rev .
mod .
phys . * 82 * 1225 ( 2010 ) .
a. srensen , l .- m .
duan , j. i. cirac , and p. zoller , nature * 409 * , 63 ( 2001 ) .
y. li , y. castin , and a. sinatra , phys .
rev .
lett . * 100 * , 210401 ( 2008 ) .
y. li , p. treutlein , j. reichel , and a. sinatra , the european physical journal b * 68 * , 365 ( 2009 ) .
m. greiner , o. mandel , t.w .
hnsch , and i. bloch , nature * 419 * , 51 ( 2002 ) .
a.v .
gorshkov , a. andr , m.d .
lukin , and a.s .
srensen , phys .
rev . a * 76 * , 033805 ( 2007 ) .
t. wang , r. ghobadi , s. raeisi , and c. simon , phys .
rev .
a * 88 * , 062114 ( 2013 ) .
g. bjrk and p.g.l .
mana , j. opt .
b : quantum semiclass . opt .
* 6 * , 429 ( 2004 ) . *
supplemental materials : proposal for the creation and optical detection of spin cat states in bose - einstein condensates * |
takagi s continuous nowhere differentiable function , shown in figure [ fig : takagi - and - gray - takagi](a ) , is defined by @xmath13 where @xmath14 , the distance from @xmath3 to the nearest integer .
a great deal has been written about this function since its initial discovery in @xcite ; see recent surveys by allaart and kawamura @xcite and lagarias @xcite for an overview of the literature . in the past few years
, interest has focused mainly on the level sets @xmath15:t(x)=y\}$ ] , which have been shown to possess a rich structure .
for instance , @xmath16 is finite for lebesgue - almost every @xmath17 @xcite , and can have any even positive integer as its cardinality @xcite .
however , the typical " level set of @xmath18 ( in the sense of baire category ) is uncountably large @xcite .
these uncountable level sets can be further differentiated according to their hausdorff dimension .
kahane @xcite showed that @xmath19 , and @xmath20 is a cantor set of hausdorff dimension @xmath10 .
de amo et al .
@xcite recently proved that @xmath10 is the maximal hausdorff dimension of any level set , settling a conjecture of maddock @xcite , who had earlier obtained a bound of @xmath21 .
two interesting papers by lagarias and maddock @xcite use novel notions of ` local level sets ' and a ` takagi singular function ' to establish several properties of the level sets of @xmath18 .
for instance , it is shown in @xcite that the set of @xmath17-values for which @xmath16 has strictly positive hausdorff dimension is a set of full hausdorff dimension 1 . in this article
we examine the level sets of a class of generalized takagi functions of the form @xmath22 where @xmath23 observe that @xmath5 can jump only at points @xmath3 where @xmath24 , so the terms of the series in are continuous . as a result , @xmath11 is a continuous function .
we denote by @xmath25 the class of all functions @xmath11 of the form , and by @xmath26 the subclass of those @xmath11 in @xmath25 for which @xmath5 is constant for each @xmath4 .
the class @xmath25 was investigated in detail by abbott , anderson and pitt @xcite , who denoted it by @xmath27 because of its relationship with zygmund s class @xmath28 of quasi - smooth functions .
but whereas @xcite studies the class @xmath25 from the perspective of abscissa or @xmath3-values , our focus here is on ordinate or @xmath17-values .
several members of @xmath25 have featured in the literature .
these include the _ alternating takagi function _
( e.g. @xcite ) , for which @xmath29 and which hence lies in @xmath26 ; and the _ gray takagi function _
@xcite , for which @xmath30 , where @xmath31 denotes the @xmath4th rademacher function defined by @xmath32 the gray takagi function is shown in figure [ fig : takagi - and - gray - takagi](b ) .
another example is the function @xmath33 of kawamura @xcite , which has @xmath34 .
all members of @xmath25 are nowhere differentiable ; billingsley s proof @xcite for the takagi function extends easily .
furthermore , their graphs have hausdorff dimension one @xcite .
all functions in the subclass @xmath26 are symmetric with respect to @xmath35 , and their level set structure is similar to that of @xmath18 : lebesgue - almost all level sets are finite , but the typical " level set is uncountably infinite ; see @xcite .
whether these properties hold for the wider class @xmath25 remains unsolved .
this paper concerns the hausdorff dimension of the level sets @xmath36 : f(x)=y\ } , \qquad f\in{\mathcal{t}}_v , \quad y\in{\mathbb{r}}.\ ] ] for a borel set @xmath37 , we denote the hausdorff dimension of @xmath38 by @xmath39 .
\(a ) ( b ) the first half of the paper gives sharp upper bounds for @xmath40 , first for @xmath41 , and then for @xmath42 .
[ thm : rigid - upper - bound ] let @xmath41 , with representation @xmath43 .
then @xmath44 the maximum in is attained at a set of @xmath17-values dense in the range of @xmath11 , and in particular , at @xmath45 the gray takagi function satisfies even though it does not belong to @xmath26 .
[ thm : gray - upper - bound ] let @xmath11 be the gray takagi function , @xmath46 where @xmath31 is the @xmath4th rademacher function . then @xmath47 . for general @xmath42 , however , the hausdorff dimension of the level sets of @xmath11 can be much greater : [ thm : flexible - upper - bound ] let @xmath48 , and put @xmath49 .
we have @xmath50 by embedding the graph of @xmath11 affinely into the graph of a suitable function @xmath51 , the above results can be extended to the intersection of the graph of @xmath11 with arbitrary lines of integer slope .
[ cor : integer - slope ] for @xmath52 , let @xmath53 denote the line with equation @xmath54 .
a. for each @xmath41 and each @xmath55 , @xmath56 b. let @xmath57 be as in theorem [ thm : flexible - upper - bound ] .
for each @xmath55 , @xmath58 we prove theorems [ thm : rigid - upper - bound ] , [ thm : gray - upper - bound ] and [ thm : flexible - upper - bound ] and corollary [ cor : integer - slope ] in section [ sec : upper - bounds ] , by modifying and extending the method of de amo et al . @xcite .
the idea is to consider the intersection of the graph of the partial sum @xmath59 of @xmath11 ( defined in below ) with a suitably chosen horizontal strip @xmath60\times j_n$ ] , where @xmath61 is a nested sequence of intervals shrinking to @xmath62 .
we then derive a system of linear recursions for the number of line segments in these strips ( differentiated according to their slopes ) .
an added complication is , that the coefficients in these recursions are dependent on @xmath4 .
thus , for instance , the greater part of the proof of consists in determining the joint spectral radius of a certain pair of @xmath63 nonnegative matrices .
it appears to be a lucky coincidence that this can be done exactly .
perhaps most interesting is the case when the signs @xmath5 are chosen at random ; we consider two natural schemes here . in model 2 below and in the rest of the paper , @xmath64 denotes the set of nonnegative integers .
let @xmath65 be a probability space large enough to accomodate an infinite sequence of bernoulli random variables with arbitrary success probabilities .
( _ random choice from @xmath26 _ ) let @xmath66 be independent and identically distributed ( i.i.d . )
random variables on @xmath65 with @xmath67 , where @xmath68 , and set @xmath69 .
( _ random choice from @xmath25 _ ) let @xmath70 be i.i.d .
random variables on @xmath65 with @xmath71 , and set @xmath72 if @xmath73 . in either model , set @xmath74 . since it seems difficult to treat the level sets in full generality , we focus on two special cases : the zero set and the set of maximum points of the random function @xmath11 .
let @xmath75 for the zero set of @xmath11 , we have the following results .
[ thm : zero - set - rigid ] assume model 1 .
a. the zero set @xmath76 is finite with probability @xmath77 and given that @xmath76 is infinite , @xmath78 a.s .
, for each @xmath79 . b. if @xmath80 , then @xmath81 a.s .
[ thm : zero - set - flexible ] assume model 2 . if @xmath80 , then @xmath82 a.s .
which functions @xmath42 , specifically , satisfy @xmath82 ?
again the gray takagi function provides an example .
[ prop : gray - takagi - zero - set ] let @xmath11 be the gray takagi function
. then @xmath82 .
next , for @xmath42 , define @xmath83 } f(x),\ ] ] and @xmath84 : f(x)=m_f\}=l_f(m_f).\ ] ] note that @xmath85 for all @xmath42 , so @xmath86 .
the following theorem was proved in @xcite and is included here for comparison with model 2 .
[ thm : maximal - set - rigid ] assume model 1 .
a. if @xmath87 , then the distribution of @xmath88 is singular continuous and @xmath89 a.s .
b. if @xmath90 , then the distribution of @xmath88 is discrete and @xmath91 is finite a.s .
in fact , the paper @xcite specifies the distributions of @xmath88 and the cardinality of @xmath91 in considerable detail under the assumption of model 1 .
the analysis appears to be much harder for model 2 , and we describe here only what happens when @xmath92 .
note the contrast with the previous theorem .
[ thm : maximal - set - flexible ] assume model 2 . a. the probability that @xmath88 attains the maximum possible value of @xmath93 is @xmath94 b. if @xmath92 , then @xmath95 and the distribution of @xmath88 is discrete , supported on the set @xmath96 it seems plausible ( by monotonicity considerations ) that @xmath97 a.s . for @xmath98 , but the author has not been able to prove this . the results for the random case are proved in section [ sec : random - case ] .
theorem [ thm : zero - set - flexible ] is proved by casting the zero set as the attractor of a mauldin - williams random recursive construction and by using properties of hitting times in a symmetric simple random walk .
the proof of theorem [ thm : zero - set - rigid ] requires a different approach . for the upper bound , we use the perron - frobenius theorem to construct a sequence of positive supermartingales which lead , via the martingale convergence theorem , to successive upper bounds for @xmath99 ; we then prove that these upper bounds converge to @xmath100 .
for the lower bound we identify a particular combination of line segments , called a @xmath101-shape " , which will appear around the @xmath3-axis in the step - by - step construction of the graph of @xmath11 with probability one given that @xmath76 is nonfinite .
we show , via the law of large numbers , that the number of @xmath101-shapes grows exponentially fast almost surely once a @xmath101-shape appears .
this , along with some additional observations , gives the lower bound .
finally , theorem [ thm : maximal - set - flexible ] is proved by considering a sequence of galton - watson branching processes associated with the random construction of @xmath11 .
there are many natural questions still unanswered ; section [ sec : open ] lists some of them .
the following notation will be used throughout . for an interval @xmath102
, @xmath103 denotes the diameter of @xmath102 , and @xmath104 denotes the interior of @xmath102 .
the cardinality of a discrete set @xmath105 is denoted by @xmath106 . for @xmath107 and @xmath11
defined by , put @xmath108 then @xmath109 is piecewise linear , and the right - hand derivative of @xmath109 at any point @xmath110 is @xmath111 where @xmath31 is defined as in .
thus , @xmath112 for all @xmath3 and all @xmath4 , and @xmath113 . in particular , @xmath114 is always even .
another important observation is that @xmath115 for integer @xmath116 and @xmath107 .
we need two more elementary facts about the functions @xmath109 : [ lem : tail - bound ] for each @xmath107 , @xmath117 .
we have the estimate @xmath118 thus , the lemma follows from the bound @xmath119 .
[ lem : even ] for @xmath107 and integer @xmath120 , @xmath121 is an even integer .
we can write @xmath122 now for each integer @xmath123 , @xmath124 is an integer with the same parity as @xmath120 . to see this , note that @xmath125 is either @xmath126 or @xmath127 , where @xmath128 , and @xmath120 , @xmath129 and @xmath130 all have the same parity
thus , the terms in the right hand side of are either all even or all odd .
since there is an even number of them , their sum is even .
we now introduce the closed intervals @xmath131 , \qquad n\in{\mathbb{z}}_+ , \quad j=0,1,\dots,2^n-1,\\ j_{n , k}:=\left[\frac{2k}{4^n},\frac{2k+2}{4^n}\right ] , \qquad n\in{\mathbb{z}}_+,\ \ k\in{\mathbb{z}}.\end{gathered}\ ] ] we denote by @xmath132 the value of @xmath5 on @xmath133 , and by @xmath134 the slope of @xmath109 on @xmath135 .
observe that @xmath136
in this section we prove the universal bounds of subsection [ subsec : sharp - bounds ] . after developing some notation , we prove theorems [ thm : rigid - upper - bound ] and [ thm : gray - upper - bound ] in subsection [ subsec : rigid - bounds ] .
the next two subsections together prove theorem [ thm : flexible - upper - bound ] . in subsection
[ subsec : extremal - function ] we construct a function @xmath42 and an ordinate @xmath17 for which @xmath137 has hausdorff dimension @xmath57 , and in subsection [ subsec : flexible - upper - bound ] we show that @xmath57 is a universal upper bound for @xmath40 .
subsection [ subsec : lines ] gives a proof of corollary [ cor : integer - slope ] .
following de amo et al .
@xcite , we divide the strip @xmath60\times{\mathbb{r}}$ ] into closed rectangles @xmath138 let @xmath139 put @xmath140 and @xmath141 lemma [ lem : even ] implies that if @xmath142 , then @xmath143 intersects the graph of @xmath59 in a nonhorizontal line segment . moreover , if @xmath143 intersects the graph of @xmath11 , then by lemma [ lem : tail - bound ] the graph of @xmath59 must intersect @xmath144 . hence , for each @xmath116 , @xmath145 so that @xmath146 let @xmath41 .
we claim that @xmath147 and @xmath148 to verify and , we consider four cases regarding @xmath149 and @xmath150 .
let @xmath151 . @xmath152 .
if @xmath153 , then @xmath154 and hence @xmath155 .
thus , @xmath156 and @xmath157 lie in @xmath158 , while @xmath159 and @xmath160 lie in @xmath161 . if @xmath142 , then @xmath162 is monotone on @xmath163 .
in particular , if @xmath164 , then @xmath165 and hence , @xmath166 . similarly , if @xmath167 , then @xmath168 . if @xmath169 , then no component of @xmath170 is equal to @xmath171 .
we see that : * for @xmath172 , each @xmath120 in @xmath173 contributes two members to @xmath174 while each @xmath120 in @xmath175 contributes none .
hence @xmath176 * for @xmath177 , each @xmath120 in @xmath173 or @xmath120 in @xmath175 contributes no members to @xmath174 while each @xmath120 in @xmath178 contributes at most one .
hence , @xmath179 * for @xmath180 or @xmath181 , neither @xmath120 in @xmath173 nor @xmath120 in @xmath175 contributes to @xmath174 ,
so @xmath182 from , and , follows in case 1 .
the verfication of is simpler : it is clear that each @xmath153 contributes at most two members to any set @xmath183 and each @xmath142 contributes at most one by the monotonicity of @xmath162 on @xmath163 . from this , follows .
@xmath184 , @xmath185 .
now if @xmath153 we have @xmath186 and so @xmath159 and @xmath160 lie in @xmath187 , while @xmath156 and @xmath157 belong to @xmath161 .
if @xmath142 then @xmath162 is monotone on @xmath163 , so @xmath120 contributes at most one member to any set @xmath183 .
if in particular @xmath164 , then @xmath188 and so @xmath189 .
similarly , if @xmath167 , then @xmath190 . if @xmath169 , then @xmath170 does not have any zero components . we conclude that : * for @xmath177 , each @xmath120 in @xmath173 contributes two members to @xmath174 while @xmath120 in @xmath175 or @xmath120 in @xmath178 contributes none .
hence @xmath191 * for @xmath181 , each @xmath120 in @xmath173 contributes no members to @xmath174 and each @xmath120 in @xmath175 contributes at most one . hence @xmath192 * finally , for @xmath172 or @xmath180 , neither @xmath120 in @xmath173 nor @xmath120 in @xmath175 contributes to @xmath174 ,
so @xmath193 from , and , follows in case 2 . and follows in the same way as in case 1 .
this case is similar to case 2 , by symmetry .
this case is similar to case 1 , by symmetry . using , and the initial conditions @xmath196 and @xmath197 , it is straightforward to verify inductively that @xmath198 hence , by , @xmath199 .
it follows that for each @xmath17 , @xmath137 is covered by at most @xmath200 intervals of length @xmath201 .
thus , @xmath202 we next show that @xmath17 given by satisfies @xmath203 .
the equalities in and show ( by induction ) that there is a sequence of ordinates @xmath204 , such that @xmath205 for each @xmath4 . specifically , @xmath206 , and @xmath207 note that in particular , @xmath208 .
now let @xmath209 .
the sets @xmath210 are compact and nested , so the intersection @xmath211 is a nonempty compact set .
the intersection of @xmath212 with any interval @xmath213 in @xmath214 consists of two ( non - overlapping ) subintervals of one - fourth the length of @xmath215 .
hence , by the basic theory of moran fractals , @xmath216 .
now let @xmath217 , which exists since @xmath218 . if @xmath219 , then @xmath220 for each @xmath4 , and taking limits gives @xmath221 .
hence , @xmath222 , and so @xmath203 .
one checks easily that the ordinate @xmath17 thus constructed is the one given by .
it remains to establish that the set @xmath223 is dense in the range of @xmath11 .
the key observation is that , when viewed as random variables on the probability space @xmath224 with borel sets and lebesgue measure , the slopes @xmath225 , @xmath107 form a symmetric simple random walk , so they take the value @xmath171 infinitely often with probability one .
this implies that the sets @xmath226 have full measure in @xmath60 $ ] , so they are in particular dense in @xmath60 $ ] .
since @xmath11 is continuous , the set @xmath227 is therefore dense in the range of @xmath11 . now if @xmath228 , the restriction of the graph of @xmath11 to @xmath135 is a similar copy of the graph of some function @xmath229 , so by the previous part of the proof there is a @xmath230 such that @xmath203 .
it therefore suffices to show that for any subinterval @xmath102 of the range of @xmath11 there is a pair @xmath231 with @xmath228 for which @xmath232 .
given such a @xmath102 , choose @xmath233 so large that @xmath234 .
then if @xmath235 and @xmath228 , we have @xmath236 the denseness of @xmath237 thus implies that at least one such interval @xmath238 must be fully contained in @xmath102 .
this completes the proof .
let @xmath11 be the gray takagi function .
then @xmath239 for all @xmath4 and @xmath120 , and it is easy to use this along with to prove inductively that for each @xmath4 , @xmath240 and @xmath241 for all @xmath120 .
this implies that @xmath242 for all @xmath4 and all @xmath120 .
thus , we have the following substitutions @xmath243 : [ cols="<,>,<,^ , < , > , < " , ] let @xmath244 , and let @xmath245 be the @xmath246-algebra generated by the random vectors @xmath247 .
let @xmath248 , @xmath249 .
since @xmath80 , the four cases above all occur with probability @xmath250 , and hence we have @xmath251 put @xmath252 . by ,
@xmath253 where @xmath254 is the @xmath255 tridiagonal matrix @xmath256 let @xmath257 denote the spectral radius of @xmath254 . since @xmath254 is nonnegative @xmath257 is an eigenvalue of @xmath254 , and since @xmath254 is irreducible , the perron - frobenius theorem guarantees the existence of a positive left eigenvector @xmath258 of @xmath254 corresponding to @xmath257 .
it follows by that the process @xmath259 is a positive supermartingale , which by the martingale convergence theorem converges almost surely to a finite nonnegative limit @xmath260 .
let @xmath261 be the smallest entry of the vector @xmath258 .
then @xmath262 , and for a given @xmath263 , implies that for all sufficiently large @xmath4 , @xmath264 thus , by the zero criterion , the number of intervals @xmath163 needed to cover @xmath76 grows at most at rate @xmath265 .
consequently , @xmath266 the proof will be complete if we can show that @xmath267 let @xmath268 be the @xmath269 matrix obtained by deleting the last row and last column of @xmath254 . for completeness , we define @xmath270 $ ] and @xmath271 .
let @xmath272 be the characteristic polynomials of @xmath254 and @xmath268 , respectively .
then @xmath273 and @xmath274 satisfies the recursion @xmath275 with initial conditions @xmath276 for fixed real @xmath277 , the solution of the system , can be written as @xmath278 where @xmath279 and @xmath280 substituting into and some rearranging finally gives @xmath281 for real @xmath277 , where @xmath282 now observe that @xmath277 implies @xmath283 and @xmath284 , with @xmath285 if and only if @xmath286 .
if @xmath287 we are done , since @xmath288 .
so assume @xmath289 ; then we can find @xmath290 such that @xmath291 for all sufficiently large @xmath116 , and hence @xmath292 for all sufficiently large @xmath116 , for some @xmath263 depending on @xmath293 .
since @xmath294 for all @xmath116 , this , together with , implies that @xmath295 must tend to zero as @xmath296 .
hence , by , either @xmath297 or @xmath298 .
it is easy to see that the latter is impossible , and so @xmath299 must be a solution of @xmath300 , as @xmath301 is continuous in @xmath302 .
routine algebra using , and shows that the only solution of @xmath300 is @xmath303 .
this proves , completing the proof of the theorem
. assume model 2 . slightly abusing notation , put @xmath304 and @xmath305 : f_n(x)=m_n\}$ ] . note ,
since @xmath306 , that @xmath307 define random index sets @xmath308 and put @xmath309 . then @xmath310 is a galton - watson ( gw ) process with initial value @xmath311 and offspring distribution @xmath312 , where @xmath313 .
the offspring distribution has mean @xmath314 and probability generating function @xmath315 .
let @xmath316 .
according to the basic theory of the gw process ( e.g. @xcite ) , @xmath317 when @xmath318 , and in that case , @xmath319 with probability one .
assume from now on that @xmath320 ; that is , @xmath92 .
then @xmath321 is the smallest positive number satisfying @xmath322 , so that @xmath323 , and @xmath324 this establishes part ( i ) of the theorem .
next , define the random set @xmath325 then @xmath326 if and only if @xmath327 , and given that @xmath328 , @xmath329 a.s . by theorem 1.1 of mauldin and williams @xcite .
put @xmath330 . proceeding inductively ,
suppose processes @xmath331 and random variables @xmath332 have been defined and that @xmath333 for @xmath334 .
let @xmath335 , and define @xmath336 since @xmath87 , @xmath337 is finite almost surely .
( put @xmath338 , and let @xmath339 be a point of maximum of @xmath340 .
the slope of @xmath341 directly to the right of @xmath342 starts with a nonpositive value and follows ( as a function of @xmath107 ) a simple random walk with parameter @xmath343 , so it will eventually reach @xmath171 . )
note that @xmath344 define the random index sets @xmath345 for @xmath107 , and put @xmath346 . by definition of @xmath337 , @xmath347 and so @xmath348 .
now @xmath349 is again a gw process with offspring distribution @xmath350 , and it depends on the preceding processes @xmath331 only through the value of @xmath351 .
thus , @xmath352 define the random set @xmath353 then @xmath354 if and only if @xmath355 as @xmath356 , and given that @xmath357 for all @xmath4 , @xmath358 a.s . now implies that with probability one , there will eventually be a @xmath359 such that @xmath360 for all @xmath4 , and for that @xmath116 , we have @xmath361 , and @xmath362 part ( ii ) of the theorem now follows .
there are many natural questions left to answer .
a few are listed here . * problem 1 . *
does there exist @xmath42 such that @xmath363 for every @xmath364 ?
* problem 2 .
* is it true for all @xmath42 that @xmath137 is finite for lebesgue - almost every @xmath17 ? is this true for the gray takagi function ? * problem 3 . *
( random case , model 1 ) prove or disprove that @xmath82 a.s . when @xmath80 . * problem 4 . * ( random case , model 2 ) what is the best ( smallest ) bound @xmath365 such that , for each @xmath364 , @xmath366 ?
more strongly , what is the smallest @xmath367 such that @xmath368 ?
( obviously , @xmath369 . ) in the case of model 1 , only the first question is of interest , as @xmath370 in view of theorem [ thm : rigid - upper - bound ] . * problem 5 . * ( random case , model 2 ) prove or disprove that @xmath97 a.s . when @xmath98 , and that @xmath91 is finite a.s .
when @xmath90 . what can one say about the distribution of @xmath88 in these cases ?
the author is grateful to the referee for a careful reading of the manuscript and for suggesting improvements to the presentation of the paper . , the takagi function and its properties . in : functions and number theory and their probabilistic aspects ( k. matsumote , editor in chief ) ,
_ rims kokyuroku bessatsu _ * b34 * ( 2012 ) , pp .
153189 , arxiv:1112.4205v2 . | this paper examines the hausdorff dimension of the level sets @xmath0 of continuous functions of the form @xmath1 where @xmath2 is the distance from @xmath3 to the nearest integer , and for each @xmath4 , @xmath5 is a @xmath6-valued function which is constant on each interval @xmath7 , @xmath8 .
this class of functions includes takagi s continuous but nowhere differentiable function .
it is shown that the largest possible hausdorff dimension of @xmath0 is @xmath9 , but in case each @xmath5 is constant , the largest possible dimension is @xmath10 .
these results are extended to the intersection of the graph of @xmath11 with lines of arbitrary integer slope .
furthermore , two natural models of choosing the signs @xmath12 at random are considered , and almost - sure results are obtained for the hausdorff dimension of the zero set and the set of maximum points of @xmath11 .
the paper ends with a list of open problems . : 26a27 , 28a78 ( primary ) , 15a18 , 15b48 , 37h10 ( secondary ) : generalized takagi function , gray takagi function , level set , hausdorff dimension , random takagi function , joint spectral radius , random fractal |
the eigenvalues of the adjacency matrix of a finite undirected graph , @xmath14 , are real and hence can be ordered @xmath15 where @xmath16 is the number of vertices in @xmath14 .
if @xmath14 is @xmath0-regular , i.e. , each vertex is of degree @xmath0 , then @xmath1 . in @xcite ,
noga alon conjectured that for any @xmath17 and @xmath5 , @xmath18 for `` most '' @xmath0-regular graphs on a sufficiently large number of vertices .
the alon - boppana bound shows that the constant @xmath19 can not be improved upon ( see @xcite ) .
the main goal of this paper is to prove this conjecture for various models of a `` random @xmath0-regular graph . ''
our methods actually show that for `` most '' @xmath0-regular graphs , @xmath20 for all @xmath21 , since our methods are variants of the standard `` trace method . ''
our primary interest in alon s conjecture , which was alon s motivation , is that fact that graphs with @xmath22 small for @xmath21 have various nice properties , including being expanders or magnifiers ( see @xcite ) . for a fixed @xmath16 we can generate a random @xmath0-regular graph on @xmath16 vertices as follows ,
assuming @xmath0 is even ( later we will give random graph models that allow @xmath0 to be even or odd ) .
take @xmath3 permutations on @xmath23 , @xmath24 , each @xmath25 chosen uniformly among all @xmath26 permutations with all the @xmath25 independent .
we then form @xmath27 yielding a directed graph @xmath28 , which we may view as undirected .
we call this probability space of random graphs @xmath29
. @xmath14 can have multiple edges and self - loops , and each self - loop contributes @xmath30 to the appropriate diagonal entry of @xmath14 s adjacency matrix .
the main goal of this paper is to prove theorems like the following , which prove alon s conjecture , for various models of a random @xmath0-regular graph ; we start with the model @xmath29 .
[ th : main ] fix a real @xmath5 and an even positive integer @xmath0 .
then there is a constant , @xmath10 , such that for a random graph , @xmath14 , in @xmath29 we have that with probability at least @xmath31 we have for all @xmath32 @xmath33 where @xmath34 .
furthermore , for some constant @xmath35 we have that @xmath36 with probability at least @xmath37 , where @xmath38 .
( so @xmath39 unless @xmath40 is an integer , in which case @xmath41 . ) left open is the question of whether or not this theorem holds with @xmath42 ( which would yield `` ramanujan graphs '' ) or even some function @xmath43 .
calculations such as those in @xcite suggest that it does , even for some negative function @xmath44 .
examples of `` ramanujan graphs , '' i.e. , graphs where @xmath45 except @xmath46 ( and , at times , @xmath47 when @xmath48 ) have been given in @xcite where @xmath0 is one more than an odd prime or prime power .
theorem [ th : main ] demonstrates the existence of `` nearly ramanujan '' graphs of any even degree .
we shall soon address odd @xmath0 , as well .
another interesting question arises in the gap between @xmath49 and @xmath50 in theorem [ th : main ] in the case where @xmath40 is an integer ; it is almost certain that one of them can be improved upon . in the language of section 4 of this paper
, @xmath49 is the smallest order of a supercritical tangle , and @xmath50 that of a hypercritical tangle ; a gap between @xmath49 and @xmath50 can only occur when there is a critical tangle of order smaller than that of any hypercritical tangle .
previous bounds of the form @xmath51 include @xmath52 of the author ( see @xcite ) , which is slight improvement over the broder - shamir bound of @xmath53 ( see @xcite ) .
asymptotically in @xmath0 , the bounds @xmath54 of kahn and szemerdi ( see @xcite , here @xmath55 is some constant ) and @xmath56 of the author ( see @xcite and see equation ( [ eq : rho ] ) for the more precise bound ) are improvements over the first two bounds . the value of @xmath49 in theorem [ th : main ] depends on the particular model of a random graph .
indeed , consider the model @xmath57 of a random graph , which is like @xmath29 except that we insist that each @xmath25 be one of the @xmath58 permutations whose cyclic decomposition consists of one cycle of length @xmath16 .
the same methods used to prove theorem [ th : main ] will show the following variant .
[ th : mainh ] theorem [ th : main ] holds with @xmath29 replaced by @xmath57 and @xmath59 and @xmath60 , except that when @xmath61 we take @xmath62 .
once again , @xmath63 , unless a certain expression , in this case @xmath64 ( excepting @xmath61 ) , is an integer .
note that for @xmath57 , the value of @xmath49 is roughly twice as large as that for @xmath29 for @xmath0 large .
next consider two more models of random @xmath0-regular graphs ; in these two models @xmath0 may be even or odd .
let @xmath65 , for positive integers @xmath66 with @xmath16 even , be the model of a random @xmath0-regular graph formed from @xmath0 random perfect matchings on @xmath4 . for an odd positive integer @xmath16 , let a _ near perfect matching _
be a matching of @xmath67 elements of @xmath4 ; such a matching becomes a @xmath68-regular graph if it is complemented by a single half - loop to a diagonal entry of the adjacency matrix ) can see section 2 of this paper or @xcite .
] at the unmatched vertex .
taking @xmath0 independent such @xmath68-regular graphs gives a model , @xmath69 , of a @xmath0-regular graph on @xmath16 vertices for @xmath16 odd .
[ th : maini ] theorem [ th : mainh ] holds with @xmath57 replaced by @xmath65 and with no @xmath61 exception ( i.e. , @xmath70 for @xmath61 ) .
theorem [ th : main ] holds with @xmath29 replaced by @xmath69 .
we can assert the truth of the alon conjecture on more models of random graphs by using results on continguity and related notions .
consider two families of probability spaces , @xmath71 and @xmath72 over the same sets @xmath73 and sigma - algebras @xmath74 ; denote @xmath75 and @xmath76 .
we say that @xmath77 _ dominates _
@xmath78 if for any family of measurable events , @xmath79 ( i.e. , @xmath80 ) , we have @xmath81 as @xmath82 implies @xmath83 as @xmath82 .
we say that @xmath77 and @xmath78 are _ contiguous _ if @xmath77 dominates @xmath78 and @xmath78 dominates @xmath77 .
fix an @xmath5 and an integer @xmath84 .
let @xmath85 be any family of probability spaces of @xmath0-regular graphs on @xmath16 vertices ( possibly defined for only certain @xmath16 ) that is dominated by @xmath29 , @xmath57 , @xmath65 , or @xmath69 .
then for @xmath14 in @xmath86 we have that with probability @xmath87 ( as @xmath82 ) for all @xmath88 with @xmath89 we have @xmath90 there are a lot of results regarding contiguity and ( at least implicitly ) domination ; see @xcite and the references there .
for example , if @xmath91 is the restriction of @xmath29 to those graphs without self - loops , then for @xmath2 it is known that ( 1 ) @xmath91 and @xmath57 are contiguous ( by @xcite and previous work ) , and ( 2 ) @xmath29 dominates @xmath91 ( easy , since a self - loop occurs in @xmath29 with probability bounded away from @xmath68 for fixed @xmath0 ) . thus the alon conjecture for @xmath29
implies the same for @xmath57 ( but this contiguity and/or domination approach does not give as tight a bound on the probablity that @xmath92 fails to hold as is given in theorem [ th : mainh ] ) .
also , @xmath29 is contiguous with the `` pairing '' or `` configuration '' model of @xmath0-regular ( pseudo)graphs ( see @xcite ) ; it follows that the alon conjecture holds for the latter model , and thus ( see @xcite , especially the beginning of section 2 and corollary 4.17 ) the conjecture holds for @xmath16 ( and @xmath0 ) even for @xmath65 or the uniform measure on all @xmath0-regular ( simple ) graphs on @xmath16 vertices .
our method for proving theorems [ th : main ] , [ th : mainh ] , and [ th : maini ] is a variant of the well - known `` trace method '' ( see , for example @xcite ) originated by wigner , especially the author s refinement in @xcite of the beautiful broder - shamir style of analysis in @xcite .
the standard trace method involves taking the expected value of the trace of a reasonably high power , where @xmath10 depends on @xmath0 and on aspects of the method .
] of the adjacency matrix . in our situation
we are unable to analyze this trace accurately enough to prove theorem [ th : main ] , as certain infinite sums involved in our analysis diverge ( for example , the infinite sum involving @xmath93 and @xmath94 just above the middle of page 351 in @xcite , for types of order @xmath95 ) .
this divergence is due to certain `` tangles '' that can occur in a random graph and can adversely affect the eigenvalues ( see sections 2 and 4 ) . to get around these `` tangles ''
we introduce a _
selective trace_. we briefly sketch what a selective trace is in the next paragraph .
recall that a closed walk about a vertex , @xmath96 , is a walk in the graph beginning and ending at @xmath96 . recall that the trace of the @xmath97-th power of the adjacency matrix equals the sum over all @xmath96 of the number of closed walks about @xmath96 of length @xmath97 .
the _ @xmath97-th irreducible trace _ ( used in both @xcite and @xcite ) is the same sum as the @xmath97-th power trace , except that we require the closed walks to be _
irreducible _ ,
i.e. , to have no edge traversed and then immediately thereafter traversed in the opposite direction .
a selective trace is a sum like an irreducible trace , but where we further require that the walk have no small contiguous piece that `` traces out '' a `` supercritical tangle '' ( the notions of `` tracing out '' and `` supercritical tangles '' will be defined later ; roughly speaking , a `` supercritical tangle '' is a small graph with many cycles ) .
since these `` tangles '' occur with probability at most proportional to @xmath98 , with @xmath99 as in theorem [ th : main ] , the selective trace usually agrees with the standard `` irreducible '' trace . analyzing the selective traces
involves a new concept of the `` new type , '' which is a refinement of the `` type '' of @xcite .
we caution the reader about the notation used here . in this paper
we work with only @xmath0-regular graphs . in @xcite @xmath100-regular graphs
were studied ; in @xcite the graphs are usually @xmath100-regular , although for a part of section 3 the graphs are @xmath0-regular .
we also caution the reader that here we use the term `` irreducible '' ( as used in @xcite and , for example , in the text @xcite ) as opposed to `` reduced '' ( which is quite common ) or `` non - backtracking '' ( sometimes used in @xcite ) in describing walks and related concepts .
we hope to generalize or `` relativize '' the theorems here to theorems about new eigenvalues of random covers ( see @xcite for a relativized broder - shamir theorem ) . in this paper
we occasionally go out of our way to use a technique that will easily generalize to this setting .
the rest of this paper is organized as follows . in section 2
we review the trace method used in @xcite and explain why it requres modification to prove alon s conjecture ; as a byproduct we establish the part of theorem [ th : main ] involving @xmath50 . in section
3 we give some background needed for some technical details in later sections . in section 4
we formalize the notion of a tangle , and discuss their properties ; we prove the part of theorem [ th : mainh ] and [ th : maini ] involving @xmath50 . in section 5
we describe `` types '' and `` new types , '' explaining how they help to estimate `` walk sums ; '' walk sums are generalizations of all notions of `` trace '' used here . in section 6
we describe the `` selective trace '' used in this paper ; we give a crucial lemma that counts certain types of selective closed walks in a graph . in section 7
we explain a little about `` @xmath0-ramanujan '' functions , giving a theorem to be used in section 14 that also illustrates one of the main technical points in section 8 .
in section 8 we prove that certain selective traces have an asymptotic expansion ( in @xmath101 ) whose coefficients are `` @xmath0-ramanujan . '' in section 9 we show that the expansion in section 8 still exists when we count selective traces of graphs not containing any finite set of tangles of order @xmath102 . in section 10
we introduce strongly irreducible traces , that simplify the proofs of the main theorems in this paper . in section 11
we prove a crucial lemma that allows us to use the asymptotic expansion to make conclusions about certain eigenvalues ; this lemma sidesteps the unresolved problem of ( even roughly ) determining the coefficients of the asymptotic expansion ( in @xcite we actually roughly determine the coefficients for the shorter expansion developed there ) . in section 12 we prove the magnification ( or `` expansion '' ) properties needed to apply the sidestepping lemma of section 11 . in section 13
we complete the proof of theorem [ th : main ] .
in section 14 we complete the proof of theorems [ th : mainh ] and [ th : maini ] , giving general conditions on a model of random graph that are sufficient to imply the alon conjecture . in section 15
we make some closing remarks .
we mention that the reader interested only in the alon conjecture for only @xmath29 ( i.e. , the first part of theorem [ th : main ] ) need not read sections 2 , 4 ( assuming a willing to believe lemma [ lm : order_increases ] ) , and 14 and subsections 3.7 , 3.8 , 5.4 , and 6.4 .
section 2 explains the problems with the trace method encountered in @xcite .
subsections 3.7 and 3.8 and section 4 concern themselves with the second part of theorem [ th : main ] ( the close to matching bound on how many graphs fail the @xmath103 bound ) .
subsection 5.4 explains the new aspects in our approach to the alon conjecture ; this subsection is not essential to the exposition ( but probably is helpful ) .
subsection 6.4 and section 14 involve the alon conjecture for @xmath104 . throughout the rest of this paper
we will work with @xmath29 unless we explicitly say otherwise , and we understand @xmath0 to be a fixed integer at least @xmath105 . at times we insist that @xmath0 be even ( for example , in dealing with @xmath29 and @xmath57 ) .
in this section we summarize the trace method used in @xcite , and why this method can not prove alon s conjecture . during this section
we will review some of the ideas of @xcite , involving asymptotic expansions of various types of traces , which we modify in later sections to complete our proof of alon s conjecture .
we begin by recalling the trace method as used in @xcite , and why it did not yield the alon conjecture .
the trace method ( see @xcite , for example ) determines information on the eigenvalues of a random graph in a certain probability space by computing the expected value of a sufficiently high power of the adjacency matrix , @xmath106 ; this expected value equals the expected value of the sum of that power of the eigenvalues , since @xmath107 now @xmath108 may also be interpreted as the number of closed walks ( i.e. , walks ( see section 3.1 ) in the graph that start and end at the same vertex ) of length @xmath97 .
now restrict our discussion to @xmath29 .
a word , @xmath109 , of length @xmath97 over the alphabet @xmath110 ( i.e. each @xmath111 ) , determines a random permutation , and the @xmath112-th entry of @xmath113 , is just the number of words , @xmath114 , of length @xmath97 , taking @xmath88 to @xmath115 . but given a word , @xmath114 , the probability , @xmath116 , that @xmath114 takes @xmath88 to @xmath88 is clearly independent of @xmath88 .
hence we have @xmath117 } = n \sum_{w\in\pi^k } p(w)\ ] ] in @xcite , broder and shamir estimated the right - hand - side of the above equality to obtain an estimate on @xmath118 .
this analysis was refined in @xcite .
we review the ideas there .
first , a word , @xmath114 , is said to be _ irreducible _ if @xmath114 contains no consecutive occurence of @xmath119 .
it is well - known that any word , @xmath114 , has a unique _ reduction _ to an irreducible word elements ( see @xcite , sections 1 and 7 of chapter 1 ) .
alternatively , see proposition 2.5 of @xcite or theorem 1 of @xcite ( this theorem says that a free group on a set , @xmath120 , is in one - to - one correspondence with the set of reduced words , @xmath121 , which means that every word over @xmath121 has a unique corresponding reduced word ; here `` reduced '' is our `` irreducible '' ) . ] ( or reduced word ) , @xmath122 , obtained from @xmath114 by repeatedly discarding any consecutive occurences of @xmath123 and @xmath124 in @xmath114 , and @xmath125 .
similarly a walk is said to be _
irreducible _ if it contains no occurrence of a step along an edge immediately followed by the reverse step along that edge .
similarly , every irreducible walk has a unique reduction .
let @xmath126 be the set of irreducible words of length @xmath97 , and let @xmath127 be the number of irreducible closed walks of length @xmath97 in @xmath14 in terms of @xmath14 , but we shall soon see ( lemma [ lm : chebyshev ] ) that @xmath127 can be defined as a polynomial in @xmath106 and @xmath0 . ] .
we shall see that to evaluate the expected value of @xmath128 it suffices , in a sense ( namely that of equation ( [ eq : other_way ] ) below ) , to evaluate @xmath129 } = n\sum_{w\in{{\rm irred}_{k } } } p(w),\ ] ] for @xmath130 .
it is easy to see that for any fixed word , @xmath114 , we have a power series expansion @xmath131 ( see , for example , theorem [ th : exp_polys ] ) . as examples , we note that for a random permutation , @xmath132 , on @xmath133 , the probability that the sequence @xmath134 first returns to @xmath68 at @xmath135 ( i.e. , the probability that @xmath68 lies on a cycle of length exactly @xmath97 ) is @xmath101 for @xmath136 .
it follows that @xmath137 , where @xmath138 is the number of positive integral divisors of @xmath139 , assuming @xmath140 . in this example
@xmath141 involves number theoretic properties of @xmath139 . for a second example , we first remark that @xmath142 maps a fixed vertex to a different vertex with probability @xmath143 , and to each of the @xmath67 different vertices with the same probability .
it is then easy to see that @xmath144 provided that @xmath145 are at most @xmath16 . if @xmath145 are at least @xmath30 , then the @xmath146 are non - zero for @xmath147 and involve number theoretic functions of @xmath145 for @xmath21 .
so set @xmath148 ( we easily see @xmath149 for @xmath150 and @xmath151 and so @xmath152 for @xmath151 ) . [
de : ram ] a function , @xmath153 , on positive integers , @xmath97 , is said to be _ @xmath0-ramanujan _ if there is a polynomial @xmath154 and a constant @xmath155 such that @xmath156 for all @xmath97 .
we call @xmath157 the _ principal term _ of @xmath158 , and @xmath159 the _ error term _ ( both terms are uniquely determined if @xmath160 ) . in @xcite it was shown ( among other things ) that for all @xmath161 we have that @xmath162 as above is @xmath0-ramanujan .
this , it turns out , gives a second eigenvalue bound of roughly @xmath163 for a universal constant , @xmath55 .
we now explain why . a standard counting and expansion argument is given in @xcite ( specifically theorem 3.1 there ) to establish the following lemma .
[ lm : counting ] for fixed even @xmath2 there is an @xmath164 such that with probability @xmath165 we have that a @xmath14 in @xmath29 has @xmath166 ( also with probability @xmath167 we have that @xmath168 ) .
next to @xmath1 , one ( or both ) of @xmath169 is the next largest eigenvalue in absolute value ; lemma [ lm : counting ] , by bounding the eigenvalues other than @xmath170 , will eventually be used to show that the @xmath162 of equation ( [ eq : g_idef ] ) are essentially determined , for small @xmath88 , by @xmath170 s `` contribution '' to @xmath127 ( see below ) .
next we establish the precise relationship between the traces of the @xmath113 and the @xmath127 .
let @xmath171 be the matrix whose @xmath112-th entry is the number of irreducible walks of length @xmath97 from @xmath88 to @xmath115 .
[ lm : chebyshev ] the @xmath171 are given by @xmath172 where @xmath173 is the degree @xmath97 polynomial given via @xmath174 ( which is a type of chebyshev polynomial ) ; alternatively we have @xmath175 , @xmath176 , and for @xmath177 we have @xmath178 also @xmath179 the proof is given in @xcite and @xcite ( specifically lemma 3.3 , page 356 , in @xcite ; the @xmath180 s there are the @xmath171 s here ) . in section 10
we shall use the fact that for fixed @xmath181 , @xmath182 satisfy the recurrence @xmath183 where @xmath184 is the `` shift in @xmath97 '' operator ( i.e. , @xmath185 ) to go the other way we note : @xmath186 where @xmath187 is the number of words of length @xmath97 that reduce to a given irreducible word of length @xmath88 .
thus @xmath188 [ lm : reducedwalks ] for @xmath189 even we have @xmath190 if @xmath191 and @xmath192 an exact formula for @xmath187 is given in @xcite .
a weaker estimate than the above lemma was used in @xcite .
the proof of this estimate is a simple spectral argument used by buck ( see @xcite ) . * proof * consider the adjacency matrix , @xmath193 , of the infinite @xmath0-regular tree , @xmath194 .
our proof requires the following sublemma .
[ lm : tree_norm ] @xmath193 has norm @xmath195 .
actually , it is well - known that the norm of @xmath193 is exactly @xmath19 ( see , for example , page 9 of @xcite , and the theorems on @xmath170 in section 3 here ) .
however , the proof below is simple and generalizes to many other situations ( and can be used in many cases to determine the exact norm of @xmath193 ) .
* proof * * ( of sublemma [ lm : tree_norm ] ) * let @xmath158 be a function in @xmath196 , i.e. , a function on the vertices of @xmath194 whose sum of squares of values is finite . fix a vertex , @xmath197 , of @xmath194 , to be viewed as the root of @xmath194 ; the _ children _ of a vertex , @xmath96 , are defined to be those vertices adjacent to @xmath96 and of greater distance than @xmath96 is to @xmath197 .
we have @xmath198 which , by cauchy - schwarz , is @xmath199 @xmath200 @xmath201 thus the norm of @xmath193 is @xmath202 .
@xmath203 ( to see that the norm of @xmath193 is exactly @xmath19 we find functions , @xmath158 , ( of finite support ) for which the applications of cauchy - schwarz in the above proof are `` usually '' tight .
namely , we can take @xmath204 for @xmath205 and @xmath206 otherwise , where @xmath50 is a parameter which tends to @xmath207 .
this technique works for some other graphs . )
we resume the proof of lemma [ lm : reducedwalks ] .
let @xmath96 be a vertex of @xmath194 , and let @xmath208 be the vertices of distance @xmath88 to @xmath50 .
then @xmath209 is the dot product of @xmath210 with @xmath211 , where @xmath212 denotes the characteristic function of @xmath213 , i.e. the function that is @xmath68 on @xmath213 and @xmath214 elsewhere .
so by cauchy - schwarz @xmath215 if @xmath216 , and otherwise @xmath217 .
@xmath203 notice that clearly @xmath218 , and so for @xmath219 lemma [ lm : reducedwalks ] is off by a multiplicative factor of roughly @xmath220 ; according to @xcite , the lemma [ lm : reducedwalks ] estimate of @xmath221 is off by roughly a factor of @xmath222 .
the roughness of lemma [ lm : reducedwalks ] is unimportant for our purposes .
now notice that by lemmas [ lm : counting ] and [ lm : chebyshev ] we have @xmath129 } = q_k(d)\bigl ( 1+n^{1-d}+o(n^{2 - 2d})\bigr ) + { \rm error},\ ] ] where @xmath223 it is easy to see ( see @xcite ) that @xmath224 , and for some @xmath225 we have @xmath226 for an absolute constant @xmath10 ( with any @xmath227 as in lemma [ lm : counting ] ) .
we wish to draw some conclusions about the principal term of the @xmath162 s .
we need the following lemma : for fixed @xmath228 there is a constant , @xmath10 , such that for @xmath151 we have that in @xmath29 @xmath129 } = g_0(k)+\frac{g_1(k)}{n } + \frac{g_2(k)}{n^2 } + \cdots + \frac{g_{r-1}(k)}{n^{r-1 } } + { \rm error},\ ] ] where @xmath229 * proof * this follows from the calculations on page 352 of @xcite ; for each @xmath88 , the @xmath230 in @xcite is the polynomial in the principal term of @xmath162 ( and we mean @xmath230 corresponds precisely to @xmath162 , not @xmath231 or @xmath232 ) .
( actually we shall later just above equation ( 21 ) ( page 352 ) could have been replaced with @xmath233 . ]
see that the @xmath234 in the error term estimate can be replaced by @xmath235 . )
@xmath203 we now take @xmath97 of order @xmath236 and use standard facts about expansion and expansion s control on eigenvalues ( namely our lemma [ lm : counting ] ) to conclude , as in @xcite , the following theorem .
[ th : f_ivalues ] let @xmath237 be @xmath0-ramanujan for some @xmath238 .
then the principal term of @xmath162 vanishes for @xmath239 , and the principal term of @xmath240 is @xmath241 .
* proof * see theorem 3.5 of @xcite .
@xmath203 we next apply lemma [ lm : reducedwalks ] to estimate the expected value of the trace of @xmath113 where @xmath97 is roughly @xmath242 as in @xcite , in order to obtain the following theorem . [
th : expansion_consequence ] with the same hypotheses as theorem [ th : f_ivalues ] , we have ( in @xmath29 ) @xmath243 } \le \rho^k\ ] ] for all @xmath244 , with @xmath245 as above , where @xmath246 and @xmath10 depends only on @xmath247 . * proof * first we take @xmath248 and find that @xmath249 can be taken as above . for smaller @xmath97 we appeal to jensen s inequality .
see @xcite for details .
@xmath203 from theorem 2.18 of @xcite we see that we can take @xmath250 as large as @xmath251 . with this value of @xmath250 , using equation ( [ eq : rho ] ) ,
we conclude that equation ( [ eq : expansion_consequence ] ) holds with @xmath252 for an absolute constant , @xmath55 , where @xmath253 is a quantity that for fixed @xmath0 tends to @xmath214 ( proportional to @xmath254 ) as @xmath82 .
whenever equation ( [ eq : expansion_consequence ] ) holds , then the expected value of @xmath255 is bounded by @xmath249 .
the alon conjecture would be implied if one could obtain @xmath256 for any @xmath5 .
in this subsection we will show that for some @xmath257 , @xmath162 is not @xmath0-ramanujan .
we similarly show the part of theorem [ th : main ] involving @xmath50 .
both these facts are due to the possible occurrence of what we call _ tangles_. tangles and avoiding them are the main themes in this paper .
we begin by describing an example of a _ tangle _ , and its effect on eigenvalues and traces .
consider @xmath29 for a fixed , even @xmath0 and a variable @xmath16 which we view as large .
fix an integer @xmath139 with @xmath258 ( assume @xmath2 ) .
consider the event , @xmath259 , that @xmath260 clearly @xmath259 occurs with probability @xmath261 .
assume @xmath259 occurs in a fixed @xmath0-regular graph , @xmath28 .
let @xmath262 be the set of vertices of distance at least @xmath30 to the vertex @xmath68 ; @xmath262 is a random set of vertices , but always of size at least @xmath263 . consider the characteristic functions @xmath264 , where @xmath212 is the function that is @xmath68 on the vertices in @xmath213 , and @xmath214 elsewhere .
let @xmath265 be the rayleigh quotient associated to the adjacency matrix , @xmath106 , of @xmath14 .
the following lemma is well - known .
[ lm : fix_bug ] let @xmath106 be a real , symmetric matrix .
let @xmath266 be nonzero vectors with @xmath96 orthogonal to @xmath267 and @xmath268 .
then @xmath269 * proof * let @xmath77 denote the @xmath270 on the right - hand - side of the above inequality . by the hypothesis of the lemma , @xmath271 ; along with the symmetry of @xmath106 , we have @xmath272 . if @xmath273 with @xmath274 scalars , we have @xmath275 it follows that the rayleigh quotient of any vector in the span of @xmath267 and @xmath96 is at least @xmath77 .
since this span is a two - dimensional subspace , the max - min principle implies that @xmath276 .
@xmath203 we intend apply the above lemma with @xmath277 and @xmath278 .
@xmath279 ( since @xmath280 counts twice the number of edges with both endpoints in @xmath213 ) .
also @xmath281 @xmath282
so @xmath283 viewing @xmath284 as fixed .
since @xmath285 and @xmath286 are supported in the neighbourhood of distance at most @xmath68 from the vertex @xmath68 , @xmath287 is orthogonal to both of them .
lemma [ lm : fix_bug ] now implies @xmath288 next consider the probability that @xmath289 for at least one value of @xmath250 .
inclusion / exclusion shows that the probability of this is at least @xmath290 @xmath291 @xmath292 we summarize the above observations .
[ th : weak ] for fixed integer @xmath139 with @xmath258 , we have that @xmath293 for sufficiently large @xmath16 with probability at least @xmath294 .
the proof of the above theorem did not exploit the fact that aside from having @xmath139 self - loops , the vertex @xmath68 is still adjacent to @xmath295 other vertices of a @xmath0-regular graph .
we seek a stronger theorem that exploits this fact .
[ th : improved_bound ] for fixed integers @xmath296 and @xmath2 , with @xmath297 and @xmath298 , we have that @xmath299 for sufficiently large @xmath16 with probability at least @xmath294 .
( notice that theorem [ th : weak ] would require @xmath300 for the same conclusion . ) we are very interested to know if one can prove theorem [ th : improved_bound ] when @xmath301 for integer @xmath139 and even integer @xmath0 .
we expect not .
( the situation where @xmath301 gives rise to what we will call a `` critical tangle , '' and @xmath302 to a `` supercritical tangle , '' in section 4 . ) * proof * note : in theorems [ th : remarkable ] and [ th : key_to_s ] we give a proof of a generalization of this theorem requiring far less calculation ( but requiring more machinery ) .
again , assume that @xmath303 for @xmath304 and some @xmath197 .
it suffices to show @xmath299 for sufficiently large @xmath16 , under the assumption that @xmath305 .
let @xmath306 by cauchy - schwarz we have @xmath307 ( equality does not hold , because @xmath308 since @xmath305 ) .
let @xmath309 denote @xmath96 s distance to @xmath197 .
for a fixed @xmath250 , let @xmath310 where @xmath311 .
it is easy to check that @xmath312 provided that @xmath313 ( this includes the case @xmath314 , since @xmath315 , but checking @xmath314 is a bit different from the other cases ) .
it follows that @xmath316 where @xmath317 but @xmath318 if @xmath319 , and also @xmath320 ( since , by induction , the number of vertices at distance @xmath250 from @xmath96 is at most @xmath321 . )
so @xmath322 can be made arbitrarily close to @xmath68 by taking @xmath250 sufficiently large ( since @xmath323 ) .
let @xmath324 be the rayleigh quotient of @xmath106 .
the last paragraph , especially equations ( [ eq : ray1 ] ) and ( [ eq : ray2 ] ) implies that for @xmath325 there is an @xmath326 such that @xmath327 .
so let @xmath328 be those vertices of distance @xmath68 or @xmath214 to the support of @xmath158 ; the size , @xmath329 , of @xmath328 is bounded as a function of @xmath0 and @xmath250 .
the function @xmath158 is orthogonal to @xmath330 and @xmath331 , and counting edges as before we see @xmath332 it follows that by taking @xmath16 sufficiently large , we can make @xmath333 ; since @xmath307 , we can choose @xmath334 , making @xmath299 .
@xmath203 theorem [ th : improved_bound ] proves the part of theorem [ th : main ] involving @xmath50 , by taking @xmath335 with @xmath139 as small as possible ( namely @xmath336 ) .
the analogous parts of theorems [ th : mainh ] and [ th : maini ] are slightly trickier , since the `` tangle '' involved has automorphisms ; we shall delay their proof ( see theorem [ th : other_ss ] ) until we give a more involved discussion of tangles in section 4 .
notice that our proof is really computing the norm of @xmath337 where @xmath338 is the @xmath0-regular graph with the vertex @xmath68 having @xmath139 self - loops , and which is a ( an infinite ) tree when these loops are removed .
the function @xmath158 as above shows that @xmath337 s norm is at least @xmath339 .
the statement and proof of sublemma [ lm : tree_norm ] for the @xmath0-regular tree applies to the above tree ( with @xmath340 replacing @xmath64 , and with @xmath339 replacing @xmath19 ) . in this way
our proof of theorem [ th : improved_bound ] is very much like one proof of the alon - boppana theorem ( see @xcite ) .
the discussion in this section leads to the following theorem .
[ th : nottreelike ] there is an absolute constant ( independent of @xmath0 ) , @xmath55 , such that the @xmath162 of equation ( [ eq : g_idef ] ) can not be @xmath0-ramanujan for all @xmath341 . * proof * we fix an integer @xmath50 to be chosen later with @xmath342 set @xmath343 and apply theorem [ th : improved_bound ] . since @xmath344 with @xmath339 as in equation ( [ eq : alpha ] ) , for @xmath97 even we have that @xmath345 with probability at least @xmath346 .
thus @xmath347 } ^{1/k } \ge \bigl(n^{-s}+o(n^{-s-1})\bigr)^{1/k } 2s.\ ] ] according to theorem [ th : expansion_consequence ] , if @xmath348 are @xmath0-ramanujan for some @xmath238 , then we have @xmath347 } ^{1/k } \le \rho,\ ] ] with @xmath249 as in equation ( [ eq : rho ] ) , provided that @xmath97 is even and bounded by @xmath349 as in theorem [ th : expansion_consequence ] .
for some constant @xmath10 we have that for any @xmath55 and for @xmath350 , equation ( [ eq : rho ] ) gives @xmath351 in other words , @xmath352 take @xmath97 even and as close to @xmath349 as possible ; note that by equation ( [ eq : hnrd ] ) , @xmath353 hence , taking @xmath82 in equation ( [ eq : long ] ) implies that for a universal constant , @xmath354 , we have @xmath355 choosing @xmath356 and dividing by @xmath30 yields @xmath357 choosing @xmath55 large enough so that @xmath358 makes this impossible .
@xmath203 we have proven that not all @xmath162 are @xmath0-ramanujan for @xmath359 where @xmath350 .
notice that in our terminology , theorem 2.18 of @xcite says that @xmath162 is @xmath0-ramanujan for @xmath360 ; again , for each @xmath88 the @xmath230 in @xcite is the polynomial in the principal part of our @xmath162 .
this leaves the question of whether theorem [ th : nottreelike ] can be improved to an @xmath250 value closer to @xmath361 ; we conjecture that it can be improved to @xmath362 , and that the tangle with @xmath336 already `` causes '' this @xmath363 ( or a lower one ) to fail to be @xmath0-ramanujan .
this also leaves open the question of what can be said about the @xmath162 that are not @xmath0-ramanujan .
perhaps such @xmath364 are a sum of @xmath365 over various @xmath78 with some added error term . in this paper
we avoid this issue , working with a modified trace ( i.e. , `` selective '' traces ) for which the corresponding @xmath162 are @xmath0-ramanujan .
in this section we review some ideas and techniques from the literature needed here .
we also give some convenient terminology that is not completely standard .
we use some nonstandard notions in graph theory , and we carefully explain all our terminology and notions here . a directed graph , @xmath14 , consists of a set of vertices , @xmath366 , a set of edges , @xmath367 , and an incidence map , @xmath368 ; if @xmath369 we will write @xmath370 , say that @xmath371 is of _ type _
@xmath372 , and say that @xmath371 originates in @xmath267 and terminates in @xmath96 .
( if @xmath88 is injective then it is usually safe to view @xmath373 as a subset of @xmath374 , and we say that @xmath14 has no multiple edges . ) the adjacency matrix , @xmath375 , is a square matrix indexed on @xmath376 , where @xmath377 counts the number of edges of type @xmath372 .
the outdegree at @xmath378 is the row sum of @xmath106 at @xmath96 , i.e. , the number of edges originating in @xmath96 ; the indegree is the column sum or number of edges terminating in @xmath96 . a graph , @xmath14 ,
is a directed graph , @xmath379 , such that each edge of type @xmath372 is `` paired '' with an `` opposite edge '' of type @xmath380 ; in other words , we have a map @xmath381 , such that @xmath382 is the identity , and if @xmath383 has @xmath370 , then @xmath384 ; in other words , the edges @xmath385 come in `` pairs , '' except that a self - loop , i.e. , an @xmath383 with @xmath386 , can be paired with itself ( which is a `` half - loop '' in the terminology of @xcite ) or paired with another self - loop at @xmath96 ( which is a `` whole - loop '' ) .
half - loops about @xmath96 contribute @xmath68 to the adjacency matrix entry at @xmath387 ( i.e. , contribute @xmath68 to @xmath388 ) , and whole - loops contribute @xmath30 . in this paper
we primarily work with whole - loops , needing half - loops only in the model @xmath69 .
we refer to the _ ( undirected ) edges _ , @xmath389 , of a graph , @xmath14 , as the set of `` pairs '' of edges , @xmath390 .
@xmath14 s vertex set and adjacency matrix are just those of the directed graph , @xmath379 , i.e. , @xmath391 and @xmath392
. a _ numbering _ of a set , @xmath208 , is a bijection @xmath393 , where @xmath394 .
a _ partial numbering _ of a set , @xmath208 , is a numbering of some subset , @xmath395 , of @xmath208 ( we allow @xmath395 to be empty , in which case none of @xmath208 is numbered ) .
we can speak of a graph , directed or not , as having numbered or partially numbered vertices and/or edges .
a numbering can be viewed as a total ordering .
each letter @xmath396 has its associated inverse , @xmath397 , and every word @xmath109 over @xmath398 has its associated inverse , @xmath399 .
if @xmath400 is any set of words over @xmath398 , then a _ @xmath400-labelling _ of an undirected graph , @xmath14 , is a map or `` labelling '' @xmath401 such that @xmath402 for each @xmath403 .
for example , any graph @xmath404 automatically comes with a @xmath398-labelling , namely @xmath405 is labelled @xmath406 , and @xmath407 is labelled @xmath408 .
an _ orientation _ of an undirected graph , @xmath14 , is the distinguishing for each @xmath409 of one of the two directed edges corresponding to @xmath371 .
the following definition is special to this paper .
[ de : structural ] fix sets @xmath410 and a set of words , @xmath400 , over @xmath398 , with @xmath411 .
a _ structural map _ is a map @xmath412 .
a structural map defines a unique @xmath400-labelled , oriented graph , @xmath14 , with @xmath413 and @xmath414 , as follows : for each @xmath415 with @xmath416 , we form a directed edge of type @xmath372 labelled @xmath123 and declare it distinguished , and pair it with a directed edge of type @xmath380 labelled @xmath124 . in this paper
we will work with graphs that have large or infinite parts of them being paths or regular trees . in this case
we can easily eliminate all the vertices in these parts by working with `` variable - length graphs . ''
this leads to simpler calculations ( in theorem [ th : infinite_edge ] , that is crucial to lemma [ lm : finiteness ] , and in theorem [ th : crucial_cycle_count ] ) .
this also gives us a notion of regular tree of non - integral degree , in theorem [ th : remarkable ] .
recall ( see @xcite ) that a _ vlg _ or _ variable - length graph _
( respectively , _ directed vlg _ or _ directed variable - length graph _ ) is a graph ( respectively , directed graph ) with an assignment of a positive integer to each edge called the edge s _ length_. the length of a walk in a vlg is the sum of the lengths of its edges ( each length is counted the number of times the edge appears in the walk ) . a graph can be regarded as a vlg with all edge lengths @xmath68 .
a vlg whose edge lengths are all @xmath68 can be identified with its underlying graph .
a _ bead _ in a directed graph ( respectively graph ) is a vertex with indegree and outdegree @xmath68 ( respectively , degree @xmath30 ) and without a self - loop .
a _ beaded path _ is a path where every vertex except possibly the endpoints are beads .
[ de : subdivision ] let @xmath14 be a directed vlg .
to _ subdivide _ an edge , @xmath371 , from @xmath267 to @xmath96 and of length @xmath417 , in @xmath14 is to replace @xmath371 with a beaded path of length @xmath417 from @xmath267 to @xmath96 in @xmath14 ( introducing @xmath418 new vertices ) .
a _ subdivided form of @xmath14 _ is a graph , @xmath419 , obtained by subdividing all edges of @xmath14 .
the same definition is made for vlg s , omitting the word `` directed '' everywhere , provided no half - loops are of length @xmath30 or greater .
actually , we can define a notion of subdivision for all half - loops of odd length , but in this paper we use half - loops only of length @xmath68 . ] .
it should be clear that countings walks of certain types in a directed vlg , @xmath14 , should translate to an appropriate similar counting in @xmath419 , and vice versa .
we next define an opposite of subdivision , supression , and a vast generalization of supression , realization .
let @xmath14 be a strongly connected directed graph , with @xmath420 .
the _ realization of @xmath14 to with vertex set @xmath262 _ denoted @xmath421 , is the directed vlg on vertices @xmath262 with the following set of edges .
we create @xmath421 one edge from @xmath267 to @xmath96 ( for @xmath422 ) of length @xmath97 for each walk from @xmath267 to @xmath96 in @xmath14 of length @xmath97 that contains no @xmath262 vertices except as the first and last vertices .
( so self - loops or edges in @xmath14 involving @xmath262 vertices appear in @xmath421 , since we regard self - loops or edges as walks with no vertices except the first and last vertices . )
the notion of realization appears constrained coding theory ( called `` fusion '' in @xcite , for example , and not given a name in @xcite ; see also @xcite ) .
we remark that if @xmath423 contains a cycle , then @xmath421 has infinitely many edges .
let @xmath14 be a strongly connected directed graph , with @xmath424 a subset of beads in @xmath14 such that @xmath213 contains no cycle .
the _ supression of @xmath213 ( in @xmath14 ) _ , denoted @xmath425_{\rm sup}}}$ ] , is the realization of @xmath14 with vertex set @xmath426 .
the subdivision ( by the supressed vertices ) of a supression returns the original directed graph .
we remark that if @xmath14 is a @xmath398-labelled graph , then any supression in @xmath14 has a natural @xmath427-labelling , where @xmath428 is the set of words on @xmath398 of length @xmath102 .
let @xmath14 be a vlg ( directed or undirected ) .
for @xmath429 and a non - negative integer , @xmath97 , let @xmath430 denote the number of walks of length @xmath97 from @xmath267 to @xmath96 in @xmath14 .
we will use standard perron - frobenius theory ( see sections 1.3 and 7.1 in @xcite or chapters 1 and 6 in @xcite ) , which includes the rest of this paragraph .
assume that @xmath14 is strongly connected , i.e. , for each @xmath429 we have @xmath431 for some @xmath97 .
let @xmath432 be the period of @xmath14 , i.e. , the greatest common divisor of the lengths of all closed walks in @xmath14 .
then all limits @xmath433 exist and are all equal ( so independent of @xmath266 ) ; we define this common limit to be @xmath434 , the _ perron value _ of @xmath14 .
it is easy to see that @xmath435 , using that @xmath436 .
if @xmath14 is a finite graph , then @xmath434 is just the usual perron - frobenius ( largest ) eigenvalue of @xmath437 .
in directed graphs , supression , realization , and subdivision preserve walk counts ( i.e. , the @xmath438 s ) between appropriate vertices ( those present in the two graphs in question ) .
therefore these operations also preserve @xmath170 .
if @xmath14 is not strongly connected , we can define @xmath434to be the supremum of @xmath170 of all the strongly connected components of @xmath14 , or equivalently as the supremum over all @xmath439 .
one can equivalently define @xmath434 with @xmath440 replacing @xmath441 , where @xmath442 is the number of walks of length at most @xmath97 .
this is a sensible definition of @xmath434 when we allow non - integral edge lengths .
one can also extend all these defitions to graphs with positively weighted edges , where the weight of a walk becomes the product of its edge weights , and where @xmath441 or @xmath440 sums the weights of the walks .
shannon gives the following algorithm ( see @xcite , chaper 1 , section 1 ) for computing @xmath434 ( or the `` valence '' or `` capacity '' ) of a finite graph : let @xmath443 be the matrix whose @xmath112 entry is the sum of @xmath444 over all edge lengths , @xmath417 , of edges from @xmath88 to @xmath115 , with @xmath445 a formal parameter .
then @xmath434 is the reciprocal of the smallest real root in @xmath445 of @xmath446 in this section we explain variants of this theorem / algorithm that hold for infinite vlg s .
we first give some conventions that we will use with formal power series . by a _ non - negative power series _ we mean a series @xmath447 , with @xmath448 non - negative reals .
we say that @xmath158 is the _ generating function _ of the @xmath448 . at times
we view @xmath158 as a formal power series , but we will also have cause to evaluate @xmath158 at non - negative reals ( as a possibly diverging infinite sum ) ; it is easy to see that if @xmath449 converges for a positive @xmath450 , then @xmath158 is continuous on @xmath451 $ ] , and if @xmath449 diverges then @xmath452 gets arbitrarily large as @xmath453 approaches @xmath450 . for such an @xmath158 ,
@xmath158 s radius of convergence is @xmath454 the function @xmath158 has a singularity at @xmath455 , and @xmath456 diverges for @xmath457 . if the singularity at @xmath455 is a pole ( e.g. , when @xmath458 is a rational function ) , then @xmath459 as @xmath460 from the left .
: we do not identitfy a formal power series with any of its analytic extensions unless we specifically say so .
for example , @xmath461 has only the value @xmath462 for real @xmath463 unless we explicitly say to the contrary . for a vlg , @xmath14 , we set @xmath464 we have @xmath465 we say that a non - negative power series , @xmath466 , _ majorizes _ another one , @xmath467 , if @xmath468 for all @xmath469 .
if so , then @xmath470 for any @xmath471 $ ] ( with appropriate convensions on the value @xmath462 ) .
given vlg s , @xmath14 and @xmath338 , we say that @xmath14 _ majorizes _
@xmath338 if @xmath472 majorizes @xmath473 entry by entry ; equivalently , there is an endpoint preserving injection from @xmath474 to @xmath389 that does not increase edge lengths .
if so , clearly @xmath475 majorizes @xmath476 entry by entry .
[ th : shannon_infinite ] let @xmath14 be a strongly connected vlg , directed or undirected , on a countable number of vertices and edges .
the following are equal : 1 .
@xmath477 , and 2 . the radius of convergence of any entry of @xmath475 .
if @xmath14 has a finite number of vertices , then the above two numbers also equal the following two : 1 . the supremum of positive @xmath445 such that @xmath472 converges ( in each entry ) and has largest ( i.e. , perron - frobenius ) eigenvalue less than @xmath68 , and 2 . the supremum of positive @xmath445 such that @xmath472 converges and @xmath478 for all @xmath479 with @xmath480 . if @xmath14 has a finite number of vertices , and if the entries @xmath472 are all rational functions of @xmath445 , then the four above quantities equal 1 . the first positive solution to @xmath481 .
finally , all entries of @xmath472 will be rational functions of @xmath445 whenever @xmath14 is finite ( i.e. , has finitely many vertices and edges ) or is a vlg realization of finite graph . * proof * ( 1)=(2 ) : clear from equation ( [ eq : m ] ) .
( 2)=(3 ) : let @xmath482 be a non - negative matrix .
according to perron - frobenius theory , all the eigenvalues of @xmath482 are of absolute value at most @xmath483 , with equality only when the eigenvalue has the same algebraic and geometric multiplicity .
it now follows from jordan canonical form that a finite dimensional , non - negative matrix , @xmath482 , has largest ( perron - frobenius ) eigenvalue less than @xmath68 iff @xmath484 converges .
( 3)=(4 ) : if @xmath485 , then @xmath486 for all @xmath487 , and hence @xmath488 for such @xmath453 .
consider the first @xmath450 for which @xmath485 fails to hold ( such a @xmath450 exists by the continuity of @xmath170 as a function of its entries ) .
either @xmath489 does not converge , or else by continuity of @xmath170 we have @xmath490 and hence @xmath491 .
( 4)=(5 ) : by the above paragraph , if suffices to show that @xmath472 can not have a pole ( in any of its entries ) at the first @xmath450 for which @xmath485 fails to hold .
but if @xmath472 has a pole in some entry at @xmath492 , then for some @xmath493 and positive integer , @xmath97 , @xmath494 has a pole at @xmath492 , by the strong connectivity of @xmath14 . but
@xmath495 for any @xmath445 , and the latter tends to @xmath462 as @xmath496 from the left .
last part of the theorem : let @xmath497 be a realization of a finite vlg , @xmath14 , on the set @xmath424 . for @xmath498 , we shall calculate the @xmath499-entry of @xmath500 .
we claim this entry is the @xmath499-entry of @xmath501 , plus the @xmath499-entry @xmath502 where @xmath503 is the induced subgraph of @xmath14 on the vertex set @xmath504 , where @xmath96 is the vector whose entries correspond to the edges from @xmath505 to the vertices of @xmath503 and similarly for @xmath114 . since @xmath506
all have polynomial entries , we conclude that the @xmath499-entry of @xmath500 is a rational function .
@xmath203 now we give two examples to show that shannon s algorithm does not literally apply as is to infinite graphs .
let @xmath14 be a directed vlg with one vertex and @xmath507 edges of length @xmath97 for all @xmath151 .
then @xmath472 is a @xmath508 matrix with sole entry is @xmath509 . in this case
we have @xmath510 and @xmath458 diverges for any positive @xmath511 .
it is not hard to see that @xmath512 ( by theorem [ th : shannon_infinite ] , since @xmath513 converges for @xmath514 , which is the case when @xmath515 , and clearly diverges when @xmath516 where the series even for @xmath158 diverges ) .
but the expression @xmath517 fails to have a zero at @xmath518 .
next consider an undirected vlg , @xmath14 , whose vertex set is the integers , with each vertex having one self loop
. then @xmath517 expanded in a power series has an infinite @xmath445 coefficient ; if we simply multiply the diagonals together ( since this is a diagonal matrix ) , we get the infinite product of @xmath519 , which is @xmath214 for any @xmath520 ( yet @xmath521 ) .
while @xmath14 is not connected , we can add edges between @xmath88 and @xmath522 for all @xmath88 ; this yields a connected graph with similar problems .
finally we mention that theorem [ th : shannon_infinite ] must be modified when @xmath14 is not strongly connected .
indeed , consider a directed vlg on three nodes , such that the first node has a single self - loop , and the only other edges are edges from the second node to the third , with @xmath448 such edges of length @xmath97 .
then the most natural way to define @xmath434 is in terms of counting closed walks , so that @xmath521 .
but if @xmath523 diverges for any @xmath524 , then theorem [ th : shannon_infinite ] fails .
let @xmath525 be a sequence of finite vlg s on the same vertex set , @xmath376 , and the same edge set , @xmath373 .
let @xmath373 be partitioned into two sets , @xmath526 , such that the following holds : 1 . for each @xmath527 ,
the length of @xmath371 in @xmath525 is independent of @xmath88 , and 2 . for each @xmath528 ,
the length of @xmath371 in @xmath525 tends to infinity as @xmath529 .
the _ limit of the @xmath525 _ is the graph , @xmath530 , which is any @xmath525 with its @xmath531 edges discarded .
this simple remark is crucial for an important finiteness lemma ( lemma [ lm : finiteness ] ) .
[ th : infinite_edge ] with notation as above , @xmath532 * proof * by counting closed walks we see that @xmath533 ; this establishes the theorem with `` @xmath534 '' replacing `` @xmath535 '' . to see `` @xmath536 '' replacing `` @xmath535 '' ,
assume that @xmath537 for some @xmath538 $ ] .
clearly @xmath539 as @xmath529 .
so the continuity of @xmath170 on its entries implies @xmath540 for @xmath88 sufficiently large , and so @xmath541 .
now take a supremum over @xmath450 with @xmath537 .
@xmath203 let @xmath14 be an undirected graph with corresponding directed graph @xmath379 .
let @xmath542 be the graph with vertices @xmath385 and an edge from @xmath543 to @xmath544 iff @xmath545 forms an irreducible path in @xmath14 ; i.e. , @xmath543 and @xmath544 are not opposites ( i.e. , paired ) in @xmath14 , and @xmath543 terminates in the vertex where @xmath544 originates .
( therefore , if @xmath371 is a half - loop in @xmath14 , then there is no edge from @xmath371 to itself in @xmath542 . )
then walks in @xmath542 give `` irreducible '' ( or `` reduced '' or `` non - backtracking '' ) walks in @xmath14 .
a closed walk of length @xmath97 in @xmath542 gives a closed walk in @xmath14 ( with specified starting vertex ) that is _ strongly irreducible _ , meaning that the closed walk is irreducible and the last step in the closed walk is not the inverse of the first step .
we define the _ irreducible eigenvalues of @xmath14 _ to be those of @xmath542 , and we define the largest or perron - frobenius eigenvalue of @xmath542 to be @xmath546 , the _ largest irreducible eigenvalue of @xmath14_. for @xmath14 an undirected vlg , we may define @xmath542 as a vlg and hence define @xmath547 . in any graph ,
an irreducible walk that enters a beaded - path must directly traverse this path to its end ( any backward step makes the walk reducible ) .
it easily follows that if we subdivide edges in a vlg that are not half - loops ( i.e. , whole - loops or edges between distinct vertices ) , then counts of irreducible walks of a given length between @xmath548 vertices remain the same invariant .
a half - loop of length @xmath417 should be traversable zero or one time in a row ( but not twice or more in a row ) in an irreducible walk . for odd lengths , @xmath417 ,
this can be achieved by adding a path ( of length @xmath549 ) with a half - loop at the end . morally speaking , subdividing an edge in a vlg , @xmath14 ,
can be viewed as subdividing the corresponding edge pair in underlying directed graph , @xmath379 , and then gluing them together with an `` opposite '' pairing ( that glues the newly introduced vertices together ) ; for half - loops , we should glue a single directed self - loop of length @xmath417 to itself , via reflection about the middle ) ; for even length half - loops , that fact that reflection leaves the middle vertex fixed creates problems if we wish to remain in the catergory of undirected graphs@xmath550 . as a consequence we get the following simple but important proposition .
[ pr : ird_invar ] we have @xmath551 of a vlg is invariant under the subdivision of any set of edges devoid of half - loops ; in particular , @xmath551 is invariant under passing to a subdivided form of any vlg with all half - loops of length one .
similarly , @xmath551 of a graph is invariant under any supression .
we now state a theorem for use later in this paper ; the theorem requires a definition .
a connected graph , @xmath14 , is _ loopy _ if @xmath552 , or equivalently if @xmath14 contains an irreducible closed walk , or equivalently if @xmath14 is not a tree .
@xmath14 is _
@xmath68-loopy _ if @xmath14 is connected and the removal of any edge from @xmath14 leaves a graph each of whose connected components are loopy . [ th : loopy ] for a connected graph , @xmath14 , the following are equivalent : 1 .
@xmath14 is @xmath68-loopy , 2 .
@xmath542 is strongly connected , and 3 .
@xmath14 is not a cycle and all vertices in @xmath14 have degree at least @xmath30 . in particular , if @xmath14 is connected and @xmath0-regular for @xmath17 , then @xmath542 is strongly connected .
condition ( 3 ) in the above theorem was pointed out to us by a referee . * proof * ( 2)@xmath553(1 ) : consider a directed edge , @xmath370 , of @xmath14 , and let @xmath554 be @xmath371 s opposite ( we permit @xmath555 , i.e. , the case of a half - loop ) .
if the removal of @xmath371 leaves the connected component of @xmath96 being a tree , then there is no irreducible walk from @xmath371 to @xmath554 . on the other hand , if this connected component is not a tree , then a cycle in this component about @xmath96 of minimum length is irreducible , which then extends to an irreducible walk from @xmath371 to @xmath554 . to summarize ,
if @xmath14 is not @xmath68-loopy , then @xmath542 is not strongly connected ; otherwise , each edge has an irreducible path to its opposite , i.e. , each edge is connected to its opposite in @xmath542 .
( 1)@xmath553(2 ) : assume @xmath14 is @xmath68-loopy .
if @xmath556 are two distinct , unpaired directed edges originating in the same vertex , @xmath96 , then an irreducible path from @xmath543 to its opposite followed by @xmath544 shows that @xmath543 and @xmath544 are connected by an irreducible path .
thus any two edges that share a vertex are strongly connected in @xmath542 . finally if @xmath556 are two undirected edges without a common vertex , then a shortest path connecting them gives an irreducible path from some orientation of @xmath543 to some of @xmath544 . by the above
we can follow them with irreducible paths to the edge of opposite orientation .
thus any two edges that do not share a vertex are strongly connected in @xmath542 .
so @xmath542 is strongly connected .
( 1)@xmath553(3 ) since an isolated vertex is loopy , a @xmath68-loopy graph has all vertices of degree at least @xmath30 .
also a @xmath68-loopy graph can not be a single cycle , since removing any one edge would give a tree .
( 3)@xmath553(1 ) any tree either ( i ) is an isolated vertex , or ( ii ) is a path with two vertices of degree one , or ( iii ) has at least three vertices of degree @xmath68 ( this follows by considering two vertices , @xmath266 , of maximum distance in a tree , and then taking a vertex of maximum distance to the unique path joining @xmath267 to @xmath96 ) .
consider the tree , @xmath194 , obtained as the non - loopy connected component by removing an edge , @xmath371 , from a graph @xmath14 that is non - loopy .
if @xmath194 is the only connected component , @xmath371 has both its endpoints in @xmath194 , and if not then @xmath194 has only one endpoint in @xmath194 ; in either situation , considering the cases ( i)(iii ) , it is easy to check that @xmath14 is a cycle or has at least one vertex ( a vertex in @xmath194 ) of degree @xmath68 .
@xmath203 in this subsection we recall some facts about @xmath170 and closed walks of graphs , either indicating the proofs or giving references .
this section is geared to infinite graphs , since most of the facts are very easy when the graph is finite .
let @xmath14 be a graph of bounded degree , i.e. there is an @xmath250 such that the degree of each vertex is at most @xmath250 .
then @xmath437 , @xmath14 s adjacency matrix , is a bounded linear operator ( bounded by @xmath250 ) on @xmath557 , the square summable functions on @xmath14 s vertices .
the next theorem shows that @xmath558 , the norm of the operator @xmath558 , equals @xmath434 as was defined by counting closed walks about a vertex of at most some given length .
assume @xmath14 is connected . for any vertex , @xmath96 , of @xmath14 and positive integer , @xmath97 , recall that @xmath559 is the number of closed walks of length @xmath97 in @xmath14 about @xmath96 .
then @xmath560 , and @xmath561^{1/(2r)}\ ] ] exists and equals @xmath558 .
since the limit above equals @xmath434 , we see that @xmath562 . * proof * see @xcite .
@xmath203 we also need the following simple fact . for every @xmath5 there is an @xmath563 such that @xmath564 , where @xmath158 has finite support . * proof * by definition of norm , there is a @xmath565 with @xmath566 . for any @xmath567 we may write @xmath568 where @xmath569 is of finite support and @xmath570 .
it is easy to see ( using the fact that @xmath106 is bounded ) that if @xmath78 is sufficiently small we can take @xmath571 to satisfy the above theroem .
@xmath203 = 2.5 in let @xmath572 be a finite connected graph with each vertex of degree at most @xmath0 for an integer @xmath160 .
by @xmath573 we mean the unique ( up to isomorphism ) undirected graph , @xmath14 , that has an inclusion @xmath574 such that @xmath14 is @xmath0-regular and such that @xmath14 becomes a forest when we remove ( the image under @xmath575 of ) @xmath572 s edges .
we have seen an example of this construction in section 2 , where @xmath572 is one vertex with @xmath139 self - loops , in the proof of theorem [ th : improved_bound ] .
see figure [ fg : dtree ] for another example . in the category of @xmath0-regular graphs with a @xmath572 inclusion
, @xmath573 is none other than the universal cover .
the methods used to prove theorems [ th : main ] , [ th : mainh ] , and [ th : maini ] suggest the following curious theorem .
[ th : remarkable ] let @xmath17 , and let @xmath572 be a finite connected graph with each vertex of degree @xmath576 .
then @xmath577 the same is true for any real @xmath160 , provided that @xmath578 is interpreted with an appropriate analytic continuation in @xmath0 ( described below ) . before proving the theorem
, we describe the analytic continuation to which we refer .
let @xmath579 be an undirected rooted tree , where every vertex has @xmath580 children , with @xmath581 an integer ( for now ) .
( @xmath579 has degree @xmath580 at the root and degree @xmath0 elsewhere , so @xmath579 is not regular . )
let @xmath582 for @xmath583 be the number of walks in @xmath579 from the root to itself that never pass through the root except at the beginning and end .
it is easy to see that @xmath584 satisfies the recurrence @xmath585 ( since the walks counted by @xmath208 take one step , in @xmath580 possible ways , to a child of the root , followed by some number ( possibly zero ) of @xmath208 walks , followed by the step back to the root ) ; so @xmath586 , so that near @xmath587 @xmath588 one can alternatively say that if @xmath338 is the realization of @xmath579 with vertex set @xmath589 , where @xmath197 is the root , then the @xmath508 matrix @xmath473 has entry @xmath590 . similarly , if @xmath579 were modified to have @xmath591 children at the root ( but every other vertex with @xmath580 children , as before ) , then @xmath473 would have entry @xmath592 . by @xmath578 for @xmath581 real ( assuming each degree in @xmath572 is @xmath576 ) we mean @xmath170 of the vlg formed from @xmath572 with an additional self - loop about each vertex of degree @xmath250 that has `` formal weight '' @xmath593 ( this can be viewed as adding an infinite set of self - loops of given weights and lengths corresponding to this power series as the sum of terms representing @xmath582 edges of length @xmath16 ( even when @xmath582 is not an integer ) . ] , or can simply be viewed as a term to add to the diagonal of @xmath472 ) .
when @xmath0 in an integer , the realization of @xmath573 with vertex set @xmath594 is exactly this vlg , and by proposition [ pr : ird_invar ] we know @xmath551 remains the same . * proof * by @xcite , exercise 13 page 72 , we know that @xmath595 is given by @xmath596 of the smallest root , @xmath479 , of @xmath597 where @xmath598 is the diagonal matrix whose entry at vertex @xmath96 is the degree of @xmath96 , and @xmath599 is the identity matrix
. now @xmath600 it is then easy to verify that @xmath601 for @xmath602 ( note that @xmath603 is an increasing function in @xmath445 for as long as @xmath604 converges ) . but
@xmath605 for all positive real @xmath606 , where @xmath607 , and also @xmath608 .
so @xmath609 becomes non - invertible for a @xmath610 precisely when @xmath611 has eigenvalue @xmath68 for a @xmath612 $ ] .
now apply the equality of quantities ( 1 ) and ( 3 ) in theorem [ th : shannon_infinite ] .
@xmath203 note that the proof shows that if @xmath613 and @xmath614 is @xmath573 , then @xmath615 provided @xmath616 .
in section 2 we saw that a vertex with @xmath139 self - loops in a @xmath29 graph , with @xmath139 `` large , '' gives rise to a `` large '' second eigenvalue ( i.e. , larger than @xmath19 for sufficiently large @xmath139 , as @xmath82 ) . here
we generalize this observation to what we call a `` tangle . ''
our proof of the alon conjecture via a trace method must somehow overcome all `` hypercritical '' tangles .
given two @xmath398-labelled graphs , @xmath14 and @xmath338 , we say @xmath14 _ contains _ @xmath338 ( or @xmath338 occurs in @xmath14 )
if there is an inclusion @xmath617 that preserves the labelling ; the _ number of times _ a graph , @xmath338 , _ occurs _ in @xmath14 is the number of distinct consists of one edge joining two distinct vertices , @xmath266 , then the identity is considered distinct from the morphism interchanging the vertices . by the same principle ,
if @xmath338 has exactly @xmath97 automorphisms , then the number of times @xmath338 occurs in a graph is always a multiple of @xmath97 . ]
such @xmath575 .
a _ tangle _ ( or _ @xmath29-tangle _ ) is a @xmath398-labelled connected graph , @xmath572 , that is contained in some element of @xmath29 .
for example , in section 2 we studied the tangle with one vertex and @xmath139 self - loops labelled @xmath618 .
we define @xmath57- and @xmath65- and @xmath69-tangles similarly , with the following modifications to the meaning of the @xmath25 s . for @xmath57 ,
each of the @xmath3 independent @xmath25 s is uniform over all permutations whose cyclic decomposition consists of a single cycle . for @xmath65 , each of @xmath0 independent @xmath25 s
are uniform over all perfect matchings , i.e. , over all permutations that are involutions without fixed points . for @xmath69 , each of @xmath0 independent @xmath25 s are uniform over all involutions with exactly one fixed point .
[ th : key_to_s ] fix a positive integer , @xmath17 , and a graph @xmath572 .
any graph , @xmath14 , on @xmath16 vertices , that contains @xmath572 has second eigenvalue at least @xmath619 , where @xmath253 is a function of @xmath16 tending to @xmath214 as @xmath82 , and where @xmath249 is the norm of the adjacency matrix of @xmath573 .
this generalizes theorem [ th : improved_bound ] . * proof * fix @xmath5
; we will show that for @xmath16 sufficiently large any such @xmath14 has @xmath620 .
first , there is a finitely supported @xmath563 on @xmath573 with @xmath621 , where @xmath106 is the adjacency matrix of @xmath573 . if @xmath594 is the set of vertices on which @xmath158 is non - zero , then replacing @xmath158 with the non - negative first dirichlet eigenfunction on @xmath594 ( see @xcite ) we may assume @xmath158 is non - negative , nonzero , and that @xmath622 ( with equality everywhere except at the boundary of @xmath594 ) .
there is a covering map ( see @xcite and @xcite ) @xmath623 ; set @xmath624 to be the function on @xmath14 defined by @xmath625 if @xmath437 is the adjacency matrix of @xmath14 , then clearly @xmath626
so , on the one hand , @xmath627 , where @xmath324 is the rayleigh quotient for @xmath437 . on the other hand , the support , @xmath328 , of @xmath628 is of size no greater than that of @xmath158 , and this size is bounded ( independent of @xmath14 and @xmath16 ) .
so the same reasoning as in the proof of theorem [ th : improved_bound ] shows that @xmath629 where @xmath630 is the set of vertices of distance @xmath214 or @xmath68 to @xmath328 .
since @xmath628 is orthogonal to both @xmath631 and @xmath632 , we are done ( by lemma [ lm : fix_bug ] ) .
@xmath203 a tangle , @xmath572 , is _ critical _ ( respectively , _
supercritical _ , _ hypercritical _ ) if @xmath633 equals ( respectively , is at least , exceeds ) @xmath64 .
according theorems [ th : remarkable ] and [ th : key_to_s ] , a fixed hypercritical tangle can only occur in a graph with sufficiently many vertices if the graph has @xmath299 .
now that we know how tangles affect eigenvalues , we want to know how often the tangles occur .
this discussion , and the particular application to @xmath57 , @xmath65 , and @xmath69 , will take the rest of this section .
a _ leaf _ on a graph is a vertex of total degree @xmath68 ( whether the graph is directed or not ) .
we say that a graph is pruned if it has no leaves .
simple pruning _ is the act of removing one leaf and its incident edge from a graph ; _ pruning _ is the repeated performance of some sequence of simple prunings ; _ complete pruning _ is the act of pruning until no more pruning can be done .
for example , completely pruning a tree results in a single vertex with no edges ; completely pruning a cycle leaves the cycle unchanged .
[ pr : lies_on_irred ] given a graph , @xmath14 , there is a unique pruned graph @xmath338 obtainable from completely pruning @xmath14 .
furthermore , @xmath338 is completely pruned iff each edge of @xmath338 lies on an irreducible cycle . * proof * let @xmath634 denote the edges pruned in one pruning of @xmath14 , in the order in which they are pruned .
let @xmath497 be a different complete pruning of @xmath14 , which we assume does not contain all the @xmath635 .
let @xmath115 be the smallest integer such that @xmath636 lies in @xmath497 .
on the one hand , the removal of @xmath637 from @xmath14 , or any subgraph of @xmath14 , allows @xmath636 to be pruned from @xmath14 , or any subgraph of @xmath14 , including @xmath497 . on the other hand ,
the prunability of @xmath636 from @xmath497 contradicts the completeness of the pruning that formed @xmath497 .
it follows that any complete pruning contains all the edges of any other , and so any two are the same .
we now address the last statement of the theorem .
if @xmath338 has a leaf , then the edge incident upon this leaf does not lie in an irreducible cycle .
conversely , if @xmath338 has no leaves , and if @xmath371 is an edge with endpoints @xmath266 , consider the graph , @xmath638 , obtained by removing @xmath371 from @xmath338 . if @xmath267 and @xmath96 are connected in @xmath638 , then a minimal length path that joins them , along with @xmath371 , gives an irreducible cycle containing @xmath371 . otherwise , since @xmath267 s connected component is not a tree ( or else @xmath338 would have leaves )
, this component has an irreducible cycle , and a shortest walk from @xmath267 to this cycle , once around the cycle , and back to @xmath267 gives an irreducible cycle beginning and ending at @xmath267 .
similarly there is such a cycle about @xmath96 .
the cycle about @xmath267 , followed by @xmath371 , followed by the @xmath114 cycle , and back through @xmath371 ( in the other direction ) , gives an irreducible cycle containing @xmath371 .
@xmath203 a morphism of tangles is a morphism of @xmath398-labelled graphs , i.e. , a graph morphism that preserves the edge labelling .
[ de : tangle_order ] the _ order _ of a tangle , @xmath572 , is @xmath639 ( so while a whole - loop is counted as one edge , for the model @xmath69 , to be considered soon , a half - loop is also counted as one edge probability events . ] ) . more generally , for a graph , @xmath14 , or any structure with an underlying graph , @xmath14 ( such as a `` form '' or `` type '' to be defined in section 5 ) , its order is @xmath640 .
[ th : tangle_count ] let @xmath572 be a tangle of non - negative order .
then the expected number of occurrences of @xmath572 in an element , @xmath14 , of @xmath29 is @xmath641 , where @xmath250 is the order of @xmath572 .
the probability that at least one occurrence occurs is at least @xmath642 , with @xmath250 as before and where @xmath10 is the number of automorphisms of the complete pruning of @xmath572 .
notice that the number of automorphisms of the complete pruning of a tangle , @xmath572 , is at least as many as that of @xmath572 , and it may be strictly greater has at most one extension to @xmath572 . on the other hand , if @xmath572 is a cycle of length @xmath643 with all edges in one `` direction '' labelled @xmath644 , then @xmath572 has @xmath643 automorphisms ; yet if we add one edge labelled @xmath645 to @xmath572 at any vertex , the new graph has only the trivial automorphism . ] .
the proof following will imply that the probability of @xmath572 s occurrence is also at most @xmath646 , which matches the lower bound to first order , provided that @xmath319 .
* proof * let @xmath647 , and for a tuple @xmath648 of distinct integers between @xmath68 and @xmath16 , let @xmath649 denote the event that the map , @xmath575 , mapping @xmath650 to @xmath651 , is an occurrence of @xmath572 in @xmath14 .
if @xmath652 is the number of @xmath572 s edges labelled @xmath25 , then each event @xmath649 involves setting @xmath652 values of @xmath25 , all of which occur with probability @xmath653 since the sum of the @xmath652 is @xmath654 , this probability is @xmath655 .
since there are @xmath656 different @xmath649 s , the expected number of occurrences is @xmath641 , where @xmath657 .
next notice that if @xmath658 is a pruning of a tangle , @xmath572 , then the probability that @xmath572 occurs is @xmath659 times the probability that @xmath658 occurs ( adding each pruned edge adds a condition that occurs with probability between @xmath68 and @xmath660 with @xmath661 constants ) .
hence we may assume @xmath572 is pruned .
an automorphism of @xmath572 can be viewed as a permutation on @xmath594 , which is the same as a permutation , @xmath123 , on @xmath662 ( identifying a @xmath663 with @xmath88 ) .
such a permutation , @xmath123 , acts by permuting the components of the @xmath664 s .
say that @xmath649 is equivalent to @xmath665 if @xmath664 and @xmath666 differ by a permutation , @xmath123 , associated to an automorphism of @xmath572 ; i.e. , if @xmath649 and @xmath665 correspond to the same subgraph of @xmath14 .
let @xmath667 be a set of representatives in the equivalence classes of all @xmath664 s . by inclusion / exclusion ,
the probability that @xmath572 occurs at least once is at least @xmath668 the first summand is @xmath669 , by the argument given for the expected number . for the second summand , we may write @xmath670 as @xmath671 , where @xmath672 is a vector comprised of the distinct components of @xmath664 and @xmath666 , and where @xmath658 is the tangle obtained by gluing two copies of @xmath572 along certain vertices ( corresponding to where the components of @xmath664 and @xmath666 coincide ) . if @xmath664 is disjoint from ( i.e. , nowhere coincides with ) @xmath666 , then @xmath658 is two disjoint copies of @xmath572 ; for fixed @xmath673 we have @xmath674 the summation being over conditional probabilities , since the conditioning of @xmath665 and summing over @xmath664 disjoint only affects equation ( [ eq : tangle_expecteds ] ) by changing @xmath675 terms into @xmath676 terms for constants @xmath677 , which is a second order change .
hence @xmath678 @xmath679 to understand the situation where @xmath664 and @xmath666 overlap somewhere , we pause for some lemmas . [ lm : inclusion_order ] let @xmath574 be an inclusion of graphs , with @xmath14 connected .
then the order of @xmath14 is at least that of @xmath572 . * proof * let @xmath680 be @xmath14 with the vertices of @xmath681 identified , and all @xmath681 edges discarded
. then @xmath680 is connected , and so has order at least @xmath682 ; on the other hand , clearly the order of @xmath680 is the order of @xmath14 minus that of @xmath572 minus @xmath68 ( for the vertex that is the identification of all @xmath681 vertices ) . hence @xmath683
@xmath203 [ lm : edge_removal ] let @xmath14 be a pruned graph and let @xmath409 .
if @xmath684 ( i.e. , @xmath14 with @xmath371 removed ) has two connected components , then each connected component has order at least @xmath214 . * proof * consider a connected component , @xmath497 , of @xmath684 .
if @xmath497 did not contain a cylce , then @xmath497 would be a tree , and then @xmath14 would not be completely pruned .
so @xmath497 contains a cycle , and we may apply lemma [ lm : inclusion_order ] to deduce that @xmath497 has order at least @xmath214 @xmath203 [ lm : order_increases ]
let @xmath574 be an inclusion of pruned graphs .
then the order of @xmath14 is at least that of @xmath572 , and @xmath14 s order is strictly greater than @xmath572 s if @xmath681 is properly contained in @xmath14 . * proof * a connected component of a graph that is pruned ( and non - empty ) has non - negative order .
so we may assume @xmath681 meets every connected component of @xmath14 .
it suffices to prove the case where @xmath681 meets one connected component of @xmath14 , i.e. , the case where @xmath14 is connected .
choosing an edge , @xmath371 , that is missed by @xmath681 , we have that @xmath575 includes @xmath572 into @xmath684 ( i.e. , @xmath14 with @xmath371 removed ) ; we apply lemma [ lm : inclusion_order ] to those components of @xmath684 containing part of @xmath681 , and to the possibly one other component of @xmath684 we apply lemma [ lm : edge_removal ] .
it follows that the order of @xmath572 is at most that of @xmath684 ; but the order of @xmath684 is one less than that of @xmath14 .
@xmath203 [ lm : order_greater ] let @xmath685 be two inclusions of a tangle , @xmath572 , in a connected ( labelled ) graph , @xmath338 , such that @xmath686 .
assume that @xmath687 .
then the order of @xmath338 is greater than that of @xmath572 .
* proof * ignoring @xmath688 , the preceeding lemma applies to @xmath689 to immediately yield this lemma .
@xmath203 lemma [ lm : order_greater ] shows that @xmath690 since the summation can be broken down into a finite number of sums over tangles of order at least @xmath691 .
thus @xmath692 @xmath693 which completes the proof of theorem [ th : tangle_count ] .
@xmath203 it is easy to see that the above proof of theorem [ th : tangle_count ] uses very little about the model of random graph , and therefore generalizes as follows .
let @xmath694 be a model of @xmath0-regular random graphs on @xmath16 vertices labelled @xmath4 , defined for some values of @xmath16 .
further assume that ( 1 ) @xmath694 is invariant under renumbering @xmath4 , and ( 2 ) any tangle , @xmath572 , has expected number of occurrences @xmath641 where @xmath250 is the order of @xmath572
. then theorem [ th : tangle_count ] holds for @xmath694 .
[ th : other_ss ] for @xmath14 drawn from @xmath57 or @xmath65 , we have that @xmath36 with probability at least @xmath695 where @xmath60 , except for when @xmath61 in @xmath57 , where we may take @xmath62 .
the same holds for @xmath69 with probability @xmath346 , where @xmath38 ( and no exceptional values of @xmath0 ) .
* proof * for @xmath69 , the tangle consisting of @xmath68 vertex with a number of self - loops ( in this case half - loops ) , proves the theorem for @xmath69 just as it did for @xmath29 in theorem [ th : improved_bound ] .
consider the tangle , @xmath572 , with two vertices and @xmath139 edges joining the two vertices labelled @xmath696 ; @xmath572 is an @xmath57-tangle provided that @xmath298 , and an @xmath65-tangle if @xmath697 .
@xmath633 is clearly @xmath698 , @xmath699 , and the automorphism group of @xmath572 is of order @xmath30 .
so if @xmath700 and if @xmath14 contains this tangle , then we have @xmath36 for @xmath16 sufficiently large .
if we take @xmath60 , then @xmath700 ; we require @xmath697 for @xmath65 , amounting to @xmath701 which is satisfied for all @xmath17 .
for @xmath57 we require @xmath702 which is satisfied for all @xmath703 . since @xmath0 is even and @xmath704 in @xmath57 , we finish by examining the cases @xmath61 and @xmath705 .
for @xmath61 consider the tangle , @xmath572 , with vertices @xmath706 , edges labelled @xmath707 from @xmath708 to @xmath709 , from @xmath709 to @xmath710 , and from @xmath710 to @xmath711 . then @xmath712 and @xmath713 where @xmath658 is the subgraph of @xmath572 induced on @xmath714 .
but in the proof of theorem [ th : taufund_chnd ] we compute @xmath715 .
hence @xmath572 is hypercritical of order @xmath30 , so we may take @xmath62 in the theorem in the case of @xmath61 and @xmath57 . for @xmath705 ,
the proof of theorem [ th : taufund_chnd ] gives a tangle of order @xmath30 with @xmath716 ( see equation ( [ eq : d6tangle ] ) and the discussion around it ) .
so we may take @xmath62 in our theorem when @xmath705 in @xmath57 .
in this section we give some general techniques that are used in estimating the expected values of all the various traces that are used in this paper .
the main idea , originated in @xcite and strengthened in @xcite , is to group contributions to the trace in the following way . consider the word , @xmath114 , and vector , @xmath717 @xmath718 @xmath719 ( see figure [ fg : form1 ] ) .
this represents a possible or potential irreducible closed walk of length 10 , from @xmath720 to @xmath30 along the edge labelled @xmath721 , from @xmath30 to @xmath722 along @xmath644 , etc .
graphically we depict this potential walk by the subgraph it traces out , called its `` form '' ( see figure [ fg : form1 ] ) .
= 2.5 in the pair @xmath723 represents a possible contribution to @xmath724 . for any word , @xmath114 , of length @xmath97 and vector , @xmath725 , of length @xmath726 ,
let @xmath727 be the probability that @xmath728 for all @xmath88 , i.e. , that the cycle does occur .
then the expected value of @xmath724 is just a sum of appropriate @xmath727 s , what we will call a `` walk sum . '' in the form of figure [ fg : form1 ] , we see that the numbers of the vertices @xmath729 , are irrelevant , since our random graph model is `` symmetric , '' or invariant under renumbering the vertices .
so we may replace the vertex @xmath720 by an abstract symbol @xmath708 , @xmath30 by @xmath709 , etc .
( for reasons to made clear later , we do want to remember the order in which the vertices were traversed ) ; this replacement is not necessary , but makes clear that there is no particular preference for any numbering of the vertices . in our example of figure [ fg : form1 ] , @xmath723 , a loop about the vertex @xmath722
is traversed twice before we return to the starting vertex @xmath30 .
broder and shamir realized that the other vertices , those of degree @xmath30 that are not the starting vertex ( here the vertices @xmath730 ) are less interesting features of the `` form . '' by supressing these `` less interesting '' vertices we get the `` type '' of the form ( see figure [ fg : form1 ] ) ; in @xcite , walk sums were grouped by their form , and the sums for each form were grouped by the type of the form . a type is a graph with certain features , but when a form gives rise to a type then the edges of the type inherit @xmath428 labels from the form , and inherit `` lengths , '' which are the lengths of the @xmath428 labels ( recall that @xmath428 is the set of nonempty words over @xmath398 ) .
for example , the edge @xmath731 of the type in figure [ fg : form1 ] inherits a length of @xmath30 and a label of @xmath732 from the form next to it ; edge @xmath733 inherits a length of @xmath105 and a label of @xmath734 .
this paper introduces a `` new type , '' used in analyzing walk sums corresponding to selective traces .
a new type is a type with some additional information , primarily fixing the lengths of certain type edges and requiring all other lengths to be sufficiently large . by a _
weakly potential @xmath735-walk _ , @xmath723 , we mean a pair consisting of a word @xmath109 of length @xmath97 over @xmath398 , and a vector , @xmath736 , with each @xmath737 ; we sometimes refer to @xmath97 as the _ length _ and @xmath16 as the _ size _ of the weakly potential walk .
given such a @xmath723 , let @xmath738 denote the event that the @xmath25 are chosen so that @xmath728 for all @xmath739 .
let @xmath727 denote the probability that @xmath740 occurs .
potential @xmath735-walk _ is a weakly potential @xmath735-walk that is _ feasible _ , meaning that @xmath741 , or equivalently that the following two feasibility conditions hold : ( 1 ) for any @xmath112 such that @xmath742 , we have @xmath743 and for any @xmath112 such that @xmath744 , we have @xmath745 for example , if @xmath746 and @xmath747 , then @xmath748 is a weakly potential walk but not a potential walk . a _
potential @xmath735-closed walk _ is a potential walk , @xmath748 as above with @xmath749 .
all our variants of traces , irreducible and not irreducible , selective and not , can be viewed as sums of @xmath727 over appropriate @xmath114 s and @xmath750 s . in this subsection
we formalize this notion and make preliminary remarks about such sums and asymptotic expansions .
a _ walk collection , @xmath751 , _ is a collection , for any two positive integers @xmath97 and @xmath16 , of @xmath735-potential walks @xmath723 as above , i.e. @xmath114 is a word over @xmath398 of length @xmath97 , and @xmath736 is a @xmath726 dimensional vector over @xmath4 , and @xmath723 is feasible .
the _ walk sum _ associated to @xmath400 is @xmath752 the main goal of this paper is to organize the various @xmath748 pairs into groups over which we can easily sum @xmath727 .
one simple organizational remark is that symmetries in the @xmath723 pairs often simplify matters .
specifically , given an permutation , @xmath12 , of the integers , and a vector @xmath725 as above , let @xmath753 say that @xmath754 and @xmath725 _ differ by a symmetry _ if @xmath755 for some @xmath12 ; in this case clearly @xmath756 for any word @xmath114 of length @xmath97 .
we use @xmath757 to denote that @xmath754 and @xmath725 differ by a symmetry .
a walk collection , @xmath400 , is _ ssiic _ if it is 1 .
_ symmetric _ , i.e. @xmath758 implies that @xmath759 for all @xmath757 such that @xmath760 for all @xmath88 , 2 .
_ size invariant _ , i.e. if @xmath748 is a potential @xmath735-walk , then for any @xmath761 , @xmath758 iff @xmath762 , 3 . _ irreducible _ , meaning that @xmath763 implies that @xmath114 is irreducible , and 4 .
_ closed _ , meaning that @xmath763 implies that @xmath764 .
the walk sums of interest here , namely traces that are irreducible or strongly irreducible and possibly selective , will all be ssiic .
we now make a series of remarks about walk sums that apply to all ssiic walk sums ; some of the remarks apply more generally .
let @xmath725 be as above , i.e. , a positive integer valued vector .
define the _ equivalence class of @xmath725 _ to be @xmath765 = \ { \vec s \mid \vec s\sim\vec t\},\ ] ] i.e. , the set of all positive integer values vectors differing from @xmath725 by a symmetry .
define the _
@xmath16-th equivalence class of @xmath725 _
@xmath765_n = \ { \vec s \mid \vec s\sim\vec t \quad\mbox{and all}\quad s_i\le n\}\ ] ] ( we may omit the @xmath16 subscript if @xmath16 is understood ) .
let @xmath16 be fixed , and set @xmath766_n } p(w;\vec s\>)\ ] ] @xmath767 where @xmath96 is the number of distinct elements of @xmath725 .
a symmetric walk sum is just the sum of certain @xmath768 s , and we can write @xmath769)\in{{\cal w}}(k , n)}{{\rm e}_{\rm symm}}(w;\vec t\>),\ ] ] where the summation over @xmath770)\in{{\cal w}}(k , n)$ ] means that we sum over one @xmath725 in each equivalence class and over all @xmath114 .
each @xmath736 has an @xmath757 where the size of @xmath771 s entries are at most @xmath726 .
so for each @xmath97 there are a finite number of equivalence classes , @xmath772 , of @xmath723 such that @xmath114 is of length @xmath97 and @xmath725 is of some finite size .
we refer to @xmath772 as the set of _ potential walk classes _ of length @xmath97 ( or _ potential closed walk classes _ when we restrict to those @xmath750 s with @xmath749 ) .
so if @xmath400 is size invariant we may write @xmath769)\in{{\cal w}}(k)}{{\rm e}_{\rm symm}}(w;\vec t\>)_n,\ ] ] where the right - hand - side summation has a fixed , finite number of summands independent of @xmath16 ( for fixed @xmath97 ) . our next step is to comment about @xmath773 .
notice that if the conditions @xmath728 involve determining @xmath774values of @xmath406 , then @xmath775 let @xmath776 .
we have @xmath777 notice that in the power series expansions about @xmath778 of @xmath779 the @xmath780 coefficient is a polynomial ( of degree at most @xmath781 ) in @xmath139 .
it follows that there exist polynomials @xmath782 in the variables @xmath783 such that @xmath784 for @xmath16 sufficiently large .
[ de : expansion_polynomials ] we define the _ expansion polynomials _
, @xmath785 { the expansion polynomials}}},\ ] ] to be the polynomials that give the expansion in equation ( [ eq : expansion_polynomials ] ) .
throughout the paper , @xmath652 denotes the number of @xmath25 values determined by the relevant structure ( in this case the potential walk , @xmath748 ) .
we remark that since @xmath776 , we have that @xmath96 is determined from the @xmath652 if @xmath786 is fixed and known ; in such situations , the expansion polynomials may be regarded as functions of the @xmath652 alone .
[ th : exp_polys ] for any @xmath787 and any integer @xmath788 we have @xmath789 where @xmath790 and the @xmath791 are the expansion polynomials . the proof is contained between lemma 2.7 and corollary 2.10 of @xcite , although the proof there has a minor error .
we will correct it here and review the entire proof , since we will later need variants of this theorem for @xmath104 . if @xmath792 where the @xmath793 and @xmath794 are positive constants , then @xmath631 s @xmath88-th derivative satisfies the bound @xmath795 where @xmath796 is the maximum of the @xmath794 ( by equation ( 6 ) in @xcite on page 339 ) .
this estimate , using taylor s theorem , expanding in @xmath797 about @xmath778 , gives the error term for theorem [ th : exp_polys ] of @xmath798 for some @xmath799 $ ] .
since @xmath800 @xmath801 we conclude that the error term is at most @xmath802 in the case of theorem [ th : exp_polys ] , the @xmath803 represents the sum of @xmath804 , which is at most @xmath805 , and the @xmath806 represents the sum over @xmath115 of all sums of @xmath807 , which is at most @xmath805 . since @xmath808 , we get an error term at most @xmath809 .
( see @xcite for more details . )
@xmath203 given a pair , @xmath748 , as above , its _ order _ is @xmath810 , with @xmath811 as above .
[ lm : big_order ] given a word , @xmath114 , of length @xmath97 over @xmath398 , we have @xmath812 which for @xmath813 is at most @xmath814 for some constant @xmath10 depending only on @xmath250 .
* proof * this can be found in @xcite , second displayed equation and discussion before on page 352 ; this is the same idea used in the @xmath815 case proven in @xcite .
( the extra factor of @xmath16 appears here but not in @xcite , since we do not fix the initial vertex of the walk . also note that the `` order , '' used here , is one less than the `` number of coincidences , '' used in @xcite . ) for the ease of reading , we shall discuss these ideas in our notation .
we may evaluate @xmath816 by considering the steps of the walk @xmath109 one by one ; i.e. , we fix a @xmath817 and consider the random walk @xmath818 , @xmath819 , etc .
we inductively consider the probability that @xmath820 is equivalent to @xmath821 for @xmath822 .
assuming @xmath820 is equivalent to @xmath821 , and assuming we wish to have @xmath823 equivalent to @xmath824 , one can divide the choice and outcome of the random variable @xmath825 into three cases : ( 1 ) @xmath826 has already been determined by previous information ( i.e. , the value @xmath827 has been determined in a previous step ) ; ( 2 ) @xmath826 has not been determined and ( 2a ) @xmath826 must occur as one of the @xmath828 ( i.e. , when @xmath829 occurs previously in @xmath830 ) , or ( 2b ) @xmath826 must must be different from @xmath828 ( i.e. , when @xmath829 does not occur previously ) .
we call case ( 1 ) a forced choice , case ( 2 ) a free choice , with case ( 2a ) a coincidence , and case ( 2b ) a generic choice . for example , if @xmath831 and @xmath832 , then @xmath708 is a coicidence ( since @xmath833 occurs previously as @xmath834 and @xmath835 has not been determined ) , @xmath709 is a forced choice ( since @xmath836 has been determined ) , @xmath710 is a generic choice ( @xmath837 does not occur in @xmath838 ) , and @xmath711 is a coincidence . a choice @xmath825 is deterministic if it is a forced choice ; a choice occurs with probability @xmath839 if it is a generic choice , and with probability @xmath840 if it is a coincidence .
we see that a word of order @xmath717 has exactly @xmath841 coincidences . if follows that each word of order at least @xmath250 and length @xmath97 has some @xmath691 positions of @xmath97 that can be marked as coincidences .
fixing a marking , each coincidence of the marking occurs with probability at most @xmath842 ( since @xmath826 must assume one of the at most @xmath843 values @xmath828 , and at most @xmath97 values of the permutation @xmath844 could have been determined up to that point ) .
finally @xmath197 can be chosen in @xmath16 ways .
we conclude the lemma .
@xmath203 [ lm : equiv_classes ] for any irreducible word , @xmath114 , over @xmath398 , of length @xmath97 , there are at most @xmath845 equivalence classes @xmath846 $ ] whose order with @xmath114 is @xmath847 . *
proof * see the third displayed equation of page 352 of @xcite and the discussion preceding . for completeness and ease of reading , we repeat the argument here
. if @xmath114 has order @xmath848 , then it has @xmath115 coincidences occurring in @xmath97 places . again , in each coincidence , @xmath825 , @xmath826 is being chosen from at most @xmath97 values .
this gives the left - hand - side of equation ( [ eq : equiv_classes ] ) .
for the right - hand - side we notice that @xmath849 for @xmath850 .
@xmath203 [ th : ssiicexpansion ] let @xmath400 be ssiic and let @xmath319 . then for all @xmath813 we have @xmath851 where @xmath852){\rm \;order\;}j,\in{{\cal w}}(k ) } p_{i - j}(w;[\vec t\>])\ ] ] ( with @xmath791 the expansion polynomials , which can be viewed as a function of @xmath770)$ ] since @xmath770)$ ] determines the @xmath652 and @xmath96 ) and where for some @xmath10 depending only on @xmath250 , @xmath853 * proof * by lemma [ lm : big_order ] , we introduce an error of at most @xmath814 per word by ignoring potential walks of order at least @xmath250 .
each word , @xmath114 , has at most @xmath854 associated potential walk classes of order at most @xmath855 ( by lemma [ lm : equiv_classes ] ) , and truncating the associated asymptotic expansion , as in equation ( [ eq : simple_expansion ] ) , of each associated potential walk class results in an error of at most @xmath854 ( by theorem [ th : exp_polys ] ) .
so each word , @xmath114 , of length @xmath97 involved in @xmath400 contributes an error of at most @xmath856 , and there are at most @xmath241 such words ( for @xmath151 ) since @xmath400 consists of only irreducible words .
@xmath203 here we analyze walk sums associated with simple loops .
this gives some ideas and a lemma to be used in section 8 .
let @xmath114 be an irreducible word over @xmath398 of length @xmath97 and @xmath725 a @xmath726 tuple over @xmath4 as in the previous subsection .
if @xmath764 and the @xmath857 s are otherwise distinct , we say @xmath723 is a _ singly traversed simple loop _ or _
stsl _ for short ; we define @xmath858 to be the collection of all stsl s . when @xmath740 occurs and @xmath859 , the closed walk from @xmath834 following @xmath114 traces out a `` simple loop '' once , that begins and ends at @xmath834 , moving through distinct edges and vertices throughout the closed walk .
clearly @xmath858 is ssiic , so according to theorem [ th : ssiicexpansion ] we have an asymptotic expansion in @xmath101 with coefficients @xmath860 for the associated walk sum .
we now briefly indicate why the @xmath230 are @xmath0-ramanujan .
this is a mildly tedious exercise , covered ( in much greater generality ) in @xcite .
we quote the main points here .
first , note that there is exactly one equivalence class @xmath846 $ ] of @xmath750 s that appear in @xmath858 .
so we may write @xmath861 for @xmath862 for any @xmath725 of the equivalence class , and we may write @xmath863 for the @xmath864 in theorem [ th : ssiicexpansion ] or [ th : exp_polys ] ( note also that @xmath865 in the notation of theorem [ th : exp_polys ] )
. let for @xmath866 , let @xmath867 denote the irreducible words of length @xmath97 beginning with @xmath123 and ending with @xmath12 .
for @xmath868 let @xmath869denote the number of @xmath25 and @xmath870 occurring in @xmath114 .
since an stsl has no forced choices , the definition of @xmath652 here agrees with that in definition [ de : expansion_polynomials ] .
[ lm : irdeigens ] let @xmath871 be a polynomial . for every @xmath872 there are polynomials @xmath873 of @xmath97 of degree at most the degree of @xmath874 such that @xmath875 * proof * this is immediate from lemma 2.11 of @xcite or corollary 2.12 ( note that our @xmath0 is @xmath100 in @xcite ) . @xmath203 note that the formula for the above lemma comes about from the fact that @xmath876 are the eigenvalues of @xmath542 where @xmath14 is the graph with one vertex and @xmath3 whole - loops ( with @xmath542 defined as in subsection [ sb : lambda_irred ] ) .
[ cr : stsl_dtreelike ] we have each @xmath860 as in theorem [ th : ssiicexpansion ] for @xmath877 is @xmath0-ramanujan .
* proof * recall that an @xmath878 is of order @xmath214 , and there is only one equivalence class of @xmath725 for stsl s .
so we have @xmath879 where the @xmath791 are the expansion polynomials .
the result now follows from lemma [ lm : irdeigens ] , with error term bounded by a polynomial in @xmath97 .
consider any irreducible @xmath748 that traces out a simple loop , i.e. , the vertices and edges visited form one closed walk , but now we do nt require that the loop is traversed only once .
corresponding to this geometric picture of a simple loop we can form the associated walk collection of _ simple loop closed walks _ ; such closed walks , being irreducible , must traverse the loop traced out some number of times .
so let @xmath880 be the set of @xmath748 pairs with 1 .
@xmath881 irreducible , 2 .
@xmath882 , 3 .
@xmath883 distinct for some @xmath250 dividing @xmath50 , 4 .
@xmath884 for @xmath885 , and 5 .
@xmath886 for some word @xmath267 with @xmath887 .
@xmath880 is the walk collection of simple loop walks .
clearly we have @xmath888 we easily conclude the following theorem .
the @xmath860 corresponding to @xmath880 are @xmath0-ramanujan , and have the same principal term as the @xmath860 corresponding to @xmath858 . * proof * consider the @xmath860 corresponding to @xmath858 . by corollary [ cr : stsl_dtreelike ] , all the @xmath230 are @xmath0-ramanujan . by equation ( [ eq : multiply_traversed ] )
, it suffices to show that @xmath889 are also @xmath0-ramanujan .
fixing an @xmath88 and setting @xmath890 as the decomposition of @xmath230 into principal and error terms , we see @xmath891 where @xmath892 for some @xmath10 as in definition [ de : ram ] .
it is clear that the right - hand - side above is bounded by @xmath893 times a polynomial in @xmath97 .
@xmath203 in this subsection we will classify potential walks , @xmath723 , or more generally potential walk classes , according to some characteristics of the subgraph that the walk traces out . a _ form _ , @xmath894 , is an oriented , @xmath398-labelled graph , @xmath895 , with edges and vertices numbered .
a _ specialization _ of a form , @xmath894 , is an injection @xmath896 . with each potential walk , @xmath723 , we can associate a form , @xmath897 , with a specialization , @xmath575 , as follows .
( the form , @xmath894 , is not unique , but is unique up to unique isomorphism , as described below ; the specialization is unique given the form . ) 1 .
set @xmath898 to be any numbered ( @xmath899 numbered @xmath88 ) set of size @xmath250 , where @xmath900 is the number of distinct elements among the @xmath857 ( where @xmath736 ) 2 .
@xmath901 is the @xmath88-th distinct element of the sequence @xmath902 , 3 .
set @xmath903 to be any numbered set of size @xmath139 ( @xmath635 numbered @xmath88 ) , where @xmath904 is the number of distinct triples @xmath905 , where we identify a triple @xmath906 with @xmath907 , 4 . if @xmath908 is the @xmath250-th distinct tuple in @xmath905 ( with the previous identification ) , then @xmath909 defines the structural map ( see definition [ de : structural ] ) of @xmath910 .
( see the example in figure [ fg : form1 ] explained at the beginning of this section ) .
in other words , the form is the subgraph traced out by @xmath748 , with some additional information ( we remember the order in which the vertices and edges are visited , and the direction each edge is first traversed ) .
we say that forms @xmath911 and @xmath912 are _ isomorphic _ if they are isomorphic as oriented , numbered , @xmath398-labelled graphs . because of the numbering , there is at most one isomorphism between any two forms ( or a form and itself ) .
we say that @xmath748 is _ of form @xmath894 _ or _ associated to @xmath894 _ , written @xmath913 , if one ( or any ) of the forms associated to @xmath748 is isomorphic to @xmath894 .
given @xmath748 , there is always an associated form , @xmath894 , with @xmath914 and associated specialization , @xmath575 , being the identity ; however , we usually view @xmath915 as any numbered set of the right size , since all of our random graph models are symmetric . if @xmath913 , then define @xmath916 } _
n = { { \rm e}_{\rm symm}}(w,\vec t\>)_n\]]which depends only on @xmath894 , and not on the particular @xmath748 to which @xmath894 is associated ; indeed , @xmath917 } _
n = \frac{n!}{(n - v)!}\prod_{i=1}^{d/2 } \frac{(n - a_i)!}{n!},\ ] ] where @xmath918 , and @xmath652is the number of edges in @xmath894 labelled with @xmath25 and @xmath870 ( this is exactly the @xmath869 of equation ( [ eq : probability ] ) of any word , @xmath114 , associated to @xmath894 ) . hence ,
if @xmath400 is symmetric and size invariant , we may write @xmath919 } _ n,\ ] ] where @xmath920is the number of potential walk classes in @xmath772 associated to @xmath894 , and we sum over one @xmath894 in each isomophism class of forms .
a _ legal walk _ in a form , @xmath894 , is a walk starting in @xmath921 that visits all the vertices of @xmath910 in order ( of their numbering ) , all the edges in order , and any edge is first traversed in the direction of its orientation .
each legal walk of length @xmath97 generates a walk class in the natural way .
the following easy proposition is worth stating formally ; it follows from the definitions .
@xmath920 is the number of legal walks on @xmath894 of length @xmath97 .
the _ order _ of a form , @xmath894 , is latexmath:[${{\rm ord}}(\gamma ) = |e_\gamma|- that any self - loop , whole - loop or half - loop , counts as a single edge .
( the order of a form equals the order of any potential walk to which it is associated . )
the `` form '' allows us to group together potential walks that determine the same information on the graph ; this can facilitate the task of studying a walk sum .
a futher tool is the grouping of forms together by their `` type , '' which we now briefly motivate in rough terms to be made precise .
a type arises from a form with all its maximal beaded paths collapsed to edges ( by the appropriate supression ) .
each collapsed edge inherits an edge length and @xmath428 labelling from the `` form , '' but in the type we forget this data .
as long as we collect forms of a given type by the first and last letters of all the collapsed edge labellings , we can apply lemma [ lm : irdeigens ] at each collapsed edge to study sums of the expansion polynomials in the @xmath652 s ( and @xmath96 ) over forms in such collections . before defining a `` type , ''
we recall that for a form , @xmath923 , with specialization , @xmath575 , the potential walk @xmath723 pulls back under @xmath924 to a walk on @xmath910 ; we remark that the vertex and edge numberings of @xmath894 serve to remember in which order the vertices and edges were traversed in the walk .
a _ type _ , @xmath194 , is a connected , oriented graph @xmath925 , with vertex and edge numberings such that all vertices except possibly the first one are of degree at least @xmath105 .
a _ labelling _ of a type means a @xmath428-labelling ( recall @xmath428 is the set of words over @xmath398 of length at least @xmath68 ) . to each form , @xmath894 , we associate a type , @xmath926 , as follows .
let @xmath262 be the set of beads of @xmath910 numbered greater than @xmath68 .
we claim @xmath262 can not contain a cycle , for then this cycle would be disconnected from the vertex numbered @xmath68 .
so we may form the supression of @xmath262 in @xmath910 .
this supression inherits a vertex and edge numbering from @xmath910 ( ordering a vertex before another in the supression if it is numbered less in @xmath910 , and ordering edges in the supression by any associated edge in @xmath927 ) . of course , the @xmath398-labelling of @xmath894 gives rise to a @xmath428-labelling of @xmath928 .
in other words , the beads of a form are `` less important '' features , and the type is just the form with these `` less important '' features supressed .
a @xmath428-labelled type uniquely determines a form , and vice versa .
we wish to group together forms corresponding to one type ( inducing different @xmath428-labellings ) , the prototypical example being stsl s or sl s discussed earlier in this section .
to do this it will be helpful to remember a small part of the labelling , namely the starting and ending letter of each @xmath428-label .
a _ lettering _ of a type , @xmath194 , is the assignment to each directed edge a _ starting letter _ in @xmath398 and an _ ending letter _ in @xmath398 ( such that opposite directed edges are lettered with the starting letter of one being the inverse of the ending letter of the other ) .
given a form , @xmath894 , with @xmath926 , or equivalently a @xmath428-labelling of @xmath194 , the associated lettering assigns to each edge the starting and ending letter of the @xmath428-label assigned to it in its orientation .
it turns out that the notion of a type is too coarse to attack the alon conjecture ( but sufficient for the results of @xcite ) .
the problem is that some types , when their edges take on certain @xmath428-labellings , contain supercritical tangles .
when such tangles occur ( which they can when their order is roughly @xmath929 ) , we must distinguish where there tangles occur .
this is what a `` new type '' does , where the type edges are partitioned into `` long '' and `` fixed '' ( in length ) edges , and where supercritical tangles lie on type edges that are `` fixed . ''
we then modify our walk sums to be `` selective '' ( see section 6 ) , a notion which requires us to know where these tangles occur in the type .
[ de : new_type ] a _ @xmath482-new type _ is a collection , @xmath930 , of ( 1 ) a lettered type , @xmath194 , ( 2 ) a partition of @xmath931 into two sets , @xmath932 , ( 3 ) for each @xmath933 an _ edge length _ , @xmath934 , with @xmath935 , and ( 4 ) a @xmath428-labelling of @xmath936 with each @xmath933 labelled with a word of length @xmath934 .
a @xmath428-labelling of @xmath194 ( or , equivalently , a form , @xmath894 ) , is said to be _ of @xmath482-new type @xmath579 _ if each @xmath937 label is of length at least @xmath482 , and each label corresponding to @xmath933 is of length @xmath934 and agrees with the label specified by @xmath579 .
@xmath579 is said to be _ based on @xmath194_. for each @xmath938 there are finitely many types ( up to isomorphism ) of order at most @xmath250 . for each type , @xmath194 , and each @xmath939
there are finitely many @xmath482-new types based on @xmath194 .
* proof * the first statement is just lemma 2.3 of @xcite , except that `` coincidence '' is used instead of `` order '' ( and the coincidence is the order plus one ) .
the second statement is clear since there are finitely many ( 1 ) letterings , ( 2 ) partitions of @xmath931 , ( 3 ) choices of @xmath934 with @xmath940 , and ( 4 ) labellings of each @xmath936 edge , @xmath635 , with a length @xmath941 .
@xmath203 so far we have defined walk sums ; we have seen that symmetric , size invariant walk sums can be organized into forms , by equation ( [ eq : fundamental ] ) ; we have seen that forms can be grouped by type and new type . in this section
we briefly explain how and why we use types and new types .
fix a symmetric , size invariant walk collection , @xmath400 .
organizing forms by type , we may write equation ( [ eq : fundamental ] ) as @xmath942 } _ n,\ ] ] where @xmath943 means the form @xmath894 is of type @xmath194 , and the summation in @xmath194 ranges over all types .
typically we need only sum over @xmath194 of at most some order , so the sum in @xmath194 will effectively be a finite sum . so fix a @xmath194 and define @xmath944 } _ { n , k}$ ] to mean @xmath945 } _ { n , k } = \sum_{\gamma\in t } w_\gamma({{\cal w}},k){\mbox{e}\left[\gamma\right ] } _ n.\ ] ] let @xmath194 s edges be @xmath946 . for a vector , @xmath947 ,
let @xmath948 denote the set of forms , @xmath894 , of type @xmath194 , such that for all @xmath88 the length of the beaded path in @xmath894 corresponding to the edge @xmath635 in @xmath194 has length @xmath949 .
for each @xmath950 fix an integer @xmath951 .
let @xmath952 denote the number of legal @xmath400 walks in @xmath894 that traverse edge @xmath635 exactly @xmath651 times .
clearly @xmath953 i.e. , @xmath954 depends only on @xmath664 and the type , @xmath194 , of @xmath894 .
this allows us to write @xmath945 } _ { n , k } = \sum_{\vec m } w_t(\vec m ) \sum_{\vec k \cdot \vec m = k } \ \ \sum_{\gamma\in
t(\vec k\ > ) } { \mbox{e}\left[\gamma\right ] } _ n \;.\ ] ] since each @xmath955 } _ n$ ] has a @xmath101 expansion , by adding expansions we get an asymptotic expansion @xmath956 } _
n = n^{-{\rm ord}(t ) } \bigl ( q_0(\vec k\ > ) + q_1(\vec k\>)/n + \cdots \bigr),\ ] ] where we ignore error terms in this subsection ; of course , @xmath957 perhaps the main technical point of section 8 ( see theorem [ th : f_vecm_est ] here , and lemma 2.14 in @xcite ) is that the @xmath958 defined by @xmath959 are @xmath0-ramanujan , and roughly speaking @xmath958 s principal part and error term decay at most like @xmath960 where @xmath961 ( actually , the principal part is shown in lemma 2.14 in @xcite to decay at most like @xmath962 ) .
the way this is done is very roughly as follows .
first , we fix a lettering of @xmath194 and apply lemma [ lm : irdeigens ] to each edge , @xmath635 , of @xmath194 , whose length is set to @xmath949 .
this shows that the @xmath963 is a sum of @xmath0-ramanujan functions whose arguments are all sums of a subset of the @xmath949 . as an example , consider a function , @xmath964 with @xmath631 being @xmath0-ramanujan .
the maximum value of @xmath965 , given @xmath966 with fixed @xmath664 and @xmath97 , is @xmath967 , which is achieved when and only when @xmath968 whenever @xmath969 ( assuming at least one @xmath970 ; otherwise @xmath964 is bounded by @xmath971 since @xmath972 ) .
this @xmath967 is where the decay exponential in @xmath139 comes from ( see section 8 and/or section 2 in @xcite for details ) .
it follows that there is an asymptotic expansion @xmath945 } _ { n , k } = n^{-{\rm ord}(t ) } \bigl ( q_0(k ) + q_1(k)/n + \cdots \bigr),\ ] ] where @xmath973 since the error term of the @xmath958 decay like @xmath974 ( roughly speaking ) , the above sum for @xmath975 turns out to be @xmath0-ramanujan provided that latexmath:[\[\label{eq : bound_w_t } @xmath5 . if all @xmath977 were @xmath0-ramanujan for all @xmath194 , then we d have a fairly simple proof of the alon conjecture .
the results of @xcite are based on the fact that equation ( [ eq : bound_w_t ] ) for all types up to a certain order ( of roughly @xmath929 ) ; past order @xmath929 , equation ( [ eq : bound_w_t ] ) generally fails to hold .
furthermore , by theorem [ th : nottreelike ] we see for certain that at least one @xmath977 fails to be @xmath0-ramanujan for an @xmath88 bounded by @xmath978 .
the new approach in this paper is as follows .
consider a constant positive integer @xmath482 , and consider @xmath979 we can divide this sum by fixing some of the @xmath949 s at fixed values less than @xmath482 ( call these @xmath980 ) , and then summing over the remaining @xmath949 s ( namely @xmath981 ) subject to the remaining @xmath949 s being at least @xmath482 ( or `` long '' ) .
this is where a `` new type '' comes from .
next we fix a constant @xmath982 and define a `` selective trace '' to be the sum of irreducible closed walks of a given length that have no subpath of length @xmath208 tracing out a graph of @xmath551 at least @xmath64 . if @xmath983 then the corresponding `` selective '' version of @xmath984 ( that depends on the new type , i.e. , on knowing the particular @xmath949 that are fixed and their values ) is bounded by roughly @xmath985 ( for appropriately large @xmath208 and @xmath482 ) . but consider the @xmath958 for the `` new type , '' i.e. , the sum @xmath986 with @xmath980 at their fixed values ; it turns out they decay like @xmath987 .
thus , after summing over all new types , the corresponding selective analogues of the @xmath975 are @xmath0-ramanujan .
unfortunately , a selective trace does not generally equal the original trace unless the graph in question is free of certain tangles .
still , in sections 9 and later we show how the asymptotic expansions with @xmath0-ramanujan coefficients of selective traces can be used to control a random graph s eigenvalues .
in this section we define a _ selective trace _ , and discuss some of its properties .
fix a graph , @xmath28 , coming from @xmath29 , so that @xmath23 and @xmath14 is @xmath398-labelled . by a _ path or @xmath870 , being taken on each step of the path .
] of length @xmath97 _ in @xmath14 we shall mean a vertex , @xmath378 , and a word of length @xmath97 , @xmath109 , over @xmath398 ( i.e. , with each @xmath111 ) . such a path determines a subgraph in @xmath14 of those vertices and edges traversed .
we say a path _ traverses _ a tangle , @xmath572 , if the subgraph traversed by the path contains the tangle , @xmath572 .
let @xmath988 be a ( finite or infinite ) collection of tangles . for positive integer , @xmath208 , the _ set of @xmath989-selective closed walks ( respectively , walks ) _ are those irreducible closed walks ( respectively , walks ) that have no subpath of length at most @xmath208 that traverses a tangle in @xmath990 .
@xmath97-th irreducible @xmath989-selective trace of @xmath14 _ ,
@xmath991 , is the number of @xmath989-selective closed walks of length @xmath97 .
intuitively , the selective trace modifies the standard irreducible trace on those graphs that have a tangle in @xmath990 , and avoids those closed walks that in some short part trace out such a tangle . what is the point of the selective trace ?
we can answer this question in two ways .
first , since hypercritical tangles give large eigenvalues , any trace with an arbitrarily long asymptotic expansion in @xmath101 with @xmath0-ramanujan coefficients must avoid hypercritical tangles ( according to theorems [ th : remarkable ] and [ th : key_to_s ] ) ; a trace must be selective or its asymptotic expansion coefficients will not all be @xmath0-ramanujan .
second , there is a crucial technical theorem , theorem [ th : crucial_cycle_count ] , about counting irreducible contributions to a selective trace .
this lemma makes certain infinite sums converge for the selective trace that would have to diverge for the standard trace for example , the infinite sum involving @xmath93 and @xmath94 just above the middle of page 351 in @xcite , for types of order @xmath95 ; for the same reason , this crucial theorem makes the @xmath101 expansion for a selective trace have @xmath0-ramanujan coefficients when they do nt for a trace that is not selective indeed , the @xmath992 bound in equation ( 24 ) of @xcite depends critically on @xmath993 , and this equation corresponds to the error in the @xmath994 term in the expansion of the expected value of the irreducible trace ( recall that @xmath100 in @xcite corresponds to our @xmath0 ) .
we shall finish this subsection with the crucial technical theorem , theorem [ th : crucial_cycle_count ] , after setting up the necessary terminology .
a _ relabelling _ of a tangle , @xmath572 , is a tangle , @xmath658 , that differs from @xmath572 only in its edge labels .
a set , @xmath990 , of tangles is called _ closed under pruning ( respectively , relabelling ) _ if @xmath995 implies @xmath996 for any pruning ( respectively , relabelling ) , @xmath658 , of @xmath572 .
note : in the definition above , @xmath572 and @xmath658 must be @xmath29-tangles ( or tangles in whatever model is discussed) a vertex with two self - loops labelled both labelled @xmath644 is not a @xmath29-tangle and is therefore not considered a relabelling of the tangle where the self - loops are labelled @xmath644 and @xmath645 . for
a positive integer , @xmath12 , let @xmath997 be the set of tangles whose order is at least @xmath12 .
for positive integers @xmath998 , let @xmath999 be the set of all tangles whose order is at least @xmath1000 and at most @xmath1001 .
we also write @xmath1002 for @xmath997
. since pruning a tangle does not affect its order , @xmath1003 is closed under pruning ; clearly @xmath999 is closed under relabelling .
consider a form , @xmath894 , of type @xmath194 , in which @xmath194 s edges , @xmath635 , have length @xmath949 ( as beaded paths arising from @xmath894 ) . for each @xmath950 fix an integer @xmath951 .
let @xmath990 be a set of tangles closed under relabelling .
let @xmath1004 be the number of legal closed walks ( in particular , beginning at the first vertex ) in @xmath894 that traverse each @xmath635 exactly @xmath651 times ( in either direction ) and that are @xmath989-selective .
since @xmath990 is closed under relabelling , @xmath1005 depends only on the length , @xmath949 , of @xmath635 in @xmath894 , not on the particular @xmath428 labels of length @xmath949 .
so we may write @xmath1006 { the number of potential walk classes or legal walks on a form , type , or new type with various additional restrictions ( such as irreducibility)}}}}.\ ] ] now given the above setting , call an edge , @xmath1007 , of @xmath194 _ long _ if @xmath1008 , and _ short _ otherwise . if a walk contains some @xmath995 in any consecutive @xmath208 steps , then by possibly pruning these consecutive steps along long edges at the beginning and end , we get a consecutive walk over short edges that contains a pruning of @xmath572 . in particular ,
if @xmath1009 and @xmath1010 is a @xmath482-new type based on @xmath194 , then @xmath1011 depends on only @xmath579 , @xmath664 , @xmath208 , and @xmath990 provided that @xmath1012 ; hence we may write @xmath1013 we say that a collection of tangles , @xmath990 , is _ @xmath250-supercritical _ if it contains all supercritical tangles of order at most @xmath250 .
next we give two examples of very natural @xmath250-supercritical tangle sets .
let @xmath49 be the smallest order of a supercritical tangle , and let @xmath1014 .
let @xmath1015 the be set of all supercritical tangles .
@xmath1016 and @xmath1015 are clearly @xmath250-supercritical for any @xmath250 ; @xmath1017 and @xmath1018={{{{\psi}}_{\rm eig}}}\cap{{{{\psi}}_{\rm ord}}}({{\tau_{{\rm fund}}}},r)$ ] are also clearly @xmath250-supercritical .
we arrive at our crucial technical theorem , that is the key to the selective trace .
[ th : crucial_cycle_count ] let @xmath194 by any type , with specified edge set partition @xmath937,@xmath936 , and a @xmath398-lettering specified .
let the edges be indexed so that @xmath1019 then there is a @xmath10 , an @xmath5 , and an @xmath1020 such that the following is true for all @xmath1021 .
let @xmath990 be a set of tangles containing all supercritical tangles included in a form of type @xmath194 ; e.g. , by lemma [ lm : inclusion_order ] we may take @xmath990 to be any @xmath250-supercritical set for @xmath1022 .
let @xmath1023 for a @xmath482-new type , @xmath579 , with @xmath1009 and with @xmath579 having edge set partition @xmath932 .
then @xmath1024 * proof * the general approach we take is based on the following crude estimate .
let @xmath1025 be a collection of non - negative power series , and let @xmath1026 be a power series that majorizes each @xmath1027 .
suppose that @xmath1028 converges for some @xmath1029 .
then the @xmath1030 coefficient of any of the @xmath1027 s is bounded by @xmath1031 .
specifically , we shall show that there is an @xmath1020 such that the following holds .
let @xmath1032 , and let @xmath579 be a @xmath482-new type subject to the conditions of the theorem .
let @xmath14 be a vlg whose underlying graph is @xmath194 , whose @xmath936 edges take their lengths from @xmath579 , and whose @xmath937 edges all have length @xmath482 .
let @xmath1033 , where @xmath1034 is the number of @xmath1035-selective irreducible walks of length @xmath97 in @xmath14 .
we shall show that there is a @xmath1036 and a @xmath1037 such that ( 1 ) @xmath631 majorizes @xmath158 , ( 2 ) @xmath631 depends only on @xmath194 , and ( 3 ) @xmath1028 converges . in that case
@xmath1038 which completes the theorem .
let @xmath1039 be the set of vlg s , @xmath338 , whose underlying graph is a subgraph of @xmath194 containing only @xmath936 edges , with the property that @xmath1040 .
let @xmath1041 be the set of elements of @xmath1039 that are not majorized by a different member of @xmath1039 .
we claim that @xmath1041 is finite ; indeed , if @xmath1042 were a distinct sequence of vlg s in @xmath1041 , then by passing to a subsequence we could assume that for every @xmath1043 either the length of @xmath371 in @xmath1044 is constant or the length tends to infinity ; but then @xmath1045 would majorize all @xmath1044 with @xmath88 sufficiently large . let @xmath1046 , and let @xmath1047 , where @xmath1048 is the number of irreducible walks of length @xmath97 in @xmath1044 .
these @xmath1048 are given as the number of walks of length @xmath1049 in @xmath1050 ; it follows that @xmath1051 has radius of convergence greater than @xmath1052 .
also , if @xmath1053 is the number of irreducible walks in @xmath1044 of length at most @xmath97 , then @xmath1054 also has radius of convergence greater than @xmath1052 .
let @xmath1055 be a value at which all @xmath1056 converge .
for any @xmath208 set @xmath1057 and set @xmath1058 consider an @xmath1020 sufficiently large so that @xmath1059 ( later we impose other lower bounds on @xmath1020 s value ) .
let @xmath1032 .
let @xmath14 be a vlg whose underlying graph is ( the graph underlying ) @xmath194 , and whose @xmath937 edges have length at least @xmath482
. let @xmath1060 ( respectively , @xmath1061 ) be the number of irreducible walks in @xmath14 of length @xmath97 that are @xmath1035-selective ( respectively , and never traverse an @xmath937 edge ) .
let @xmath1062 be the generating functions of @xmath1063 , respectively .
the functions @xmath1062 are clearly majorized by the @xmath1062 in the case where @xmath1064 and all @xmath937 edges have length @xmath1065 .
we shall assume this to be the case .
first , we claim that @xmath158 is majorized by @xmath1066 this is because any walk in @xmath14 can be broken into alternating subwalks that remain in @xmath936 and steps along @xmath937 edges ; each time an @xmath937 edge is taken its length is at least @xmath1065 , and there are at most @xmath0 such @xmath937 edges from which to choose after finishing the @xmath936 walk such @xmath937 edges from which to choose except possibly at the very first step of a walk . ] . to prove the theorem it therefore suffices to show that @xmath1067 is less than one for sufficiently large @xmath1020 .
next , we claim that @xmath1068 is majorized by @xmath1069 assuming this , it is clear that for sufficently large @xmath1020 we have @xmath1070 and the theorem is proven . for @xmath1071 ,
let @xmath1072 be the number of @xmath1035-selective irreducible walks , @xmath1073 , in @xmath14 of length @xmath97 through only @xmath936 edges such that @xmath1074 is of length less than @xmath1020 .
the walk @xmath122 is contained in a subgraph of @xmath14 that is majorized by one of the @xmath1044 . clearly @xmath1075 when @xmath1076 , then from @xmath1077 the walk has at most @xmath0 possible directions to take ( or at most @xmath580 possible directions if @xmath1078 ) , and so @xmath1079 thus @xmath1080 is majorized by @xmath1081 . but any walk over @xmath936 edges can be broken into a series of walks of the form @xmath114 as above , plus a final walk of length less than @xmath1020 . the generating function for such walks is clearly majorized by @xmath1082 .
@xmath203 in order to use the selective trace , we must determine @xmath49 .
we begin by doing so for the model @xmath29 ; next subsection we use similar techniques for the models @xmath57 , @xmath65 , and @xmath69 . more generally , for a given @xmath12 , consider the task of finding the tangle , @xmath572 , in @xmath29 , of order at most @xmath12 with @xmath633 as large as possible . to simplify this task , notice that pruning leaves the order and @xmath551 invariant ( an irreducible closed walk can never visit a leaf , so pruning a leaf does nt affect the number of irreducible closed walks ) ; hence we may restrict our search to those @xmath572 s that are completely pruned .
[ lm : contract ] let @xmath14 be a graph with edge @xmath1083 with @xmath1084 .
let @xmath1085 be the _ contraction _ of @xmath14 along @xmath371 , i.e. the graph obtained by discarding @xmath371 and identifying @xmath267 with @xmath96
. then @xmath1086 . * proof * consider an irreducible closed walk , @xmath10 , about @xmath267 in @xmath14
. then we can associate to this closed walk one in @xmath1085 , @xmath1087 , by discarding all occurrences of @xmath371 .
this association , @xmath575 , is an injection , since given a @xmath1085 irreducible closed walk about @xmath267 of the form @xmath1087 , we can infer when @xmath371 was taken ( since @xmath1083 with @xmath1084 ) in the @xmath14 closed walk , giving rise to ( at most ) a single @xmath14 closed walk .
since this injection does not increase the length of the closed walks , we conclude that the number of irreducible closed walks about @xmath267 in @xmath14 of length @xmath1088 is no more than the number in @xmath1085 .
hence the conclusion of the lemma .
@xmath203 since edge contraction reduces the number of vertices and of edges by one each , edge contraction leaves the order invariant .
so in looking for a @xmath551 tangle of a given order , we may always assume the tangle has no edge contraction that leaves it a tangle , two vertices joined by between @xmath30 and @xmath3 edges is a tangle ( with apropriate @xmath398-labelling ) , but contracting any edge gives self - loops , which are not feasible in @xmath57 .
therefore edge contraction can take graphs that can be tangles to graphs that can not , at least for certain random graph models . ] .
we now claim ( by lemma [ lm : contract ] ) that for @xmath29 and @xmath1089 , a vertex with @xmath13 whole - loops has the largest @xmath551 of all tangles of order @xmath12 ( recall that each self - loop is counted as one edge , according to definition [ de : tangle_order ] ) . for this graph
we clearly have @xmath1090 ; hence @xmath49 is the smallest integer @xmath12 with @xmath1091 , provided that this @xmath12 is @xmath1092 .
but we easily verify that this @xmath12 , @xmath1093 is indeed at most @xmath1094 for all @xmath2 .
we have just established the following theorem . for the model @xmath29 , we have @xmath1095 . for @xmath57
we have to remember that tangles ca nt have self - loops .
thus contractions can only be done along non - multiple edges , and @xmath49 will not generally be the same for @xmath57 and @xmath29 .
[ lm : contract_distance_two ] let @xmath266 be vertices of distance two in a graph , @xmath14 , i.e. , there are no edges joining @xmath267 and @xmath96 , but there is a @xmath114 with edges to each of @xmath266 .
let @xmath497 be the graph obtained by identifying @xmath267 and @xmath96 and deleting one of the edges from @xmath114 to @xmath267 ( or to @xmath96 ) ( so that the order of @xmath497 is the same as that of @xmath14 ) .
then @xmath1096 .
* proof * let @xmath213 be the vertex in @xmath497 which is the identification of @xmath96 and @xmath267 .
let the edges from @xmath267 to @xmath114 be enumerated @xmath1097 , and those from @xmath96 to @xmath114 enumerated @xmath1098 .
the edges from @xmath213 to @xmath114 are @xmath1099 , where @xmath1100 .
first consider the case when @xmath1101 , and consider the irreducible closed walks about @xmath114 ( which are necessary of even length ) .
such a closed walk begins in @xmath114 and takes two steps , visiting either @xmath267 or @xmath96 , in , respectively , @xmath1102 or @xmath1103 ways .
after coming back from a @xmath267 vertex , another step of length @xmath30 can either ( 1 ) visit a @xmath267 vertex , in @xmath1104 ways , or ( 2 ) visit a @xmath96 vertex , in @xmath1103 ways ; similarly for coming back from a @xmath96 vertex .
thus `` coming back from a @xmath267 vertex '' and `` coming back from a @xmath96 vertex '' forms a markov chain , and the total number of irreducible closed walks of length @xmath97 about @xmath114 is @xmath1105 \left [ \begin{array}{cc}(s-1)^2&t(t-1)\\s(s-1)&(t-1)^2\end{array}\right]^{(k-2)/2 } \left [ \begin{array}{c}1\\1\end{array}\right]\ ] ] we wish to compare this to the number of irreducible @xmath497 closed walks about @xmath114 , of which there are clearly @xmath1106 for starters , we see @xmath1107 which is non - negative , since both @xmath1108 .
now since the maximum row sum in the @xmath1109 matrix of equation ( [ eq : w_cycles ] ) is @xmath1110 we have @xmath1111 for all @xmath97 . but @xmath1112 and @xmath1113 which is positive unless @xmath50 or @xmath717 is @xmath68 .
thus , provided that @xmath1114 and @xmath1115 , we have @xmath1116 and @xmath1117 for all even @xmath97 .
if @xmath1118 we calculate @xmath1119 and similarly when @xmath70 .
we shall use the above calculation below .
we can now assume that @xmath548 has a vertex , @xmath1120 , different from @xmath1121 .
there is a natural bijection of edges , @xmath575 from @xmath1122 to @xmath1123 .
extend @xmath575 to a map on all of @xmath389 by defining @xmath1124 and @xmath1125 to be a formal symbol @xmath208 .
for any irreducible @xmath14 closed walk about @xmath1120 specified by its edges , @xmath1126 with @xmath1127 , we associate a sequence @xmath1128 we claim that the number of @xmath10 with a given image @xmath1087 is no more than the number of @xmath1129 closed walks corresponding to @xmath1087 by changing all @xmath162 edges into @xmath208 s .
indeed , consider a block of consecutive @xmath208 s in @xmath1087 , i.e. @xmath1130 , and @xmath1131 and @xmath1132 ; @xmath1087 can not begin or end with an @xmath208 , since the closed walk begins at @xmath1120 , and so we can assume @xmath1133 and @xmath1134 . by looking at @xmath1135 and
@xmath1136 we can determine whether or not the @xmath208-block begins in @xmath267 , @xmath96 , or @xmath114 , and ends in @xmath267 , @xmath96 , or @xmath114 .
if the @xmath208-block begins in @xmath114 and ends in @xmath114 , then equations ( [ eq : compare_i ] ) and ( [ eq : compare_i_t1 ] ) show that there are no fewer @xmath497 sequences for the corresponding @xmath208-block than @xmath14 sequences .
next compare those @xmath208-blocks that begin in a @xmath267 and end in a @xmath114 .
the number of such sequences in @xmath14 is @xmath1137 \left [ \begin{array}{cc}(s-1)^2&t(t-1)\\s(s-1)&(t-1)^2\end{array}\right]^{(b - a)/2 } \left [ \begin{array}{c}1\\1\end{array}\right],\ ] ] whereas the number in @xmath497 is @xmath1138 ( since the non-@xmath208 edge @xmath1135 can be followed by any @xmath213 to @xmath114 edge in @xmath497 ) ; so the @xmath497 number is no less than the @xmath14 number for @xmath1139 ( since @xmath1140 ) , and each time @xmath1141 is increased by @xmath30 , the former number gets multiplied by an @xmath1142 , the latter gets multiplied by no more than @xmath1143 , where @xmath1144 , provided that @xmath1114 and @xmath1115 ; the @xmath70 or @xmath1118 case is easily checked to result in equality .
the same arguement holds for @xmath96 to @xmath114 @xmath208-blocks . for an @xmath208-block starting and ending in @xmath267 ,
we wish to compare @xmath1137 \left [ \begin{array}{cc}(s-1)^2&t(t-1)\\s(s-1)&(t-1)^2\end{array}\right]^{(b - a-1)/2 } \left [ \begin{array}{c}s-1\\0\end{array}\right],\ ] ] with @xmath1138 .
again , it suffices to compare when @xmath1145 , which is immediate , and to check @xmath70 or @xmath1118 separately .
we argue for @xmath208-blocks starting in either @xmath267 or @xmath96 and ending in either @xmath267 or @xmath96 similarly .
@xmath203 [ th : taufund_chnd ] for the model @xmath57 , we have @xmath59 .
* proof * as before , we consider a @xmath12 and search for those @xmath572 of order at most @xmath12 with @xmath633 as large as possible . by lemma [ lm : contract_distance_two ] , and by contractions ( in lemma [ lm : contract ] ) , we may restrict our search to those @xmath572 with two or more edges between every pair of nodes .
first assume that @xmath1146 .
if @xmath572 has two vertices , then @xmath572 has @xmath1147 edges joining the two vertices ( since there are no self - loops in @xmath57 ) . in this case @xmath1148 .
we claim that this is as large a @xmath551 as possible ( again , assuming @xmath1146 ) . indeed ,
if @xmath572 has @xmath1149 vertices , then the maximum degree of a vertex is @xmath1150 minus the edges not involved with that particular vertex , which is at least @xmath30 for each pair of the @xmath855 other vertices .
so the maximum degree is at most @xmath1151 since @xmath551 is no greater than the maximum degree minus @xmath68 , we have @xmath1152 it follows that if @xmath1149 , @xmath551 is strictly less than @xmath13
. to achieve @xmath1148 with our @xmath572 having two vertices , we required @xmath1146 . to get @xmath1148 to equal or exceed @xmath64 , we require @xmath1153 , for which we must have @xmath1154 since @xmath3 is an integer , this is equivalent to @xmath1155 which we easily see holds for all even @xmath160 except @xmath1156 .
we conclude that @xmath1157 for even @xmath1158 .
it suffices to analyze the cases @xmath1156 . for each order , @xmath12 , and @xmath1156
, we must examine those tangles of order @xmath12 and determine the largest possible @xmath633 .
let us note that if @xmath572 is a tangle of order @xmath682 , then it is a tree and has @xmath1159 .
if @xmath572 is a completely pruned tangle of order @xmath214 , then @xmath572 is a closed walk and has @xmath1160 .
if @xmath61 , then consider the tangle of order @xmath68 with three vertices , consisting of one `` middle '' vertex joined by two edges to each of two vertices .
( this is a tangle by labelling the left to middle edges and the middle to right edges @xmath707 . )
we easily compute @xmath1161 , as this graph is bipartite and the number of irreducible walks of length @xmath1162 from the middle vertex , all such walks being closed walks , is clearly @xmath1163 .
so for @xmath61 , @xmath1164 .
for @xmath705 , consider the tangle , @xmath1165 with @xmath1166 with three edges connecting @xmath708 to @xmath709 ( labelled @xmath1167 ) and two edges connecting @xmath709 to @xmath710 ( labelled @xmath707 ) .
we claim that @xmath1168 .
say that a closed walk about @xmath709 ends in `` state a '' if the last vertex before @xmath709 was @xmath708 , and otherwise in `` state b '' ( i.e. the second to last vertex is @xmath710 ) . from state
a , taking two additional irreducible steps , there are 4 ways to reach another state a , and two ways to reach another state b. from state b , taking two irreducible steps , there is one way to reach another state b and six ways to reach another state a. it easy follows that @xmath633 is the square root of the largest eigenvalue of @xmath1169.\ ] ] but @xmath170 of this matrix is @xmath1170 , and this @xmath170 is just @xmath633 .
it follows that @xmath1171 , and hence @xmath1172 .
we wish to rule out @xmath1164 when @xmath705 . since we are considering only completely pruned graphs , @xmath572 , each vertex has degree @xmath1173 .
such a graph , @xmath572 , of order @xmath68 has all vertices of degree @xmath30 except for one of degree @xmath722 or two of degree @xmath105 . in the case where there are vertices , @xmath266 , of degree @xmath105 , therefore joined by three disjoint beaded paths ,
then @xmath633 is greatest when the beaded paths are each of length @xmath68 ( by setting up an obvious map from irreducible closed walks about @xmath267 from the general graph to the one with beaded paths of length @xmath68 ) ; hence @xmath1174 in this case , since the graph of two vertices joined by three edges has @xmath1175 .
similarly , in the case with @xmath267 of degree @xmath722 , therefore having two beaded closed walks from @xmath267 , @xmath633 is greatest when the lengths of the two closed walks are two ( they can not be one since @xmath57 does not permit self - loops ) ; hence @xmath1176 in this case .
hence @xmath1177 and therefore @xmath1178 .
we conclude that @xmath1179 also when @xmath1156 .
@xmath203 for the model @xmath65 , we have @xmath1180 for all @xmath17 . *
proof * we argue as with @xmath57 .
the only difference is that in @xmath65 , two vertices can have as many as @xmath0 edges between them in a tangle ( as opposed to @xmath3 edges in an @xmath57 tangle ) .
so the argument in the previous thoerem shows that the two - vertex tangles give that @xmath1181 equals @xmath49 provided that @xmath1182 ( as opposed to @xmath1146 for @xmath57 ) .
but we easily verify that @xmath1183 for all @xmath17 ( indeed , we have equality for @xmath1184 , and each time we increase @xmath17 by one , @xmath64 increases by less than one ) .
@xmath203 for the model @xmath69 , we have @xmath1180 for all @xmath17 . * proof * as in @xmath65 , for any @xmath1185 there is a tangle @xmath1186 that is a pair of vertices with @xmath1147 edges joining them .
@xmath1186 has order @xmath12 and @xmath1187 ; since when @xmath1188 we have @xmath1189 giving a supercritical tangle , we need worry only about whether or not there is a tangle of order @xmath1185 that can beat the @xmath551 of @xmath1186 . again , as with @xmath65 , lemma [ lm : contract_distance_two ] can be applied to graphs with half - loops , and so by the same argument as for @xmath65 we have that only graphs on one or two vertices can possibly beat @xmath1186 .
so consider a graph on vertices @xmath266 with @xmath1190 half - loops about @xmath267 , @xmath10 half - loops about @xmath96 , and @xmath1191 edges from @xmath267 to @xmath96 .
an irreducible path traverses edges of four different states : ( 1 ) half - loops about @xmath267 , ( 2 ) edges from @xmath267 to @xmath96 , ( 3 ) edges from @xmath96 to @xmath267 , and ( 4 ) half - loops about @xmath96 .
now we write a transition matrix about the states : for example in state ( 1 ) we may either continue on one of @xmath1192 half - loops in state ( 1 ) or continue on one of @xmath1191 edges in state ( 2 ) .
we find the transition matrix @xmath1193,\ ] ] and @xmath551 of our graph is this matrice s largest eigenvalue .
the order of the graph is @xmath1194 ( recall , each half - loop contributes one to the order of a graph ) .
but the row sum is never greater than @xmath1195 ( and always less unless @xmath1190 or @xmath10 vanishes ) , and so if this graph has order @xmath12 its @xmath551 is no more than @xmath13 .
hence no @xmath69 tangle of order @xmath12 beats @xmath1186 , provided that @xmath1185 .
thus @xmath49 is the smallest number with @xmath1196 .
in this section we discuss ramanujan functions in order to ( 1 ) explain their significance , and ( 2 ) give some intuition on some very technical issues surrounding the asymptotic expansion for irreducible traces ( as in section [ se : expansion ] ) .
a function , @xmath153 , on positive integers , @xmath97 , is said to be _
@xmath0-ramanujan of order @xmath225 _ if there is a polynomial @xmath154 and a constant @xmath155 such that @xmath1197 for all @xmath97 .
we call @xmath157 the _ principal term _ of @xmath158 , and @xmath159 the _ error term _ ( both terms are uniquely determined if @xmath1198 ) . a function is _
super-@xmath0-ramanujan _ if it is @xmath0-ramanujan of order @xmath68 . a @xmath0-ramanujan function as defined before , in definition [ de : ram ] , is just a @xmath0-ramanujan of order @xmath64 .
let @xmath1199 be the number of irreducible cycles of length @xmath97 in a @xmath0-regular graph .
then in @xcite it is shown that if @xmath1199 is @xmath0-ramanujan , then any eigenvalue , @xmath1200 , of the graph satisfies @xmath1201 .
the discussion there also shows that in any case , if @xmath181 is the eigenvalue of largest absolute value @xmath1202 , then @xmath1199 is @xmath0-ramanujan of order @xmath339 with @xmath1203 ( and not for any smaller an @xmath339 ) .
any discussion of irreducible traces and eigenvalues is bound to be tied to @xmath0-ramanujan functions .
one important property of @xmath0-ramanujan functions of order @xmath339 is that they are closed under addition .
another very important property is that they are closed under _
convolution _ , which we now formally explain .
this property will be used in section [ se : finish ] , and refined versions of it will be used in section [ se : expansion ] .
[ th : baby_convolute ] let @xmath1204 be @xmath0-ramanujan of order @xmath339 with @xmath1198 . then their convolution , @xmath1205 is also @xmath0-ramanujan of order @xmath339 . the techniques in section [ se : expansion ]
prove a more precise version of this theorem ( keeping track of the sizes of the the error term and the coefficients of the principal part ) ; for this reason we keep the argument below concise .
* proof * for @xmath1206 let @xmath1207 where @xmath791 are polynomials and the @xmath1208 are bounded by @xmath1209 for some @xmath97 .
we may also write @xmath1210 since convolution is bilinear , we easily see @xmath1211 it suffices to show that @xmath1212 are @xmath0-ramanujan of order @xmath339 .
according to sublemma 2.15 of @xcite , @xmath1213 is a polynomial .
next @xmath1214 where @xmath1215 writing @xmath1216 we see that @xmath1217 is a polynomial , and @xmath1218 is bounded by @xmath1219 ( see section 8 , especially lemma [ lm : useful_aux ] , for details ) .
this shows @xmath1220 is @xmath0-ramanujan of order @xmath339 .
similarly , so is @xmath1221 ; @xmath1222 is easily also seen to be so ( with zero principal term ) .
in this section we prove the first crucial expansion theorem .
our second such theorem , theorem [ th : with ] , will extend these ideas .
[ th : main_expansion ] let @xmath250 be a positive integer , and let @xmath990 be a set of tangles containing all supercritical tangles of order less than @xmath250 .
then there is an @xmath1223 such that for all @xmath1021 the following holds .
we have @xmath1224 } = f_0(k)+\frac { f_1(k)}{n}+\cdots+ \frac { f_{r-1}(k)}{n^{r-1}}+\frac{{\rm error}}{n^r},\ ] ] where the @xmath230 are @xmath0-ramanujan and the error term satisfies the bound given in theorem [ th : ssiicexpansion ] .
we begin by explaining why this theorem is an easy consequence of theorem [ th : ssiicexpansion ] and the following theorem .
[ th : reduce1 ] fix a lettering , @xmath1225 , of type @xmath194 , and fixed non - negative integers @xmath1226 .
let @xmath1227 where the sum is over all tuples of words @xmath1228 such that each @xmath1229 is irreducible and of length @xmath949 and is compatible with @xmath1225 .
let @xmath579 be a @xmath482-new type based on @xmath194 , and let @xmath1230 ( with @xmath262 as in theorem [ th : crucial_cycle_count ] ) .
then @xmath158 is @xmath0-ramanujan for all @xmath1231 .
assume theorem [ th : reduce1 ] for the moment .
let @xmath400 ( respectively , @xmath1232 and @xmath1233 ) be the walk collections corresponding to irreducible , @xmath989-selective walks ( respectively , and that are associated to the type @xmath194 and @xmath579 ) .
these walk collections are all ssiic . according to theorem [ th :
ssiicexpansion ] @xmath851 where @xmath1234){\rm \;order\;}j,\in{{\cal w}}(k ) } p_{i - j}\bigl ( a_1(w;[\vec t\>]),\ldots , a_d(w;[\vec t\>])\bigr).\ ] ] it suffices to show that these @xmath230 are @xmath0-ramanujan .
we know there are finitely many types of order at most @xmath855 .
now fix @xmath1020 to be the max over @xmath1235 over types , @xmath194 , of order at most @xmath855 ( and @xmath1236 as in theorem [ th : reduce1 ] ) . for any @xmath1021 , choose @xmath1237 ; of course , @xmath1238 for any type , @xmath194 , of order at most @xmath855 .
we know there are finitely many @xmath482-new types based on any type .
so the sum involving @xmath400 in equation ( [ eq : want_to_be_ram ] ) decomposes as a finite sum over @xmath1232 s or @xmath1233 s .
furthermore , each expansion polynomial @xmath1239 , involving a fixed new type , @xmath579 , is just a function of @xmath1240 over appropriate forms , and each @xmath652 of the form is just the sum of the @xmath652 along each edge of the form
. therefore theorem [ th : reduce1 ] just says that each @xmath230 , when summing over a @xmath1233 , is @xmath0-ramanujan .
summing over all @xmath1233 shows that the @xmath230 corresponding to @xmath400 are also @xmath0-ramanujan .
* proof ( of theorem [ th : reduce1 ] ) * clearly it suffices to prove the following theorem with @xmath1241 replaced by @xmath1242 with @xmath1243 any set of non - negative integers .
our lemma [ lm : irdeigens ] reduces the above theorem to the following .
[ th : reduce3 ] with notation as in theorem [ th : reduce1 ] , let @xmath1244 be a partition of @xmath1245 , and let @xmath1246 for @xmath1247 denote the sum of the @xmath1248 in @xmath1249 .
then for fixed non - negative integers @xmath1250 , theorem [ th : reduce1 ] holds with @xmath1241 replaced by @xmath1251 more generally , theorem [ th : reduce1 ] holds with @xmath1241 replaced by @xmath1252 where the edges are ordered so that @xmath1253 and where @xmath1254 is a function such that @xmath1255 for some constant @xmath10 .
the @xmath667 of equation ( [ eq : specific_r ] ) is all that is needed for @xmath29 ; it will be convenient ( if not necessary ) to use the @xmath667 of equation ( [ eq : general_r ] ) for @xmath69 ( see section 14 ) . * proof * it suffices to deal with the @xmath667 of equation ( [ eq : general_r ] ) .
let @xmath1256 we may assume @xmath1257 and @xmath1258 ; set @xmath1259 clearly it suffices to prove the theorem for @xmath1260 understanding that @xmath1261 are fixed by @xmath1262 .
the _ ( coefficient ) norm _ , @xmath1263 , of a polynomial , @xmath874 ( which is possibly multivariate ) , is the largest absolute value among its coefficients . working with this notion of a norm
is a bit `` weak , ''
( i.e. , sometimes much stronger statements would hold with other norms ) , but this notion is sufficient for our purposes .
let @xmath1264 [ th : f_vecm_est ] for any vector of positive integers , @xmath664 , we have @xmath1265 is @xmath0-ramanujan with principal term @xmath1266 and error term @xmath1267 satisfying @xmath1268 and @xmath1269 where @xmath10 depends only on the @xmath1270 and @xmath1254 .
* proof * fix a value of @xmath664 . without loss of generality we may assume @xmath1271 and @xmath1272 . for now assume that @xmath1273 ; we will later indicate the minor changes needed for the situation @xmath1274 ( i.e. when there are no @xmath651 belonging to @xmath1275 that equal one ) .
let @xmath1276 [ lm : crucial_b_type_est ] if @xmath579 is a @xmath482-new type , then @xmath1277 for some constant @xmath10 depending only on @xmath579 .
* proof * since each @xmath949 is at most @xmath250 , @xmath1278 is bounded by @xmath1279 , and it suffices to prove the estimate for @xmath1280 replaced with @xmath1281 but there are only @xmath1282 ways of writing @xmath250 as the sum of @xmath1283 positive integers .
so it suffices to show @xmath1284 now we have @xmath1285 so @xmath1286 @xmath1287 since @xmath969 for @xmath88 between @xmath1288 and @xmath267 , and since @xmath1289 for @xmath1290 ( and @xmath1291 for all @xmath88 ) , we conclude @xmath1292 @xmath1293 @xmath1294 @xmath1295 hence @xmath1296 but @xmath1297 are bounded by the number of edges in @xmath579 . @xmath203 we will need another lemma .
[ lm : q ] for any non - negative integers @xmath1298 there is a polynomial @xmath1299 such that for all @xmath1300 we have @xmath1301 * proof * this is a special case of sublemma 2.15 of @xcite ( proven in a straightforward induction on @xmath50 ) .
@xmath203 now let @xmath1302 ( so that @xmath1303 ) .
in the notation of the above two lemmas , letting @xmath1304 , we have @xmath1305 @xmath1306 here we sum until @xmath1307 since lemma [ lm : q ] requires @xmath1300 ( and @xmath1308 vanishes for @xmath1309 ) , and we sum from @xmath815 to simplify the expression , despite the fact that @xmath1310 clearly vanishes for @xmath1311 the sum in equation ( [ eq : qgsum ] ) is clearly @xmath1312 , where @xmath1313 assuming these sums converge .
we claim @xmath1314 will be the principal part of @xmath1315 , and @xmath1316 will be the error term .
first we need the following lemma .
[ lm : useful_aux ] for any positive integer , @xmath598 , there is a @xmath1317 such that the following holds .
let @xmath1318 be a function defined on non - negative integers , @xmath250 , such that @xmath1319 , with @xmath1320 .
let @xmath1321 be any polynomial of degree at most @xmath598 .
then ( 1 ) the infinite sum @xmath1322 is convergent ( in coefficient norm ) , ( 2 ) the degree of @xmath245 is that of @xmath1299 , and ( 3 ) we have @xmath1323 the same is true for the sum @xmath1324 for any positive integer @xmath267 , except that we replace the last claim with the estimate @xmath1325 ( however , the @xmath1317 in this equation might need to be larger than that in the estimate for @xmath245 ) . * proof * first we observe that since @xmath1326 we have @xmath1327 where @xmath1328 is some polynomial of degree at most @xmath1329 .
we first prove the claims on @xmath245 . by the binomial theorem , and the fact that @xmath1299 s degree is bounded , it suffices to examine only the cases where @xmath1330 is replaced by @xmath1331 for @xmath1332 . in this case
@xmath1333 becomes @xmath1334 and we have @xmath1335 this establishes the claim on @xmath245 .
the claim on @xmath1336 is reduced to @xmath245 via @xmath1337 where @xmath1338 and @xmath1339 .
so @xmath1340 satisfies the same estimate as does @xmath631 , except with an extra factor of @xmath1341 ; the binomial theorem implies that @xmath1342 is at most @xmath1343 times a constant depending on @xmath598 . @xmath203 we continue with the proof of theorem [ th : f_vecm_est ] .
we have @xmath1344 = \sum_{r=1}^\infty q(k - r ) \tilde g(r),\ ] ] where @xmath1345 .
now @xmath1299 is fixed in the theorem , so @xmath1346 can be regarded as a constant .
also , since @xmath1347 according to lemma [ lm : crucial_b_type_est ] , we have @xmath1348 it follows that @xmath1349 , where @xmath245 is a polynomial with @xmath1350 assuming that @xmath160 ( so that @xmath1351 with @xmath1352 is strictly positive ) .
furthermore , lemma [ lm : useful_aux ] also implies that @xmath1353 @xmath1354 now we see that @xmath1355 is the decomposition of @xmath1315 into principal and error terms , as claimed before , and that these terms satisfy the bounds stated in theorem [ th : f_vecm_est ] .
finally we indicate the minor changes when @xmath1274 . in this case
we take @xmath1299 to be the function @xmath1356 and @xmath1299 vanishing elsewhere .
then @xmath1357 , so lemma [ lm : crucial_b_type_est ] shows that @xmath1265 is @xmath0-ramanujan with zero principal part .
@xmath203 we continue with the proof of theorem [ th : reduce3 ] .
we are studying @xmath1358 set @xmath1359 so that @xmath1360 theorem [ th : crucial_cycle_count ] combined with theorem [ th : f_vecm_est ] gives that for fixed @xmath1361 we have that @xmath1362 is @xmath0-ramanujan with principle term @xmath1363 and error term @xmath1364 where @xmath1365 @xmath1366 and the degree of @xmath1367 is bounded independent of @xmath1361 .
so we sum over all @xmath1361 to conclude that @xmath158 is @xmath0-ramanujan .
let us review our general approach to the alon conjecture .
we are interested in expansions in @xmath101 of the expected value of the @xmath97-th irreducible trace .
unfortunately these expansions have some coefficients that fail to be @xmath0-ramanujan , and , as explained in section 2 , this prevents us from proving the alon conjecture .
replacing irreducible traces with selective traces gives @xmath0-ramanujan coefficients up to any desired power of @xmath101 .
however , selectivity , in the presence of appropriate tangles , modifies the irreducible trace in a way that seems hard to control .
thus we do nt know how to use the results of the last section to conclude the alon conjecture .
the last main idea of the proof is to get an expansion for the expected value of selective traces counted only when appropriate tangles are present ( i.e. , the selective trace multiplied by a characteristic function over those graphs in @xmath29 with appropriate tangles ) .
the methods of the last section generalize , rather tediously , to such expansions .
these expansions will also have @xmath0-ramanujan coefficients up to any desired power of @xmath101 .
it follows that we also get such expansions for the expected value of the selective trace counted only when appropriate tangles are _ not _ present ; for this count , the selective trace and irreducible trace are the same .
this information turns out to be enough to prove the alon conjecture ( with an auxilliary lemma proven in section 11 ) . before doing the above
, it is crucial to know that a certain set of tangles is finite .
recall that @xmath1015 is the set of supercritical tangles , i.e. , whose @xmath551 is at least @xmath64 . recall that @xmath1018 $ ] is the subset of elements of @xmath1015 of order at most @xmath250 .
let @xmath1368 $ ] be the set of tangles of @xmath1018 $ ] that are minimal with respect to inclusions , i.e. , that do nt have another element of @xmath1018 $ ] properly included in it .
[ lm : finiteness ] the set @xmath1368 $ ] is finite .
this means that containing a supercritical tangle of order at most @xmath250 is equivalent to containing one of a finite set of tangles .
* proof * assume that @xmath1368 $ ] is not finite .
with each tangle we associate a type which is the labelled graph obtained by supressing the degree two vertices .
since there are finitely many types of order at most @xmath250 , there must be an infinite number of @xmath1368 $ ] tangles of some type , @xmath194 . by passing to a subsequence
we may assume there is an infinite sequence of @xmath1368 $ ] elements , @xmath1369 , such that for each edge of @xmath194 the associated labelling is either constant or has length tending to infinity ; furthermore , the length must tend to infinity along at least one @xmath194 edge .
let @xmath1370 be the limiting tangle , where we discard all edges with length tending to infinity .
we claim @xmath1371 , where @xmath1372 is the vlg with underlying directed graph @xmath1373 , and where @xmath1374 has length equal to the length , @xmath1375 , of @xmath708 in @xmath1376 ( recall @xmath708 can be viewed as a directed edge of @xmath194 ) ; indeed , with a vertex path @xmath1377 in @xmath1373 with @xmath1378 for @xmath1379 , we associate the walk @xmath1380 , which has @xmath1376 length equal to the sum of the @xmath1381 . for a closed walk , where @xmath1382 , its length ( in @xmath1372 ) @xmath1383 , which corresponds to a unique @xmath1384 closed walk of the same length , arising from the subdivided @xmath1385 .
this correspondence is clearly a length preserving bijection between @xmath1372 closed walks and @xmath1384 closed walks . hence @xmath1371 .
similarly @xmath1386 with @xmath1387 defined similarly .
now by theorem [ th : infinite_edge ] , @xmath1388 , and so @xmath1389 .
also @xmath1370 s order is less than that of the @xmath1376 ( because of the edge removal(s ) ) . hence @xmath1370 is again a @xmath1018 $ ] tangle . but
@xmath1370 properly contains ( all ) @xmath1376 , which contradicts the supposed minimality of the @xmath1376 .
hence @xmath1368 $ ] is finite .
@xmath203 we illustrate the above lemma with an example .
let @xmath1376 be a sequence of tangles whose underlying graph is the same except for one beaded cycle of length @xmath88 about some vertex .
this infinite collection of tangles would prove troublesome to the methods of this section .
however either ( 1 ) @xmath1390 for some @xmath88 , at which point only finitely many of the @xmath1376 are relevant , or ( 2 ) the limiting tangle , @xmath1370 , has @xmath1391 , in which case a @xmath1376 inclusion implies a @xmath1370 inclusion .
[ th : with ] let @xmath990 be a finite set of pruned ( nonempty ) tangles of order at least @xmath68 .
let @xmath1392 be the indicator function of the event that @xmath404 contains a ( i.e. , at least one ) tangle from @xmath990 , i.e. , @xmath1393 let @xmath1394 be a set of tangles including all supercritical tangles of order less than @xmath250 .
then for any @xmath250 there is an @xmath1395 such that for all @xmath1021 we have an expansion @xmath1396 } = f_0(k)+\frac { f_1(k)}{n}+\cdots+ \frac { f_{r-1}(k)}{n^{r-1}}+\frac{{\rm error}}{n^r},\ ] ] where the @xmath230 are @xmath0-ramanujan and the error term satisfies @xmath1397 with @xmath10 and @xmath1398 depending only on @xmath250 , @xmath990 , and @xmath1394 .
we believe this theorem is true even if @xmath990 contains cycles , i.e. , tangles of order @xmath214 . but to prove this would be more difficult , since the last part of lemma [ lm : truncate ] would not be true ( see the remark that follows this lemma ) . before giving the proof of theorem [ th : with ]
, we give an important corollary of it and theorem [ th : main_expansion ] . with notation and conditions as in theorem
[ th : with ] , we have that @xmath1399 } \ ] ] also has an expansion of the form given by the right - hand - side of equation ( [ eq : with_expansion ] ) .
* proof * * ( of theorem [ th : with ] ) * we want to generalize walk sums , forms , types , etc . into structures that incorporate the presence of a @xmath990 tangle .
we shall first explain why we run into inclusion / exclusion and tangle automophisms ( as in theorem [ th : tangle_count ] ) .
potential tangle specialization _ as a pair , @xmath1400 , of a tangle , @xmath1401 , and an inclusion @xmath1402 .
let @xmath1403 denote the indicator function of the event that @xmath1401 is contained in @xmath404 via the map @xmath123 .
let @xmath1404 be the indicator function of the event , @xmath1405 , of the potential walk @xmath1406 .
we plan to study sums involving various @xmath1407 .
for now consider @xmath1403 alone .
since the word `` form '' was used earlier to mean `` look at the graph traced out and forget the specific values of the vertices , '' it makes sense to view @xmath1401 itself as the `` form '' of @xmath1400 , and let @xmath1408 } _
n = \sum_\sigma { \mbox{e}\left[\chi_{\omega,\sigma}\right ] } = \frac{n!}{(n - v)!}\prod_{i=1}^{d/2 } \frac{(n - a_i)!}{n!},\ ] ] where , as usual , the @xmath652 are the number of @xmath1401 edges labelled @xmath25 and @xmath96 is the number of vertices in @xmath1401 .
the problem is that @xmath1409 } _
n$ ] gives the expected number of times @xmath1401 is included into a random graph ; when computing traces and walk sums , we do wish to count ; but when looking at tangle inclusions , when do nt wish to count we want to compute only the function @xmath1392 , i.e. , we want a @xmath68 when a @xmath990 tangle is present , and a @xmath214 otherwise
. given two tangles , @xmath1410 , let @xmath1411 denote the number of inclusions of @xmath572 into @xmath658 .
of course , @xmath1412 is the number of automorphisms of @xmath572 .
also , @xmath1409 } _
n$ ] is just the expected value of @xmath1413 for a @xmath404 ( viewing @xmath14 as a tangle ) .
let @xmath1414 ( respectively , @xmath1415 ) denote that @xmath572 has an inclusion ( respectively , proper inclusion ) into @xmath658 . by a @xmath990-tangle
we mean any tangle isomorphic to an element of @xmath990 .
a _ derived tangle of @xmath990 _ is a tangle that is the nonempty union of @xmath990-tangles .
let @xmath1416 be a collection of one derived tangle of @xmath990 in every tangle isomorphism class .
there exist reals numbers @xmath1417 such that for any @xmath1418 we have @xmath1419 * proof * this is just generalized mbius inversion : @xmath1416 is a partially ordered set , and given @xmath1420 there are only finitely many @xmath1401 with @xmath1421 . furthermore @xmath1422 is positive for all @xmath1420 .
so we can inductively solve for @xmath1423 .
@xmath203 it follows that @xmath1424 } & = & \sum_{(w,\vec t\>)\in{{\cal w } } } { \mbox{e}\left[\chi_{(w;\vec t\;)}\chi_{{\psi}}\right ] } \\ \label{eq : mother2 } & = & \sum_{(w,\vec t\>)\in{{\cal w } } } \quad\sum_{(\omega,\sigma),\omega\in{{\psi}}^+ } { \mbox{e}\left[\chi_{(w;\vec t\;)}\chi_{\omega,\sigma}\right ] } { \mu}_\omega,\end{aligned}\ ] ] where @xmath400 is the walk collection corresponding to the irreducible @xmath1425 selective walks .
[ lm : truncate ] let @xmath1426 be those elements of @xmath400 of order less than @xmath250 , and similarly for @xmath1427 . in equation ( [ eq : mother2 ] ) , by replacing the summation over @xmath400 and @xmath1416 by summation over @xmath1426 and @xmath1427 ,
the difference is at most @xmath1428 .
furthermore , the set @xmath1427 is finite for each @xmath250 .
we remark that if @xmath990 could contain a tangle of order @xmath214 , i.e. , one whose underlying graph is a cycle , then @xmath1429 would be infinite , containing arbitrarily large disjoint unions of tangles whose underlying graph is a cycle . * proof * we know that we can ignore @xmath1406 s of order at least @xmath250 by using theorem [ th : ssiicexpansion ] and the fact that @xmath1430 in equation ( [ eq : mother1 ] ) .
this leaves us with a sum over @xmath1426 and @xmath1416 , and it is left to see what happens when @xmath1416 is replaced by @xmath1427 . for the @xmath1400 s , notice that every tangle in @xmath1416 of order at least @xmath250 must contain a tangle of order between @xmath250 and @xmath1431 , where @xmath50 is the maximum number of edges in a @xmath990-tangle ; this follows because taking the union of any graph with a @xmath990-tangle increases the number of edges by at most @xmath50 .
lemma [ lm : order_increases ] shows that there are only finitely many elements of @xmath1416 of order at most @xmath1431 ( and that any such element is the union of at most @xmath1431 elements of @xmath990 ) .
theorem [ th : tangle_count ] now shows that there is a @xmath1432 such that a graph contains a @xmath1416-tangle of order at least @xmath250 with probability at most @xmath1433 .
let @xmath1427 be the subset of tangles of order less than @xmath250 in @xmath1416 . then @xmath1434 where @xmath1435 is a function bounded by a @xmath68 plus the finite sum of all @xmath1436 over @xmath1437 .
truncating the sum in equation ( [ eq : mother2 ] ) to @xmath1437 therefore introduces an error of at most @xmath1438 } \le d(d-1)^{k-1 } c n^{-r } \max(h).\ ] ] @xmath203 according to lemma [ lm : truncate ] , it suffices to fix an @xmath1439 and to show that @xmath1440 } \ ] ] has an expansion like the right - hand - side of equation ( [ eq : with_expansion ] ) .
so fix an @xmath1439 ; we now define the `` form '' of @xmath723 , but we incorporate into the form the information of how @xmath1441 overlaps with @xmath723 by allowing @xmath1401 and @xmath894 to share vertices and edges . [ de : omega_isom ] an @xmath1401-specialization of a form , @xmath894 , is an inclusion @xmath1442 .
an @xmath1401-isomorphism of forms is an isomophism of forms @xmath1443 which is the identity from @xmath1444 to @xmath1445 .
we introduce the notation @xmath1446 } _
n = \frac{n!}{(n - v)!}\prod_{i=1}^{d/2 } \frac{(n - a_i)!}{n!},\ ] ] where @xmath1447 and @xmath652 is the number of @xmath25 labels occurring in @xmath1448 . given @xmath1406 and @xmath1441 as above
, we set @xmath1449 $ ] to be the set of all pairs , @xmath1450 , such that there is a permutation of the integers taking @xmath725 to @xmath754 and @xmath123 to @xmath12 ; we set @xmath1449_n$ ] to the same , with the additional requirement that the components of @xmath754 and the image of @xmath12 lie in @xmath4 .
if we let @xmath894 be the form of @xmath114 , sharing vertices and edges with @xmath1401 where @xmath123 and @xmath725 `` overlap , '' then @xmath1446 } _ n = \sum _ { ( \vec s,\tau)\in [ \vec t;\sigma]_n } { \mbox{e}\left [ \chi_{(w;\vec s ) } \chi_{\omega;\tau } \right ] } .\ ] ] it follows that the expression in equation ( [ eq : omega_aim ] ) is just @xmath1451 } _ n,\ ] ] summed over one @xmath894 from each @xmath1401-isomophism class ( compare equation ( [ eq : fundamental ] ) ) .
we now go through the rest of sections 5 , 6 , and 8 , indicating how to modify our results to allow deal with sums in equation ( [ eq : omega_fundamental ] ) .
let us begin by remarking that the expansion of theorem [ th : exp_polys ] generalizes easily here , in that @xmath1446 } _
n = n^{v - e } \biggl ( p_0+\frac{p_1}{n}+\cdots+\frac{p_q}{n^q } + \frac{{\rm error}}{n^{q+1 } } \biggr)\ ] ] where the @xmath1452 are the expansion polynomials , and @xmath1453 ( by equation ( [ eq : error_term ] ) ) . since
our potential walk is of length @xmath97 , we have @xmath1454 and @xmath1455 since @xmath1401 is fixed , it follows that the error is bounded by @xmath1456 in the range @xmath813 , since the order of @xmath1448 is at least @xmath68 ( since @xmath1448 is pruned and contains an element of @xmath990 ) .
we next claim that given @xmath114 ( and our fixed @xmath1401 ) , the number of equivalence classes @xmath1449 $ ] corresponding to @xmath114 and @xmath1401 is bounded by @xmath1457 ; indeed , lemma [ lm : equiv_classes ] shows there at most @xmath1458 @xmath725 classes .
the additional @xmath1459 choices of @xmath123 values can be chosen from at most @xmath1460 `` old '' values plus one `` new '' value each time , for a total number of equivalence classes of at most @xmath1458 times @xmath1461 , which is the claimed bound .
we obtain that the sum in equation ( [ eq : omega_fundamental ] ) is @xmath1462\ ] ] with @xmath1463 the expansion polynomials , with @xmath1464 to attack the expansion polynomial sums in equation ( [ eq : omega_gamma_exp ] ) , we introduce straightforward generalizations of types and new types . an @xmath1401-type is an oriented graph @xmath925 , with vertex and edge partial numberings such that ( 1 ) @xmath1401 is a subgraph of @xmath1465 , ( 2 ) all vertices of @xmath1465 except possibly the first one and possibly @xmath1466 vertices are of degree at least @xmath105 , and ( 3 ) all vertices and edges not in @xmath1401 are numbered . to a form , @xmath894 , we associate its @xmath1401-type , @xmath194 , by taking @xmath1448 and supressing the degree @xmath30 vertices in @xmath1467 .
@xmath194 s edges are partially numbered since we number the edges in they order they must be traversed by a corresponding walk , and we do nt number an edge if the walk does nt traverse the edge ( such edges are the edges of @xmath1468 ) . a @xmath482-new @xmath1401-type is a collection @xmath930 , ( 1 ) a lettered @xmath1401-type , @xmath194 , ( 2 ) a partition of @xmath931 into two sets , @xmath932 , ( 3 ) for each @xmath933 an _ edge length _ , @xmath934 , with @xmath935 , and ( 4 ) a @xmath428-labelling of @xmath936 with each @xmath933 labelled with a word of length @xmath934 .
furthermore , we require that @xmath1469 is contained in @xmath936 , and that each @xmath1469 edge has length @xmath68 in @xmath579 and is labelled with its @xmath1401 label in @xmath579 .
so conditions ( 2)(4 ) are just as for a new type ( see definition [ de : new_type ] ) , and condition ( 1 ) here involves an @xmath1401-type instead of a type .
again , it is easily seen that there are finitely many ( isomorphism classes ) of @xmath1401-types of a given order , and finitely many @xmath482-new @xmath1401-types belonging to a type , @xmath194 , for a given @xmath482 ( and @xmath194 ) .
it suffices to show that for @xmath482 and @xmath208 sufficiently large with @xmath1009 we have that for any @xmath482-new @xmath1401-type , @xmath579 , and any polynomial , @xmath874 , in the @xmath652 s we have @xmath1470 is @xmath0-ramanujan . for integers
@xmath1471 indexed over @xmath635 edges of @xmath931 , set @xmath1472 to the number of walk classes of new type @xmath579 , traversing the edge @xmath635 @xmath651 times , that are irreducible @xmath1425 selective cycles and that respect the @xmath194 partial numbering and orientation ; we assume @xmath951 or @xmath1473 according to whether or not @xmath635 is numbered ( i.e. , @xmath635 is to be traversed on our walk ) ; assuming @xmath579 is a @xmath482-new type with @xmath1009 , this number does not depend on the `` long '' ( i.e. , @xmath937 ) edge lengths . index the edges so that @xmath1474 and @xmath1475 , with @xmath1476 the numbered edges ( so @xmath1477 ) .
we claim that the proof of theorem [ th : crucial_cycle_count ] shows that @xmath1478 satisfies the bound @xmath1479 for @xmath1021 for some @xmath10 , @xmath1020 , and @xmath1480 depending only on @xmath579 .
this is because the argument of theorem [ th : crucial_cycle_count ] is unaffected by the two essential new features that @xmath194 has over types , which are the possible presence of ( 1 ) some more degree @xmath30 vertices ( other than just the vertex numbered @xmath68 ) , and ( 2 ) some edges whose lengths are fixed at @xmath68 ( namely @xmath1469 edges ) . but each @xmath1481 , which is the number of edges labelled @xmath25 in @xmath1448 , has @xmath1482 and @xmath1483 is fixed in a new @xmath1401-type , @xmath579 ( by the edge partial numbering ) .
thus for any polynomial @xmath874 and new typoid , @xmath579 , there is a polynomial @xmath1484 such that @xmath1485 but it is easy to see that theorem [ th : reduce1 ] holds if we extend the definition of type and new type to allow any fixed number of vertices of degree @xmath30 .
furthermore the condition that @xmath894 belongs to @xmath579 is the same as @xmath894 belonging to this extended notion of new type .
we conclude that either side of equation ( [ eq : almost_done ] ) is @xmath0-ramanujan .
we wish to use theorem [ th : with ] to estimate eigenvalues . however , it is easier to use _ strongly irreducible traces _ rather than irreducible traces .
we explain this and develop the properties of the strongly irreducible trace in this section .
a word @xmath1486 is _ strongly irreducible _ if @xmath114 is irreducible and @xmath109 with @xmath1487 . for any of our irreducible traces , selective irreducible traces , irreducible walk sums , etc .
, we can form its `` strongly irreducible '' version where we discard contributions from words that are not strongly irreducible . in any graph , labelled or not , one can speak of strongly irreducible closed walks as those closed walks that are irreducible and whose last step is not the opposite of its first step .
the @xmath97-th strongly irreducible trace of a graph , @xmath14 , is the number of strongly irreducible closed walks of length @xmath97 for a positive integer @xmath97 ; we denote it @xmath1488 or @xmath1489 if @xmath106 is the adjacency matrix of @xmath14 .
if @xmath14 has half - loops , we consider each half - loop to be a strongly irreducible closed walk ( of length 1 ) and include it in our count for @xmath1490 or @xmath1491 .
half - loops are only a concern for us in the model @xmath69 ; the reason that we count half - loops as strongly irreducible is to make lemma [ lm : main_sit ] hold . with a closed walk of length @xmath97 in @xmath542 about a directed edge , @xmath371 , of @xmath14 , we may associate the strongly irreducible closed walk in @xmath14 about the vertex in which @xmath371 originates .
it follows that if @xmath1492 are the eigenvalues of @xmath542 , we have @xmath1493 for all @xmath850 ; for @xmath1494 we must add the number of half - loops to the right - hand - side of the above equation , since there is no edge in @xmath542 from a vertex to itself when the vertex corresponds to a half - loop .
we will study the relationship between @xmath1495 and @xmath1488 and its consequences .
the most important consequence is the following theorem .
[ th : strongtrace ] for @xmath1496 , let @xmath1497 and set @xmath1498 if @xmath14 is a @xmath0-regular graph with no half - loops and adjacency matrix eigenvalues @xmath1499 , then @xmath1500 furthermore , if instead @xmath14 has no whole - loops , then the same is true with @xmath1501 replaced by @xmath1502 where @xmath1503 furthermore , we shall see that @xmath1501 , like the @xmath173 of lemma [ lm : chebyshev ] , are polynomials of degree @xmath97 that may alternatively be expressed as a simple linear combination of chebyshev polynomials ( plus the @xmath1504 eigenvalue contribution for the @xmath1501 ) .
( in a sense , equation ( [ eq : q_k_sum ] ) says that to each eigenvalue , @xmath181 , of @xmath14 , there correspond eigenvalues @xmath1505 of multiplicity one each and eigenvalues @xmath68 and @xmath682 of multiplicity @xmath1506 each in @xmath542 . )
[ lm : main_sit ] let @xmath14 be a @xmath0-regular graph on @xmath16 vertices with @xmath245 half - loops .
then for all integers @xmath850 we have the summation in the formula to follow is ignored , since it ranges from @xmath46 to @xmath1507 . ]
@xmath1509 therefore for all @xmath1510 ( and for @xmath1511 when @xmath1512 ) we have @xmath1513 * proof * the last equation follows from the previous one , with a little care when @xmath1511 ; indeed , for @xmath1511 we have @xmath1514 , where @xmath1515 are the number of whole- and half - loops .
so @xmath1516 and @xmath1517 hence @xmath1518 thus equation ( [ eq : sit_recur ] ) holds with @xmath1511 if @xmath1512 .
we similarly show that this equation holds regardless of @xmath245 for @xmath1510 .
so it suffices to prove the first equation of the lemma .
each irreducible closed walk about a vertex , @xmath96 , begins by traversing a path , @xmath874 , to a vertex , @xmath114 , then follows a strongly irreducible ( nonempty ) closed walk about @xmath114 , and then backtracks over @xmath874 ( and this statement is only true if we count half - loops as strongly irreducible ) ; each irreducible closed walk has a uniquely determined such @xmath874 and @xmath114 .
so we may count irreducible closed walks of length @xmath97 in @xmath14 by counting how many paths of length @xmath88 there are from @xmath114 that when combined with a strongly irreducible closed walk about @xmath114 of length @xmath1519 yield an irreducible closed walk , @xmath55 , of length @xmath97 .
the strongly irreducible closed walk s length , @xmath1519 , must be positive , or else @xmath55 is nt irreducible . if @xmath55 is of length at least @xmath30 or is a whole - loop , then there are @xmath1520 possibilities for the first edge of the path ( since two edges are ruled out by @xmath55 in either case ) , and @xmath580 possibilities for all edge choices thereafter . for a half - loop
there are @xmath580 possibilities for the first edge ( since only the single half - loop edge is ruled out ) .
so the contribution per strongly irreducible closed walk of length @xmath1519 is @xmath68 for @xmath216 , @xmath1521 for @xmath147 , except in a half - loop , where the contribution per half - loop is @xmath1522 , i.e. , an additional @xmath1523 beyond the standard contribution .
since @xmath97 is odd and @xmath1524 in the case of a half - loop , the additional amount beyond the standard contribution is @xmath1525 per half - loop .
@xmath203 we return to the proof of the theorem .
first assume that @xmath14 has no half - loops .
we will prove by induction on @xmath1526 that @xmath1527 where @xmath1501 are polynomials of degree @xmath97 .
clearly @xmath1528 and so @xmath1529 exists as desired with @xmath1530 .
of the closed walks of length @xmath30 , all irreducible closed walks are strongly irreducible , so @xmath1531
. lemmas [ lm : main_sit ] and [ lm : chebyshev ] now imply ( by induction on @xmath97 ) that polynomials @xmath1532 exist of degree @xmath97 satisfying equation ( [ eq : tilde_trace ] ) , and that the @xmath1501 , for @xmath850 , are annihilated by @xmath1533 where @xmath184 is the `` shift in @xmath97 '' operator , i.e. , @xmath1534 ( here we use the fact that the @xmath173 are annihilated by @xmath1535 , mentioned below lemma [ lm : chebyshev ] ) . since @xmath1536 are the four roots in @xmath184 of equation ( [ eq : recurrence ] ) , we have @xmath1537 where the @xmath1538 assuming that the four roots are distinct .
there are now two ways to finish the theorem . the first method to finish
the theorem is to calculate @xmath1539 and verify equation ( [ eq : tilde_q ] ) holds for @xmath1540 , i.e. , that @xmath1541 and @xmath1542 work in those cases .
then by uniqueness ( i.e. , the nonvanishing of a vandermonde determinant ) , those @xmath677 s must be the unique @xmath677 s that work for all @xmath97 . another way to finish
the theorem is to use the fact that @xmath1543 ( see the remark after lemma [ lm : irdeigens ] ) , and then argue that @xmath1544 by analytic continuation .
first , the @xmath677 s are the unique solutions to a @xmath1545 system of equations with coefficient analytic in @xmath181 ; hence the @xmath677 are , indeed , analytic in @xmath181 .
next , notice that @xmath1546 at @xmath1547 analytically continues to @xmath1548 at @xmath1547 by one loop about @xmath19 ; thus is suffices to prove that @xmath1549 near @xmath0 .
next notice that different @xmath181 s give different @xmath77 s ( indeed , @xmath1550 and @xmath1551 for the same @xmath250 implies @xmath1552 ) , so if @xmath1553 is an eigenvalue of multiplicity @xmath97 in a graph , then @xmath1554 times @xmath97 must be an integer .
but there is a sequence , @xmath1555 , with @xmath1556 an integer or half - integer ( namely a cycle of length @xmath16 where each edge has multiplicity @xmath3 has @xmath1557 as an eigenvalue of multiplicty two ) .
so by continuity @xmath1558 for sufficiently large @xmath16 , and thus @xmath1559 is identically @xmath68 .
it suffices to determine @xmath1560 , which from @xmath1561 we find are ( the constant functions ) @xmath1542 . finally ,
if @xmath14 has only half - loops ( no whole - loops ) , then the @xmath97 even formula and @xmath1494 formula are unchanged .
we easily see by induction on odd @xmath177 that the polynomial @xmath1562 works ( we use the fact that @xmath245 , the number of half - loops , is the trace of @xmath437 , i.e. , @xmath1563 ) .
@xmath203 fix an integer @xmath160 and a real @xmath5 .
there is an @xmath164 such that if @xmath14 is a @xmath0-regular graph with @xmath1564 for all @xmath32 , then the @xmath1492 of equation ( [ eq : sit_sum ] ) satisfy @xmath1565 and @xmath1566 for all @xmath32 .
* proof * the @xmath1492 must be @xmath1504 or roots of the equation in @xmath77 @xmath1567 or @xmath1568 for @xmath46 we have @xmath1 and the corresponding @xmath77 s are @xmath1569 .
the other @xmath77 s are either @xmath68 or come from @xmath1570 with @xmath32 .
but for @xmath1564 it is easy to see that the corresponding @xmath77 s are bounded away from @xmath580 .
@xmath203 we can form a selective , strongly irreducible trace by taking @xmath1571 to be the number of strongly irreducible closed walks of length @xmath97 that are @xmath1425-selective .
notice that there are no more strongly irreducible closed walks than irreducible closed walks in any lettered type , and the strong irreducibility of a potential walk can be determined from its image in the corresponding lettered type .
hence the expansion theorems of sections 69 , especially theorem [ th : with ] , carry over to @xmath1572 replacing @xmath1573 , by simply replacing @xmath1574 by the same number of walk classes of the new @xmath1401-type , @xmath579 , except requiring that the walks are strongly irreducible .
[ lm : ugly ] fix integers @xmath1575 with @xmath160 , polynomials @xmath1576 , a constant , @xmath10 , and an integer @xmath598 . assume that for each @xmath16 we have complex - valued random variables @xmath1577 such that @xmath1578 .
assume that @xmath1579 is of absolute value at most @xmath68 , and is purely real if its absolute value is greater than @xmath1052 .
furthermore assume that for all integers @xmath151 we have @xmath1580 } = \sum_{j=0}^{r-1 } p_j(k)n^{-j } \ ; + \ ; o\bigl(k^{\widetilde r}n^{-r } + k^c ( d-1)^{-k/2}\bigr).\ ] ] then for sufficiently large @xmath16 we have @xmath1581 } = o(dn^{1-(r/3)}+k^c(d-1)^{-k/2})\ ] ] for all @xmath97 with @xmath1582 for some constant @xmath1583 , where the constant @xmath1584 and the constant in the @xmath1585 notation depends only on @xmath1575 and the maximum degree of the @xmath791 .
after this section we will apply this lemma with the @xmath1586 being the eigenvalues of @xmath542 , with @xmath97 proportional ( for fixed @xmath250 ) to @xmath1587 . if @xmath184 denotes the `` shift with respect to @xmath97 , '' i.e. , @xmath1588 , then some fixed power of @xmath1589 annihilates the @xmath1590 , and also @xmath1591 this allows us to say a lot about the @xmath1592 by applying some power of @xmath1589 to equation ( [ eq : ugly_start ] ) . for the irreducible trace ,
the @xmath1593 are replaced by certain chebyshev polynomials of @xmath1592 , and applying powers of @xmath1589 to them seems more awkward ; this is why we have introduced strongly irreducible traces . in @xcite
, we worked with irreducible traces , not strongly irreducible traces .
there we had the @xmath1593 replaced by chebyshev polynomials of @xmath1579 ; however ( 1 ) we knew that @xmath1594 and @xmath1592 was bounded away from @xmath214 with probability @xmath1595 , and ( 2 ) we could only prove the asymptotic expansion up to @xmath250 which was roughly propotional to @xmath1596 .
so we could directly apply the analogue of equation ( [ eq : ugly_start ] ) with @xmath97 roughly @xmath1597 to determine that @xmath1598 and the higher @xmath1599 vanish ( up to @xmath115 roughly proportional to @xmath1596 ) . in this paper
the arbitrary length of the asymptotic expansion for a type of trace comes at the cost of having far less control over the @xmath1592 , and we have no ability to determine the @xmath1599 exactly .
fortunately lemma [ lm : ugly ] allows us to control the @xmath1592 bounded away from @xmath214 , and fortunately we will see that the @xmath1592 for @xmath32 are bounded away from @xmath214 with probability @xmath1600 .
we wish to comment that one expects polynomials @xmath1601 to arise from the binomial expansion .
namely ( by taylor s theorem ) , @xmath1602 for some @xmath1603 $ ] .
so for those @xmath1604 for some constant @xmath1605 , we may take @xmath50 roughly @xmath1606 and get an error term in @xmath1607 bounded by roughly @xmath1608 ; we get a similarly bounded error term when taking expected values of @xmath1609 where @xmath373 is the event that @xmath1604 . in this way ,
equation ( [ binom_expansion ] ) could give rise to the terms of an asymptotic expansion .
* proof * ( of lemma [ lm : ugly ] ) .
let @xmath50 be a fixed even integer such that the maximum degree of the @xmath1599 is at most @xmath1610 .
we apply @xmath1611 to equation ( [ eq : ugly_start ] ) to conclude that @xmath1612 } = o\bigl(k^{\widetilde r}n^{-r } + k^c ( d-1)^{-k/2}\bigr)\ ] ] for any @xmath97 , where the constant in the @xmath1585 notation depends on @xmath0 and @xmath50 .
we conclude that for @xmath1613 we have @xmath1614 } = o(n^{-r/2}).\ ] ] applying this for @xmath1615 and @xmath1616 , where @xmath1617 , and subtracting we conclude @xmath1618 } = o(n^{-r/2}).\ ] ] since @xmath1592 is real unless @xmath1619 , we conclude that @xmath1620 } \ ] ] @xmath1621 now for any @xmath1622 we have @xmath1623 is at least roughly @xmath1624 for large @xmath16 ; also for @xmath1625 for a constant @xmath225 we have @xmath1626 is near @xmath214 for large @xmath16 provided @xmath1627 , and also @xmath1628 . we conclude that @xmath1629 } = o(n^{-r/2}),\ ] ] and hence , since @xmath1592 is real for @xmath1630 for @xmath16 large , @xmath1631 } = o(n^{s\alpha-(r/2)}).\ ] ] let @xmath225 be fixed with @xmath1632 , and set @xmath1633 & = & \;\mbox{the event that $ \theta_i < n^{-\alpha}$},\\ a_2[i , n ] & = & \;\mbox{the event that $ n^{-\alpha}\le \theta_i\le\log^{-2}n$},\\ a_3[i , n ] & = & \;\mbox{the event that $ \log^{-2}n<|\theta_i|$}.\end{aligned}\ ] ]
equation ( [ eq : midrange ] ) implies that @xmath1634 \right\ } } = o(n^{-r/3}).\ ] ] since the number of @xmath1592 is linear in @xmath16 , we conclude that @xmath1635 } ( 1-\theta_i)^k \right ] } = o(n^{1-(r/3)}).\ ] ] by the comment just before the proof , we have ( using taylor s theorem ) @xmath1636 } ( 1-\theta_i)^k \right ] } = \sum_{j=0}^{j\le 2r/\alpha } \binom{k}{j } { \mbox{e}\left [ \chi_{a_1[i , n]}(-\theta_i)^j\right ] } \;+\;o(k^{1+(2r/\alpha)}n^{-2r}).\ ] ] summing over @xmath88 in the above involves summing over @xmath88 in the expected values and multiplying the error term by a number linear in @xmath16 .
so let @xmath1637}(-\theta_i)^j\right ] } , \ ] ] which is a polynomial of fixed degree in @xmath97 whose coefficients depend on @xmath16 ( and the @xmath1592 which are given for each value of @xmath16 ) .
fix a @xmath1584 for which @xmath1638 for all @xmath1639 , i.e. , fix a @xmath1584 with @xmath1640 then for all @xmath1639 we have @xmath1641 } ( 1-\theta_i)^k \right ] } = q(k , n ) + o(n^{-r/3}).\ ] ] now combine @xmath1642 } = \sum_{j=1}^3 { \mbox{e}\left [ \sum_i \chi_{a_j[i , n ] } ( 1-\theta_i)^k \right ] } \ ] ] with equations ( [ eq : a1 ] ) and ( [ eq : big_expected ] ) to conclude that for @xmath1639 we have @xmath1642 } = q(k , n ) + { \mbox{e}\left [ \sum_i \chi_{a_3[i , n ] } ( 1-\theta_i)^k \right ] } + o(n^{1-(r/3)}).\ ] ] on the other hand , equation ( [ eq : ugly_start ] ) just says @xmath1642 } = p(k , n ) + o\bigl(k^{\widetilde r}n^{-r } + k^c ( d-1)^{-k/2}\bigr),\ ] ] where @xmath1643 is the polynomial in @xmath97 given as the sum of the @xmath1644 . therefore @xmath1645}(1-\theta_i)^k \right ] } \ ] ] @xmath1646 for all @xmath1639 .
since @xmath1647 $ ] implies @xmath1648 and thus @xmath1649 for @xmath1650 , we have @xmath1651 } for @xmath16 sufficiently large .
we conclude that @xmath1652 for all @xmath97 with @xmath1653 .
let @xmath1654 be a polynomial in @xmath97 of degree @xmath1655 such that @xmath1656 for integers @xmath1657 for some integers @xmath1658 with @xmath1659 .
then @xmath1660 for integers @xmath88 with @xmath1661 * proof * we have @xmath1662 , and therefore @xmath1663 for any @xmath1120 and @xmath245 . given @xmath1664 ,
let @xmath1665 and @xmath1666 in the above ; @xmath1667 are integers between @xmath1190 and @xmath1191 provided that @xmath1668 so @xmath1669 or @xmath1670 .
if so , then @xmath1671 @xmath203 recall equation ( [ eq : pq_close ] ) and the fact that @xmath874 and @xmath643 are polynomials in @xmath97 ( for fixed @xmath16 ) of bounded degree .
so applying the above sublemma for @xmath1672 and @xmath1673 implies that @xmath1674 for @xmath1675 .
equation ( [ eq : pq_close ] ) now holds for all @xmath1582 .
we conlude that equation ( [ eq : for_the_end ] ) holds for all @xmath1582 . adding this to equation ( [ eq : big_expected ] ) yields the desired equation ( [ eq : ugly_finish ] ) .
in this section we use standard counting arguments to prove theorems implying `` magnification '' or `` expansion '' for `` most '' random graphs ; here `` most '' means all graphs excepting a set of probability @xmath1676 .
these theorems will then be used with lemma [ lm : ugly ] to prove theorems [ th : main ] , [ th : mainh ] , and [ th : maini ] .
a graph , @xmath14 , with @xmath16 vertices is said to be a @xmath1584-magnifier if for all subsets of vertices , @xmath106 , of size at most @xmath1677 we have @xmath1678 where @xmath1679 denotes those vertices connected to some member of @xmath106 by an edge .
alon has shown that any @xmath0 regular @xmath1584-magnifier has @xmath1680 ( see @xcite ; see @xcite for related `` edge magnification '' results ) .
say that a @xmath0-regular graph on @xmath16 vertices is a _ @xmath1584-spreader _
if for every subset , @xmath106 , of at most @xmath1677 vertices we have @xmath1681 [ th : separation ] let @xmath14 be a @xmath0-regular @xmath1584-spreader . then for all @xmath32 we have @xmath1682 * proof * since the graph is @xmath0-regular , we have @xmath1683 for all subsets of vertices ,
@xmath482 . taking @xmath1684 yields @xmath1685 hence @xmath1686 , the graph on @xmath548 whose edges are paths in @xmath14 of length @xmath30 ( and whose adjacency matrix is @xmath1687 ) , is a @xmath1688-regular @xmath1584-magnifier .
now apply alon s result on magnification and eigenvalues to @xmath1686 , whose eigenvalues are @xmath1689 .
@xmath203 we now establish that for all our models , a graph will be a @xmath1584-spreader for some fixed @xmath1690 with probability @xmath1600 .
[ th : magnify_cgnd ] for any @xmath5 and even @xmath2 there is a @xmath1583 such that @xmath404 is a @xmath1584-spreader with probability @xmath1600 .
later we shall prove this theorem for other models of random graphs , by very similar calculations .
this theorem is easy for @xmath0 sufficiently large ; but when @xmath61 ( or later possibly @xmath1184 ) one has to calculate fairly carefully . * proof * fix @xmath1691 , and consider the event that @xmath1692 .
we will impose the condition that @xmath1693 and @xmath1694 .
fix constants , @xmath1695 with @xmath1696 .
consider the situation where @xmath1697 . in this case @xmath1698 .
but since @xmath14 is @xmath0-regular and we have @xmath1699 edges leaving @xmath106 , these edges comprise all edges incident upon @xmath482 ( since @xmath1700 ) .
thus @xmath1701 is a union of connected components of @xmath14 . but this can not occur if @xmath14 has no supercritical tangles of size at most @xmath1702 ( since each connected component of @xmath14 has @xmath1703 ) . for a constant @xmath55
there are only a constant number of tangles of size at most @xmath1702 .
thus , by forsaking a probability of @xmath1676 , we may assume that @xmath1704 for any fixed constant , @xmath55 ( provided that we then take @xmath1705 ) for sufficiently large @xmath16 .
so consider a random permutation , @xmath1706 , and consider the event that @xmath132 and @xmath1707 map @xmath106 to @xmath482 .
let @xmath1708 , @xmath1709 , @xmath1710 , and let @xmath1711 .
we view @xmath132 as determined by a perfect matching of a bipartite graph on inputs , @xmath599 , and outputs , @xmath1712 , with @xmath1713 being copies of @xmath4 ( and @xmath1714 mapped to @xmath1715 ) . viewing @xmath132 as a bipartite matching , it consists of ( 1 ) @xmath250 edges from @xmath1716 to @xmath1716 ( i.e. , the @xmath599 vertices corresponding to @xmath1716 to those @xmath1712 vertices corresponding to @xmath1716 ) , ( 2 ) @xmath1717 edges from @xmath1716 to @xmath1718 , ( 3 ) @xmath1717 edges from @xmath1718 to @xmath1716 , ( 4 ) @xmath1719 edges from @xmath1317 to @xmath1718 , and ( 5 ) @xmath1719 edges from @xmath1718 to @xmath1317 .
( this is true since a @xmath1317 vertex , either input or output , must be paired with a @xmath1718 vertex , and a @xmath1716 vertex must be paired with either a @xmath1716 or @xmath1718 vertex . )
so the event that @xmath132 and @xmath1707 map @xmath106 to @xmath482 with @xmath1720 all held fixed has probability @xmath1721 \left [ \binom{c_3}{c_1-r } ( c_1-r ) !
\right]^2 \times\ ] ] @xmath1722 ^ 2 \left [ n(n-1)\cdots(n-2c_1 - 2c_2+r+1)\right]^{-1}\ ] ] ( the first expression in square brackets corresponds to choosing @xmath250 @xmath1716 to @xmath1716 edges ; the second expression corresponds to choosing @xmath1717 @xmath1716 to @xmath1718 edges , and is squared to include choosing the @xmath1718 to @xmath1716 edges ; etc . )
the probability taken over all @xmath1723 of a given @xmath1724 ( and with @xmath250 fixed ) is therefore at most @xmath1725 it suffices to show that this expression is @xmath1726 with @xmath1727 , provided that @xmath1190 is sufficiently large ( and at most @xmath1677 ) , since then we can sum equation ( [ eq : bimon_and_p ] ) over the at most @xmath1728 relevant values of @xmath1729 .
we should remind ourselves that @xmath1729 range over integers with @xmath1730 furthermore , considering the expression defining @xmath874 , we have @xmath1731 .
we now write @xmath1732 @xmath1733 and @xmath1734 we make some general remarks about analyzing the factorials in the above two equations : 1 .
all factorials in the above equations are of the form @xmath1735 for some @xmath1736 $ ] .
stirling s formula @xmath1737 implies that @xmath1738 = \mu\log(n / e ) + \mu\log\mu + o\left(\frac{\log n } { n}\right),\ ] ] where the constant in the @xmath1585 is independent of @xmath16 and @xmath1736 $ ] .
2 . in analyzing @xmath1191 and @xmath874 above , we may ignore the @xmath1739 term in equation ( [ eq : stirling ] ) .
this is because @xmath1740 are _ balanced _ in that the sum of the numbers to which factorials are applied is the same in the numerator and denominator ; in other words , the @xmath1739 terms in the numerator will exactly cancel those in the denominator .
3 . let @xmath1741 .
we claim that for @xmath1742 $ ] we have @xmath1743 indeed , since @xmath1744 for @xmath1745 , @xmath158 is concave in @xmath1746 $ ] , and so @xmath1747 is decreasing in @xmath1748 for @xmath1749 fixed ; so @xmath1750 s maximum over an interval is taken at its endpoints , and since @xmath1751 , the above claim is established . + next , a taylor expansion shows that @xmath1752 for sufficiently small @xmath5 . hence there is an @xmath1753 such that latexmath:[\[\label{eq : difference } @xmath1742 $ ] with @xmath1755 .
let @xmath1756 ( @xmath1247 ) be the non - negative reals given by @xmath1757 we have that @xmath1758 we conclude that @xmath1759 it follows from equation ( [ eq : stirling ] ) , remark ( 2 ) below it , and equation ( [ eq : difference ] ) , that we may replace @xmath1760 with @xmath1761 and @xmath249 with @xmath1762 in calculating @xmath1763 and incur an additive error term of at most @xmath1764 .
thus we get @xmath1765 where @xmath1766 similarly we calculate @xmath1767 i.e. , we have the exact same equation ( ! ) for @xmath1768 replaced by @xmath1769 ( this `` coincidence '' happens for the other models as well ) .
hence @xmath1770 since @xmath1771 , we have either ( or both ) @xmath1772 are @xmath1773 . hence @xmath1774 now we claim that for any constant @xmath1775 there is a constant @xmath1583 such that for all @xmath1776 $ ] we have @xmath1777 indeed , for @xmath1778 fixed we have @xmath1779 is increasing for @xmath1776 $ ] . hence it suffices to choose a @xmath1583 sufficiently small so that @xmath1780 so that @xmath1781 which along with equation ( [ eq : g_alpha ] ) yields equation ( [ eq : alphalog ] )
. it follows that for sufficiently small @xmath1583 we have @xmath1782 and , since @xmath1783 , this expression is @xmath1784 hence for any constant , @xmath1716 , there is a @xmath1317 such that if @xmath1785 ( i.e. , @xmath1786 ) then @xmath1787 for all @xmath16 sufficiently large . in other words
@xmath1788 , i.e. , the expression in equation ( [ eq : bimon_and_p ] ) , is at most @xmath1789 ; this , by the discussion after equation ( [ eq : bimon_and_p ] ) , completes the proof . @xmath203 theorem [ th : magnify_cgnd ] holds in the models @xmath57 , @xmath65 , and @xmath69 . * proof * in @xmath57 each permutation occurs with probability at most @xmath16 times its probability in @xmath29 .
therefore the same analysis goes through , except that @xmath874 is multiplied by at most a factor of @xmath16 .
this changes the expression for @xmath1790 by an @xmath1791 factor , so the same proof carries over . for @xmath65 we again set @xmath1792 and @xmath677 as before .
a perfect matching in @xmath4 will have ( 1 ) @xmath250 vertices of @xmath1716 paired amongst themselves , ( 2 ) @xmath1717 vertices of @xmath1716 paired with @xmath1718 vertices , and ( 3 ) @xmath1719 vertices of @xmath1317 paired with @xmath1718 vertices .
this data determines the pairing for @xmath1793 vertices .
the expression for @xmath1191 , representing the number of ways the @xmath1792 can be chosen , is the same as before .
we now derive an expression for @xmath874 , the probability that a single perfect matching matches all @xmath106 vertices to those in @xmath482 . for
an even integer , @xmath139 , let _
@xmath139 odd factorial _
be @xmath1794 which is just the number of perfect matchings of @xmath139 elements . stirling s formula yields @xmath1795 ( so that for our purposes @xmath1796 can be regarded as replacable by the square root of @xmath1797 ) .
we have @xmath1798\ ; \left [ \binom{c_3}{c_1-r } ( c_1-r ) ! \right]\;\ ] ] @xmath1799\ ; \frac { ( n-2c_1 - 2c_2-r){!_{\rm odd}}}{n{!_{\rm odd}}}.\ ] ] we get that @xmath1769 is @xmath1800 with @xmath245 as in equation ( [ eq : ubiqu_h ] ) . since @xmath1191 is unchanged , by analyzing as before we see that there is a fixed @xmath1583 such that for any constant @xmath1716 there is a @xmath1317 such that @xmath1801 provided that @xmath1785 .
next we consider @xmath69 .
let @xmath14 be a random graph in @xmath69 , so @xmath1802 .
consider the graph @xmath497 formed by adding one new vertex , @xmath1803 , to @xmath14 and replacing each half - loop about a vertex , @xmath96 , in @xmath14 by an edge from @xmath96 to @xmath114 .
then @xmath497 is precisely distributed as an element of @xmath1804 ; indeed , a perfect matching on @xmath1805 matches @xmath114 to some element of @xmath1802 and then randomly matches the remaining @xmath67 elements of @xmath548 .
now we know that @xmath497 is a @xmath1584-spreader with probability @xmath1600 .
but for any @xmath1806 , @xmath1807 consists of at most one more vertex than @xmath1808 .
hence for @xmath1809 and @xmath497 being a @xmath1584-spreader , we have @xmath1810 where @xmath1811 , provided that @xmath1812 .
hence @xmath14 is a @xmath1584-spreader on sets , @xmath106 , of size @xmath1813 . on smaller sets , @xmath106
, we have @xmath14 is a @xmath1814-spreader with probability @xmath1600 , assuming @xmath1815 , by the argument given before for @xmath29 .
( notice that the @xmath49 for @xmath65 and @xmath69 are the same . )
hence a random graph in @xmath69 is a @xmath1814-spreader with probability @xmath1600 .
here we quickly finish the proof of theorem [ th : main ] , which proves alon s conjecture for @xmath29 . fix a value of @xmath250 to be specified later .
let @xmath1816 $ ] be the set of supercritical tangles of order less than @xmath250 , and let @xmath1817 $ ] ( which we recall is the set of minimal @xmath1818 $ ] elements with respect to inclusion ) ; we know that @xmath990 is finite by lemma [ lm : finiteness ] , and we recall that if @xmath14 contains a @xmath1394 tangle then it contains an element of @xmath990 .
the probability that @xmath1819 , i.e. , that @xmath14 contains at least one element of @xmath990 , is at most @xmath1676 , since every one of the finitely many tangles in @xmath990 occurs with probability proportional to @xmath101 to the order of tangle ( theorem [ th : tangle_count ] ) , and each tangle order is at least @xmath49 .
given that @xmath1820 , we have that @xmath14 contains no supercritical tangle of order less than @xmath250 , and hence no irreducible closed walk can fail to be @xmath1425-selective for any @xmath208 .
hence for all @xmath208 and @xmath97 we have @xmath1821 thus @xmath1822 } = { \mbox{e}\left [ ( 1-\chi_{{\psi } } ) { { \rm sit}}(g;k)\right ] } .\ ] ] now according to theorem [ th : strongtrace ] we have @xmath1823 for even @xmath850 , where @xmath1824 in other words , there are @xmath1825 numbers @xmath1772 , such that @xmath1826 is the sum of the @xmath97-th powers of these numbers .
also for each @xmath88 we have @xmath1772 is not real only if it is of absolute value @xmath64 . combining this and theorem [ th : with ]
we see that @xmath1827 are random variables that satisfy the conditions of lemma [ lm : ugly ] for each @xmath88 ( and @xmath14 ) .
it follows that @xmath1828 } = o(dn^{1-(r/3)}+k^c(d-1)^{-k/2})\ ] ] for all @xmath97 with @xmath1582 for some constant @xmath1583 depending only on @xmath250 . according to theorems [ th : separation ] and [ th : magnify_cgnd ]
there is an @xmath5 such that with probability @xmath1600 we have @xmath1564 for all @xmath1829 ; in this case there is an @xmath1830 such that @xmath1831 for all @xmath1832 and @xmath1829 .
we claim that for any @xmath14 and an even integer @xmath97 we have @xmath1833 indeed , if @xmath1834 is not real , it is of absolute value @xmath64 ; if @xmath1834 is real then its @xmath97-th power is non - negative .
now let @xmath106 be the event that @xmath1835 and that @xmath1831 for all @xmath1832 and @xmath1829 .
let @xmath1836 be the event that for some @xmath115 and some @xmath1829 we have @xmath1837 for an arbitrary fixed @xmath164 .
@xmath1838 implies that for even integer @xmath97 we have @xmath1839 it follows , using equation ( [ eq : verybig ] ) , that for even @xmath97 , @xmath1840 } \ ] ] @xmath1841 } - { \mbox{e}\left [ \sum_{i\;{\rm s.t.}\;\mu_{1,2}(\lambda_i)\;{\rm not\;real}}\quad \sum_{j=1}^2 \mu_j(\lambda_i)^k\right ] } \ ] ] @xmath1842 we now take @xmath1843 we have @xmath1844 hence @xmath1845 and so @xmath1846 @xmath1847 where @xmath1848 , i.e. @xmath339 is a positive constant ( depending only on @xmath227 and @xmath0 ) .
choosing @xmath250 so that @xmath1849 , we have @xmath1850 but we have already seen ( theorems [ th : separation ] and [ th : magnify_cgnd ] ) that @xmath1851 where @xmath1852 is the complement of @xmath106 . hence @xmath1853 for any @xmath5 there is an @xmath164 such that @xmath1854 implies @xmath1855 for at least one @xmath88 , which is the event @xmath1836 above .
it follows that for any @xmath5 we have @xmath1856 this ( and theorem
[ th : improved_bound ] ) proves theorem [ th : main ] .
we now complete the proofs of theorems [ th : mainh ] and [ th : maini ] , i.e. , we establish the alon conjecture for @xmath57 , @xmath65 , and @xmath69 .
the proofs of the theorems are as the proof for @xmath29 .
we only need to establish the following results for the different models of random graph : 1 .
labelling : the model comes with edges labelled from a set @xmath398 such that to each @xmath1857 we associate a @xmath397 such that @xmath1858 ( in other words , the elements of @xmath398 are paired , with the possibility that an element is paired with itself ) .
coincidence : if @xmath97 of the random edges have been determined , and if we fix any two vertices , @xmath1859 , in the graph , then the probability that an edge of a given label takes @xmath96 to @xmath114 is at most @xmath1860 for some constant @xmath10 .
we have only briefly mentioned coincidences in this paper , but our lemmas [ lm : big_order ] and [ lm : equiv_classes ] , proven in @xcite , require a property like this .
3 . expansion with error : consider a @xmath398-labelled graph , @xmath338 , with vertices a subset of @xmath4 , that can occur as a subgraph of a graph in our model .
the probability that @xmath338 occurs must depend only on the number of edges , @xmath1861 , of each label , @xmath132 ( of course , @xmath1862 ) .
furthermore , this probability times the number subsets of @xmath4 of size @xmath1863 is , for every positive integer @xmath250 , @xmath1864 where @xmath791 are polynomials in @xmath1865 ( where @xmath1865 is the collection of all @xmath1861 ) and where @xmath1866 for all @xmath1867 , where @xmath1868 depend only on @xmath250 .
furthermore , @xmath1869 if @xmath88 is less than the order of @xmath338 .
simple word sum : let @xmath867 be those words that begin with @xmath123 , end in @xmath12 , and are _ irreducible _ ( meaning no consecutive occurrence of @xmath132 and @xmath1707 ) .
then for any polynomial , @xmath1870 ( with @xmath1865 as above ) , we require @xmath1871 for a polynomial , @xmath1872 , and a function @xmath373 with @xmath1873 for some constant @xmath10 ( i.e. , the above sum is super-@xmath0-ramanujan ) .
@xmath49 determination : we must determine @xmath49 for the model . 6 . spreading : there is a constant @xmath1583 such that the probability that a random graph has @xmath1874 for some @xmath1829 is of order at most @xmath1875 .
we have already shown spreading and determined @xmath49 for all three models .
the labelling of the models is quite simple : @xmath57 is labelled like @xmath29 ; @xmath65 is labelled with @xmath1876 where @xmath1877 ( each @xmath1878 represents a perfect matching ) ; @xmath69 is labelled with @xmath1879 with @xmath1880 as before and @xmath1881 with @xmath1882 , and where the @xmath1878 represent the near perfect matching and the @xmath1883 represents the single completing half - loop for @xmath1878 .
coincidence is easily checked for all three models .
we address the issue of simple word sum .
the word sum for @xmath57 is the same as for @xmath29 . for @xmath65 , the technique of lemma 2.11 of @xcite reduces the matter to the irreducible eigenvalues of a vertex with @xmath0 half - loops ; since these eigenvalues are the eigenvalues of a @xmath1884 matrix which is @xmath214 on the diagonal and @xmath68 s elsewhere , the eigenvalues are @xmath580 with multiplicity @xmath68 and @xmath682 with multiplicity @xmath580 .
hence the simple word sum of equation ( [ eq : simple_word ] ) is given by @xmath1885 where @xmath977 are polynomials .
for @xmath69 we can break the sum by how many half - loops are involved . for a fixed set of half - loops involved in the irreducible word , the sum is a convolution of functions of the form in equation ( [ eq : matching_walk ] ) , which by theorem [ th : baby_convolute ] is again super-@xmath0-ramanujan .
we now establish expansion with error for the three models .
equation ( [ eq : probability ] ) has the @xmath57 analogue @xmath1886 now recall the proof of theorem [ th : exp_polys ] , especially equations ( [ eq : g_rational ] ) and ( [ eq : error_term ] ) . for @xmath57 , we have @xmath1887 @xmath1888 with @xmath631 as in equation ( [ eq : g_rational ] ) with @xmath1889 being @xmath804 and @xmath1890 being the collection of the sequences @xmath1891 .
hence for a walk of length at most @xmath97 we have @xmath1892 accordingly expansion with error holds for @xmath57 with expansion polynomials determined by equation ( [ eq : chnd_form ] ) , and with error term bounded by @xmath1893 this bound is @xmath1894 for all @xmath1895 with @xmath1896 and @xmath1897 . similarly for @xmath65
we have the analogue @xmath1898 the analysis goes through essentially as before ; in the error bound we have @xmath803 is again @xmath805 , but this time the @xmath806 is as large as @xmath1899 ( taking one @xmath1900 and the rest @xmath214 ) .
so expansion with error holds for @xmath65 with error term bounded by @xmath1901 for @xmath69 , consider a random @xmath68-regular graph , @xmath497 , consisting of a near perfect matching plus one complementing half - loop on the vertex set @xmath4 .
notice that the number of such graphs is @xmath1902 .
hence the probability of occurrence of a specified half - loop and @xmath1190 other matchings in @xmath497 is @xmath1903 and the probability of @xmath1190 specified matchings ( with no specified half - loop ) is @xmath1904 so for any specification of half - loops in @xmath338 , i.e. , any fixing of each @xmath1905 to @xmath214 or @xmath68 , @xmath1906 is a polynomial in the @xmath1907 s ; this makes @xmath1906 a polynomial in the @xmath1865 , namely @xmath1908 we also see that , in the terminology above , @xmath1909 and @xmath1910 .
hence expansion with error holds for @xmath69 as well .
this establishes the six required results mentioned at the beginning of this section for the models @xmath57 , @xmath65 , and @xmath69 .
theorems [ th : mainh ] and [ th : maini ] follow .
we make a number of final remarks . [ [ stronger - conjectures ] ] stronger conjectures : + + + + + + + + + + + + + + + + + + + + + as mentioned before , numerical experiments indicate that the average ( and median ) @xmath118 for a random graph is @xmath1911 , where @xmath44 is a negative function ( tending to @xmath214 as @xmath82 ) . by the results of friedman and kahale ( extending the alon - boppana result ) ,
@xmath1912 ( see @xcite ) . however , the trace method , even with selective traces , seems to require some fundamental new idea in order to have any hope of achieving @xmath44 that is zero or negative .
[ [ critical - d ] ] critical @xmath0 : + + + + + + + + + + + + + + + + + + + + + + + + + as mentioned before , when there is a critical tangle of order strictly less than that of any hypercritical tangle , then our techniques leave a gap in that we can only prove @xmath299 with probability at least @xmath1913 where @xmath1914 .
this case is extremely interesting , since it seems that there should be a theorem that closes this gap , and such a theorem would either get around a poorly bounded @xmath1915 or improve the very interesting theorem [ th : remarkable ] ( or do something else ) . [
[ relative - alon - conjecture ] ] relative alon conjecture : + + + + + + + + + + + + + + + + + + + + + + + + + following @xcite , it seems quite possible to relativize the main theorems in this paper .
namely , fix a `` base '' graph , @xmath482 , ( or , more generally , a `` base '' pregraph , in the sense of @xcite ) .
fix an @xmath5 .
then we believe that most random coverings of @xmath482 of degree @xmath16 have all `` new '' eigenvalue @xmath1916 , where @xmath249 is the spectral radius of the universal cover of @xmath482 .
similarly , we can ask for @xmath1480 to be zero or even a negative function of @xmath16 . see @xcite for further discussion and a result in this direction . [ [ alternate - proof - with - trace - see - the - end - of - section2 ] ] alternate proof with trace ( see the end of section 2 ) : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + it may be possible to analyze the expected irreducible trace over all of @xmath29 . as remarked in theorem [ th : nottreelike ] and the discussion thereafter , the coefficients of @xmath1917 there could no longer be @xmath0-ramanujan
. it may be possible to analyze selective traces without discarding contributions from tangled graphs .
in other words , if we better understood how selectivity affected irreducible traces , we might not need section 9 ( and certain parts of our understanding of these traces might improve ) .
clearly selectivity in @xmath14 can be expressed in terms of walks in an induced subgraph of a `` higher block presentation '' of @xmath14 ( see @xcite ) .
however , it is not clear what can be said about the eigenvalues of induced subgraphs of a higher block presentation ; the author has only some weak results in this directions ( see @xcite ) .
_ this glossary contains a term or a piece of notation , followed by a colon ( :) , followed by a brief description , followed by the page number(s ) where the term / notation is explained . _ | a @xmath0-regular graph has largest or first ( adjacency matrix ) eigenvalue @xmath1 .
consider for an even @xmath2 , a random @xmath0-regular graph model formed from @xmath3 uniform , independent permutations on @xmath4 .
we shall show that for any @xmath5 we have all eigenvalues aside from @xmath1 are bounded by @xmath6 with probability @xmath7 , where @xmath8 .
we also show that this probability is at most @xmath9 , for a constant @xmath10 and a @xmath11 that is either @xmath12 or @xmath13 ( `` more often '' @xmath12 than @xmath13 ) .
we prove related theorems for other models of random graphs , including models with @xmath0 odd .
these theorems resolve the conjecture of alon , that says that for any @xmath5 and @xmath0 , the second largest eigenvalue of `` most '' random @xmath0-regular graphs are at most @xmath6 ( alon did not specify precisely what `` most '' should mean or what model of random graph one should take ) . |
secure network coding offers a method securely transmitting information from the authorized sender to the authorized receiver .
cai and yeung @xcite discussed the secrecy for the malicious adversary , eve , wiretapping a subset @xmath0 of all channels in the network . using the universal hashing lemma @xcite , the papers @xcite showed the existence of a secrecy code that universally works for any types of eavesdroppers under the size constraint of @xmath0 . also , the paper @xcite discussed construction of such a code . as another attack to information transmission via the network ,
the malicious adversary contaminates the communication by contaminating the information on a subset @xmath1 of all channels in the network . using the method of error correction , the papers @xcite proposed a method to protect the message from the contamination .
the correctness of the recovered message by the authorized receiver is called robustness .
now , for simplicity , we consider the unicast setting . when the transmission rate from the authorized sender , alice to the authorized receiver , bob is @xmath2 and the rate of noise injected by eve is @xmath3 , using the result of the papers @xcite the paper @xcite showed that there exists a sequence of asymptotically correctable code with the rate @xmath4 if the rate of information leakage to eve is less than @xmath4
. however , there is a possibility that the malicious adversary makes a combination of eavesdropping and the contamination .
that is , contaminating a part of channels , the malicious adversary might improve the ability of eavesdropping . in this paper , we discuss the secrecy when eve eavesdrops the information on the channels in @xmath0 , and adds artificial information to the information on the channels in @xmath1 sequentially based on the obtained information .
we call such an active operation a strategy . when @xmath5 , under this assumption eve is allowed to arbitrarily modify the information on the channels in @xmath1 sequentially based on the obtained information
. the aim of this paper is the following .
firstly , we show that any strategy can not improve eve s information when the any operations in the network are linear .
then , to clarify the necessity of the linearity assumption , we give a counterexample for the non - linear network , in which , there exists a strategy to improve eve s information .
this example shows the importance of the assumption of the linearity .
also , when the transmission rate from alice to bob is @xmath2 , the rate of noise injected by eve is @xmath3 , and the rate of information leakage to eve is @xmath6 , we discuss a code satisfying the secrecy and the robustness . in the asymptotic
setting , we show the existence of such a secure protocol with the rate @xmath7 . the remaining part of this paper is organized as follows .
section [ s2 ] formulates our problem and shows the impossibility of eve s eavesdropping under the linear network .
section [ s4 ] gives a counterexample in the non - linear case .
section [ s3 ] discusses the asymptotic setting , and show the achievability of the asymptotic rate @xmath7 .
we consider the unicast setting .
assume that the authorized sender , alice , and the authorized receiver , bob , are linked via a network with the set of edges @xmath8 , where the operations on all nodes are linear on the finite filed @xmath9 with prime power @xmath10 .
alice inputs the input variable @xmath11 in @xmath12 and bob receives the output variable @xmath13 in @xmath14 .
we also assume that the malicious adversary , eve , wiretaps the information @xmath15 in @xmath16 on the edges of a subset @xmath17 .
now , we fix the topology and dynamics ( operations on the intermediate nodes ) of the network .
when we assume all operations on the intermediate nodes are linear on @xmath12 , there exist matrices @xmath18 and @xmath19 such that the variables @xmath11 , @xmath13 , and @xmath15 satisfy their relations @xmath20 that is , the matrices @xmath21 and @xmath22 are decided from the network topology and dynamics .
we call this attack the _ passive attack_. to address the active attack , we consider stronger eve , i.e. , we assume that eve adds an error @xmath23 in @xmath24 on the edges of a subset @xmath25 . using matrices @xmath26 and @xmath27 ,
we rewrite the above relations as @xmath28 which is called the _ wiretap and addition model_. now , to consider the time ordering among the edges in @xmath8 , we assign the numbers to all elements of @xmath8 such that @xmath29 .
we assume that the information transmission on each edge is done with this order . in this representation ,
the elements of @xmath23 and @xmath15 are arranged in this order .
hence , the elements of the subsets @xmath0 and @xmath1 are expressed as @xmath30 and @xmath31 by using two strictly increasing functions @xmath32 and @xmath33 .
the causality yields that @xmath34 it is natural that eve can choose the information to be added in the edge @xmath35 based on the information obtained previously on the edges in the subset @xmath36 .
that is , the added error @xmath23 is given as a function @xmath37 of @xmath38 , which can be regarded as eve s strategy .
we call this attack the _ active attack _ with the strategy @xmath39 .
now , we consider the @xmath40-transmission setting , in which , alice uses the same network @xmath40 times to send the message to bob .
alice s input variable ( eve s added variable ) is given as a matrix @xmath41 ( a matrix @xmath42 ) , and bob s ( eve s ) received variable is given as a matrix @xmath43 ( a matrix @xmath44 ) .
we assume that the topology and dynamics of the network and the edge attacked by eve are not changed during @xmath40 transmissions .
their relation is given as @xmath45 to discuss the secrecy , we formulate a code .
let @xmath46 and @xmath47 be the message set and the set of values of the scramble random number .
then , an encoder is given as a function @xmath48 from @xmath49 to @xmath50 , and the decoder is given as @xmath51 from @xmath52 to @xmath46 .
our code is the pair @xmath53 , and is denoted by @xmath54 .
then , we denote the message and the scramble random number by @xmath55 and @xmath56 .
the cardinality of @xmath46 is called the size of the code and is denoted by @xmath57 . here
, we treat @xmath58 as deterministic values , and denote the pairs @xmath59 and @xmath60 by @xmath61 and @xmath62 , respectively . in the following , we fix @xmath63 . as a measure of the leaked information
, we adopt the mutual information @xmath64 between @xmath55 and eve s information @xmath65 and @xmath66 since the variable @xmath66 is given as a function of @xmath65 , we have @xmath67 .
since the leaked information is given as a function of @xmath63 in this situation , we denote it by @xmath68 $ ] .
when we always choose @xmath69 , the attack is the same as the passive attack .
this strategy is denoted by @xmath70 . when @xmath71 are treated as random variables independent of @xmath72 , the leaked information is given as the expectation of @xmath68 $ ] .
this probabilistic setting expresses the situation that eve can not necessarily choose her position to attack by herself while she knows the position , and chooses her strategy dependently of the position .
now , we have the following theorem . for any eve s
strategy @xmath39 , eve s information @xmath65 with strategy @xmath39 and that with strategy @xmath70 can be simulated by each other .
hence , we have the equation @xmath73 = i(m;y_e^n)[\phi_n,\bm{k},\bm{h},\alpha].\end{aligned}\ ] ] this theorem shows that the information leakage of the active attack with the strategy @xmath39 is the same as the information leakage of the passive attack .
hence , to guarantee the secrecy under an arbitrary active attack , it is sufficient to show the secrecy under the passive attack .
we define two random variables @xmath74 and @xmath75 .
eve can simulate her outcome @xmath76 under the strategy @xmath39 from @xmath77 , @xmath78 , and @xmath66 .
so , we have @xmath79 .
conversely , since @xmath77 is given as a function of @xmath76 , @xmath66 , and @xmath78 , we have the opposite inequality . to compare the passive attack with the active attack , we count the number of choices of both attacks . in the passive attack ,
when we fix the rank of @xmath22 ( the dimension of leaked information ) , by taking account into the equivalent class , the number of the possible choices is upper bounded by @xmath80 . in the active attack case , this calculation is more complicated . for simplicity
, we consider the case with @xmath81 . now
, we do not count the choice for the inputs on the edge @xmath82 with @xmath83 because it does not effects eve s information . then , even when we fix the matrices @xmath84 , the number of choices of @xmath39 is @xmath85 where @xmath86 .
notice that @xmath87 when @xmath5 . if we count the choice on the remaining edges , we need to multiply @xmath88 on . for general @xmath40 ,
the number of choices of @xmath39 is @xmath89 the above analysis discusses the case when eve adds the error @xmath23 .
however , eve might replace the information on the edges in @xmath1 by another elements @xmath90 , which is called the _ wiretap and replacement model_. this situation has different network structure . by using matrices @xmath91 , @xmath92 , @xmath93 and @xmath94 different from @xmath21 , @xmath22 , @xmath95 and @xmath78 ,
the observations @xmath13 and @xmath15 are given as @xmath96 however , when the set @xmath1 is the same as the set @xmath0 , it is allowed to employ the original model and because any strategy with the model and can be written as a strategy with the original model and .
theorem [ t1 ] discusses the unicast case .
it can be trivially extended to the multicast case because we do not discuss the decoder .
it also can be extended to the multiple unicast case , in which , there are several pairs of sender and receiver . when there are @xmath97 pairs in this setting , the random variables @xmath55 and @xmath56 have the forms @xmath98 and @xmath99 .
so , we can apply the discussion of theorem [ t1 ] .
to clarify the impossibility of theorem [ t1 ] under the non - linear network , we show a counter example composed of non - linear operations . we consider the network given in fig .
[ f1 ] , whose edges are @xmath100 . each edge @xmath101 is assumed to send the binary information @xmath102 .
the intermediate node makes the non - linear operation as @xmath103 to send the binary information @xmath104 , we prepare the binary uniform scramble random variable @xmath105 .
we consider the following code .
the encoder @xmath106 is given as @xmath107 the decoder @xmath108 is given as @xmath109 . since
@xmath110 and @xmath111 are given as follows under this code ; @xmath112 the decoder can recover @xmath55 nevertheless the value of @xmath56 .
now , we consider the leaked information for the passive attack .
eve is allowed to attack two edges of @xmath8 except for the pairs @xmath113 and @xmath114 .
the mutual information of these cases are calculated to @xmath115 in this section , we choose the base of the logarithm to be @xmath116 .
now , we consider the active attack with @xmath5 of the above four cases .
( i ) : : when @xmath117 , eve replaces @xmath118 by @xmath119 . then , @xmath120 because @xmath121 .
( ii ) : : when @xmath122 , eve replaces @xmath118 by @xmath70 .
then , @xmath120 because @xmath123 .
( iii ) : : when @xmath124 or @xmath125 , eve has no good active attack . to see the difference from the linear case ,
we adopt a weak secrecy criterion in this section .
when @xmath15 is eve s information and @xmath126 for all of eve s possible attack , we say that the code is secure . otherwise , it is called insecure . when eve is allowed to use the above passive attack ,
shows that the code is secure .
when eve is allowed to use the above active attack , the above calculation shows that the code is insecure .
therefore , this example is a counter example of theorem [ t1 ] .
our analysis covers the optimal codes . as another encoder
, we can consider @xmath127 replacing @xmath128 by @xmath56 , the analysis can be reduced to the presented analysis .
other encoders clearly leaks the message @xmath55 to @xmath129 or @xmath130 . in this model , eve can
perfectly contaminates the message @xmath55 .
when eve take the choice ( i ) , and replace @xmath110 by @xmath131 , bob s decoded message is @xmath132 . under the choice ( ii ) ,
eve can perfectly contaminates the message @xmath55 in a similar way .
next , under the same assumption as section [ s2 ] , we consider the asymptotic setting by taking account into robustness as well as secrecy .
we have assumed that the topology and dynamics of network and the edge attacked by eve are not changed during @xmath40 transmission .
now , we assume that eve knows these matrices and alice and bob know none of them because alice and bob often do not know the topology and dynamics of the network the edge attacked by eve .
when eve adds the error @xmath66 , there is a possibility that bob can not recover the original information @xmath55 .
this problem is called the _ robustness _ , and may be regarded as a kind of error correction . under the conventional error correction ,
the error @xmath66 is treated as a random variable subject to the uniform distribution . however , our problem is different from the conventional error correction .
since the decoding error probability depends of the strategy @xmath39 .
so , we denote it by @xmath133 $ ] . in the following ,
we assume that @xmath134 as well as . since any algebraic operation on the finite field @xmath9 can be regarded as an algebraic operation on @xmath135-dimensional algebraic extension @xmath136 of the finite field @xmath9 , where @xmath137 .
so , the matrices @xmath138 on @xmath9 can be regarded as matrices on @xmath136 . by choosing @xmath40 to be @xmath139 , the matrices @xmath140 , @xmath141 , @xmath65 , @xmath66 on @xmath9 is converted to matrices @xmath142 , @xmath143 , @xmath144 , @xmath145 on @xmath136 , which also satisfy and by regarding the same matrices @xmath138 on @xmath9 as matrices on @xmath136 .
assume that @xmath146 . when the size @xmath147 of the finite field increases such that @xmath148 , ( e.g. , @xmath149 , i.e. , @xmath150 . ) , there exists a sequence of codes @xmath151 of block - length @xmath152 on finite field @xmath136 whose message set is @xmath153 such that @xmath154=0\end{aligned}\ ] ] where the maximum is taken with respect to @xmath155 with and .
the optimality of the rate @xmath156 was also shown in @xcite .
proposition [ t2 ] requires the finite field @xmath9 with infinitely large .
the paper ( * ? ? ?
* appendix d ) discussed the construction of @xmath157 whose multiplication and inverse multiplication have calculation complexity @xmath158 and @xmath159 of @xmath157 is essentially given in ( 124 ) of @xcite by using the fourier transform via the calculation on circulant matrices .
for the inverse multiplication of an element @xmath160 of @xmath157 , we calculate @xmath161 $ ] instead of @xmath162 $ ] in ( 124 ) , where @xmath163 is the discrete fourier transform . ] . by choosing @xmath164 and @xmath165 , proposition [ t2b ]
can be rewritten as follows .
assume that @xmath146 .
there exists a sequence of codes @xmath166 of block - length @xmath167 on finite field @xmath9 whose message set is @xmath168 such that @xmath169=0,\end{aligned}\ ] ] where the maximum is taken in the same way as proposition [ t2b ] . combining theorem [ t1 ] and corollary [ t2 ] , we obtain the following theorem .
assume that @xmath146 .
there exists a sequence of codes @xmath170 of block - length @xmath167 on finite field @xmath9 whose message set is @xmath171 such that @xmath172=0 \\ & \lim_{n \to \infty } \max _ { \bm{k},\bm{h } } \max_{\alpha } i(m;y_e^n)[\phi_n,\bm{k},\bm{h},\alpha]=0,\end{aligned}\ ] ] where the maximum is taken in the same way as proposition [ t2b ] . before our proof , we prepare basic facts of information - theoretic security .
we focus on a random hash function @xmath173 from @xmath174 to @xmath175 with random variable @xmath176 deciding the function @xmath173 .
it is called _ universal2 _ when @xmath177 for any @xmath178 and @xmath179 . for @xmath180 $ ]
, we define the conditional rnyi entropy @xmath181 for the joint distribution @xmath182 as @xcite @xmath183 which is often denoted by @xmath184 in @xcite .
when @xmath11 obeys the uniform distribution , we have @xmath185 @xcite ( * ? ? ?
* theorem 1 ) @xmath186 for @xmath187 $ ] .
we choose a sequence of codes @xmath188 given in corollary [ t2 ] .
we fix @xmath189 .
now , we choose a universal2 linear random hash function @xmath173 from @xmath171 to @xmath190 .
to make our code , we consider a virtual protocol as follows .
first , alice sends her larger message @xmath55 by using the code @xmath54 , and bob recovers it .
second , alice randomly chooses @xmath176 deciding the hash function @xmath173 and sends it to bob via public channel .
finally , alice and bob apply the hash function @xmath173 to their message , and denote the result value by @xmath191 so that alice and bob share the information @xmath191 with probability almost close to @xmath119 .
since the conditional mutual information between @xmath191 and @xmath192 depends on @xmath63 , we denote it by @xmath193 $ ] .
theorem [ t1 ] shows @xmath193= i(\bar{m};y_e^{l_n}|r)[\phi_n,\bm{k},\bm{h},0]$ ] , which does not depend on @xmath194 and depends only on @xmath22 .
now , we evaluate this leaked information via a similar idea to the paper @xcite . since inequality implies that @xmath195 , proposition [ t4 ] yields that @xmath196 \le \frac{e^{s ( \bar{k}_n\log q - h_{1+s}(m|y_e^{l_n}))}}{s } \nonumber \\ \le & \frac{q^{s ( \bar{k}_n- k_n + { l_n } m_{2})}}{s } \le \frac{q^{-s \lceil \sqrt{l_n}\rceil } } { s}.\end{aligned}\ ] ] we set @xmath197 . since the number of matrices @xmath22 satisfying @xmath198 is bounded by @xmath199 , markov inequality guarantees that the inequality @xmath200|_{r = r } \le q^{- \lceil \sqrt{l_n}\rceil + m_{2}(m_{{3}}-m_{2})+1 } \label{e20}\end{aligned}\ ] ] holds for any matrix @xmath22 satisfying for any matrix @xmath201 at least with probability @xmath202
. therefore , there exists a suitable hash function @xmath203 such that @xmath204|_{r = r } \le q^{- \lceil \sqrt{l_n}\rceil + m_{2}(m_{{3}}-m_{2})+1},\end{aligned}\ ] ] which goes to zero as @xmath40 goes to infinity because @xmath205 is a constant .
now , we back to the construction of real codes .
we choose the sets @xmath46 and @xmath47 to be @xmath190 and @xmath206 , respectively .
since the linearity and the surjectivity of @xmath203 implies that @xmath207 for any element @xmath208 , we can define the invertible function @xmath209 from @xmath210 to the domain of @xmath203 , i.e. , @xmath211 such that @xmath212 for any element @xmath208 .
this condition implies @xmath213 for @xmath214 .
then , we define our encoder as @xmath215 , and the decoder as @xmath216 . the sequence of codes @xmath217 satisfies the desired requirements .
we discuss an efficient construction of our code from a code @xmath53 given in corollary [ t2 ] with @xmath218 . a modified form of the toeplitz matrices is also shown to be universal@xmath219 , which is given by a concatenation @xmath220 of the @xmath221 toeplitz matrix @xmath222 and the @xmath223 identity matrix @xmath224 @xcite , where @xmath225 is the random seed to decide the toeplitz matrix and belongs to @xmath226 .
the ( modified ) toeplitz matrices are particularly useful in practice , because there exists an efficient multiplication algorithm using the fast fourier transform algorithm with complexity @xmath227 . when the random seed @xmath225 is fixed , the encoder for our code is given as follows . by using the scramble random variable @xmath228 , the encoder @xmath229 is given as @xmath230 because @xmath231 .
( the multiplication of toeplitz matrix @xmath222 can be performed as a part of circulant matrix .
for example , the reference ( * ? ? ?
* appedix c ) gives a method to give a circulant matrix . ) .
more efficient construction for univeral2 hash function is discussed in @xcite .
so , the decoder @xmath232 is given as @xmath233 .
if we do not apply theorem [ t1 ] in , we have to multiply the number of the choices of the strategy @xmath39 . as the generalization of ,
this number is given in , which glows up double - exponentially .
hence , our proof of theorem [ t3 ] does not work without use of theorem [ t1 ] .
we have discussed the effect by sequential error injection to eve s obtained information . as the result
, we have shown that there is no improvement when the network is composed of linear operations . however ,
when the network contains non - linear operations , we have found a counterexample to improve eve s obtained information .
further , we have shown the achievability of the asymptotic rate @xmath7 for a linear network under the secrecy and robustness conditions when the transmission rate from alice to bob is @xmath2 , the rate of noise injected by eve is @xmath3 , and the rate of information leakage to eve is @xmath6 .
the converse part of this rate is an interesting open problem .
the works reported here were supported in part by the jsps grant - in - aid for scientific research ( c ) no . 16k00014 and ( b ) no .
16kt0017 , the okawa research grant and kayamori foundation of informational science advancement .
j. kurihara , r. matsumoto , and t. uyematsu , `` relative generalized rank weight of linear codes and its applications to network coding , '' _ ieee trans .
theory _ , vol .
61 , no . 7 , pp . 39123936 ( 2013 ) .
t. ho , b. leong , r. koetter , m. mdard , m. effros , and d. r. karger , `` byzantine modification detection for multicast networks using randomized network coding , '' _ proc .
2004 ieee int .
information theory ( isit 2004 ) _ , chicago , il , june / july 2004 , p. 144 .
s. jaggi , m. langberg , t. ho , and m. effros , `` correction of adversarial errors in networks , '' _ proc .
2005 ieee int .
information theory ( isit 2005 ) _ , adelaide , australia , sept . 2005 , pp .
1455 - 1459 .
s. jaggi , m. langberg , s. katti , t. ho , d. katabi , m. medard , and m. effros , `` resilient network coding in the presence of byzantine adversaries , '' _ ieee transactions on information theory _ , vol .
54 , no . 6 , pp .
25962603 , ( 2008 ) .
s. jaggi and m. langberg , `` resilient network coding in the presence of eavesdropping byzantine adversaries , '' _ proc .
2007 ieee int .
information theory ( isit 2007 ) _ , nice , france , june 2007 , p. 541
m. hayashi and t. tsurumaru , `` more efficient privacy amplification with less random seeds via dual universal hash function , '' _ ieee transactions on information theory _
62 , no . 4 , pp .
2213 2232 , ( 2016 ) . | in the network coding , we discuss the effect by sequential error injection to information leakage .
we show that there is no improvement when the network is composed of linear operations .
however , when the network contains non - linear operations , we find a counterexample to improve eve s obtained information .
further , we discuss the asymptotic rate in the linear network under the secrecy and robustness conditions . secrecy analysis , secure network coding , sequential injection , passive attack , active attack |
the physical processes associated with interactions between the intracluster medium ( icm ) and the interstellar medium ( ism ) play critical roles driving the evolution of spiral galaxies in clusters ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
many basic effects related to icm - ism interactions ( e.g. , ram pressure stripping ) are still not well understood .
these effects include the fate of star - forming molecular clouds , the rates of triggered star formation versus gas removal , and the possible reconfiguration of a galaxy s large - scale magnetic field and/or cosmic - ray ( cr ) particles .
it is also not known how far from cluster centers the stripping occurs ; for example , gas - deficient galaxies are found in cluster outskirts @xcite yet it remains unclear if they are stripped in cluster cores or locally in the outskirts as seems to be the case in ngc 4522 @xcite .
if we can find clear tracers of ongoing ram pressure and identify galaxies in different stages of the stripping process , then we can further understand the consequences of stripping . since ram pressure will more easily affect lower density constituents of the ism
, it seems that the diffuse radio continuum halos of galaxies may be sensitive tracers of active icm pressure as indicated by observations for a number of virgo cluster spirals including ngc 4522 @xcite , ngc 4402 @xcite , ngc 4254 @xcite , and a few others @xcite .
more specifically , large regions of enhanced polarized radio continuum emission have been found within a number of cluster spirals @xcite ; the maxima of the polarized radio continuum within these galaxies are located along outer edges and thought to arise from external influences of the cluster environment .
however , without a good idea of the galaxy s unperturbed appearance in the radio , these observations alone make it difficult to quantify the ram pressure effects .
we now address this by predicting the radio distribution expected in the absence of ram pressure by using the nearly universal correlation between the far - infrared ( fir ) and non - thermal radio continuum emission of normal galaxies ( e.g. * ? ? ?
* ; * ? ? ?
studies of the fir - radio correlation within field galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) have found that the dispersion in the fir / radio ratios across a galaxy disk is as small as the dispersion of the global correlation ( i.e. @xmath00.3 dex ) on scales down to @xmath1100 pc . while a number of other authors have studied how global fir / radio ratios behave among cluster galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ) , finding them to be systematically lower than those for field galaxies , it is only now by using
_ spitzer _ that we can study the influence of the cluster environment on the fir / radio behavior _ within _ galaxies allowing us to better explain the global fir / radio observations .
the fir - radio correlation is driven by the process of massive star formation .
massive stars both heat dust , yielding fir emission , and end their lives as supernovae ( sne ) whose remnants ( snrs ) , via shocks , accelerate cr electrons in a galaxy s magnetic field bringing about synchrotron radiation . using a phenomenological smearing model , in which a galaxy s fir map is smoothed by a parameterized kernel to compensate for
the fact that the mean free path of dust heating photons is much shorter than the diffusion length of cr electrons , @xcite has shown for a sample of 15 non - virgo spiral galaxies that the dispersion in the fir / radio ratios on sub - kiloparsec scales _ within _ galaxies can be reduced by a factor of @xmath12 , on average .
accordingly , it is possible to derive a good first order approximation of an undisturbed galaxy s non - thermal radio continuum morphology from its fir image . assuming that dust and gas are well mixed within galaxies , the fir emission distribution provides a reasonable picture for the molecular gas distribution as modulated by heating by starlight .
the non - thermal radio continuum emission , on the other hand , is a function of a galaxy s magnetic field and cr electron distributions ( i.e. the relativistic phase of the ism ) .
systematic departures from the nominal fir / radio ratios across galaxy disks undergoing icm - ism interactions can provide insight into differential influence of these interactions on the gaseous and relativistic phases of the ism .
these differences may help to quantify the icm pressure acting on the ism of such galaxies . with the capabilities of current observatories
the virgo cluster remains an ideal laboratory allowing for many resolution elements across a large number of galaxy disks , especially at fir wavelengths . coupling fir observations taken by the _
space telescope with vla radio continuum imaging , obtained as part of the vla imaging of virgo in atomic gas ( viva ; a. chung et al .
2008 , in preparation ) survey , we study how the relativistic and gaseous phases of the ism are affected by icm - ism interactions for a sample of virgo cluster galaxies .
this is done using fir _
spitzer _ maps to predict how the radio morphology should appear if the galaxy were not experiencing an interaction ; in this paper we make the case that significant deviations from such an appearance are directly related to icm - ism interactions .
the paper is organized as follows : in @xmath2 we briefly describe the galaxy sample and observations .
we then present our analysis techniques for comparing the fir and radio continuum imaging in @xmath33 .
results from this comparison are presented in @xmath34 . in @xmath35
we compare our results in detail with other observational evidence of icm effects for ngc 4522 , a well studied galaxy experiencing ram - pressure stripping , and put together a physical scenario for our observations .
then , in @xmath36 , we compare and discuss this physical picture for the rest of the sample . finally , we summarize our conclusions in @xmath37 . ccccccccccccc ngc 4254&12 18 49.4 & + 14 25 07 & sac & n & & -20.66 & 2407 & 5.4 & 0.87 & 0 & 3.6&-0.10 + ngc 4321&12 22 55.2 & + 15 49 23 & sabbc & t /
n & l / h & -21.05 & 1571 & 7.4 & 0.85 & 30 & 4.0 & 0.35 + ngc 4330&12 23 16.5 & + 11 22 06 & scd & t / n @xmath4 & & -18.01 & 1565 & 4.5 & 0.20 & 59 & 2.1 & 0.80 + ngc 4388&12 25 47.0 & + 12 39 42 & sab & t / n [ s]@xmath4 & sy2 & -19.34 & 2524 & 5.6 & 0.23 & 92 & 1.3 & 1.16 + ngc 4396&12 25 59.3 & + 15 40 19 & sad & n@xmath4 & & -18.04 & -128 & 3.3 & 0.29 & 125 & 3.5 & 0.30 + ngc 4402&12 26 07.9 & + 13 06 46 & sb & t / n@xmath4 & & -18.55 & 232 & 3.9 & 0.28 & 90 & 1.4 & 0.74 + ngc 4522&12 33 40.0 & + 09 10 30 & sbcd & t / n [ s ] & & -18.11 & 2307 & 3.7 & 0.27 & 33 & 3.3 & 0.86 + ngc 4569&12 36 50.1 & + 13 09 48 & sabab & t / n [ s ] &
l / sy & -20.84 & -235 & 9.6 & 0.46 & 23 & 1.7 & 1.47 + ngc 4579&12 37 44.2 & + 11 49 11 & sabb & t / n & l / sy1.9&-20.62 & 1519 & 5.9 & 0.79 & 95 & 1.8 & 0.95 + ngc 4580&12 37 48.4 & + 05 22 09 & saba pec & t / n [ s ] & & -19.27 & 1034 & 2.1 & 0.78 & 165 & 7.2 & 1.53
we present _ spitzer _ far - infrared and vla radio continuum imagery of 10 virgo cluster spirals galaxies included in the vla imaging of virgo in atomic gas ( viva ; a. chung et al .
2008 , in preparation ) survey .
a total of @xmath140 of the 53 viva sample galaxies are included in the _ spitzer _ survey of virgo ( spitsov ; j. d. p. kenney et al .
2008 , in preparation ) imaging program .
the sub - sample of 10 galaxies were the first few observed with _
spitzer _ that also had high quality radio maps as well as fir and radio emission detected out about as far as the h i extent . because we are interested in studying the properties of the radio continuum halo to determine if it is interacting with the icm , we use this criterion to ensure we are detecting infrared and radio continuum emission near the edge of the ism .
basic data for each galaxy can be found in table [ tbl - galdat ] .
galaxy diameters ( @xmath5 ) , morphological types , position angles ( pa ) , and absolute @xmath6-band magnitudes , assuming a distance of 16.6 mpc to the virgo cluster center @xcite , were taken from the third reference catalog of bright galaxies ( rc3 ; * ? ? ?
we also include distances to the virgo cluster center ( @xmath7 , @xmath8 ; * ? ? ?
* ) , in degrees , for which the sample ranges between @xmath9 ( i.e. spanning a factor of @xmath105 ) . since the hubble morphological classification appears to have limited meaning for cluster spirals ( e.g. stripped sc s tend to be classified as sa s ; * ? ? ?
* ) , we include the h@xmath11 star formation type of @xcite .
type - independent h i deficiencies , taken from a. chung et al .
( 2008 , in preparation ) , are included in table [ tbl - galdat ] and have a typical uncertainty of 0.2 dex arising from the uncertainty in morphological classifications .
the present sample spans a factor of @xmath150 in h i deficiency and nearly the entire range in h@xmath11 star formation type . to determine if the fir and radio properties of these cluster galaxies differ systematically from those in the field , we compare with 6 non - virgo galaxies ( ngc 2403 , ngc 3031 , ngc 3627 , ngc 4631 , ngc 5194 , and ngc 6946 ) , chosen to act as a control sample .
these galaxies were selected to be less distant than virgo , span a range in spiral type and inclination , and are a sub - sample of the galaxies included the fir and radio study of @xcite ; this investigation describes the physical link between the fir and radio properties of normal galaxy disks on kiloparsec scales .
information on the observations and general characteristics of these galaxies can be found there . _
spitzer _ imaging was carried out for each galaxy using the multiband imaging photometer for _ spitzer _ ( mips ; * ? ? ?
the strategy of the imaging campaign was based on that of the spitzer infrared nearby galaxies survey ( sings ; * ? ? ?
* ) with the only difference being a factor of 2 increase in exposure times to better detect diffuse emission arising in the outer regions of the virgo sample galaxies .
the mips data were reduced as in @xcite with the following modifications . since most of the galaxies were observed twice , the data had to be combined into a single mosaic of the two observations . for the 70 and 160@xmath12 data this was a straightforward process , combining the filtered basic calibrated data frames ( bcds ) , which have near zero residual background , from the two observations using the mosaicking and point source extraction ( mopex ) software package .
the target galaxy was masked during the 70@xmath12 filtering , as in @xcite , while the pipeline filtered bcds ( fbcds ) were used for 160@xmath12 observations .
the individual 24@xmath12 observations were first reduced separately in order to measure the background level .
the background level was then subtracted from each bcd , and a combined mosaic was constructed using mopex as above .
although the two observations for a given galaxy were often performed in close succession , this was not always the case . measuring and subtracting the two backgrounds was the most reliable method of placing the two observations on a common scale for mosaicking . for galaxies which were observed once at the time of this writing ,
preliminary results were obtained by combining the bcds ( as above , filtered in the case of 70@xmath12 and 160@xmath12 ) using mopex .
the resulting @xmath13 m mosaics have small regularly spaced coverage gaps at the locations of bad pixels on the array .
this is a consequence of the single redundancy of 160@xmath12 scan mode observations , although such occurrences did not affect the field of view of our galaxies in any observation .
the final calibration uncertainties of 24 , 70 , and 160 @xmath14 data are 5 , 10 , and 15% respectively . for galaxies in common with the viva survey and sings , we use mips data included in the sings data release 5 ( dr5 ) ( see table [ tbl - obsdat ] ) .
these data were processed using the mips data analysis tool ( dat ; * ? ? ?
* ) version 3.06 .
additional steps beyond the standard reduction procedure of the mips dat are described in @xcite .
the full width at half maximum ( fwhm ) of the mips 24 , 70 , and 160 @xmath14 point spread functions ( psfs ) are @xmath15@xmath15 , 17@xmath16 and 38@xmath16 while the final calibration uncertainties are @xmath12 , 5 , and 9% , respectively .
the mean rms noise values of the 24 , 70 , and 160 @xmath14 data for the sings galaxies are @xmath10.048 , 0.45 , and 0.57 mjy sr@xmath17 while our deeper imaging campaign reaches mean rms noise values of @xmath1@xmath18 better . in some cases ( e.g. ngc 4254 , ngc 4321 , ngc 4579 ) , bright nuclei created negative latent images in the 70 @xmath14 data resulting in
bright and dark streaking on opposite sides of the nuclei after background subtraction . while visible in some of the images , these streaks are generally low amplitude and should not significantly affect the analysis .
the 1.4 ghz radio continuum maps were created from the line - free channels of h i data cubes collected as part of of the viva survey ; a detailed description of the vla observations along with the h i reductions and associated data products can be found in a. chung et al .
( 2008 , in preparation ) . in order to reduce the effects of interfering sources , which causes substantial sidelobes especially at low frequencies ,
we have used the aips procedure peelr .
it iteratively attempts to calibrate on multiple fields around bright continuum sources ( self - calibration ) , subtract those fields from the self - calibrated data , undo the field - specific calibration from the residual data , and it finally restores all fields to the residual data .
the quality of our continuum data varies depending on the number of line free channels of individual galaxies .
the same weighting scheme ( robust@xmath19 ) used for the the h i maps was also used here for the continuum maps .
the rms of each 20 cm radio continuum map is given in table [ tbl - obsdat ] .
the mean rms noise value of the viva radio maps is @xmath10.008 mjy sr@xmath17 ( 0.1 mjy bm@xmath17 ) and the range in rms noise values spans a factor of @xmath13 .
the fwhm of the clean beam major and minor axes , @xmath20 and @xmath21 respectively , are also given for each galaxy in table [ tbl - obsdat ] ; the major axes range from @xmath1@xmath22 . in the case of ngc 4569 ,
a sings galaxy observed as part of the westerbork synthesis radio telescope ( wsrt)-sings survey @xcite , we use the wsrt radio map in our analysis since it is an order of magnitude more sensitive than what was achieved by the viva survey for this object .
ccccc ngc 4254 & @xmath23 & @xmath24 & 0.0066 & 0.45 + ngc 4321 & @xmath23 & @xmath25 & 0.0066 & 0.24 + ngc 4330 & 1 & @xmath26 & 0.0045 & 0.30 + ngc 4388 & 2 & @xmath27 & 0.014 & 0.37 + ngc 4396 & 2 & @xmath28 & 0.0047 & 0.18 + ngc 4402 & 1 & @xmath29 & 0.0097 & 0.54 + ngc 4522 & 2 & @xmath30 & 0.0066 & 0.26 + ngc 4569@xmath31 & @xmath23 & @xmath32 & 0.0015 & 0.22 + ngc 4579 & @xmath23 & @xmath33 & 0.0048 & 0.26 + ngc 4580 & 1 & @xmath34 & 0.0094 & 0.50 + ngc 2403 & @xmath23 & @xmath35 & 0.0020 & 0.13 + ngc 3031 & @xmath23 & @xmath36 & 0.0028 & 0.13 + ngc 3627 & @xmath23 & @xmath37 & 0.0039 & 0.45 + ngc 4631 & @xmath23 & @xmath38 & 0.0023 & 0.24 + ngc 5194 & @xmath23 & @xmath39 & 0.0015 & 0.24 + ngc 6946 & @xmath23 & @xmath40 & 0.0036 & 0.35 to perform an accurate comparative analysis between the mips and radio data we match the resolution of the final calibrated images using the mips psf .
after cropping each set of galaxy images to a common field of view we cleaned the radio data and convolved the resulting clean components with a model of the mips 70 @xmath14 psf .
we restore with the mips beam first because it has a non - gaussian shape with significant side - lobes .
since the fwhm of the original radio clean beam @xmath20 was in all cases larger than the fwhm of the mips 70 @xmath14 psf , we convolved the 70 @xmath14 and 20 cm data with an additional gaussian beam to account for this difference .
these images were then re - gridded to a common pixel scale and properly aligned .
the observed 70 @xmath14 and resolution - matched 20 cm maps for each galaxy are presented in the first and second columns of figure [ fig - maps ] . the effective resolution used in the comparison of the infrared and radio data is given in table [ tbl - obsdat ] . in table [ tbl - flux ]
we give the global mips and 20 cm flux densities .
we include extended source aperture corrections for the mips fluxes by adopting the median values derived by @xcite which are 1.01 , 1.04 , and 1.10 at 24 , 70 , and 160 @xmath14 respectively .
because our sample does not include dwarf galaxies , for which the corrections were found to be large , simply taking the median value should be adequate .
since we are using the wsrt - sings radio continuum map of ngc 4569 , we also work with the global flux density at 22 cm reported by @xcite which has been scaled to what is expected at 20 cm assuming a synchrotron power law of the form @xmath41 with a radio spectral index @xmath42 ( i.e. typical for star - forming galaxies ; * ? ? ?
this was also done for the 22 cm flux densities of the wsrt - sings field galaxies where appropriate .
we also provide measures of the total infrared flux ( tir ; @xmath43 ) along with logarithmic far - infrared ( fir ; @xmath44)-to - radio ratios in the form of the commonly used parameter @xmath45 @xcite such that , @xmath46 the tir fluxes were measured using a linear combination of the 3 mips bands given by equation 4 of @xcite .
the corresponding fir fractions were then calculated using the same spectral energy distribution ( sed ) models of @xcite as given by equation 3 of @xcite , i.e. , @xmath47 where @xmath48 = [ 0.5158 , 0.1619 , -0.3158 , -0.1418]$ ] and @xmath49 .
ccccccc ngc 4254 & 0.439 & 4.17 & 45.0 & 147 & 60.0 & 2.13 + ngc 4321 & 0.273 & 3.44 & 36.8 & 143 & 54.9 & 2.26 + ngc 4330 & 0.020 & 0.107 & 1.05 & 5.65 & 1.97 & 1.92 + ngc
4388 & 0.164 & 2.64 & 10.6 & 20.0 & 13.7 & 2.00 + ngc 4396 & 0.022 & 0.132 & 1.74 & 5.79 & 2.29 & 2.00 + ngc 4402 & 0.067 & 0.689 & 7.52 & 25.5 & 10.2 & 2.17 + ngc 4522 & 0.027 & 0.217 & 1.43 & 6.60 & 2.55 & 1.91 + ngc 4569&0.157@xmath4&1.41 & 11.2 & 41.5 & 16.8 & 2.00 + ngc 4579 & 0.167 & 0.816 & 9.34 & 42.8 & 15.4 & 1.89 + ngc 4580 & 0.005 & 0.148 & 1.83 & 7.75 & 2.84 & 2.73 + ngc 2403 & 0.33@xmath4 & 5.84 & 86.4 & 246 & 102 & 2.50 + ngc 3031 & 0.55@xmath4 & 5.09 & 85.2 & 360 & 129 & 2.31 + ngc 3627 & 0.46@xmath4 & 7.42 & 92.6 & 230 & 103 & 2.38 + ngc 4631 & 1.20@xmath4 & 8.15 & 130 & 290 & 132 & 2.09 + ngc 5194 & 1.15@xmath50&12.7 & 147 & 495 & 198 & 2.22 + ngc 6946 & 1.57@xmath4 & 20.4 & 207 & 503 & 234 & 2.21 ccc ngc
4254 & 2.3 & 0.927 + ngc 4321 & 0.7 & 0.253 + ngc 4330 & 0.7 & 0.381 + ngc 4388 & 0.2 & 0.078 + ngc 4396 & 0.9 & 0.668 + ngc 4402 & 0.6 & 0.581 + ngc 4522 & 0.8 & 0.835 + ngc 4569 & 0.8 & 0.933 + ngc
4579 & 0.0 & 0.000 + ngc 4580 & 0.8 & 0.762 + ngc 2403 & 1.0 & 0.664 + ngc 3031 & 1.8 & 0.464 + ngc 3627 & 0.5 & 0.541 + ngc 4631 & 0.6 & 0.768 + ngc 5194 & 0.5 & 0.343 + ngc 6946 & 0.6 & 0.590
a comparison of the 70 @xmath14 and 20 cm images in the first two columns of figure [ fig - maps ] shows that the radio continuum images for some of the galaxies are quite disturbed , appearing asymmetric compared to the infrared images ( e.g. ngc 4330 and ngc 4522 ) . in some cases we find regions along edges having fir emission without corresponding radio emission (
these are discussed further in @xmath36.4 ) ; regions opposite of these edges often have radio continuum tails without fir counterparts . to illustrate this we plot pixel identification maps for ngc 4402 and ngc 4522 in figure [ fig - pixid ] ; regions of cyan indicate that both the radio and 70 @xmath14 emission was detected at the 3-@xmath51 rms level while region of red and yellow indicate areas where only the radio or only the 70 @xmath14 emission has been detected at the 3-@xmath51 rms level .
the fir surface brightness depends on the ism density and heating intensity by starlight . assuming that the fir dust emission is , to first order , tracing the distribution of molecular gas within galaxies ( e.g. * ? ? ?
* ) , the observed differences in the radio continuum and fir emission distributions suggest that the relativistic ism is perturbed in cluster galaxies more easily than molecular material .
we would like now to interpret and quantify these effects .
detailed studies of the fir - radio correlation within nearby field galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ) have shown that the dispersion in the fir / radio ratios on @xmath10@xmath52 kpc scales within galaxy disks is similar to the dispersion in the global fir - radio correlation ( i.e. @xmath00.3 dex ) .
this strong correlation between the fir and radio images of spirals has been found to improve by a factor of @xmath12 using an image - smearing technique which approximates a galaxy radio image as a smoother version of its fir image due to the diffusion of cr electrons . while the star - forming disks of cluster galaxies are often truncated due to gas stripping events , the images of the remaining 70 @xmath14 disks for the entire sample do not appear to be otherwise more strongly disturbed than galaxies in the field ; we do not observe any obvious sharp edges , tails , or evidence for strong extraplanar emission at 70 @xmath14 .
therefore , we apply the image - smearing technique of @xcite ( using a single smoothing function ) to create models for the expected radio distribution of a galaxy assuming that the fir / radio ratio map is like that of a field galaxy ; the modeled radio continuum images are denoted as @xmath53 .
since the image - smearing description is applicable purely for non - thermal radio emission , we include an estimate of the thermal radio emission @xmath54 derived from the 24 @xmath14 flux density distribution ( see * ? ? ?
* ; * ? ? ?
* for details ) such that , @xmath55 where we have assumed an average h ii region temperature of @xmath56 k. we compare the observed radio continuum images with the model radio continuum images , @xmath53 , such that @xmath57 where @xmath54 is an estimate of the thermal radio emission distribution using equation [ eq - rctherm ] , @xmath58 is the 70 @xmath14 image , @xmath59 denotes a convolution with an exponential kernel @xmath60 having scale - length @xmath61 , and @xmath62 is the global @xmath63 flux density ratio after correcting the global 20 cm flux density for thermal radio emission using the galaxy s 24 @xmath14 flux density and equation [ eq - rctherm ] .
the scale - length @xmath61 is determined by minimizing the residuals between the radio and smoothed infrared maps as described by @xcite ; only areas of the galaxy disk which are detected above the 3-@xmath51 rms noise in both of the observed radio and infrared images are considered .
thus , the residuals are calculated over the non - disturbed regions of the radio disks . to quantify the improvement in the spatial correlation between the radio and smoothed 70 @xmath14 images we measure @xmath64 which is the logarithmic decrement from the squared sum of the residuals between the _ observed _ ( unsmoothed ) radio and infrared images to the minimum residuals which defines the best - fit scale - length .
the values of the best - fit scale - lengths used for creating the modeled radio maps along with the improvement in the residuals between the two maps , as defined by @xmath64 , are given in table [ tbl - scl ] .
we find that the the image - smearing improves the dispersion between the infrared and radio maps by a factor of @xmath12 , on average ; this is the same quantitative improvement quoted by @xcite using largely field galaxies .
we also find best - fit scale - lengths of 0.2 kpc for ngc 4388 and 0.0 kpc for ngc 4579 , which likely result from their bright seyfert - type nuclei .
our aim is to determine if differences between the observed and modeled radio continuum images arise from icm - ism interactions and , if so , whether quantifying and comparing such differences within our sample can help improve our current physical picture of such interactions .
we therefore create ratio maps between the observed and modeled radio maps in the following manner .
we divide the observed radio map by the modeled radio map , keeping only those pixels ( @xmath65 ) detected at @xmath665-@xmath51 rms level of either map .
we also construct a map giving the associated uncertainty at each pixel of the ratio map ( @xmath67 ) by numerically propagating the uncertainties of the input images as measured by the 1-@xmath51 rms of their maps .
any background radio galaxies or artifacts arising from streaking in the 70 @xmath14 data ( e.g. ngc 4254 , ngc 4321 , and ngc 4579 ) outside the galaxy disk are removed .
the ratio maps are displayed in column 3 of figure [ fig - maps ] .
next , we identify regions for which the ratio of observed to modeled radio flux density departs significantly from unity . to do this we compute the interquartile standard deviation of the ratio map which is the standard deviation computed after removing values included in the first and fourth quartiles ; we denote this value as @xmath68 which , for a gaussian distribution , is equal to @xmath1@xmath69 the population standard deviation .
radio deficit regions are constructed using pixels which simultaneously obey the following two conditions : @xmath70 and @xmath71 .
the first condition selects pixels which are significantly deviant with respect to the rms noise of the maps while the second condition selects pixels which are significantly deviant with respect to the intrinsic dispersion of the ratio map within the galaxy disk .
similarly , radio excess regions are constructed from all those pixels in the ratio map which simultaneously obey the conditions such that : @xmath70 and @xmath72 .
radio excesses appear largely coincident with regions associated with synchrotron tails like those observed in ngc 4330 , ngc 4402 and ngc 4522 . in the cases of ngc 4524 ,
ngc 4321 , and ngc 4579 , the radio excess regions appear to be possibly associated with the streaking in the 70 @xmath14 data . on the other hand , for most of our sample galaxies
we detect radio deficit regions which are located along a single edge of their disks .
the radio deficit regions are generally found opposite radio excess regions that are associated with synchrotron tails ; this is illustrated in figure [ fig - tails ] for ngc 4330 , ngc 4402 , and ngc 4522 .
ngc 4254 is an exception , for which an excess region is also found interior to the deficit region .
since these synchrotron tails are likely identifying the direction of the icm wind , we focus our attention toward the radio deficit regions as they are probing directly the most intense effects of the ongoing icm - ism interactions . we define the maximum ratio , or fractional deficit , used for identifying radio deficient pixels as @xmath73 and find @xmath74 among the sample galaxies .
we note that the lack of high - quality radio continuum data for the entire sample is a limiting factor for this study .
the large range in the rms noise of the radio maps ( see table [ tbl - obsdat ] ) prohibits a proper comparison between the deficit regions for each galaxy .
however , since there is such small scatter in @xmath75 , we choose to use a constant value of @xmath76 , or 50% of the expected flux , for creating the deficit region maps of all galaxies ; this allows for the most reliable inter - galaxy comparison . doing the same for excess regions we find @xmath77 and choose a constant value of @xmath78 , or 1.3 times the expected flux , for creating the excess region maps of all galaxies
the radio continuum deficit ( black ) and excess ( cyan ) regions are given in column 4 of figure [ fig - maps ] and are overlaid by the contours of the viva h i images .
to compare quantitatively the radio deficiencies among the sample galaxies we define two parameters .
first , we measure the ratio of the deficit region area ( @xmath79 ) to that of the total galaxy area given by the size of our ratio maps ( @xmath80 ) which are determined by the regions of either significant fir or radio emission .
next , we calculate the severity of the deficit as measured by the parameter @xmath81 which measures the difference between the observed and modeled radio flux densities within the radio deficit region normalized by the global radio flux density of the galaxy ( i.e. the flux densities given in table [ tbl - flux ] ) . these values along with the flux densities and areas of the deficit regions are given in table [ tbl - rcdefres ] .
the ratio maps in column 3 of figure [ fig - maps ] exhibit gradients near one edge of the galaxy disks for most of the sample ; the lowest ratios lie preferentially along these edges .
a comparison of these gradients with our beam demonstrates that they are in fact resolved .
furthermore , these gradients are opposite to what is expected if we were not smoothing the infrared images enough to estimate properly the cr - electron diffusion lengths since , as observed , the radio disks of galaxies are generally found to be characterized by a larger scale - length than their infrared disks ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
unlike the rest of the sample , we do not identify statistically significant radio deficiencies within the disks of ngc 4321 and ngc 4580 .
we do detect radio deficit regions in ngc 4396 and ngc 4579 , however these regions may arise from non - icm related effects .
the radio deficit region for ngc 4396 is detected to the north next to a number of nearby background galaxies .
it is unclear if the observed radio deficit region partly due to a confusion artifact resulting from the background galaxies , or if it is real . unlike the infrared map ,
the radio continuum and h i images appear to have a sharp edge coincident with the radio deficit region ; however , this region is not opposite the location of extended h i emission . in the case of ngc 4579
we find a radio deficit region interior to the edge of the h i disk .
the region is consistent with a streaking artifact in the 70 @xmath14 image resulting from the galaxy s bright , seyfert - type nucleus . for the remaining six galaxies ( i.e. 60% of the sample ) , regions identified to be radio deficient are always at an edge of the disk and opposite of any identified h i tails .
for example , h i tails are clearly identified by the contour overlay on the radio continuum deficit region maps of ngc 4330 and ngc 4522 in column 4 of figure [ fig - maps ] . h i tails are also observed for , and opposite to the radio deficit regions of ngc 4254 , ngc 4388 , ngc 4402 , and ngc 4569 , however they are not easily visible in figure [ fig - maps ] either because the viva data is not sensitive enough to pick up the low level emission ( i.e. ngc 45254 and ngc 4388 ) or because of smoothing the data for resolution matching ( i.e. ngc 4402 and ngc 4569 )
. the radio continuum deficit regions also appear opposite any observed radio continuum tails which is clearly illustrated for ngc 4330 , ngc 4402 , and ngc 4522 in figure [ fig - tails ] .
the combination of these observations suggests that the radio deficit regions likely arise from the same gravitational or gasdynamical effects which have displaced the galaxy s h i gas and relativistic plasma .
since we think that the h i tails are created by the icm wind , we therefore believe that the radio deficit regions identify the zone in which the icm wind is actively working on each galaxy s ism . for those sample galaxies with radio deficit regions , the fractional area of the disk defined to be radio deficient ranges from @xmath15 to 30% .
we also find that the difference between the observed and modeled flux densities within the deficit region , normalized by the global flux density ( i.e the parameter @xmath82 ) , range from @xmath11 to 15% .
not surprisingly , we find that deficit area ratio increases with increasing @xmath82 ( see table [ tbl - rcdefres ] ) . to determine whether the identification of the radio continuum deficit regions is unique to these cluster galaxies we repeat the above analysis on six non - virgo galaxies observed as part of sings : ngc 2403 , ngc 3031 , ngc 3627 , ngc 4631 , ngc 5194 , and ngc 6946 .
a detailed comparison of the fir and radio properties of these galaxies , along with the rest of the wsrt - sings sample , can be found in @xcite .
we note that @xcite do not report finding asymmetric gradients in the fir / radio ratio maps for their sample of mostly field galaxies , suggesting that the fir and radio properties of these virgo galaxies do differ systematically from those in the field . to test this further , we repeat the exercise of looking for radio deficit regions among these six galaxies .
the smoothing scale - lengths used to create the modeled radio maps are given in table [ tbl - scl ] .
however , before constructing model radio and ratio maps , the infrared and radio images of these galaxies were smoothed appropriately to simulate their appearance at the distance of virgo .
this ensures distance effects , such as differences in the physical resolution of each disk , will not introduce any biasses into the comparison .
the final beam sizes , along with the corresponding noise in the radio and infrared maps , are given in table [ tbl - obsdat ] .
the ratio maps of each galaxy are given in figure [ fig - singsmaps ] using the same color scale as for the ratio maps of the virgo galaxies in the third column of figure [ fig - maps ] .
unlike some of the ratio maps for the virgo sample ( e.g. ngc 4402 and ngc 4522 ) we do not find strong gradients near the edges of the disks for the non - virgo galaxies .
we also note that it was not necessary to show maps of radio continuum deficit regions for the wsrt - sings galaxies since , except for two cases , radio deficits were not detected .
the first exception is ngc 3031 where we find a few pixels ( i.e. @xmath00.8% of the total area ) to be radio deficient just surrounding its nucleus ; this likely arises from its bright agn .
the other exception is ngc 3627 , in which the north - west region of the galaxy is significantly radio deficient ( i.e. @xmath83 and @xmath84 as noted in table [ tbl - rcdefres ] ) ; among the virgo galaxies with radio deficit regions detected , only ngc 4524 has a lower @xmath82 value .
the radio deficit region is located opposite of its observed h i tail @xcite and just north of an area observed to have highly polarized radio emission @xcite .
the consistency in the location of the deficit region with other tracers for ongoing icm - ism interactions , combined with the fact that ngc 3627 is a member of the leo triplet , suggests that the radio deficit region may arise from an interaction within the group .
the control sample behavior clearly demonstrates that the radio deficiency morphology observed in virgo galaxies is singular and most naturally explained as a result of icm - ism interactions .
the virgo spiral ngc 4522 is one of the clearest cases where the effects of ram pressure , resulting from the galaxy s rapid motion through the icm , are directly observed ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
a combination of observational data including asymmetric and extraplanar emission in the h@xmath11 @xcite and h i @xcite lines are highly suggestive of such icm - ism interactions .
while the icm wind has stripped away atomic gas , apparent by the extraplanar h i gas , the extraplanar h@xmath11 emission is thought to arise from extraplanar h ii regions ; so it appears that virtually the entire ism has been stripped away from the outer disk , including molecular clouds .
the stripping appears to have occurred very recently .
both the stellar population study of the stellar disk beyond the gas truncation radius @xcite and a comparison of the observed gas distribution and kinematics with simulations @xcite indicate that the galaxy was stripped within the last @xmath85 myr .
thus , it is likely that the current ram pressure is still significant .
the h i contour overlay on the radio deficit region of ngc 4522 in figure [ fig - maps ] shows that the deficit region runs along the eastern edge of the galaxy , opposite the extraplanar h i gas .
furthermore , the radio continuum emission itself appears asymmetric ; the radio disk has a sharp cutoff along the eastern edge and extraplanar emission past the western edge similar to that of the h i gas .
this is consistent with the picture in which the icm is acting on the eastern ( leading ) edge of the galaxy disk , where we find the radio continuum to be deficient , and is pushing the ism ( gaseous and relativistic phases ) toward and beyond the western edge .
radio polarization and spectral index data of ngc 4522 provide evidence for an ongoing icm - ism interaction @xcite ; the eastern edge of ngc 4522 is found to be highly polarized and has the flattest spectral index , which steepens as one moves towards the western side of the galaxy . in the first column of figure [ fig - n4522 ]
we plot the radio continuum deficit regions and overlay the 20 to 6 cm spectral index and 6 cm polarized radio continuum maps of @xcite .
we find that the regions of high polarization and flattest spectral index ( i.e. @xmath86 ) lie just interior to the radio deficit region . while the peak in the polarized radio continuum is coincident with the flattest part of the spectral index distribution , this is not where the total radio intensity ( or fir ) peaks as pointed out by @xcite .
this is also evident in figure [ fig - n4522b ] where we show the 6 cm radio continuum emission together with the vectors of the magnetic field uncorrected for faraday rotation @xcite , on the radio deficit map .
the magnetic field vectors appear to be elongated parallel to the radio deficit region and the icm - ism interaction surface .
these observations suggest that , except for synchrotron tails or the radio deficit regions , to first order , star formation in the disk still drives the appearance of the fir / radio ratios . however , by inspecting the observed - to - modeled
radio continuum ratio map of ngc 4522 displayed in the second column of figure [ fig - n4522 ] , which was created using our _ smoothed _ 70 @xmath14 image ,
icm - ism effects become more apparent .
we find a localized peak in the ratios just outside the radio deficit region , and near the regions of high radio polarization and flat spectral indices ; we attribute this peak to a local enhancement in the radio brightness .
we now try to determine the most plausible physical scenario to explain this combination of observations .
ram pressure from the icm can in principle work to both compress the ism , possibly triggering star formation ( e.g. * ? ? ?
* ; * ? ? ?
* ) , as well as truncate a galaxy s star - forming disk by sweeping away interstellar material ( e.g. * ? ? ?
similarly , ram pressure may act on a galaxy s relativistic ism by compressing or stretching its magnetic field , redistributing cr particles , as well as by re - accelerating cr particles via icm triggered shocks .
while the high level of radio polarization in ngc 4522 could arise from a compressed magnetic field , such an occurrence is not clearly supported by our observations .
since the synchrotron emissivity goes roughly as the square of the magnetic field strength , we would expect a distinct region of enhanced radio continuum emission , where the field is being compressed and amplified , to lie just interior to the radio deficit region .
if there is such an enhancement in the total radio continuum just behind the radio deficit region we are probably not resolving it , but its amplitude must be modest since a strong signature in the ratio map is not detected .
on the other hand , we also observe a large region of relatively flat spectral indices ( i.e. @xmath86 ) interior to the deficit region .
while @xcite report an average radio spectral index of @xmath87 for a sample of 74 star - forming galaxies , the radio spectral indices of snrs , whose associated shocks are thought to be the primary acceleration sites for crs , are found to be slightly flatter , typically ranging between @xmath1@xmath88 both in the galaxy ( e.g. * ? ? ?
* ; * ? ? ?
* ) and m 33 @xcite .
these values are also consistent with those predicted by diffusive shock acceleration theory ( e.g. * ? ? ?
thus , it seems plausible that the flat indices interior to the deficit region of ngc 4522 is the signature of cr electron acceleration sites within ngc 4522 .
we would therefore expect such areas to exhibit moderate enhancements in total continuum emission .
these observations are consistent with the interpretation that the icm wind is sweeping up cr electrons in a low pressure outer - region of the disk as well as driving a series of shocks into the ism which will re - accelerate cr electrons there .
the latter secondary shocks , or shocklets , will likely originate at the interface between the icm and ism , arising from inhomogeneities and turbulence .
this would then explain the large region of flat spectral radio indices and area of modest increase to the radio continuum , as observed
. some of the cr electrons , including those that are newly accelerated , will be swept behind the galaxy creating a synchrotron tail that limit the enhancement to the radio surface brightness even though acceleration is occurring . a second explanation for the existence of the high radio polarization could be that of shear as discussed by @xcite which does not require total magnetic field amplification . in this scenario
the icm wind sweeps up crs and the magnetic field ; the ism magnetic field lines are stretched thereby increasing the fraction of polarized radio continuum emission without significantly increasing the total field strength .
this process alone , however , does not explain the observed flat spectral indices .
we therefore prefer the scenario in which the magnetic field of ngc 4522 is being sheared along the interacting region between its relativistic disk and the icm as ram pressure induced shocks propagate into its ism .
the locally depressed radio continuum emission in the deficit region is the result of ram pressure from the icm sweeping up low density cr electrons , associated with past star formation activity which have diffused significantly far away ( @xmath1kpc ) from their acceleration sites in supernovae .
while we do not find significant enhancements to the total radio continuum behind the deficit regions , the magnetic field may still be moderately compressed .
some of the cr electrons which have been swept - up by the icm are redistributed downstream producing the synchrotron emission tails as observed very prominently here , as well as for ngc 4402 @xcite ; this effect will limit the enhancement to the radio surface brightness .
another subset of particles , however , are re - accelerated interior to the radio continuum deficit region raising the total radio power as evidenced by the flat spectral indices and modest increase in the radio continuum found in the ratio map of ngc 4522 .
the preference for this scenario is discussed further in the context of the observations from the entire sample in @xmath36.2 .
we have shown that a large fraction of our sample of cluster galaxies exhibits statistically significant radio deficit regions relative to the prediction of the phenomenological image - smearing model of @xcite .
such deficit regions are not found within normal field galaxies , suggesting that cluster environment processes are at work . comparing our results with other observational data for ngc 4522 , a galaxy which is clearly experiencing the effects of ram pressure
, we find our deficit region agrees with such a scenario and adds new insight into the interaction .
similar to the example of ngc 4522 , radio polarization observations of other virgo cluster galaxies exhibit highly polarized radio emission along the side of the galaxy presumed to be the leading edge of the icm - ism interaction .
for the galaxies in our sample having polarization data and which show evidence for being disturbed ( i.e. ngc 4254 , ngc 4388 and ngc 4402 ) , the regions of highly polarized continuum emission appear coincident with the modest local radio continuum enhancements found in our ratio maps ( i.e. ratios of @xmath1@xmath89 ) and interior to our radio deficit regions . when averaged over these kpc - sized regions , such enhancements in the total radio continuum translate , roughly , into a compression and amplification of the magnetic field by @xmath1@xmath90% , clearly a modest effect .
thus , the scenario described for ngc 4522 seems to be applicable for these galaxies as well based on the polarized radio continuum results . with this picture
we will use the findings from our analysis to quantitatively assess the strength of the ram pressure from the intracluster medium ( icm ) .
since the radio deficit regions detected for ngc 4396 and ngc 4579 differ empirically from what we find for the rest of the sample , and show evidence for being artificial , we exclude them from the following discussion ( see @xmath3[sec - ratmaps ] ) .
cccccccccc n4254 & 1.578 & 4.728 & 0.007 & 33.0 & 0.04 & 3.9 & 8.3 & 0.46 & + n4330 & 1.891 & 4.939 & 0.153 & 20.7 & 0.27 & 2.0 & 2.1 & 0.97 & + n4388 & 1.733 & 7.607 & 0.036 & 22.6 & 0.21 & 2.1 & 5.6 & 0.38 & 225 + n4402 & 1.120 & 4.134 & 0.045 & 18.8 & 0.13 & 2.3 & 4.3 & 0.53 & 200 + n4522 & 1.676 & 4.228 & 0.096 & 18.3 & 0.17 & 2.5 & 3.2 & 0.80 & 100 + n4569 & 1.852 & 4.702 & 0.023 & 35.6 & 0.12 & 2.6 & 7.1 & 0.36 & 300 + ngc 3627 & 2.207 & 7.318 & 0.011 & 43.3 & 0.08 & 2.9 & 6.8 & 0.43 & ram pressure is expected to vary by several orders of magnitude as galaxies orbit within clusters @xcite , but there is no known way to measure the strength of current ram pressure on a galaxy .
the radio deficit offers the potential to do this .
figure [ fig - tup ] shows that the strength of the radio deficit is closely related to the time since peak pressure as inferred from stellar population studies and gas stripping simulations @xcite .
this strongly suggests that the strength of the radio deficit is good indicator of the strength of the current ram pressure . in simple models ( e.g. *
* ) the ram pressure drops by roughly an order of magnitude 300 myr after peak pressure , and by a factor of @xmath15 from 100 myr to 300 myr after peak pressure .
as shown in figure [ fig - tup ] , this is similar to the change in the radio deficit parameter between ngc 4522 and ngc 4569 , suggesting that differences from galaxy to galaxy in the radio deficit parameter are reflecting differences in the strength of current ram pressure .
we can get a rough estimate of the strength of ram pressure acting on these galaxies from the equipartion energy density of the radio - emitting plasma .
ram pressure from the icm is simply defined as , @xmath91 where @xmath92 is the icm mass density and @xmath93 is velocity of a galaxy relative to the icm .
if the icm ram pressure exceeds the internal pressure of a galaxy s relativistic ism ( crs @xmath94 magnetic field ) , then it should be possible to redistribute and even strip them from the galaxy disk . using the predicted radio flux density for each deficit region
, we can approximate a minimum value for the icm ram pressure needed to cause the observed depression in the radio .
taking the predicted flux density of the deficit region along with its area we use the revised equipartition and minimum energy formulas of @xcite to calculate the minimum energy magnetic field strength of the deficit regions before they were stripped .
this calculation assumes a proton - to - electron number density ratio , for particles in the energy range corresponding to ghz synchrotron emission , of 100 , a radio spectral index of @xmath95 , and a path length through the emitting medium of @xmath96 kpc where @xmath97 is the galaxy inclination .
the inclination is derived using the method of @xcite such that , @xmath98 where @xmath99 and @xmath100 are the observed semi - major and semi - minor axes and the disks are oblate spheroids with an intrinsic axial ratio @xmath101 for morphological types earlier than sbc and @xmath102 otherwise . using these calculated magnetic field strengths @xmath6 ( i.e. calculated using equation 4 of @xcite ) we compute the magnetic field energy densities @xmath103 of the deficit regions . assuming minimum energy between the magnetic field and cr particle energy densities , @xmath104 and @xmath105 respectively
, we can use the values of @xmath104 over the deficit region in our model to determine the minimum @xmath106 necessary to create the deficit regions . from minimum energy arguments we find that @xmath107 . for a relativistic gas the magnetic pressure , @xmath108 , and cr pressure @xmath109
are related to energy density such that @xmath110 and @xmath111 leading to the relation that @xmath112 .
then , the pressure of the relativistic ism is found to be @xmath113 the relativistic ism pressure of the deficit regions before they were stripped , as estimated from our models and denoted as @xmath114 , are given in table [ tbl - rcdefres ] ; these estimates set the minimum icm ram pressure necessary to create the observed radio continuum deficit regions ( i.e. @xmath115 ) .
all of the virgo galaxies listed in table [ tbl - rcdefres ] have @xmath114 values ranging between @xmath1@xmath116 dyn @xmath117 . using the projected linear distances to the cluster center for these sample galaxies along with a measured density profile of virgo @xcite
yields a range in icm density of @xmath118 @xmath119 .
taking a virgo galaxy velocity of 1000 km s@xmath17 , which is comparable to the @xmath1930 km s@xmath17 velocity dispersion among virgo spirals @xcite and typically achieved by cluster members during some portion of their orbits @xcite , the associated range in icm ram pressure is @xmath120 dyn @xmath117 .
this range of icm pressures is generally greater than or similar to the values for @xmath114 which is consistent with icm ram pressure being able to create the observed deficit regions .
these pressures agree with those produced in the 3d hydrodynamical simulation of @xcite over the same projected linear distances to the cluster center among our sample .
we also note that , given the icm density range , the corresponding velocities required to reach a minimum ram pressure of @xmath1@xmath121 dyn @xmath117 range between 530 and 980 km s@xmath17 , quite comparable to the velocity dispersion of virgo spirals . while this value only sets the minimum ram pressure strength required to create the deficit region , it is perhaps more interesting to try to get a better handle on the equilibrium pressure at the working surface between the deficit region and the edge of the galaxy s observed radio disk . since it is difficult , and probably not very accurate , to calculate the relativistic ism pressure on a point - by - point basis within the disk , we instead measure the average internal relativistic plasma pressure of observed radio disk @xmath122 .
we define @xmath122 using the observed total radio flux density and the area given within the 3-@xmath51 rms isophote of each radio image .
the values of @xmath122 , along with the ratios between @xmath114 and @xmath122 , are given in table [ tbl - rcdefres ] .
while ngc 4321 and ngc 4580 were excluded in table [ tbl - rcdefres ] since they do not have radio deficit regions , we note that their values of @xmath122 are 7.1 and @xmath123 dyn @xmath117 , respectively .
therefore , the full range of relativistic plasma pressures are similar to ram pressures calculated above for a typical virgo galaxy and a range of icm densities
. the internal relativistic plasma pressures can not be vastly different from that of the icm ram pressure since then we would either not observe the radio deficit regions or the entire relativistic ism would be swept out of the disk .
we expect to see more significant radio deficit regions ( i.e. larger values of @xmath82 ) for cases where the internal relativistic ism pressure is low . comparing the values of @xmath82 with @xmath122 in table [ tbl - rcdefres ]
we find that this is indeed the case .
our ram pressure estimates from minimum energy arguments give a good rough approximation for the strength of ram pressure within a factor of a few , but can not be expected to measure precisely the differences in ram pressure among the sample galaxies .
further work is required to more closely and quantitatively relate the radio deficit to the strength of ram pressure . in figure [ fig - tirup ]
we find that there is a general trend of decreasing values of @xmath82 with increasing tir surface brightnesses as measured by the ratio of each galaxy s global tir flux to @xmath80 , the deprojected area of their ratio map .
since tir surface brightness and @xmath122 are directly related , this trend is not terribly surprising .
interpreting the tir surface brightness to be a proxy for the mean ism density , we find that galaxies having a more dense ism appear to be less affected by the icm .
this seems reasonable given that the ram pressure would need to be stronger to disturb their ism and is consistent with higher values for the internal relativistic plasma pressure for small values of @xmath82 .
however , this trend may be fortuitous given that the external ram pressure likely varies tremendously among the sample .
using our results on the spatially resolved fir / radio distributions of these cluster spirals , along with their global fir / radio properties , we will now investigate whether the icm wind plays a significant role in enhancing the escape and re - acceleration of cr electrons .
after considering escape , we explore a number of physical scenarios to explain simultaneously the local and global fir / radio ratio behavior among these galaxies before coming to our preferred picture .
the only galaxy with a high global @xmath45 value is ngc 4580 .
while the cr electrons are being moved around by the icm wind , it is unclear as to whether they easily escape the galaxy disks as global @xmath45 ( logarithmic fir / radio ) ratios do not appear systematically high with respect to the nominal value of @xmath1@xmath124 dex ( i.e. * ? ? ? * ) .
in fact , by plotting @xmath45 against @xmath82 ( our parameter defining the severity of the deficit region ) in figure [ fig - qup ] we find a trend of decreasing fir / radio ratios with increasing @xmath82 and that nearly all of the @xmath45 values are lower than the nominal value .
other , more detailed studies on the global fir and radio properties of cluster galaxies have pointed out notably lower fir / radio ratios than expected from the fir - radio relation in the field @xcite .
this suggests that there is either depressed fir emission or an enhancement in the global radio emission among our sample galaxies . and while the trend in figure [ fig - qup ] may simply be the result of small number statistics
, it also might suggest that @xmath45 is sensitive to the strength of the icm ram pressure and perhaps a galaxy s stripping history .
we do not detect a radio deficit region for ngc 4580 .
this galaxy is highly h i deficient , having the largest deficiency value in table [ tbl - galdat ] ( 1.53 dex ; a factor of @xmath15 larger than the median value among the sample ) , and has been classified as an anemic spiral @xcite , though better described as having a severely truncated star - forming disk @xcite .
it appears to have been stripped long ago ( @xmath1475 myr ) and is the only post - strongly stripped galaxy in our sample observed so long after peak pressure @xcite .
we therefore speculate that its high @xmath45 value is the result of cr electron escape as the icm wind has swept out most of the galaxy s gaseous and relativistic ism long ago leaving only the central disk .
the only other galaxy for which we do not detect a radio deficit region is ngc 4321 which is much different from ngc 4580 .
ngc 4321 has not been strongly stripped and has a near - average @xmath45 value ; this galaxy appears much more similar to the wsrt - sings field galaxies . as already discussed in @xmath3[sec - cr_b ] , we have attributed local enhancements to the radio continuum arising from the re - acceleration of cr electrons by icm induced shocklets .
while moderate local enhancements are found in the radio continuum interior to the radio deficit regions via the ratio maps , they do not appear significant enough to explain @xmath45 ratios which are @xmath00.4 dex lower than average ( i.e. a factor of @xmath02.5 ) .
the @xmath45 values of taffy systems are found to be a factor of @xmath12 low due to excess radio continuum emission associated with the synchrotron bridges connecting the galaxy pairs @xcite . looking to see
if synchrotron tails may be causing a similar effect in ngc 4522 ( where the observed synchrotron tail appears most striking ) we find that this can not be the case .
only @xmath125% of the total radio continuum emission is associated with the tail . while we have argued that it does not seem likely that the magnetic field has been compressed significantly ( i.e. no more than @xmath030% as suggested by the radio excesses ) at the leading edge between the ism and the icm wind , leaving no clear signatures of a gradient in the ratio map , perhaps the field strength is raised globally due to thermal compression by the surrounding icm as suggested by @xcite .
models of such an effect seem to be able to produce the factor of @xmath126 excess synchrotron emission and reproduce the observed @xmath45 ratios @xcite .
if such a scenario were true , and the virgo galaxies with low @xmath45 ratios are simply compressed versions of field galaxies , we would expect galaxies with the lowest @xmath45 ratios to appear compressed in all directions and exhibit relatively larger internal relativistic plasma pressures , @xmath122 , compared to virgo galaxies with larger @xmath45 ratios .
however , the galaxies having the lowest @xmath45 ratios both exhibit extraplanar radio continuum emission along their trailing edges in the form of diffuse synchrotron tails and do not have larger @xmath122 values relative to the rest of sample , inconsistent if compression was the dominant mechanism .
another possibility is that the fir emission is depressed in these systems through the removal of a dust component .
for instance , the icm wind may have removed a substantial amount of the cool dust ( cirrus ) component , a significant contributor to the total fir emission . however , fir color differences between the stripped virgo galaxies and the wsrt - sings spirals are too small to induce differences in the seds and total fir emission large enough to account for the differences in @xmath45 . the mips @xmath127 and @xmath128 flux density ratios have medians that are @xmath129 20% lower and higher , respectively , than those found for the wsrt - sings spirals .
both samples have fir colors in the range of quiescent normal star - forming galaxies , where the 20% color differences translate into small changes in the sed , and changes of only a few percent in fir normalized either to the 24 or 70 @xmath14 emission , based on the @xcite sed templates .
the 24 @xmath14 emission component is relevant for normalization because of its close association with hii regions and dense , star - forming regions @xcite , making it much more resistant to stripping than the longer wavelength components .
this result is consistent with the study of @xcite who also found that the fir properties ( i.e. luminosities and colors ) of cluster galaxies having excess radio emission are similar to those of field galaxies .
the fir emission in these systems may also be depressed , without systematic changes in their fir colors , as a result of depressed star formation activity per unit area .
if ram pressure from the icm wind is driving shocklets into the ism , mechanical energy would be dissipated into molecular clouds as turbulence , possibly reducing their efficiency to collapse and form new stars . however , in order for a galaxy to exhibit @xmath45 values similar to those reported here , the decrease in star formation activity must have begun within the last @xmath150 myr ( i.e. roughly the expected cr electron cooling time to synchrotron losses ) . at later epochs ,
the synchrotron emission associated with the previous star formation episodes fades away , and @xmath45 values drift back towards more normal values even if the star formation activity remains depressed . since 50
myr is more than an order of magnitude smaller than the cluster crossing time , it is unlikely that we would be observing so many galaxies during this special period of time .
it therefore appears that depressed fir emission is not the dominant effect driving the systematically low @xmath45 values for these galaxies .
while it is still not completely unambigous , we believe the most plausible explanation is that the icm wind raises the global cr electron energy and synchrotron power by driving shocklets into the ism which are re - accelerating cr electrons . taking again @xmath130 km s@xmath17 and the range in @xmath92 values discussed in @xmath3 [ sec - pram ] , and assuming the average surface area of each galaxy disk exposed to the icm wind is 25 kpc@xmath131 ( i.e. the mean size of the radio deficit regions given in table [ tbl - rcdefres ] ) , we compute the kinetic energy per unit time a galaxy will inherit from the icm wind . comparing this number to that of the mean radio continuum luminosity among those galaxies listed in table [ tbl - rcdefres ] , and assuming a number density ratio of protons - to - electrons , for particles in the energy range corresponding to ghz synchrotron emission , of 100 , we find that @xmath1@xmath132% of the total kinetic energy available must be given to the crs in order to double the radio luminosity .
while a value @xmath111% may be slightly large , there are many uncertainties involved with this calculation including whether the cr electrons are more efficient than cr nuclei at absorbing this energy . in any case , it is certainly clear that the available mechanical luminosity of the system is enough to cause the observed global excess in synchrotron emission . the fact that we only find relatively moderate local enhancements to the total continuum behind the radio deficit region , save perhaps ngc 4254 where a less modest radio excess is observed ( i.e. ratios peaking at @xmath11.7 ) , suggests that the shocks run through the entire galaxy disk rather quickly .
the shock speed through the galaxy , @xmath133 , which must be super - alfvnic to re - accelerate cr electrons , should have a value around @xmath134 where @xmath106 is the ram pressure , as defined in equation [ eq - pram ] , and @xmath135 and @xmath92 are the ism and icm densities , respectively . again taking @xmath130 km s@xmath17 , the range in @xmath92 values discussed in @xmath3 [ sec - pram ] , and @xmath136 @xmath119 leads to shock velocities ranging between @xmath115 to 25 km / s . assuming a thin disk thickness of 500 pc
, the shocks should run through each disk within @xmath120 to 40 myr ; indeed this is very short compared to the dynamical timescale of these systems .
as previously noted , this proposed mechanism may work to slow down star formation itself , decreasing the infrared surface brightness , either from an increase in cold molecular cloud heating by the re - accelerated cr particles or similar gas - phase shocklets perturbing the clouds .
however , it is unclear how significant this effect may be relative to the increase in synchrotron emissivity via the associated re - acceleration of cr electrons .
since many cluster galaxies have highly radial orbits , the direction to the cluster center can help provide some insight as to whether a galaxy may be traveling into or away from the cluster center .
the direction toward the cluster center is indicated on the radio deficit maps in figure [ fig - maps ] .
some galaxies have deficit regions facing toward the cluster center ( ngc 4254 , and ngc 4402 ) which is suggestive of a galaxy entering the cluster and at a phase of pre - peak pressure . on the other hand
, we also find an example for which the radio deficit region faces away from the cluster center ( ngc 4569 ) which may suggest that the galaxy is on their way out of the cluster and past peak - pressure .
there are also galaxies which face neither directly towards or away from the cluster center ( ngc 4330 and ngc 4522 ) leading to a more complex scenario ; such objects may be stripped by interactions with sub - cluster structures ( e.g. ngc 4522 with m 49 ) .
we also note that ngc 4388 may be interacting with the m 86 sub - cluster ; its h i tail points neither towards or away from the cluster center but rather towards m 86 @xcite .
another interesting case is that of ngc 4254 which is thought to be entering virgo for the first time at high speed .
it has recently been suggested by @xcite and @xcite that a large h i tail associated with ngc 4254 is a tidal arm resulting from galaxy harassment @xcite .
however , no interacting counterpart has been identified . also , the h i tail is located opposite the observed radio deficit region and the direction to the cluster center .
since it appears that the radio deficit regions are related to the effects of ram pressure , it therefore seems plausible that the large h i tail could be the result of ram pressure stripping .
however , it is also possible that the radio deficit region may be created via galaxy harassment or that , more plausibly , both physical processes are currently at work .
projection effects , specifically the orientation of the galaxy disk with respect to the icm wind direction and our line - of - sight , likely play a significant role in determining the location of deficit region centroids . for largely face - on systems ,
the relationship between the deficit region location and the direction of the wind is probably more easily predicted than for more highly inclined galaxies .
the highly inclined galaxies ngc 4330 , ngc 4402 , and ngc 4522 , have radio deficit regions consistent with lunes along one side of their disks ; these could occur over a large range of wind angles .
furthermore , the size of the observed deficit region is subject to projection effects as well , likely appearing much smaller for disks viewed face - on , especially if they are moving through the icm along the line - of - sight .
therefore it is hard in general to gauge whether galaxies may be pre- or past - peak pressure from the direction of the cluster center with respect to the deficit region . in figure
[ fig - dup ] we plot @xmath82 against the distance to the cluster center .
the lack of observed trend suggests that the 3d structure of the cluster along with presence of sub - cluster structures are important in understanding icm - ism interactions and the picture of a smoothly distributed and static icm is overly simplistic for characterizing icm - ism interactions . while we can explain regions having radio continuum emission without corresponding fir emission as the result of cr electrons being swept into a synchrotron tail by the icm , the observation of fir emission without corresponding radio emission is a bit more interesting since it suggests a strong departure from typical ism conditions .
even though such regions are close to the noise limits of the infrared and radio maps , this result appears to be significant . in figure [ fig - pixid ]
we show two examples ( i.e. ngc 4402 and ngc 4522 ) for which we detect 70 @xmath14 emission more radially extended than the radio ; these regions are found _ only _ along the galaxy edge associated with the radio deficit region .
furthermore , if @xmath45 were constant across these galaxy disks , we would expect to detect the radio continuum in these regions at @xmath103-@xmath51 rms level of the radio maps .
however , we know that @xmath45 does not remain constant across galaxy disks and actually decreases with increasing galactocentric radius @xcite , implying even higher levels of radio continuum should be expected , easily exceeding @xmath106-@xmath51 rms noise levels .
we therefore conclude that the non - detection of radio emission is significant at least at the @xmath16-@xmath51 level or , in other words , @xmath015% of the expected radio continuum emission is present .
these observations then suggest that the diffuse relativistic and gaseous ism have been stripped away along these leading edges leaving the denser molecular material as traced by the 70 @xmath14 dust emission .
presumably , the low density relativistic phase is quickly removed due to an increase in open magnetic field lines leading to the quick escape of cr particles . while our observations do not rule out that synchrotron emission associated with the remaining molecular clouds is still present and simply below our detection limit , since the diffuse component will dominate the total radio continuum emission in these regions , it is still interesting to speculate how quickly this synchrotron emission associated with the remaining molecular clouds would fade as cr electrons diffuse away . assuming a flux freezing scaling of @xmath137 ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , a solar neighborhood ism density and magnetic field strength of @xmath138 @xmath119 and @xmath139 g @xcite , and an average giant molecular cloud ( gmc ) density of @xmath140 @xmath119 leads to a gmc magnetic field strength of @xmath141 g .
this value is actually quite consistent with field strengths of molecular clouds inferred from the chandrasekhar - fermi technique @xcite using polarized dust continuum emission ( e.g * ? ? ?
* ) . rewriting equation 24 from @xcite for diffusion alone , such that @xmath142 where @xmath143 is the time it takes a cr electron emitting at frequency @xmath144 to diffuse a distance @xmath145 in a magnetic field of strength @xmath6 assuming random walk diffusion characterized by an energy - dependent diffusion coefficient , we can calculate the time it would take the cr electrons to diffuse away from the molecular clouds . taking @xmath146 g and @xmath147
kpc leads to a diffusion time of @xmath11.7@xmath148 yr for cr electrons to leave the molecular clouds ; this is considerably short compared to the stripping timescales and plausible given our observations .
such a scenario may have implications for star formation in these regions as well . according to equation [ eq - prism ] and assuming minimum energy arguments ,
the complete lack of cr pressure , which works to support a cloud from collapse , will decrease the total relativistic ism pressure by @xmath130% .
this situation may encourage cloud collapse in these regions leading to enhanced star formation .
on the other hand , the icm wind will also perturb molecular clouds introducing additional turbulence further discouraging collapse . determining which of these competing effects will actually dominate
is beyond the scope of this paper .
we have studied the interstellar medium ( ism ) of 10 virgo galaxies included in the vla imaging of virgo in atomic gas ( viva ) survey using _ spitzer _ mips and vla 20 cm imagery . by comparing the observed radio continuum images with modeled distributions ( i.e. appropriately smoothed fir images ) , created using a phenomenological image - smearing model described by @xcite , we find that the edges of many cluster galaxy disks are significantly radio deficient .
these radio deficit regions are consistent with being leading - edge regions affected by intracluster medium ( icm ) induced ram pressure as suggested by the location of h i and radio continuum tails . from our results we are able to conclude the following : 1 . the distributions of radio / fir ratios within cluster galaxies thought to be experiencing icm - ism effects are systematically different from the distributions in field galaxies .
radio / fir ratios are found to be systematically low , due to a deficit of radio emission , along the side of the galaxy experiencing the icm wind .
2 . in the case of ngc 4522 , a clear example of ongoing strong stripping , we find that the radio deficit region lies exterior to a region of high radio polarization and a flat radio spectral index .
we interpret this to suggest that cr electrons in the halos of galaxies are being swept up by the icm wind .
the icm wind drives shocklets into the ism of the galaxy which re - accelerate cr particles interior to the working surface at the icm - ism interface .
some cr electrons may also be redistributed downstream creating synchrotron tails as observed for ngc 4522 .
the high radio polarization is probably mostly the result of shear as the icm wind stretches the magnetic field ; compression may also play a role , though a modest one , since the total radio continuum in these regions does not appear significantly enhanced .
the severity of the current icm - ism interaction , as measured by the normalized difference between the modeled and observed radio flux densities in the radio deficit regions , appears to be strongly correlated with the time since peak pressure as measured by stellar population studies and gas stripping simulations .
thus the radio deficit appears to be a good tracer for the strength of current ram pressure .
4 . using the identified radio deficit regions we are able to get a quantitative estimate of the minimum strength of the icm pressure required to affect a galaxy disk ; we find values in the range of @xmath1@xmath149 dyn @xmath117 .
these pressures are generally smaller than , but similar to , those estimated for typical values of the icm gas density and galaxy velocities , as well as the range of ram pressures calculated by 3d hydrodynamical simulations . therefore , our estimates are consistent with the scenario of ram pressure creating the observed deficit regions .
the internal relativistic plasma pressures range between @xmath12 and @xmath150 dyn @xmath117 and appear to be lowest in some of the galaxies with the strongest radio deficits .
thus the strength of the radio deficit may also be affected by the internal relativistic plasma pressures of the galaxies as well as external ram pressure .
further work will be required to learn how much the radio deficits depend on these 2 variables .
. the global radio / fir ratios of these cluster galaxies are systematically higher than the average value found for field galaxies and appear to increase with increasing severity of the local radio deficit , which may trace current ram pressure .
we attribute this to an increase in the cr energy density which is greater when the interaction between the icm - ism is more energetic .
we also show that the amount of mechanical luminosity available from a galaxy s motion through the icm is significantly larger than what is required to heat the cr plasma .
this paper describes a preliminary study comparing the fir and radio emission distributions of virgo cluster galaxies using the first 10 out of @xmath140 galaxies for which _ spitzer _ data was available .
we have clearly shown that by comparing the fir and radio continuum distributions of these cluster galaxies we can address a number of physically important characteristics affecting their evolution including the strength and directions of the icm wind .
the inclusion of the remaining @xmath130 sample galaxies along with other ancillary data sets will help to better distinguish among scenarios presented here and better illuminate our understanding of how the icm affects the relativistic phase of the ism .
we would like to thank the anonymous referee for their useful suggestions that helped to improve the content and presentation of this paper .
we are grateful to the sings team for producing high quality data sets used in this study .
this work is based in part on observations made with the _ spitzer _ space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
support for this work was provided by nasa through an award issued by jpl / caltech . | we present a study on the effects of the intracluster medium ( icm ) on the interstellar medium ( ism ) of 10 virgo cluster galaxies using _ spitzer _ far - infrared ( fir ) and vla radio continuum imaging . relying on the fir - radio correlation within normal galaxies ,
we use our infrared data to create model radio maps which we compare to the observed radio images . for 6 of our sample galaxies
we find regions along their outer edges that are highly deficient in the radio compared with our models .
we also detect fir emission slightly beyond the observed radio disk along these outer edges .
we believe these observations are the signatures of icm ram pressure . for ngc 4522
we find the radio deficit region to lie just exterior to a region of high radio polarization and flat radio spectral index , although the total 20 cm radio continuum in this region does not appear strongly enhanced .
these characteristics seem consistent for other galaxies with radio polarization data in the literature .
the strength of the radio deficit is inversely correlated with the time since peak pressure as inferred from stellar population studies and gas stripping simulations , suggesting the strength of the radio deficit is good indicator of the strength of the current ram pressure .
we also find that galaxies having _ local _ radio _ deficits _ appear to have _ enhanced global
_ radio fluxes .
our preferred physical picture is that the observed radio deficit regions arise from the icm wind sweeping away cosmic - ray ( cr ) electrons and the associated magnetic field , thereby creating synchrotron tails as observed for some of our galaxies .
we propose that cr particles are also re - accelerated by icm - driven shocklets behind the observed radio deficit regions which in turn enhances the remaining radio disk brightness .
the high radio polarization and lack of precisely coincident enhancement in the total synchrotron power for these regions suggests shearing , and possibly mild compression of the magnetic field , as the icm wind drags and stretches the leading edge of the ism . |
it is well known that an electrical current @xmath0 circulating in a ring with a non - zero resistance @xmath1 is maintained @xmath2 by the fadaday electric field @xmath3 .
the current would rapidly decay @xmath4 during the relaxation time @xmath5 when a magnetic flux inside the ring @xmath6 does not change with time @xmath7 .
@xmath8 is the circumference of the ring ; @xmath9 is the inductance of the ring .
however , the measurements @xcite demonstrate that an electrical current can not decay in spite of @xmath10 and @xmath7 . these mesoscopic quantum phenomena , known as persistent current ,
are observed in a normal metal ring @xcite at temperatures enough low and at temperatures corresponding to the fluctuation region of the superconducting transition @xcite .
the quantum theories @xcite describe these phenomena @xcite as a consequence of the momentum quantization @xmath11 of quantum particles , i.e. particles described by the wave function @xmath12 , and the aharonov - bohm effect , connected with the dependence of the canonical momentum @xmath13 of a charged @xmath14 particle on the magnetic vector potential @xmath15 . the direction and value of the persistent current corresponding to thermodynamic equilibrium change periodically with a magnetic flux @xmath16 inside the ring of an area @xmath17 @xcite .
the @xmath18 oscillation period corresponds to the flux quantum @xmath19 @xcite . the persistent current of electrons ( which are fermions with the charge @xmath20 ) observed in a normal metal ring @xcite oscillates with the period @xmath21 and its amplitude can not exceed @xmath22 @xcite , where @xmath23 is the number of one - dimensional channels @xcite and @xmath24 is the current of each channel .
it can be said that the current @xmath24 at @xmath25 is the current of single electron @xmath26 on the fermi level @xmath27 @xcite because the oppositely directed currents of electrons occupying other permitted levels ( 1 ) compensate for each other @xcite .
the amplitude of the @xmath18 oscillations observed in measurements of real normal metal rings @xcite is appreciably lower than the current of a single electron ( @xmath28 at typical radius @xmath29 and the fermi velocity @xmath30 ) because of electron scattering and non - zero temperature of measurements @xcite .
the amplitude of the oscillations @xmath18 with the period @xmath31 observed in measurements of a superconducting ring is appreciably higher @xmath32 even near the critical temperature @xmath33 @xcite because all @xmath34 pairs ( bosons with the charge @xmath35 ) in the ring of a macroscopic volume @xmath36 have the same velocity @xmath37 corresponding to the minimum energy @xmath38 .
the quantum theories @xcite describe the quantum oscillations @xmath18 and temperature dependence of their amplitudes both for normal metal @xcite and superconducting @xcite rings .
but these theories do not answer two important questions : `` why does the persistent current not decay in spite of non - zero resistance ? '' and `` could the persistent current induce a persistent voltage in the ring arms of different resistance ? ''
two possible answers to the first question are considered in the paper @xcite . according to the first answer , assumed by the authors of @xcite and @xcite , the persistent current can flow without dissipating energy in spite of non - zero electric resistance .
contrary to this preposterous assumption @xcite the author @xcite emphasized , as far back as 1970 , that weak energy dissipation should not result in the disappearance of the persistent current because this equilibrium phenomenon corresponds to the minimum free energy . according to this second answer
the persistent current observed at non - zero resistance is the brownian motion @xcite similar to the nyquist @xcite ( or johnson @xcite ) equilibrium electric current .
it is argued in @xcite that this second answer rather than the first one agrees with experimental data .
the antithetical answers to the first question also presume opposing answers to the second one .
the authors @xcite use the analogy to stationary electron orbits of an atom as a sole argument for their interpretation of the persistent current as dissipationless .
but the rings used for the observation of the persistent current @xcite basically differ from atoms .
rings have an electric resistance @xmath39 and it is possible to make a ring with different resistance of its arms @xmath40 , @xmath41 .
it is well known that the potential difference @xmath42 is observed in both arms of the ring @xmath43 , @xmath44 when a conventional electric current @xmath45 maintained by the fadaday electric field @xmath46 circulates along the ring circumference @xmath47 .
the sign of the voltage @xmath48 on the arm with smaller resistance @xmath49 is opposite to the @xmath50 sign .
but the current is observed in accordance with the ohm s law @xmath51 and its direction corresponds to the direction of an electric field @xmath52 in both arms , @xmath53 and @xmath54 , due to the non - zero fadaday voltage @xmath55 .
the persistent current , in contrast to the conventional circular current , is observed at @xmath56 and it should be directed opposite to the electric field @xmath57 in one of the arms at @xmath58
. therefore it may seem that no persistent voltage can arise , basing on the idea that no direct electric current can flow against the direct electric field @xmath57 in the ring arm .
but the experimental data @xcite have disproved this .
the persistent current flowing against the electric field @xmath57 is observed in measurements of the little - parks oscillations @xcite .
these oscillations of the resistance observed by w. a. little and r. d. parks as far back as 1962 @xcite is the first experimental evidence of @xmath59 at @xmath10 .
another phenomenon disproving the idea that the persistent current can not flow opposite to the electric field was discovered forty years later @xcite .
the observations @xcite of the dc voltage @xmath60 on the arms of asymmetric rings , the sign and value of which oscillate in magnetic field like the persistent current @xmath61 , give unambiguous evidence that the averaged electric current @xmath61 flows against the dc electric field @xmath62 in one of the arms . in terms of quantum mechanics
these paradoxical phenomena can be described as a consequence of switching between superconducting states with different connectivity of the wave function @xcite . in order to verify the possibility of persistent voltage existence the amplitude of oscillations @xmath63
should be increased to a measurable value .
in one of the first attempts to observe the persistent current in normal metal rings the authors @xcite measured a magnetization response of an array of 10 million copper rings .
their strategy was to add together many small signals to create a much larger , measurable signal .
this attempt was not quite successful because the persistent current in each normal metal ring has a random sign and therefore the total signal was proportional only to the square root @xmath64 of the number of rings @xmath65 .
in contrast to this case the persistent current in each superconducting ring of identical radius has the same sign at any value of magnetic field @xmath66 .
therefore it can be expected that the persistent voltage of a system of asymmetric superconducting rings connected in series can increase in proportion to the number of rings @xmath65 .
the proportionality @xmath67 implies that the voltage @xmath68 can be made measurable even at very small persistent voltage @xmath69 of a single ring .
for example , in a system of @xmath70 rings the oscillations @xmath68 with an amplitude exceeding the measurable voltage @xmath71 can be observed even at the extremely small persistent voltage value @xmath72 .
our experimental facilities has allowed the fabrication of systems with a much smaller number of asymmetric aluminium rings , @xmath73 and @xmath74 , fig.1 .
the measurements of these systems showed that the number of rings @xmath74 is not sufficient to observe the persistent voltage above the superconducting transition .
nevertheless we could observe oscillations of the dc voltage at temperatures corresponding to the lower part of this transition .
the results of our measurements corroborate the proportionality @xmath75 and allow the estimation of the number @xmath65 required for the observation of persistent voltage above the superconducting transition .
the structures consisting of 110 and 1080 series - connected asymmetric rings were fabricated by the lift - off method by depositing a thin aluminium film @xmath76 thick onto a si substrate .
the lithography was performed using a jeol-840a scanning electron microscope transformed into a laboratory electron lithograph by the nanomaker program package .
all rings are of the same inner radius @xmath77 and semi - ring widths @xmath78 , @xmath79 , fig.1 .
the distance between the rings and the width of the strips connecting them is @xmath80 .
the cross sections of the semi - rings are @xmath81 and @xmath82 . with the london penetration depth and the correlation length in aluminium film being @xmath83 and @xmath84 @xcite
, the cross sections correspond to a weak screening @xmath85 and the structure is one - dimensional @xmath86 at the temperature of measurements @xmath87 .
the midpoint of the resistive transition corresponds to @xmath88 for a 110 ring system and @xmath89 for a 1080 ring system ; the width of the transition is @xmath90 for the booth systems , and the maximum slope is @xmath91 and @xmath92 .
the resistance of the 110 and 1080 rings structures in the normal state is @xmath93 ( the resistance of each ring is @xmath94 ) and @xmath95 ( the resistance of each ring is @xmath96 ) .
the resistance per square is @xmath97 , @xmath98 , and the resistance ratio is @xmath99 .
system of 1080 series - connected asymmetric aluminum rings with the same inner radius @xmath77 and the semi - ring widths @xmath78 , @xmath79 .
the voltage was measured between the v - v contact pads , and for the current the i - i contact pads were used .
the inset at the centre shows a system fragment of a large scale . ]
the measurements were performed in a glass cryostat using @xmath100 as a cooling agent . by pumping
, the temperature could be reduced to @xmath101 .
the voltage was measured across the potential contact pads @xmath102 ( fig.1 ) by an instrumental amplifier with a gain of 1000 and an input - normalised noise level of @xmath103 in the frequency band from 0 to 1 hz . then , the signal was supplied to an sr560 preamplifier ( stanford research systems ) , which was used for additional amplification and formation of the required signal frequency band by low - pass and high - pass filters .
because of the galvanic connection between the measured sample located at low temperature @xmath104 and the measuring system located at room temperature @xmath105 , a thermal electric noise disturbs the thermodynamic equilibrium of ring system . in this work
, we used a filtration system consisting of low - temperature @xmath106-filters and coaxial resistive twisted pairs with attenuation at high frequencies to reduce this noise and to thermalize samples at low temperatures . owing to the filtration , the frequency of the transmission band of each conductor leading to the sample was equal to @xmath107 , and
the suppression 70 to 80 db was achieved at frequencies in the range @xmath108 to @xmath109 .
the temperature was measured by a calibrated resistance thermometer ( @xmath110 ) with a measuring current of @xmath111 .
an external current @xmath112 , constant , sine or noise , was supplied to the current contact pads @xmath113 ( fig.1 ) .
the constant @xmath112 current from @xmath114 to @xmath115 was used to measure the resistance versus magnetic field @xmath116 ( the little - parks oscillations ) and temperature @xmath117 .
the sine @xmath118 and noise @xmath119 currents were used to calibrate the ring systems as a noise detector i.e. in order to obtain the dependence of the amplitude of dc voltage oscillations in magnetic field @xmath120 on the amplitude @xmath121 or @xmath122 . the amplified voltage taken from the sample and the signals proportional to the current passing through the sample , to the magnetic field , and to the temperature ,
were simultaneously digitized by a 16-digit a / d converter with eight differential inputs .
the persistent voltage , i.e. the dc voltage @xmath63 at the thermodynamic equilibrium , may be observed only in the temperature region near @xmath123 corresponding to the resistive transition where @xmath124 because of switching by thermal fluctuations of ring segments between superconducting and normal states .
the resistance of homogeneous one - dimensional superconductor changes from zero @xmath125 to the resistance in the normal state @xmath126 in a narrow interval estimated with the standard scaling factor @xmath127 for the temperature variable @xmath128 , see fig.19 of @xcite and fig.1 @xcite , where @xmath129 is sometimes called the ginzburg number .
a mean free path @xmath130 of the aluminum systems used in our work is short @xmath131 and therefore their resistive transition should be described by a theoretical dependence obtained in the dirty limit , fig.1a @xcite .
according to this dependence the resistance should change , for example , from @xmath132 to @xmath133 in the interval @xmath134 .
the ginzburg number of the one - dimensional aluminum structure used in our work is approximately equal to @xmath135 at its parameters : @xmath136 , the average cross sections of the semi - rings @xmath137 , the thermodynamic critical field @xmath138 and the correlation length @xmath139 @xcite extrapolated to the zero temperature .
our measurements showed that the width of resistive transition of a single ring agrees with the expected value @xmath140 whereas the resistive transition of ring systems is wider .
the latter suggests that these systems are not homogeneous .
the persistent voltage should exponentially decrease above the superconducting transition @xmath141 because the persistent current decrease with temperature increase @xmath142 @xcite . below the transition at @xmath143 the amplitude @xmath144 of the persistent current oscillations @xmath18 increases with temperature decrease @xmath145 @xcite .
the persistent voltage could be expected to exponentially decrease with going away from @xmath123 also at @xmath143 because of an exponential decrease of the resistance calculated in the zero - current limits @xcite , i.e. under equilibrium conditions . according to the relation @xmath146 obtained in @xcite
the resistance of a homogeneous aluminium structure ( with parameters of our structure @xmath147 ) should fall down to @xmath148 at @xmath149 , down to @xmath150 at @xmath151 and so forth .
the equilibrium resistance should fall because the probability of the jumps in the normal state @xmath152 diminishes with an increase of the energy difference between the normal and superconducting states @xmath153 @xcite .
the probability @xmath154 is negligibly small in the superconducting state at @xmath155 under equilibrium conditions .
it increases @xcite and reaches @xmath156 when a non - zero external current @xmath157 exceeds the critical current @xmath158 .
therefore , a non - equilibrium ac current , rather than thermal fluctuations , can switch segments of a one - dimensional superconductor between the superconducting and normal states at @xmath155 when its amplitude exceeds the critical current @xmath159 .
it was discovered that such switching can induce a dc voltage @xmath160 oscillating in magnetic field @xmath66 in superconducting asymmetric rings @xcite and percolating films @xcite .
the oscillations @xmath161 observed in asymmetric aluminium rings @xcite are similar to the oscillations of the persistent current observed on such rings . the period of these alternating - sign oscillations @xmath162 corresponds to the flux quantum @xmath163 inside the ring with the area @xmath164 whereas the @xmath60 sign @xcite and direction of the persistent current @xcite change at @xmath165 and @xmath166 .
the observations of the oscillations @xmath167 suggest that the persistent voltage @xmath168 can arise but they can not be interpreted as an unambiguous experimental evidence of this equilibrium phenomena because the dc voltage @xmath60 at @xmath155 was induced by a non - equilibrium noise @xmath169 @xcite or sine current @xmath170 @xcite .
this non - equilibrium phenomenon may be considered as a rectification effect , i.e. the transformation of the ac current power @xmath171 into the power of the dc voltage @xmath172 .
the dc voltage , as well as the average value of the persistent current @xcite , equals zero at @xmath173 , @xmath174 and has a maximum @xmath175 between these values of a magnetic flux , see fig.2 .
the amplitude @xmath175 of the @xmath167 oscillations does not depend on the frequency @xmath176 of the sine current @xmath170 ( at least in the frequency range @xmath177 at @xmath178 ) @xcite and the maximum @xmath179 of the rectification efficiency @xmath180 has the same value for both noise current with the amplitude @xmath122 and sine current with the same amplitude @xmath181 .
voltage oscillations in magnetic field , @xmath167 , measured in 110 rings at the temperature @xmath182 corresponding to a maximum amplitude for this system with @xmath183 and in 1080 rings at @xmath184 below this temperature for this system with @xmath185 . the dc voltage was induced in the both cases with the same controllable ac current with the amplitude @xmath186 .
the inset shows the dependence of the maximum rectification efficiency @xmath187 ( vertical reference axis ) on the amplitude @xmath188 of the controllable ac current ( horizontal reference axis , microampere , @xmath189 ) measured on the 110 ring system . ]
the dc voltage @xmath167 is observed at the amplitude of the ac current @xmath190 @xcite when the sine or noise current can switch a ring or its segments to the normal state .
the critical current @xmath191 of the ring equal to @xmath158 at @xmath192 changes periodically in magnetic field and decreases at high @xmath193 , see fig.6 and 9 of @xcite .
the oscillations @xmath167 appear initially at a high magnetic flux inside the rings , see fig.2 of @xcite and fig.11 of @xcite , because @xmath191 decreases with increasing @xmath193 .
the amplitude @xmath175 of these oscillations is a nonmonotonic function of both ac current amplitude @xmath188 ( at a given temperature @xmath194 ) , see fig.6 of @xcite and fig.12 of @xcite , and temperature ( at a given @xmath188 value ) , see fig.6 of @xcite .
the oscillations @xmath167 are not observed at low amplitudes of ac current , fig.6 of @xcite , fig.12 of @xcite , and low temperatures , fig.6 of @xcite , because the rings can not be switched in the normal state at @xmath195 .
the amplitude @xmath175 decreases down to an immeasurably small value at high temperatures @xmath196 , fig.6 of @xcite , because of a decrease of the persistent current @xcite .
the maximum amplitude @xmath197 of the oscillations @xmath167 observed at @xmath198 corresponds to the maximum rectification efficiency @xmath199 .
the rectification efficiency @xmath200 allows the estimation of the minimum amplitude @xmath201 of the ac current capable of inducing @xmath167 oscillations with an amplitude @xmath202 , high enough for real observations .
the measurements @xcite have shown that the rectification efficiency of asymmetric aluminium rings is abnormally high in the superconducting state @xmath203 at @xmath204 because of the hysteresis of the current - voltage characteristics in this temperature region . a noise with the amplitude down to @xmath205 could be detected with the help of a 110 ring system ( @xmath93 ) at the measurable voltage @xmath206 and the above value of the rectification efficiency .
but the critical current of this system @xmath207 at @xmath204 is much higher than @xmath208 .
the rectification efficiency decreases at @xmath209 , fig.15 of @xcite , and thus at @xmath210 , fig.2 .
in contrast to the critical current , the value @xmath200 does not fall down to zero at @xmath209 .
therefore it is important that the amplitude @xmath211 of the oscillations @xmath167 induced by ac current of the same amplitude @xmath188 should increase with the number of rings , fig.2 .
both the amplitude @xmath211 and the resistance @xmath212 increase approximately proportionally to the number of rings .
therefore the systems of 110 and 1080 rings have the same maximum rectification efficiency @xmath213 at the same amplitude @xmath214 .
our measurements confirmed that the system of 1080 rings can detect a noise with a lower amplitude @xmath215 than the system of 110 rings .
magnetic dependence of the dc voltage , @xmath167 , ( on the left side ) and the fourier transform ( on the right side ) .
the peak of the fourier transform at @xmath216 is evidence of the quantum oscillations observed on the rings with radius @xmath217 .
the curve 1 ( on the left side ) is an original dependence measured in the 1080 rings .
the oscillations 1 and 2 are the reverse fourier transform taken in a limit region @xmath218 .
the temperature of 1 and 1 measurement @xmath219 corresponds to the maximum amplitude of the quantum oscillations , fig.4 .
the dependence 2 measured at @xmath220 is typical for a lower temperature .
the amplitude of the oscillations @xmath221 was measured at low @xmath66 .
the curves 1 and 2 are vertically shifted . ]
the @xmath167 oscillations with an amplitude to @xmath222 observed in @xcite were due to an uncontrollable noise existing in any real measuring system .
this non - equilibrium noise is a consequence of the galvanic connection between the parts of the measuring system located at low temperature @xmath104 and at room temperature @xmath105 .
additional filtration with the help of low - temperature @xmath106-filters and coaxial resistive twisted pairs reduced the uncontrollable noise .
the diminution of the uncontrollable noise allowed the expansion of the calibration @xmath223 of the ring system as a noise detector @xcite to the lower amplitude @xmath188 , fig.2 . in order to obtain the dependence @xmath223 shown in fig.2 , the maximum @xmath211 of the temperature dependence @xmath224
was measured at each value of the amplitude @xmath188 of the controllable ac current . according to this calibration the uncontrollable noise inducing the oscillations @xmath167 with the maximum amplitude @xmath222 in @xcite had the amplitude @xmath225 .
the noise reduced with the help of the additional filtration was undetectable with the 110 ring system but the system of 1080 rings detected it . to pick out barely visible quantum oscillations @xmath167 in the real dependence of the voltage @xmath161 we used the fourier transformation , fig.3 .
this method allowed the detection of quantum oscillations with the amplitude down to @xmath226 .
our measurements made with the additional filtration showed that the amplitude @xmath211 of quantum oscillations @xmath167 induced by an uncontrollable noise is @xmath103 lower in the 110 ring system and reaches @xmath227 in the system of 1080 rings , fig.4 .
the amplitude @xmath228 is by @xmath229 time lower than the maximum amplitude @xmath230 of the oscillations @xmath167 observed at the controllable noise with @xmath231 .
consequently , the amplitude @xmath232 of the uncontrollable noise should be higher than @xmath233 and lower than @xmath234 .
a rough extrapolation of the dependence @xmath223 , fig.2 , gives @xmath235 .
thus , low - temperature @xmath106-filters and coaxial resistive twisted pairs allowed the reduction of the amplitude of the uncontrollable noise approximately by a factor of ten .
the detection of the weak noise only with the help of the 1080 ring system demonstrates that a system of asymmetric superconducting rings is a promising detector of an uncontrollable noise .
this noise in the low - temperature part of a typical measuring system is much higher than @xmath235 .
for example , the observations of @xmath167 oscillations in a single aluminum ring ( resistance @xmath236 ) with the amplitude @xmath237 @xcite and in an asymmetric superconducting quantum interference device ( @xmath238 ) with @xmath239 @xcite indicate that the amplitude of an uncontrollable noise exceeds @xmath240 in the first case and @xmath241 in the latter .
screening and filtration reduce the uncontrollable noise down to @xmath225 @xcite .
additional filtration with low - temperature @xmath106-filters and coaxial resistive twisted pairs reduce its amplitude to @xmath235 .
each time the systems with a larger number of asymmetric rings allowed the detection of a reduced non - equilibrium noise .
the proportionality @xmath242 corroborated by our measurements ensures the detection of a weaker noise down to equilibrium by a system with a number of rings larger than 1080 .
it can be expected , for example , that @xmath167 oscillations with the maximum amplitude @xmath243 , fig.3 , will be observed in the system of 10000 rings of the same type at @xmath235 and that the noise with a lower amplitude @xmath244 can be detected by this system .
temperature dependence of the oscillation amplitude @xmath175 of a noise - induced voltage @xmath167 measured in the 1080 rings ( indicated by circles ) and expected in 10000 rings .
the data are shown on the background of the resistive transition ( @xmath245 ) .
@xmath246 is the resistance of the 1080 ring system in the normal state . ]
the noise current with the amplitude @xmath235 detected with the help of 1080 rings can not greatly exceed the nyquist current @xmath247 .
the frequency @xmath176 of the nyquist current in the normal state is limited by the relaxation time @xmath248 and the quantum limit @xmath249 .
the quantum limit @xmath250 is a critical factor at @xmath251 , the ring inductance @xmath252 and the resistance along the circumference @xmath253 .
the amplitude @xmath254 in the whole frequency band from @xmath255 to the quantum limit @xmath256 exceeds the amplitude @xmath235 of the noise current detected by the 1080 ring system .
however there are some factors limiting the frequency of the noise current , both non - equilibrium and equilibrium , capable of inducing of both @xmath167 and @xmath168 .
one of these factors is the time of relaxation to the equilibrium superconducting state .
this time , equal to @xmath257 in the linear approximation , increases near the superconducting transition , limiting the frequency band much stronger in the region where a weak noise current can induce @xmath167 or @xmath168 .
other factors can also contribute to the decrease of the frequency band and consequently the amplitude of the persistent voltage .
nevertheless , the proportionality @xmath242 should allow the detection of the persistent voltage with any amplitude when a system is made with a sufficiently large number of asymmetric superconducting rings .
we assume that the quantum oscillations @xmath167 observed in the 1080 rings , fig.3 , are induced mainly by non - equilibrium noise because of the temperature dependence of the amplitude @xmath224 , fig.4 .
the maximum of the @xmath224 dependence is observed at @xmath219 corresponding to the lower part of the resistive transition @xmath258 whereas the maximum of the persistent voltage @xmath168 amplitude should be expected at @xmath259 corresponding to the middle of the resistive transition @xmath260 .
we could observe the quantum oscillations @xmath167 up to @xmath261 where @xmath262 . however
this result does not unambiguously prove that we observed the persistent voltage in the fluctuation region because of some non - homogeneity of the system used in our work .
its resistive transition @xmath263 is wider than the one @xmath264 expected for homogeneous system .
the rings of our real system have different critical temperatures @xmath265 and different values @xmath266 at @xmath261 .
some rings with lower @xmath265 and @xmath267 may contribute to the resistance @xmath262 whereas other rings with @xmath268 may contribute to the quantum oscillations @xmath167 .
the non - homogeneity of our system can partly explain the decrease of the rectification efficiency @xmath200 with @xmath269 , fig.2 .
a ring @xmath270 with its critical temperature @xmath271 can contribute to the dc voltage @xmath167 only at @xmath272 and @xmath273 .
therefore , all rings can not contribute to the total voltage @xmath167 at any temperature @xmath194 when @xmath274 and our real system of 1080 rings with different @xmath123 could not detect the persistent voltage even if the maximum amplitude of @xmath69 appreciably exceeds @xmath275 .
in summary , we have experimentally studied the possibility to observe the persistent voltage in asymmetric superconducting rings using systems of 110 and 1080 such rings connected in series .
the results of our measurements unambiguously point to the possibility that very weak electric noise can be detected with the help of such systems having a large enough number of asymmetric rings connected in series .
we succeeded in reducing the non - equilibrium noise in the low - temperature part of our measuring system down to a very small power .
nevertheless , the system of 1080 rings could detect this weak noise even in spite of some imperfections of this system : its non - homogeneity , too high critical current of the rings and so on .
our results have corroborated that systems with a larger number of asymmetric rings connected in series can detect a weaker noise down to the equilibrium one .
a possible observation with help of such systems the quantum oscillations of the dc voltage in the temperature region corresponding to the upper part of the resistive transition will be unambiguous evidence of the persistent voltage because non - equilibrium noise can induce a resistance but it can not induce the persistent current .
our results suggest that the persistent voltage can be observed with the help of @xmath276 rings .
dubonos , v.i .
kuznetsov , a.v .
nikulov , in proceedings of 10th international symposium `` nanostructures : physics and technology '' , st petersburg : ioffe institute , ( 2002 ) p.350 ; arxiv : 1112.6157 ( 2011 ) . v.l .
gurtovoi , s.v .
dubonos , a.v .
nikulov , n.n .
osipov , v.a .
tulin , in the proceedings of spie vol .
6260 , micro- and nanoelectronics - 2005 , edited by kamil a. valiev and alexander a. oplikovskii , ( 2006 ) p. 62600t1 ; arxiv : cond - mat/0603009 | a possibility to observe the persistent voltage in a superconducting ring of different widths of the arms is experimentally investigated .
it was earlier found that switching of the arms between superconducting and normal states by ac current induces the dc voltage oscillation in magnetic field with a period corresponding to the flux quantum inside the ring .
we use systems with a large number of asymmetric rings connected in series in order to investigate the possibility to observe this quantum phenomenon near the superconducting transition where thermal fluctuations switch ring segments without external influence and the persistent current is much smaller than in the superconducting state . |
the theory of subfactors ( @xcite ) has experienced several new developments in the last few years through the introduction of planar algebra technology ( @xcite ) .
every subfactor comes with a very rich mathematical object , the _ standard invariant _ or _
planar algebra _ of the subfactor , which in nice situations is a complete invariant of the subfactor ( @xcite , @xcite ) .
it can be described in many interesting ways , as for instance a certain category of bimodules ( @xcite , see also @xcite ) , as lattices of multi - matrix algebras ( @xcite ) , or as a planar algebra ( @xcite ) .
the planar algebra approach is particularly powerful since it allows one to use algebraic - combinatorial methods in conjunction with topological ones to investigate the structure of subfactors .
a number of examples of explicit planar algebras associated to subfactors have been computed ( see for instance @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) but there is a need for more concrete examples .
this is what we accomplish in this paper .
we give a description of the planar algebra of the diagonal subfactors associated to a @xmath0-kernel .
let @xmath3 be a ii@xmath2 factor and let @xmath4 , @xmath5 , @xmath6 be automorphisms of @xmath3 ( we may assume without loss of generality that @xmath7 ) .
consider the subfactor @xmath8 , where @xmath9 denote matrix units in @xmath10 .
@xmath11 is then called the _ diagonal subfactor _ associated to @xmath12 .
these subfactors were proposed by jones in 1985 to provide examples of potentially non - classifiable subfactors , since this construction allows one to translate problems on classification of group actions into problems on subfactors .
popa used them to prove vanishing of 2-cohomology results for cocycle actions of finitely generated , strongly amenable groups on an arbitrary ii@xmath2 factor ( @xcite ) .
ocneanu had proved such a result for cocycle actions of amenable groups on the hyperfinite ii@xmath2 factor using different techniques ( @xcite ) .
the diagonal subfactors are of course reducible and have jones index @xmath13 .
they provide a wealth of simple examples of infinite depth subfactors whose structure theory is well understood .
in particular , the standard invariant or planar algebra of these subfactors has been determined in ( @xcite , @xcite , @xcite ) .
let @xmath0 be the group generated by the images @xmath14 of @xmath15 , @xmath16 , in @xmath17 .
popa showed that analytical properties of these subfactors are reflected in the corresponding properties of the group @xmath0 .
for instance , if @xmath3 is hyperfinite , then the diagonal subfactor is amenable ( in the sense of popa ) if and only if @xmath0 is an amenable group ( @xcite ) .
the subfactor has property ( t ) in the sense of popa if and only if @xmath0 has property ( t ) of kazhdan ( @xcite ) .
the principal graphs of these subfactors are then cayley - like graphs of @xmath0 with respect to the generators @xmath18 , @xmath5 , @xmath19 and their inverses ( see @xcite or @xcite for the precise statement ) .
the higher relative commutants , jones projections and conditional expectations have all been worked out in ( @xcite , @xcite , @xcite ) .
a well - known variation of the diagonal subfactors is obtained as follows ( see e.g. @xcite ) .
consider a @xmath0-kernel , that is an injective homomorphism @xmath20 from a ( countable , discrete , finitely generated ) group @xmath0 into @xmath21 .
denote by @xmath22 the canonical homomorphism , and let @xmath23 be a lift of @xmath20 such that @xmath24 , for all @xmath25 .
it follows that @xmath26 , for all @xmath27 , @xmath28 , and some unitaries @xmath29 .
associativity of composition of automorphisms implies that @xmath30 . hence there is a function @xmath31 with @xmath32 .
it is easy to check that @xmath1 is a 3-cocycle and that its class in @xmath33 does not depend on the choices made .
one usually denotes the class by @xmath34 or @xmath35 , the _ obstruction _ of @xmath20 .
it is an obstruction to lifting @xmath0 to an _ action _ on the ii@xmath2 factor @xmath3 . clearly , if two @xmath0-kernels @xmath20 and @xmath36 are conjugate ( in @xmath21 ) , then @xmath37 .
it was shown in @xcite that @xmath38 is a complete conjugacy invariant for @xmath0-kernels if @xmath3 is the hyperfinite ii@xmath2 factor and @xmath0 is a countable , discrete , amenable group . note that in general , even if @xmath39 , there may be no lifting of the @xmath0-kernel to an action on @xmath3 .
connes and jones found in @xcite the first such example of a non - liftable @xmath0-kernel with vanishing obstruction , where @xmath0 is a group with property ( t ) .
vanishing of the obstruction is a necessary and sufficient condition for @xmath0 to lift to a _ cocycle action _ on the ii@xmath2 factor @xmath3 .
see @xcite , @xcite , @xcite , @xcite , @xcite for more on this .
connes showed that if @xmath0 is cyclic , one can construct @xmath0-kernels in @xmath40 with arbitrary obstructions , where @xmath41 denotes the _ hyperfinite _ ii@xmath2 factor ( @xcite ) .
it was an open problem whether this result is true for more general groups , and jones showed in @xcite that this is indeed the case for @xmath0 an arbitrary discrete group .
thus , given a discrete group @xmath0 and a class @xmath42 , there is @xmath0-kernel @xmath43 with @xmath44 .
sutherland constructed @xmath0-kernels with arbitrary obstructions in non - hyperfinite ii@xmath2 factors ( @xcite ) . given a finitely generated , countable group @xmath0 and a 3-cocycle @xmath1 , we can associate a ( hyperfinite ) subfactor to @xmath45 as follows .
let @xmath46 be a @xmath0-kernel with @xmath47 \in h^3(g , \mathbb t)$ ] ( choose for instance @xmath48 , and use @xcite ) .
fix generators @xmath49 of g , let @xmath50 be a lift of @xmath20 , and consider the diagonal subfactor associated to the automorphisms @xmath51 , @xmath5 , @xmath52 ( let us choose @xmath53 , the identity of @xmath0 , and @xmath54 ) . if @xmath36 is another @xmath0-kernel with lift @xmath55 , and automorphisms @xmath56 , @xmath5 , @xmath57 , then the diagonal subfactors associated to @xmath58 and @xmath59 are isomorphic if and only if there is an automorphism @xmath60 of the underlying ii@xmath2 factor @xmath3 such that @xmath61 , where @xmath62 is a permutation of the indices ( fixing @xmath63 ) ( see e.g. @xcite ) . thus , in particular , isomorphism of these diagonal subfactors implies that @xmath64 ( up to a possible modification of @xmath20 by the permutation @xmath62 ) .
isomorphism of standard invariants is weaker than isomorphism of subfactors , but we still have the following . if @xmath65 and @xmath66 are two @xmath0-kernels as above with @xmath67 , then the standard invariants of the associated diagonal subfactors are isomorphic ( see ( @xcite , page 228 ff . ) .
the converse is not true due to the fact that group isomorphisms can change the class of a 3-cocycle .
if one constructs diagonal subfactors where the automorphisms are repeated with distinct multiplicities , one can show a converse , e.g. if @xmath0 is strongly amenable and @xmath3 is hyperfinite , using popa s classification results of amenable subfactors ( @xcite , page 229 ff . ) .
clearly , the 3-cocycle @xmath1 giving rise to the obstruction of the @xmath0-kernel will appear in the standard invariant of these diagonal subfactors ( with cocycle ) and the purpose of this paper is to give a precise description of this occurrence .
here is a more detailed outline of the sections of this paper .
we review group cocycles in section 2 . in section 3 ,
we define an abstract planar algebra @xmath68 associated to a finitely generated group @xmath0 with a fixed finite generating set @xmath69 , and a 3-cocycle @xmath70 .
the vector spaces underlying this planar algebra are spanned by multi - indices in @xmath71 such that the corresponding alternating word on generators and their inverses is the identity in @xmath0 .
the action of jones planar operad is defined explicity , and of course @xmath1 appears prominently in this definition .
it should be noted that the definition of the action of a tangle involves a labelling of the _ strings _ in the tangle whereas the planar algebra description of the group - type subfactors in @xcite involved a labelling of _ boundary segments _ ( called `` openings '' in @xcite ) of the internal and external discs of the tangle .
we would like to point out that the 3-cocycle @xmath1 does not appear in our definition of the action of the multiplication , inclusion , jones projection and right conditional expectation tangles .
it does appear in the definition of the action of the left conditional expectation tangles ( and hence the rotation tangles ) .
we verify in this section that our definition of the action of tangles is indeed natural with respect to composition of tangles .
this takes a little work , but some interesting identities for 3-cocycles appear along the way . in section 4
we give a model for the higher relative commutants of the diagonal subfactor with cocycle and describe the associated concrete planar algebra .
we choose an appropriate basis of the higher relative commutants which allows us to identify this concrete planar algebra with the abstractly defined one in section 3 .
this isomorphism is obtained in the usual way by constructing a filtered @xmath72-algebra isomorphism between the abstract planar algebra of section 3 and the concrete one of section 4 , that preserves jones projection and conditional expectation tangles .
the main feature of this planar algebra is the fact that the distinguished basis of the higher relative commutants that we found here , matches with the one coming from the description of the planar algebra as a path algebra associated to the principal graphs of the subfactor ( see e.g. @xcite , @xcite ) .
conversely , we prove that any finite index extremal subfactor whose standard invariant is given by the abstract planar algebra ( in section 3 ) , must necessarily be ( isomorphic to ) a diagonal subfactor .
for the convenience of the reader , we recall in this brief section the definition of a cocycle of a group @xmath0 .
@xmath0 will denote throughout this article a countable , discrete group , and we will denote the identity of @xmath0 by @xmath73 . define @xmath74 the space of functions from @xmath75 to @xmath76 and @xmath77 by + @xmath78 it follows that @xmath79 where @xmath63 denotes the constant function @xmath63 .
denote @xmath80 by @xmath81 ( whose elements are called @xmath82-cocycles ) and @xmath83 by @xmath84 ( whose elements are called coboundaries ) .
note that @xmath85 .
in this paper we will be dealing mostly with a @xmath86-cocycle @xmath1 .
thus @xmath1 satisfies the identity @xmath87 we call @xmath1 _ normalized _ if @xmath88 whenever either of @xmath18 , @xmath89 , @xmath90 is @xmath73 .
any cocycle @xmath1 is coboundary equivalent to a normalized cocycle . in particular
, @xmath91 is a normalized @xmath86-cocycle where @xmath92 is defined as @xmath93 for all @xmath94 .
in this section we will construct an abstract planar algebra which , in section [ subfact ] , will be shown to be isomorphic to the planar algebra of a diagonal subfactor with cocycle .
let @xmath0 be a countable , discrete group generated by a finite subset @xmath69 , and let @xmath70 be a normalized @xmath86-cocycle .
we will construct a planar algebra @xmath68 (= @xmath3 ) as follows .
let @xmath73 denote the identity of @xmath0 .
we define first a map @xmath95 from multi - indices @xmath96 to @xmath0 by @xmath97 where @xmath95 of the empty multi - index is defined to be the identity element of the group . to define the planar algebra we need to define vector spaces @xmath98 and an action of jones planar operad on these vector spaces .
we refer to @xcite for the planar algebra terminology used in this paper .
for @xmath99 , define @xmath100 let @xmath101 be an @xmath102-tangle having internal discs @xmath103 with colors @xmath104 respectively ( or no internal discs of course ) .
a _ state _
@xmath105 on @xmath101 is a map from @xmath106 to @xmath107 such that @xmath108 for all @xmath109 where latexmath:[$\sigma obtained by reading the elements of @xmath107 induced at the marked points on the boundary of @xmath111 by the strings via the map @xmath105 , starting from the first marked point and moving clockwise .
this has been illustrated in figure [ fig : diagstate ] , where @xmath112 and @xmath113 are just the products @xmath114 and @xmath115 respectively , and are thus required to be the identity element .
it is a consequence that @xmath116 in the figure ) holds for the external disc .
let @xmath117 denote the states on @xmath101 .
in order to define the action @xmath118 of @xmath101 , it is enough to define the coefficient @xmath119 of @xmath120 in the linear expansion of @xmath121 where @xmath122 such that @xmath123 for @xmath109 . for this
, we choose a picture @xmath124 in the isotopy class of @xmath101 and then choose a simple path @xmath125 in @xmath126 $ ] starting from the @xmath72 of a disc @xmath127 is a point chosen on the boundary of @xmath127 strictly between the last and the first marked points , moving clockwise .
] of @xmath128 to that of @xmath111 for @xmath109 such that : \(i ) @xmath125 intersects the strings of @xmath124 transversally for @xmath129 , \(ii ) @xmath130 and @xmath131 intersects exactly at the @xmath72 of @xmath128 for @xmath132 .
+ note that any state @xmath105 on @xmath101 gives an element @xmath133 obtained by reading the elements of @xmath107 induced by @xmath105 at the crossings of the path @xmath125 and the strings along the direction of the path where @xmath134 ( necessarily even ) is the number of strings cut by @xmath125 .
define @xmath135 where @xmath136 and @xmath137 for @xmath138 , @xmath139 .
if there is no compatible state on @xmath101 , then we take the coefficient to be @xmath140 and if there is no internal disc in @xmath101 , then the scalar inside the sum is considered to be @xmath63 .
( note that @xmath141 depends only on @xmath142 and @xmath143 . )
we need to show first that the multi - linear map @xmath118 is well - defined .
two configurations of paths @xmath144 and @xmath145 in @xmath124 can be obtained from each other using a finite sequence of the following moves : + i. _ isotopy _
_ cap - sliding moves _ \iii .
_ disc - sliding moves _ \iv . _ rotation moves _ it is enough to check that the definition of the action is independent under each of the above moves .
invariance under isotopy moves are the easiest to check since @xmath146 for all @xmath147 . to see invariance under the remaining three moves , we show that @xmath148 for @xmath109 . for a cap - sliding move , note that the cap induces the same index at the two consecutive crossing with the path but after applying the @xmath95 map , the corresponding group elements cancel each other since they inverses of each other . for the disc - sliding ( resp .
, rotation moves ) , the invariance follows from the fact that @xmath149 ( resp . , @xmath150 )
. we will show next that the action is compatible with composition of tangles . let @xmath151 be an @xmath152-tangle containing the internal discs @xmath153 with colors @xmath154 and @xmath101 be an @xmath155-tangle containing internal discs @xmath156 with colors @xmath104 .
let @xmath157 and @xmath128 denote the external discs of @xmath151 and @xmath101 respectively .
we need to show that for all @xmath158 and @xmath159 where @xmath160 and @xmath161 @xmath162 the left - hand side of ( [ comp ] ) can be expanded as @xmath163 where @xmath164 s and @xmath125 s are paths in the tangles @xmath151 and @xmath101 respectively , required to define their actions . for the action of @xmath165 , we consider the paths @xmath164 for @xmath166 and @xmath167 for @xmath109 .
( strictly speaking , one has to disjointify the @xmath168-portion of the paths @xmath169 for different values of @xmath170 in order to define the action of @xmath165 . )
so , the right - hand side of ( [ comp ] ) becomes @xmath171 observe that @xmath172 is clearly in bijection with @xmath173 .
a bijection is obtained by sending @xmath174 to the state defined by @xmath105 ( resp .
@xmath175 ) on the @xmath151-part ( resp .
@xmath101-part ) of @xmath176 and the well - definedness of such a state is a consequence of the condition @xmath177 ; we denote this state by @xmath178 .
if these sets are empty , equation ( [ comp ] ) holds trivially since both sides have value @xmath140 .
let us assume that the sets are nonempty .
it is enough to prove for @xmath179 and @xmath180 such that @xmath181 for @xmath182 and @xmath183 for @xmath184 and @xmath185 we have @xmath186 we prove this in two cases .
+ * case 1 : * @xmath101 has no internal disc or closed loop , that is , @xmath101 is a temperley - lieb diagram .
then the right - hand side of equation ( [ comp2 ] ) is @xmath63 .
it remains to show @xmath187 .
this follows from the next lemma and the fact @xmath188 is a sequence of non - crossing matched pairings of indices from @xmath107 .
[ pair1 ] if @xmath189 , @xmath139 , @xmath190 and @xmath191 , then we have @xmath192 .
* proof : * note that + ( i ) @xmath193 for @xmath194 , and + ( ii ) @xmath195 for @xmath196 .
+ we compute then , @xmath197 [ pair2 ] in lemma [ pair1 ] , if @xmath143 is a sequence of indices with non - crossing matched pairings , then we can apply the lemma several times to reduce all the consecutive matched pairings to get @xmath198 .
* case 2 : * @xmath101 has at least one internal disc .
any unlabelled tangle @xmath101 can be expressed as composition of elementary annular tangles of four types as described in @xcite , namely , capping , cap - inclusion , left - inclusion and disc - inclusion tangles .
it is enough to prove equation ( [ comp2 ] ) for any tangle @xmath151 and compatible tangle @xmath101 in @xmath199 ( @xmath200 the set of all elementary tangles ) .
if @xmath101 is of capping or cap - inclusion type annular tangle , the proof directly follows from lemma [ pair1 ] and is left to the reader .
if @xmath101 is a left - inclusion annular tangle , equation ( [ comp2 ] ) is implied from the following lemma .
[ leftinc1 ] for all @xmath201 , @xmath202 and @xmath203 such that @xmath204 , we have @xmath205 where @xmath206 is the sequence of indices from @xmath207 in the reverse order . *
proof : * we rearrange the terms of the right - hand side of the identity in the lemma in the following way : @xmath208 let @xmath209 and @xmath210 . note that for @xmath211 , + @xmath212 since @xmath204 and @xmath213
thus the first product of the terms in the rearrangement vanishes . for the second product , if @xmath214 is odd , then @xmath215 where we used the defining equation ( [ defcocycle ] ) of the @xmath86-cocycle @xmath1 for the second equality . similarly ,
if @xmath214 is even , then @xmath216 thus , @xmath217 since @xmath204 and @xmath1 is normalized .
[ leftinc2 ] the proof of lemma [ leftinc1 ] also implies the following identity : @xmath218 now , suppose @xmath101 is a disc - inclusion tangle as shown in figure [ path ] . note that @xmath219 . without loss of generality
, we can assume that @xmath101 be given by the following picture in which we also indicate the paths @xmath220 and @xmath221 .
observe that since @xmath220 does not intersect any string , we have that @xmath222 is the empty multi - index .
so it is enough to prove @xmath223 let us denote @xmath224 by @xmath225 . since @xmath175 is a state , the following relations clearly follow from figure [ path ] : + ( i ) @xmath226 for @xmath227 and @xmath228 , + ( ii ) @xmath229 for @xmath230 , + ( iii ) @xmath231 for @xmath227 , + ( iv ) @xmath232 for @xmath230 .
+ we now express @xmath233 as a product of three terms with which we work separately .
@xmath234 first term : for @xmath227 , @xmath235 second term : for @xmath236 , @xmath237 third term : note that @xmath238 thus , for @xmath239 , @xmath240 combining the three terms , we get @xmath241 now , for @xmath236 , we compute @xmath242 hence , we obtain @xmath243 we can now proceed with the proof of equation ( [ discinc ] ) : @xmath244 where we applied lemma [ leftinc1 ] for the last identity .
this completes the proof that the action of tangles defined above is compatible with composition of tangles ( called _ naturality of composition _ in @xcite ) .
hence , @xmath68 is a planar algebra .
we will define next a @xmath72-_structure _ on @xmath245 . note that if @xmath246 , then @xmath247 if and only if @xmath248 .
extend the operation @xmath249 conjugate linearly to define a @xmath72-structure on @xmath250 .
clearly , this is an involution .
we need to check whether the action of tangles preserves @xmath72 , that is , @xmath251 .
it is enough to check this identity when @xmath101 has no internal discs or closed loops , and when @xmath101 is an elementary annular tangle . if @xmath101 has no internal discs or closed loops , then it is a temperley - lieb diagram and hence @xmath118 is the sum of all sequences of indices from @xmath107 which have non - crossing matched pairings where the position of the pairings are given by the numberings of the marked points on the boundary of @xmath101 which are connected by a string .
now , in the tangle @xmath252 , the @xmath253-th and the @xmath82-th marked points are connected by a string if and only if the @xmath253-th and the @xmath82-th marked points starting from the last point in @xmath101 reading anticlockwise , are connected .
so , @xmath254 is indeed the sum of all sequences featuring in the linear expansion of @xmath118 in the reverse order ( that is , applying @xmath249 ) . if @xmath101 is an elementary annular tangle of capping ( resp .
cap - inclusion ) type with @xmath253-th and @xmath255-th marked points of the internal ( resp .
external ) disc being connected by a string , then @xmath252 is also the same kind of elementary annular tangle but the ` capping ' occurs at the @xmath253-th and @xmath255-th marked points of the internal ( resp .
external ) disc starting from the last point reading anticlockwise .
the identity easily follows from this observation . if @xmath101 is an elementary tangle of left - inclusion type , then @xmath256 .
the identity will then follow from the next lemma .
[ lambda * ] if @xmath139 and @xmath257 such that @xmath247 , then @xmath258 .
* proof : * the proof is an immediate consequence of @xmath259 , which holds since @xmath247 .
lastly , if @xmath101 is an elementary tangle of disc - inclusion type given by figure [ path ] , then we need to show @xmath260 for @xmath261 with @xmath262 .
it is easy to verify that @xmath143 , @xmath207 and @xmath263 define a state on @xmath101 if and only if @xmath264 , @xmath206 and @xmath265 define the same on @xmath252 .
if they fail to define a state , then both sides are zero .
if they define a state , then the scalars appearing on both sides can be made equal by applying lemma [ lambda * ] . if @xmath1 is trivial , that is , a coboundary , then @xmath266 is isomorphic to example 2.7 in @xcite .
jones constructed this example of a planar algebra by considering a certain subspace of the tensor planar algebra ( tpa ) over the vector space with the indexing set @xmath107 as a basis .
he then showed that this subspace is closed under the tpa - action of tangles , thus showing that the subspace is indeed a planar algebra .
one can view our planar algebra @xmath68 as a subspace of the tpa in an obvious way but the action of planar tangles induced by tpa will not be the same as our action which involves the extra data of a @xmath86-cocycle .
it does not seem clear if our planar algebra @xmath68 can be viewed as a planar subalgebra of the tensor planar algebra over the vector space with basis @xmath107 if @xmath1 is nontrivial .
in this section we will compute the relative commutants of the diagonal subfactor associated to a @xmath0-kernel .
we will determine the filtered @xmath72-algebra structure , jones projections and the conditional expectations .
note that some of this can already be found in @xcite , @xcite , @xcite .
the main point here is that we are able to choose an appropriate basis of the higher relative commutants such that the action of planar tangles on these allows us to identify this concrete planar algebra with the abstract one defined in the previous section .
1em let @xmath267 be a ii@xmath2 factor , @xmath107 be a finite set and for @xmath189 , choose @xmath268 .
set @xmath269 where @xmath270 denotes the matrix algebra whose rows and columns are indexed by @xmath107 . as in the introduction , consider the _ diagonal subfactor _ @xmath271 given by @xmath272 that is , an element @xmath273 of @xmath267 sits in @xmath274 as a diagonal matrix whose @xmath275th diagonal is @xmath276 . if we have another collection of automorphisms @xmath277 for @xmath189 such that @xmath278 , for all @xmath189 ( up to a permutation of @xmath107 ) , then the associated diagonal subfactors are isomorphic .
it is therefore natural to associate a diagonal subfactor to a collection of elements @xmath279 , @xmath189 .
we consider the subgroup @xmath280 of @xmath281 , which can be viewed as a @xmath0-kernel in the obvious way , and choose a lift @xmath282 such that @xmath283 .
set @xmath284 for all @xmath189 .
consider the diagonal subfactor @xmath285 where the @xmath275th diagonal entry of an element of @xmath267 viewed in @xmath274 is twisted by the action of @xmath286 .
the index of this subfactor is @xmath287 .
set @xmath288 for @xmath99 .
we will often identify @xmath289 with @xmath290 for @xmath291 and @xmath292 .
now , @xmath293 is included in @xmath294 in the following way : @xmath295 for all @xmath296 and @xmath297 where @xmath298 is defined as : @xmath299 it is easy to check ( see @xcite , @xcite ) that @xmath300 is isomorphic to the jones tower of ii@xmath2 factors associated to @xmath11 where the jones projections and conditional expectations are given by : for all @xmath296 , @xmath302 , @xmath303 .
moreover , @xmath304 forms a pimsner - popa basis of @xmath274 over @xmath267 .
this basis will be used to write the conditional expectation of commutant of @xmath267 onto the commutant of @xmath274 ( see @xcite ) .
thus , for @xmath296 , if @xmath311 , then @xmath312 and @xmath313 which implies @xmath314 . similarly , if there exist @xmath296 such that @xmath314 , then @xmath315 where @xmath316 satisfies @xmath317 .
thus , @xmath318 the elements in this set do not yet form a basis since the unitary @xmath319 can be modified by a scalar of absolute value @xmath63 . to get a good basis of @xmath305
we need to choose @xmath319 in such a way that the planar algebra associated to @xmath11 can easily be identified with the abstract one defined in section [ plnalg ] .
@xmath321 such that @xmath322 whenever either of @xmath18 or @xmath89 is @xmath73 . for @xmath323 , define @xmath324 and set @xmath325 .
next , we prove a several useful lemmas involving @xmath326 .
the first lemma motivates our choice of the basis .
* proof : * first we expand @xmath334 and @xmath335 into products of unitaries arising from the definition of @xmath326 , and then we consider that the product of the @xmath336-th unitary of @xmath334 from the right and @xmath336-th unitary of @xmath335 from the left , that is , @xmath337 for @xmath338 . * proof : * using an argument similar to the proof of lemma [ v * ] , one can prove that the product of the @xmath336-th unitary of @xmath342 from the right and the @xmath336-th unitary of @xmath343 from the left , is @xmath63 for @xmath344 .
again , for @xmath345 , @xmath346 thus , @xmath347 + by applying lemma [ unichoice ] , we see that the set @xmath348 is a basis for @xmath305 .
we will use this basis to establish an isomorphism between the planar algebra associated to @xmath349 and the abstract planar algebra defined in section [ plnalg ] .
let @xmath350 be the @xmath86-cocycle associated to @xmath0 , that is , for all @xmath351 we have @xmath352 as before , this is a consequence of associativity of the multiplication in @xmath0 .
let us recall the following well - known fact about isomporphisms of two planar algebras which will be used in the next theorem ( @xcite ) .
let @xmath362 and @xmath363 be two planar algebras . then @xmath364 ( as planar algebras ) if and only if there exist a vector space isomorphism @xmath365 such that : [ mainiso ] the planar algebra @xmath367 associated to the diagonal subfactor obtained from a ii@xmath2 factor @xmath267 and a finite collection of automorphisms @xmath368 for @xmath189 , is isomorphic to @xmath369 where @xmath370 \in { \hbox{out\,}}n$ ] for all @xmath189 , and @xmath1 is the normalized @xmath86-cocycle associated to @xmath371 as above . define the map @xmath375 by first defining it on basis elements as @xmath376 for all @xmath296 such that @xmath314 , and then extending it linearly .
clearly , @xmath377 is a vector space isomorphism . we will show that @xmath377 is @xmath72-planar algebra isomorphism .
we make first the following observation : for @xmath189 , @xmath308 and @xmath191 , we have the identity @xmath378 .
the proof is similar to that of lemma [ pair1 ] .
thus , if @xmath143 is a sequence of indices with non - crossing matched pairings , then using this identity several times to reduce all consecutive matched pairings , we get @xmath379 .
\(b ) _ @xmath377 preserves jones projection tangles _ : by theorem 4.2.1 in @xcite , the @xmath82-th jones projection tangle @xmath383 acts as @xmath384 . since @xmath385 is a sequence of indices with non - crossing matched pairings , by the above note @xmath386 .
so , @xmath387 .
\(c ) _ @xmath377 preserves the action of conditional expectation tangle _ : by theorem 4.2.1 in @xcite , the action of conditional expectation tangle @xmath388 is given by @xmath389 for all @xmath296 , @xmath302 such that @xmath390 .
from the action of tangles defined in section [ plnalg ] , it is easy to check @xmath391 .
\(d ) _ @xmath377 preserves @xmath72 _ : applying @xmath72 on a basis element of @xmath392 , we get @xmath393 for @xmath296 such that @xmath394 .
note that we used lemma [ unichoice ] for the third equality and lemma [ v * ] for the fourth one .
@xmath377 preserves multiplication _
: suppose @xmath395 such that @xmath396 .
then , @xmath397 on the other hand , one can easily deduce from the action of the multiplication tangle in @xmath3 that @xmath398 .
\(f ) _ @xmath377 preserves the action of the left conditional expectation tangle _ : by theorem 4.2.1 in @xcite , the action of the left conditional expectation tangle @xmath399 is given by @xmath400 . using the basis of @xmath274 over @xmath267 mentioned before , the conditional expectation onto @xmath401 can be expressed as ( see @xcite ) @xmath402 for @xmath403 .
hence , for @xmath404 , @xmath405 and @xmath406 such that @xmath407 , we have @xmath408 on the other hand , one can easily check that the action of @xmath399 is given by @xmath409 given a group @xmath0 generated by a finite collection @xmath14 for @xmath189 and given a normalized @xmath86-cocycle @xmath410 , there exists a hyperfinite subfactor whose associated planar algebra is isomorphic to @xmath411 . note that the isomorphism @xmath377 in the proof of theorem [ mainiso ] uses the @xmath86-cocycle @xmath1 only in the step involving the conditional expectation onto the commutant of @xmath274 .
in particular , the filtered @xmath72-algebra structure does not involve @xmath1 . analyzing the filtered @xmath72-algebra structure of our planar algebra
, one can easily find that the principal graph @xmath412 of @xmath11 is a cayley - like graph .
more precisely , if @xmath413 for @xmath414 , and @xmath415 , then @xmath416 denotes the set of vertices of @xmath412 which are at a distance @xmath82 from the distinguished vertex for @xmath414 , and @xmath417 .
the number of edges between @xmath418 and @xmath419 is @xmath420 ( resp .
@xmath421 ) if @xmath82 is odd ( resp . even ) . note that this is well - known .
the most elegant feature of the planar algebra @xmath411 is that the distinguished basis forms the ` loop - basis ' of the filtered @xmath72-algebra arising from paths on the principal graph .
note that the @xmath86-cocycle @xmath1 does not enter in the definition of the actions of multiplication , inclusion and unit tangles ( defined in section [ plnalg ] ) or in the @xmath72-operation .
of course , we found the abstract planar algebra by first computing the action of tangles on the relative commutants .
we then deduced from it an abstract prescription of the planar algebra associated to a @xmath0-kernel and a @xmath86-cocycle , which is the one presented in section 3 .
the path algebra associated to the principal graph can always be used to obtain a description of the filtered @xmath72-algebra structure of a subfactor planar algebra ( see for instance @xcite ) .
the extra information encoded in the planar algebra which the principal graph can not provide , is the action of the left conditional expectation tangle ( or equivalently , the rotation tangle or the left - inclusion tangle with an even number of strings ) .
the main issue in this paper was the choice of the unitaries @xmath422 satisfying the conclusion of lemma [ unichoice ] in such a way that the basis elements of @xmath423 correspond to loops on the principal graph with the correspondence extending to a filtered @xmath72-algebra isomorphism . note that this choice of @xmath422 is unique up to a scalar in @xmath76 .
it is a delicate choice in the sense that another choice would very likely make the @xmath86-cocycle @xmath1 appear in the description of the filtered @xmath72-algebra structure , whereas @xmath1 does not feature in the path algebra on the principal graph @xmath412 .
we will prove next a converse of theorem [ mainiso ] .
we will refer to the abstract planar algebra defined in section [ plnalg ] as _ diagonal planar algebra_. [ conv ] any finite index extremal subfactor @xmath11 whose standard invariant is given by a diagonal planar algebra is a diagonal subfactor .
moreover , if the associated group and its generators of the diagonal planar algebra is given by @xmath0 and @xmath424 repectively , then for every @xmath425 , there exists @xmath426 , @xmath189 , such that : * proof : * let @xmath3 be the planar algebra associated to @xmath11 , @xmath432 be a diagonal planar algebra associated to @xmath0 and @xmath424 , and @xmath433 a @xmath72-planar algebra isomorphism . + _ setting up matrix units : _
+ for all @xmath434 such that @xmath435 , set @xmath436 .
note that @xmath437 .
so , to create other off - diagonal matrix units @xmath438 , we partition @xmath439 such that : \(ii ) @xmath441 for some @xmath442 . + for each @xmath443 , choose @xmath444 and partial isometry @xmath445 such that @xmath446 and @xmath447 .
( note that @xmath448 and @xmath449 have the same trace @xmath450 ) .
extend @xmath451 by defining @xmath452 by @xmath453 it is completely routine to check ( using properties of partial isometry and action of multiplication tangle in @xmath432 ) that ( i ) @xmath451 is well - defined , ( ii ) @xmath454 , and ( iii ) @xmath455 .
+ _ finding automorphisms : _ + using extremality of @xmath11 , for each @xmath189 , we get @xmath456=1 & \rightarrow n q^i_i = q^i_i m q^i_i\\ & \rightarrow n q^i_j \subset q^i_i m q^j_j \subset q^i_j n \text { and }
n q^ i_j \supset q^i_i m q^j_j \supset q^i_j n\\ & \rightarrow n q^i_j = q^i_i m q^j_j = q^i_j n \rightarrow m \cong \underset { i , j \in i}{\bigoplus } q^i_j n\end{aligned}\ ] ] where we use @xmath457=0 $ ] in the second implication . for each @xmath189 , define @xmath458 by @xmath459 for all @xmath297 . since @xmath267 is a factor and @xmath460=0 $ ] for @xmath461 , @xmath429 is well - defined and injective ; surjectivity follows from @xmath462 .
linearity and homomorphism property of @xmath429 follow immediately , and we also have the identity @xmath463 . to show @xmath464 , note that @xmath465 using this relation , it is easy to show that @xmath429 is @xmath72-preserving .
two other easy consequences are @xmath466 if and only if @xmath435 and @xmath467 .
+ _ structure of diagonal subfactor : _ + define : _ matrix units for the tower of basic construction : _
+ let @xmath294 denotes the ii@xmath2 factor obtained from @xmath11 by iterating the basic construction @xmath82 times .
we will first define @xmath479 for @xmath296 and @xmath480 satisfying : suppose we have defined such @xmath492 s for all @xmath493 and @xmath494 .
now , for @xmath495 and @xmath296 , set @xmath496 for some @xmath497 , where @xmath498 is the element in @xmath294 given by the @xmath82-th jones projection tangle . to show that the definition of @xmath499 is independent of @xmath500 , observe that @xmath501 for all @xmath502 . properties ( i ) and ( v ) hold trivially . for
( ii ) , note that @xmath503 it remains to establish property ( iii ) . now , for @xmath511 , @xmath189 and @xmath497 , @xmath512 since @xmath513 ( by property ( ii ) ) and @xmath514 is a minimal projection of @xmath515 , therefore @xmath516 .
the proof for the initial case of @xmath517 is similar and is left to reader .
+ _ correspondence between relations satisfied by @xmath429 s and @xmath14 s : _ + in this part , we will prove that for @xmath482 , @xmath518 if and only if @xmath519 .
following the construction of the isomorphism @xmath520 between @xmath274 and @xmath521 , one can define an isomorphism @xmath522 as follows : @xmath523 where @xmath524 is the map given by the relation @xmath525 thus , @xmath526 , for @xmath414 .
note that @xmath527 .
conversely , if @xmath535 , that is , @xmath536 for some @xmath537 , then @xmath538 .
now , @xmath539 implies @xmath540 which is a contradiction .
hence , @xmath541 .
+ _ the group generated by @xmath429 s : _ + let @xmath542_{{\hbox{out\,}}n } : i \in i \rangle \leq { { \hbox{out\,}}n}$ ] , @xmath543 , @xmath544 .
define maps @xmath545 and @xmath546 by @xmath547 where @xmath548 for @xmath549 .
define @xmath550 by @xmath551 . for @xmath552 to be an isomorphism , it is enough to show @xmath552 is well - defined and injective .
suppose the map @xmath553 sends @xmath554 ( resp .
@xmath555 ) to @xmath556 ( resp .
@xmath557 ) . extend @xmath558 to @xmath559 entrywise .
note that @xmath560 and @xmath561 .
this implies @xmath562 hence , @xmath563 and @xmath564 are isomorphic .
d. bisch , _
bimodules , higher relative commutants and the fusion algebra associated to a subfactor _ , operator algebras and their applications , 13 - 63 , fields inst .
commun . , 13 , amer .
soc . , providence , ri , 1997 .
a. ocneanu , _ quantized groups , string algebras and galois theory for algebras _ , operator algebras and applications , vol .
2 , 119172 , london math . soc .
lecture note ser .
, 136 , cambridge univ . press , cambridge , 1988 . | there is a natural construction which associates to a finitely generated , countable , discrete group @xmath0 and a 3-cocycle @xmath1 of @xmath0 an inclusion of ii@xmath2 factors , the so - called diagonal subfactors ( with cocycle ) . in the case
when the cocycle is trivial these subfactors are well studied and their standard invariant ( or planar algebra ) is known .
we give a description of the planar algebra of these subfactors when a cocycle is present .
the action of jones planar operad involves the 3-cocycle @xmath1 explicitly and some interesting identities for 3-cocycles appear when naturality of the action is verified . |
gillessen et al . ( 2012 ) reported the detection of a very red infrared object approaching the supermassive black hole ( smbh ) at the galactic center on an orbit with a predicted closest approach of only 3000 times the radius of the event horizon .
their detection of brackett - gamma emission led them to interpret it as a dusty , @xmath4 earth mass ( @xmath5 ) gas cloud .
if this object ( g2 ) is indeed a gas cloud , it would be disrupted by the tidal forces of the smbh during closest approach and some of it would be accreted ( @xcite ; @xcite ; @xcite ) . consequently
, g2 has generated tremendous interest since it can be followed through the predicted accretion event and , possibly , provide new insight into accretion physics ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . the identification of g2 as a pure gas cloud , however , is controversial . in the pure disrupting gas cloud scenario ,
g2 is required to have formed relatively recently ( @xmath11995 - 2000 ) , close to the moment when it first became detectable with new adaptive optics technologies , and at a position that is well inside the formal apoapse position ( @xcite ) .
since this makes the unusual demand that g2 has been fortuitously observed during the exact decade of its entire existence , many alternative models containing a central stellar source have been proposed ( e.g. , @xcite , @xcite , @xcite , @xcite , @xcite , ; see also @xcite ) .
the presence of a central star would allow g2 to survive its closest approach and would not demand a recent formation event .
it would also , in most scenarios , reduce the amount of gas expected to be accreted onto the central black hole following closest approach ( e.g. , @xcite ) .
observationally , very little is known about g2 .
it has been imaged as a very red point source in the near - infrared with adaptive optics systems , where it has been detected at wavelengths of 3 - 5 @xmath0 m ( _ l _ - m ) , but is very faint at 2 @xmath0 m ( @xmath6 mag , @xcite , @xcite ; see also ) .
spectroscopically , it is a faint emission - line object , best detected in br-@xmath3 ( @xcite , @xcite ) , which shows a slightly elongated , rather compact core and some low surface brightness emission that appears to form leading and trailing tidal tails .
the interpretation of exactly how much of the low surface brightness ionized gas is associated with g2 is complicated by the highly structured gas streams , also seen in emission , that abound in the projected vicinity of sgr a * ( @xcite ; @xcite , @xcite ) . with
a predicted closest approach having occurred in spring 2014 , the pure gas model can now be tested . in this model
the _ l_-band component of g2 is interpreted as dust embedded in the gas .
it should follow the spatial evolution of the br-@xmath3 component , and thus , lose its compactness at _ l_. in this paper , we present new _ l _ imaging observations of g2 at the predicted moment of closest approach and during the following few months .
new near - infrared images of g2 were obtained on 2014 march 20 , may 11 , july 3 , august 4 and 5 , 2006 may 21 , and 2005 july 30 using the lgsao system ( @xcite , @xcite ) and nirc2 ( p.i .
k. matthews ) at the w. m. keck observatory as part of our long - term study of the central supermassive black hole and its environs ( @xcite ) . at the time of our observations ,
g2 and sgra * , the emissive source associated with the black hole , are expected to be spatially unresolved from each other in our nir observations .
the observational set - up used during these measurements enables us to disentangle the emission of g2 from that of sgr a * , which is a highly variable source at infrared wavelengths ( @xcite , @xcite , , @xcite ) . each night
, images were obtained in nirc2 s narrow field mode ( 10 mas / pix ) , interleaving observations through the _ k _ [ 2.1 @xmath0 m ] and _ l _ [ 3.8 @xmath0 m ] broadband filters .
individual exposure times of 28 sec ( 10 coadds @xmath7 2.8 sec ) and 30 sec ( 60 coadds @xmath7 0.5 sec ) at _ k _ and _ l _ , respectively , resulted in a duty cycle time of 134 sec for the two - wavelength cycle . to optimize the efficiency of operations and performance , the images on g2 were obtained in a stare - mode described in @xcite and
l _ sky images were taken over a range of rotator angles according to @xcite .
table 1 summarizes all of the new data collected , and the historic data used in this analysis .
all the images were analyzed with standard image reduction techniques , as laid out in detail for our group s earlier work ( e.g. , @xcite , @xcite , @xcite ) .
each image was sky - subtracted , flat - fielded , bad - pixel and image - distortion corrected .
all the individual images from a given night and at the same wavelength were then averaged to obtain a deep combined map for each night , and were de - convolved frame by frame with a lucy - richardson algorithm ( @xcite ) to obtain time - series of flux density for both bands . to determine statistical uncertainties beyond the formal fitting errors , we additionally created three sub - maps from three simultaneous subsets of frames . since g2 and sgr a * are spatially unresolved , we disentangled the measurement of their brightness spectrally . at
k _ , the source at g2 s predicted location is assumed to be dominated by sgr a * 20 ; @xcite , , @xcite ) . ] . at _
l _ , the source is expected to be the combination of g2 and sgr a*. while sgr a * s brightness is highly variable , our interleaved _
k _ measurements of sgr a * , and the well measured and constant _ k_-_l _
color for sgr a * ( @xcite ; @xcite ) allow sgr a * s _ l _ flux to be estimated and removed .
the details of this approach are described below .
photometric estimates of g2 and sgr a * are extracted using two different approaches : point - spread - function ( psf ) fitting and aperture photometry on deconvolved images . in the first method , _ starfinder _ , a psf - fitting program ( @xcite ) ,
is used to identify and characterize point sources in each combined map , and in the corresponding sub - maps .
this resulted in astrometric and photometric values for sgr a * in _ k _ and for the unresolved g2/sgr a * source in _
l_. in the second method , aperture photometry is carried out on the individual deconvolved frames .
the psf for the deconvolution process is obtained by running _
starfinder _ on individual frames .
the restoring beam had the fwhm half the resolution at each wavelength , and the aperture diameter was 60 and 120 mas for _ k _ and _ l _ , respectively .
the aperture photometry values are obtained at each wavelength by averaging over the resulting time - series of flux densities .
while this second procedure was intended to assist with any confusion with additional sources that might be near g2 or sgr a * , the photometry in both approaches agree well with each other .
the final values are the average of the two approaches , and the differences are treated as an additional , albeit negligible , source of uncertainty .
photometric calibration was accomplished with non - variable calibrators ( @xcite , @xcite ) in the immediate surrounding of sgr a*. the exact set of stars used is identical to that in @xcite for psf - fitting results , and to that in @xcite and witzel et al .
( in prep . ) for the results from aperture photometry on deconvolved images .
the different sets were chosen to guarantee comparability between the results here and in earlier measurements , but both establish consistent zero - points . to infer the flux density of sgr a * in _
l _ from our _ k _ measurements , we applied the known spectral index and the extinction values in the infrared .
sgr a * s spectral index has been shown to be constant with brightness and constant in time to within @xmath80.1 ( @xcite ) , and we adopt a value of @xmath9 , which includes the systematics of the extinction correction ( @xcite , @xcite ) and which is based on the same photometric calibration applied here . for this analysis , an extinction law published by , and zero - points from @xcite were applied .
we subtract the inferred _
l_-band , reddening - corrected flux density for sgr a * from the reddening - corrected value for the combined g2 + sgr a * point source .
the final g2 brightness values are reported without reddening correction , for ease of comparison with earlier observed photometry .
we report the photometric values for each individual observation in 2014 as well as the variance - weighted average .
the uncertainties for the individual nights incorporate the statistical errors of the zero - point calibration and the sub - map photometry .
the average value additionally includes the systematic differences between the psf - fitting and the deconvolution method , the error of the spectral index , and the uncertainties of the extinction values . an upper limit on g2 s size for each 2014 epoch
is obtained from radial profiles of g2 s emission . to create the radial profile from the epochal images
the inferred _
l _ flux of sgr a * and the closest neighboring stars ( s0 - 2 , @xmath10 ; s0 - 8 , @xmath11 ) are removed through psf subtraction . the positional and brightness information for s0 - 2 and s0 - 8 are obtained from the _ starfinder _ analysis discussed above . the exact location of sgr a * is best inferred from a dataset in which sgr a * is particularly bright . since sgr a * was faint ( @xmath12 2 mjy at _
dereddened ) during all of the _
k_/_l _ interleaved datasets , we make use of the _
k_-band data obtained on 2014 august 5 when sgr a * was bright ( 7 mjy , dereddened ) .
this allows for a positional accuracy for sgr a * of @xmath1 1 mas , which is dominated by the reference frame alignment uncertainty between the dates .
the radial profile of g2 is compared with the psf and the reduced @xmath13-value is calculated . additionally , to derive an upper limit for the size of g2 the psf is convolved with a gaussian .
the upper limit is the fwhm of the gaussian that corresponds to a @xmath13 probability of @xmath14 .
the centroid of g2 with sgra * and nearby sources removed gives the position of g2 , which can be determined to within 10 mas ( uncertainty dominated by dust contributions ; see @xcite ) .
fig . [ klmaos ] shows that g2 is easily detected in our _ l _ images .
the following key results emerge from the analysis of these images : * _ g2 has survived its closest approach to the central black hole as a compact , unresolved source at _
l__. our observations took place very near to g2 s closest approach of 2014 march 16 ( @xmath15 2 months ) and extended well after this date ( 5 months ) .
all images reveal g2 and there is no evidence for this object to be extended at _
l _ in any 2014 epoch , as implied by the @xmath13 values in tab .
we place a @xmath16 upper limit on the diameter of g2 s emission at _
l _ of @xmath17 mas , which correspond to 260 au at a distance of @xmath18 kpc ( see tab . [ tab2 ] and
[ s]a , b ) . * _ g2
s l brightness measurements are consistent with those made over the past decade .
_ as the inset to fig .
[ s]b displays , all our g2 _
l _ brightness estimates from 2014 are in agreement with one another .
furthermore , the 2014 average value of @xmath19 is also consistent with earlier measurements ( fig .
[ s]b ) . tab .
[ tab2 ] summarizes the quantitative results for g2 s brightness . * _
g2 s motion continues to be consistent with a keplerian orbit model . _
the _ l _ position of g2 in 2014 is consistent with the predictions of our orbital model ( @xcite ) . within these predictions ,
the new position favors orbital solutions with shorter periods ( @xmath20 yr ) and reduces the uncertainty in the lower range of the periapse distance @xmath21 by a factor of 3 ( @xmath21 = 215 @xmath15 30 au ) .
our _ l _ measurements present a very different view of g2 from what has been seen through the br-@xmath3 line emission measurements .
the br-@xmath3 line emission traces hot gas ( @xmath22k ; @xcite ) that appears to be _ externally _ heated by ionizing photons from massive stars in the vicinity of g2 . in br-@xmath3
, g2 shows clear evidence of tidal interaction with the black hole ( @xcite , @xcite , @xcite ) .
this interaction and its evolution have been seen both in the increasing linewidth associated with the spatially compact ` head ' as well as the increasing extent of the low surface brightness tail(s ) associated with g2 . however , it is important to note that the br-@xmath3 line emission measurements imply only that _ some _ gas associated with g2 has a size that exceeds its tidal radius . unlike br-@xmath3 , the _ l
_ emission remains spatially unresolved , continues to follow a well - defined keplerian orbit , and is constant in brightness .
this allows us to rule out a pure gas cloud model as such a gas cloud model predicts that g2 s brightness and size should undergo substantial changes ( e.g. , @xcite , @xcite ) . instead , g2 s
emission appears to be coming from an optically thick dust shell surrounding an underlying star .
two lines of evidence point to this conclusion .
* the _ l _ flux has been invariant at @xmath23 mjy ( reddening - corrected , see tab .
[ tab2 ] ) since 2005 , in spite of a factor of 10 change in distance from the central black hole . * in 2004 , g2 had an _ l - m _ color of @xmath24 ( reddening - corrected , @xcite ) , which corresponds to a blackbody temperature of @xmath1 560 k. * no _ k _ detection has been made in any epoch .
* the _ l _ emission is much more compact than the emission in br-@xmath3 , and thus originates in a different region .
we conclude that this is most likely explained by optically thick dust that has had constant temperature and size over the past decade and is _ internally _ heated . modeling the dust emission as a blackbody results in an inferred luminosity @xmath25 , which can easily be generated by a 2 @xmath26 main - sequence star or a somewhat lower mass post- or pre - main - sequence star that is temporally very close to the main - sequence ( see fig . [
model]a ) ; higher mass stars are ruled out with the inferred luminosity .
the challenge is to explain the unusually large size inferred from the blackbody .
this size is @xmath12.6 au and is @xmath27 times larger than the photospheric size of 30 @xmath28 stars .
several scenarios can account for such a large size , but can be excluded based on other properties . a single young star surrounded by a protoplanetary disk , as proposed by @xcite , would have to be observed edge - on , which is rather unlikely .
the corresponding face - on systems would not appear as such highly obscured objects .
a common envelope surrounding a binary system containing a giant star would be too luminous . as we had suggested in @xcite and
was followed up by @xcite , a binary merger is a natural model for g2 . as fig .
[ model]b shows , g2 has a radius in the range calculated for several putative stellar merger products (; also and @xcite ) .
a second line of evidence that g2 indeed is a @xmath1 2 @xmath26 merger product comes from tidal radius arguments .
the constancy of the _ l _ flux and the compactness of g2 lead us to conclude that the tidal radius of the source , which is proportional to its distance from the black hole , has not become smaller than the source size for most of the time . for a black hole mass of @xmath29 ,
the tidal radius is @xmath30 where @xmath31 is the 3d - distance of g2 from the black hole .
[ model]b shows the time development of the tidal radius @xmath32 . for a mass of @xmath12 @xmath26
the derived _
l _ size of g2 does not show any tidal interaction with the black hole , except possibly near periapse passage .
this is consistent with g2 s _
l _ size and photometry not evolving during its approach to the black hole .
we note that the mass of g2 , in the case in which it is indeed a merger product , can be different from the main - sequence star mass assumed here . however , the tidal radius is only weakly dependent on the mass ( a factor 10 in mass corresponds to a factor @xmath33 in tidal radius ) . in this picture ,
in which g2 s _ l _ emission originates within the tidal radius , gas and dust beyond the tidal radius are removed by the tidal forces , creating extended optically thin tidal tails . while this extended emission is faint at _ l _ , it is competitive with the central source at br-@xmath3 .
thus , the br-@xmath3 emission is a by - product of the tidal interaction whereas the _
l _ emission traces the properties of the merger product .
the interpretation of g2 as a merger product suggests that it will eventually look like a typical member of the s - star cluster after the extended atmosphere contracts on a kelvin - helmholtz time scale .
the environs of the central supermassive black hole might be particular conducive to mergers as close binary stars in orbit around the smbh could experience kozai oscillations that increase the eccentricity of their orbits ( @xcite and references therein ; @xcite ) .
if this mechanism is indeed at work , it could dramatically increase the rate of stellar mergers .
the plausibility of such binary interactions contributing to the central stellar population relies on the poorly known distribution and fraction of stars in binary systems at the gc ( e.g. , @xcite ) .
further investigations of the stellar dynamics are underway to address the question whether it is possible that an important fraction of the s - stars has resulted from such mergers .
we thank the anonymous referee .
support for this work was provided by nsf grant ast-0909218 and ast-1412615 , the levine- leichtman family foundation , the preston family graduate fellowship ( held by ab ) , and the janet merrot student travel awards .
the w. m. keck observatory is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration .
the observatory was made possible by the generous financial support of the w. m. keck foundation .
llllllllll & & & & & & & & & + & & & & obs .
photometry & + & & & & & & & & & + date & @xmath34 & @xmath35 & @xmath36 & sgr a * & sgr a * & g2 + sgr a * & g2 & @xmath13/dof & size limit + & _ k_/_l _ & _ k_/_l _ & _ k_/_l _ & @xmath37 [ mjy ] & @xmath38 [ mjy ] & @xmath38 [ mjy ] & @xmath39 & & au + & & & & measured & inferred & measured & inferred & & + & & & & & & & & & + + & & & & & & & & & + 2014 march 20 & 22/21 & 67/91 & 0.21/0.44 & 1.1 @xmath15 0.1 & 1.5 @xmath15 0.2 & 3.7 @xmath15 0.4 & 13.88 @xmath15 0.16&0.48 & @xmath12370 + 2014 may 11 & 9/9 & 67/90 & 0.21/0.44 & 1.5 @xmath15 0.2 & 2.2 @xmath15 0.2 & 4.7 @xmath15 0.4 & 13.70 @xmath15 0.14 & 1.08 & @xmath12350 + 2014 july 3 & 64/64 & 70/108 & 0.19/0.32 & 1.6 @xmath15 0.3 & 2.3 @xmath15 0.4 & 4.9 @xmath15 0.5 & 13.67 @xmath15 0.23&1.14 & @xmath12380 + 2014 august 4 & 28/28 & 63/92 & 0.20/0.42 & 1.8 @xmath15 0.2 & 2.5 @xmath15 0.4 & 4.6 @xmath15 0.6 & 13.92 @xmath15 0.23&0.65 & @xmath12260 + & & & & & & & & & + average & & & & & & & 13.8 @xmath15 0.2 & & + & & & & & & & & & + & & & & & & & & & + + & & & & & & & & & + 2005 july 30 & -/56 & -/81 & -/0.66 & & & & 13.96 @xmath15 0.05 & & + 2006 may 21 & -/19 & -/82 & -/0.56 & & & & 14.05 @xmath15 0.10 & & + 2009 july 22 & 4/- & -/86 & -/0.46 & & & & 14.1 @xmath15 0.4 & & + 2012 july 20 - 23 & 1314/- & -/91 & -/0.48 & & & & 13.9 @xmath15 0.3 & & + & & & & & & & & & + 2014 august 5 & 127/- & 57/- & 0.26/- & 7.0 @xmath40 & & & & & + & & & & & & & & & + and @xmath4 m@xmath41 and evolve them according to the stellar evolution code sse ( @xcite ) .
over - plotted is the luminosity of g2 as measured from _
l _ emission .
b : tidal radius of g2 as a function of time assuming as mass of @xmath12 solar masses , and the radius of g2 derived for a optically thick blackbody . for comparison
we show the _
l _ radius limit derived directly from our observations and the radius of a model binary star merger . ] | we report new observations of the galactic center source g2 from the w. m. keck observatory .
g2 is a dusty red object associated with gas that shows tidal interactions as it nears closest approach with the galaxy s central black hole .
our observations , conducted as g2 passed through periapse , were designed to test the proposal that g2 is a 3 earth mass gas cloud .
such a cloud should be tidally disrupted during periapse passage .
the data were obtained using the keck ii laser guide star adaptive optics system ( lgsao ) and the facility near - infrared camera ( nirc2 ) through the _ k _ [ 2.1 @xmath0 m ] and _ l _ [ 3.8 @xmath0 m ] broadband filters .
several results emerge from these observations : 1 ) g2 has survived its closest approach to the black hole as a compact , unresolved source at _ l _ ; 2 ) g2 s _ l _ brightness measurements are consistent with those over the last decade ; 3 ) g2 s motion continues to be consistent with a keplerian model .
these results rule out g2 as a pure gas cloud and imply that g2 has a central star .
this star has a luminosity of @xmath130 @xmath2 and is surrounded by a large ( @xmath12.6 au ) optically thick dust shell .
the differences between the _ l _ and br-@xmath3 observations can be understood with a model in which _
l _ and br-@xmath3 emission arises primarily from internal and external heating , respectively .
we suggest that g2 is a binary star merger product and will ultimately appear similar to the b - stars that are tightly clustered around the black hole ( the so - called s - star cluster ) . |
generalized @xmath2-gons were introduced by tits in a famous work on triality @xcite of 1959 , in order to propose an axiomatic and combinatorial treatment for semisimple algebraic groups ( including chevalley groups and groups of lie type ) of relative rank @xmath0 .
they are the central rank @xmath0 incidence geometries , and the atoms of the more general `` tits - buildings . ''
if the number of elements of a generalized @xmath2-gon is finite , a celebrated result of feit and higman @xcite guarantees that @xmath2 is restricted to the set @xmath3 .
note that projective planes are nothing else than generalized @xmath4-gons .
generalized @xmath5-gons are also called _ generalized quadrangles _ , and certainly in the finite case , they are considered as being the main players in the class of generalized @xmath2-gons , and one of the most studied types of incidence geometries . the most fruitful way to construct finite generalized quadrangles is through a now standard group coset geometry construction , in which a group @xmath6 provided with certain sets of subgroups @xmath7 @xmath8 and satisfying some strong intersection properties , is used to represent a generalized quadrangle .
such a system of groups @xmath9 is called a _ kantor family _ for @xmath6 , and the defining properties are as follows . * for some @xmath10 , @xmath11 , @xmath12 , each @xmath13 has order @xmath14 and each @xmath15 has order @xmath16 . * for each @xmath17 ( in @xmath18 ) , we have @xmath19 . * for distinct @xmath20 and @xmath21 ( in @xmath18 )
, we have @xmath22 . * for distinct @xmath17 and @xmath23 ( in @xmath18 ) , we have @xmath24 .
generic points of the quadrangle are elements of the group , generic lines are left ( or right | the choice does not change the isomorphism class of the geometry ) cosets of the subgroups of type @xmath13 , and some other special points and lines exist .
a fundamental feature of this construction is that @xmath6 naturally acts as an automorphism group of the geometry , sharply transitively on the generic points .
see @xcite for the details . call generalized quadrangles with such a group coset representation `` elation generalized quadrangles '' ( `` egqs '' ) ; the group @xmath6 is the _ elation group _ , and there is a special point @xmath25 through which all the lines are fixed by @xmath6 ( the `` elation point '' ) . in the long literature of generalized quadrangles , it has appeared that one special type of egq plays a central role .
such egqs are called _ skew translation generalized quadrangles _ ( `` stgqs '' ) , and they are specializations of egqs in the sense that @xmath25 satisfies an additional combinatorial property called regularity . " except for the quadrangles associated to hermitian varieties in @xmath5-dimensional projective space , all known classes of generalized quadrangles are stgqs , up to a combination of duality and payne - integration .
this observation strongly motivates the necessity to understand stgqs , and perhaps aiming for classification results in every which way . only very partially , mostly in an influential paper by payne @xcite ,
such results have been obtained for particular types of stgqs up till recently | see also [ hpmgq ] .
this classifcation program is the main purpose of this note .
in fact , more precisely , we want to understand the category @xmath26 , where objects are triples @xmath27 , with @xmath28 an stgq , and @xmath29 a regular elation point with respect to an elation group @xmath6 .
morphisms are natural : if @xmath30 and @xmath31 , then @xmath32 consists of morhisms @xmath33 which map @xmath29 to @xmath34 and @xmath6 to @xmath35 ( the latter meaning that for @xmath36 , @xmath37 ) . for now , we imagine the stgqs being finite ( although the infinite case should eventually be handled as well ) .
several subtleties arise : * _ one _ fixed stgq @xmath28 could have _ several _ ( collinear ) elation points @xmath29 and @xmath34 ( if the points would be non - collinear , a classification of such stgqs is known @xcite ) - it can be easily shown however that there _ always _ exists automorphisms @xmath38 of @xmath28 mapping @xmath29 to @xmath34 ; * even if one fixes the couple @xmath39 , by recent work of rostermundt and independently the author of this paper , see @xcite , it is known that examples exist which have _ different _ , even _
non - isomorphic _ elation groups @xmath6 and @xmath35 . as @xmath40 , @xmath27 and @xmath41 can not be elements of the same isomorphism class in @xmath26 .
so classifying isomorphism classes in @xmath26 is a strictly finer job than `` just '' classifying isomorphism classes of stgqs .
the examples of stgqs in ( b ) arise as the rational points and lines on a hermitian variety in @xmath42 ( @xmath43 a power of @xmath0 ) ( the point @xmath29 is arbitrary due to the transitivity of its automorphism group ) .
this quadrangle is denoted by @xmath44 .
payne asked in 2004 ( cf .
@xcite ) whether these examples are the _ only ( finite ) stgqs _ with different elation groups for the same point .
we will come back to this question in the last section of this paper .
in this paper , a _ generalized quadrangle _ ( `` gq '' ) is a point - line incidence geometry @xmath45 for which the following axioms are satisfied : * @xmath28 contains no ordinary @xmath21-gon ( as a subgeometry ) , for @xmath46 ; * any two elements @xmath47 are contained in some ordinary @xmath5-gon in @xmath28 ; * there exists an ordinary @xmath48-gon in @xmath28 . here , @xmath49 is the point set , @xmath50 the line set , both non - empty and disjoint , and @xmath51 is a symmetric relation on @xmath52 called `` incidence , '' which tells us how points and lines are related .
so @xmath53 , with @xmath54 and @xmath55 , means that @xmath56 is incident with @xmath57 ( and @xmath57 is incident with @xmath56 ) .
an `` ordinary @xmath4-gon '' ( e.g. ) is a set of three points , two by two collinear but not all incident with the same line . by ( iii )
, generalized quadrangles have at least three points per line and three lines per point .
note that points and lines play the same role in the axioms ; this is the principle of `` duality . ''
all generalized quadrangles have an _ order _
@xmath58 ; there exist constants @xmath59 such that the number of points incident with a line is @xmath60 , and the number of lines incident with a point is @xmath61 , cf .
see also @xcite for a detailed discussion regarding parameters of generalized @xmath2-gons .
note that an ordinary quadrangle is just a `` generalized quadrangle without ( iii ) , '' of order @xmath62 | we call such a subgeometry also `` apartment '' ( of @xmath28 ) .
an _ automorphism _ of a generalized quadrangle @xmath45 is a bijection of @xmath63 which preserves @xmath49 , @xmath50 and incidence .
the full set of automorphisms of a gq forms a group in a natural way | the _ automorphism group _ of @xmath28 , denoted @xmath64 .
it is one of its most important invariants .
in @xcite , tits proved roughly that there is a one - to - one correspondence between buildings of irreducible spherical type and rank @xmath65 , and the algebraic absolutely simple groups of relative rank @xmath66 . in order to have a similar statement in the rank @xmath0 case
| the case of generalized @xmath2-gons that is to say | one must impose an extra condition , called the _
moufang condition_. we describe it for generalized quadrangles .
let @xmath67 be an apartment of a gq @xmath28 .
@xmath68 of @xmath67 is a set of @xmath48 different elements @xmath69 in @xmath67 such that @xmath70 ( where the indices are taken in @xmath71 ) , and @xmath72 are the _ extremal elements _ of @xmath68 .
there are two types of roots , depending on whether the extremal elements are lines or points ; in the second case we speak of _ dual roots _ to make a distinction between the types . also , a ( dual ) root @xmath68 without its extremal elements | the _ interior _ of @xmath68 | is denoted by @xmath73 and called ( dual ) _ i - root_. if @xmath74 is a subgeometry of @xmath28 , by @xmath75}$ ] we denote the subgroup of the automorphism group @xmath64 of @xmath28 which fixes every line incident with a point of @xmath74 and every point incident with a line of @xmath74 . now a ( dual ) root @xmath68 is _ moufang _ if @xmath76}$ ] acts transitively on the apartments containing @xmath68 . in fact , @xmath76 } = : a(\dot{\gamma})$ ] then acts _ sharply _ transitively on these apartments .
once a ( dual ) root @xmath68 is moufang , all ( dual ) roots with interior @xmath73 are also moufang , with respect to the same group @xmath77 .
( the latter groups are uniquely defined by @xmath73 and the moufang property . ) in a natural way , we also use the terms `` moufang i - root '' and `` dual moufang i - root '' , and the elements of @xmath77 are called _ root - elations_. now @xmath28 is _ half moufang _ if all roots or all dual roots are moufang .
it is _ moufang _ if all roots _ and _ dual roots are .
all moufang generalized quadrangles were classified in the classical work @xcite , as the main and hardest step in the classification of all moufang buildings of rank @xmath0 | see @xcite for a historical sketch . in the finite case ,
this already followed from work of fong and seitz @xcite , see also chapter 8 - 9 of @xcite .
historical details can be found in the survey @xcite ; see also @xcite . in @xcite ( and the references therein ) ,
the strength of the local automorphic theory for generalized quadrangles became clear : to eventually end up with a geometric treatment of the classification of moufang quadrangles , the authors developed a theory which studies local moufang conditions for quadrangles , and eventually the theory reached far beyond the eventual goal .
one instance is the theory of egqs | see chapter 8 of @xcite and the recent monograph @xcite .
other instances of local moufang theory are surveyed in , e.g. , @xcite , see also the more recent paper @xcite .
an important class of stgqs consists of those stgqs ( called `` flock quadrangles , '' and introduced in @xcite ) which arise through an intricate construction as a group coset geometry , from a _ flock _ of a quadratic cone @xmath78 in @xmath79 ( @xmath80 a finite field ) .
a _ flock _ is a partition of the @xmath80-rational points of @xmath81 into @xmath43 disjoint irreducible conics . in the fundamental paper @xcite ,
payne studied local moufang conditions in gqs , partly to understand and generalize flock quadrangles .
this motivated him to introduce _ skew translation generalized quadrangles _
( `` stgqs '' )
. one can prove that if @xmath27 is an stgq of finite order @xmath58 , all dual i - roots on @xmath29 are _ moufang for the same group _
@xmath82 ; this means that @xmath83 has size @xmath84 , and that all its elements fix every point collinear with @xmath29 .
the elements of @xmath83 are called _ symmetries _ with center @xmath29 .
the existence of @xmath83 forces @xmath29 to have a certain combinatorial property called _ regularity _ , and conversely , one can show that an egq with regular elation point is an stgq . in @xcite
, payne introduced and studied a particular class of stgqs , called `` mstgqs . ''
they are stgqs @xmath27 with the following properties .
* each i - root containing @xmath29 is moufang ( and the corresponding root group is a subgroup of @xmath6 ) . * a redundant property , since it follows from ( mstgq1 ) . * no line @xmath85 is the unique center of a triad @xmath86 with @xmath87 .
he then showed that all flock quadrangles are mstgqs , and defined his now famous `` property ( g ) '' @xcite .
a combination of @xcite and @xcite eventually led to the discovery of most of the presently known examples of gqs of order @xmath88 .
recall that all known finite generalized quadrangles except the hermitian quadrangles in projective @xmath5-space have the property that up to a combination of duality and payne - integration , they are stgqs .
this observation follows from the main result of @xcite . besides fundamental work in especially @xcite ,
not much is known in the classification theory for stgqs . by results of hachenberger @xcite and independently chen ( unpublished ) the parameters of a finite stgq
are always power of the same prime .
the _ general heisenberg group _ @xmath89 ( sometimes also written as @xmath90 if we do not want to specify @xmath43 ) of dimension @xmath91 over @xmath80 , with @xmath2 a positive integer , is the group of square @xmath92-matrices with entries in @xmath80 , of the following form ( and with the usual matrix multiplication ) : @xmath93 where @xmath94 , @xmath95 and with @xmath96 being the @xmath97-identity matrix .
the group @xmath98 is isomorphic to the group @xmath99 , where the group operation @xmath100 is given by @xmath101 ( here , @xmath102 is a notation for transposition ) .
the following properties hold for @xmath98 ( defined over @xmath80 ) .
* @xmath98 has exponent @xmath103 if @xmath104 with @xmath103 an odd prime ; it has exponent @xmath5 if @xmath43 is even .
* the center of @xmath98 is given by @xmath105 .
* @xmath106 = z = \phi({\mathcal{h}}_n)$ ] and @xmath98 is nilpotent of class @xmath0 ( @xmath107 is the _ frattini subgroup _ of @xmath108 , that is , the intersection of all its maximal subgroups ) .
let @xmath109 be the elementary abelian @xmath103-group @xmath110 .
the map @xmath111 @xmath112\ ] ] naturally defines a non - singular bilinear alternating form over @xmath113 .
so @xmath109 can be seen as a @xmath5-dimensional space over @xmath80 , and in the corresponding projective @xmath4-space over @xmath80 , @xmath111 defines a symplectic polar space @xmath114 of rank @xmath0 ( projective index @xmath115 ) . here
, @xmath114 can be defined as the generalized quadrangle which arises as the points of @xmath116 , @xmath80 a finite field , together with the totally isotropic projective lines of a non - degenerate alternating bilinear form on @xmath116 .
all its points are regular elation points .
the importance of alternating forms in stgq theory can be read from the following theorem .
its proof uses the connection explained in [ fsp ] .
suppose @xmath108 is a special @xmath103-group of order @xmath117 for which @xmath118 $ ] is elementary abelian of order @xmath43 .
suppose @xmath108 admits a kantor family of type @xmath119 , and suppose @xmath111 defines a non - singular bilinear alternating form over @xmath80 .
then @xmath120 , and the corresponding generalized quadrangle @xmath28 of order @xmath119 is a flock quadrangle .
roughly put ( see the citation @xcite for more details ) , if for an stgq @xmath27 , @xmath6 is isomorphic to a general heisenberg group , then @xmath28 is a flock quadrangle : @xmath121 so enough structural knowledge of the group in this case , leads to determination of @xmath28 .
this result / idea is one of the main models for the classification started in @xcite .
we are ready to describe several new results in stgq theory . some of them settle long - standing open problems .
proofs will be published elsewhere ( see @xcite ) .
the first result was obtained in 2009 , and already announced ( a.o . ) on the 2010 conference `` combinatorics 2010 '' in verbania .
it was also mentioned ( without proof ) in the proceedings paper of my talk @xcite .
it can be found in the preprint @xcite .
i explained several proofs of this result in a lecture at the `` buildings 2012 '' conference in mnster .
if ( c ) is true for @xmath27 , a slightly more general version of the moufang property holds for any i - root containing @xmath29 .
this allows one to control the situation to quite a far extent @xcite ; @xmath6 then comes in a class of abstract groups which share several distinguished properties with general heisenberg groups .
property ( * ) : _ let @xmath27 be an stgq .
let @xmath127 .
then @xmath28 has @xmath128 _ at @xmath57 _ , if for some @xmath53 , @xmath129 , @xmath130 is a normal subgroup of @xmath6 . in that case , @xmath130 is independent of the choice of @xmath56 .
the stgq has @xmath128 if it has @xmath128 at every line on @xmath29 .
_ + payne s mstgqs always satisfy ( * ) .
an important corollary is that all i - roots on @xmath29 are moufang , if ( * ) is satisfied ( and its order is not @xmath133 if @xmath84 is even ) .
this enables one to understand the elation groups much better | see @xcite .
when the order is @xmath133 and @xmath84 is even , the situation is ( very ) different , and an altogether different approach is needed to attack this case .
unfortunately ( depending on the viewpoint ) , not all stgqs have ( * ) at every line through the elation point ; a first class of counter examples ( related to suzuki groups ) is displayed in the preprint @xcite | see also 8.4 of @xcite . in @xcite it is conjectured that for a finite stgq @xmath27 , either zero , one or all lines incident with @xmath29 have ( * ) .
let @xmath27 be an stgq , and let @xmath134 be the frattini subgroup of @xmath6 .
define a point - line geometry @xmath135 as follows .
its lines are the @xmath136-orbits on the lines incident with @xmath29 ( where the trivial orbit @xmath137 is excluded ) ; its points are the @xmath136-orbits in the set of points not collinear with @xmath29 .
a point @xmath138 is incident with a line @xmath109 if at least one @xmath28-point of the orbit @xmath109 is collinear with some point of the orbit @xmath138 .
if @xmath138 then is incident with @xmath109 , it is easy to see that @xmath109 is surjectively projected on @xmath138 , so that for each @xmath139 , a @xmath135-point is incident with precisely one line which is a @xmath136-orbit on @xmath140 .
so each @xmath135-point is incident with precisely @xmath61 @xmath135-lines .
let @xmath27 be a generic stgq of order @xmath58 , with kantor family @xmath9 .
let @xmath141 , and suppose that the only extension of @xmath142 in @xmath6 is @xmath6 itself .
suppose also that at least one of the following properties is satisfied .
it is then shown that @xmath27 always contains ideal sub stgqs ( that is , it contains sub stgqs containing @xmath29 and all the lines on @xmath29 ) . whence the parameters are always of type @xmath132 for this type of generic stgqs . | we describe new classification results in the theory of generalized quadrangles (= tits - buildings of rank @xmath0 and type @xmath1 ) , more precisely in the ( large ) sub theory of skew translation generalized quadrangles ( `` stgqs '' ) . some of these involve , and solve , long - standing open problems . |
the collection @xmath1 of subgroups of direct products of free groups is surprisingly rich and has been studied by many authors . in the early 1960s stallings
@xcite exhibited a subgroup of @xmath2 , where @xmath3 is the rank @xmath4 free group , as the first known example of a finitely presented group whose third integral homology group is not finitely generated .
bieri @xcite demonstrated that stallings group belongs to a sequence of groups @xmath5 , the stallings - bieri groups , with @xmath6 being of type @xmath7 but not of type @xmath8 .
( see @xcite for definitions and background concerning finiteness properties of groups . ) in the realm of decision theory , mihalova @xcite and miller @xcite exhibited a finitely generated subgroup of @xmath9 with undecidable conjugacy and membership problems .
it is thus seen that even in the @xmath4-factor case fairly wild behaviour is encountered amongst the subgroups of direct products of free groups .
in contrast to this , baumslag and roseblade @xcite showed that in the @xmath4-factor case this wildness only manifests itself amongst the subgroups which are not finitely presented .
they proved that if @xmath10 is a finitely presented subgroup of @xmath11 , where @xmath12 and @xmath13 are free groups , then @xmath10 is itself virtually a direct product of at most @xmath4 free groups .
this work was extended by bridson , howie , miller and short @xcite to an arbitrary number of factors .
they proved that if @xmath10 is a subgroup of a direct product of @xmath14 free groups and if @xmath10 enjoys the finiteness property @xmath8 , then @xmath10 is virtually a direct product of at most @xmath14 free groups .
further information on the finiteness properties of the groups in class @xmath1 was provided by meinert @xcite who calculated the bns invariants of direct products of free groups .
these invariants determine the finiteness properties of all subgroups lying above the commutator subgroup .
several authors have investigated the isoperimetric behaviour of the finitely presented groups in @xmath1 .
elder , riley , young and this author @xcite have shown that the dehn function of stallings group @xmath15 is quadratic .
the method espoused by bridson in @xcite proves that the function @xmath16 is an upper bound on the dehn functions of each of the stallings - bieri groups .
in contrast , there have been no results which give non - trivial lower bounds on the dehn functions of any groups in @xmath1 . in other words , until now there has been no finitely presented subgroup of a direct product of free groups whose dehn function was known to be greater than that of the ambient group . the purpose of this paper is to construct such a subgroup : we exhibit a finitely presented subgroup of @xmath2 whose dehn function has the function @xmath16 as a lower bound .
let @xmath17 be the kernel of a homomorphism @xmath18 whose restriction to each factor @xmath3 is surjective .
by lemma [ lem1 ] below the isomorphism class of @xmath17 is independent of the homomorphism chosen .
results in @xcite show that @xmath17 is finitely presented .
[ thm1 ] the dehn function @xmath19 of @xmath17 satisfies @xmath20 .
note that theorem [ thm1 ] makes no reference to a specific presentation of @xmath17 since , as is well known , the dehn function of a group is independent ( up to @xmath21-equivalence ) of the presentation chosen .
we refer the reader to section 2 for background on dehn functions , including definitions of the symbols @xmath22 and @xmath21 .
the organisation of this paper is as follows .
section 2 gives basic definitions concerning dehn functions , van kampen diagrams and cayley complexes .
we expect that the reader will already be familiar with these concepts ; the purpose of the exposition is principally to introduce notation . in section 3
we define a class of subgroups of direct products of free groups of which @xmath17 is a member .
the subsequent section gives finite generating sets for those groups in the class which are finitely generated and section 5 proves a splitting theorem which shows how certain groups in the class decompose as amalgamated products .
we then prove in section 6 how , in certain circumstances , the distortion of the subgroup @xmath23 in an amalgamated product @xmath24 gives rise to a lower bound on the dehn function of @xmath25 .
theorem [ thm1 ] follows as a corollary of this result when applied to the splitting of @xmath17 given in section 5 .
in this section we recall the basic definitions concerning dehn functions of finitely presented groups .
all of this material is standard . for further background and a more thorough exposition , including proofs ,
see , for example , @xcite or @xcite . given a set @xmath26 , write @xmath27 for the set of formal inverses of the elements of @xmath26 and write @xmath28 for the set @xmath29 .
denote by @xmath30 the free monoid on the set @xmath28 .
we refer to elements of @xmath30 as words in the letters @xmath28 and write @xmath31 for the length of such a word @xmath32 .
given words @xmath33 we write @xmath34 if @xmath35 and @xmath36 are freely equal and @xmath37 if @xmath35 and @xmath36 are equal as elements of @xmath38 . [ def1 ]
let @xmath39 be a finite presentation of a group @xmath10 .
a word @xmath40 is said to be _ null - homotopic _ over @xmath41 if it represents the identity in @xmath10 .
null-@xmath41-expression _ for such a word is a sequence @xmath42 in @xmath43 such that @xmath44 define the _ area _ of a null-@xmath41-expression to be the integer @xmath45 and define the _
@xmath41-area _ of @xmath32 , written @xmath46 , to be the minimal area taken over all null-@xmath41-expressions for @xmath32 .
the dehn function of the presentation @xmath41 , written @xmath47 , is defined to be the function @xmath48 given by @xmath49 although the dehn functions of different finite presentations of a fixed group may differ , their asymptotic behaviour will be the same .
this is made precise in the following way .
given functions @xmath50 , write @xmath51 if there exists a constant @xmath52 such that @xmath53 for all @xmath14 . write @xmath54 if @xmath51 and @xmath55 . if @xmath56 and @xmath57 are finite presentations of the same group then @xmath58 . for
a proof of this standard result see , for example , ( * ? ? ?
* proposition 1.3.3 ) .
a useful tool for the study of dehn functions is a class of objects known as van kampen diagrams . roughly speaking , these are planar cw - complexes which portray diagrammatically schemes for reducing null - homotopic words to the identity .
such diagrams , whose definition is recalled below , allow the application of topological methods to the calculation of dehn functions . for background and further details
see , for example , ( * ? ? ?
* section 4 ) . for the definition of a combinatorial cw - complex see , for example , ( * ? ? ?
* appendix a ) . a _ singular disc diagram _
@xmath59 is a finite , planar , contractible combinatorial cw - complex with a specified base vertex @xmath60 in its boundary .
the _ area _ of @xmath59 , written @xmath61 , is defined to be the number of @xmath4-cells of which @xmath59 is composed .
the _ boundary cycle _ of @xmath59 is the edge loop in @xmath59 which starts at @xmath60 and traverses @xmath62 in the anticlockwise direction .
the interior of @xmath59 consists of a number of disjoint open @xmath4-discs , the closures of which are called the _
disc components _ of @xmath59 .
each @xmath63-cell of @xmath59 has associated to it two directed edges @xmath64 and @xmath65 , with @xmath66 .
let @xmath67 be the set of directed edges of @xmath59 .
a _ labelling _ of @xmath59 over a set @xmath26 is a map @xmath68 such that @xmath69 .
this induces a map from the set of edge paths in @xmath59 to @xmath38 .
boundary label _ of @xmath59 is the word in @xmath38 associated to the boundary cycle .
let @xmath70 be a finite presentation .
@xmath41-van kampen diagram _ for a word @xmath71 is a singular disc diagram @xmath59 labelled over @xmath26 with boundary label @xmath32 and such that for each @xmath4-cell @xmath72 of @xmath59 the anticlockwise edge loop given by the attaching map of @xmath72 , starting at some vertex in @xmath73 , is labelled by a word in @xmath74 .
a word @xmath75 is null - homotopic over @xmath41 if and only if there exists a @xmath41-van kampen diagram for @xmath32 . in this case
the @xmath41-area of @xmath32 is the minimal area over all @xmath41-van kampen diagrams for @xmath32 . for
a proof of this result see , for example , ( * ? ? ?
* theorem 4.2.2 ) . associated to a presentation @xmath76 of a group @xmath10
there is a standard combinatorial @xmath4-complex @xmath77 with @xmath78 .
the complex @xmath77 is constructed by taking a wedge of copies of @xmath79 indexed by the letters in @xmath26 and attaching @xmath4-cells indexed by the relations in @xmath80 .
the @xmath4-cell corresponding to a relation @xmath81 has @xmath82 edges and is attached by identifying its boundary circuit with the edge path in @xmath83 along which one reads the word @xmath84 .
the cayley @xmath4-complex associated to @xmath41 , denoted by @xmath85 , is defined to be the universal cover of @xmath77 .
if one chooses a base vertex of @xmath85 to represent the identity element of @xmath10 then the @xmath63-skeleton of this complex is canonically identified with the cayley graph of @xmath41 .
given a @xmath41-van kampen diagram @xmath59 there is a unique label - preserving combinatorial map from @xmath59 to @xmath85 which maps the base vertex of @xmath59 to the vertex of @xmath85 representing the identity .
in this section we introduce a class of subgroups of direct products of free groups of which the group @xmath17 defined in the introduction will be a member .
we first fix some notation which will be used throughout the paper .
given integers @xmath86 , let @xmath87 be the rank @xmath45 free group with basis @xmath88 . given an integer @xmath89 , let @xmath90 be the rank @xmath84 free abelian group with basis @xmath91 .
given positive integers @xmath92 and @xmath93 , we wish to define a group @xmath94 to be the kernel of a homomorphism @xmath95 whose restriction to each factor @xmath87 is surjective . for fixed @xmath14 , @xmath45 and @xmath84 ,
the isomorphism class of the group @xmath94 is , up to an automorphism of the factors of the ambient group @xmath96 , independent of the homomorphism @xmath97 .
this is proved by the following lemma .
[ lem1 ] let @xmath98 be a rank @xmath45 free group .
given a surjective homomorphism @xmath99 , there exists a basis @xmath100 of @xmath98 so that @xmath101 the homomorphism @xmath102 factors through the abelianisation homomorphism @xmath103 , where @xmath104 is the rank @xmath45 free abelian group @xmath105 $ ] , as @xmath106 for some homomorphism @xmath107 .
since @xmath108 is surjective , @xmath104 splits as @xmath109 where @xmath108 is an isomorphism on the first factor and @xmath110 on the second factor . there
thus exists a basis @xmath111 for @xmath104 so as @xmath112 we claim that the @xmath113 lift under @xmath114 to a basis for @xmath98 . to see this
let @xmath115 be any basis for @xmath98 and let @xmath116 be its image under @xmath114 , a basis for @xmath104 .
let @xmath117 be the change of basis isomorphism from @xmath116 to @xmath111 .
it suffices to show that this lifts under @xmath114 to an automorphism of @xmath98 . but
this is certainly the case since @xmath118 is generated by the elementary transformations and each of these obviously lifts to an automorphism .
[ def2 ] for integers @xmath92 and @xmath93 , define @xmath94 to be the kernel of the homomorphism @xmath119 given by @xmath120 note that @xmath121 is the @xmath122 stallings - bieri group @xmath123 . by a result in section 1.6 of @xcite , if @xmath124 and @xmath125 then @xmath94 is of type @xmath7 but not of type @xmath126 .
in particular the group @xmath127 defined in the introduction is finitely presented .
we give finite generating sets for those groups @xmath94 which are finitely generated .
we make use of the following notational shorthand : given formal symbols @xmath128 and @xmath129 , write @xmath130 $ ] for @xmath131 and @xmath132 for @xmath133 .
[ prop1 ] if @xmath134 then @xmath94 is generated by @xmath135 where @xmath136 \ , : \ , 1 \leq i < j \leq r \}. \end{aligned}\ ] ] partition @xmath137 as @xmath138 where @xmath139 and @xmath140 project @xmath141 onto the last @xmath142 factors to give the short exact sequence @xmath143 .
note that @xmath144 projects to a set of generators for @xmath145 and that @xmath146 is the normal closure in @xmath147 of @xmath148 . if @xmath149 and @xmath150 is a word in the generators of @xmath147 then
@xmath151 thus @xmath152 generates @xmath94 .
the following result gives an amalgamated product decomposition of the groups @xmath94 in the case that @xmath153 .
note that with slightly more work one could prove a more general result without this restriction .
we introduce the following notation : given a collection of groups @xmath154 with @xmath155 for each @xmath156 , we denote by @xmath157 the amalgamated product @xmath158 . [ thm2 ] if @xmath134 and @xmath159 then @xmath160 , where @xmath161 and , for each @xmath162 , the group @xmath163 is the kernel of the homomorphism @xmath164 given by @xmath165 projecting @xmath166 onto the factor @xmath167 gives the short exact sequence @xmath168 .
this splits to show that @xmath166 has the structure of an internal semidirect product @xmath169 where @xmath170 is the subgroup of @xmath171 generated by @xmath172 since the action by conjugation of @xmath173 on @xmath174 is the same as the action of @xmath175 , we have that @xmath176 define a homomorphism @xmath177 by @xmath178 and note that @xmath179 is the kernel @xmath180 of the standard homomorphism @xmath181 given in definition [ def2 ] .
considering the restriction of @xmath182 to @xmath183 gives the short exact sequence @xmath184 which demonstrates that @xmath185 .
in this section we will be concerned with finitely presented amalgamated products @xmath24 where @xmath23 , @xmath186 and @xmath187 are finitely generated groups and @xmath23 is proper in each @xmath188 .
suppose each @xmath188 is presented by @xmath189 , with @xmath190 finite .
note that we are at liberty to choose the @xmath191 so as each @xmath192 represents an element of @xmath193 .
indeed , since @xmath23 is proper in @xmath188 there exists some @xmath194 representing an element of @xmath193 and we can replace each other element @xmath195 by @xmath196 if necessary .
let @xmath197 be a finite generating set for @xmath23 and for each @xmath198 choose words @xmath199 and
@xmath200 which equal @xmath201 in @xmath25 . define @xmath202 to be the finite collection of words @xmath203 .
then , since @xmath25 is finitely presented , there exist finite subsets @xmath204 and @xmath205 such that @xmath25 is finitely presented by @xmath206 [ thm3 ] let @xmath207 be a word representing an element @xmath208 and let @xmath209 and @xmath210 be words representing elements @xmath211 and @xmath212 respectively . if @xmath213 = [ \beta , h ] = 1 $ ] then @xmath214 ) \geq 2n \ , { \rm d}_\mathcal{b}(1 , h)\ ] ] where @xmath215 is the word metric on @xmath23 associated to the generating set @xmath197 . for each @xmath217 ,
define @xmath218 to be the vertex in @xmath62 such that the anticlockwise path in @xmath62 from the basepoint around to @xmath218 is labelled by the word @xmath219 .
similarly define @xmath220 to be the vertex in @xmath62 such that the anticlockwise path in @xmath221 from the basepoint around to @xmath220 is labelled by the word @xmath222 .
we will show that for each @xmath156 there is a @xmath197-path ( i.e. an edge path in @xmath59 labelled by a word in the letters @xmath197 ) from @xmath218 to @xmath220 .
we assume that the reader is familiar with bass - serre theory , as exposited in @xcite .
let @xmath223 be the bass - serre tree associated to the splitting @xmath224 .
this consists of an edge @xmath225 for each coset @xmath226 and a vertex @xmath227 for each coset @xmath228 .
the edge @xmath225 has initial vertex @xmath229 and terminal vertex @xmath230 .
we will construct a continuous ( but non - combinatorial ) map @xmath231 as the composition of the natural map @xmath232 with the map @xmath233 defined below .
there is a natural left action of @xmath25 on each of @xmath85 and @xmath223 and we construct @xmath234 to be equivariant with respect to this as follows .
let @xmath45 be the midpoint of the edge @xmath23 of @xmath223 and define @xmath234 to map the vertex @xmath235 to the point @xmath236 , the midpoint of the edge @xmath225 .
define @xmath234 to map the edge of @xmath85 labelled @xmath237 joining vertices @xmath238 and @xmath239 to the geodesic segment joining @xmath236 to @xmath240 . since @xmath241 this segment is an embedded arc of length @xmath63 whose midpoint is the vertex @xmath227 .
define @xmath234 to collapse the edge in @xmath85 labelled @xmath242 joining vertices @xmath238 and @xmath243 to the point @xmath244 .
this is well defined since @xmath245 .
this completes the definition of @xmath234 on the @xmath63-skeleton of @xmath59 ; we now extend @xmath234 over the @xmath4-skeleton .
let @xmath72 be a @xmath4-cell in @xmath85 and let @xmath238 be some vertex in its boundary .
assume that @xmath72 is metrised so as to be convex and let @xmath246 be some point in its interior .
the form of the relations in @xmath41 ensures that the boundary label of @xmath72 is a word in the letters @xmath247 for some @xmath156 and so every vertex in @xmath73 is labelled @xmath248 for some @xmath249 .
thus @xmath234 as so far defined maps @xmath73 into the ball of radius @xmath250 centred on the vertex @xmath251 ; we extend @xmath234 to the interior of @xmath72 by defining it to map the geodesic segment @xmath252 $ ] , where @xmath253 , to the geodesic segment @xmath254 $ ] .
this is independent of the vertex @xmath255 chosen and makes @xmath234 continuous since geodesics in a tree vary continuously with their endpoints .
we now define @xmath256 to be the map given by composing @xmath234 with the label - preserving map @xmath257 which sends the basepoint of @xmath59 to the vertex @xmath258 .
since @xmath32 commutes with @xmath259 and @xmath260 we have that @xmath261 ; define @xmath262 to be the preimage under @xmath263 of this point . by construction ,
the image of the interior of each @xmath4-cell in @xmath59 and the image of the interior of each @xmath190-edge is disjoint from @xmath264 .
thus @xmath262 consists of vertices and @xmath197-edges and so finding a @xmath197-path from @xmath218 to @xmath220 reduces to finding a path in @xmath262 connecting these vertices .
let @xmath113 and @xmath265 be the vertices of @xmath62 immediately preceding and succeeding @xmath218 in the boundary cycle .
unless @xmath266 , in which case the theorem is trivial , the form of the word @xmath216 $ ] , together with the normal form theorem for amalgamated products , implies that all the vertices @xmath218 , @xmath113 and @xmath265 lie in the boundary of the same disc component @xmath267 of @xmath59 .
furthermore , since @xmath259 and @xmath260 are words in the letters @xmath268 and @xmath269 respectively , the points @xmath270 and @xmath271 are separated in @xmath223 by @xmath272 . thus @xmath113 and @xmath265 are separated in @xmath267 by @xmath262 and so there exists an edge path @xmath273 in @xmath262 from @xmath218 to some other vertex @xmath274 .
since @xmath273 is a @xmath197-path it follows that the word labelling the sub - arc of the boundary cycle of @xmath59 from @xmath218 to @xmath275 represents an element of @xmath23 , and , by considering subwords of @xmath216 $ ] , we see that the only possibility is that @xmath276 . thus , for each @xmath277 , the path @xmath273 gives the required @xmath197-path connecting @xmath218 to @xmath220 .
we choose each @xmath273 to contain no repeated edges .
for @xmath278 , the two paths @xmath273 and @xmath279 are disjoint , since if they intersected there would be a @xmath197-path joining @xmath218 to @xmath280 and thus the word labelling the subarc of the boundary cycle from @xmath218 to @xmath280 would represent an element of @xmath23 .
observe that no two edges in any of the paths @xmath281 lie in the boundary of the same @xmath4-cell in @xmath59 since each relation in @xmath282 contains at most one occurrence of a letter in @xmath197 . because the word labelling @xmath221 contains no occurrences of a letter in @xmath197 the interior of each edge of a path @xmath273 lies in the interior of @xmath59 and thus in the boundary of two distinct @xmath4-cells . since each path @xmath273 contains no repeated edges we therefore obtain the bound @xmath283 .
but the word labelling each @xmath273 is equal to @xmath284 in @xmath25 and so the length of @xmath273 is at least @xmath285 whence we obtain the required inequality .
by proposition [ prop1 ] and theorem [ thm2 ] we have that @xmath289 where , as subgroups of @xmath290 , @xmath291 is generated by @xmath292 , @xmath293 is generated by @xmath294 and @xmath295 is generated by @xmath296 \}$ ] . to obtain the generating set for @xmath297 we have here
implicitly used the automorphism of @xmath290 which interchanges @xmath298 with @xmath299 and realises the isomorphism between @xmath297 and @xmath300 . for @xmath301 , define @xmath302 to be the element @xmath303 \in k^2_2(2)$ ] and define @xmath304 to be the word @xmath305 \in { \mathcal{a}}_1^{\pm{{\displaystyle\ast}}}$ ] representing @xmath302 .
note that @xmath302 commutes with both @xmath306 and @xmath307 so by theorem [ thm3 ] the word @xmath308 $ ] , which has length @xmath309 , has area at least @xmath310 .
we claim that @xmath311 .
suppose that in @xmath290 the element @xmath302 is represented by a word @xmath312)$ ] in the generators @xmath197 .
let @xmath313 be the number of occurrences of the third variable in the word @xmath32 .
we will show that @xmath314 .
observe that , as a group element , the word @xmath315)$ ] is equal to the word @xmath316 ) \
, w(x_2^{-1 } , y_2^{-1 } , 1).$ ] thus we have that @xmath303 $ ] is freely equal to @xmath316)$ ] and that @xmath317 , and thus @xmath318 , is freely equal to the empty word .
it follows that @xmath303 $ ] can be converted to the empty word by free expansions , free contractions and deletion of @xmath313 subwords @xmath319 $ ] .
hence @xmath303 $ ] is a null - homotopic word over the presentation @xmath320 \rangle$ ] with @xmath41-area at most @xmath313 . but @xmath41 presents the rank @xmath4 free abelian group , and basic results on dehn functions give that @xmath321 $ ] has area @xmath322 over this presentation .
thus @xmath314 . | we establish a cubic lower bound on the dehn function of a certain finitely presented subgroup of a direct product of @xmath0 free groups .
department of mathematics + university walk + bristol , bs8 1tw + united kingdom + ` [email protected] ` |
it is now generally accepted that elliptical galaxies are complicated objects , containing a variety of stellar populations , with a still poorly constrained dark matter content and a multi - component interstellar medium .
the latter , and in particular the interstellar dust component , is not well understood .
nevertheless , knowledge about the dust component in galaxies is of fundamental importance . on the one hand ,
interstellar dust plays an active role in e.g. interstellar chemistry and star formation , and is therefore a key ingredient to understand galaxy structure and evolution . on the other hand , dust is very effective in absorbing and scattering uv and optical light , and knowledge about its presence , quantity and properties is necessary to correctly interpret any observable . the far - infrared fluxes detected in the eighties by the _ iras _ satellite at 60 @xmath0 m and 100 @xmath0 m unveiled the presence of a substantial amount of dust in elliptical galaxies ( jura 1986 ; bally & thronson 1989 ; knapp , gunn & wynn - williams 1989 ) . at that time , dust was already observed in the form of dust lanes and patches for a number of early - type galaxies ( hawarden et al . 1981 ; ebneter & balick 1985 ; vron - cetty & vron 1988 ) .
recent surveys indicate that dust extinction features are present in a large fraction of early - type galaxies ( van dokkum & franx 1995 ; ferrari et al . 1999
; tomita et al .
2000 ; tran et al . 2001 ) .
these dust features , however , can not account for the high _ iras _ fluxes : the dust masses estimated from the _ iras _ measurements exceed those estimated from integrating the extinction features by nearly an order of magnitude ( goudfrooij & de jong 1995 ) .
moreover , the _ iras _ dust mass estimates are a lower limit for the true dust masses , because _ iras _ is not sensitive to cold dust , which emits the bulk of its radiation longwards of 100 @xmath0 m .
calculations with a more realistic dust temperature distribution ( merluzzi 1998 ) and submillimeter observations ( fich & hodge 1993 ; wiklind & henkel 1995 ) may indicate dust masses up to a magnitude higher than the _ iras _ estimates . the only plausible way to solve this dust mass discrepancy is to assume that the interstellar dust in elliptical galaxies exists as a two - component medium : the less massive component is optically visible in the form of dust lanes , whereas the more massive one is distributed diffusely over the galaxy ( goudfrooij & de jong 1995 ) . such a diffusely distributed dust distribution can be traced in two ways .
the most obvious way is to map the thermal emission of the dust with spatially resolved far - infrared or submillimeter observations . a first effort to do this
was undertaken by leeuw , sansom & robson ( 2000 ) , who used _ scuba _ imaging at 850 @xmath0 m to look for thermal emission from dust in ngc4374 .
unfortunately , they could not detect a diffuse dust component , and found that the bulk of the 850 @xmath0 m core emission is more likely the result of synchrotron radiation .
deeper imaging covering a larger wavelength range is necessary to constrain the spatial distribution of dust this way .
hopefully , the new generation of far - infrared and submillimeter instrumentation such as _ sirtf _ and _ alma _ will help to clarify this issue .
a second way in which the diffuse dust component in elliptical galaxies can be traced is by colour gradients : because the extinction efficiency of dust grains decreases with wavelength , a diffuse dust component is expected to generate bluer colours at larger projected radii .
goudfrooij & de jong ( 1995 ) and wise & silva ( 1996 ) found that the dust distributions , necessary to create the colour gradients observed in a sample of elliptical galaxies , are in reasonable agreement with their observed integrated fir fluxes .
however , various arguments seriously complicate the interpretation of colour gradients .
foremost , dust attenuation is not the only process that generates colour gradients : also age and metallicity variations can cause broadband colour gradients . recently , michard ( 2000 ) argued that the mean observed colour gradient ratios of a sample of elliptical galaxies are more likely the result of metallicity than of diffuse dust . and even if the dust attenuation were the only process responsible for the generation of colour gradients , tracing and quantifying the amount of dust would still be complicated .
first , the dust is not located between the source and the observer , but well mixed with the stars .
as a result , the amount of dust is not simply proportional to the amount of reddening ( disney , davies & phillipps 1989 ) .
second , bluing due to scattering partly compensates the effects of reddening due to absorption , which suppresses the formation of large broad band colour gradients even if a substantial amount of dust is present ( witt , thronson & capuano 1992 ) .
we have set up a program to investigate the effects of diffuse dust on the observed kinematics of elliptical galaxies . in the first two papers of this series ( baes & dejonghe 2000 , hereafter paper i
; baes , dejonghe & de rijcke 2000 , hereafter paper ii ) we investigated how the light profile and the observed kinematics of elliptical galaxies are affected by dust absorption . in this series
we extend our models by including the effect of scattering . in section [ rte.sec ]
we explain how the processes of dust absorption and scattering can affect the observables of galaxies .
in particular , we explain how kinematical information can be included in the radiative transfer equation . in section [ montecarlo.sec ]
, we argue that a monte carlo method is the most straightforward way to do this .
we describe the code we developed , with a special emphasis on the calculation of the observed kinematics .
we use this code to investigate the effects of dust attenuation on the observed kinematics in elliptical galaxies . in section
[ galaxymodels.sec ] we present a set of simple elliptical galaxy models , consisting of a stellar and a dust component .
we demonstrate how dust attenuation affects the light profile and observed kinematics of these models in section [ results.sec ] .
section [ discussion.sec ] is devoted to a discussion of our results , and finally , section [ conclusions.sec ] sums up .
the basis for any study of attenuation is the radiative transfer equation ( rte ) , which statistically describes the interaction between matter and light . in a general form
, the time - independent rte can be written as ( chandrasekhar 1960 ; mihalas 1978 ) @xmath1 where @xmath2 is the path length , and @xmath3 represents the intensity at a position @xmath4 into a direction @xmath5 .
the right - hand side of the equation contains two terms that represent how the radiation field changes as a result of interactions with the matter . the first term , the total emissivity @xmath6 accounts for the sources of the radiation field .
the second term , where @xmath7 is the total opacity coefficient , represents the sinks .
the specific form of both terms depends on which physical processes are taken into account , ( e.g. stellar emission , absorption , scattering ) and they can be complicated functions that depend on the intensity itself . in a galaxy without dust attenuation , there are no sink terms , and the only source in the radiation field is the ( isotropic ) emission of photons by stars , such that @xmath8 where @xmath9 represents the stellar emissivity . if we take absorption by dust grains into account , a sink term must be added , which accounts for the loss of photons from the radiation field .
the total opacity then equals the absorption coefficient , i.e.@xmath10 . in either of these situations ,
the rte is a ordinary differential equation , and it can be solved by an integration along the line of sight .
dust grains do not only absorb photons , they also scatter them , i.e. a number of photons are , as a result of an interaction with a dust grain , removed from their path and sent into another direction .
this physical process is not a rare phenomenon : for typical milky way dust grains , the probability for scattering even slightly exceeds the probability for absorption in the optical wavelength range ( table [ dustprop.tab ] ) .
it is therefore obvious that scattering should be included in radiative transfer calculations .
this will add two extra terms to the rte .
the first one is a sink term that accounts for the loss of photons scattered out of the beam .
the total opacity will hence be the sum of the absorption coefficient and the scattering coefficient , @xmath11 .
the second extra term is a source term that characterizes the gain of photons scattered into the beam .
more precisely , this term will contain the contribution of photons that had another direction @xmath12 , but are now scattered into the direction @xmath5 .
generally , the distribution of angles after a scattering process are described by a scattering phase function @xmath13 , which describes the probability that a photon which comes from the direction @xmath12 and is scattered at @xmath4 , will have @xmath5 as its new direction . by convention ,
it is normalized as @xmath14 the fraction of the intensity scattered from a solid angle @xmath15 around an arbitrary direction @xmath12 into the direction @xmath5 will therefore equal @xmath16 , such that the extra source term that has to be added to the rte reads @xmath17 it is convenient to introduce the scattering albedo @xmath18 as the ratio of the scattering coefficient to the total opacity coefficient , @xmath19 we finally find for the rte , @xmath20 the inclusion of scattering turns the rte into a integro - differential equation , far more complicated than the ordinary differential equation if only absorption is taken into account . in particular , the scattering term is responsible for the coupling of the rte along different paths : due to the integration over the angle , we can not solve the rte for a single path , but we have to solve it for all paths at the same time
. the complexity of the rte does not only depend on which physical processes are taken into account ( the right - hand side ) , but also on the geometry of the system .
indeed , the path length appearing in the left - hand side of the rte is in general a function of position and direction , i.e. @xmath21 . as a consequence ,
the ( time - independent ) rte is a partial differential equation with five independent coordinates ( see e.g. mihalas 1978 ) .
this complexity , however , is reduced if the system has symmetries .
for example , in an axially symmetric geometry , the azimuthal dependence vanishes , such that only four independent coordinates need to be considered .
particularly interesting is the spherical geometry , because in that case only two coordinates remain : the radius @xmath22 and the cosine @xmath0 of the angle between the direction @xmath5 and the local radial direction .
the aim of this paper is to provide a global picture of the effects of dust attenuation on the observed kinematics of elliptical galaxies , rather than modelling a specific , possibly geometrically complex object . for this goal ,
the assumption of spherical symmetry is satisfactory .
there are no ways to solve the general rte ( [ rte_dust ] ) analytically , but several techniques have been developed to solve it numerically . and
because many astrophysical systems can in a first approximation be considered spherically symmetric , many efforts have yet been spent on the rte in a spherical geometry .
the pioneering work started nearly 70 years ago ( kosirev 1934 ; chandrasekhar 1934 ) , and , in particular since the 1970s , a vast number of different approaches has been developed ( e.g. hummer & rybicki 1971 ; schmid - burgk 1975 ; witt 1977 ; flannery , roberge & rybicki 1980 ; yorke 1980 ; rowan - robinson 1980 ; rogers & martin 1984 ; peraiah & varghese 1985 ; gros , crivellari & simonneau 1997 ) .
most of these techniques , however , are not suitable for our needs , because it seems hard ( and at least not obvious ) to extend them such that they can handle kinematical information . only for monte carlo methods
, the kinematical information could be included in an elegant way . as long as scattering
is not taken into account , the inclusion of velocity information in the rte is rather straightforward .
indeed , instead of taking into account the entire stellar emissivity , we can just consider the light from those stars whose velocity component in the direction of the observer equals @xmath23 .
the emissivity of these stars is given by @xmath24 , where @xmath25 is the spatial losvd , i.e. the probability for a star at position @xmath4 to have a velocity component @xmath23 in the direction @xmath5 ( appendix a ) .
solving the rte with this emissivity , we obtain the losvds observed in the plane of the sky . hence ,
if either dust attenuation is completely neglected or only absorption by dust grains is taken into account , the observed losvds can be calculated from the spatial losvds through a single integration along the line of sight .
similar relations hold between the moments of the distribution function , and the moments of the losvds , e.g. the observed velocity dispersion profile can be determined from the intrinsic velocity dispersions by a simple integration along the line of sight . for more details , we refer to paper i. when scattering is included , however , the inclusion of kinematical information becomes much more complicated , because the velocities of both the stars and the dust grains that scatter their light need to be taken into account .
consider , for example , fig.[kininfo.eps ] .
a photon with rest wavelength @xmath26 is emitted by a star moving with a velocity @xmath27 into a direction @xmath28 ( in a reference frame centered on the galaxy center ) .
assume this photon is scattered by a dust grain moving with velocity @xmath29 , and then scattered towards the observer .
the wavelength `` detected '' by the dust grain is determined by the relative velocity between star and dust grain , i.e. the dust grain detects a wavelength @xmath30.\ ] ] the dust grains scatters the photon coherently into the direction @xmath31 , such that the wavelength detected by the observer equals @xmath30 \left[1-\frac{({{\boldsymbol{v}}}_d+{{\boldsymbol{v}}}_{\text{sys}})\cdot{{\boldsymbol{k}}}_{\text{obs}}}{c}\right ] , \label{lam1}\ ] ] where @xmath32 is the system velocity with respect to the observer .
the line - of - sight velocity @xmath23 detected by the observer can be found from the relation @xmath33 , \label{lam2}\ ] ] which defines @xmath23 and already incorporates the system velocity .
because the velocities of dust grains and stars in galaxies are very small compared the speed of light , second order terms in @xmath34 can be safely neglected , such that with ( [ lam1 ] ) and ( [ lam2 ] ) , @xmath35 when more scattering events are involved in the photon s trajectory , the relative velocities of each pair of subsequent dust grains have to be taken into account .
if we denote the total number of scattering events with @xmath36 , the velocities of the dust grains with @xmath37 and the propagation directions with @xmath38 ( for @xmath39 ) , we obtain @xmath40 in general , therefore , the inclusion of kinematical information into radiative transfer problems is very complex .
for our problem of dust attenuation in elliptical galaxies , however , we can make one assumption that reduces the complexity considerably : the summation in formula ( [ vvv ] ) , i.e. the common contribution of the dust velocities , is in general negligible with respect to the first term , the contribution of the stellar velocity .
several arguments support this assumption . [ 1 ] it is reasonable to assume that the dust grains have smaller velocities than the stars in elliptical galaxies .
indeed , if the dust grains were to have large velocities , they would collide and heat up .
however , the lion s share of the dust in elliptical galaxies is assumed to be cold ( fich & hodge 1993 ; goudfrooij & de jong 1995 ; wiklind & henkel 1995 ; merluzzi 1998 ) . [ 2 ] scattering off interstellar dust is generally anisotropic , with a larger probability for forward scattering ( see section [ phafu.sec ] ) . in appendix
b we show that the anisotropic nature of scattering contributes to reducing the importance of the dust grain velocity terms in equation ( [ vvv ] ) .
[ 3 ] there is no reason why the velocity of dust grains in an elliptical galaxy would have a preferential direction .
the individual terms in the summation in formula ( [ vvv ] ) will therefore have a random sign , and the summation will in the mean be washed out by multiple scattering . as a consequence
, we can write equation ( [ vvv ] ) simply as @xmath41 .
each photon hence carries the velocity component of the star in the direction of the emission of the photon .
we constructed a monte carlo code to solve our radiative transfer problem .
usually , the main argument against monte carlo methods is that it is computationally rather expensive compared to other methods ( see e.g. baes & dejonghe 2001a ) .
it has other advantages , however , which make it very competitive , in particular in an era when cpu time is not the most stringent limitation anymore .
important qualities of the monte carlo method are the possibility for a proper error analysis and a very wide flexibility .
nice examples of this flexibility are the ability to handle arbitrary geometries ( wolf , fischer & pfau 1998 ; wood & reynolds 1999 ; gordon et al . 2001 ) , the polarization of the scattered radiation ( code & whitney 1995 ; bianchi , ferrara & giovanardi 1996 ) , the clumpiness of the interstellar medium ( witt & gordon 1996 ; bianchi et al .
2000 ) , and the self - consistent heating and re - emission of the absorbed radiation ( wolf , henning & stecklum 1999 ; bianchi , davies & alton 2000 ; misselt et al.2001 ) .
we make use of another possibility of the monte carlo method : its ability to include kinematical information in an elegant way .
this possibility has recently been explored by matthews & wood ( 2001 ) , who studied the effect of dust attenuation on the observed rotation curve in the spiral galaxies .
our modelling opens up the possibility to construct all losvds and hence to investigate the entire observed kinematical structure of galaxies .
the principles of the monte carlo technique are outlined in detail by various authors ( cashwell & everett 1959 ; mattila 1970 ; witt 1977 ; yusef - zadeh , morris & white 1984 ; fischer , henning & yorke 1994 ; bianchi et al .
basically , the routine consists of following the individual trajectory of a very large number of photons through the galaxy .
a photon s history is given by a number of quantities such as the position and propagation direction at birth , the distance the photon travels before it interacts with a dust grain , the kind of this interaction , etc .
each of these quantities is described statistically by a random variable , taken from a particular probability density .
our approach is based on witt ( 1977 ) and bianchi et al .
it includes the use of a continuous three - dimensional cartesian reference system ( no grid ) .
furthermore , we use the classical tricks to optimize the routine : we assign a weight to each photon in order to avoid the loss of photons due to absorption , and we apply the forced first scattering mechanism to improve the statistics of the scattered radiation ( cashwell & everett 1959 ; witt 1977 ) .
the novelty of our monte carlo code is that it can calculate both photometric and kinematic data : not only the light profile , but also the losvds and the projected kinematic moments .
it simultaneously calculates these data in three modes : without dust attenuation , with only absorption taken into account and with dust attenuation fully taken into account .
it is a monochromatic code , i.e. it calculates the observables in one single wavelength .
if desired , the wavelength dependence of the results can be investigated by running the code various times at different wavelengths .
we implemented the central wavelengths of the optical @xmath42 , @xmath43 , @xmath44 , @xmath45 , @xmath46 and the near - infrared @xmath47 , @xmath48 and @xmath49 broadbands .
the light profile of the galaxy is constructed by simulating an imaging process .
if a photon leaves the galaxy , we record the position of the last scattering event and the final propagation direction of the photon . because in spherical symmetry , a path can be determined by its projected radius
, we only need to record the projected radius of the photon s final path .
the registration of the photons is done by constructing a histogram of the photons leaving the galaxy as a function of @xmath45 , where the contribution of each photon is measured by its weight ( see witt 1977 ) .
the monte carlo method provides an elegant possibility to include kinematical information .
indeed , we argued that the velocity information that a photon carries is essentially the velocity component of the star in the direction of the emission .
this extra information can easily be included in the monte carlo routine .
the only extension is that we do not only have to generate an initial position and propagation direction for each photon , but also a stellar velocity .
the initial position and the stellar velocity have to be extracted in accordance with the phase space distribution function @xmath50 , which represents the probability of finding a star at the position @xmath4 with velocity @xmath51 . a practical way to do so , is following the strategy of wybo & dejonghe ( 1996 ) , who generated sets of phase space coordinates of stars to create @xmath52-body representations of globular clusters . first
, a random position @xmath53 is determined from the stellar emissivity , and then a random stellar velocity @xmath27 is generated from the three - dimensional probability density @xmath54 , using the acceptance - rejection method ( abramowitz & stegun 1965 ; press et al .
next , we determine a random emission direction @xmath28 , and determine the velocity component of @xmath27 in the direction @xmath28 , i.e.@xmath55 . given these initial conditions , we follow the photon until the exit conditions are satisfied , and calculate the projected radius of the final line of sight .
the photon will then be stacked , according to its weight , in the appropriate bin in a two - dimensional array with axes @xmath45 and @xmath23 .
this process is repeated this for a large number of photons and two - dimensional histograms are formed .
this method can be optimized in two ways .
firstly , we see that the velocity vector @xmath27 does not show up in the monte carlo procedure itself , but only in the classification process .
moreover , we actually do nt need the entire information contained in the velocity @xmath27 : we only need the component @xmath23 of the velocity in the direction @xmath28 .
therefore it is sufficient to generate a random @xmath23 from the spatial losvd @xmath56 , which represents the probability density of line - of - sight velocities for a star at position @xmath53 in the direction @xmath28 . instead of generating a three - dimensional velocity vector for each photon in the beginning of the monte carlo cycle ,
it is hence sufficient to generate one single line - of - sight velocity component .
this saves many random number generations and distribution function evaluations , which are costly processes .
a second optimization eliminates the generation a @xmath23 altogether . instead of generating one single random @xmath23 ,
we can , for each bin in the velocity direction , calculate the probability that the line - of - sight velocity will fall in that bin .
if the boundaries of a bin are given by @xmath57 and @xmath58 , this probability equals @xmath59 for each photon , we assign a weighted value to each velocity bin corresponding to the final projected radius , which gives better statistics than dropping the photon in a single bin
. of course , we are not able anymore to write down a direct connection between the intrinsic moments of the distribution function and the projected moments on the plane of the sky , contrary to the absorption - only case ( paper i ) .
instead , we will calculate the projected moments directly from the obtained losvds .
the projected kinematics we consider are the mean projected velocity @xmath60 , the projected velocity dispersion @xmath61 , the skewness @xmath62 , the kurtosis @xmath63 , and the lowest order gauss - hermite coefficients @xmath64 and @xmath65 .
note , however , that for spherically symmetric non - rotating galaxies all odd moments vanish , i.e. @xmath66 .
obviously , a reliable error analysis is necessary to estimate the accuracy of the results .
but it will be even more important in a later stage .
indeed , we are presently incorporating this radiative transfer mechanism into the qp program ( dejonghe 1989 ) , a code for the dynamical modelling of gravitating systems .
therefore , it is necessary that we can dynamically evaluate the error bars on any observable : the program has to be able to calculate any observable with an preset accuracy , required by either the user or another part of the code itself .
monte carlo methods satisfy this need : they can keep on adding photons until any accuracy requirement is satisfied .
the determination of the errors is easy for observables that are directly proportional to the number of photons detected in the bins , e.g. the intensity . because of
the poisson character of the noise , the error bars on these quantities can be directly calculated from the square root of the number of photons .
the observed kinematics we calculate , however , are not directly proportional to number of photons obtained in the bins , such that we can not adopt this simple procedure . therefore , we estimate the error on our observables using an alternative method .
assume we have calculated a value @xmath67 for a given observable ( either photometrical or kinematical ) after a run with @xmath52 photons .
we divide the total number of photons in @xmath36 subsets , and for each of these subsets , we calculate the corresponding observable @xmath68 .
the uncertainty on @xmath67 is estimated as , @xmath69^{1/2}.\ ] ] typically , we set @xmath70 , and we adopt a minimum number of @xmath71 photons , i.e. @xmath72 photons per subset .
to investigate the effects of dust attenuation on the light profile and the observed kinematics , we will adopt a set of dusty galaxy models , similar to those in paper i and paper ii .
they are spherically symmetric , and consist of a stellar and a dust component .
we adopt a core radius @xmath74kpc , a total mass @xmath75m@xmath76 and a ( constant ) mass - to - light ratio @xmath77@xmath78 , such that the total emitted luminosity equals @xmath79l@xmath76 .
as in paper i , we consider different plummer models , each with a different value for the parameter @xmath80 , which describes the internal anisotropy of the stellar orbits .
we adopt three different models : a radial model ( @xmath81 ) , the isotropic model ( @xmath82 ) and a tangential model ( @xmath83 ) .
these are the same values as in paper i , except that we replace the radial model with @xmath84 by the model with @xmath81 .
the reason for preferring @xmath81 above @xmath84 has a computationally nature : the spatial losvds can be calculated analytically for @xmath81 , whereas this is not possible for the @xmath84 model ( appendix a ) . as shown in section [ genrte.sec ] , a complete characterization of the dust component
requires a specification of the opacity @xmath85 , the albedo @xmath86 and the scattering phase function @xmath87 .
the opacity sets the total amount of dust and its spatial distribution , whereas the other two quantities describe the optical properties at a given distance to the galaxy centre . in order to limit the number of parameters in our model
, we will assume that the optical properties of the dust grains are the same all over the galaxy , such that the albedo and the phase function become independent of @xmath22 .
we will return to this assumption in section 5.3.3 .
we adopt the same family of dust components considered in paper i and paper ii , characterized by the dust number density @xmath88 where @xmath89 is the dust density at the center of the galaxy , and @xmath90 and @xmath91 determine the actual shape of the dust distribution . because the effect of varying @xmath91 and @xmath90 are roughly similar , we restrict ourselves to varying the dust exponent @xmath91 , and we fix the dust core radius to @xmath74 kpc , equal to the stellar core radius .
the variation of @xmath91 still allows us to explore a large variety of star - dust geometries : small values of @xmath91 correspond to centrally concentrated distributions , whereas for large @xmath91 the distribution is shallower ( paper i ) . at a fixed wavelength @xmath92
, the opacity function @xmath93 can be derived from the dust density by multiplication with the ( wavelength - dependent ) dust cross - section @xmath94 . to characterize the dust distribution at a given wavelength ,
we thus need three parameters : the dust exponent @xmath91 , the central dust number density , and the dust cross - section @xmath94 .
it is more convenient , however , to use another , equivalent , set of parameters .
indeed , if the opacity function is known at one particular wavelength , it can be determined at any other wavelength . usually , the _ v_-band is used as the reference wavelength .
we can then write @xmath95 , where @xmath96 is the extinction coefficient , relative to the _ v _ band .
it is then customary to introduce the total _
v_-band optical depth as is defined as the optical depth from the center of the galaxy to the edge .
we want to stress that this definition differs a factor 2 with the definition of the optical depth we adopted in paper i and paper ii , where we defined @xmath97 as the total optical depth along the central line of sight , i.e. from one edge to the other .
the definition of @xmath97 we adopt here , however , is more commonly used , and it will allow the reader to link the obtained results more easily to those found in the literature ( e.g. witt et al . 1992 ; wise & silva 1996 ) .
when we refer to the results of paper i and paper ii , we will adopt the convention used here . ] @xmath98 if we use the triplet @xmath99 as parameters , the opacity at an arbitrary wavelength @xmath92 can be written as @xmath100 we can safely assume that the phase function does not depend independently on the four variables @xmath101 , but only on the angle between the two directions @xmath5 and @xmath12 .
setting @xmath102 , we can write the phase function as @xmath103 .
probably the most widely adopted phase function that can describe anisotropic scattering is the one named after henyey & greenstein ( 1941 ) .
it has been derived empirically , as a description of the scattering of light in reflection nebulae in the galaxy , and has a simple analytical form , @xmath104 this family of phase functions contains the so - called asymmetry parameter @xmath105 , which is a measure for the anisotropy of the scattering . in particular , for @xmath106 the scattering is isotropic , whereas for @xmath107 the scattering is completely forward .
.the adopted optical properties of the dust grains , at the central wavelengths of the standard optical and near - infrared wavebands . tabulated are the relative extinction coefficient @xmath108 , the scattering albedo @xmath109 and the asymmetry parameter @xmath110 . [ cols="^,^,^,^,^",options="header " , ] each term in the summation in expression ( [ vvv ] ) is a inner product of the velocity vector @xmath37 and the relative direction vector @xmath111 . it is obvious that the contribution of one such term does not depend only on the dust grain velocity @xmath112 , but also on the scattering angle .
the distribution of scattering angles will therefore also play a role in the importance of the dust grain velocity terms in equation ( [ vvv ] ) .
if , for example , the propagation direction of a photon is on average only slightly changed during a scattering event , we find that @xmath113 , which reduces the magnitude of the dust grains velocity terms .
we can illustrate this nicely by calculating the statistical distribution of the modulus of the vector @xmath111 , i.e. we will calculate the probability density @xmath114 for @xmath115 . since the directions @xmath116 and @xmath38 represent unit vectors , elementary trigonometry learns that @xmath117 where @xmath91 is the scattering angle .
the scattering angle is distributed according to the phase function @xmath103 , such that @xmath118 with @xmath119 taking values between 0 and 2 .
if the scattering is isotropic , the phase function has the simple form @xmath120 , such that @xmath121 .
larger values for @xmath119 will hence be preferred .
if we adopt henyey - greenstein scattering ( see section [ phafu.sec ] ) , we can calculate the corresponding values for different degrees of anisotropy . in fig .
[ zeta.eps ] , the probability density @xmath114 corresponding to the henyey - greenstein phase function ( [ hg ] ) is shown for different values of anisotropy parameters @xmath105 . in the limit @xmath106
, this function of course reduces to the isotropic equivalent @xmath121 .
if the anisotropy parameter increases , however , smaller values for @xmath119 will gradually become dominant .
we can quantify this trend by calculating the mean value @xmath122 , the median @xmath123 or the value @xmath124 for which @xmath114 reaches its maximum value .
for the isotropic phase function we find @xmath125 , @xmath126 and @xmath127 , whereas for the henyey - greenstein phase function , we obtain @xmath128 \\ \zeta_{\text{med } } = \sqrt{g^3 - 3g+2 } \\
\zeta_{\text{peak } } = \min\left(2,\frac{1-g}{\sqrt{2g}}\right).\end{gathered}\ ] ] in table [ zeta.tab ] we tabulate @xmath122 , @xmath123 and @xmath124 as a function of the asymmetry parameter .
this table confirms that smaller values of @xmath119 quickly start to dominate as @xmath105 grows larger .
in particular , for the anisotropy parameters appropriate in the optical regime , @xmath129 ( see table [ dustprop.tab ] ) , the density @xmath114 clearly favours fairly small values for @xmath119 ( black curve in fig . [ zeta.eps ] ) . as a consequence
, @xmath130 will be fairly small in the majority of the scattering events .
the anisotropic nature of scattering hence contributes to reducing the importance of the dust grain velocity terms in equation ( [ vvv ] ) . | this paper is the third one in a series , intended to investigate how the observed kinematics of elliptical galaxies are affected by dust attenuation . in paper
i and paper ii , we investigated the effects of dust absorption ; here we extend our modelling in order to include the effects of scattering .
we describe how kinematical information can be combined with the radiative transfer equation , and present a monte carlo code that can handle kinematical information in an elegant way .
compared to the case where only absorption is taken into account , we find that dust attenuation considerably affects the observed kinematics when scattering is included .
for the central lines of sight , dust can either decrease or increase the central observed velocity dispersion . the most important effect of dust attenuation , however , is found at large projected radii .
the kinematics at these lines of sight are strongly affected by photons scattered into these lines of sight , which were emitted by high - velocity stars in the central regions of the galaxy .
these photons bias the losvds towards high line - of - sight velocities , and significantly increase the observed velocity dispersion and losvd shape parameters .
these effects are similar to the expected kinematical signature of a dark matter halo , such that dust attenuation may form an alternative explanation for the usual stellar kinematical evidence for dark matter halos around elliptical galaxies .
we apply our results to discuss several other topics in galactic dynamics , where we feel dust attenuation should be taken into account .
in particular , we argue that the kinematics observed at various wavelengths can help to constrain the spatial distribution of dust in elliptical galaxies .
dust , extinction galaxies : elliptical and lenticular , cd galaxies : kinematics and dynamics radiative transfer scattering |
all graphs considered in this paper are undirected and simple , i.e. no loops or multiple edges are allowed . given a graph @xmath2 , @xmath7 is called the @xmath8 @xmath9 @xmath10 of @xmath2 , where @xmath11 with @xmath12 being the degree of vertex @xmath13 @xmath14 , and @xmath15 is the adjacency matrix of @xmath2 .
the @xmath0-@xmath1 of @xmath2 is the largest eigenvalue @xmath3 of its signless laplacian @xmath4 . from spectral graph theory ,
we know that if graph @xmath2 is connected , there is a unit positive eigenvector ( called perron eigenvector ) of @xmath4 corresponding to @xmath3 . in recent years ,
the study of the @xmath0-index of a graph attracted much attention , and the reader may consult @xcite-@xcite . on the study of the @xmath0-index , a hot topic is that given a class of graphs with fixed order , what is the maxima of the @xmath0-index .
denote by @xmath16 , @xmath17 , @xmath18 a complete graph , a cycle and a path of order @xmath6 respectively .
the join @xmath19 of disjoint graphs @xmath2 and @xmath20 is the graph obtained from @xmath21 by joining each vertex of @xmath2 to each vertex of @xmath20 .
it has been conjectured in @xcite that the graph @xmath5 has the maximal adjacency spectral index among all outer - planar graphs . in @xcite , for a connected outer - planar graph of order @xmath22 , it has been shown that @xmath23 . by comparisons from some examples shown in @xcite ,
it appears plausible that the graph @xmath5 has the maximal @xmath0-index among all outer - planar graphs of order @xmath6 . in this paper , we confirm that the graph @xmath5 really has the maximal @xmath0-index among all outer - planar graphs of order @xmath6 .
the reader is referred to @xcite for the facts about outer - planar graphs .
a graph @xmath2 is @xmath24-@xmath25 if it has a planar embedding , called @xmath26 @xmath27 , in which all vertices lie on the boundary of its outer face .
a simple outer - planar graph is ( edge ) @xmath28 if no edge can be added to the graph without violating outer - planarity . in the standard embedding of a maximal outer - planar graph @xmath2 of order @xmath29 , the boundary of the outer face is a hamiltonian cycle ( a cycle contains all vertices ) of @xmath2 , and each of the other faces is triangle . obviously , a maximal outer - planar graph is 2-connected , and in a maximal outer - planar graph , the least vertex degree is at least @xmath30 ( in fact , a maximal outer - planar graph has at least 2 vertices with degree @xmath30 ) . from a nonmaximal outer - planar graph @xmath2 , by adding edges to @xmath2 , a maximal outer - planar graph @xmath31 can be obtained .
denote by @xmath32 the edge number of a graph @xmath2 .
for an outer - planar graph @xmath2 , we have @xmath33 with equality if and only if it is maximal . from spectral graph theory , for a graph @xmath2 , it is known that @xmath34 if @xmath35 .
consequently , when we consider the maxima of the @xmath0-index among outer - planar graphs , it is sufficient to consider the maximal outer - planar graphs directly .
we introduce some notations .
denote by @xmath36 the vertex set and @xmath37 the edge set for a graph @xmath2 .
if there is no ambiguity , we use @xmath38 instead of @xmath39 .
we use @xmath40 to denote the maximum vertex degree of a graph . in a graph
, the notation @xmath41 denotes that vertex @xmath13 is adjacent to @xmath42 .
denote by @xmath43 a complete bipartite graph with one part of size @xmath44 and another part of size @xmath45 .
next we introduce some working lemmas .
* @xcite*[le2,1 ] let @xmath46 be a vertex of a maximal outer - planar graph on @xmath22 vertices .
then @xmath47 .
* @xcite*[le2,3 ] let @xmath2 be a graph
. then @xmath48 . *
@xcite*[le2,4 ] let @xmath2 be a connected graph containing at least one edge .
then @xmath49 with equality if and only if @xmath50 .
[ le3,1 ] let @xmath2 be a maximal outer - planar graph with order @xmath51 and @xmath52
. then @xmath53 .
let @xmath62 ( see fig .
3.1 ) . by lemma [ le2,4 ]
, we see that @xmath63 . from this
, we see that among all outer - planar graphs of order @xmath6 , the maxima of the @xmath0-index is more than @xmath6 . combining with lemma [ le3,1 ]
, we find that among outer - planar graphs of order @xmath51 , the maximal degree of the graph with the maxima of the @xmath0-index is more than @xmath64 .
next , we consider the outer - planar graphs of order @xmath51 with @xmath65 @xmath66 respectively .
( 545,235 ) ( 21,146)(21,178)(44,201)(44,201)(68,225)(102,225)(102,225)(135,225)(159,201 ) ( 159,201)(183,178)(183,146)(183,146)(183,113)(159,90)(159,90)(135,67)(102,67 ) ( 102,67)(68,67)(44,90)(44,90)(20,113)(21,146 ) ( 99,69 ) ( 182,143 ) ( 99,69)(140,106)(182,143 ) ( 171,186 ) ( 171,186)(135,128)(99,69 ) ( 22,133 ) ( 99,69)(60,101)(22,133 ) ( 25,174 ) ( 99,69)(62,122)(25,174 ) ( 47,204 ) ( 99,69)(73,137)(47,204 ) ( 115,188 ) ( 132,188 ) ( 96,188 ) ( 158,90 ) ( 46,89 ) ( 96,54)@xmath67 ( 29,79)@xmath68 ( 1,133)@xmath69 ( 3,179)@xmath70 ( 27,209)@xmath71 ( 175,188)@xmath72 ( 188,145)@xmath73 ( 164,82)@xmath74 ( 95,24)@xmath75 ( 383,141)(383,171)(404,192)(404,192)(425,214)(456,214)(456,214)(486,214)(507,192 ) ( 507,192)(529,171)(529,141)(529,141)(529,110)(507,89)(507,89)(486,67)(456,67 ) ( 456,68)(425,68)(404,89)(404,89)(383,110)(383,141 ) ( 386,124 ) ( 451,68 ) ( 386,124)(418,96)(451,68 ) ( 398,185 ) ( 451,68)(424,127)(398,185 ) ( 424,207 ) ( 451,68)(437,138)(424,207 ) ( 491,203 ) ( 451,68)(471,136)(491,203 ) ( 526,154 ) ( 451,68)(488,111)(526,154 ) ( 522,113 ) ( 451,68)(486,91)(522,113 ) ( 424,207)(457,205)(491,203 ) ( 498,81 ) ( 513,164 ) ( 400,129 ) ( 463,214 ) ( 400,152 ) ( 400,141 ) ( 505,173 ) ( 498,183 ) ( 456,220)@xmath67 ( 414,215)@xmath68 ( 446,54)@xmath76 ( 409,84 ) ( 383,192)@xmath69 ( 345,124)@xmath77 ( 375,76)@xmath78 ( 501,71)@xmath79 ( 530,111)@xmath80 ( 530,155)@xmath81 ( 492,210)@xmath74 ( 445,26)@xmath2 ( 208,-11)fig . 3.1 .
@xmath75 , @xmath2 note that @xmath94 , @xmath95 .
then @xmath96 from ( 1 ) and ( 2 ) , we also get that @xmath97 by ( 4)-(3 ) , we get that @xmath98@xmath99 it follows that @xmath100=(n-10)x_k+3(x_k - x_{0})+\sum_{3\leq i\leq n-3 , i\neq k } x_i .
\hspace{0.5 cm } ( 5)\ ] ] ( 5 ) tells us that if @xmath101 , then @xmath102 .
let @xmath103 .
note the relation between the rayleigh quotient and the largest eigenvalue of a non - negative real symmetric matrix , and note that @xmath104 .
it follows that if @xmath101 , then @xmath105 . because @xmath106 , if @xmath101 , then @xmath107 .
( 665,304 ) ( 49,208 ) ( 10,278 ) ( 49,208)(29,243)(10,278 ) ( 93,278 ) ( 49,208)(71,243)(93,278 ) ( 11,230 ) ( 94,233 ) ( 49,303 ) ( 41,182)@xmath108 ( 238,207 ) ( 195,262 ) ( 220,294 ) ( 238,207)(229,251)(220,294 ) ( 282,226 ) ( 206,223 ) ( 20,175.1)(20,175.3)(20,175.2 ) ( 267,293 ) ( 238,207)(252,250)(267,293 ) ( 289,262 ) ( 238,207)(263,235)(289,262 ) ( 220,294)(213,259)(206,223 ) ( 230,181)@xmath109 ( 434,209 ) ( 389,270 ) ( 434,209)(411,240)(389,270 ) ( 458,298 ) ( 434,209)(446,254)(458,298 ) ( 482,268 ) ( 434,209)(458,239)(482,268 ) ( 476,230 ) ( 414,298 ) ( 389,270)(423,284)(458,298 ) ( 395,230 ) ( 426,182)@xmath110 ( 268,-9)fig
@xmath108-@xmath110 ( 93,278)(51,278)(10,278 ) ( 4,255)(4,235)(18,221)(18,221)(32,207)(52,207)(52,207)(71,207)(85,221 ) ( 85,221)(100,235)(100,255)(4,255)(4,274)(18,288)(18,288)(32,303)(52,303 ) ( 52,303)(71,303)(85,288)(85,288)(100,274)(100,255 ) ( 195,253)(195,233)(209,219)(209,219)(223,205)(243,205)(243,205)(262,205)(276,219 ) ( 276,219)(291,233)(291,253)(195,253)(195,272)(209,286)(209,286)(223,301)(243,301 ) ( 243,301)(262,301)(276,286)(276,286)(291,272)(291,253 ) ( 388,256)(388,236)(402,222)(402,222)(416,208)(436,208)(436,208)(455,208)(469,222 ) ( 469,222)(484,236)(484,256)(388,256)(388,275)(402,289)(402,289)(416,304)(436,304 ) ( 436,304)(455,304)(469,289)(469,289)(484,275)(484,256 ) ( 569,253)(569,233)(582,219)(582,219)(596,206)(616,206)(616,206)(635,206)(649,219 ) ( 649,219)(663,233)(663,253)(569,253)(569,272)(582,286)(582,286)(596,300)(616,300 ) ( 616,300)(635,300)(649,286)(649,286)(663,272)(663,253 ) ( 577,280 ) ( 654,280 ) ( 577,280)(615,280)(654,280 ) ( 617,300 ) ( 662,247 ) ( 649,220 ) ( 587,216 ) ( 570,247 ) ( 570,247)(612,264)(654,280 ) ( 570,247)(609,234)(649,220 ) ( 2,104)(2,84)(15,70)(15,70)(29,57)(49,57)(49,57)(68,57)(82,70 ) ( 82,70)(96,84)(96,104)(2,104)(2,123)(15,137)(15,137)(29,151)(49,151 ) ( 49,151)(68,151)(82,137)(82,137)(96,123)(96,104 ) ( 10,132 ) ( 86,132 ) ( 10,132)(48,132)(86,132 ) ( 95,101 ) ( 84,73 ) ( 14,73 ) ( 2,101 ) ( 49,151 ) ( 621,207 ) ( 49,57 ) ( 14,73)(50,103)(86,132 ) ( 14,73)(54,87)(95,101 ) ( 14,73)(49,73)(84,73 ) ( 14,73)(12,103)(10,132 ) ( 39,31)@xmath111 ( 232,80 ) ( 144,103)(144,83)(157,69)(157,69)(171,56)(191,56)(191,56)(210,56)(224,69 ) ( 224,69)(238,83)(238,103)(144,103)(144,122)(157,136)(157,136)(171,150)(191,150 ) ( 191,150)(210,150)(224,136)(224,136)(238,122)(238,103 ) ( 237,103 ) ( 191,57 ) ( 156,71 ) ( 144,101 ) ( 190,149 ) ( 156,135 ) ( 225,135 ) ( 156,135)(190,135)(225,135 ) ( 191,57)(173,96)(156,135 ) ( 191,57)(208,96)(225,135 ) ( 191,57)(214,80)(237,103 ) ( 181,30)@xmath112 ( 287,104)(287,84)(300,70)(300,70)(314,57)(334,57)(334,57)(353,57)(367,70 ) ( 367,70)(381,84)(381,104)(287,104)(287,123)(300,137)(300,137)(314,151)(334,151 ) ( 334,151)(353,151)(367,137)(367,137)(381,123)(381,104 ) ( 191,57)(167,79)(144,101 ) ( 300,137 ) ( 366,137 ) ( 300,137)(333,137)(366,137 ) ( 335,151 ) ( 380,112 ) ( 374,82 ) ( 347,60 ) ( 311,63 ) ( 291,84 ) ( 288,112 ) ( 288,112)(327,125)(366,137 ) ( 288,112)(334,112)(380,112 ) ( 288,112)(299,88)(311,63 ) ( 288,112)(317,86)(347,60 ) ( 288,112)(331,97)(374,82 ) ( 324,29)@xmath113 ( 430,103)(430,83)(443,70)(443,70)(456,57)(476,57)(476,57)(495,57)(508,70 ) ( 508,70)(522,83)(522,103)(430,103)(430,122)(443,135)(443,135)(456,149)(476,149 ) ( 476,149)(495,149)(508,135)(508,135)(522,122)(522,103 ) ( 442,135 ) ( 508,135 ) ( 442,135)(475,135)(508,135 ) ( 475,150 ) ( 520,108 ) ( 515,82 ) ( 500,64 ) ( 472,58 ) ( 441,75 ) ( 430,108 ) ( 441,75)(441,105)(442,135 ) ( 441,75)(474,105)(508,135 ) ( 441,75)(480,92)(520,108 ) ( 441,75)(478,79)(515,82 ) ( 441,75)(470,70)(500,64 ) ( 467,30)@xmath114 ( 575,106)(575,87)(588,74)(588,74)(601,61)(620,61)(620,61)(638,61)(651,74 ) ( 651,74)(665,87)(665,106)(575,106)(575,124)(588,137)(588,137)(601,151)(620,151 ) ( 620,151)(638,151)(651,137)(651,137)(665,124)(665,106 ) ( 610,183)@xmath115 ( 587,136 ) ( 652,136 ) ( 587,136)(619,136)(652,136 ) ( 620,150 ) ( 664,107 ) ( 655,78 ) ( 629,62 ) ( 598,68 ) ( 580,86 ) ( 576,112 ) ( 598,68)(587,90)(576,112 ) ( 598,68)(592,102)(587,136 ) ( 598,68)(625,102)(652,136 ) ( 598,68)(631,88)(664,107 ) ( 598,68)(626,73)(655,78 ) ( 610,30)@xmath116 ( 570,247)(616,247)(662,247 ) ( 570,247)(595,227)(621,207 ) when @xmath119 , then @xmath120 . from case 1 , we know that for @xmath119 , @xmath121 . when @xmath122 , @xmath2 is isomorphic to one in @xmath123
; when @xmath124 , @xmath2 is isomorphic to one in @xmath125 ( see fig . 3.6 ) . by computation with computer , we get the approximation of the @xmath0-index for each @xmath117 @xmath126 ( see table 4 ) . combining with the results about @xmath127 for order @xmath122 , @xmath128 in the proof of lemma [ le3,2 ] , by a simple comparison
, we get that for each @xmath117 ( @xmath129 and @xmath130 ) of order @xmath6 ( @xmath131 ) , @xmath132 . from the narration in section 2 , we know that by adding edges , a maximal outer - planar graph can be obtained from a nonmaximal outer - planar graph ; and know that for a graph @xmath2 , @xmath34 if @xmath35
. then the result follows from lemma [ le3,4 ] .
this completes the proof . | the @xmath0-@xmath1 of graph @xmath2 is the largest eigenvalue @xmath3 of its signless laplacian @xmath4 . in this paper
, we prove that the graph @xmath5 has the maximal @xmath0-index among all outer - planar graphs of order @xmath6 .
* ams classification : * 05c50 * keywords : * signless laplacian ; @xmath0-index ; outer - planar graph 18.6pt |
nlo qcd corrections to one and two - jet production cross sections and distributions , for @xmath6 exchange , are implemented in the fully differential @xmath7 jets event generators mepjet @xcite and disent @xcite . full neutral current exchange ( @xmath8 and/or @xmath9 ) is now available in mepjet 2.0 for tree level cross sections up to @xmath10 , _ i.e. _ up to 4-jet final states .
nlo corrections to two - jet production including @xmath9 and @xmath11 exchange will be available soon . both programs allow to study arbitrary experimental cuts and jet definition schemes .
previous calculations @xcite were limited to a jade type algorithm and important single jet - mass effects were neglected . because of the disagreement with these earlier calculations a comparison of the two new calculations is particularly important . in table
[ table_comp ] we compare results from mepjet and disent for 2- and 3-jet rates at hera ( @xmath12 gev , @xmath13 gev ) in the @xmath14-scheme with @xmath15 and a hard scattering scale @xmath16 .
parton distributions are used @xcite , we fix @xmath18 , and require @xmath19 gev@xmath20 . factorization and renormalization scales are set to @xmath21 .
no further cuts are imposed in the first column while in the second column reconstructed jets must have transverse momenta @xmath22 gev and pseudo - rapidities @xmath23 .
we find agreement at the 23% level , which is slightly larger than , but still compatible with the statistical errors determined by the monte carlo integration routines .
for all practical applications this agreement is satisfactory .
l|cc|cc & & & & + & & & & + + & 395.6 @xmath24 1.5 pb & 392.6 @xmath24 0.5 pb & 339.0 @xmath24 0.4 pb & 337.5 @xmath24 0.4 pb + & 33.0 @xmath24 0.5 pb & 32.5 @xmath24 0.2 pb & 27.7 @xmath24 0.2 pb & 27.2 @xmath24 0.2 pb + & 561 @xmath24 7 pb & 559 @xmath24 6 pb & 463 @xmath24 8 pb & +
the internal structure of jets is first modeled at nlo where two massless partons may be recombined to form a jet .
the dependence of jet cross sections on the details of the internal jet structure can be investigated by studying different recombination schemes : the @xmath25-scheme where parton 4-momenta are added to form massive jet 4-momenta , and the @xmath26 and @xmath27-schemes in which the jets are made massless by rescaling either the 3-momentum or the energy of the resulting cluster .
we use mepjet for numerical studies with mrs d- parton distribution functions @xcite .
the renormalization and factorization scales are set to one half the sum of parton transverse momenta in the breit frame .
a minimal set of kinematical cuts is imposed .
we require 40 gev@xmath28 gev@xmath20 , @xmath29 , an energy cut of @xmath30 gev on the scattered lepton , and pseudo - rapidities @xmath31 for the scattered lepton and jets .
also , jets must have transverse momenta of at least 2 gev in the lab and the breit frame . within these
general cuts four different jet definition schemes are considered .
i ) a cone algorithm with a jet separation cut of @xmath32 and a minimum jet transverse momentum of @xmath22 gev in the lab frame .
ii ) the @xmath14 algorithm ( in the breit frame ) as defined in ref . @xcite with a hard scattering scale @xmath33 gev@xmath20 and a resolution parameter @xmath15 for resolving the macro - jets .
in addition , jets are required to have a minimal transverse momentum of 5 gev in the breit frame .
iii ) the @xmath11-scheme where parton / cluster pairs ( including the proton remnant ) with invariant mass squared , @xmath34 are recombined @xcite .
iv ) the `` jade '' algorithm @xcite which is obtained from the @xmath11-scheme by replacing the invariant definition @xmath35 by @xmath36 . in the @xmath11 and jade schemes we set @xmath37 .
l|ccccc & & & & & + & & ( @xmath25 ) & ( @xmath25 ) & ( @xmath26 ) & ( @xmath27 ) + + & 1107 pb & 1047 pb & 1203 pb & 1232 pb & 1208 pb + @xmath14 & 1067 pb & 946 pb & 1038 pb & 1014 pb & 944
pb + @xmath11 & 1020 pb & 2061 pb & 2082 pb & 1438 pb & 1315 pb + & 1020 pb & 1473 pb & 1507 pb & 1387 pb & 1265 pb + the resulting cross sections ( see table [ table_dijet ] ) show large nlo corrections for the jade and @xmath11-schemes while @xmath38-factors close to unity are found for the @xmath14 and cone algorithms .
another disadvantage of the @xmath11 and jade schemes is their strong recombination scheme dependence . since even in a nlo calculation
the internal structure of a jet is only calculated at tree level , these strong variations point to large uncertainties from two - loop effects in the @xmath11 and jade schemes . in both schemes widely separated but relatively soft partons tend to be clustered to what is then defined as a jet , even though these clusters can be very massive and quite distinct from the pencil - like and low - mass objects which one starts out with in the parton model or at lo @xcite .
these differences then result in the large higher order corrections to jet cross sections shown in table [ table_dijet ] .
these effects are much smaller in the cone and the @xmath14 schemes which are hence favored for precision qcd studies in dis .
recently much interest has been focused on the small bjorken-@xmath4 region , where one would like to distinguish bfkl evolution @xcite , which resums the leading @xmath39 terms , from the more traditional dglap evolution equation @xcite , which resums leading @xmath40 terms .
bfkl evolution can be enhanced and dglap evolution suppressed by studying dis events which contain an identified jet of longitudinal momentum fraction @xmath41 ( in the proton direction ) which is large compared to bjorken @xmath4 @xcite .
when tagging a forward jet with @xmath0 this leaves little room for dglap evolution while the condition @xmath1 leaves bfkl evolution active .
this leads to an enhancement of the forward jet production cross section proportional to @xmath42 over the dglap expectation .
= 4.0 in = 3.0 in a conventional fixed order qcd calculation up to @xmath5 does not yet contain any bfkl resummation and must be considered a background for its detection ; one must search for an enhancement in the forward jet production cross section above the expectation for two- and three - parton final states .
the full calculation of the forward jet inclusive cross section in dis , at @xmath5 , has been performed in ref .
ordinarily , such a calculation would contain 3-parton final states at tree level , 1-loop corrections to 2-parton final states and 2-loop corrections to 1-parton final states .
however , these 2-loop contributions vanish identically , once the condition @xmath43 is imposed . the remaining 2-parton and 3-parton differential cross sections , and the cancellation of divergences between them , are the same as those entering a calculation of 2-jet inclusive rates .
these elements are already implemented in the mepjet program , which , therefore , can be used to determine the inclusive forward jet cross section at @xmath5 . in fig .
[ fig : h1comp ] numerical results are compared with recent data from h1 @xcite . here the conditions @xmath0 and @xmath1 are satisfied by selecting events with forward jets ( in the angular region @xmath44 ) with @xmath45 clearly h1 observes substantially more forward jet events than expected from nlo qcd . however , since h1 accepted jets of rather low transverse momentum , @xmath46 gev , the comparison of our parton level calculation to the data may be subject to sizable hadronization corrections .
a recent bfkl calculation @xcite ( dashed histogram ) agrees better with the data , but here the overall normalization is uncertain and the agreement may be fortuitous .
a very rough estimate of the uncertainty of the nlo calculation is provided by the two dotted lines which correspond to variations by a factor 10 of the renormalization and factorization scales @xmath47 and @xmath48 .
very similar results are found by zeus @xcite .
we conclude that the hera data show evidence for bfkl dynamics in forward jet events via an enhancement in the observed forward jet cross section above nlo expectations .
however , additional data , with harder forward jets , are needed to make the comparison of qcd calculations with data more reliable .
we thank m. seymour for his help comparing mepjet and disent .
this research was supported by the university of wisconsin research committee with funds granted by the wisconsin alumni research foundation and by the u. s. department of energy under grant no .
de - fg02 - 95er40896 .
the work of e. m. was supported in part by dfg contract ku 502/5 - 1 .
t. brodkorb and j.g .
krner , z. phys .
* c54 * ( 1992 ) 519 ; t. brodkorb and e. mirkes , z. phys .
* c66 * ( 1995 ) 141 ; d. graudenz , phys . lett . *
b256 * ( 1992 ) 518 ; phys . rev .
* d49 * ( 1994 ) 3291 .
martin , w.j .
stirling , r.g .
roberts , phys .
* b306 * ( 1993 ) 145 .
mueller , nucl .
b ( proc . suppl . )
* 18c * ( 1990 ) 125 ; j. phys . *
g17 * ( 1991 ) 1443 ; j. kwiecinski , a.d .
martin and p.j .
sutton , phys .
* d46 * ( 1992 ) 921 ; w.k .
tang , phys . lett .
* b278 * ( 1992 ) 363 ; j. bartels , a. de roeck and m. loewe , z. phys .
* c54 * ( 1992 ) 635 . | dijet production in dis is an important laboratory for testing our understanding of perturbative qcd .
flexible nlo monte carlo programs allow to investigate general jet definition schemes . for forward jet production at @xmath0 and @xmath1 the nlo predictions fall well below hera data , thus providing evidence for bfkl dynamics .
multi - jet production in dis is an important topic of study at hera .
good event statistics allow for precise measurements and thus for a variety of tests of our understanding of qcd dynamics .
such tests include a ) the determination of @xmath2 from dijet production over a range of renormalization scales @xcite , b ) the measurement of the gluon density in the proton ( via @xmath3 ) @xcite , c ) forward jet production in the low-@xmath4 regime as a signal of bfkl dynamics @xcite , d ) the determination of @xmath2 and power corrections in dis event shapes @xcite . for quantitative studies
one clearly needs to compare data with calculations which include next - to - leading order ( nlo ) qcd corrections .
here we concentrate on issues related to dijet production at hera .
the present status of nlo monte carlo programs is reviewed .
second , we compare predictions in different jet definition and parton recombination schemes . finally , the @xmath5 predictions for forward jet production are discussed . |
almost two decades ago it was shown that the generalized leading logarithmic approximation in the regge limit of qcd can be described by a spin chain @xcite , which was later shown in @xcite to be completely integrable , i.e. have an infinite tower of conserved charges allowing an analytic diagonalization of its hamiltonian .
the spin chain is defined by a nearest - neighbor interaction given by the bfkl hamiltonian , which has been known since the 70 s as the kernel of the bfkl equation @xcite , giving the leading logarithmic approximation to scattering with color singlet exchange . from the viewpoint of qcd
it is surprising that such an infinite amount of symmetry is found in the regge limit .
the simplest explanation probably lies in that the regge limit of qcd coincides to lla with the regge limit of the maximally supersymmetric extension of pure yang - mills , @xmath0 super yang mills , for which there is a large amount of evidence that it is integrable ( in the large @xmath1 limit ) to all orders in the gauge coupling @xcite .
having an exact solution of @xmath0 sym will induce a solution in the regge limit , which would explain its integrability . to date , however , there is little understanding of how the integrable structure of @xmath0 implies the symmetries of bfkl and its extensions .
part of the difficulty in making the connection lies in that the symmetries in the regge limit are naturally understood in terms of a two dimensional effective theory ( see for example @xcite ) , in which the longitudinal components of the momenta have decoupled and the dynamics of the theory have been reduced to the transverse plane .
an essential component of the symmetry found in this limit is given by the two - dimensional conformal @xmath2 symmetry discovered by lipatov @xcite .
the running of the coupling constant does not enter the lla , and one can thus suspect that this symmetry is a consequence of the classical conformal symmetry of the qcd lagrangian . when effects from the running of the coupling are included at nlla , the symmetry is broken . in @xmath0 sym ,
by contrast , the coupling does not run , and recent work indeed seems to indicate that the @xmath2 symmetry remains exact at nlla @xcite .
it therefore seems natural to understand the @xmath2 as what remains of the four - dimensional conformal symmetry of @xmath0 when taking the regge limit .
the infinite yangian algebra @xcite of conserved charges of @xmath0 is generated by two copies of the superconformal group @xmath3 @xcite , the first being the ordinary superconformal symmetry of the lagrangian , while the second is a novel , non - lagrangian symmetry , coined dual superconformal symmetry @xcite . if it were possible , in the same way
that the ordinary conformal symmetry would seem to imply the ordinary @xmath2 , to identify a non - trivial remnant of the dual conformal symmetry in the regge limit , one could quite possibly explain the integrability of the generalized lla as being generated by the ordinary @xmath2 and this new dual symmetry .
of this article , used in @xcite to find the spectrum of an integrable , open spin chain in the octet channel . ] in @xcite we took a step in this direction , identifying a dual @xmath2 symmetry of bfkl .
however , as discussed in that article , the most straightforward application of the dual @xmath2 turns out to be broken by ir effects , which produce anomalous terms when iterating the bfkl equation . in this article
, we discuss whether it is possible that such terms can be reabsorbed into a deformation of the symmetry .
this would parallel the way that the dual conformal symmetry of @xmath0 is made exact by correcting for ir anomalies @xcite .
we also initiate a study of the algebra that is obtained by commuting the ordinary and dual @xmath2 symmetries , extracting a piece of the dual symmetry which is close to a bi - local form frequently appearing in yangian algebras .
although not entirely surprising , the serre relations , which define a yangian algebra , are however not satisfied .
in this section we will motivate the definition of the dual @xmath2 symmetry .
since we are looking for what is left in the regge limit of the dual conformal symmetry of @xmath0 scattering amplitudes , let us first recall how this four - dimensional symmetry is constructed .
if one introduces a set of variables @xmath4 , related to the external ( all taken as incoming ) particle momenta @xmath5 , @xmath6 , through p_i = x_i - x_i+1x_i i+1 , [ eq : xi ] the amplitudes exhibit , at tree level , covariance under a superconformal group acting on the @xmath7 variables in the same way as the ordinary superconformal symmetry acts on spatial coordinates @xcite .
loop integrals are then found to be formally ( before ir regularization ) invariant under the same symmetry . as an example , which will help motivate our version of the symmetry in the regge limit , let us look at the one - loop correction to the four - particle amplitude . in @xmath0 sym
it is simply proportional to the scalar box diagram @xcite , shown in figure [ fig : scalarbox ] .
super yang mills , in terms of incoming momenta @xmath5 , and an integral over the loop momentum @xmath8 . ] the integral which one obtains when evaluating this diagram is , which becomes , after introducing @xmath7-variables , and defining a new integration variable @xmath9 through @xmath10 , .
[ eq : xint ] if the external particles are on - shell , with @xmath11 , this integral diverges , but ignoring this for the moment one has a formal conformal symmetry acting on the @xmath7 .
this is easily seen by noting that under conformal inversions @xmath12 implying that the squared differences @xmath13 , and the measure @xmath14 transform as x ^2_ij , d^4x_i , [ eq : diffandmeasure ] respectively .
let us now make contact with the regge limit . in
non - abelian gauge theories the gluon reggeizes , meaning that when the quantum numbers of the external states so permit , the amplitudes can be described in terms of effective particles propagating in the @xmath15-channel called reggeized gluons , which carry the quantum numbers of ordinary gluons , but for which the propagators ( in feynman gauge ) have been replaced as ( ) ^(t ) , [ eq : reggeprop ] where @xmath16 is called the gluon regge trajectory , and the scale @xmath17 is of order @xmath15 but whose value does otherwise not matter at lla . to leading order ,
the gluon trajectory reads ( ^2 ) = -d^2 , [ eq : trajectoryreggegluon ] where @xmath18 , with @xmath19 the gauge theory coupling and @xmath1 the number of colors .
the form of the trajectory reflects , as mentioned in the introduction , that the dynamics are reduced to the two - dimensional transverse plane in the regge limit , so that @xmath20 , where @xmath21 is the transverse part of the momentum exchanged in the scattering process .
the substitution incorporates an all - loop resummation , but knowing that the gluon reggeizes it is enough to perform a one - loop computation to calculate the trajectory . expanding in the coupling ,
it is clear that the one - loop amplitude for gluon quantum number exchange should in the regge limit be given by the tree amplitude multiplied by @xmath22 .
perhaps the simplest way to perform this calculation and obtain , as reviewed nicely in @xcite , is to use unitarity , the cutkosky rules and dispersion relations , to obtain the 1-loop amplitude as a product of two tree amplitudes .
this corresponds to setting the upper and lower propagators of figure [ fig : scalarbox ] on - shell .
doing so removes the propagators @xmath23 and @xmath24 from the denominator of and introduces two delta functions , thereby reducing the dimensions of the integral from four to two .
furthermore , in the regge limit , the integral is dominated by the transverse components of the momenta , replacing the @xmath7 by two - dimensional vectors @xmath25 . taking into consideration the tree - level factors ( which corrects numerator of ) one then obtains the reggeized gluon trajectory as ( ^2_24)=-d^2 _ i .
[ eq : trajectoryreggegluonk1x ] the same representation of the trajectory would have been obtained by shifting the integration variable @xmath26 of as @xmath27 , where @xmath28 is again taken as the new integration variable , and performing the replacement @xmath29 .
this last replacement is precisely what one would obtain from a two - dimensional version of .
furthermore , the integral representation has a formal ( ignoring ir divergences ) two - dimensional inversion symmetry _
i [ eq : xinversion ] since the squared differences @xmath30 and the measure transforms , similarly to , as ^2_ij , d^2_i .
[ eq : diffandmeasure2d ] this suggest that a remnant of the dual conformal symmetry can be found in the effective two - dimensional theory that arises in the regge limit by simply introducing @xmath25-variables , related to the incoming , transverse momenta @xmath31 by _
-_i+1 , [ eq : xi2d ] and then acting on the @xmath25-variables in the standard way . together with the inversions ,
one also has translation symmetry ( since the @xmath25 only enter as the differences ) as well as rotation symmetry . in total , this constitutes the group @xmath2 .
since our main motivation for looking at the dual @xmath2 was to try to understand the integrability of the generalized lla spin chain , we should see what form it takes for color singlet exchange . as a first step one
is led to search for a dual @xmath2 symmetry of the bfkl equation .
this equation can be interpreted , in the effective two - dimensional theory that arises in the regge limit , as giving the leading approximation to the scattering amplitude @xmath32 of 2 to 2 reggeized gluons , with color singlet exchange in the @xmath15-channel .
if we label the incoming momenta of these reggeized gluons as @xmath33 , we can introduce @xmath25-variables using , in terms of which the bfkl equation takes the form @xmath34 with the integration kernel given by k(_1 , _ 2 , _ 3 , _
i ) = + ^(2)(_i4 ) , [ eq : kernelx ] where k_r ( _ 12 , _ 23 ; _ 3i , _
i1 ) = - [ eq : kernelrx ] and @xmath35 is the gluon regge trajectory .
the variable @xmath36 is , despite the notation , which we keep for historical reasons , not the gluon trajectory , but rather the mellin transform conjugate of @xmath37 . in @xcite we showed that is invariant under a formal dual @xmath2 symmetry , which acts on the @xmath7 variables in the same way that the @xmath2 of lipatov acts on impact parameter space .
the symmetry requires that @xmath36 be invariant , which either means that we should not transform the longitudinal components of the momenta at all , or should require that the scale @xmath17 transform in the same way as @xmath38 .
the dual @xmath2 is only a formal symmetry of the bfkl equation and the gluon regge trajectory , and is expected to be affected by the ir divergences present . in bfkl
the divergences cancel , so one could hope that the symmetry remains unaltered , but as mentioned in @xcite this is unfortunately not the case .
the gluon regge trajectory is given , in dimensional regularization , by ( ^2_12)=-(4 ) ^2d^2 - 2 _
i -(4e^-)^ ( - ) . [
eq : trajectorydimreg ] in the bfkl equation the poles in @xmath39 and the logarithms in the scale @xmath40 cancel , but factors such as @xmath41 do not cancel and rather add up to an anomalous , non - invariant expression . perhaps the most direct way to see this is to write out the first orders to the solution @xmath32 of the equation , since a consequence of the symmetry , if it where to remain exact , is that f ^2_2_4 ^ 2 f [ eq : ftransform ] under dual inversions .
the solution to bfkl can be constructed , order by order , by iterating the equation , regularizing and canceling infrared divergences at each order . in general , the solution can be written f = f_1 ( ) ^(2)(_24)+f_2 ( ) , where @xmath42 , @xmath43 is obtained by repeatedly iterating the trajectory part of the kernel , and where @xmath44 must be invariant under dual inversions in order for to hold .
the lowest order of @xmath45 is trivial to calculate , simply being the result of applying the kernel to the inhomogenous delta - function term of . at the next order ,
apart from integrals similar to the trajectory , non - trivial integrals of the form appear , but as shown in appendix [ app : nfintegral ] their form can be considerably restricted by studying their symmetry properties under dual @xmath2 transformations .
one finds that f_2 ( ) = ( 1 + u -v ) + + , [ eq : f2 ] where u = v = are the two independent conformal invariants that can be formed from @xmath46 .
we see that the lowest order is invariant , while the next - to - lowest order , where the trajectories start to matter , is not , containing an anomalous logarithmic term . , but it is simple to check numerically that even for such configurations @xmath47 is not invariant at higher orders . ]
such a logarithm is not completely unexpected , though , since the dual conformal symmetry of @xmath0 sym is broken by infrared divergences . in that case
it seems , fortunately , to be broken in a controlled way .
an @xmath48-particle color - ordered @xmath0 scattering amplitude @xmath49 can always be factorized as _
n = _ n , _
n , where @xmath50 is the tree - level amplitude .
there is evidence that the finite part of @xmath51 satisfies to all orders an `` anomalous ward identity '' @xcite given by k^ _
( ) _ i=1^n x_i , i+1^ , [ eq : ward ] where @xmath52 is the cusp anomalous dimension .
if we instead apply the generator @xmath53 ( to be defined in the next section ) of the dual @xmath2 corresponding to the special conformal transformations to @xmath47 we obtain ( to order @xmath54 ) i k_^(d ) f_2 = 2_1 ( ) ( x_1 + x_3 - x_2 - x_4 ) , [ eq : ward2d ] where @xmath55 here is a two - dimensional index , where @xmath56 , and where we omit the bold font from the @xmath25 when writing them out in components .
this is a rather simple structure , which stems from the fact that the anomalous term is a simple logarithm , to lowest order , while the anomalous terms in the finite part of @xmath0 amplitudes consist of squares of logarithms .
the important question is now if this structure is exact to all loops , as seems to be the case in the four - dimensional gauge theory .
unfortunately , it is not .
perhaps the simplest way to see this analytically is to consider the forward limit , defined by @xmath57 , or equivalently @xmath58 , where calculations simplify considerably .
it is then not difficult to calculate the third order of @xmath47 . in the forward limit , @xmath59 and @xmath60 , and becomes i k_^(d ) f_2 , = 2 _ 1 ( ) ( 2 x_1 ^ - x_2 ^ - x_4^ ) , [ eq : ward2dforward ] for this to hold to all orders , in the forward limit @xmath61 would have to be of the form
_ 1 ( ) ( ) + _ 2 ( ) , where @xmath62 and @xmath63 only depend on @xmath54 . instead , one obtains f_2 , ( ) = + ( ) + + .
[ eq : f2forward ] as a result does not hold to all orders in bfkl .
this does not mean that the dual @xmath2 symmetry is broken beyond repair , however .
it simply means that the way that the representation of the symmetry is deformed by the coupling is not as simple as in the case of @xmath0 .
one way to understand that the structure of the representation of the dual @xmath2 may be complicated even though it may stem from the simple dual conformal symmetry is to visualize , in the standard way , the iteration of the bfkl equation as a sum over effective ladder diagrams .
these diagrams can then , in turn , be written , by unitarity , as a product of tree - level amplitudes . summing over the rungs of the ladder
will then sum over products of tree - amplitudes with a different number of legs , and consequently different dual conformal weights . an alternative way to see
this is to regularize the bfkl equation in such a way that the dual @xmath2 invariance is preserved by adapting the procedure of @xcite , introducing different scales which are allowed to transform independently under inversions .
then one finds that new scales must be introduced when iterating the equation .
incidentally , this means that without a more detailed description of how the dual @xmath2 is represented at higher orders it can not be used to constrain the form of @xmath32 .
still , we believe that it is possible to deform the representation of the dual @xmath2 in such a way to make it an all order symmetry . at the algebraic level , the dual @xmath2 would then be a perfectly well defined symmetry of the bfkl equation . also , as discussed at the end of the next section , at least the lowest order correction will not change the commutation of the original and dual @xmath2 , keeping open the possibility that the algebra obtained at leading order describes an exact symmetry of bfkl .
in @xmath0 sym , when taken together , the ordinary and dual superconformal symmetries generate an infinite dimensional yangian algebra @xcite .
the same yangian appears in the study of the spectrum of the dilatation operator @xcite , which is believed to define the hamiltonian of an integrable system @xcite , and could therefore explain the great amount of structure observed also for scattering amplitudes . indeed , it has recently been shown @xcite that the grassmannian integral @xcite conjectured to give the amplitudes leading singularities to all orders is the most general invariant of the yangian algebra of this kind . continuing this line of thought
, perhaps the original and dual @xmath2 algebras of bfkl may explain its integrability when taken together .
in this section we will check what form the closure of the two @xmath2 algebras takes .
in the case of @xmath0 ( where a supersymmetric version of the serre relations apply ) , the ordinary @xmath3 can be taken to make up the level 0 generators , while the level 1 generators can be extracted from the dual superconformal algebra .
in @xcite it was shown that in the representation acting on the scattering amplitudes , where the level 0 generators are written as a sum over external particles as j_a^(0 ) = _ i^n j_ia^(0 ) , the level 1 generators could be written in the bi - local form j_a^(1)=f_a^bc_1j
< i nj_jb^(0)j_ic^(0 ) , [ eq : ja1 ] where the indices of the structure constants have been raised by the metric of the algebra
. this bi - local form guarantees the commutation relation , and together with some additional requirements satisfied by the representation in question the serre relations are also implied @xcite .
recently , it has also been shown that the role of the ordinary and dual algebras can be interchanged , taking the dual superconformal algebra to make up the level 0 generators @xcite .
let us first review the action of the original @xmath2 symmetry of the bfkl hamiltonian , discovered by lipatov in @xcite , which we will take to play the role of the level 0 generators @xmath64 .
it is uncovered by performing a fourier transform of @xmath32 into impact parameter space : f ( ) = d^2 _ ad^2_bd^2e^i ( _ 1 _ a + _ 2 ( -_a)-_3 ( -_b ) -_4 _ b ) , [ eq : fourierf ] the additional factors @xmath70 and @xmath71 , appearing in correspond to propagators removed from @xmath32 in the normalization of the bfkl equation we use .
this can be rewritten in the form of an ordinary fourier transform in terms of incoming momenta @xmath5 as f ( ) = _ i d^2 _ i e^i _ i _ i f ( ) , where f()= ^(2 ) ( _ i ) .
[ eq : f ] in terms of the complex coordinate @xmath72 one finds an invariance of @xmath73 when , [ eq : sl2cl ] where @xmath74 and @xmath75 are complex parameters satisfying @xmath76 .
the transformations thus represent the group @xmath2 .
this group is generated by the transformations a + , corresponding to translations , b , corresponding to dilatations ( when @xmath77 is real ) and rotations ( when @xmath77 is a phase ) , , which is a complex inversion . all of these transformations are standard two - dimensional conformal transformations . with
the exception of the complex inversion they all have analogous transformations in the four - dimensional conformal group , acting on a four - vector @xmath78 .
the reason that the inversion is different is that in four - dimensions it appears as x^ which would correspond in the two - dimensional case to , which is , written in complex notation . it is
this last transformation that we will refer to as the two - dimensional inversion .
it is also a symmetry of bfkl since the kernel is real , implying invariance under @xmath79 . in order to better see the relation with the four - dimensional conformal group , where we can view the @xmath2 symmetry as the subgroup comprised of dilatations , and rotations , translations and special conformal transformations with indices taking values in the transverse plane
, we will now drop the convenient complex notation and instead use two - dimensional vector notation .
when we write @xmath80 , the superscript is a two - dimensional index taking the values @xmath7 and @xmath81 , while the subscript labels the particle , taking values @xmath82 . in the case of bfkl , @xmath48 is 4 , while it may be higher when we study its extensions .
we will now construct the infinitesimal generators corresponding to the translations , rotations , dilatations and special conformal transformations , following the convention that an infinitesimal generator @xmath83 produces a finite transformation through @xmath84 . in impact parameter space
the generators are given by the standard expressions .
infinitesimal two - dimensional translations , result in the usual expression p _
= -i _ i [ eq : plipatov ] for the momentum operator , while an infinitesimal dilatation induced the change gives the generator for dilatations as d = -i _ i _
an infinitesimal ( counterclockwise ) rotation gives the generator of rotations as r = -i _ _ i _
i^ , with @xmath85 .
the special conformal transformations are defined by the usual iti ( inversion - translation - inversion ) transformation giving k_= -i _ i ( _ i^2 - 2_^ ) .
[ eq : klipatov ] before ending this section we will fourier transform the generators .
the reason is that the dual @xmath2 is more naturally expressed in the momentum representation , and in order to be able to combine the two symmetries they must be expressed in a common language .
the way we perform the transformation is simply to act with the generators on f ( ) = _ ie^i_i _ if ( ) [ eq : fourier ] and rewrite , by partial integration , this as an action on @xmath86 . for simplicity we will denote the
fourier transformed generators by the same symbol as before .
we obtain d = i _ i ( 2 + p_i ) , [ eq : dlipatovp ] were the constant 2 stems from being in two dimensions , r = -i _ _ i p_i^ , [ eq : rlipatovp ] and k_= _ i ( 4 + 2 p_i^ - p_i ) . [
eq : klipatovp ] here the coefficient in front of @xmath87 is twice the number of dimensions , and is therefore 4 in our case . and finally , the momentum will of course be p _ = _
[ eq : plipatovp ] we will now show how a piece can be extracted from the dual @xmath2 symmetry that is ( almost ) of the form . as explained above , starting from the set of incoming momenta @xmath88 , @xmath89 , the the dual @xmath2 symmetry is uncovered by performing the change of variables _ i -_i+1=_i .
[ eq : xp ] momentum conservation is automatically satisfied by identifying @xmath90 with @xmath91 , but since we are interested in the algebraic structure that results from commuting the ordinary and dual symmetries , we will let @xmath90 and @xmath91 be independent variables , and include a factor @xmath92 in the green s function to impose momentum conservation . in terms of the @xmath25-variables , the generators of the dual @xmath2
take the same form as the generators - do in terms of the @xmath93-variables .
it should be noted , though , that just as for the @xmath0 scattering amplitudes , not all the dual generators are invariances .
the translations and the rotations are so , while the dilatations and the special conformal transformations are covariances .
one can easily make them them into invariances , however , by shifting them by a constant term .
this must be done before they can be combined with the generators - .
first , let us rewrite the generators of the dual algebra in terms of momenta @xmath94 .
the inverse of the change of variables is _
i = _ 1 - _ j=1^i-1_j , where we have kept @xmath90 as an independent variable , together with the momenta . requiring that the @xmath95 all be independent , in the sense that = ^i_j ^ _ then implies that the derivatives with respect to the @xmath7 should be replaced
, when going to the @xmath94 variables , as @xmath96 performing these substitutions , the dilatation operator becomes @xmath97 where we have introduced the superscript @xmath98 to distinguish this set of generators from that of the ordinary conformal symmetry .
we see that when acting on an object independent of @xmath90 ( or , equivalently , which only depends on the momenta ) , this dilatation operator is the same , up to a change of sign and a shift by a constant , as the original dilatation operator .
we do not obtain anything new from translations and rotations , either . up to terms containing derivatives with respect to @xmath99 the former
are represented by the identity , while the latter give the same generator as the original symmetry .
the only new symmetry comes from the special conformal generators , mimicking the structure appearing in the scattering amplitudes of @xmath0 sym , in the form of the original and dual conformal generators .
the special conformal generators take the form @xmath100 again , when acting on a physical object , which only depends on the momenta , all of the derivatives with respect to @xmath99 can be dropped .
furthermore , the third term can also be dropped due to the rotational invariance .
the remaining @xmath99-dependence is given by - 2x_1_i _ i .
[ eq : partd ] as mentioned above , @xmath101 is not an invariance of the gluon green s function , and must be corrected by a constant piece . in order to do
so we will hereafter restrict our attention to the case @xmath102 of bfkl for which we know the transformation properties . under dual inversions @xmath103 , and from
we get f ( ) = ^(2)(_15 ) _ 1 ^ 6_2 ^ 4_3
^ 2_4 ^ 4 f ( ) .
[ eq : finv ] this implies that acting on @xmath86 with @xmath104 produces the factor 6 x_1 + 4x_2 + 2x_3 + 4x_4= 16 x_1 -10p_1 - 6 p_2 - 4p_3 , and we see that the gluon green s function will be invariant under the combination
ik^(d)_+ 10p_1 + 6 p_2 + 4p_3 - 16x_1 . here
, we see that the @xmath105 term is precisely what is needed in order to eliminate the remaining @xmath90 dependence completely . when added to ,
one obtains a term which is proportional to and can be dropped . in the end
, we extract the new generator @xmath106 from the dual @xmath2 .
we will now go on to remove a part of , which itself annihilates the green s function , so that what is left is almost of the bi - local form .
the relevant bi - local operators turn out to be those where the index @xmath107 corresponds to the two - dimensional momentum @xmath108 .
using the metric @xmath109 for the algebra to raise indices , where the constant @xmath110 is chosen for convenience , and inserting the level 0 generators - into we have the operator @xmath111 = \nonumber \\ & = \sum_{1\leq j < i \leq n}\left [ p_{j\mu}\left ( 2 + p_i^\rho { \frac{\partial}{\partial p_i^\rho } } \right ) + \epsilon _ { \mu \nu } p_{j}^\nu\epsilon_{\rho \lambda}p_i^\rho { \frac{\partial}{\partial p_{i\lambda } } } - ( i \leftrightarrow j ) \right ] \label{eq : j13 } \ , \end{aligned}\ ] ] where a tilde is added to indicate that we do not know yet if these are symmetries of the green s function . in order to arrive at let us start by splitting the last sum in into two equal pieces .
one of the pieces we leave in it s current form , while using @xmath112 we rewrite the second piece as @xmath113 the nested sums run from index @xmath114 to @xmath48 , and we can cancel the terms corresponding to index @xmath114 with terms from the first line of . the only part that then remains from the sum in the first line is the term corresponding to @xmath115 , @xmath116 these terms complete the sum in so that it starts from @xmath115 . altogether
, we find @xmath117 the first sum will itself annihilate the green s function .
in fact , it will annihilate any function of the form @xmath118 , which includes a momentum conserving delta function .
the factors of @xmath119 make all terms in which the derivatives act on @xmath32 vanish so only the terms in which the derivatives act on the delta function remain .
for the same reason , when the derivatives are moved off the delta function by partial integration they must act on the factors of @xmath119 in order to get a non - vanishing contribution .
we are then left with -_i=1^n(p_i2 + p_i - p_i ) ^(2 ) ( ) f(\ { _ i } ) , which once again vanishes due to the delta function .
extracting the first term from , what remains is _
1j < i n ( p_i_j + p_j_i - _ i _ j - ( i j ) ) + 10p_1 + 6 p_2 + 4p_3 , which can be rewritten ( in two dimensions ) as _
1j < i n ( p_jp^_i + _ p_j^_p_i^ - ( i j ) ) + 10p_1 + 6 p_2 + 4p_3 , [ eq : khat4 ] reproducing a large portion of the terms in .
the only one that remains unaccounted for is _
1j < i n 2 ( p_j - p_i)= 4 ( ( n-1)p_1 + ( n-2)p_2 + + p_n-1 - ( n-1)p _ ) . for the case that interests us , which is @xmath102 , this becomes , after discarding the term proportional to @xmath108 , 12 p_1 + 8 p_2 + 4p_3 , which is almost , but not exactly , equal to the term @xmath120 , present in .
the difference is @xmath121 .
we must thus conclude that , in contrast to the case of @xmath0 sym , the bi - local operators @xmath122 are in general not symmetries of the green s function .
we instead have the symmetries j^(1)__^(1 ) - 2 q _ .
[ eq : p1 ] commuting these generators with the level 0 algebra gives the rest of the `` level 1 '' generators , all of which can be written as as j^(1)_a ^(1)_a - 2j^(0)_(12)a , [ eq : ja1sym ] where j^(0)_(12)a_i=1 ^ 2j^(0)_ia are the level 0 generators restricted to the first two ( upper ) momenta , and where the @xmath123 are defined by the bi - local formula .
it should be noted , that the reason that we are forced to deform the @xmath124 generators is the non - symmetrical action of the dual conformal inversions .
the @xmath125 would have been symmetries if @xmath126 had produced a factor @xmath127 under inversions .
for which the inversion produces the factor @xmath128 , consistent with the bi - local formula because the relevant constants appearing in the generators ( such as the constant in the dilatation operator ) are half of what they are in the present case . ]
notable is that in the forward limit @xmath57 , and then the bi - local operators will indeed be symmetries .
this case is particularly important , since it is related , via the optical theorem , to the total cross section . in @xcite
it was discussed how ir safe quantities , which includes the total cross section , retain the superconformal symmetry of @xmath0 .
by contrast , in order to promote the superconformal symmetry of tree amplitudes beyond tree level , an analysis carried out in @xcite , the generators must be non - trivially deformed .
this suggests that a more direct link to the integrable structure of @xmath0 might be found by looking at the forward limit .
so , does the original @xmath2 together with generate a yangian ?
it follows from the form of the bi - local formula that the yangian commutation relations are satisfied with the @xmath129 , implying that they are also obeyed by the full @xmath130 , since the additional @xmath131 piece simply gives the level 0 algebra restricted to particles 1 and 2 .
the key issue is therefore whether the serre relations - are satisfied . for many algebras , such as @xmath3
the first serre relation implies the second one @xcite .
and in our case the first serre relation is indeed satisfied .
however , for @xmath2 it is trivially satisfied ( its structure constants imply it take the form @xmath132 ) , meaning that it is the second serre relation that must be checked .
unfortunately , it turns out to not be satisfied . from the form of and it is clear that if the right hand side of the first relation is zero , the right hand side of the second relation must also be so .
but , a quick check reveals that the left hand side becomes non - zero when using the generators .
so it does not seem that the closure of the original and dual @xmath2-algebras give a yangian algebra .
this does not mean that the algebra can not be infinite dimensional , though .
the generators @xmath133 do not close under commutation , and from their form it seems unlikely that their closure would be finite - dimensional .
an important issue is also whether the algebra itself is affected when the representation of the dual @xmath2 is deformed to @xmath134 in order to take into consideration the anomaly .
the full form of the deformed generator becomes j^(1)_a ( ) = ^(1)_a - 2j^(0)_(12)a - 2 j^(0)_(13)a + ( ^2 ) , [ eq : ja1symdef ] where @xmath135 is defined analogously to @xmath136 .
interestingly , in the same way that the shift in by @xmath137 does not alter the commutation relation , neither does the change in . also when considering the rest of the algebra , generated by @xmath64 and @xmath134 it does not seem that the commutation relations are altered by the introduction of the coupling dependence , meaning that the algebra obtained at lowest order may very well be exact to all orders of bfkl .
we will , however , leave the exact determination of this algebraic structure to future studies .
in this article we have continued the study of how the symmetries of @xmath0 sym may explain the integrability found in the regge limit of qcd , examining the bfkl equation .
the conclusion is that even though much of the structure present in the four - dimensional gauge theory remains ( original and dual @xmath2 as remnants of ordinary and dual conformal symmetry , generators in the bi - local form , etc ) , it seems that , algebraically , things get considerably more complicated in this limit . in particular , the yangian symmetry is broken , and the dual @xmath2 does not satisfy a simple all - loop anomalous ward identity , although the possibility remains that the algebra itself be coupling independent to all orders
. these difficulties probably result from breaking the symmetries corresponding to the longitudinal and fermionic generators upon taking the regge limit , since the structure of the yangian of @xmath0 seems intimately related to supersymmetry ( the bi - local form of the level 1 generators is , for example , only compatible with the cyclicity of the amplitudes for a set of superalgebras @xcite . ) .
there are several possible directions for future research .
it would , of course , be nice to find an all - order representation of the dual @xmath2 .
but perhaps a more immediate issue is to extend this analysis to the full generalized lla spin chain of @xcite , and see if the structure of conserved charges can be related to the closure of the two @xmath2 algebras .
another possibility is to study the bk equation @xcite , giving unitarity corrections to bfkl by taking into account transitions from 2 to 4 reggeized gluons . since it was shown in @xcite that the @xmath138 to @xmath139 reggeized gluon vertex has the same dual @xmath2 covariance as the @xmath138 to @xmath138 vertex present in bfkl , and in @xcite that it has the original @xmath2-invariance
, it seems likely that the entire equation has this symmetry .
a more fundamental step that remains in order to understand the symmetries of the regge limit is precisely to elucidate exactly how , not only the dual , but also the original @xmath2 symmetry of lipatov is implied by the symmetries of the four dimensional gauge theory .
the solution to this problem is bound to be interesting , especially at nlla , since the conformal symmetry is broken at loop level in @xmath0 .
one source of the breaking is the introduction of an infrared regulator .
since bfkl is ir finite one could think that this is not a problem , but we have seen in the form of the breaking of the dual @xmath2 that ir effects can still cause problems . a second
, trickier problem is that the generators of the superconformal symmetry must be deformed in a non - trivial way at one - loop and beyond @xcite .
this deformation must somehow disappear before arriving at the @xmath2 symmetry of the regge limit , since the latter is exact to all orders , also at nlla , with no deformation present .
* acknowledgments * we are grateful to j. drummond , c. gmez and a. sabio vera for comments and discussions . the work of j. g. is supported by a spanish fpu grant .
in this appendix we will show how the form of the integral i_1 [ eq : nfintegral ] can be restricted by using dual @xmath2 symmetry , thus exemplifying how this symmetry can play a role in simplifying calculations .
first , we note that the integrand of @xmath140 has the same behavior close to the singularities at @xmath141 , @xmath142 and @xmath143 , and the same transformation properties under dual @xmath2 transformations as the integrand of i_2 d^2_i , which can be evaluated directly in dimensional regularization using , where u = v = are the two independent dual @xmath2 invariants that can be formed from @xmath144 .
this implies that the integral @xmath145 will be finite and that we can apply the dual conformal symmetry to conclude that i_1-i_2 = i(u , v ) , [ eq : i1mi2 ] for some function @xmath146 of the conformal invariants .
next , we observe that both @xmath140 and @xmath147 are symmetric under the interchanges @xmath148 , producing u , v v , [ eq:23swap ] as well as @xmath149 which gives uv .
[ eq:34swap ] when applied to these symmetries give i(u , v)=v i ( , ) i(u , v)=i(v , u ) , [ eq : symi ] which implies that i(u , v ) = ( 1+u+v)g(_1 , _ 2 ) , for some function @xmath150 , where _ 1 = _ 2 = are two independent invariants of and .
the difference between the two integrals @xmath140 and @xmath147 is further restricted by noting that for some values of @xmath144 , their integrands coincide for all values of the integration variable .
this occurs , for example , if @xmath90 equals one of the other @xmath95 , translating to @xmath151 .
in fact , it can be shown that the two integrands are equal for all @xmath28 precisely when = 1/4 , [ eq : integrandequal ] implying that @xmath152 , for arbitrary @xmath153 . as a consequence we can write i(u , v ) = ( 1+u+v)(4 -1)h(_1 , _ 2 ) = h(_1 , _ 2 ) , [ eq : ifinal ] where @xmath154 must be finite , for all @xmath153 , when @xmath155 .
it does not seem that we can obtain anything more from symmetry considerations alone .
still , from equations and we see that the form of @xmath140 is greatly restricted .
it should then not come as a surprise that in fact @xmath156 , as can easily be checked for any arbitrary choice of the @xmath95 .
this analysis can also be applied to restrict the form of more complicated integrals .
for example , at the third iteration of the bfkl equation , integrals such as i_3 = ( ) appear .
if we define @xmath157 by exchanging @xmath158 for the integrand of @xmath147 , then @xmath159 will once again be restricted by dual @xmath2 symmetry , symmetry under the exchange of @xmath148 , and the condition that the difference vanishes when is satisfied .
j. bartels , _ high - energy behavior in a nonabelian gauge theory .
first corrections to t(n@xmath160 m ) beyond the leading lns approximation _ , nucl . phys .
b * 175 * , 365 ( 1980 ) .
@xmath161 j. kwiecinski and m. praszalowicz , _ three gluon integral equation and odd c singlet regge singularities in qcd _ , phys .
b * 94 * , 413 ( 1980 ) .
@xmath161 l. n. lipatov , _ high - energy asymptotics of multicolor qcd and exactly solvable lattice models _ , arxiv : hep - th/9311037 . @xmath161
l. n. lipatov , _ asymptotic behavior of multicolor qcd at high energies in connection with exactly solvable spin models _ , jetp lett .
* 59 * ( 1994 ) 596 , [ pisma zh .
. fiz .
* 59 * ( 1994 ) 571 ] .
l. d. faddeev and g. p. korchemsky , _ high - energy qcd as a completely integrable model _ , phys .
b * 342 * ( 1995 ) 311 , arxiv : hep - th/9404173 .
l. n. lipatov , _ reggeization of the vector meson and the vacuum singularity in nonabelian gauge theories _ , sov .
j. nucl .
* 23 * ( 1976 ) 338 , [ yad .
* 23 * ( 1976 ) 642 ] .
@xmath161 v. s. fadin , e. a. kuraev and l. n. lipatov , _ on the pomeranchuk singularity in asymptotically free theories _ , phys .
b * 60 * ( 1975 ) 50 .
@xmath161 e. a. kuraev , l. n. lipatov and v. s. fadin , _ multi - reggeon processes in the yang - mills theory _ , sov .
jetp * 44 * ( 1976 ) 443 , [ zh . eksp
. teor .
* 71 * ( 1976 ) 840 ] .
@xmath161 e. a. kuraev , l. n. lipatov and v. s. fadin , _ the pomeranchuk singularity in nonabelian gauge theories _ ,
jetp * 45 * ( 1977 ) 199 , [ zh . eksp .
teor . fiz .
* 72 * ( 1977 ) 377 ] . @xmath161 i. i. balitsky and l. n. lipatov , _ the pomeranchuk singularity in quantum chromodynamics _
j. nucl .
* 28 * ( 1978 ) 822 , [ yad .
* 28 * ( 1978 ) 1597 ] .
j. a. minahan and k. zarembo , _ the bethe - ansatz for n = 4 super yang - mills _
, jhep * 0303 * ( 2003 ) 013 arxiv : hep - th/0212208 .
@xmath161 n. beisert , c. kristjansen and m. staudacher , _ the dilatation operator of n = 4 super yang - mills theory _ , nucl .
b * 664 * , 131 ( 2003 ) arxiv : hep - th/0303060 .
@xmath161 n. beisert and m. staudacher , _ the n=4 sym integrable super spin chain _
b * 670 * , 439 ( 2003 ) arxiv : hep - th/0307042 .
@xmath161 a. v. belitsky , s. e. derkachov , g. p. korchemsky and a. n. manashov , _ quantum integrability in ( super ) yang - mills theory on the light - cone _ , phys .
b * 594 * , 385 ( 2004 ) arxiv : hep - th/0403085.@xmath161 n. beisert , v. dippel and m. staudacher , _ a novel long range spin chain and planar @xmath162 super yang - mills _
, jhep * 0407 * ( 2004 ) 075 , arxiv : hep - th/0405001 .
@xmath161 m. staudacher , _ the factorized s - matrix of cft / ads _ , jhep * 0505 * ( 2005 ) 054 , arxiv : hep - th/0412188 .
@xmath161 n. beisert , v. a. kazakov , k. sakai and k. zarembo , _ complete spectrum of long operators in n = 4 sym at one loop _ , jhep * 0507 * , 030 ( 2005 ) arxiv : hep - th/0503200 .
@xmath161 n. beisert and m. staudacher , _ long - range @xmath3 bethe ansaetze for gauge theory and strings _ , nucl .
b * 727 * ( 2005 ) 1 , arxiv : hep - th/0504190 . @xmath161
n. beisert , _ the @xmath163 dynamic s - matrix _ , adv .
* 12 * , 945 ( 2008 ) , arxiv : hep - th/0511082 .
r. a. janik , _ the @xmath164 superstring worldsheet s - matrix and crossing symmetry _ , phys .
d * 73 * ( 2006 ) 086006 , arxiv : hep - th/0603038 .
@xmath161 n. beisert , r. hernndez and e. lpez , _ a crossing - symmetric phase for @xmath165 strings _ , jhep * 0611 * ( 2006 ) 070 , arxiv : hep - th/0609044 .
@xmath161 n. beisert , b. eden and m. staudacher , _ transcendentality and crossing _ , j. stat .
mech . * 0701 * ( 2007 ) p021 arxiv : hep - th/0610251 .
@xmath161 a. v. kotikov , l. n. lipatov , a. rej , m. staudacher and v. n. velizhanin , _ dressing and wrapping _ , j. stat .
* 0710 * ( 2007 ) p10003 , arxiv:0704.3586 .
@xmath161 n. beisert , t. mcloughlin and r. roiban , _ the four - loop dressing phase of n=4 sym _ ,
d * 76 * , 046002 ( 2007 ) , arxiv:0705.0321 .
@xmath161 f. fiamberti , a. santambrogio , c. sieg and d. zanon , _ wrapping at four loops in @xmath166 sym _ , phys .
b * 666 * ( 2008 ) 100 , arxiv:0712.3522 .
@xmath161 f. fiamberti , a. santambrogio , c. sieg and d. zanon , _ anomalous dimension with wrapping at four loops in @xmath166 sym _ , nucl .
b * 805 * ( 2008 ) 231 , arxiv:0806.2095 .
@xmath161 z. bajnok and r. a. janik , _ four - loop perturbative konishi from strings and finite size effects for multiparticle states _ ,
b * 807 * ( 2009 ) 625 arxiv:0807.0399 .
@xmath161 z. bajnok , r. a. janik and t. lukowski , _ four loop twist two , bfkl , wrapping and strings _ , nucl
. phys .
b * 816 * , 376 ( 2009 ) , arxiv:0811.4448 .
l. n. lipatov , _ high - energy scattering in qcd and in quantum gravity and two - dimensional field theories _ , nucl .
b * 365 * ( 1991 ) 614 .
@xmath161 r. kirschner , l. n. lipatov and l. szymanowski , _ effective action for multi - regge processes in qcd , _ nucl .
b * 425 * ( 1994 ) 579 arxiv : hep - th/9402010 l. lipatov , _ the bare pomeron in quantum chromodynamics _ , sov.phys.jetp * 63 * ( 1986 ) , 904 - 912 , zh.eksp.teor.fiz .
* 90 * ( 1986 ) 1536 - 1552 .
i. balitsky and g. a. chirilli , _ nlo evolution of color dipoles in n=4 sym _ , nucl .
b * 822 * ( 2009 ) 45 arxiv:0903.5326 .
@xmath161 v. s. fadin , r. fiore and a. v. grabovsky , _ matching of the low - x evolution kernels _ , nucl .
b * 831 * ( 2010 ) 248 arxiv:0911.5617 .
v. g. drinfeld , _ hopf algebras and the quantum yang - baxter equation , _ sov . math .
* 32 * ( 1985 )
254 [ dokl . akad .
nauk ser .
* 283 * ( 1985 ) 1060 ] .
@xmath167 v. g. drinfeld , _ a new realization of yangians and quantized affine algebras , _ sov . math . dokl .
* 36 * ( 1988 ) 212 .
j. m. drummond , j. henn , v. a. smirnov and e. sokatchev , _ magic identities for conformal four - point integrals _ , jhep * 0701 * , 064 ( 2007 ) arxiv : hep - th/0607160 @xmath161 l. f. alday and j. m. maldacena , _ gluon scattering amplitudes at strong coupling _ , jhep * 0706 * ( 2007 ) 064 arxiv:0705.0303 .
@xmath161 j. m. drummond , g. p. korchemsky and e. sokatchev , _ conformal properties of four - gluon planar amplitudes and wilson loops _ ,
b * 795 * ( 2008 ) 385 arxiv:0707.0243 .
@xmath161 a. brandhuber , p. heslop and g. travaglini , _ mhv amplitudes in n=4 super yang - mills and wilson loops _ ,
b * 794 * , 231 ( 2008 ) arxiv:0707.1153 . @xmath161 j. m. drummond , j. henn , g. p. korchemsky and e. sokatchev , _ on planar gluon amplitudes / wilson loops duality _ , nucl .
b * 795 * , 52 ( 2008 ) arxiv:0709.2368 .
@xmath161 j. m. drummond , j. henn , g. p. korchemsky and e. sokatchev , _ dual superconformal symmetry of scattering amplitudes in n=4 super - yang - mills theory _ , nucl .
b * 828 * , 317 ( 2010 ) , arxiv:0807.1095 j. m. drummond and l. ferro , the yangian origin of the grassmannian integral , arxiv:1002.4622 .
@xmath161 g. p. korchemsky and e. sokatchev , _ superconformal invariants for scattering amplitudes in n=4 sym theory _
, arxiv:1002.4625 .
i. balitsky , _ operator expansion for high - energy scattering _ , nucl .
b * 463 * ( 1996 ) 99 arxiv : hep - ph/9509348 .
@xmath161 y. v. kovchegov , _ small - x f2 structure function of a nucleus including multiple pomeron exchanges _ , phys .
d * 60 * ( 1999 ) 034008 arxiv : hep - ph/9901281 | in this paper we continue our study of the dual sl(2,c ) symmetry of the bfkl equation , analogous to the dual conformal symmetry of n=4 super yang mills .
we find that the ordinary and dual sl(2,c ) symmetries do not generate a yangian , in contrast to the ordinary and dual conformal symmetries in the four - dimensional gauge theory .
the algebraic structure is still reminiscent of that of n=4 sym , however , and one can extract a generator from the dual sl(2,c ) close to the bi - local form associated with yangian algebras .
we also discuss the issue of whether the dual sl(2,c ) symmetry , which in its original form is broken by ir effects , is broken in a controlled way , similar to the way the dual conformal symmetry of n=4 satisfies an anomalous ward identity .
at least for the lowest orders it seems possible to recover the dual sl(2,c ) by deforming its representation , keeping open the possibility that it is an exact symmetry of bfkl .
ift uam / csic10 - 12 |
since most stars are formed in clusters , it would be useful to have quantitative and objective statistical measures of their structure , with a view to comparing clusters formed in different environments , and tracking changes in structure as clusters evolve .
this is particularly important for young , embedded clusters , where the structure may yield important clues to the formation process but is changing rapidly .
it is also important for comparing observed clusters with numerical simulations . at present
, we do not have sufficiently robust statistical measures for this purpose .
features which are easily identified by the human eye , such as sub - clusters , or linear features , can be strangely elusive to objective statistical analysis .
for example , it is difficult to distinguish , statistically , between a degree of fractal or random sub - clustering , and the existence of a density gradient ( bate , clarke & mccaughrean 1997 ) .
this paper explores some possible measures , and evaluates their usefulness .
in particular , we find a robust objective measure which both quantifies , and distinguishes between , a smooth large - scale radial density gradient and multi - scale ( fractal ) sub - clustering . in section 2
we describe our methodology . in section 3
we look again at the mean surface density of companions ( msdc ) , a tool pioneered by larson ( 1995 ) and subsequently used by several others ( e.g. simon 1995 ; bate , clarke and mccaughrean 1997 ; nakajima et al .
1998 ; brandner & khler 1998 ; gladwin et al . 1999 ; klessen & kroupa 2001 ) .
we focus on measures which reflect the clustering regime ( wide separations ) rather than the binary regime ( close separations ) . in section 4 we explore the use of the minimal spanning tree ( barrow , bhavasar & sonoda 1985 ) and its derivatives . in section 5
we combine the msdc and the mst to derive a single measure @xmath0 which is able both to quantify , and to distinguish between , a smooth radial density gradient and multi - scale ( fractal ) sub - clustering .
all the measures are tested and calibrated on multiple realizations of artificial star clusters , and applied to @xmath14 ophiuchus , chamaeleon , taurus , ic348 and ic2391 .
our results are discussed in section 6 , and the main conclusions are summarized in section 7 .
three different types of artificial star cluster have been created , using random numbers @xmath27 to generate the individual star positions . the first type ( 2d@xmath4 )
are circular clusters ( i.e. two - dimensional disc - like clusters ) with surface density @xmath28 and @xmath29 .
the second type ( 3d@xmath4 ) are spherical clusters ( i.e. three - dimensional clusters ) having volume density @xmath3 with @xmath30 .
the third type ( f@xmath7 ) are fractal star clusters ( again three dimensional ) with fractal dimension @xmath31 .
the different types are listed in column 1 of table [ measures ] .
all of the artificial clusters are created with 100 to 300 stars , as the numbers of stars within the five real clusters lie within that range .
the data for the five real clusters used are illustrated in appendix [ rawdata ] , and the sources listed in table 2 . [
tbp]lcccc cluster type & & & + or name & @xmath2 & @xmath8 & @xmath32 & @xmath0 + & & & & + 2d0(@xmath33 ) & @xmath34 & @xmath35 & @xmath36&@xmath37 + 2d1(@xmath38 ) & @xmath39 & @xmath40 & @xmath41&@xmath42 + & & & & + 3d2.9(@xmath43 ) & @xmath44&@xmath45 & @xmath46&@xmath47 + 3d2(@xmath48 ) & @xmath49 & @xmath50 & @xmath51&@xmath52 + 3d1(@xmath53 ) & @xmath54 & @xmath40 & @xmath55&@xmath56 + 3d0(@xmath57 ) & @xmath58 & @xmath59 & @xmath36&@xmath60 + & & & & + f3.0(@xmath61 ) & @xmath62 & @xmath63 & @xmath64 & @xmath58 + f2.5(@xmath65 ) & @xmath66 & @xmath67 & @xmath68 & @xmath69 + f2.0(@xmath70 ) & @xmath71 & @xmath72 & @xmath73 & @xmath74 + f1.5(@xmath75 ) & @xmath76 & @xmath77 & @xmath78 & @xmath79 + & & & & + ic2391 & 0.74 & .49 & .30 & .66 + chamaeleon & 0.63 & .42 & .45 & .67 + taurus & 0.55 & .26 & .56 & .47 + @xmath14 ophiuchus & 0.53 & .45 & .39 & .85 + ic348 & 0.49 & .48 & .46 & .98 + [ measures ] a cluster of type 2d@xmath4 is created by positioning the stars according to @xmath80 where @xmath81 and @xmath82 are random numbers in the range 0 - 1 . a cluster of type 3d@xmath4 is created by positioning the stars according to @xmath83 where @xmath81 , @xmath84 and @xmath82 are random numbers in the range 0 - 1 .
clearly this method can not be used for @xmath4 = 3 , so to have a cluster type approximating to @xmath4 = 3 we use @xmath4 = 2.9 .
a cluster of type f@xmath7 is created by defining an ur - cube with side 2 , and placing an ur - parent at the centre of the ur - cube .
next , the ur - cube is divided into @xmath85 equal sub - cubes , and a child is placed at the centre of each sub - cube ( the first generation ) .
normally we use @xmath86 , in which case there are 8 sub - cubes and 8 first - generation children .
the probability that a child matures to become a parent in its own right is @xmath87 , where @xmath7 is the fractal dimension . for lower @xmath7 , the probability that a child matures to become
a parent is lower , and the cluster is more ` porous ' .
children who do not mature are deleted , along with the ur - parent .
a little noise is then added to the positions of the remaining children , to avoid an obviously regular structure , and they then become the parents of the next generation , each one spawning @xmath85 children ( the second generation ) at the centres of @xmath85 equal - volume sub - sub - cubes , and with each second - generation child having a probability @xmath87 of maturing to become a parent .
this process is repeated recursively until there is a sufficiently large generation that , even after pruning to impose a spherically symmetric envelope of radius 1 within the ur - cube , there are still more children than the required number of stars .
children are then culled randomly until the required number is left , and the surviving children are identified with the stars of the cluster . at each generation , the survival of a child is determined by generating a random number @xmath27 in @xmath88 ; survival then requires that @xmath89 .
clusters of type 2d@xmath4 are investigated for two purposes .
first , we wish to clarify the effect of a sharply defined circular edge on an otherwise statistically uniform , two - dimensional distribution of stars .
clusters of type 2d0 enable us to isolate this effect .
second , we wish to explore how readily two - dimensional and three - dimensional distributions can be distinguished
. this could be important if stars are being formed in layers , for example at a shock front . for each type of artificial cluster ,
100 realisations are analysed , so that means and standard deviations can be obtained for the parameters extracted .
three - dimensional clusters ( types 3d@xmath4 and f@xmath7 ) are projected onto an arbitrary plane prior to analysis .
two - dimensional clusters are viewed face - on .
a widely used tool for analysing the structure of star clusters is the log / log plot of the mean surface - density of companions , @xmath90 against separation , @xmath91 .
this tool has been pioneered by larson ( 1995 ) , building on earlier work by gomez et al .
( 1993 ) , who used the two point correlation function .
several papers have confirmed larson s finding that a plot of @xmath92 $ ] against @xmath93 $ ] hereafter a larson plot can be fitted with two power law sections , corresponding to two distinct regimes . at the smaller separations , @xmath94 ,
a star s companions are mainly in binary and higher multiple systems , and the slope of the larson plot is @xmath95/d\ell og[s ] \simeq -2 $ ] . at larger separations , @xmath96 , companions are simply other members of the overall cluster , and may only be close due to projection .
the slope here is generally larger ( i.e. still negative but smaller in magnitude ) , @xmath97/d\ell og[s ] \ga -1 $ ] .
larson has suggested that @xmath98 might be related to the fractal dimension of the sub - clustering , @xmath99 .
in addition , he has proposed that the break point between the two straight sections , at @xmath100 , might correspond to the jeans length .
however , recent analysis has cast some doubt on these interpretations .
first , the break point is strongly influenced by the overall surface - density of stars ( and hence by the depth of the cluster along the line of sight ) , as pointed out by simon ( 1997 ) and bate et al ( 1997 ) .
second , fitting @xmath101 objectively is difficult , because at the low-@xmath91 end it is distorted by the switch to the binary regime , and more importantly at the high-@xmath91 end it is distorted by edge effects .
consequently , one is left with at best a range of order @xmath102 to @xmath103 and the result is sensitive to how the range is actually chosen ; if the range is shortened or extended arbitrarily , the slope of the fitted line may change dramatically .
third , @xmath98 is not necessarily related to the fractal dimension of the clustering . as shown by bate et al .
( 1997 ) and klessen & kroupa ( 2001 ) , it may simply reflect a large - scale density gradient in the cluster .
an alternative way of evaluating the data from which larson plots are derived is to calculate the distribution function @xmath1 , where @xmath104 gives the probability that the projected separation between two cluster stars chosen at random is in the interval @xmath105 . to do this empirically ,
we define @xmath106 equal @xmath91-bins in the range @xmath107 , so that all the bins have width @xmath108 , and the @xmath109th bin corresponds to the interval @xmath110 , with mean value @xmath111 .
@xmath112 is the overall radius of the cluster , and is defined by finding the mean position of all the stars in the cluster and then setting @xmath112 equal to the distance to the furthest star .
then we count the number of separations @xmath113 falling in each bin , and put @xmath114 where @xmath11 is the total number of stars in the cluster , and hence @xmath115 is the total number of separations .
figure [ arti / p(s)](a ) presents the results obtained from 100 clusters of type 2d0 , i.e. a disc having statistically uniform surface - density .
the plotted points give the mean @xmath116 from the 100 realizations , and the error bars give the width of the bin and the @xmath117 standard deviation . if there were no edge effects ( i.e. if the uniform surface - density extended to infinity in two dimensions ) , we would have @xmath118 , and this is indeed a good fit to @xmath116 at small @xmath119 values , as indicated by the dashed line on fig . [ arti / p(s)](a ) .
_ departures from this straight line are entirely due to edge effects . _ in fact , @xmath120 can be calculated semi - analytically for a disc having uniform surface - density : @xmath121 where @xmath122\end{aligned}\ ] ] the solid line on fig .
[ arti / p(s)](a ) shows that this function fits the plotted points well , and it is included on all the other plots for reference , i.e. to emphasize the features which are not due to edge effects .
when derived in this way , the @xmath120 plot contains little information about the distribution of binary separations , since they are all in the first bin .
however , it seems to be well established that the distribution of binary separations is approximately scale free over a wide range of separations ( @xmath123 ) .
the more critical issue the one with which we are concerned here is the distribution of separations in the clustering regime and what it tells us about the overall structure of the cluster .
this information is well represented by @xmath120 , as can be seen from figs .
[ arti / p(s)](b ) through [ arti / p(s)](f ) , which show the results obtained for the other five types of non - fractal artificial star cluster .
figure [ arti / p(s)](b ) shows how @xmath120 is slewed towards smaller @xmath91 values for a disc with a centrally concentrated surface - density , @xmath124 .
figures [ arti / p(s)](c ) to [ arti / p(s)](f ) show spherical clusters having volume - density gradients @xmath3 with @xmath125 . again
the distribution slews to smaller @xmath91 values as the sphere becomes more centrally concentrated ( i.e. with increasing @xmath4 ) .
one feature which distinguishes the plots is the location of the maximum , i.e. the separation @xmath126 at which @xmath120 is largest . as
a cluster becomes more centrally condensed , @xmath126 moves to smaller values , and the amplitude of the maximum increases .
however , for an individual cluster @xmath126 will not be well defined , and so it is not a robust measure .
a better measure of this trend is the normalized correlation length for each cluster .
the correlation length is the mean separation @xmath2 between stars in the cluster , and it is normalized by dividing by @xmath112 .
the second column of table [ measures ] gives mean values of @xmath2 and their standard deviations , for the various artificial cluster types .
the @xmath2 values for the five real star clusters are also given .
the shapes of the @xmath1 plots , and hence also the @xmath2 values , are independent of the number of stars in the cluster . in trials with cluster sizes of 100 to 1000 stars ,
@xmath2 stays within one standard deviation of the mean value for 200 stars .
this is at first sight surprising .
a 1000-star cluster is so much more dense than a 100-star cluster , that one might expect the mean separation of stars to be smaller .
however , although each star has more close neighbours , it also has more distant neighbours , and the value of @xmath2 remains constant .
this is an attractive feature of the normalised correlation length as a statistical descriptor for clusters . from table 1 , we see that @xmath2 decreases monotonically with increasing @xmath4 , and can therefore be used to estimate @xmath4 for star clusters which are presumed _ a priori _ to have radial density gradients .
importantly , cluster types 2d1 and 3d2 are easily distinguished by their @xmath2 values and their @xmath120 plots , despite the widespread but fallacious assumption that a three dimensional cluster with volume - density @xmath127 is , when projected on the sky , similar to a two dimensional cluster with surface - density @xmath124 .
in fact it is clusters of types 2d1 and 3d1 ( i.e. with the same exponent , @xmath128/d \ell n[r ] \sim -1 $ ] , and @xmath129/d \ell n[r ] \sim -1 $ ] ) which are hard to distinguish .
@xmath1 plots for the real clusters are shown on fig .
[ real / p(s ) ] .
ic348 and @xmath14 ophiuchus resemble clusters of type 3d2 , both on the basis of their @xmath2 values ( table 1 ) , and the shapes of their @xmath1 plots ( figs .
[ real / p(s)](a ) and [ real / p(s)](c ) ) .
for ic2391 the @xmath2 value and the @xmath1 plot ( fig .
[ real / p(s)](b ) ) are most like those for clusters of type 3d1 .
chamaeleon and taurus have correlation lengths intermediate between types 3d1 and 3d2 , but their @xmath1 plots are clearly not generic .
this is because they contain sub - clusters , as illustrated in figs .
[ realfield](d ) and [ realfield](e ) .
consequently @xmath1 has multiple maxima . in some cases
these maxima can be identified with ( i ) separations between stars in the same sub - cluster ( the maximum at the smallest separations ) and ( ii ) separations between stars in two distinct sub - clusters ( maxima at larger separations , corresponding to the separation between the two sub - clusters ) .
if there are @xmath130 sub - clusters , there can be up to @xmath131 maxima , but fewer if there is degeneracy in the distances between sub - clusters . after smoothing , the @xmath1 plot for chamaeleon ( fig .
[ real / p(s)](e ) ) shows two distinct maxima , suggesting at least two sub - clusters , and fig .
[ realfield](e ) does indeed show two sub - clusters .
they are separated by @xmath132 , hence giving rise to the maximum in @xmath1 at @xmath133 . however , after smoothing , the @xmath1 plot for taurus ( fig .
[ real / p(s)](d ) ) shows only three well defined maxima , suggesting at most three sub - clusters , whereas fig . [ realfield](d ) shows at least eight well defined sub - clusters .
evidently the @xmath1 plot is not a robust diagnostic of sub - clustering . if we now consider artificial fractal star clusters with the same fractal dimension d , we find that there is so much variance in their individual @xmath120 plots that we can not sensibly define a mean @xmath120 plot
. however , we can still compute the mean normalised correlation length @xmath2 and its variance .
the results are given in table 1 .
we see that @xmath2 increases monotonically with increasing d and can therefore be used to estimate d for star clusters which are presumed _ a priori _ to be fractal .
moreover , the value of @xmath2 for star clusters of type f3.0 is essentially the same as for clusters of type 3d0 , as it should be .
the small difference is attributable to the fact that in constructing clusters of type f3.0 the positioning of the individual stars is not completely random , whereas for type 3d0 it is .
however , the range of @xmath2 for d in ( 1.5,3.0 ) is almost identical to that for @xmath4 in ( 0,2 ) .
therefore @xmath2 is degenerate and can not on its own be used to distinguish multi - scale ( fractal ) sub - clustering from a large - scale radial density gradient .
the minimal spanning tree ( mst ) is the unique network of straight lines joining a set of points , such that the total length of all the lines hereafter ` edges ' in the network
is minimised and there are no closed loops .
the construction of such a tree is described by gower and ross ( 1969 ) .
starting at any point , an edge is created joining that point to its nearest neighbour .
the tree is then extended by always constructing the shortest link between one of its nodes and an unconnected point , until all the points have been connected .
figure [ mstreal ] shows the msts for the real star clusters @xmath14 ophiuchus , taurus , chamaeleon , ic348 and ic2391 .
the use of minimal spanning trees ( msts ) as a probe of cosmological structure was explored by barrow , bhavsar and sonoda ( 1985 ) , and a further refinement , the self avoiding random walk , was described by baugh ( 1993 ) . although the approach seemed promising as a means of picking out clumps and filaments , the only statistical analysis of the mst for cosmological purposes of which we are aware is due to graham , clowes and campusano ( 1995 ) , who adopted methods developed by hoffman and jain ( 1983 ) and dussert et al(1987 ) and applied them to the distribution of quasars .
we describe their analysis in appendix [ graham ] , and show that it does not work well for star clusters .
once the mst of a cluster has been constructed it is straightforward to compute the mean edge length , @xmath8 . unlike the normalised correlation length @xmath2 , @xmath8 is not independent of the number of stars in the cluster , @xmath11 .
as @xmath11 increases , more short edges are created on the mst and @xmath8 decreases .
the expected total length of the mst of a random array of @xmath11 points , uniformly distributed over a two - dimensional area a , is asymptotically proportional to @xmath134 ( hammersley et al .
as there are @xmath135 edges , the mean edge length is asymptotically proportional to @xmath136 , and so this factor should be used to normalise the mean edge length of clusters having different areas @xmath10 and/or different numbers of stars @xmath11 .
the resulting normalised mean edge length @xmath8 has been evaluated for 100 realisations of each type of artificial star clusters , and for the real star clusters , and the results are recorded in table 1 ( column 3 ) .
also recorded in table 1 ( column 4 ) is the mean of the standard deviations of the mst edge lengths , @xmath32 , this quantity is used in appendix [ graham ] .
table [ measures ] shows that for artificial clusters of type 2d@xmath4 , 3d@xmath4 and @xmath137 , both @xmath8 and @xmath2 decrease monotonically as @xmath4 increases ( i.e. the degree of central concentration becomes more severe ) or as @xmath7 decreases ( i.e. the degree of sub - clustering become more severe ) .
however , @xmath2 decreases more quickly than @xmath8 as @xmath4 is increased , while @xmath8 decreases more quickly than @xmath2 as @xmath7 is decreased .
thus , the ratio @xmath138 yields a measure which not only quantifies , but also distinguishes between , a smooth overall radial density gradient and multi - scale fractal sub - clustering .
mean values of @xmath0 for the various types of artificial star cluster are recorded in table 1 ( column 5 ) . for artificial clusters with a smooth large - scale radial density gradient ( type 3d@xmath4 ) , @xmath139 increases from @xmath140 to @xmath141 as the degree of central concentration increases from @xmath142 ( statistically uniform number - density ) to @xmath143 ( @xmath144 ) . for artificial clusters with fractal sub - structure ( type f@xmath7 ) ,
@xmath139 decreases from @xmath140 to @xmath145 as the degree of sub - clustering increases from @xmath146 ( uniform number - density , no sub - clustering ) to @xmath147 ( strong sub - clustering ) .
we can therefore construct a plot ( figure [ qplot ] ) of @xmath7 against @xmath0 for @xmath148 , and @xmath4 against @xmath0 for @xmath149 . for any real cluster
we can compute its @xmath0 value , and then use figure [ qplot ] to read off its notional fractal dimension @xmath25 ( if @xmath150 , implying sub - clustering ) , or its radial density exponent @xmath4 ( if @xmath151 , implying a large - scale radial density gradient ) .
the small kink at @xmath152 is due to the fact in constructing a cluster of type f3.0 the stars are positioned regularly ( in the sense that at each generation , each subcube of space is occupied ) and therefore the number - density is artificially uniform ; in contrast , when we construct a cluster of type 3d0 the stars are positioned randomly , so that the density is only uniform in a statistical sense and there are poisson fluctuations in the local density .
fractal dimensions obtained from fig . [ qplot ] in this way are only notional , because @xmath0 ( or any other single measure ) can reflect sub - clustering , but can not capture whether the sub - clustering is hierarchically self - similar . using figure [ qplot ]
, we infer that taurus , ic2391 and chamaeleon have substructure with notional fractal dimensions @xmath153 .
in contrast , @xmath14 ophiuchus and ic348 appear to be centrally concentrated , with radial density exponents @xmath154 .
these inferences agree well with an intuitive reading of the raw data shown in fig .
[ realfield ] .
the mst will normally link a star to its binary companion , as this will usually be the shortest way of adding one or other of the stars to the tree .
binaries create very short edges and therefore a large population of binary stars will cause a noticeable reduction in the mean edge length , @xmath8 . of the five real clusters considered in this paper , taurus has been subjected to particularly close scrutiny and has a larger identified population of binaries than any of the others .
as the binaries are not part of the clustering regime , it is important to establish whether they are distorting the result . using the mst , all pairs of stars lying closer together than @xmath155 of the cluster radius
were pruned , leaving single stars .
for chamaeleon , @xmath14 ophiuchus , ic348 and ic2391 , only 3 , 0 , 1 and 2 such pairs were found ; taurus , however , was pruned from 215 down to only 137 primary stars .
for the pruned version of taurus , @xmath2 increased from 0.55 to 0.57 , while @xmath8 increased from 0.26 to 0.33 and @xmath0 increased from 0.47 to 0.58 .
removal of the binaries resulted in the notional fractal dimension for taurus being increased from 1.5 to 1.9 .
this demonstrates that in a cluster with a large binary population , it is important to prune the close companions before evaluating @xmath2 , @xmath8 and @xmath0 .
lccclname & members & age&@xmath156 & sources + & & myr & myr & + & & & & + ic2391 & 166 & 53&2.5 & barrado et al .
( 2001 ) + cham .
& 136 & 0.1 - 40 & 2.7&lawson et al .
( 1996 ) + & & & & ghez et al .
( 1997 ) + taurus & 215 & 1.0 & 10.0 & briceno et al .
( 1993 ) + & & & & ghez et al .
( 1993 ) + & & & & gomez et al .
( 1992 ) + & & & & hartmann et al .
( 1991 ) + & & & & herbig et al .
( 1988 ) + & & & & leinert et al .
( 1993 ) + & & & & simon et al .
( 1995 ) + & & & & waer et al .
( 1988 ) + & & & & luhman et al .
( 2003 ) + @xmath14 oph . & 199 & 0.3 - 2.0 & 1.35 & bontemps et al .
( 2001 ) + ic348 & 288 & 2.0 & 2.0 & luhman et al .
( 2003 ) + [ sources ]
the ratio of the normalized mean edge length to the normalized correlation length , @xmath0 , is effective in distinguishing between a smooth large - scale radial density gradient and multi - scale fractal sub - clustering , because it is sensitive not only to the frequency of small separations between stars , but also to their spatial distribution .
the mst edge length @xmath8 is a simple average of the distances of stars to their ( usually ) closest neighbours .
if one star is moved very close to another , the change in @xmath8 will be diluted by the total number of stars , @xmath11 .
however , when calculating the mean distance of companions for all stars , only one star out of @xmath11 in the cluster has had one of its @xmath157 companions moved very close .
the change in @xmath2 , is therefore diluted by @xmath158 .
thus , when small separations are scattered all over the cluster , increasing the number of small separations causes both @xmath8 and @xmath2 to decrease , but @xmath2 decreases more slowly than @xmath8 .
this is the case for decreasing fractal dimension
. for radially concentrated clusters , by contrast , increasing the clustering creates more small separations between stars , but these are all in the central region of the cluster . moving another star to this area affects @xmath8 in the normal way , the star having a newly short edge length between it and its nearest neighbour , and the change in the mean distance being diluted by @xmath11 .
however , the large number of other stars in the centre also gain another close neighbour .
the decrease in @xmath2 is therefore compounded and exceeds that in @xmath8 .
consequently , the quotient @xmath159 successfully distinguishes between clusters which have a smooth large - scale radial density gradient and clusters which have multi - scale fractal sub - clustering , in a way which agrees with an intuititive analysis but which can not be accomplished using existing methods such as larson plots or box dimension plots .
an additional advantage over these methods is that the calculation of @xmath0 is quantitative and objective , as no intervention is required in the normalisation process , in the construction of the mst , or in choosing a range over which to calculate a slope .
we should emphasize that classical methods for evaluating the density profile of a cluster , or its fractal dimension , are not viable for clusters with @xmath160 200 members , primarily because of low - number statistics .
for example , if one attempts to define the projected radial density profile for a real cluster of stars by counting stars in different annuli , the result is very noisy .
alternatively , if one attempts to determine the mean projected radial density profile for a 200-member artificial cluster having a given radial density profile in three dimensions , using many different realizations and with a view to comparing this with a real cluster , one finds that the standard deviation is very large , and so the diagnostic power of this profile is poor . in the same spirit
one might attempt to construct the box dimension plot ( bdp ) of a real cluster and compare it with the mean bdp of artificial star clusters having a given fractal dimension . to construct a bdp one
divides the projected image of the star cluster into a grid of square cells of side @xmath161 and counts the number of cells , @xmath162 which are occupied by stars .
then , by repeating this for different values of @xmath161 , one obtains a plot of @xmath163 against @xmath164 . for a true fractal
this plot is a straight line with slope equal to the fractal dimension .
however , for a star cluster with only @xmath165 members , the plot is not linear . by treating many realisations of artificial clusters all having the same fractal dimension _ and the same number of stars _ , one can define a mean bdp .
however , the mean bdp is not very strongly dependent on the fractal dimension and it has a large standard deviation
. therefore the box dimension plot of a real cluster does not give a useful constraint on its fractal dimension .
it is for this reason that we have sought integral measures of cluster structure .
the same philosophy informs the use of equivalent width when evaluating noisy spectral lines ( for example ) .
we also note that a cluster can not have a large - scale radial density gradient , and at the same time be fractally sub - clustered .
a cluster could have a large - scale radial density gradient and _ non - fractal _ sub - clustering
but then it would require more parameters to characterize the structure , and its diagnosis would become correspondingly more difficult ( if not impossible for clusters with @xmath160 200 stars ) . in table 2
we list estimates for the ages and the crossing times of the clusters we have analyzed . on the basis of simple arguments
, we might expect the @xmath0 value of a cluster to increase with time , as the substructure dissolves and the overall cluster relaxes to a radially concentrated density profile . however , this is not evident in the small sample treated here .
taurus has an age much less than its crossing time , which is consistent with its small @xmath0 value and low fractal dimension .
on the other hand , ic2391 and chamaeleon have ages much greater than their crossing times and yet they are still fractal with relatively low @xmath0 values . in contrast , @xmath14 ophiuchus and ic348 , which have ages comparable with their crossing times , are both centrally condensed , with no discernible substructure .
we should , however , caution against drawing firm conclusions from such a small sample
. we also note that young clusters observed at short wavelengths ( i.e. in the optical ) , may appear to have substructure due to patchy observation . therefore long wavelength surveys are preferable for embedded young star clusters .
we have explored two statistical measures for analysing objectively the observed ( i.e. projected ) structures of star clusters .
these measures are based on the mean surface density of companions ( msdc ) , and the minimal spanning tree ( mst ) .
the measures are @xmath2 , the normalised mean separation between stars , and @xmath8 , the normalised mean edge - length of the mst , both of which are independent of the number of stars in the cluster . for artificial star clusters , created with a smooth large - scale radial density profile ( @xmath3 ) , and for artificial star clusters created with sub - structure having fractal dimension @xmath7
, @xmath2 and @xmath8 both decrease with increasing @xmath4 and/or decreasing @xmath7 but at different rates . hence a cluster with a radial gradient can be distinguished from one with sub - structure by evaluating @xmath159 . for a cluster of uniform volume - density ( i.e. @xmath142 and @xmath146 ) , @xmath166 .
if the cluster is made more centrally condensed by increasing @xmath4 , @xmath0 increases monotonically , reaching @xmath167 at @xmath143 .
conversely , if the cluster is given sub - structure by reducing @xmath7 , @xmath0 decreases monotonically , reaching @xmath168 at @xmath147 . on the basis of their @xmath0 values , @xmath14 ophiuchus and ic348
have radial gradients with @xmath169 , and @xmath170 , respectively .
chamaeleon and ic2391 have sub - structure with notional fractal dimension @xmath171 .
taurus has even more sub - structure , with @xmath172 , and if the binaries in taurus are treated as single systems , @xmath25 increases to @xmath26 .
@xmath25 is only a notional fractal dimension , because the integral measures we have defined do not give any indication of whether the sub - structure is hierarchically self - similar .
( indeed , for clusters having only @xmath160 200 stars the range of separations is too small to possess hierarchical self - similarity . )
barrado y navascues d. , stauffer j.r .
, briceno c. , patten b. , hambly n.c .
, adams j.d .
, 2001 , apjs , 134 , 103 barrow j.d .
, bhavsar s.p . ,
sonoda d.h . , 1985 , mnras , 216 , 17 bate m.r .
, clarke c.j . , mccaughrean m.j . , 1997 , mnras , 297 , 1163 baugh c. , mnras , 264 , 87 bontemps s. , nadre p. , kaas a.a . , nordh l. , olofsson g. , huldtgren m. , abergel a. , blommaert j. , boulanger f. , burgdorf m. , cesarsky c.j . , cesarsky d. , copet e. , davies j. , falgarone e. , lagache g. , montmerle t. , perau m. , persi p. , prusti t. , puget j.l . , sibille f. , 1979 , a&a , 372 , 173 brandner w. , khler r. , 1998 , apj , 499 , l79 briceno c. , calvet n. , gomez m. , hartmann l. , kenyon s.j . , whitney b.a .
, 1993 , pasp , 105 , 686 dussert c. , rasigni m. , palmari j. , rasigni g. , llebaria a. , marty f. , 1987 , j theor .
125,317 fry n. , 1979 , tectonophysics , 60 , 89 ghez a.m. , neugebauer g. , matthews k. , 1993 , aj , 106 , 2005 gladwin p.p .
, kitsionas s. , boffin h.m.j . , whitworth a.p . , 1999 ,
mnras , 302 , 305 gomez m. , jones b.f . , hartmann l. , kenyon s.j .
, stauffer j.r . , hewett r. , reid i.n , 1992 , aj , 104 , 762 gomez m. , hartmann l. , kenyon s.j . , hewett r. , 1993 , aj , 106 , 2005 gower j.c . , ross g.j.s
stat . , 18 , 54 graham m.j . , clowes r.g.,campusano l.e . , 1979 , mnras , 275,790 hammersley j.m . , halton j.h .
, beardwood j. , 1959 , camb .
proceedings 55 , 299 - 327 hanna s.s . , fry n. , 1979 , jsg , 1.2 , 155 hartmann l. , stauffer j.r . , kenyon s.j . , jones b.f . , 1991 ,
aj , 101 , 1050 herbig g.h . , bell k.r . , 1988 , lick obs .
1111 hoffman r.,jain a. k. , 1983 , patt .
lett.,1,175 klaoudatou , i. , 2002 , mphil thesis ( cardiff university ) klessen r.s .
, kroupa p. , 2001
, a&a , 372 , 105 larson r.b . , 1995 , mnras , 272 , 213 lawson w.a . , fiegelson e.d . ,
huenemoerder d.p .
, 1996 , mnras , 280 , 1071 leinert ch .
, zinnecker h. , weitzel n. , christou j. , ridgway s.t .
, jameson r. , haas m. , lenzen r. , 1993 , a&a , 278 , 129 luhman k.l . ,
stauffer j.r . ,
muench a.a . ,
rieke g.h . ,
lada e.a . , bouvier j. , lada c.j . , 2003 , astro - ph/0304414 najita j.r . , tiede g.p .
, carr j.s . , 2000 , apj , 541 , 977 simon m. , ghez a.m. , leinert ch . , cassar l. , chen w.p . , howell r.r
jameson r.f .
, matthews k. , neugebauer g. , richichi a. , 1995 , apj , 443 , 625 simon m. , 1997 , apj , 482 , l81 steele j.m . , 1988 , ann .
, 16,1767 waer f.m .
, brown a. , mathieu r.d . , myers p.c .
, vrba f.j . , 1988 ,
aj , 96 , 297
table [ sources ] gives the sources of the positions of stars or , in the case of @xmath14 ophiuchus , protostars used in the analysis of sections 3 and 4 .
these positions are plotted on fig .
[ realfield ] .
graham et al ( 1995 ) have applied the mst to quasar clustering on very large scales , using a method which was developed by dussert et al ( 1987 ) for characterising biological structures . in dussert
s method , the mean @xmath8 and standard deviation @xmath173 of the edge lengths of the mst are first computed and normalised by dividing by the factor @xmath9 , then plotted on the @xmath174-plane .
[ mstbar](a ) , reproduced from dussert et al .
( 1987 ) , shows the theoretical locations on the @xmath174-plane for different types of clustering in two - dimensions .
the region of the @xmath174-plane around the central star represents the locus of a random distribution .
the region of the @xmath174-plane around ` 1 ' represents the locus of distributions dominated by sub - clustering .
the region of the @xmath174-plane around ` 2 ' represents the locus of distributions dominated by radial concentration gradients .
the region of the @xmath174-plane around ` 3 ' represents the locus of distributions dominated by quasi - periodic tilings . and
the region of the @xmath174-plane around ` 4 ' represents the locus of highly organized distributions ( i.e. lattices ) .
[ mstbar](b ) shows the loci on the @xmath174-plane for the various artificial star cluster types and the five real clusters , and reveals some drawbacks to this plot .
the locus for artificial clusters with a radial density gradient do indeed tend towards region 2 with increasing @xmath4 ( i.e. greater degree of central concentration ) , although only for @xmath175 are they clearly distinguishable from a purely random distribution .
similarly , the locus for artificial clusters with fractal sub - clustering tend towards region 1 with decreasing @xmath7 ( i.e. greater degree of sub - clustering ) for @xmath176 .
however , for @xmath177 , this trend is abandoned , and the locus moves towards region 2 ; in other words , a cluster with a low fractal dimension and hence a high degree of sub - clustering masquerades on the @xmath174-plane as a cluster with a strong radial density gradient , albeit it not precisely of the form @xmath3 .
moreover , taurus , which to the human eye appears to have the most well defined sub - clustering of all five real clusters , masquerades on the @xmath174-plane as a cluster with a strong radial density gradient , @xmath178 . | we review a range of stastistical methods for analyzing the structures of star clusters , and derive a new measure @xmath0 which both quantifies , and distinguishes between , a ( relatively smooth ) large - scale radial density gradient and multi - scale ( fractal ) sub - clustering .
the distribution of separations @xmath1 is considered , and the normalised correlation length @xmath2 ( i.e. the mean separation between stars , divided by the overall radius of the cluster ) , is shown to be a robust indicator of the extent to which a smooth cluster is centrally concentrated . for spherical clusters
having volume - density @xmath3 ( with @xmath4 between 0 and 2 ) @xmath2 decreases monotonically with @xmath4 , from @xmath5 to @xmath6 .
since @xmath2 reflects all star positions , it implicitly incorporates edge effects .
however , for fractal star clusters ( with fractal dimension @xmath7 between 1.5 and 3 ) @xmath2 decreases monotonically with @xmath7 ( from @xmath5 to @xmath6 ) .
hence @xmath2 , on its own , can quantify , but can not distinguish between , a smooth large - scale radial density gradient and multi - scale ( fractal ) sub - clustering . the minimal spanning tree ( mst )
is then considered , and it is shown that the normalised mean edge length @xmath8 ( i.e. the mean length of the branches of the tree , divided by @xmath9 , where @xmath10 is the area of the cluster and @xmath11 is the number of stars ) , can also quantify , but again can not on its own distinguish between , a smooth large - scale radial density gradient and multi - scale ( fractal ) sub - clustering .
however , the combination @xmath12 does both quantify , and distinguish between , a smooth large - scale radial density gradient and multi - scale ( fractal ) sub - clustering .
ic348 has @xmath13 and @xmath14 ophiuchus has @xmath15 , implying that both are centrally concentrated clusters with , respectively , @xmath16 and @xmath17 .
chamaeleon and ic2391 have @xmath18 and @xmath19 respectively , implying mild substructure with a notional fractal dimension @xmath20 .
taurus has even more sub - structure , with @xmath21 implying @xmath22 .
if the binaries in taurus are treated as single systems , @xmath23 increases to @xmath24 and @xmath25 increases to @xmath26 .
open clusters and associations : general |
the breathing circle is the time - dependent plane region bounded , instantaneously , by a circle whose radius is a periodic function of time .
the billiard problem on the breathing circle consists on the free motion of a point particle inside this region , colliding elastically with the moving boundary . as j. koiller , r. markarian and ourselves have showed in @xcite , time varying billiard computations can be performed in the same way as done on rigid billiards .
but , beside the usual coordinates describing static billiards ( for the impact point on the boundary and the direction of the movement ) , one must also introduce the energy and time .
if , for each fixed time , the frozen boundary is a strictly convex plane curve , the associated time - dependent billiards can be modeled by a 4-dimensional , volume preserving diffeomorphism which maps successive impacts with the moving boundary . in the special case of the breathing circle ,
the angular momentum is conserved , implying a reduction of the model to a 2-dimensional diffeomorphism .
then , using a corollary of herman s _ thorme des courbes translates _ @xcite we prove that if the motion of the moving boundary is sufficiently smooth then , for any admissible initial condition , the particle will move with bounded velocity and therefore will have a bounded gain of energy .
a similar case was studied by levi in ( @xcite 1 ) , where he shows examples of _ pulsating soft circular billiards _ whose energies stay bounded for all time .
the study of time - dependent billiards is motivated by two main questions . from one side ,
it is a natural extension to higher dimensions of the fermi accelerator , studied for instance in @xcite , @xcite and @xcite . from another side , among many branches of physics , it is of special interest on the study of the motion of particles inside collectively excited nuclei . on the attempts to understand the origin of dissipation of collective motion in finite fermi systems ( see , for instance ,
@xcite and @xcite ) models were constructed where the classical part corresponds to time - dependent billiards , especially with multipole resonances ( with the breathing circle corresponding to the monopole mode ) .
let @xmath0 be a strictly positive @xmath1-periodic @xmath2 function , @xmath3 and @xmath4 be the ( moving ) boundary of the breathing circle .
suppose that at an instant @xmath5 a particle is on @xmath6 at a point @xmath7 .
let the unitary tangent vector at that point be @xmath8 and the unitary outward normal vector be @xmath9 .
the point on @xmath6 is then given by @xmath10 .
suppose that it leaves the circle from @xmath7 with velocity @xmath11 .
it travels on a straight line and hits the boundary again at an instant @xmath12 at @xmath13 with r_n + 1 = r_n + ( t_n + 1 - t_n ) v_n [ eq : r ] and will then rebound with velocity v_n + 1 = v_n - 2 < v_n , _
n + 1 + 2 . |_t = t_n+1 _ n + 1 .
[ eq : reb ] let @xmath14 denote the angular momentum with respect to the center of the breathing circle , then @xmath15 this shows that the angular momentum is conserved at impacts with the moving boundary . as it is also obviously conserved between impacts ( since the particles moves on a straight line with constant velocity ) , angular momentum is conserved along all the movement .
a first easy consequence of this conservation is that a particle never changes orientation with respect to the breathing circle , i.e. , it always rotates in the same direction . in other words , if we introduce the notation @xmath16 , is is easy to verify that @xmath17 is conserved .
another consequence of the conservation of angular momentum is that a particle never stops if its angular momentum is different from zero .
this is because if @xmath18 , as @xmath19 then @xmath20 and @xmath21 , where @xmath22 max @xmath23 .
it also follows that the traveling time of a particle ( between two impacts with the moving boundary ) is bounded , because t_n+1 - t_n = [ eq : trav ] for every @xmath24 .
we introduce the variable @xmath25 .
it is , in a certain sense , a natural variable , because it is the normal counterpart of the conserved quantity @xmath17 . for given @xmath26 and @xmath27 , @xmath28 initial impact variables , let @xmath29 and @xmath30 be the values of @xmath31 and @xmath32 at the next impact with the breathing circle
let @xmath33 [ lem:1 ] for any fixed @xmath34 and for every @xmath35 with @xmath36 , the map @xmath37 is defined by the formulae : & & i_1 = i_0 - w_1 + ( ( t_1 - t_0 ) - 2 i_0 ) [ eq : i ] + & & ( t_1 - t_0 ) ( ( t_1 - t_0 ) - 2 i_0 ) = r_1 ^ 2 - r_0 ^ 2 [ eq : t ] equation ( [ eq : t ] ) above is easily obtained after computing @xmath38 from ( [ eq : r ] ) .
now , if the particle moves inside the breathing circle region , it s normal velocity at the impacts must be bigger than the normal velocity of the boundary , i.e. , @xmath39 or , using the variables introduced above , @xmath40 .
this condition implies that ( [ eq : t ] ) has a solution @xmath41 and thus the map is well defined under this assumption .
( in @xcite we have in fact proved a little more : @xmath40 implies @xmath42 . ) on the other hand , from ( [ eq : r ] ) we have that @xmath43 .
this relation together with the rebound formula ( [ eq : reb ] ) yields ( [ eq : i ] ) . formulae ( [ eq : i ] ) and ( [ eq : t ] ) above are invariant under the translations @xmath44 , where @xmath45 is the period of @xmath0 and @xmath24 is any integer .
so , we can take @xmath46 and then the domain of @xmath47 is contained in the cylinder @xmath48 .
it is worthwhile to point out that , although time - dependent billiards are described by four variables , the breathing circle only takes two , namely @xmath31 and @xmath49 , the other two independent variables being the ( conserved ) angular momentum and its conjugate variable . moreover , as @xmath50 appears only squared in the calculations , we can suppose @xmath51 .
the case @xmath52 corresponds to the movement on the diameter of the breathing circle and is equivalent to the 1-dimensional fermi accelerator , studied , for instance , in @xcite , @xcite and @xcite .
now let @xmath53 and define @xmath54 by @xmath55 .
let a_1 & = & \ { ( i_0 , t_0 ) | i_0 + > 0 i_1 + = 0 } + a_-1 & = & \ { ( i_0 , t_0 ) | i_0 + > 0 i _ - 1 + = 0 } .
( with our assumptions on @xmath0 , the lebesgue measure on @xmath56 of these sets is zero . )
let @xmath57 .
[ lem:2 ] if @xmath0 is a strictly positive @xmath45-periodic @xmath2 function , @xmath58 then for each fixed @xmath34 the map @xmath59 , given by ( [ eq : i ] ) and ( [ eq : t ] ) , is a @xmath60 diffeomorphism , preserving the area @xmath61 .
the proof follows from theorem 1 in @xcite
from the section above , it seems natural to introduce the variable @xmath62 .
the advantages of this choice will become clear in what follows . in the new
coordinates @xmath63 lemma [ lem:1 ] rewrites : * lemma [ lem:1] * _ for a fixed @xmath34 and given an initial condition @xmath64 with @xmath65 , its iterated by the billiard map , @xmath66 , is given by the formulae : _ & & j_1 = j_0 - ( w_0 + w_1 ) + 2 [ eq : j ] + & & ( t_1 - t_0 ) 4 l_0 ^ 2 + ( j_0 - w_0 ) ^2 4 r_0 ^ 2 = j_0 - w_0 + r_1 ^ 2 - r_0 ^ 2 t_1 - t_0 .
[ eq : t1 ] since @xmath0 is at least @xmath67 , there exist @xmath68 and @xmath69 such that r_1 ^ 2 - r_0 ^ 2 = ( t_1 - t_0 ) + ( t_1 - t_0)^2 - ( t_0 - t_1)^2 , [ eq : int ] and @xmath70 for @xmath26 and by inequality ( [ eq : trav ] ) we have | j_1 - j_0 | [ eq : adi ] where @xmath71\}$ ] . * remark
* : suppose that the breathing circle moves slowly , i.e. , @xmath72 , for a small parameter @xmath73 and @xmath74 a @xmath2 , @xmath1-periodic function .
then @xmath75 . let @xmath76 be the value of @xmath77 at the @xmath78 impact .
for @xmath79 @xmath80 which shows that @xmath77 is an adiabatic invariant ( see @xcite for the definition of adiabatic invariant ) .
let @xmath81 denotes the cylinder @xmath82 .
[ lem:3 ] suppose that @xmath0 is a strictly positive @xmath45-periodic @xmath2 function , @xmath3 .
let @xmath83 be defined by formulae ( [ eq : j ] ) and ( [ eq : t1 ] ) .
then @xmath84 is a @xmath60 diffeomorphism , preserving the measure @xmath85 . for a fixed @xmath34
, we have automatically from ( [ eq : adi ] ) that @xmath86 the proof follows immediately from this fact and lemma [ lem:2 ] . given any curve @xmath87 in the cylinder @xmath88 homotopic to @xmath89 then @xmath90 .
in other words , @xmath91 has the property of intersection .
@xmath91 preserves a measure , absolutely continuos with respect to the lebesgue measure .
this implies the property of intersection @xcite .
the purpose of this section is to investigate the behaviour of the breathing circle billiard map in the neighbourhood of @xmath92 .
we shall show that a map can be defined which exhibits some properties of twist maps . more specifically , using a result proved by r. douady @xcite we will prove the existence of invariant curves for values of @xmath93 sufficiently large . for a fixed @xmath34 , and since @xmath94 the change of variables @xmath95 is well defined on @xmath96 and is @xmath97 .
let @xmath98 .
then the map @xmath99 defined by z_1 = z_0 g(t_1 , t_0 ) = - ( w_0 + w_1 ) + 2 [ eq : z ] + ( t_1 - t_0 ) + 4r_0 ^ 2 z_0 ( w_0 z_0 - 1 ) = 4 r_0 ^
2 z_0 ^ 2 [ eq : t2 ] is a @xmath60 diffeomorphism , with the property of intersection .
[ lem : n ] @xmath100 can be extended to a @xmath60 diffeomorphism @xmath101 on a neighbourhood of @xmath102 . @xmath101 has the property of intersection and a development of the form : & z_1 = & z_0 + o ( z_0 ^ 2 ) + & t_1 = & t_0 + 4 r_0 ^ 2 z_0 + o ( z_0 ^ 2 ) first of all , let us show that @xmath103 can be extended for @xmath104 .
@xmath0 being @xmath2 , @xmath3 then @xmath105 is a @xmath106 function with @xmath107 . by hadamard s lemma @xmath108 , where @xmath109 is a @xmath60 function . using ( [ eq : t2 ] ) and discarding the trivial solution @xmath110
, we have : @xmath111 ( t_1 - t_0 ) + 4r_0 ^ 2 z_0 ( w_0 z_0 - 1 ) = 4 r_0 ^ 2 z_0 ^ 2 \left [ w_0 + h(t_1 - t_0 ) \right].\ ] ] let @xmath112 ( t_1 - t_0 ) + 4r_0 ^ 2 z_0 ( w_0 z_0 - 1 ) - 4 r_0 ^ 2 z_0 ^ 2 \left [ w_0 + h(t_1 - t_0 ) \right].\ ] ]
@xmath113 is a @xmath60 function on @xmath114 such that @xmath115 and @xmath116 . if @xmath58 and for each fixed @xmath117 , by the implicit function theorem we have @xmath118 on a neighbourhood of @xmath119 .
moreover , as @xmath120 is compact and @xmath121 for every @xmath122 , we can find a neighbourhood of @xmath102 where @xmath118 is a @xmath60 function .
this implies that @xmath123 has a @xmath60 extension on a neighbourhood of @xmath102 .
clearly it is bounded and , perhaps on a smaller neighbourhood , we can take @xmath124 .
so , @xmath125 has a @xmath60 extension on a neighbourhood of @xmath102 .
analogously , @xmath126 can be extended on a neighbourhood of @xmath102 .
@xmath100 is then extended to a @xmath60 diffeomorphism @xmath101 on a neighbourhood of @xmath102 .
since the formulae remain unchanged , the property of intersection is preserved for curves on @xmath127 or @xmath104 . but
@xmath128 for any @xmath117 .
then , if a curve @xmath87 cuts the circle @xmath102 , @xmath129 , and @xmath101 has the property of intersection .
finally , a simple calculation leads to the development of @xmath101 .
if @xmath130 , we have a family of invariant curves for @xmath101 that _ protect _
@xmath131 , in the following sense : every orbit of @xmath101 with no intersection with @xmath131 , stay at a bounded distance , above and bellow , from @xmath131 .
those bounds are arbitrarily small strictly positive constants .
lemma [ lem : n ] give the conditions to apply a corollary of herman s _ thorme des courbes translates _ , proved by r. douady ( @xcite , page iii - 8) and the result follows .
coming back to the coordinates @xmath63 and the diffeomorphism @xmath91 , let us denote by @xmath132 the points of the orbit of @xmath64 under @xmath91 .
an admissible initial condition for @xmath91 is the set of @xmath64 such that @xmath133 ( which is the condition of existence of the next impact given in lemma [ lem:1] ) .
given the billiard on the breathing circle @xmath134 , with @xmath0 a strictly positive @xmath45-periodic @xmath2 function , @xmath130 , then , for any admissible initial condition , a particle will move with bounded velocity .
for @xmath34 , since @xmath101 has a family of invariant curves that _ protect _ @xmath131 and @xmath135 for @xmath127 , we have a family of invariant curves of @xmath91 that _ protect _
@xmath92 , i.e. , there exist a family of functions @xmath136 , approaching infinity with norm @xmath60 , whose graphs are invariant under @xmath91 .
given @xmath64 such that @xmath137 for some @xmath138 and @xmath139 , we will have that @xmath140 , for every @xmath24 .
therefore @xmath64 is an admissible initial condition and @xmath76 will remain bounded . on the other hand ,
if @xmath64 is bellow all the invariant curves , then @xmath76 will be bounded above although it may be equal to zero .
anyway , for every admissible initial condition , @xmath76 will remain bounded .
if @xmath142 , we have the motion along the diameter of the breathing circle , that corresponds to the 1-dimensional fermi accelerator , which has bounded velocities for any admissible initial condition ( see , for instance , @xcite ) .
* remark * : this result is also true if we consider the billiard in any higher dimension breathing sphere @xmath143 , instead of the circle @xmath120 , since the conservation of the angular momentum implies that the movement will occur on a plane and so , the effective boundary will be the breathing circle . in terms of the dynamics of the breathing circle billiard
, we can translate theorem 1 in the following way : for each fixed @xmath50 , there exists a bound @xmath144 , depending on @xmath0 and decreasing with @xmath50 , such that for @xmath145 , there are infinitely many spanning curves that may not exist for @xmath146 .
this fact is illustrated on the pictures bellow where we plot the phase - space for @xmath147 and different choices of @xmath148 .
( because of the special form of @xmath0 , @xmath77 is , in fact , an eternal adiabatic invariant and , for big @xmath77 s , all curves are invariant . )
clearly our approach does not apply to the low energy regime ( i.e. , for small @xmath148 s and/or small @xmath93 s ) . for the 1-dimensional case ,
a simplified model was introduced by lichtemberg and liebermann @xcite .
this approximation and some comparisons with the standard map seems to give some valuable insights about this energy regime . in the simplified model
one allows the boundary to interact with the particle through momentum exchange ( via a given @xmath149 periodic in time ) , but assuming that the boundary does not change in time and so @xmath150 . in this case , the associated billiard map is given by & & j_1 = j_0 - ( w_0+w_1 ) + & & t_1 = t_0 + ( j_0-w_0 ) .
this simplified model allows unphysical negative values of @xmath93 , which are forbidden in the complete model . on the next set of pictures
we show the phase - space of the simplified model with the same conditions and @xmath149 we took in figure [ fig : mov ] . except for a small neighbourhood of @xmath151 ,
the dynamical behaviour seems astonishingly similar .
finally we point out that the existence of a conserved quantity ( the angular momentum ) is the clue of our proof , since it reduces the dimension to two .
unfortunately , it is easy to prove that a 2-dimensional moving billiard , with convex frozen boundary and with only normal deformation , preserves the angular momentum if and only if it is the breathing circle .
so , the extension of this result to other time dependent billiards will depend on the existence of some constant of motion or to the extension of the techniques to higher dimensions .
99 v. i. arnold : small denominators and problems of stability of motion in classical and celestial mechanics , russ . math .
surveys * 18:6 * , 85 - 191 ( 1963 ) .
j. blocki , y. boneh , j. r. nix , j. randrup , m. robel , a. j. sierk , w. j. swiatecki : one - body dissipation and the super - viscidity of nuclei , annals of physics * 113 * -2 , 330 - 386 ( 1978 ) . g. f. burgio , m. baldo , a. rapisarda : chaoticity in vibrating nuclear billiards , phys.rev.c , * 52 * , 2475- ( 1995 ) .
r. douady : applications du thorme des tores invariants , thse de 3me cycle , univ .
paris vii ( 1982 ) .
j. koiller , r. markarian , s. oliffson kamphorst , s. pinto de carvalho : time - dependent billiards , nonlinearity * 8 * , 983 - 1003 ( 1995 ) .
m. levi : quasiperiodic motions in superquadratic time - periodic potentials , comm .
phys , * 143 * 43 - 83 ( 1991 ) .
a.j . lichtenberg , m.a .
liebermann : _ regular and stochastic motion _ , springer - verlag appl .
sci . * 38 * , berlin , heidelberg , new york , springer - verlag ( 1983 ) .
t. krger , l.d .
pustylnikov , s.e .
troubetzkoy : acceleration of bouncing balls in external fields , nonlinearity , * 8 * , 397 - 410 ( 1995 ) .
pustylnikov : on ulam s problem , theor .
phys . * 57 * , 1035 - 1038 ( 1983 )
pustylnikov : on the fermi - ulam model , soviet math
* 35:1 * , 88 - 92 ( 1987 ) . | the breathing circle is a 2-dimensional generalization of the fermi accelerator .
it is shown that the billiard map associated to this model has invariant curves in phase space , implying that any particle will have bounded gain of energy .
= 0 cm 16truecm = 0truecm = 0truecm = 22truecm plus 1pt minus 1pt [ proposition ] lemma [ proposition ] corollary = 5 mm = 5 mm |
leading order ( lo ) qcd predicts photon interactions to have a two - component nature . in direct
photon processes the whole of the photon takes part in the hard subprocess with a parton from the proton whereas in resolved photon processes , the photon acts as a source of partons and one of these enters the hard subprocess ( see figure [ fig : diresfig ] ) .
both lo processes are characterized by having two outgoing partons of large transverse energy .
previous studies of dijet photoproduction at hera have shown that both classes of process are evident for the case of quasi - real photons ( those of negligible virtuality @xmath0 ) @xcite .
the parton content of photons is neither well constrained theoretically nor well known experimentally , particularly for photons with small but non - zero virtualities .
various theoretical predictions exist for the behaviour of the photon structure as a function of the photon virtuality @xcite .
the general expectation is that the contribution to the dijet cross section of resolved photon processes should decrease relative to the contribution from direct photon processes ( _ i.e. _ that the partonic content of the photon is suppressed ) as the virtuality of the photon increases .
the data used in this analysis were collected during the 1994 run when hera collided 27.5 gev positrons with 820 gev protons .
the zeus detector is described elsewhere@xcite .
a tungsten - silicon sampling calorimeter was installed for the 1994 running period .
this beampipe calorimeter tagged positrons scattered through small angles ( 17 - 35 mrad ) and gave a sample of events with photon virtualities in the range @xmath7 gev@xmath2 .
a sample of events with quasi - real photons ( @xmath3 gev@xmath2 with a median of @xmath8 gev@xmath2 ) was obtained by requiring that the scattered positron be detected in the downstream luminosity calorimeter@xcite .
jets were found in the main uranium - scintillator@xcite calorimeter using a cone algorithm @xcite in @xmath9 - @xmath10 space , where @xmath11 is azimuth and the pseudo rapidity @xmath12 , @xmath13 being defined with respect to the proton direction . only those events with two or more jets of transverse energy @xmath4 gev in the range @xmath14 were selected as dijet events . to reduce contamination from beam gas and deep inelastic scattering events , a cut of @xmath15 was applied , where the sum is over calorimeter cells with deposits of total and longitudinal energy @xmath16 and @xmath17 respectively and where @xmath18 is the positron beam energy . in photoproduction , @xmath19 is the fraction of the positron energy carried by the photon .
these cuts left a sample of 375 events with virtual photons and a sample of 14181 events with quasi - real photons , corresponding to respective integrated luminosities of 2.07 pb@xmath20 and 2.19 pb@xmath20 .
for each dijet event the fraction of the photon momentum manifest in the two highest @xmath6 jets , @xmath5 was calculated .
@xmath5 is defined by @xmath21 we measure @xmath22 and @xmath23 using the raw calorimeter energies and use the jaquet - blondel method , @xmath24 above , to measure @xmath19 .
no corrections are made for detector effects .
uncorrected distributions of @xmath5 are shown in figure [ fig : xgfig ] a ) for events with virtual photons and figure [ fig : xgfig ] b ) for events with quasi - real photons .
events at high @xmath5 are associated with direct photon processes while those at low @xmath5 are associated with resolved photon processes .
it is clear that both classes of event are present in both @xmath0 ranges . to quantify the relative contributions of these classes of event
, we have calculated the ratio @xmath25 , defined as the number of events at low @xmath5 ( @xmath26 ) divided by the number of events at high @xmath5 ( @xmath27 ) .
figure [ fig : resdirfig ] shows @xmath25 as a function of @xmath0 .
@xmath25 is independent of @xmath24 for the sample of events with quasi - real photons passing all the cuts applied .
this implies that the difference in the @xmath5 distributions for the two samples is not due to the differing @xmath24 acceptances of the two positron detectors .
the acceptance corrections that will eventually be applied to these data are only weakly dependent on @xmath5 and we therefore expect the corrections to @xmath25 to be small .
more work is required to understand fully these corrections however . in conclusion
we find the preliminary result that for events with photons of virtuality @xmath28 gev@xmath2 there is a contribution from resolved photon processes .
the size of this contribution relative to that from direct photon processes seems to decrease with increasing photon virtuality .
99 zeus collab .
, m.derrick et al .
, , m.drees and r.godbole , t. walsh , f. borzumati and g. schuler , zeus collab . , the zeus detector , status report ( 1993 ) m.derrick et al .
, a.andressen et al .
, a.bernstein et al .
, j.andruszkw et al . ,
desy 92 - 066 ( 1992 ) j.huth et al . , _
proc 1990 dpf summer study on h.e.phys_ , snowmass , ed e.l.berger , 1992 , p. 134 | preliminary results are presented from a study of dijet photoproduction in ep collisions with both virtual and quasi - real photons at the zeus detector .
samples of events with photons of virtuality @xmath0 in the ranges @xmath1 gev@xmath2 and @xmath3 gev@xmath2 having two jets of @xmath4 gev in the final state have been obtained . for both quasi - real and virtual photons , uncorrected distributions of the quantity @xmath5 , the fraction of the photon momentum manifest in the two highest @xmath6 jets ,
are presented .
these distributions are sensitive to the relative contributions of the direct and resolved processes .
resolved photon processes are evident in both data sets , with an apparent decrease in the relative contribution from resolved processes as photon virtuality increases .
# 1#2#3 _ ann . phys .
( ny ) _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ astrophys .
j. _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ astrophys .
j. lett .
_ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ acta .
phys . pol .
_ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ ann .
rev .
nucl . part .
sci . _ * # 1 * ( # 2 ) # 3 . # 1#2#3 _ computer phys .
comm . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ erratum _ * # 1 * ( # 2 ) # 3 . #
1#2#3 _ ibid . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ j. math .
phys .
_ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ int .
j. mod .
phys . _
* # 1 * ( # 2 ) # 3 .
# 1#2#3 _ jetp lett . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ j. phys
. g. _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ mod .
phys .
lett . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ nature ( london ) _ * # 1 * ( # 2 ) # 3 . # 1#2#3 _ nuovo cim . _ * # 1 * ( # 2 ) # 3 . # 1#2#3 _ nucl .
instr .
meth .
_ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ nucl .
phys . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ proc .
cam .
phil .
soc .
_ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ phys .
lett . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ phys .
rep . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ phys .
rev . _
* # 1 * ( # 2 ) # 3 .
# 1#2#3 _ phys .
rev .
lett . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ proc .
roy .
soc . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ prog .
th .
phys .
_ * # 1 * ( # 2 ) # 3 . ps . #
1#2#3 _ physica scripta _ * # 1 * ( # 2 ) # 3 . # 1#2#3 _ rev .
mod .
phys . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ rep .
prog .
phys . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ sov .
j. nucl .
phys . _ * # 1 * ( # 2 ) # 3 .
# 1#2#3 _ sov .
phys .
jept _
* # 1 * ( # 2 ) # 3 .
# 1#2#3 _ sov .
phys .-
usp . _
* # 1 * ( # 2 ) # 3 .
# 1#2#3 _ zeit .
phys . _ * # 1 * ( # 2 ) # 3 .
glas
ppe/9503 talk given at the europhysics conference on high energy physics 27th july - 2nd august 1995 on behalf of the zeus collaboration |
co - nonsolvency ( insolubility of polymer in a mixture of two good solvents ) is one of the most intriguing phenomena in physical chemistry of polymers . despite the great efforts in both experimental @xcite and theoretical @xcite investigations in understanding the co - nonsolvency ,
the mechanism of this phenomenon remains unclear untill now @xcite .
basing on the results of the experimental researches devoted to a behavior of the single pnipam polymer chain @xcite as well as a behavior of pnipam hydrogels @xcite , tanaka et al .
@xcite formulated a quasi - chemical model of a single pnipam chain in a mixed water - methanol solvent .
authors showed that the co - nonsolvency occurs due to a competition of water and methanol molecules for hydrogen bonding with polymer backbone .
thus , it seemed to be that the co - nonsolvency is caused by the hydrogen bonding between the solvent molecules and monomers .
however , in recent papers of mukherji et al .
@xcite by using md computer simulations of lennard - jones polymer chain dissolved in two - component lennard - jones mixed solvent was shown that co - nonsolvency can take place even in the polymer solutions without hydrogen bonding , but it may be driven by the universal van - der - waals and excluded volume interactions only .
authors established that a microscopic parameter which mainly determines an availability of the co - nonsolvency is a difference between energetic parameters of polymer - co - solvent and polymer - solvent attraction , i.e. , @xmath0 .
moreover , they showed that at the sufficiently large value of this difference ( or sufficiently low temperature ) the coil - globule transition occurs as a first - order phase transition @xcite .
thus , one can conclude that the co - nonsolvency is a generic physical phenomenon which can be caused by only universal van - der - waals and excluded volume interactions between molecules of solvent species and monomers @xcite .
the avalability of co - nonsolvency in the mixture n , n dimethylformamide / cyclohexane / polystyrene additionally indicates on the correctness of the latter conclusion @xcite .
mukherji et al . interpreted results of their md simulation by using a simple analytical lattice adsorption model @xcite , taking into account so - called bridging mechanism which implies that one co - solvent molecule can be strongly associated with two monomers . in other words , from authors point of view the co - nonsolvency usually occurs due to an association of co - solvent molecules with the polymer backbone .
it should be noted that within such interpretation bridging-induced coil - globule transition is similar to the electrostatically driven coil - globule transition in polyelectrolyte solutions @xcite .
recently , basing on the full atomistic md simulation of the pnipam chain in mixed water - methanol solvent , thorough analysis of entropy and enthalpy contributions to the solvation free energy at the level of linear response approximation for the frozen coil and globule states of polymer chain was provided @xcite .
authors showed that in the region of co - nonsolvency , when a collapse of polymer coil takes place the energetics of electrostatic , hydrogen bonding , or bridging - type interactions with the globule is found to play no role .
instead , preferential methanol binding results in a significant increase of the globule configurational entropy , stabilizing methanol - enriched globular structures over wet globular structures in neat water @xcite .
thus , there is an ambiguity in the interpretation of the co - nonsolvency microscopic mechanism .
however , the main goal of this paper is not to consider the microscopic mechanism of the co - nonsolvency , but to understand its thermodynamic nature more deeply . to reach our goal
, we calculate the solvation free energy of the polymer chain as well as its enthalpic and entropic contributions as the functions of mixed solvent composition in a region of co - nonsolvency . to the best of our knowledge
, this problem has not been considered from the first principles of the statistical thermodynamics till now . in order to consider the above - mentioned problem we develop the off - lattice statistical model of the single polymer chain in mixed binary solvent .
taking into account the conformational entropy of polymer chain and renormalization of the co - solvent mole fraction near the polymer backbone , we show that the co - nonsolvency can be successfully described within our self - consistent field theory .
we show that , when the collapse of polymer chain takes place , the entropy and enthalpy contributions to the solvation free energy of globule strongly grow , almost compensating each other . thus we obtain that from thermodynamic point of view the co - nonsolvency is the essentially enthalpic - entropic effect and caused by so - called enthalpy - entropy compensation @xcite . we also show that at sufficiently high temperature a reentrant coil - to - globule - to - coil transition proceeds smoothly . on the contrary ,
when the temperature drops below a certain threshold value , the coil - globule transition occurs in the regime of first - order phase transition , i.e. , discontinuous changes of the radius of gyration and the local co - solvent mole fraction that confirms earlier results of md simulations @xcite .
we calculate the solution heat capacity change due to the solvation of polymer chain as a function of the temperature and show that it is in qualitative agreement with the experimental differential scanning calorimetry ( dsc ) data for pnipam in aqueous methanol @xcite .
we consider an isolated polymer chain with a degree of polymerization @xmath1 immersed in a low - molecular weight two - component solvent at a specified number density @xmath2 and temperature @xmath3 that are located at fluid state region .
so the polymer chain in our model is dissolved in a mixture of solvent and co - solvent which are good ones for the polymer chain .
thus , a co - solvent concentration in the bulk solution is @xmath4 , while a solvent concentration is @xmath5 , where @xmath6 is a co - solvent mole fraction in the bulk . like in recent works @xcite we assume for convenience that the volume of system consists of two parts : the gyration volume @xmath7 ( @xmath8 is a radius of gyration of the polymer chain ) containing predominantly monomers of the polymer chain and the bulk solution . to exclude from the consideration a number density change ( that can take place near the polymer chain @xcite ) which unimportant for this research , we assume that the entire polymer solution is incompressible , so that the solvent number density in the gyration volume can be determined by the relation @xmath9 , where @xmath10 is a monomer number density . moreover
, we introduce a local co - solvent mole fraction @xmath11 by the relations @xmath12 and @xmath13 , where @xmath14 and @xmath15 are the local number densities of the solvent and co - solvent in the gyration volume , respectively .
we also assume that the pair interaction potentials for monomer - monomer , monomer - solvent , monomer - co - solvent , solvent - solvent , co - solvent - co - solvent , and solvent - co - solvent have a following form @xmath16 where @xmath17 ; @xmath18 ; @xmath19 and @xmath20 are effective diameters and energetic parameters , respectively . as well as in our previous work @xcite , we assume that @xmath21 , whereas each energetic parameter @xmath20 is considered as independent . within the present study we do not introduce the second virial coefficients as the parameters of interaction @xcite , but as in the work @xcite we construct the total free energy by using different expressions which are straightforwardly related to repulsive and attractive parts of interaction potentials ( [ eq : pot ] ) .
moreover , in contrast to the works @xcite we take into account the both solvent species explicitly
. a conditional solvation free energy of the polymer chain can be written as : @xmath22 @xmath23 where @xmath8 is the radius of gyration of the polymer chain , @xmath24 and @xmath25 are molecule numbers of the solvent and co - solvent in the gyration volume , respectively ; @xmath26 is the ideal free energy of the polymer chain and mixed solvent which can be calculated in the following way @xmath27 @xmath28 where @xmath29 is the expansion factor , @xmath30 is the mean - square radius of gyration of the ideal gaussian polymer chain , @xmath31 is the kuhn length of the segment , @xmath32 is the boltzmann constant , @xmath3 is the absolute temperature , @xmath33 and @xmath34 are the de broglie wavelengths of the solvent species .
the first term in ( [ eq : fid ] ) is the free energy of the ideal gaussian polymer chain within the fixman approximation @xcite ; @xmath35 is a pressure in the bulk solution which will be determined below .
the excess free energy of polymer solution takes the form @xmath36 where @xmath37 is a contribution of the repulsive interactions in the gyration volume due to the excluded volume of the monomers and molecules of solvent species which we determine through the mansoori - carnahan - starling - leland equation of state for the hard - spheres mixture ( see supporting information ) @xcite .
the use of the latter equation of state instead the virial equation of state @xcite allows us to take into account more precisely the packing effects at the large density of solution @xmath2 . the contribution of attractive interactions @xmath38 we determine within the standard mean - field approximation as follows @xmath39 where the interaction parameters @xmath40 can be determined by the standard rule : @xmath41 where the van - der - waals volumes @xmath42 are introduced ; @xmath17 .
we determine the equilibrium values of the expansion factor @xmath43 and of the local co - solvent mole fraction @xmath11 by the minimization of solvation free energy @xmath44 ( see supporting information ) .
it is worth noting that for the practical calculations the local co - solvent mole fraction @xmath11 may be related to the excess coordination number of co - solvent @xmath45 which is usually used to quantify the stability of given polymer state and describe the preferential binding of co - solvent _ vs. _ solvent to the polymer surface in real polymer solutions @xcite .
turning to the numerical calculations , we introduce the dimensionless parameters : @xmath46 , @xmath47 , @xmath48 , @xmath49 , @xmath50 . following the papers of mukherji et al .
@xcite , we use the effective diameters of species : @xmath51 , @xmath52 .
the latter choice approximately corresponds to the pnipam chain dissolved in the water - methanol mixed solvent . to get the co - nonsolvency regime ,
is needed to take the polymer - co - solvent energetic parameter larger than other ones .
thus , we choose the following values of the energetic parameters : @xmath53 , @xmath54 , so that @xmath55 .
we also assume that the degree of polymerization of the polymer chain @xmath56 .
figures 1(a , b ) illustrate the dependences of expansion factor @xmath43 and co - solvent mole fraction @xmath11 in the gyration volume on the bulk co - solvent mole fraction @xmath6 at the different values of temperature @xmath57 under the fixed pressure @xmath58 . as is seen , at sufficiently high temperature at increasing co - solvent mole fraction
@xmath6 the reentrant coil - to - globule - to - coil transition proceeds smoothly , while the co - solvent mole fraction @xmath11 in the gyration volume monotonically increases .
however , when the temperature drops below a certain threshold value , the polymer chain undergoes the coil - globule transition in a regime of first - order phase transition .
namely , when the discontinuous decrease in the expansion factor takes place , the local mole fraction of co - solvent in the gyration volume simultaneously abruptly increases . at further increase in the mole fraction of co - solvent in the bulk solution , the expansion factor and the mole fraction of co - solvent in the gyration volume
smoothly increase .
it is worth noting that increase in the temperature leads to the shift of threshold co - solvent mole fraction at which the coil - globule transition takes place to higher values .
this trend qualitatively agrees with the results of the both experiment and md computer simulation @xcite .
we would also like to stress that an availability of the region where the increase in bulk co - solvent mole fraction has only minor effect on the local solvent / co - solvent composition is a natural consequence of the co - solvent molecules preferential binding with the polymer backbone . indeed , in the case of strong attractive interactions between polymer and co - solvent , the co - solvent - induced coil - globule transition accompanied by the significant increase of the co - solvent concentration within the polymer volume @xcite .
so when the polymer chain adopts a compact globular conformation , further increase in the bulk co - solvent mole fraction should lead to only minor increase in the local co - solvent concentration due to the excluded volume effects .
moreover , the above - mentioned behavior of the local co - solvent mole fraction is in qualitative agreement with the results of md simulations of conformation transition of pnipam hydrogel as a function of the methanol mole fraction in water / methanol mixtures @xcite .
as already pointed out above , the abrupt increase in the local co - solvent mole fraction additionally confirms the results of works @xcite that in the case of strong polymer - co - solvent attraction the co - solvent molecules must be enriched in the globule .
nevertheless , an implicit accounting for the solvent molecules does not allow us to obtain the reentrant coil - to - globule - to - coil transition , but only co - solvent induced coil - globule transition @xcite . in our opinion , it may be related to the fact that implicit solvent models do not take into account the preferential binding co - solvent _ vs _ solvent to the polymer chain which , it seems , should play a crucial role in the co - nonosolvency .
moreover , presence of the abrupt decrease in expansion factor predicted by present theory confirms the earlier md simulations results @xcite . finally before we pass to the discussion of thermodynamic aspects of the co - nonsolvency , it is instructive to discuss how the difference between the energetic parameters of attraction polymer - co - solvent and polymer - solvent @xmath59 influences on the coil - to - globule - to - coil transition .
we obtain ( see fig.2 ) that increase in the difference @xmath59 leads to more pronounced coil - to - globule - to - coil transition that also confirms the recent md simulations results @xcite .
it should be noted that increase in the degree of polymerization @xmath1 also leads to more pronounced the coil - to - globule - to - coil transition . in order to understand a thermodynamic nature of co - nonsolvency
, we discuss a behavior of entropic and enthalpic contributions to the solvation free energy of the polymer chain in the region where the reentrant coil - to - globule - to - coil transition takes place .
we shall discuss below the solvation free energy per monomer @xmath60 , where @xmath61 and @xmath62 are entropy and enthalpy of solvation per monomer , respectively .
on the figures 3a , b are depicted the dependences of solvation entropy and solvation enthalpy on the co - solvent mole fraction at the fixed temperatures @xmath63 ( fig .
3a ) and @xmath64 ( fig .
3b ) and the pressure @xmath58 at a region of co - nonsolvency . as is seen , enthalpy @xmath65 and entropy @xmath66 contributions are strongly oscillating functions of the co - solvent mole fraction @xmath6 within the region of co - nonsolvency in both presented cases . for instance , when the co - solvent mole fraction increases , the solvation enthalpy @xmath65 at first monotonically decreases , attains a minimum , abruptly increases to a maximum , and further monotonically decreases .
the entropy contribution @xmath67 behaves analogously .
it should be noted that an abrupt increase ( decrease ) of the enthalpy ( entropy ) contribution corresponds to the coil - globule transition , occuring in the regime of first - order phase transition .
in contrast to the enthalpy and entropy of solvation , the free energy of solvation in the co - nonsolvency region close to zero .
the latter means that entropy and enthalpy contributions almost compensate each other .
thus the co - nonsolvency has to be considered as complex thermodynamic process driven by entropy - enthalpy compensation that confirms the speculations presented in works @xcite .
it is interesting to discuss the change of solution heat capacity @xmath68 due to the polymer chain solvation as a function of the temperature at different solvent composition @xmath6 in the co - nonsolvency region .
we obtain ( see fig.4 ) that each curve @xmath69 has pronounced minimum which decreases with increase in the bulk co - solvent mole fraction @xmath6 .
such behavior of @xmath70 is in qualitative agreement with the experimental data on the heat flow of pnipam in aqueous methanol obtained by differential scanning calorimetry ( dsc ) method @xcite .
taking into account the effects of conformational entropy , renormalizing the solvent composition near the polymer backbone , and universal intermolecular excluded volume and van - der - waals interactions within the self - consistent field theory , we have described the reentrant coil - to - globule - to - coil transition of polymer chain in mixture of the good solvents .
we show that , when the collapse of polymer chain takes place , the entropy and enthalpy contributions to the solvation free energy of globule strongly grow almost compensating each other . from the first principles of statistical thermodynamics we confirm earlier speculations based on the md simulations results that the co - nonsolvency is the essentially enthalpic - entropic effect and caused by enthalpy - entropy compensation .
however , we would like to discuss the limitations of the present self - consistent field theory .
first , in this work we use the unrealistic potentials of interaction between the particles of solution . in order to apply this theory to real polymer solutions
, one can take the more realistic lennard - jones potentials using standard weeks - chandler - andersen procedure @xcite .
however , in present work we have demonstrated only the principle possibility to describe the co - nonsolvency within self - consistent field theory , retaining its application to the experimental systems for the future researches .
second , our theory in present form is based on the assumption that the entire polymer solution is incompressible .
such assumption may be correct , when the polymer solution is under ambient pressure , whereas the incompressibility approximation must be invalid at the region of extremely high pressures @xcite .
recently was experimentally observed @xcite and confirmed by full atomistic md simulations @xcite that co - nonsolvency of pnipam in aqueous methanol can be suppressed by application of sufficiently high pressure ( order of 500 @xmath71 ) . to describe this very interesting phenomenon
theoretically it is necessary to go beyond the incompressibility approximation that is a subject of the forthcoming publications .
here we present some calculation details omitted in the main text .
we start from the conditional solvation free energy of polymer chain in the mixed solvent media @xmath72 where @xmath73 is the volume of gyration of the polymer chain , @xmath24 and @xmath25 are molecule numbers of the solvent and co - solvent in the gyration volume , respectively ; @xmath26 is the ideal free energy of the polymer chain and mixed solvent which can be calculated in the following way @xmath27 @xmath28 where @xmath29 is the expansion factor , @xmath30 is the mean - square radius of gyration of the ideal gaussian polymer chain , @xmath31 is the kuhn length of the segment , @xmath32 is the boltzmann constant , @xmath3 is the absolute temperature , @xmath33 and @xmath34 are the de broglie wavelengths of the solvent species .
the first term in ( [ eq : fid ] ) is the free energy of the ideal gaussian polymer chain within the fixman approximation ; @xmath35 is the pressure in the bulk solution which will be determined below .
the excess free energy of polymer solution takes the form @xmath36 where @xmath37 is the contribution of the repulsive interactions in the gyration volume due to the excluded volume of the monomers and molecules of solvent species which we determine through the mansoori - carnahan - starling - leland equation of state for the hard - spheres mixture ( see below ) . the contribution of attractive interactions @xmath38 we determine within the standard mean - field approximation as : @xmath39 where the interaction parameters @xmath40 can be determined by the standard rule : @xmath41 where the van - der - waals volumes @xmath42 are introduced ; @xmath17 . choosing the local mole fraction of co - solvent @xmath11 in the gyration volume and the expansion factor @xmath43 as the order parameters
, one can rewrite the solvation free energy in the following way @xmath74 @xmath75 @xmath76 where @xmath77 is a monomer number density and @xmath78 is a density of excess free energy which has a form @xmath79 @xmath80 where the following short - hand notations are introduced
@xmath81 @xmath82 @xmath83 @xmath84 @xmath85 @xmath86 @xmath87 @xmath88 @xmath89 the local solvent composition @xmath11 in the gyration volume is introduced by the following relations @xmath90 the pressure in the bulk solution @xmath35 in our model is determined by the following equation of state : @xmath91 @xmath92 where the first term in eq .
( [ eq : p ] ) determines a pressure of the two - component hard spheres mixture within the mansoori - carnahan - starling - leland equation of state ; the second term determines the contribution of attractive interactions to the pressure within the mean - field approximation .
the chemical potentials of the solvent species can be calculated by the following obvious thermodynamic relations @xmath93 @xmath94 where @xmath95 is a density of helmholtz free energy of the bulk solution which can be calculated as @xmath96 @xmath97 this work was supported by grant from the president of the russian federation ( no mk-2823.2015.3 ) .
the part concerning development of theoretical model has been supported by russian scientific foundation ( grant no 14 - 33 - 00017 ) . | we present an off - lattice statistical model of a single polymer chain in mixed solvent media .
taking into account a polymer conformational entropy , renormalization of solvent composition near the polymer backbone , the universal intermolecular excluded volume and van - der - waals interactions within the self - consistent field theory the reentrant coil - to - globule - to - coil transition ( co - nonsolvency ) has been described in this paper . for convenience
we split the system volume in two parts : the volume occupied by the polymer chain and the volume of bulk solution .
considering the equilibrium between two sub - volumes , the polymer solvation free energy as a function of radius of gyration and co - solvent mole fraction within internal polymer volume has been obtained . minimizing the free energy of solvation with respect to its arguments , we show two qulitatively different regimes of co - nonsolvency .
namely , at sufficiently high temperature a reentrant coil - to - globule - to - coil transition proceeds smoothly . on the contrary ,
when the temperature drops below a certain threshold value a coil - globule transition occurs in the regime of first - order phase transition , i.e. , discontinuous changes of the radius of gyration and the local co - solvent mole fraction near the polymer backbone .
we show that , when the collapse of polymer chain takes place , the entropy and enthalpy contributions to the solvation free energy of globule strongly grow . from the first principles of statistical thermodynamics we confirm earlier speculations based on the md simulations
results that the co - nonsolvency is the essentially enthalpic - entropic effect and caused by enthalpy - entropy compensation .
we show that the temperature dependences of solution heat capacity change due to the solvation of polymer chain are in qualitative agreement with the differential scanning calorimetry data for pnipam in aqueous methanol . |
gamma ray burst mystery lays in its huge energy fluence , sharp variability , extreme cosmic distances and very different morphology .
a huge isotropic explosion disagree with the shortest millisecond time scales and non - thermal spectra .
the huge grbs powers as grb@xmath24 made the final collapse of the fireball model .
new families of fire - ball jet ( and their different label names like hyper - nova , supra - nova , collapsar ) models alleviated the energy budget , by a mild jet beaming . however in this compromise attitude the puzzle of the grb@xmath6/sn@xmath7bw ( which require a very thin jet observed off - axis ( fargion & salis 1995),(fargion1999 ) , has been cured by most skeptical authors by a tenacious cover - up ( neglecting or refuting the grb - sn existence ) or claiming the co - existence of an new zoo of grbs ( bloom 2000),(kulkarni ) .
these new compromised fountain - like fireball model has been collimated in a jet within a @xmath25 angle beam , as a soft link between past fireball and emerging jet
. however the apparent required grb power output is still huge ( @xmath26 erg @xmath10 ) , nearly @xmath27 more intense than other known maximal explosion power ( the super - nova one ) .
more and more evidences in last years and more recently have shown that super - nova might harbor a collimated jet gamma ray burst ( grb@xmath6/sn@xmath7bw , grb@xmath4/sn@xmath5dh ) . to combine the super - nova luminosity and the apparent huge grbs power one need a very much thinner beam jet , as small as a solid angle @xmath28 @xmath29 or @xmath30 respect to @xmath31,(corresponding to a jet angle @xmath32 ) .
there is a statistical need ( fargion 1999 ) to increase the grb rate inversely to the beam jet solid angle .
the needed sn rate ( to explain grbs ) may even exceed the observed one ( at least sn type ib andic ) event in our observable universe ( @xmath33 @xmath34 ) . indeed assuming that only a fraction of the sn ( with optimistic attitude @xmath35 of all known sn ) experience an asymmetric jet - sn explosion , than the corresponding observed rate @xmath36 @xmath37 and @xmath38 @xmath39 imply @xmath40 a result nearly @xmath41 order of magnitude larger than the observed sn rate . in this frame
one must assume a grb jet with a continuous active , decaying life - time much larger than grbs duration itself at least by a corresponding scale @xmath42 .
indeed we considered grbs ( as well as soft gamma repeaters sgr ) as very thin blazing ( @xmath43 ) precessing gamma jets spinning and precessing ( fargion & salis 1995,1996,fargion 1999 ) ; in this scenario grbs are born within a super - nova power collimated inside a very thin beam able to blaze us by an apparent grb intensities .
the inner angle geometrical dynamics while spinning and precessing induce the wide @xmath0 burst variability able to fit the very different observed grbs ones .
the averaged @xmath0 jet deflection from the axis of sight defines a main early power law decay ; an inner damped oscillatory substructure may be observed , as the amazing damping oscillatory afterglows in grb@xmath4 .
the very thin and collimated and long life decaying jet ( opening angle @xmath44 , whose decay power law life - time of a few hours occurs with an exponent @xmath45 ) , while spinning and precessing at different scale times , it may trace and may better explain the wobbling of the @xmath0 grbs and the long train of damped oscillations of the x tail afterglows within hours , the optical transient during days and weeks later .
the grbs re - brightening are no longer a mystery as in a one - shoot model .
these wobbling signatures may be also be found in rarest and most powerful and studied sgrs events .
the spread and wide conical shape of these precessing twin jets may be recognized in a few relic snrs as in the twin sn 1987a wide external rings , the vela arcs and the spectacular egg nebula dynamical shape .
we imagine the grb and sgr nature as the early and the late stages of jets fueled first by sn event and later by an asymmetric accretion disk or a companion ( wd , ns ) star .
the binary angular velocity @xmath46 bends the angle as @xmath47 or more generally a multi - precessing angle @xmath48:@xmath49 + @xmath50 + @xmath51 + where @xmath52 is the minimal impact angle parameter of the jet toward the observer , @xmath53 , @xmath54 , @xmath55 are , in the order , the maximal opening precessing angles due to the binary , spinning pulsar , nutation mode of the multi - precessing jet axis .
the arbitrary phase @xmath56 , @xmath57 , @xmath58 , for the binary , spinning pulsar and nutation , are able to fit the complicated grbs flux evolution in most grb event scenario .
naturally it is very possible to enlarge the parameter to a fourth precession angular component whose presence may better fit the wide spread of scale variability ; here we shall constrains to a three parameter precession beam . for a 3d pattern and its projection along the vertical axis in an orthogonal 2d plane see following descriptive pictures .
the different angular velocities are combined in the multi - precession wobbling .
each bending component is keeping memory either of the pulsar jet spin angular velocity ( @xmath59 ) and its opening angle @xmath54 , its nutation speed ( @xmath60 ) and nutation angle @xmath55 ( due to possible inertial momentum anisotropies or beam - accretion disk torques ) ; a slower precession by the binary @xmath46 companion ( and its corresponding open angle @xmath53 ) will modulate the overall jet precession . on average , from eq.(3 ) the @xmath0 flux and the @xmath61 , optical afterglow are decaying on time as @xmath62 , where @xmath63 ; however the more complicated spinning and precessing jet blazing is responsible for inner small scales wide morphology of grbs and sgrs as well as their partial internal periodicity . the consequent @xmath0 time evolution and spectra derived in this ideal models
may be compared successfully with observed grb data evolution .
the @xmath0 jet is born mainly by inverse compton scattering by gevs electron pairs onto thermal photons ( fargion 1995 - 96 - 98 - 99 ) in nearly vacuum space .
therefore these electron pairs are boosted in the jet at lorentz factor @xmath64 .
their consequent inverse compton scattering will induce a parallel @xmath0 jet whose beam angle is @xmath65 and a wider , less collimated x , optical cone .
these beaming angles agree with the one assumed to explain the required beamed grbs - sn powers . indeed
the electron pair jet may generate a secondary beamed synchrotron radiation component at radio energies , in analogy to the behaviour of bl lac blazars whose hardest tev @xmath0 component is made by inverse compton scattering while its correlated x band emission is due to the synchrotron component . anyway the inner jet is dominated by harder photons while the external cone contains softer @xmath61 , optical and radio waves . a jet angle related by a relativistic kinematics
would imply @xmath66 , where @xmath67 is found to reach @xmath68 ( fargion 1999 ) . at first approximation the gamma constrains is given by inverse compton relation : @xmath69 for @xmath70 and @xmath71 leading to characteristic x-@xmath0 grb spectra .
the origin of @xmath72 electron pairs are very probably decayed secondary related to primary inner muon pairs jets , able to cross dense stellar target ( fargion 2003 ) .
the same gevs electron pair jet may generate a secondary beamed synchrotron radiation component at radio energies , in analogy to the behaviour of bl lac blazars whose hardest tev @xmath0 component is made by inverse compton scattering while its correlated radio bumps is due to the synchrotron component . anyway the inner jet is dominated by harder photons while the external cone contains softer @xmath61 , optical and radio waves .
the very peculiar oscillating grb@xmath73 optical variability did show both a re - brightening and remarkable multi - bump variability in radio wave - lenght .
for this reason and we believe that this fluctuations were indeed to be related to the jet precession and not to any interstellar scintillation .
there is not any direct correlation between the @xmath0 jet made up by i.c .
scattering and the radio jet because the latter is dominated by the external magnetic field energy density : there maybe a different beaming opening and a consequent different time modulation respect to the inner @xmath0 jet .
however the present wide energy power emission between sn2002ap and grb@xmath4 radio light curves makes probable a beaming angle comparable : @xmath74 radiant .
these grbs jets are originated by nss or bh in binary system or disk powered by infall matter ; their relics ( or they progenitors ) are nearly steady x - ray pulsars whose fast blazing is source of sgrs .
this external @xmath0 jet has a chain of progenitor identities : it is very probably originated by a very collimated inner primary muon jet pairs at tevs - pevs energies .
these muons could cross with negligible absorption the dense target lights along the sn explosions , nearly transparent to photon - photon opacities . the high energy relativistic muons ( tens tevs - pevs energies ) decay in flight in electron pairs where the baryon density is still negligible
; these muons are source , by decay in flight to tevs - gev electron pair showering whose final inverse compton scattering with nearby thermal photon is the final primary of the observed hard @xmath61 - @xmath0 jet .
the cost of this long chain of reactions is a poor energy conversion , but the benefit is the possibility to explain the @xmath0 escape from a very dense explosive and polluted ( by matter and radiation ) narrow volume . its inner jet ruled by relativistic inverse compton scattering , has the hardest and rarest beamed gevs - mevs photons ( as the rare and long @xmath75 s life egret grb@xmath76 one ) but its external jet cones are dressed by softer and softer photons .
the complex variability of grbs and sgrs are simulated successfully by the equations above ( fargion & salis 1995,1996,1998 , fargion 1999 ) ; the consequent geometrical beamed jet blazing may lead also to the observed widest morphology @xmath77 signatures and rapid re - brightening as x - ray precursors .
the mystery therefore is not longer in an apparent huge grb luminosity , but in an extreme beam jet collimation and precession .
coburn w. , boggs s. e. nature 423 ( 2003 ) 415 - 417 blackman , e : g. , yi , i. , field g. b. : 1996 , _
apj_,*479 * , l79-l82 j.s.bloom et all .
fargion , d. , salis , a.:1995a,_nuclear phys b ( proc.suppl.)_,*43*,269-273 .
fargion , d. , salis , a.:1995b,_astrophysics & space science_,*231*,191 - 194 .
fargion , d.,et all .
: 1996a , _
3rd huntsville grbaip.conf._ : * 384*,749 - 753 .
fargion , d. , salis , a. : 1998,physics - uspekhi,*41(8)*,823 - 829 .
fargion , d.1999c,_aa & ss_,*138*,507 ; see astro - ph/9903433 ; p. hoeflich , j.c .
wheeler , l. wang . : 1999,apj.521,179 .
kulkarni , s. r. , et al .
, 1998,_nature_,*393*,35 . | a clear gamma polarization in the @xmath0 signals from grb@xmath1 probes the presence of a very thin collimated jet ( opening angle @xmath2 ; @xmath3 ) in gamma ray burst , grbs . the last and well proved grb@xmath4/sn@xmath5dh time and
space coincidence confirms definitively the earliest grb@xmath6/sn@xmath7bw connection among grb and supernova .
the apparent extreme grbs luminosity is just the extreme beamed blazing gamma jet observed in axis during a supernova birth .
the maximal isotropic sn power , @xmath8@xmath9 erg @xmath10 , because of its probable energy equi - partition , should be collimated even into an even thinner jet @xmath11 , in order to explain at once the apparent observed maximal grbs output , @xmath12 @xmath8@xmath13 @xmath14 erg @xmath10 .
consequently one - shoot thin jet grbs needs many more @xmath15 @xmath16 event rate than any spread isotropic fireballs .
such a rate exceed even the known supernova one . to overcome the puzzle a persistent precessing decaying jet ( life - time @xmath17 @xmath18 @xmath19 ) is compelling .
the late relic grbs sources may be found in compact snrs core , as ns or bh jets ; at much later epoch , their lower power @xmath0 jets may be within detectability only from nearby galactic distances , as soft gamma repeaters ( sgrs ) or anomalous x - ray pulsars , axps .
this common jet nature explain some connection between grbs and sgrs spectra .
the geometrical multi - precessing beaming and their reappearance to the observer may explain the puzzling presence of x - ray precursors in grbs as well as in sgrs . a mysterious re - brightening afterglows observed in early and late grb @xmath4 optical transient , like in the @xmath20 august @xmath7 and @xmath21 april @xmath22 sgr @xmath23 events , might be understood as the damped oscillatory imprint left by such a multi - precessing @xmath0-x - optical and radio jet .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
elastic objects pinned by a quenched random potential are a relevant model for many experimental systems .
it describes interfaces in magnets @xcite which experience either short - range disorder ( random bond ) , or long range ( random field ) disorder , the contact line of a liquid wetting a rough substrate @xcite , vortex lines in superconductors @xcite .
it also provides powerful analogies , via mode coupling theory , to complex systems such as structural glasses @xcite .
one important observable is the roughness exponent @xmath3 of the pinned manifold . from the theoretical side ,
this problem still offers considerable challenges .
it is the simplest example of a class of disordered systems , including random field magnets , where the so called dimensional reduction @xcite renders conventional perturbation theory trivial and useless at zero temperature .
the elastic object is usually parameterized by a @xmath1 component vector @xmath4 in the embedding space @xmath5 , and @xmath6 is the coordinate in the internal space .
apart from the case of the directed polymer ( dp ) in @xmath7 dimensions ( @xmath8 , @xmath9 ) , where some exact results were obtained @xcite , analytical results are scarce . one important challenge is to understand the dp for any @xmath1 , due to its exact relation to the kardar - parisi - zhang growth equation whose upper critical dimension is at present not known , and even its very existence is debated @xcite .
two main analytical approaches have been devised so far .
each succeeds in evading dimensional reduction , providing an interesting physical picture , but comes with its limitations .
the first one is the mean field theory , the replica gaussian variational method ( gvm ) @xcite in the statics and the off equilibrium dynamical version @xcite .
the gvm approximates the replica measure by a replica symmetry broken ( rsb ) gaussian , equivalently , the gibbs measure for @xmath10 as a random superposition of gaussians @xcite , and is argued to be exact for @xmath11 .
it yields flory values for the exponent @xmath3 . as for spin glasses , computing the next order corrections ( i.e. in @xmath12 ) at the rsb saddle point
is very arduous @xcite .
one may question whether it is the most promising route , since it is as yet unclear whether the huge degeneracy of states encoded in the parisi rsb is relevant to describe finite @xmath1 .
there seems to be some agreement that this type of rsb does not occur for low @xmath0 and @xmath1 .
certainly , in the simpler but still - non trivial @xmath13 limit , parisi type rsb found in the gvm should exist only at @xmath11 , apart from the interesting so - called marginal case of logarithmic correlations @xcite .
for the dp , another exactly solvable mean field limit is the cayley tree and there too it is not clear how to meaningfully expand around that limit @xcite .
the second main analytical method is the functional renormalization group ( frg ) which performs a dimensional expansion around @xmath14 and was originally developped only to one loop , within a wilson scheme @xcite .
its aim is to include fluctuations , neglected in the mean - field approaches .
there too , the dynamics @xcite has been investigated .
the frg follows the second cumulant of the random potential @xmath15 under coarse graining , a full function since the field is dimensionless in @xmath14 .
it was found that @xmath15 becomes non - analytic already in the 1-loop equation at @xmath16 after a finite renormalization , at the larkin scale .
both methods circumvent dimensional reduction by providing a mechanism which is non - perturbative in the bare disorder .
the gvm evades dr thanks to the rsb saddle point .
the frg escapes via the generation of a cusp - like non - analyticity in @xmath17 at @xmath18 .
indeed , while the bare disorder correlator is an analytic function , frg fixed points for the renormalized @xmath15 , perturbative in @xmath2 , are found only in the space of non - analytic functions , and subject to the condition that the resulting exponent @xmath3 is non - trivial . both methods are disconcertingly different in spirit and it is an outstanding question in the theory of disordered systems how to compare and reconcile them .
comparisons were made between some predictions of the 1-loop frg and of the gvm @xcite .
balents and fisher obtained the 1-loop frg equation for any @xmath1 restricted to @xmath19 , and found that its solution reproduces the flory value of @xmath3 for lr disorder , but yields subtle corrections for sr disorder , exponential in @xmath1 .
physically both methods capture the metastable states beyond the larkin scale @xmath20 and it is tempting to compare how they describe them . in @xcite
a coarse grained random potential was defined and it was found within the gvm that its correlator mimics the one in the frg , exhibiting some non - analyticity which was interpreted in terms of shock - like singularities in the coarse grained disorder . unfortunately , this analogy was demonstrated only around the larkin scale , while a quantitative and more general connection able to reach perturbatively the true large scale behaviour , as is achieved in the field theoretic frg , is still missing .
the need for a study of the frg at large @xmath1 is all the more pressing since we have developped systematic higher loop approaches within the @xmath21-expansion @xcite . within these studies
, we have found that higher loop frg equations for @xmath15 at @xmath22 contain non - trivial , potentially ambiguous `` anomalous terms '' involving the non - analytic structure of @xmath15 at @xmath23 .
we have proposed a solution to lift these ambiguities in the statics at two loops @xcite .
since the large-@xmath1 limit allows in principle to handle higher - loop corrections ( i.e. to treat any @xmath21 ) it should be useful to understand the many - loop structure of the field theory .
stated differently , we want to understand which physical quantity precisely does the frg computes ? finally , developping a systematic @xmath12 expansion within the frg for any @xmath0 should provide a novel handle to attack problems such as kpz , maybe avoiding the need for spontaneous rsb altogether if it proves to be non - essential .
the aim of this paper is to study the frg at large @xmath1 . for this purpose
we first perform an exact calculation of the effective action @xmath24 $ ] of the replicated field theory at large @xmath1 .
its value for a uniform mode and further expansion in cumulants yields a definition of the renormalized disorder consistent with field theoretic approaches .
the second disorder cumulant is found to obey a closed self - consistent equation .
all higher cumulants can be constructed recursively from the lower ones .
it can be easily inverted below the larkin scale and there the solution is analytic and corresponds to the replica symmetric solution of mp .
varying with respect to an infrared scale , here the mass , we obtain the frg @xmath25-function in any @xmath0 at dominant order , @xmath11 .
the continuation beyond the larkin scale is remarkably easier to perform on the resulting frg equation .
its solution reveals that the frg _ exactly _ reproduces the non - trivial result of the gvm with full rsb for small overlap .
we also give a formula which yields the complete rsb solution for all overlaps . at no point in our derivation parisi - rsb
is invoked , as replica symmetry is broken explicitly here . since our saddle point equations for the effective action contain both the mp equations and the frg , it can be used to describe the crossover from frg to rsb .
a qualitative analysis of this crossover is given , as well as a comparison with previous attempts to relate the frg to the gvm @xcite .
finally , applications to other problems and new perspectives are discussed .
a short version of this work has appeared in @xcite . in a related paper @xcite ,
we give all details of the calculation of the @xmath26 corrections , with the aim of understanding finite but large @xmath1 .
the outline of the paper is as follows . in section [ sec : model and program ]
we define the model , the effective action and its physical interpretation . in section [ sec : calculgamma ] we compute the effective action at large @xmath1 , using the saddle point method and perform a cumulant expansion ( section [ sec : self - consistent equation for the renormalized disorder ] ) .
a graphical interpretation is given in section [ s : graphinterprete ] . in section [
sec : functional renormalization group equations ] we establish the frg equation at large @xmath1 ( the @xmath25-function of the theory ) .
then in section [ detailed ] we perfom a detailed analysis of the frg equation for a specific class of disorder correlators , both below and above the larkin scale . in section [ sec : comparisonrsbfrg ] we compare the frg with the mp solution using rsb .
first we recall the mp approach and find agreement with the predictions of the frg calculation .
next we extend these results to an arbitrary disorder correlator for which the gvm gives full rsb .
finally we discuss the physical interpretation and compare our approach with the one of ref .
section [ sec : discussion and conclusion ] presents the conclusion .
the appendices contain several generalizations , the calculation of the third and fourth disorder cumulant , finite temperature fixed points , and an analysis and comparison with the effective action in more conventional field theories .
we consider the general model for an elastic manifold of internal dimension @xmath0 embedded in a space of dimension @xmath1 .
the position of the manifold in the embedding space is described by a single valued displacement field @xmath27 , where @xmath28 belongs to the internal space and @xmath10 is a @xmath1 component vector which belongs to the embedding space .
( its components @xmath29 , @xmath30 , are specified below only when strictly necessary . )
a well studied example is that of an interface ( e.g. a domain wall in a magnet ) where @xmath31 and @xmath9 .
there @xmath27 denotes the height of the interface .
other examples are the directed polymer ( @xmath8 ) in a @xmath1 dimensional space , which can be mapped to the @xmath1-dimensional burgers and kardar - parisi- zhang ( kpz ) equations @xcite , or a vortex lattice in the absence of dislocations described by @xmath32 , @xmath33 , where @xmath27 is there the deformation from the ideal crystal @xcite .
we will study here the equilibrium statistical mechanics of such an elastic manifold in presence of quenched disorder , modeled by a random potential @xmath34 .
it is described , in a given realization of the random potential , by the partition function @xmath35\ , { { \mathrm{e}}}^{- { \cal h}_v[u]/t } \ , \ ] ] where @xmath36}{t } = \int_q ~ \frac{1}{2 } ~ c(q)^{-1 } ~ u(-q ) \cdot u(q ) + \frac{1}{t } \int_x v(x , u(x))\ ] ] consists out of an elastic energy ( expressed here in fourier space and taken to be isotropic ) , and of a pinning energy due to disorder . here and below we denote @xmath37 and @xmath38 . throughout , square brackets as e.g. in @xmath39 $ ]
denote a functional , here @xmath40 of the field @xmath41 , while parenthesis as in @xmath42 denote functions .
a convenient form for the inverse bare propagator , used below , is : @xmath43 where @xmath44 is the temperature and the elastic constant is set to unity by a choice of units .
the role of the additional mass term @xmath45 will be discussed below . an additional small scale ( ultraviolet , uv ) cutoff @xmath46 is implied here and will be made explicit when needed .
this model is highly non - trivial and , apart from the cases of @xmath9 and @xmath47 , very few exact results are known @xcite . to obtain exact results for large embedding space @xmath48 , we need to consider a fully isotropic version of the model with @xmath49 symmetry such that the model remains non - trivial in that limit . as in standard large-@xmath1 treatment ( as for instance of the @xmath50 @xmath49 model )
one defines the rescaled field @xmath51 we will freely switch from one to the other in the following .
one also chooses the distribution of the random potential to be @xmath49 rotationally invariant .
it can be parameterized by its set of connected cumulants , of the form @xmath52 this adequately models the case of uncorrelated ( or short - range correlated ) disorder in the internal space , studied here .
the second cumulant , which plays the central role , is thus defined in terms of a function @xmath53 .
the higher cumulants are not strictly necessary in the bare model , but they appear , as we will see , under coarse graining .
the distribution of disorder being translationally invariant , these functions satisfy @xmath54 for any @xmath55 .
the model studied here is thus a slight generalization of the model studied by mezard and parisi @xcite , henceforth also referred to as mp , in the same limit .
although we will consider the general case , it is useful , as in mp @xcite to define two sets of simple models for which more specific results will be given .
these are , respectively , the gaussian , short - range ( sr ) disorder , correlator @xmath56 and the power - law correlations @xmath57 which , for infinite @xmath1 always corresponds to long - range ( lr ) disorder , a different universality class , as we will see below . for finite @xmath1 , the long - range disorder corresponds , at the bare level , to @xmath58 ; but this is modified at the renormalized level , and the true frontier lr - sr for finite @xmath1 is non - trivial . having defined the model , and before turning to calculations , let us first outline what we aim at .
all the considerations in the present section are valid for any @xmath1 , but , since in the next section we will consider the large @xmath1 limit explicitly , we already make apparent the rescalings . the model defined above has already been studied in mp @xcite .
one of the aims of this study was to compute the roughness exponent of the manifold , defined from the 2-point function as @xmath59 besides the roughness exponent @xmath3 , the amplitude @xmath40 is also of interest whenever it is universal , as it is the case e.g. for long range disorder . to this
aim the model was replicated ( @xmath60 ) , averaged over disorder and self - consistent saddle point equations where derived for the 2-point function @xmath61 this can always be done in a large-@xmath1 limit , and is then solved via a rsb ansatz .
our goal is in a sense broader .
we want to understand the full structure of the field theory , i.e. all correlation functions and not only the 2-point one .
we will thus instead study the generating function of correlations as well as the effective action functional which yields the renormalized vertices .
this program , defined here , will be carried out in the following sections explicitly for large @xmath1 . in this article
we will restrict ourselves to dominant order , but the aim is to understand large but finite @xmath1 , including calculating of @xmath12 corrections .
this is deferred to @xcite .
all physical observables for any @xmath1 can be obtained from the replicated action in presence of a source , i.e. an external force @xmath62 acting on each replica : @xmath63 = \overline { \int \prod_a { \cal d } [ u_a ] e^{- \sum_a h_v[u_a]/t + \int_x \sum_a j_a(x)\cdot u_{a}(x ) } } \ , \ ] ] where @xmath41 , @xmath64 are the replicated fields ( each one being an @xmath1 component vector @xmath65 ) .
differentiating with respect to the replicated source @xmath62 in the limit @xmath66 yields all correlation functions .
the finite-@xmath67 information is also interesting .
for instance from @xmath68 = \overline{\exp(- n { \cal f}_v / t)}\ ] ] one can retrieve the sample to sample distribution of the free energy @xmath69 , as was done e.g. in a finite size system for @xmath8 @xcite . thus , unless specified we will keep @xmath67 arbitrary .
one can explicitly perform the disorder average in ( [ repzj ] ) : @xmath70&= & \int \prod_a { \cal d } [ u_a ] { { \mathrm{e}}}^ { - n { \cal s}[u , j ] } \\ { \cal s}[u , j ] & = & \frac{1}{2 } \int_q c(q)^{-1 } v_a(-q ) \cdot v_a(q ) \nonumber \\ & & + \int_x [ u(\chi(x ) ) - j_a(x ) \cdot v_a(x ) ] \qquad \label{lf9}\ , \end{aligned}\ ] ] where here @xmath71 and here and below summations over repeated replica indices are implicit .
we have rescaled the source in a manner complementary to the field : @xmath72 we have also introduced the bare interaction potential @xmath73 which is a function of a @xmath67 by @xmath67 replica matrix @xmath74 and has a cumulant expansion in terms of sums with higher numbers of replicas . because of translational symmetry and @xmath49 invariance it depends only on the matrix @xmath75 and the form of each cumulant is restricted .
for instance one has @xmath76 etc .. the matrix potential @xmath77 can thus be considered as a convenient way to parameterize the disorder ( here the bare disorder ) .
the physical object which contains the information about the field theory at large scale is the effective action .
it is the generating function of the 1-particle irreducible diagrams and in conventional field theories its formal expansion in powers of the field yields the renormalized vertices . all correlation functions are then obtained simply as tree diagrams from these renormalized vertices .
in particular it is known that within a @xmath78 expansion at zero temperature to at least 2-loop order the theory can be renormalized ( i.e. rendered uv finite and yielding universal results ) by considering counter - terms only to the second cumulant .
the latter is a function @xmath15 , and can be viewed as the set of all coupling constants which simultaneously become marginal in @xmath14 . to probe renormalizability to any number of loops ,
we want to compute the effective action from first principles .
the effective action functional is defined as a legendre transform : @xmath79 + { \cal w}[j ] & = & \int_x \sum_a j_a(x ) \cdot u_a(x ) \\ { \cal w } [ j ] & = & \ln { \cal z}[j ] \ .\end{aligned}\ ] ] strictly speaking the definition is the convex envelope @xmath24=\min_j ( \int_x \sum_a j_a(x ) \cdot u_a(x ) - { \cal w}[j])$ ] . here
we apply the definition to the replicated action , and will content ourselves with the differential definition @xmath80}{\delta u_a(x ) } & = & j_a(x ) \\
\frac{\delta { \cal w}[j]}{\delta j_a(x ) } & = & u_a(x ) \ , \end{aligned}\ ] ] which relates a pair of values @xmath81 , later also denoted by @xmath82,u)$ ] . since @xmath24 $ ] defines the renormalized vertices , its zero momentum limit defines the _
renormalized disorder_. thus in order to compute the renormalized disorder , we only need to compute @xmath24 $ ] ( per unit volume ) for a _ uniform _ configuration of the replica field @xmath83 ( a so - called fixed background configuration ) . because of the statistical tilt symmetry @xcite , i.e. invariance of disorder term in the replicated action ( [ lf9 ] ) under the translation @xmath84 , and of the @xmath49 invariance one can argue , and this is what we find below , that for the model ( [ choice ] ) the scaled effective action per unit volume ( which for a uniform mode is simply a function of @xmath85 ) should have the following form @xmath86 where @xmath87 is the volume of the system , and here and below we use the notation : @xmath88 for the @xmath67 by @xmath67 replica matrix .
this defines the renormalized disorder .
furthermore , whenever @xmath89 can be expanded , up to a constant , in the form : @xmath90 where here and in the following we denote @xmath91 then ( [ cumulants ] ) defines the _ renormalized cumulant _ functions @xmath92 etc .. as we will see below this is correct up to some very subtle behavior at coinciding replica vectors ( i.e.@xmath93 for some pair @xmath94 ) .
also note that the constant part @xmath95 is the free energy .
the main result of the following sections will be the exact calculation of the uniform part of the effective action , i.e. of the function @xmath89 .
this will be performed within a large @xmath1 expansion : @xmath96 and here we will obtain the dominant order @xmath97 ; the corrections @xmath98 are calculated in @xcite . it will be a function of a scale parameter . we choose to add a mass - term @xmath45 which provides such a scale .
it is a convenient choice since for @xmath99 one has @xmath100 : fluctuations are totally suppressed and the effective action equals the action .
one can then progressively lower the mass down to zero , starting from this initial condition , since ultimately one is interested in the massless limit .
another choice is to change the uv cutoff , as will be discussed again below .
it is now useful to give a more direct physical interpretation of this quantity , in addition to the above field theoretic interpretation . the effective action for a uniform background is also known to be related to the distribution of the order parameter .
let us recall the relation for a simple pure ferromagnet .
the unnormalized probability distribution of the order parameter @xmath101 where @xmath102 is the local magnetization is by definition @xmath103\ , \delta\left(\phi - \frac{1}{l^d } \int_x \phi(x)\right ) { { \mathrm{e}}}^ { - s[\phi ] } \ , \ ] ] where @xmath104 $ ] is the action which describes the ferromagnet ( e.g. a @xmath50 theory or a landau ginsburg model ) .
the functional @xmath105 evaluated for a uniform @xmath106 reads : @xmath107 in the large - volume limit , the saddle point can be taken and since the legendre transform is involutive , this yields the relation between the effective action at @xmath108 per unit volume and the probability distribution of the order parameter as : @xmath109 in the thermodynamic limit the effective action per unit volume can very well be a non - analytic function .
this is the case e.g. in the ferromagnetic phase where its left and right second derivatives at @xmath110 do not coincide ( @xmath111 is the spontaneous magnetization per unit volume ) .
while the right derivative at @xmath110 is related to the inverse susceptibility , the left one is zero , mathematically due to the prescription to take the convex envelope , and physically because one can always lower the magnetization at no cost in free energy per unit volume by introducing a domain wall .
the above property ( [ lf17 ] ) can be extended to a given @xmath112 mode .
finally , note that in @xmath13 the above does not hold since there is no large factor @xmath113 , and the probability distribution is directly given by the action @xmath114 .
what is then the physical meaning of the quantity that we will be computing in the next sections ? let us in analogy to the magnetization for a ferromagnet define the center of mass of an interface : @xmath115 since we have added a mass in the elastic energy ( [ choice ] ) , which acts as an extra quadratic well bounding the fluctuations of the interface , the disorder - induced fluctuations of the center of mass are always finite .
one expects that they diverge typically as @xmath116 as @xmath117 , thus their behavior as a function of @xmath45 is of high interest and yields e.g. the information about the roughness exponent .
one can then define the probability distribution @xmath118 of the center of mass of the interface in a given realization of the random potential @xmath119 ( and in presence of the quadratic well induced by the mass ) .
one can see that by definition the generating function for a uniform @xmath120 is the laplace transform of the probability distribution of @xmath121 , namely @xmath122 then by the same saddle point argument as for the ferromagnet one expects , at least naively , that @xmath123 = - \lim_{l \to \infty } \frac{1}{n l^d } \ln \overline{p_v(w_1 ) \dots p_v(w_n ) } \ .\end{aligned}\ ] ] symbolically one can write : @xmath124 } \ , \end{aligned}\ ] ] provided this is taken with a grain of salt .
thus one can also think of the renormalized disorder @xmath125 as parameterizing the set of correlations of an effective equivalent toy model ( @xmath13 ) which has the same set of correlations as the center of mass variable in the original model .
the @xmath126-th connected moment of the center of mass is identical , up to a volume factor , to the zero momentum limit of the connected m point correlator of the @xmath10 field , e.g. @xmath127 and , once the effective action is known , both can thus be obtained in principle as the sum of all tree graphs made from @xmath128 $ ] vertices .
for instance the 2-point function should be obtainable from : @xmath129^{-1}_{ab}\ ] ] and the connected 4-point function from : @xmath130^{-1}_{ae } [ \hat{\gamma}''[v=0]]^{-1}_{bf } \nonumber \\ & & \times [ \hat{\gamma}''[v=0]]^{-1}_{cg } [ \hat{\gamma}''[v=0]]^{-1}_{dh } \nonumber \\ & & \times \hat{\gamma}''''[v=0]_{efgh}\end{aligned}\ ] ] this however assumes analyticity , which as we will see below , does not always hold .
another integral relation holds @xmath131 } \ .\end{aligned}\ ] ]
let us now consider explicitly the large @xmath1 limit .
one can rewrite for any @xmath1 the starting generating function ( [ lf8]-[lf9 ] ) as : @xmath132 = \int { \cal d } [ u ] { \cal d } [ \chi ] { \cal d}[\lambda ] { { \mathrm{e}}}^ { - n { \cal s}[u,\chi,\lambda , j ] } \\ & { \cal s}[u,\chi,\lambda , j ] = \frac{1}{2}\ ! \int_q c(q)^{-1 } v_a(-q ) \cdot v_a(q ) -\ ! \int_x j_{a}(x ) \cdot v_{a}(x ) \qquad \nonumber \\ & + \int_x u(\chi(x ) ) - \frac{1}{2 } i \lambda_{ab}(x ) \left[\chi_{ab}(x ) - v_{a}(x ) \cdot v_{b}(x)\right ] \
, \hspace{-1 cm}\end{aligned}\ ] ] where the replica matrix field @xmath133 has been introduced through a lagrange multiplier matrix @xmath134 . here and below summations over repeated replica indices are implicit .
one can then explicitly perform the functional integration over the field @xmath27 and obtain : @xmath135 & = & \int d \chi d\lambda e^ { - n s[\chi,\lambda , j ] } \\ s[\chi,\lambda ,
j ] & = & \frac{1}{2 } { { \mathrm{\bf tr}}}\ln ( c^{-1 } + i \lambda ) \nonumber \\ & & + \int_x u(\chi(x ) ) - \frac{i}{2 }
\lambda^{ab}(x ) \chi^{ab}(x ) \nonumber \\ & & - \frac{1}{2 } \int_{x , x ' } j_a(x ) ( c^{-1 } + i \lambda)^{-1}_{a x , b x ' } j_b(x ' ) \ , \qquad \end{aligned}\ ] ] where the inversion and trace are performed in both replica space and spatial coordinate space .
it has now the standard form for a saddle point evaluation of the functional @xmath136 except that the saddle point is not , in general , uniform in space .
it is useful to define the scaled functional @xmath137 through @xmath138 = n \tilde{w}[j = j/\sqrt{n } ] \ , \ ] ] which has a well defined large-@xmath1 limit and can be expanded in @xmath12 as : @xmath139 = w^0[j ] + \frac{1}{n } w^1[j ] + \dots \ .\ ] ] deferring the calculation of the corrections to a future publication @xcite , we obtain here the dominant order in @xmath12 as : @xmath140 = - s[\chi_{j},\lambda_{j},j ] \ , \ ] ] where @xmath141 and @xmath142 depend on @xmath143 and are the solutions of the saddle point equations obtained respectively by setting to zero the functional derivatives ( at fixed @xmath143 ) : @xmath144 } { \delta \lambda_{ab}(x)}\bigg| _ { \chi=\chi_j , \lambda = \lambda_{j } } & = & 0 \\
\frac{\delta s[\chi,\lambda , j ] } { \delta \chi_{ab}(x)}\bigg |_{\chi=\chi_j , \lambda = \lambda_{j } } & = & 0 \label{spgenw } \ .\end{aligned}\ ] ] the result is @xmath145 where @xmath146 is a matrix with both replica indices and spatial coordinates and inversion is carried out for both . here and below ,
replica indices are raised whenever explicit dependency is given , e.g. @xmath147 .
the notation for the @xmath1-component vector @xmath148 is a shorthand for a matrix product , and everywhere we denote by @xmath149 the simple derivative of the function @xmath150 with respect to its matrix argument @xmath151 . of course
, if , for a given @xmath143 there are several solutions to these equations , then one must sum over all saddle points , to the same order @xmath152 \approx \ln \big ( \sum_{\mathrm{sp } } { { \mathrm{e}}}^ { - n s[\chi_{\mathrm{sp}}(j),\lambda_{\mathrm{sp}}(j),j ] } \big)\ .\ ] ] this case will be discussed below , for now we ignore this possible complication , as well as issues of stability of the saddle point .
now we want to take the legendre transform and trade the variable @xmath120 for the variable @xmath55 to obtain the effective action @xmath24 $ ] .
one also defines the scaled functional , and its @xmath12 expansion through : @xmath79 & = & n \tilde \gamma[v = u/\sqrt{n } ] \\
\tilde{\gamma}[v ] & = & \gamma_0[v ] + \frac{1}{n } \gamma_1[v ] + \dots \ .\end{aligned}\ ] ] then ( @xmath153$],@xmath154 $ ] ) and ( @xmath155$],@xmath156 $ ] ) are also two pairs of legendre transforms .
thus the dominant order of the effective action functional in the large-@xmath1 limit is given by @xmath157 = \int_x v_a(x ) \cdot j^a_v(x ) - w_0[j_v ] \label{legendre}\ ] ] with @xmath156 $ ] given by ( [ w0 ] ) , ( [ spw ] ) and where @xmath158 is the @xmath159-dependent source solution of : @xmath160}{\delta j^a_v(x ) } = v_a(x ) \ .\ ] ] one can now derive a self - consistent functional saddle point equation for @xmath161 $ ] . first we establish the relation between @xmath55 and @xmath162 , namely @xmath163 where from now on we define @xmath164 eq . ( [ vj ] ) is obtained noting that @xmath165}{\delta j^a_v(x ) } = - \frac{{{\mathrm{d}}}}{{{\mathrm{d}}}j_a(x ) } s[\chi_j,\lambda_j , j]\bigg|_{j = j_v } \nonumber \\ & = & - \int_y \left [ \frac{\partial \chi_{j}(y ) } { { \partial}j_a(x ) } \frac{{\partial}s}{{\partial}\chi_j(y ) } + \frac{{\partial}\lambda_j(y)}{{\partial}j_a(x ) } \frac{{\partial}s}{{\partial}\lambda_j(y ) } \right]\bigg|_{j = j_v } \nonumber \\ & & - \frac{{\partial}s}{{\partial}j_a(x)}\bigg|_{j = j_v } = - \frac{{\partial}s}{{\partial}j_a(x)}\bigg|_{j = j_v } = ( g_{j_v}:j_v)_{a x } \nonumber \ , \\\end{aligned}\ ] ] where we have used the saddle - point equations ( [ spgenwb ] ) , ( [ spgenw ] ) . we can now use ( [ vj ] ) in the saddle point equations ( [ spgenwb ] ) , ( [ spgenw ] ) and defining @xmath166 this yields a self - consistent equation for @xmath167 @xmath168 which is also a self - consistent equation for @xmath169 .
since the legendre transform is involutive , one can also write : @xmath170}{\delta
v_a(x ) } = j_a(x ) = ( g_v^{-1 } : v)_{ax } \ , \ ] ] which determines the derivative of @xmath161 $ ] once ( [ scg ] ) is solved . one can however do better . using ( [ vj ] ) in ( [ legendre ] ) one obtains the effective action for a spatially varying field @xmath159 : @xmath171 = v:(g_v^{-1}):v + s[\chi_v,\lambda_v , j_v ] \ , \ ] ] which gives @xmath172 & = & \frac{1}{2}\nonumber \int_{xy } c^{-1}_{a x , b y } v_a(x ) v_b(y ) \\ & & + \frac{1}{2 } { { \mathrm{\bf tr}}}\ln ( c^{-1 } + 2 \partial u(\chi_v ) ) + \int_x u(\chi_v(x ) ) \nonumber \\ & & { + } \ !
\int_x\ !
v_a(x ) \partial_{ab } u(\chi_v(x ) ) v_b(x ) { - } \chi^{ab}_v(x ) \partial_{ab } u(\chi_v(x ) ) \ . \nonumber
\\ & & \end{aligned}\ ] ] it is interesting to rewrite it with the help of ( [ scg ] ) as a functional of @xmath169 and @xmath55 only : @xmath173 & : = & \gamma_0[v ] = - \frac{1}{2 } { { \mathrm{\bf
tr}}}\ln g_v \nonumber \\ & & + \frac{1}{2 } \int_{xy } c^{-1}_{a x , b y } \left[v_a(x ) v_b(y ) + ( g_v)_{a x , b y } \right ] \nonumber \\ & & + \int_x u(vv(x ) + ( g_v)_{x , x } ) \label{var}\ .\end{aligned}\ ] ] we have dropped a constant @xmath174 .
( [ var ] ) has the property @xmath175 |_{g = g_v } = 0 \label{saddleg } \ , \ ] ] where the derivative acts only on @xmath176 , leaving fixed all @xmath55 , since this coincides with the saddle point equation ( [ scg ] ) .
this makes apparent that it can also be obtained from a _ variational method _ where the average of the field is fixed , as we detail in appendix [ sec : var ] .
since the explicit non - trivial @xmath55-dependence in ( [ var ] ) using ( [ scg ] ) is purely in terms of the bilinears @xmath177 at the same space points , it also shows that one can write : @xmath178 = \frac{1}{2}\ , v : c^{-1}:v + \tilde u_0[v\cdot v ] \ , \ ] ] where the interaction ( i.e. disorder ) part satisfies @xmath179}{\delta ( v_a(x ) \cdot v_b(y ) ) } = 0\ , \qquad x \neq y\ ] ] and is the solution of a self - consistent functional equation : @xmath180}{\delta ( v_a(x ) \cdot v_b(x ) ) } = \partial_{ab } u(vv(x ) + ( g_v)_{x , x } ) \label{scfunctional } \\ & & \!\!\ ! ( g_v^{-1})_{ax , by } = ( c^{-1})_{x , y } \delta_{ab } + \frac{\delta \tilde u_0[v\cdot v]}{\delta ( v_a(x ) \cdot v_b(x ) ) } \ , \delta^{d}(x - y ) \nonumber \ .\end{aligned}\ ] ] a generalization of this equation is presented in appendix [ sec : appb ] .
let us now consider the simpler problem of computing the effective action for a uniform field configuration , which can be solved self - consistently .
to be more specific we will focus on the form ( [ choice ] ) for the elastic energy . also , to simplify notations and since we will restrict ourselves to dominant order in @xmath12 , we drop the index @xmath181 , so we set : @xmath182 and so on . for a uniform field
@xmath183 the effective action ( [ form ] ) per unit volume takes the form : @xmath184 note that these are now simply functions ( not functionals ) of a @xmath185-component vector , and @xmath89 is a function of the @xmath67 by @xmath67 matrix @xmath186 .
( [ var ] ) yields also a formula for @xmath187 ( up to a constant ) : @xmath188^{-1 } \big\}\ .\qquad \label{freeenergy}\end{aligned}\ ] ] the trace acts in replica space , and the log is a function of a matrix , to be defined as usual . since ( [ freeenergy ] ) contains the derivative @xmath189 we must first determine the latter .
one finds that analogous to ( [ scfunctional ] ) @xmath190_{ab}^{-1 } \qquad \label{saddle } \ .\end{aligned}\ ] ] since one can replace the matrix @xmath191 by @xmath192 in the denominator of ( [ saddle ] ) , this is also a self - consistent equation , which involves only @xmath193 . here
inversion is simple @xmath67 by @xmath67 matrix inversion and @xmath194 is the kronecker @xmath67 by @xmath67 identity matrix @xmath195
. one must be careful that @xmath196 is a _ first derivative _ of @xmath197 with respect to the matrix element @xmath198 .
one can also check that taking the derivative of ( [ freeenergy ] ) with respect to @xmath198 correctly reproduces ( [ saddle ] ) .
a direct derivation uses @xmath199 from eq .
( [ saddle ] ) . a more clever way is to remember that
because of ( [ saddleg ] ) , one is allowed to differentiate only with respect to the explicit @xmath200 in @xmath201 in the first term , and that the remaining terms can be written as a function of @xmath202 only , and using again ( [ saddleg ] ) .
this self - consistent equation for @xmath203 , i.e. for the uniform part of the effective action is one of our main results and the remainder of this paper is devoted to analyze it .
it contains a huge amount of information , since it encodes the _ full distribution _
( i.e. all cumulants ) of the renormalized disorder , and is thus quite non - trivial to analyze .
it includes both the gaussian variational method ( gvm ) of mezard - parisi@xcite and the functional renormalization group ( frg ) . for simplicity ,
we now consider the bare disorder to be gaussian and set all bare cumulants except the second cumulant @xmath53 to zero .
the gvm is recovered upon setting @xmath204 which is one limit in which the equation `` simplifies '' .
one sees that ( [ saddle ] ) then reproduces the mezard parisi equations , the self energy @xmath205 and two point function @xmath206 in ref .
@xcite being : @xmath207 in the glass phases , these exhibit spontaneous replica symmetry breaking ( rsb ) with multiple solutions corresponding to saddle points obtained via replica permutations , and the above equations are solved by a hierarchical parisi ansatz for @xmath208 where @xmath209 is the overlap between replicas @xmath210 and @xmath211 .
we will give more details about this correspondence in the following .
for now we will study the opposite limit of `` strong '' explicit symmetry breaking field ( all @xmath212 ) .
then we expect that the renormalized disorder @xmath89 is given by a single saddle point and can be expanded in replica sums in terms of unambiguous renormalized cumulants , i.e. up to a constant @xmath213 this is the limit solved here , which we will show below is the natural limit in the frg , and amounts , as we will discuss , to forcing the manifold in distant states within the rsb picture .
the rich crossover to rsb contained in ( [ saddle ] ) , when some of the @xmath214 are set to zero will be discussed below .
we now transform equation ( [ saddle ] ) for the formal function @xmath215 in a set of equations for the second , third , fourth , a.s.o . cumulants .
this is performed through an expansion in sums over an increasing number of free replica indices , and _ is not an approximation_. the such obtained equations are as exact as ( [ saddle ] ) , i.e. exact to dominant order at large @xmath1 , albeit more explicit .
in fact , they allow a recursive exact calculation of _ all cumulants_. their increasing complexity will illustrate the wealth of information summarized in ( [ saddle ] ) . let us first rewrite ( [ saddle ] ) using an infinite series : @xmath216 where the @xmath67-th power here denotes the matrix product . since we consider a gaussian bare model ( [ u ] ) where only the second cumulant is non - zero one has : @xmath217 \label{chipower1}\ ] ] using that @xmath218 @xcite .
the same quantity for the renormalized disorder reads : @xmath219 \nonumber \\ & & + \cdots \ , \label{power1}\end{aligned}\ ] ] where we denote @xmath220 , @xmath221 and @xmath222 denotes a derivative with respect to the first argument of the function @xmath223 ( @xmath224 has the symmetries implied by replica permutation symmetry ) .
all matrices we will encounter can be parameterized as : @xmath225 where @xmath226 do not contain any explicit kronecker @xmath195 , the upper index denotes the number of free replica sums , e.g. @xmath227 , @xmath228 .
since under matrix product @xmath229 or hadamar product @xmath230 the number of sums can only increase , one gets only a finite number of terms in projecting out on terms with a given number of free replica sums . if we parameterize in the same way
: @xmath231 then one easily sees that : @xmath232 + b'(\tilde{\chi}_{ab } ) \nonumber \\ & & b'(\tilde \chi^{ab}_v )
- \delta_{ab } \sum_c b'(\tilde \chi^{ac}_{v } ) = b'(\tilde{\chi}_{ab } ) - \delta_{ab } \sum_c b'(\tilde{\chi}_{ac } ) \nonumber\ .
\\ & & \label{simpli}\end{aligned}\ ] ] we can now expand in number of sums : @xmath233 and the equivalence of ( [ chipower1 ] ) and ( [ power1 ] ) using ( [ simpli ] ) , implies : @xmath234 and so on for higher cumulants . thus to obtain the second renormalized cumulant we only need to compute the part @xmath235 of @xmath236 which contains zero sum and no explicit @xmath195 .
one has in general : @xmath237 thus for the second cumulant we need only @xmath238 and @xmath239 .
since one has , to be explicit : @xmath240_{ab } & = & \frac{4}{t^2 } \!\left [ \delta_{ab } \sum_{ef } \tilde{b}'_{ae } \tilde{b}'_{af } - \tilde{b}'_{ab } \sum_f ( \tilde{b}'_{af } + \tilde{b}'_{bf } ) + \sum_c \tilde{b}'_{ac } \tilde{b}'_{cb } \right ] \nonumber \\ & & + \frac{4}{t^3 } \!\left [ 2 \delta_{ab } \sum_{egh } \tilde{b}'_{ae } \tilde{s}'_{agh } - \tilde{b}'_{ab } \sum_{gh } ( \tilde{s}'_{bgh } { + } \tilde{s}'_{agh } ) - \sum_{eh } ( \tilde{b}'_{ae } \tilde{s}'_{abh } + \tilde{b}'_{be } \tilde{s}'_{abh } ) + \sum_{h c } ( \tilde{b}'_{ac } \tilde{s}'_{cbh } { + } \tilde{b}'_{bc } \tilde{s}'_{cah})\right]\!\!\nonumber \\ & & + \cdots\end{aligned}\ ] ] where all terms not written have at least three free replica sums ( this is the case for @xmath241 as well as terms involving the fourth cumulant and higher ) .
similarly @xmath242_{ab}$ ] has at least @xmath243 free replica sums ( from the @xmath244 term ) .
this is much more what we need , which comes only from ( [ power1 ] ) and , using ( [ series ] ) : @xmath245 this yields : @xmath246 thus we find that the renormalized second cumulant satisfies a closed equation at any @xmath44 : @xmath247 with no other contributions from higher cumulants at any @xmath44 .
appendix [ scnonloc ] contains a non - local extension of this formula .
( [ saddlepointforb ] ) can be integrated with the result @xmath248\right)\nonumber \\ & & - 2 i_{2 } \left\ { b ' \!\left(v^{2}+2 t i_{1}+ 4 i_{2}[\tilde b ' ( v^{2})-\tilde b ' ( 0)]\right)\right\}^{2}\ .\nonumber \\ & & \end{aligned}\ ] ] a direct derivation from ( [ lf85 ] ) is also possible . ,
the second diagram the terms proportional to @xmath249 in eq . ( [ 3cum]).,title="fig : " ] higher cumulants of the renormalized disorder can be obtained by the same method using ( [ eqcum ] ) and its extensions .
they can also be obtained by the graphical method . for simplicity here we give only the expression of the third cumulant . the complete expression for the fourth cumulant , together with all calculational details and an introduction to the graphical method , can be found in appendix [ sec : highercum ] .
the third cumulant is found to be : @xmath250 \nonumber \\ & & + 24 i_3 \text{sym}_{x , y , z } \big
[ ( \tilde{b}'(x ) - \tilde{b}'(0 ) ) \tilde{b}'(x ) \tilde{b}'(y ) \big ] \nonumber \\ & & - 8 i_3 \tilde{b}'(y ) \tilde{b}'(z ) \tilde{b}'(x ) \label{3cum } \ , \end{aligned}\ ] ] where @xmath251 is @xmath252 times the sum of all permutations of @xmath253 .
note that this relation is exact for all values of the mass @xmath45 , and not just a fixed point form .
the only input in the derivation is the absence of a third cumulant for the bare model ( @xmath99 )
. it would be interesting to include an additional bare third cumulant .
the fourth cumulant is derived in appendix [ sec : highercum ] , where also details for the graphical method are given .
.,title="fig : " ] @xmath254 @xmath255 @xmath256 , a solid line the field @xmath10 , and its correlator .
a dashed line attaches two fields @xmath10 to a vertex @xmath257 .
we do not draw replica - indices.,title="fig : " ] , as explained in the main text.,title="fig : " ] .,title="fig : " ] .
the wiggly line denotes a derivative , and is combinatorially equivalent to choosing one @xmath257 . at finite @xmath44
one can attach an additional arbitrary number of tadpoles to any @xmath257.,title="fig : " ] in this section , we sketch how the central results at large @xmath1 can be obtained graphically , first the saddle - point equation ( [ saddlepointforb ] ) , which gives the effective disorder @xmath258 as a function of the bare disorder @xmath257 , and second the @xmath25-function ( [ betafinal ] ) . the graphical rules for the perturbation theory of the replicated model have been described in detail in @xcite for @xmath9 and we refer the reader to this work for elementary details .
here there are in addition @xmath1 components of the field @xmath259 , the propagator being diagonal in all indices . for the present purpose
we are mostly interested in the counting in @xmath1 , and since it is difficult to represent graphically both vector- and replica - indices , we work with unsplitted vertices ( see @xcite ) and specify the replica content only when needed .
disorder vertices may contain arbitrary number of derivatives and some examples are represented on fig.[fig : vert+tadpole ] .
as usual there is a factor of @xmath12 per derivative ( i.e. per dashed line ) at each vertex ( see e.g. fig.[fig : vert+tadpole ] , using @xmath260 ) , @xmath1 per vertex , and @xmath1 per loop .
we consider the effective action , i.e. the sum of all 1-particle irreducible diagrams ( 1pi ) , and later focus on its 2-replica part .
we start our analysis at @xmath16 with the three possible 1-loop diagrams , as presented on figure [ fig:1loopncomp ] .
they are obtained from contracting @xmath261 in order to simplfy the calculation we omit the terms taken at coinciding replicas ( e.g. @xmath262 ) , they can be added at the end . contracting ( [ iosdu ] )
twice between points @xmath28 and @xmath263 gives @xmath264\nonumber \\ & & \hspace{-1.8 cm } + \,\mbox{higher replica terms}\end{aligned}\ ] ] this is graphically depicted on figure [ fig:1loopncomp ] .
the important observation is that only the first diagram , with a closed @xmath10-loop is contributing in the limit of large @xmath1 .
this analysis can be repeated to higher loop order .
again , only diagrams as the first one on figure [ fig:1loopncomp ] contribute . especially , there are no loops with three propagators or more , as loops 4 or 6 on figure [ fig : badloops ] . also , there are no `` meta - loops '' , i.e. loops formed by loops , as loop 5 on figure [ fig : badloops ]
. finally , only diagrams as those on figure [ fig : tree1 ] survive , which as building block have only the elementary 1-loop diagram with a closed loop contributing a factor of @xmath1 , as the first diagram on figure [ fig:1loopncomp ] .
these are tree - like diagrams , where the nodes are made out of @xmath257 and the links out of the before - mentioned 1-loop diagram ( two parallel replica lines in the splitted diagrammatics @xcite which produce the desired 2-replica term ) . at junction
points the replica lines branch also in parallel .
these are of course not tree diagrams , i.e. they are 1pi and contribute to the effective action .
note on fig .
[ fig : tree1 ] that since there is a @xmath265 factor per vertex , but that each vertex ( except one ) comes with two propagators ( factor @xmath266 ) the counting in temperature is right to produce a 2-replica term with the expected @xmath265 global factor ( the 3-replica terms proportional to @xmath44 have been discarded etc . ) .
we are now in a position to derive the self - consistent equation ( [ saddlepointforb ] ) .
the key - observation is that deriving @xmath267 once with respect to its argument , amounts in the graphical intepreation of figure [ fig : tree1 ] to choose one of the bare vertices @xmath268 , and deriving it .
@xmath269 thus is @xmath270 , with as many branches attached as one wants .
every branch consists out of a 1-loop integral @xmath271{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex}}}}\renewcommand{{0.35}}{{{0.35}}}}=i_{2}$ ] times another tree ; the latter is again @xmath272 , given that one of the bare vertices is chosen , i.e. again @xmath269 .
since attaching loops to @xmath273 amounts to deriving @xmath273 once for every loop , we arrive at @xmath274^{\ell } } { \ell ! }
\nonumber \\ & = & \label{resumtildeb } b'\ ! \left(v^{2 } + 4 i_{2 } [ \tilde b ' ( v^{2})-\tilde b ' ( 0 ) ] \right)\ .\end{aligned}\ ] ] note that we have added the term with coinciding replica - indices , dropped previously .
the combinatorial factor comes from the expansion of the exponential function in @xmath275 .
that it indeed resums to @xmath276 with a shifted argument is natural : for a function @xmath277 taking the expectation value @xmath278 in a theory with only a first moment @xmath279 is equivalent to calculating @xmath280 .
taylor - expanding the latter leads to the above combinatorics . by the same arguments the full effective action can be written as the sum over tree - like ( but not tree ) diagrams represented in fig .
[ fig : tree1 ] where , in addition , each vertex can be dressed by an arbitrary number of tadpoles ( see fig.[fig : vert+tadpole ] ) .
each tadpole brings an additional factor of @xmath44 , thus tadpoles contribute to the two replica term only at @xmath281 . at finite temperature ,
any of the @xmath282 s could be contracted , leading to the relacement @xmath283 ( this offers another possibility to verify the combinatorics in [ lf86 ] . )
thus the final result is @xmath284\right)\ ] ] we now illustrate how to recover the @xmath25-function . applying @xmath285 to @xmath258 implies to derive each integral w.r.t .
@xmath45 appearing in each loop of fig [ fig : tree1 ] .
diagrammatically this amounts to choosing in the tree of fig [ fig : tree1 ] one of the bonds ( loop @xmath286 ) which connects two @xmath257 s . summing over all trees , it gives a term @xmath287\ ] ] since the two trees attached to the loop @xmath286 are nothing but @xmath288 , derived once , and again itself with things attached , i.e. @xmath269 as given in ( [ tildebfinal ] ) .
this reproduces the @xmath16 term in ( [ betafinal ] ) .
the second contribution comes from deriving @xmath289 .
the graphical derivation is complicated , and we refer the interested reader to @xcite where a more complete , but much more involved , diagrammatic method is presented .
we will now study the self - consistent equation , exact for @xmath11 , for the second cumulant correlator of the random potential that we have derived in the previous section : @xmath290 which involves only the two one loop integrals : @xmath291 where we have indicated symbolically that a short scale uv cutoff is needed for @xmath292 to be finite if @xmath293 and for @xmath294 for @xmath295 .
there is a simple way to obtain directly the solutions of ( [ starting ] ) which we will detail below .
it is also interesting to turn this equation into a frg equation for the function @xmath296 as a function of the scale parameter @xmath45 .
indeed this yields the @xmath25-function of the field theory in the limit of infinite @xmath1 , which is our main goal .
let us show first how one does it .
let us first take a derivative of ( [ starting ] ) with respect to @xmath28 .
one obtains : @xmath297 taking the derivative @xmath298 of ( [ starting ] ) and using ( [ secder ] ) gives : @xmath299 \ .\qquad
\label{5.5}\end{aligned}\ ] ] regrouping the terms one obtains : @xmath300 \ .
\qquad \label{5.7}\end{aligned}\ ] ] from ( [ 5.5 ] ) one also has @xmath301 inserting ( [ lf37 ] ) into ( [ 5.7 ] ) finally yields @xmath302 \label{frg1 } \ .\qquad\end{aligned}\ ] ] this equation is valid for any space dimension @xmath0 .
it can be integrated once w.r.t .
@xmath28 to obtain the final result @xmath303\ , \qquad\end{aligned}\ ] ] where we have dropped an @xmath45-dependent integration constant .
a general method to study ( and solve ) the frg equation ( [ frg1 ] ) is then to start from @xmath304 where the initial condition is @xmath305 in the presence of a uv momentum cutoff @xmath46 , or a lattice with lattice constant @xmath306 .
then one studies how @xmath296 evolves as @xmath45 is slowly decreased .
there are thus two possible paths to solve the problem , namely the direct inversion of the self - consistent equation and the solution of ( [ frg1 ] ) with the above initial condition .
both are studied below .
these two methods are clearly equivalent when the solution @xmath296 is analytic at @xmath307 .
indeed , in the above derivation , we have assumed that @xmath308 exists .
this will not always hold , as we now discuss .
what the proper ensuing modifications are is a subtle point which will be examined later . before solving this equation
let us first find the conditions under which there exists an analytic solution .
this will give us insight in the phases of the model .
one notes from ( [ secder ] ) that : @xmath309 for @xmath99 the starting value is @xmath310 , in any dimension @xmath0 .
( the force correlator decays for small distances . ) as @xmath45 is decreased several things can happen .
let us start with @xmath16 .
then for @xmath311 , since @xmath292 diverges for small @xmath45 , one sees from ( [ secder0 ] ) that @xmath308 becomes infinite as @xmath312 , where the larkin mass @xmath313 is the solution of : @xmath314 with @xmath315)$ ] and has the standard dependence @xmath316 of the inverse larkin length on the bare disorder ( a larkin length @xmath317 can be defined ) .
since @xmath318 is like @xmath319 positive , this divergence is the usual one of the frg , as also found in 1- and 2-loop studies @xcite , where it signals that the function @xmath320 becomes non - analytic and that a cusp singularity forms at @xmath18 in the second derivative @xmath321 , i.e. in the correlator of the pinning force .
this is usually interpreted as a glass phase with many metastable states beyond the larkin length .
thus for @xmath322 the function always becomes non - analytic at large scale ( small mass ) , and there is a single glass phase .
for @xmath323 , since @xmath292 is convergent , the cusp occurs only if the bare disorder is sufficiently large . at non - zero temperature @xmath281 ( [ secder0 ] )
shows that for @xmath324 thermal fluctuations do not change the scenario .
since @xmath294 remains finite , temperature only slightly renormalizes the value of @xmath313 downward , as @xmath325 for @xmath326 . for @xmath327 the effect of thermal fluctuations
is more important . for definiteness let us consider the set of models with power law correlations ( [ corrlr ] ) .
then ( [ secder0 ] ) becomes : @xmath328 since both integrals diverge for small mass as @xmath329 , @xmath330 , one can distinguish three cases : * if disorder correlations decay fast enough @xmath331 then the @xmath294 term wins and as @xmath117 one has @xmath332 , indicating that disorder is subdominant , resulting in a high - temperature phase . in that case
the solution is analytic as @xmath117 .
there is however a more complicated behavior for intermediate values of @xmath45 ( see appendix [ app : sr ] ) . *
if disorder correlations decay slower , i.e. @xmath333 , the term proportional to @xmath292 wins and the solution always becomes non - analytic at some larkin mass .
* in the marginal case , @xmath334 there is a transition at some critical temperature @xmath335 between a high temperature phase and a glass phase .
these features are very general and each of these cases will be studied in more details below .
one can immediately see that the existence of an analytic solution for @xmath336 is in one to one correspondence to the existence of a locally stable replica symmetric solution of the mp equations .
indeed the condition for the stability of the rs saddle point is precisely that the replicon eigenvalue be positive , namely that @xcite : @xmath337 be positive for all @xmath126 .
the rsb instability occurs when the lowest eigenvalue , which corresponds to @xmath338 , vanishes .
the condition @xmath339 is equivalent to the vanishing of ( [ secder0 ] ) , i.e. of the divergence of @xmath318 and the emergence of non - analytic behavior .
thus the generation of a cusp in the frg coincides at large @xmath1 exactly with the instability of the rs solution .
it is easy to see that an analytic solution @xmath296 of ( [ starting ] ) and ( [ frg1 ] ) can not describe the glass phase at @xmath16 .
indeed when @xmath296 is analytic , eq .
( [ 3cum ] ) and similar results for higher cumulants indicate that the full effective action is analytic .
it is then immediate to obtain correlations from its derivatives .
for instance , from ( [ lf20 ] ) the 2-point function at @xmath108 is simply : @xmath340 on the other hand , setting @xmath307 in ( [ starting ] ) one finds : @xmath341 thus at @xmath16 one recovers the dimensional reduction ( dr ) result @xmath342 with @xmath343 instead of a non - trivial value for @xmath3 expected in the glass phase .
furthermore since the effective action is analytic , all higher connected cumulants will trivially vanish at @xmath16 ( or be equal to the bare ones if the bare model contains such higher cumulants ) from the dr property .
clearly , in the glass phase , the dr scaling is expected to be incorrect and a non - analytic solution should be found , as well as a way to escape ( [ dr ] ) .
below we find how such a mechanism occurs within the frg .
it will emerge from our study that for the case where disorder is relevant in the large scale limit ( i.e. the long range case @xmath333 mentioned above ) the non - analytic solution of the frg equation will correspond to the full replica symmetry breaking solution of mp .
the situation for the short range case is more delicate .
both are discussed below .
the equation ( [ frg1 ] ) is valid ( for @xmath11 ) in any spatial dimension @xmath0 .
since one has the exact relation : @xmath344 one sees that the frg equation ( [ frg1 ] ) has a well defined limit @xmath345 for @xmath322 .
it makes formulae somewhat simpler so we will start by considering this case ; the case @xmath293 will be studied later .
note that although the equation has a well - defined limit , its solution may require a uv cutoff ( e.g. as is manifestly the case in integrating ( [ rel ] ) above ) .
thus from now on we study @xmath311 and consider the infinite uv cutoff limit .
then one has @xmath346 with @xmath2 .
it is convenient to define the rescaled dimensionless function : @xmath347 where @xmath3 is a fixed number , but for now arbitrary .
note that whether one works with @xmath348 or the rescaled @xmath349 does not make any difference for the possibility of a non - analyticity or a divergence of the second derivative .
then @xmath349 satisfies _ the frg equation in the infinite-@xmath1 limit _ : @xmath350 \nonumber \\ & = & ( \epsilon - 4 \zeta ) b(x ) + 2 \zeta x b'(x)\nonumber \label{rgscaled } \\ & & + \frac{1}{2 } b'(x)^2 - b'(x )
b'(0 ) + t_m \frac{b'(x ) } { 1 + \frac{b''(0)}{\epsilon } } + c_m \ . \nonumber \\ & & \label{frg2}\end{aligned}\ ] ] the rescaled temperature , and the energy exponent @xmath351 are defined as @xmath352 to obtain ( [ frg2 ] ) we have also integrated ( [ frg1 ] ) once , so there is a priori a @xmath45-dependent integration constant .
we emphasize that this frg equation ( [ frg2 ] ) that we have derived is valid , to dominant order in @xmath12 , _ in any dimension _ @xmath322 and at any temperature @xmath44 . in a previous study
@xcite balents and fisher studied another limit : arbitrary @xmath1 but only to first order in @xmath2 and @xmath16 .
if we consider the dominant order in @xmath1 of their equation , we find that it is identical to the @xmath16 part of ( [ frg2 ] ) ( up to some changes in notation ) .
equation ( [ frg2 ] ) however is valid to _ all orders _ in @xmath21 , an important point which the method used in @xcite could not address .
comparison of ( [ frg2 ] ) to our recent 2-loop , i.e. @xmath353 studies requires to expand to next order in @xmath12 and is performed in @xcite .
furthermore ( [ frg2 ] ) includes the effect of temperature to all orders in @xmath21 . expanding the term proportional to @xmath44 to lowest order in disorder @xmath211 , one finds the term @xmath354 .
this is the large-@xmath1 limit of the tadpole term obtained in the 1-loop frg at @xmath281 @xcite : @xmath355 where for infinite @xmath1 the last term drops out .
( it appears however to next order in @xmath12 @xcite . )
the form and the effect of the temperature term in ( [ frg2 ] ) to all orders in @xmath21 is radically different from its 1-loop truncation .
indeed , in the 1-loop frg the temperature is known to smoothen the cusp and render the function @xmath320 analytic in a boundary layer @xmath356 ( e.g. for @xmath9 @xcite ) with @xmath357 . here however , as further analysis confirms below , for @xmath358 the divergence of @xmath359 is self - reinforcing since it kills the term proportional to @xmath360 .
we find that it usually occurs at a finite ( larkin ) scale . in the marginal case @xmath361
, we will find non - trivial analytic finite - temperature fixed points .
let us now show how one can invert the self - consistent equation ( [ starting ] ) .
we first rewrite it in terms of the rescaled correlator @xmath362 where in the term proportional to temperature , for @xmath363 we mean @xmath364 choosing a bare temperature @xmath365 ( this choice is known to be necessary to give a universal and finite @xmath25 function , see e.g. the discussion in ref .
one can of course keep an explicit @xmath46 dependence everywhere , but that leads to needless complications without changing the result .
the above equation ( [ self ] ) is easily inverted into @xmath366 + \frac1{\epsilon } ( y - y_0 ) - \tilde{t}_m \label{self2 } \ , \ ] ] where we define @xmath367 with @xmath368 for @xmath327 , and @xmath369 is the inverse function of @xmath370 i.e. @xmath371 this means in turn that the frg equation ( [ frg2 ] ) is fully integrable , a feature not immediately obvious if one does not know that it originates from a self - consistent equation ( an observation not made in ref .
@xcite ) . to better understand this integrability property
let us show that ( [ frg2 ] ) can be transformed into a _ linear _ equation .
let us first take a derivative of ( [ frg2 ] ) and express it in terms of the new function @xmath372 ( [ new ] ) @xmath373 where we denote @xmath374 .
converting this into an equation for the inverse function @xmath375 one finds : @xmath376 with @xmath377 .
we have used that @xmath378 and have canceled a factor of @xmath379 on both sides .
( the validity near @xmath307 beyond the larkin length is reexamined below ) .
one recovers now that the general solution of this linear equation is ( [ self ] ) since it is the sum of the general solution of the homogeneous part @xmath380 \ , \ ] ] where @xmath381 is an arbitrary function , and of a particular solution @xmath382 the @xmath263 dependence obviously satisfies ( [ linear ] ) and for the constant part to work we use : @xmath383 the first line comes from evaluating ( [ y ] ) at @xmath307 and assuming analyticity , i.e. that @xmath384 , and equality which will not work beyond the larkin length ( @xmath385 ) , as found below . now that we have clarified the connections between the two approaches ( self - consistent equation and frg ) we can try to find solutions valid in the small mass limit . to analyze the solutions of the large-@xmath1 frg equation ( [ rgscaled ] ) ,
two approaches are legitimate , corresponding to different points of view .
the first , natural in mean field , is exact integration .
but then one discovers that the solution becomes non - analytic upon reaching the larkin mass .
it thus raises the non - trivial question on how to continue this solution beyond the larkin length . before doing so
, we will first examine a second point of view , more familiar from standard rg arguments .
the standard rg approach amounts to construct and compute the @xmath25-function of the theory , and then search for a fixed point ( function ) which describes the large scale physics .
usually , finding the basin of attraction of the fixed point , or relating arbitrary initial conditions to the final approach of the fixed point is an unmanageably difficult task .
it is fortunately also besides the goal of the rg which is to compute universal large scale physics independently of the irrelevant details of the bare model . here
, however , because of the large-@xmath1 limit , we can integrate the rg flow exactly and in principle `` solve '' any bare model .
let us temporarily ignore this integrability feature and focus on finding the zeroes of the @xmath25-function .
the @xmath25-function was derived previously within an @xmath21 expansion and non - analytic fixed points were found to one loop @xcite and also to two loops @xcite . in the latter case
additional `` anomalous '' terms are present in the @xmath25-function for the non - analytic theory to be renormalizable and a meaningful fixed point to exist . viewing the right hand side of ( [ frg2 ] ) as the large-@xmath1 limit of the true @xmath25-function ,
let us follow the same strategy and ask whether we can find non - trivial fixed points .
let us study @xmath16 and use the equivalent linear form of the frg equation .
we want to find the solutions @xmath372 of @xmath386 @xmath387 is a fixed number ( we want to impose @xmath388 ) , since we are looking for a fixed point function .
keeping @xmath387 arbitrary , one first tries a linear solution @xmath389 which yields @xmath390 and @xmath391 . writing @xmath392
one finds a homogeneous equation for @xmath381 and thus @xmath393 imposing now @xmath388 , i.e. @xmath394 , fixes the value of @xmath395 and one finds the _ family of zero temperature fixed point _
functions , parameterized by @xmath3 : @xmath396 since @xmath397 , @xmath398 one must have @xmath399 and thus @xmath400 the case @xmath401 corresponds to a larkin random force model . for the same reason , we must exclude the branch @xmath402 and thus @xmath403 is given by the unique solution of ( [ fixedpoints ] ) with @xmath397 and @xmath404 .
finally , for @xmath405 we find the fixed point : @xmath406 an important observation is that all of these fixed points exhibit automatically the expected cusp .
indeed one finds that @xmath407 , i.e. @xmath375 in ( [ fixedpoints ] ) vanishes and has a minimum at @xmath408 : @xmath409 this gives @xmath410 with @xmath411 , implying that the second derivative diverges as @xmath412 @xmath413 recalling that @xmath414 we see that all fixed points with @xmath415 correspond to a power - law long - range correlator @xmath349 , while @xmath416 corresponds to a gaussian short range disorder . if we follow the standard rg arguments , we can now sort the models ( [ correlator ] ) into these universality classes . since for the bare model @xmath417 and since the decay of @xmath15 in ( [ correlator ] ) at large @xmath10 can be argued to be identical for @xmath257 and @xmath348 ( for lr fixed points ) we find @xmath418 or @xmath416 for short range correlations .
these values are valid to dominant order in @xmath12 . in @xcite
the effect of the @xmath26 terms in the 1-loop frg equation was studied , i.e. the corrections of @xmath3 were estimated to order @xmath19 and at zero temperature . for sr disorder
it was found that the result of the gvm ( i.e. flory ) is corrected by terms @xmath419 exponentially small in @xmath1 , i.e. @xmath420 . for lr disorder with @xmath421 the result ( [ zetag ] )
was found to be uncorrected to @xmath19 .
( the crossover sr to lr occurs at @xmath422 such that @xmath423 ) .
one can in fact argue that ( [ zetag ] ) is always exact in the lr case ( see e.g. discussion in ref .
@xcite ) .
several important remarks are in order .
first we have found the fixed points of the inverted linear form ( [ linear ] ) of the frg equation .
a valid question is whether this is equivalent to finding the fixed points of the initial form of the @xmath25-function ( [ frg2 ] ) .
second we have found fixed points _ assuming _ that @xmath424 . since this is _ different _ from what has been found previously in ( [ dy0 ] ) at @xmath16 , one can ask whether these result are compatible .
these two questions have a common answer .
examining more closely what has really been done in this section , we note that it is equivalent to declaring both ( [ frg2 ] ) and ( [ linear ] ) valid for any @xmath397 and interpreting everywhere @xmath425 in ( [ linear ] ) and , equivalently @xmath426 as @xmath427 defined by continuity as @xmath412 .
this is legitimate since the transformation from ( [ frg2 ] ) to ( [ linear ] ) is certainly valid for @xmath397 and we note that this answer the second question above since eq .
( [ y ] ) , i.e. the derivative of ( [ frg2 ] ) , evaluated at @xmath412 yields : @xmath428 which works both in the regime @xmath429 where the solution is analytic @xmath430 and in the fixed point regime @xmath117 when the cusp has developed and the last term in ( [ cusp ] ) has a non - zero limit according to ( [ bp],[bpp ] ) .
we expect these fixed points to be the physically correct solutions at small @xmath45 .
we now investigate whether we can confirm this by providing the solution at infinite @xmath1 , for arbitrary mass @xmath45 , i.e continue our solution ( [ self ] ) beyond the larkin length . we now show that one can connect the two regimes , i.e. the regime for @xmath431 where an analytic solution exists to the asymptotic one , for @xmath117 , studied in the previous section . this can be done here because of the full integrability of the infinite-@xmath1 limit and provides a rare and non - trivial insight into what happens around the larkin scale .
it is instructive to start our analysis with the specific power law models with lr correlations ( [ corrlr ] ) , together with the case of sr correlations ( [ corrsr ] ) , in the form of a gaussian .
the solution for an arbitrary bare potential @xmath53 is more subtle , and will be given in section [ sec : solarbdocor ] , and appendix [ app : convergence ] . for the power law
correlators the inverse function in ( [ self ] ) is : @xmath432 for gaussian correlations it is : @xmath433 we can now insert this result into the general solution ( [ self2 ] ) of the self - consistent equation .
@xmath3 is arbitrary , but the convenient choice ( to later obtain a fixed point ) is @xmath434 such that the @xmath45 dependence of the first term drops . let us define @xmath435 we then obtain , for power law models : @xmath436 since we want @xmath437 i.e. @xmath394 .
this solution is valid for @xmath429 and the value of @xmath426 is the dr result ( [ dr ] ) .
for short range disorder the solution for @xmath429 is @xmath438 having set @xmath416 in that case .
we recall that @xmath439 .
note that the bare disorder is recovered for @xmath440 .
we have kept temperature , but here we discuss only the case where @xmath441 i.e. @xmath442 , or @xmath443 with @xmath444 . in that case
@xmath445 decreases as @xmath45 decreases , and , as mentioned above the role of temperature is minor . let us plot the r.h.s of ( [ solu1 ] ) , ( [ solu1exp ] ) on fig . [ fig2 ] .
the curve @xmath375 has the indicated shape in all cases .
it cuts the axis @xmath307 at @xmath408 and has a minimum @xmath446 at @xmath447 with @xmath448 _ independent _ of @xmath45 , and @xmath449 for sr disorder . for @xmath429
the minimum occurs at negative @xmath28 and the slope at @xmath450 is non - zero , indicating an analytic solution @xmath451 . for large @xmath45
only the first term on the r.h.s . of ( [ self2 ] ) contributes and one recovers essentially the bare disorder @xmath257 .
decreasing @xmath45 simply amounts to translate the curve upward along positive @xmath28 , and @xmath387 increases as the curve @xmath375 cuts the axis @xmath307 closer to the minimum .
it reaches it at the larkin mass , solution of @xmath452 , i.e. @xmath453 for sr disorder @xmath454 gives @xmath455 .
exactly as @xmath456 the solution acquires a cusp and one finds : @xmath457 i.e. the same result as ( [ bp ] ) . given by ( [ solu1 ] ) or ( [ solu1exp ] ) .
the physical branch is the one with @xmath458.,title="fig:",scaledwidth=50.0% ] although it is a priori not obvious how to follow this solution for @xmath385 , the following remarkable property indicates how to proceed . if we compute the @xmath25-function , i.e. the r.h.s.of ( [ linear ] ) using ( [ solu1 ] ) at @xmath459 and @xmath434 we find that it _
exactly vanishes_. similarly the @xmath25-function for @xmath460 also _ exactly vanishes _ for all @xmath397 provided we use also ( [ cusp ] ) , i.e all @xmath426 are defined as @xmath427 .
thus at @xmath459 the function _ has already reached its fixed point form _ @xmath461 , and _ freezes _ for @xmath385 . for the disorder correlators studied here
, @xmath349 evolves according to ( [ solu1 ] ) or ( [ solu1exp ] ) until @xmath313 where it reaches its fixed point @xmath462 , and does not evolve for @xmath463 . in particular @xmath464 freezes at @xmath313 and one has @xmath465 for @xmath463 , exactly as was discussed in the previous section .
the solution for @xmath466 is thus : @xmath467 where the parameter @xmath468 is defined in ( [ deftc ] ) , thus it exactly identifies with the zero temperature fixed point ( [ fixedpoints ] ) with @xmath434 , as can be explicitly verified .
this is easily understood a posteriori , since the same functions appear and in both cases we have two conditions to fix the two undetermined amplitudes @xmath469 .
it does however heavily rely on the exact power law form of the model , so it is not immediately obvious how it will extend to an _ arbitrary _ bare model @xmath53 .
one clearly can not expect in the general case that convergence to the fixed point will be completed within a finite scale .
the solution to this puzzle is given below .
similarly the solution for the gaussian sr disorder correlator for @xmath470 is given by setting @xmath459 in ( [ solu1exp ] ) , ( [ solu2exp ] ) with @xmath471 ( which determines @xmath313 ) .
the result of this section thus provides unambiguously a solution beyond the larkin scale which connects with the zero temperature fixed point .
it justifies the previous section and the value obtained for @xmath3 .
we found that for power law and gaussian models the freezing mechanism apparent in ( [ cusp ] ) leads to : @xmath472 the fixed point is reached at @xmath459 . in the case where disorder is relevant i.e. for @xmath473 ( i.e. @xmath442 ; @xmath327 for @xmath474 ) we found in the previous section that temperature plays only a minor role since the convergence to the non analytic zero temperature fixed point occurs on a finite ( larkin ) rg scale . whether it should be called a zero temperature fixed point
can also be debated since it is reached when @xmath475 .
a proper definition of the renormalized temperature may then include the denominator in ( [ frg2 ] ) .
let us now examine the marginal case @xmath476 , @xmath477 and @xmath31 for sr disorder .
we give here the main results , further details are examined in the appendix [ app : finitet ] the analytic solution is given by ( [ solu1 ] ) and @xmath387 given by ( [ solu2 ] ) , where here @xmath478 does not flow as @xmath45 is lowered .
let us examine the second derivative , @xmath479 \label{div } \ , \end{aligned}\ ] ] which is a rescaled version of ( [ b2 ] ) . the first line in ( [ div ] )
holds more generally ( in the infinite uv cutoff limit ) and to obtain the second we have set @xmath434 , @xmath480 and assumed @xmath327 . setting now @xmath481 , i.e. @xmath482 , we find that there is a transition at a temperature @xmath483 defined by @xmath484 such that for @xmath485 the solution is analytic for all @xmath45 down to @xmath486 , given by ( [ solu1 ] ) and @xmath359 remains finite and given by ( [ div ] ) .
this is a line of analytic fixed points which terminates at @xmath335 . for @xmath487
the solution freezes as in the previous section , and becomes non - analytic at and below the larkin mass @xmath488 the case @xmath13 , @xmath489 corresponds to the logarithmically correlated disorder @xmath490 .
it has been studied for finite @xmath1 in @xcite where it was shown that there is a transition _ for any @xmath1 _ at @xmath491 ( @xmath492 in the notations of ref .
the above result is in agreement with this value for @xmath335 .
there , for @xmath493 there is also a line of fixed points for @xmath494 with a continuously varying dynamical exponent ( and also one for @xmath495 with a different dynamical exponent and some form of rsb ) . since the dynamical exponent
is perturbatively related to @xmath359 , obtained above for infinite @xmath1 , it would be particularly interesting to study the @xmath12 corrections in this case .
let us now examine the case of sr disorder ( [ corrsr ] ) in @xmath31 .
more details are given in the appendix [ app : finitet ] .
one has @xmath496 .
the analytic solution ( [ solu1exp ] ) , ( [ solu2exp ] ) becomes @xmath497 with @xmath498 .
thus there is a transition at @xmath499 . for @xmath500
, we find @xmath387 to increase as @xmath45 decreases and reach @xmath501 at the larkin mass . for @xmath502
the solution remains frozen to ( [ solu0 ] ) with @xmath471 . for @xmath494
, we find that @xmath387 flows to zero and disorder is irrelevant .
the physics is the same as the one contained in the variational method for the periodic model in @xmath31 @xcite which exhibits a ( so - called marginal ) 1-step rsb solution .
the case @xmath503 , ( @xmath504 ) is discussed in appendix [ app : sr ] .
although an analytic solution exists as @xmath117 and disorder is formally irrelevant , there are some freezing phenomena at intermediate @xmath45 .
it corresponds to the case where mp find , in addition to a rs solution , a 1-step rsb solution which is so called non - marginal ( different in nature from the one step solutions obtained in the case @xmath361 ) .
in this section we compare the frg approach at large @xmath1 with the gvm using rsb .
since the two methods study the same model in the same limit ( large @xmath1 ) a precise connection should exist .
we start by comparing the two methods at the level of the results of the calculations .
we first perform the comparison for power law models .
then we generalize the frg solution to arbitrary bare disorder correlator .
based on these results , we address the deeper connections between the two methods , and emphasize what we learn from them about the physical consequences . our main result up to now is a non - trivial solution for the renormalized disorder correlator @xmath92 as a function of the scale parameter @xmath45 , i.e. the effective action for the zero momentum mode .
since this function is once differentiable , i.e. @xmath505 , we can extract from its first derivative the 2-point correlation function at zero momentum ( see section [ orderparam ] ) : @xmath506 where in the last equation we have used the definition ( [ lf43 ] ) for the rescaled function @xmath211 , and added the index @xmath45 to recall its dependence on the mass . in the case
@xmath507 , for the power - law models ( [ corrlr ] ) , we thus find , using @xmath508 from ( [ solu2 ] ) @xmath509 and , for @xmath431 the dr result ( [ lf41 ] ) , ( [ dr ] ) ( where @xmath313 is determined by ( [ deftc ] ) ) and , we recall , @xmath510 .
let us recall the rsb solution at large @xmath1 and resolve carefully the mp saddle - point equation in presence of a mass .
we only assume that there is indeed full rsb , to be checked a posteriori .
let us first reexpress the general solution , valid for an arbitrary @xmath257 , in a rather compact form . in the rsb method one
first parameterizes the correlation matrix as @xmath511 and the self - energy matrix @xmath512 , in terms of the overlap @xmath513 between ( distinct ) replicas @xmath210 and @xmath211 ( and denote @xmath514 ) .
the saddle point equations then read @xmath515({\sf w}))^2}\qquad \label{sadsig2 } \\
\tilde{\chi}({\sf u}^c ) & = & 2 t \int_k \frac{1}{k^2 + m^2 + \sigma_c } \label{sadsig3}\end{aligned}\ ] ] with @xmath516({\sf u})={\sf u } \sigma({\sf u } ) - \int_0^{\sf u } { { \mathrm{d}}}{\sf w}\ , \sigma({\sf w})\ ] ] and @xmath517({\sf u } \geq { \sf u}_c)$ ] . the last two equations are the rsb - matrix inversion formulae ; @xmath518 is assumed to be continuous . taking a derivative of ( [ sadsig ] ) w.r.t .
@xmath519 gives @xmath520({\sf u}))^2}\ .\ ] ] this equation admits two solutions : either @xmath518 is constant , or satisfies the _ marginality condition _ @xmath521({\sf u}))^2 } \ .\ ] ]
we thus look for a solution of the full rsb equations ( see fig
. [ mpfunction ] ) with a non - trivial function @xmath518 for @xmath522 joined by two plateaus @xmath523 similar forms are valid for @xmath524 and @xmath525 .
the breakpoint @xmath526 is related to the physics at the larkin scale @xmath527 , which , at weak disorder , can be much smaller than the uv scale @xmath46 , while @xmath528 depends on the ir cutoff @xmath45 .
( [ marginality ] ) also yields by continuity a closed equation which determines @xmath529 @xmath530 as well as @xmath531 since @xmath532({\sf u})=0 $ ] for @xmath533 . to solve these equations
one firsts determines the function @xmath532({\sf u})$ ] ( see below ) , then finds @xmath526 and @xmath528 .
( { \sf { \sf u}})+ m^2 $ ] and a finite mass @xmath45 .
@xmath518 has identical behavior , with two plateaus at values @xmath534 and @xmath535 . in both cases , upon increasing the mass only , @xmath528 varies ( increases ) and the lower plateau moves up , the rest of the function being unchanged , see eqs .
( [ rgrsb1 ] ) f. in the text .
the dashed line is the zero mass solution .
the lower breakpoint @xmath528 reaches the upper one @xmath526 at the larkin mass @xmath459 above which the solution becomes rs .
the frg gives exactly the lower plateau value for @xmath536 and its @xmath45-dependence . from it ,
the full rsb solution can be reconstructed , see section [ reconstruct].,title="fig : " ] one can already note at this stage that ( [ scsc ] ) is exactly the condition ( [ lf39 ] ) which determines the larkin mass @xmath313 , equivalent to the vanishing of the replicon : @xmath537 and @xmath538 ( no rsb ) for @xmath431 . to find @xmath532({\sf u})$ ] for arbitrary @xmath257 and cutoff , one notes @xcite that with the help of ( [ marginality ] ) and ( [ sadsig ] ) @xmath539
can be expressed as a function of @xmath532 $ ] as @xmath540({\sf u}))^2 } } \bigg)\right ) \ , \ ] ] where @xmath541 is the inverse function of @xmath542 .
then one notes that @xmath519 as a function of @xmath539 of @xmath543 $ ] is from ( [ sigmasimgabracket ] ) simply @xmath544 $ ] .
this yields immediately , using the chain rule : @xmath545({\sf u}))^2}\big]^3}{\int_k \frac{1}{(k^2 + m^2 + [ \sigma]({\sf u}))^3 } } \nonumber \\ & & \times b'''\left ( ( b'')^{-1}\bigg ( \frac{1}{4 \int_k \frac{1}{(k^2 + m^2 + [ \sigma]({\sf u}))^2 } } \bigg)\right ) \ .\qquad \end{aligned}\ ] ] upon inversion one obtains the exact function @xmath532({\sf u})$ ] , and inserting into ( [ sigmausol ] ) @xmath518 .
more precisely , we see that the sum @xmath532({\sf u } ) + m^2 $ ] is a @xmath45-independent function of @xmath546 , and thus from ( [ sigmausol ] ) @xmath547 is also @xmath45-independent .
then one solves the self - consistent equation ( [ scsc ] ) for @xmath529 , and finally obtains @xmath526 from the above .
the result can be written using ( [ scsc ] ) in the simple form @xmath548 ^ 3}{\int_k \frac{1}{(k^2 + m^2_c)^3 } } ~~ b'''\left ( ( b'')^{-1}\bigg ( \frac{1}{4 \int_k \frac{1}{(k^2 + m^2_c)^2 } } \bigg ) \right)\ .\ ] ] thus @xmath526 depends only on the larkin mass and is independent of @xmath45 ( see appendix [ app : mprg ] for another derivation and a discussion of this useful property ) . similarly one obtains : @xmath549 let us apply these considerations to the power law model ( [ corrlr ] ) .
for this model the larkin mass is determined by ( [ deftc ] ) .
next one has : @xmath550 in the limit of infinite uv cutoff @xmath46 limit , using @xmath551 and @xmath552 we obtain from ( [ usolu ] ) @xmath553({\sf u } ) & = & ( \tilde a { \sf u})^{2/\theta } \\ \tilde a & = & \frac{1}{4 t a_d ( 1 + \gamma ) } \epsilon^{\frac{1 + 2 \gamma}{1 + \gamma } } ( \gamma \tilde g)^{\frac{1}{1 + \gamma}}\qquad \\ { \sf u}_m & = & m^{\theta}/\tilde a \\ { \sf u}_c & = & m_c^{\theta}/\tilde a \end{aligned}\ ] ] with @xmath554 , @xmath555 . using ( [ sigmausol ] ) one finds the @xmath45-independent result @xmath556 in particular one has the value of the lower plateau ( see fig .
[ mpfunction ] ) @xmath557 let us already note the relation @xmath558 which will be demonstrated to hold more generally below .
the inversion formula yielding the diagonal correlation from the rsb solution is @xmath559({\sf u})}{m^2 + [ \sigma]({\sf u } ) } \right]\ ] ] and is a sum of contributions from all overlaps @xmath560 . in particular
the contribution from states with _
zero overlap _ , i.e. the most distant states , is : @xmath561 we can now compare with the frg .
one has , using @xmath562 , @xmath563 : @xmath564 as given by ( [ resfrg ] ) .
thus , for this power - law model , we found that the frg gives exactly and only the contribution from the most distant states ( the lower plateau in the rsb solution ) . before discussing the reasons and consequences , let us show that this feature is much more general than power law models , and holds in any case where full rsb holds . in section [ fullsolumc ]
we found how to continue the solution of the frg equation beyond the larkin scale .
it involved freezing of the @xmath45 dependence of @xmath565 at @xmath459 and worked only for two special forms of disorder correlators , which happened to be already fixed point forms .
it is important to find the solution for a more general form of the bare correlator @xmath53 , and this is what we achieve here .
let us examine whether we can find a solution for any @xmath45 of the frg equation ( [ linear ] ) in inverted variables @xmath566 which correspond to a more general function @xmath53 .
we take special care here to indicate that @xmath567 is an @xmath45 dependent function of @xmath263 ( we note @xmath568 and we recall that @xmath569 ) .
the idea is to play with the @xmath45 dependence of @xmath570 since this is really all the freedom we have .
let us restrict our analysis for simplicity to @xmath16 , the generalization being straightforward .
the definition of @xmath571 is given implicitly by @xmath572 for all @xmath45 .
the total derivative thus vanishes : @xmath573 together with ( [ frgeqlin ] ) at @xmath574 , it yields ( recall that @xmath575 ) : @xmath576 there are only two possible solutions : @xmath577 the first holds before the larkin scale and the second , which implies a non - analytic @xmath578 , beyond .
we now want to find the solution beyond the larkin scale , i.e. assuming that @xmath579 , together with @xmath580 , which of course implies @xmath581 .
equation ( [ frgeqlin ] ) with @xmath570 is trivially separable and admits the general solution @xmath582 where for now @xmath583 is arbitrary and so is the function @xmath584 .
( it will be identified below with @xmath585 as in section [ detailed ] ) .
the first condition one must impose is the definition @xmath586 , i.e. @xmath587 which should be valid both for @xmath429 and @xmath463 . taking @xmath588 of ( [ cond1 ] ) yields , using ( [ cond1 ] ) again @xmath589
\left [ m \partial_m y_0 + ( \epsilon - 2 \zeta)y_{0 } \right ] = 0 \ .\ ] ] in order to satisfy this equation , at least one of the factors must vanish .
the regime @xmath590 corresponds to the first , the regime @xmath591 to the second factor being zero . for @xmath429
one has @xmath592=0 $ ] leading to @xmath593 and the above solution becomes : @xmath594 this can clearly be identified with the analytic solution of the self - consistent equation ( [ self ] ) found before in section [ detailed ] , and thus implies that @xmath369 is the reciprocal function of @xmath595 .
( [ cond1 ] ) is trivially satisfied by @xmath596 applying @xmath595 to ( [ 6.80 ] ) fixes @xmath40 to be @xmath597 and one recovers the dimensional reduction result .
the interesting new information is obtained for @xmath463 .
then the first factor in ( [ master ] ) vanishes , i.e.@xmath598 deriving ( [ 6.78 ] ) w.r.t .
@xmath263 one sees that ( [ cuspcond ] ) correctly implies @xmath599 thus the solution for @xmath600 has a cusp .
( [ cond2 ] ) determines the function @xmath583 for @xmath463 .
note that if the power law in the correlator holds only asymptotically , @xmath583 will nicely converge to a constant ( for the right choice of @xmath3 ) due to the asymptotic power law tail , but may vary arbitrarily according to the irrelevant corrections to power law .
this is studied in more details in appendix [ app : convergence ] .
it is convenient to rewrite the final result , i.e. eqs.([cond1 ] ) , ( [ cuspcond ] ) in the form : @xmath601 where we use the notation @xmath602 .
the connection with the rsb solution becomes obvious in this form .
comparing with ( [ lower ] ) , the equation ( [ rep2 ] ) of the frg solution identifies with the marginality condition at @xmath603 , the lower plateau of the rsb solution , see fig .
[ mpfunction ] .
it allows to determine @xmath604 ; the two other equations are self - consistently obeyed and give @xmath605 . comparing with ( [ sadsig ] ) at @xmath606 yields the identification @xmath607 and thus we obtain : @xmath608 it thus holds for an arbitrary disorder correlator , provided a solution to eqs .
( [ rep3 ] ) , ( [ rep2 ] ) exists , i.e. for the class of functions @xmath609 which yield full rsb ( also called continuous rsb ) within the mp approach . of course , equations ( [ rep3 ] ) , ( [ rep2 ] )
were derived _ without any assumption about replica symmetry breaking_. extension to @xmath281 is obvious . adding the last term of ( [ linear ] ) and following the same steps as above
, one finds : @xmath610 x_m'(y_0 ) = 0 \ .\ ] ] vanishing of the first factor yields the finite @xmath44 analytic solution studied in the previous section ( equivalent to the rs solution of mp ) .
continuation beyond the larkin mass implies @xmath611 , in which case the additional temperature term in ( [ linear ] ) vanishes and one is back to the @xmath16 equations ( [ rep2 ] ) , ( [ rep3 ] ) : thus only the value of the larkin mass depends on temperature , everything else is independent of @xmath44 .
in the previous section we have shown that the frg yields @xmath612 ( via @xmath605 ) i.e. the value of the rsb function only at @xmath613 .
in fact , as we now discuss , by varying the mass one can scan the whole function @xmath518 of mp for any @xmath546 , and thus the frg yields the same information as contained in the function @xmath518 .
remarkably , we can obtain an explicit expression for @xmath518 , even though the argument of this function , the `` overlap '' is not obviously related to any quantity in the frg .
furthermore , we can also compute the full correlation - function of mezard parisi , if one knows only @xmath614 for all @xmath45 , which is given by the frg .
thus from now on we assume that we know only @xmath612 as a function of @xmath45 through the frg , together with some general properties of the mp solution . as we have already found in section [ mpfull ] , and is shown more directly in appendix [ app : mprg ] , the gvm saddle point equations , upon assuming full continuous rsb , satisfy the two `` rg equations '' @xmath615({\sf u } ) + m^2 ) & = & 0 \label{rgrsb2 } \ , \end{aligned}\ ] ] valid for any @xmath546 such that @xmath616 .
one can thus relate the solution @xmath617 at finite @xmath45 to the solution at zero mass @xmath618 .
note that eqs .
( [ rgrsb1 ] ) and ( [ rgrsb2 ] ) have been hypothesized by parisi and toulouse for the sk - model @xcite .
however , it has been shown that there they are only approximately satisfied , see e.g. .
one notes that the upper breakpoint @xmath626 is independent of @xmath45 . as shown in section ( [ mpfull ] ) , @xmath528 increases upon increase of @xmath45 , and reaches @xmath526 at the larkin mass , i.e. for @xmath627 .
let us show how one can recast the correlation function of mp , given in ( [ diagcorr ] ) at zero momentum , entirely using frg data .
@xmath628({\sf u})}\right ) { \nonumber}\\ & & + \int_{{\sf u}^c}^{1 } \frac{{{\mathrm{d}}}{\sf u}}{{\sf u}^2 } \left(\frac{1}{m^2 } - \frac{1}{m^2 + \sigma^c } \right ) \ .\end{aligned}\ ] ] using our previous results gives : @xmath629({\sf u})}\right)\nonumber \\ & & + \left ( \frac{1}{{\sf u}_c } -1\right ) \left(\frac{1}{m^2 } - \frac{1}{m^2_c}\right ) \ .
{ \nonumber}\end{aligned}\ ] ] shifting from the variable @xmath546 to the variable @xmath630 defined by @xmath631 , one finds that the correlation function can be expressed _
entirely _ from the knowledge of @xmath614 . to see this , note that @xmath632 this gives @xmath633 where we have used that @xmath634(u_\mu ) = \mu^2 $ ] . after an integration by part and using again ( [ umu ] ) at @xmath635 one finds the remarkably simple formula @xmath636 valid for @xmath637
recalling the relation between @xmath614 and @xmath638 we see that the mp result is a simple average of the correlations corresponding to masses between @xmath45 and @xmath313 .
note that one can derive a similar formula for @xmath639 obtained by mp as a function of @xmath640 . in summary ,
although strictly speaking our frg result gives only the contribution of distant states to the 2-point correlation function , it does allow to obtain the whole mp result , although we do not yet have a derivation within the framework of the frg alone .
one should also note that the formula ( [ inversionchi ] ) is in a sense equivalent to the inversion formula of hierarchical matrices which relates @xmath641 to the self energy @xmath518 .
this raises the question of whether the frg equations `` know '' about ultra - metric matrix inversion .
these result hold for continuous rsb and the case of non - marginal rsb ( when the marginality condition is not obeyed ) is discussed in appendix [ app : sr ] .
let us examine what has been achieved and how it compares with other works .
we are interested in the behavior of the effective action of the replicated field theory for @xmath1 large .
let us focus here on the uniform configuration , for which @xmath642 , where we denote @xmath643 the effective action per unit volume .
@xmath644 then represents the free energy per unit volume , depending on a set @xmath85 , @xmath645 in presence of external sources @xmath646 which impose field averages @xmath647 . the usual free energy @xmath648 is recovered for @xmath18 , @xmath649 .
the saddle point method allows to write in the limit of infinite @xmath1 : @xmath650 where @xmath651 satisfies the saddle point equations ( [ lf33]),([freeenergy ] ) , and ( [ saddle ] ) , and the @xmath652 denotes a sum over saddle points whenever more than one solution exists . ) yields ( [ superpos ] ) . ] to be accurate the saddle - point method computes @xmath653 where the limit is taken _ at fixed @xmath55_. as mentioned in section [ sec : sp ] , the saddle point equations ( [ saddle ] ) contain _ both _ the frg and the gvm .
they depend on the set of @xmath654 , and when expanded in cumulants , taking _ all _ @xmath655 , they lead to the frg equations .
this approach clearly consists in imposing an _ explicit _ breaking of replica symmetry .
also we expect that in that case a single saddle point exists .
this is indicated by the fact that the quantity @xmath656 plays the role of a replicon eigenvalue and remains frozen and positive for @xmath657 .
by contrast , mp studied the case where all @xmath93 and found _ spontaneous _ replica symmetry breaking , i.e. multiple saddle points , differing by permutations @xmath658 of replicas .
we can now make contact between the two approaches and understand why we have obtained via the frg the correlations of mp corresponding to the distant states .
let us focus on the mode @xmath108 , and define the center of mass variable @xmath659 ( i.e. without rescaling in @xmath1 ) and consider : @xmath660 @xmath661 is the probability distribution of @xmath662 in a given disorder configuration : @xmath663 \delta\left ( \tilde{u } - \frac{1}{l^d } \int_x u(x ) \right ) { { \mathrm{e}}}^{- { \cal h}_v[u]/t}\ .\ ] ] in the present paper we have computed ( [ zj ] ) , scaling @xmath664 , and taking all @xmath646 different in order to impose all @xmath665 .
because of this scaling with @xmath1 we obtained a _
different _ saddle point than mp ( shifted by @xmath120 , see ( [ spw ] ) ) , and since all @xmath646 are different , this saddle point has explicit rsb .
according to section [ sec : calculgamma ] this gives @xmath24 $ ] when @xmath10 scales as @xmath666 , i.e. we determined the averaged probability ( [ prob1 ] ) , ( [ prob2 ] ) for fixed @xmath667 .
on the other hand mp found that : @xmath668 one should in principle be able to recover the 2-point correlation function ( [ diagcorr ] ) obtained by mp adding small sources @xmath646 as in ( [ zj ] ) and taking derivatives at @xmath669 .
clearly , to reproduce the mp result , these should be taken as @xmath670 and not @xmath671 so as to maintain the unperturbed mp saddle point .
for instance the diagonal 2-point correlation function is obtained using @xmath672 , all others @xmath673 , and differentiating twice w.r.t @xmath106 .
( the off diagonal one involves @xmath674 ) .
equivalently it should be obtainable from the effective action for @xmath675 . thus scaling @xmath676 and @xmath677 as was done here
selects the distant states in the mp solution .
the fact that we obtain exactly the mp result for these states shows that there is no intermediate scaling regime .
we emphasize that our primary aim here is not to recover the mp result , but to understand what exactly the frg predicts , in view of getting a better understanding of frg within e.g. the @xmath21-expansion .
extension of the frg beyond the larkin scale requires giving a meaning to the limit @xmath678 .
we find here that what the frg actually computes ( from @xmath427 ) is a second moment of @xmath121 in presence of a small extra external source @xmath679 such that all @xmath680 , i.e. an average , such that when there are several states , the different replicas are chosen in maximally separated states ( @xmath613 ) .
note that the quantity computed by the frg specifies the system s preparation , while such a procedure still has to be worked out for the mp solution . in presence of a broken symmetry
this is an important issue , and the frg gives a natural solution
. it would be interesting to understand further the limit @xmath681 coming from our solution , which one can call the crossover from frg to rsb .
it is clearly non - trivial .
for instance , one question is what we get if we take a source @xmath682 so that we still have spontaneous rsb in @xmath683 copies , or if we divide the replicas in two groups of @xmath684 each , @xmath685 so that rsb persists within each packet . another important issue is what happens at large but finite @xmath1 . for any @xmath1 , if one parameterizes the 2-replica part of the effective action using @xmath686 , one can write the 2-point correlation function as @xmath687 for @xmath688 . we have determined the function @xmath689 for @xmath690 , i.e. @xmath691 . to obtain @xmath692 for finite @xmath1
however , one needs a priori to know @xmath693 for @xmath694 , i.e. @xmath689 for @xmath695 .
the two could be the same , or there could be a boundary layer of size @xmath696 .
a priori the knowledge of this requires including the @xmath12 corrections in the frg equation ( as is examined in @xcite ) .
this may help to better understand the connection of this regime to rsb .
this is important since there are cases ( e.g. for @xmath13 , @xmath697 ) where we know that paris - type rsb can not survive at finite @xmath1 . a previous study @xcite aimed at connecting the rsb solution to the frg .
the authors defined , for each configuration of the disorder , an `` effective random potential '' @xmath698 for a given mode ( e.g. the zero mode ) . starting from the mp solution ( [ mpproba ] ) , they computed the second cumulant of @xmath699 and showed that it exhibits a non - analyticity , reminiscent of the one found in the frg .
a parallel was drawn with a @xmath13 , @xmath9 toy model where @xmath699 satisfies an exact rg equation of the burgers - kpz type with random initial conditions .
shocks , well known to develop in this equation , provide an appealing physical picture for the singularities in the energy landscape responsible for the non - analyticity in the frg beyond the larkin length @xmath700 .
comparison between this study and the present one shows several important differences , with interesting physical consequences .
the scaling in @xmath701 , in @xmath1 , and the definition of the `` renormalized '' disorder are different .
as here , the authors of ref.@xcite focus on the zero mode , but with a different scaling with system size : they define @xmath702 , such that fluctuations of @xmath703 remain of order one .
other quantities are : @xmath704 this definition means that the mode @xmath703 sees an equivalent @xmath13 toy model , with random potential @xmath698 . comparing with ( [ prob2 ] ) and ( [ lf21 ] ) , we see that since the rescaling in @xmath701 is different there can be no relation between @xmath705 or @xmath706 and the two point correlation function , neither the one of mp , nor the one obtained here in the frg . to obtain the 2-point correlation of mp one would still have to _ solve _ the toy model defined by @xmath698 , i.e. compute : @xmath707 this task is difficult , since it requires not only the second cumulant , but also higher ones ( not computed in @xcite ) .
more importantly , it requires the large argument behavior @xmath708 of @xmath709 , not obtained in @xcite , were attention was focused on the larkin scale ( see below ) .
thus the information contained in @xmath710 is physically interesting but not obviously related to large scale correlations .
it is in a sense ( e.g. for the @xmath13 case discussed in @xcite ) closer in spirit to wilson - polchinski type rg @xcite versus an rg based on the effective action ( see @xcite ) .
reexpressed in the variables of the present work , the result of @xcite reads : @xmath711 where @xmath712 is a constant .
because of the different rescaling , the non - analytic term has a coefficient growing with the system size , which expresses again that it is not an effective action .
however , since the @xmath713 term is simply the bare disorder , and the non - analytic term involves only the larkin scale @xmath700 , it seems that this carries information for and only for the physics below and around the larkin length , and does not contain any information about large scale behavior .
thus , despite exhibiting shock behavior at the larkin scale , we think it has little to do with the frg as a perturbative method to obtain large scale behavior . not surprisingly then , @xmath710 is non - perturbative in @xmath2 , contrarily to the one obtained in standard frg , which is of order @xmath714 .
another important difference with the present approach is the scaling with @xmath1 .
the approach of @xcite used the unperturbed mp saddle point and thus , as was extensively discussed in the previous section , it focuses on @xmath715 while we focus on @xmath716 , ( i.e. @xmath717 there and @xmath718 here ) .
further work is needed to connect these regimes . on the other hand it seems that the thermal boundary layer can be found within this approach .
in this paper we have studied the frg at large @xmath1 . from an exact saddle point calculation of the replicated effective action at large @xmath1 we have derived the exact renormalization group equation , valid in any dimension @xmath0 for infinite @xmath1 , for the field theory of pinning .
it is expressed as the @xmath25-function for the second cumulant of the disorder correlator , and is exact as the second cumulant satisfies a remarkably simple closed equation . to order @xmath19
it agrees with the one derived by balents and fisher .
this result teaches us a lot about how the frg works and helps put the frg approach to the @xmath21 expansion on more solid grounds . since here
the frg flow equation is formally equivalent to a self - consistent saddle point equation , it is fully integrable , i.e. one can follow the rg flow from any initial condition .
it is thus possible to examine in detail what happens around the larkin length and how the disorder correlator develops the non - analyticity .
let us emphasize that this is the first time that the emergence of non - analytic behaviour in the frg is proven rigourously , beyond perturbative calculations .
indeed , the 1-loop frg is insufficient per se to provide such a proof since the runaway of @xmath719 could very well be argued to be the analog of the famous landau ghost , i.e. a flow to a strong coupling fixed point without the need , or better the possibility , for renormalization within a non - analytic functional space .
here we demonstrate that this is not the case , at least for infinite @xmath1 .
if we had restricted the analysis to the self - consistent equation , the continuation beyond the larkin scale would have seemed quite problematic .
remarkably , the frg equation , equivalent below the larkin scale , provides an unambiguous way to continue the flow equation beyond the larkin scale .
even more remarkably , its solution reproduces exactly the small overlap result of the full rsb solution of mp , a non - trivial result which , within mp , can not be obtained without constructing the full rsb solution .
the mechanism for this seems to be that the frg solution in that case naturally satisfies the so - called marginality condition .
in fact , it turns out to be equivalent to it , and we were able to find a formula yielding the complete rsb solution for all overlaps .
this is striking since we did not make any assumption about parisi rsb .
we avoided the issue altogether by using a method where rsb is not spontaneous , but explicit .
given that the validity of the parisi ansatz , e.g. for the sk model , has not yet been proven ( despite recent progress @xcite ) , it is interesting to find a method which does _ not _ rely on rsb .
in fact there may well be deeper connections to be unveiled between the parisi algebra of ultrametric matrices and the type of singular differential equations arising in the frg .
another example where a rsb solution can equivalently be obtained via an rg type equation is the derrida spohn solution of the dp on the cayley tree @xcite .
this has inspired a similar solution for a model with translationally invariant disorder in @xcite .
we have thus shown agreement with the main results of the full and the marginal one - step rsb solutions of mezard - parisi .
this is also interesting since it has been widely debated @xcite whether the rsb method captures the physics : our results raise no doubt for infinite @xmath1 .
more puzzling is the situation for sr disorder . there mp find both a stable replica symmetric solution and a 1-step solution where minimization over the breakpoint has to be enforced ( marginality condition violated ) . for gaussian disorder
both solutions of mp have @xmath720 .
similarly the frg naturally finds the finite temperature rs solution with @xmath720 ( and one fixed point solution at @xmath16 with @xmath416 ) .
a non - analytic solution also seems to exist in the frg , and work is in progress to analyze it further and elucidate whether it is related to the non - marginal 1-step solution of mp .
let us note that in @xmath13 we essentially know that ( apart from @xmath489 ) rsb does not hold at finite @xmath1 ( the phase transition predicted by the gvm in that case must go away at finite @xmath1 , @xmath281 ) . so there is little doubt that the correct branch at finite @xmath1 is the rs one , as also given by the frg .
for the dp problem with @xmath8 on the other hand , it is not yet clear whether both branches ( a @xmath16 fixed point starting from @xmath416 or a finite @xmath44 non - analytic solution with @xmath720 ) can coexist .
one scenario is that they would cross over at some lower value of @xmath721 yielding the upper critical dimension of kpz .
the calculation of the frg @xmath25-function to next order in @xmath12 should shed light on this question , and is thus of high interest .
it is presented in @xcite .
our method thus provides a unique candidate for a field theory of the strong - coupling phase .
to summarize , the present method is promising in solving mean field models , by using explicit rather than spontaneous rsb .
it would be of interest to investigate whether other models like the sk - model could be solved via the same route .
more importantly , it may open an alternative road to tackle disordered systems from a different direction than expanding around rsb saddle points , a task which still has to be accomplished .
of course , in the end , the same difficulties may be in store .
they could hide in the subtelties due to the non - analytic behaviour of the @xmath25-function at large @xmath1 .
however we are optimistic , since we have understood the infinite-@xmath1 limit , at least in the full - rsb case .
also , a solution has been found for @xmath9 to two @xcite and three loop order @xcite , and for finite @xmath1 at 2-loop order .
let us close by indicating that many extensions of this work are possible and some in progress .
one example is the random field problem , still under intense debate @xcite , for which we have also computed @xcite the effective action at large @xmath1 , and at 2-loop order .
finally , the same method applies to the dynamics , classical or quantum : it has been shown @xcite that the mode - coupling approximation in glasses @xcite identifies with ( non - marginal ) mean field ( large @xmath1 ) dynamics , exhibiting aging solutions .
however this picture leaves out thermally activated processes , and our @xmath12 method may be promnising there too .
we gratefully acknowledge discussions with l. balents , e. brzin , j.p .
bouchaud , l. cugliandolo , y.y .
goldschmidt , j. kurchan and m. mzard .
is supported by the deutsche forschungsgemeinschaft through heisenberg grant wi 1932/1 - 1 , and in part by the national science foundation under grant phy99 - 07949 .
let us extend the variational method of ref .
@xcite to the case where the average of the field @xmath41 is fixed to a non - zero value denoted here @xmath722 .
one defines the variational hamiltonian and free energy : @xmath723 & = & \frac{1}{2 } \int_{xy } \left[u_a(x ) - \overline{u}_a(x ) \right ] \nonumber \\ & & \qquad \times ( g_{\mathrm{var}})_{xa , yb } \left[u_a(y ) - \overline{u}_b(y ) \right ] \nonumber \\ { \cal f}_{\mathrm{var}}[g_{\mathrm{var}},\overline{v } ] & = & - t \ln \int { \cal d}[u ] \exp\left ( - { \cal h}_{\mathrm{var}}[u]/t\right ) \nonumber \\ & & + \left < { \cal h } - { \cal h}_{\mathrm{var } } \right>_{{\cal h}_{\mathrm{var } } } \ , \end{aligned}\ ] ] which satisfies ( for @xmath67 positive integer ) the usual bound @xmath724 . here
$ ] defined in the text , is the replicated hamiltonian . comparing with ( [ var ] )
one finds that @xmath726/t = \gamma_0[g_{\mathrm{var}},\overline{v } , \hat{u}]\ , \ ] ] where the last argument indicates that for @xmath1 finite @xmath77 should be replaced by @xmath727 ; in the infinite-@xmath1 limit @xmath728 . restricting to a bare model with only a second cumulant
one finds ( omitting the bars on @xmath55 ) : @xmath729 in general , @xmath730 is a function of two variables , which becomes a function of the sum @xmath731 only as @xmath732 , since in that limit @xmath733 and @xmath734 , without fluctuations , in the gaussian measure @xmath735 .
in some applications bilocal terms may already be present in the starting action .
let us thus give a more general and compact result , which also includes that case .
it is derived by a simple extension of the methods of section [ sec : calculgamma ] .
let us consider a @xmath1-component field @xmath736 , with components @xmath737 , @xmath738 , which can carry other indices , coordinates , or be a set of fields , etc . a general @xmath49-invariant form for the action functional is @xmath739 & = & \frac{1}{2 } \phi : \ !
c^{-1}\ ! : \phi + n s_{\mathrm{int}}[\psi ] \\
\psi_{xy } & = & \frac{1}{n } \phi_x \cdot \phi_y \label{psi } \ , \end{aligned}\ ] ] where @xmath740 $ ] is a functional of the _ bilocal field _ @xmath741 ( which is also a bi - index matrix if the field @xmath381 carries other indices , etc . ) .
then the effective action associated to @xmath742 can be written as @xmath743 = \frac{1}{2 } \phi : \ !
c^{-1}\ ! : \phi + n \gamma^0[\psi ] + \gamma^1[\psi ] + \dots \ , \ ] ] where @xmath744 $ ] is a functional of the bilocal field @xmath741 in ( [ psi ] ) and satisfies the self - consistent equation : @xmath745 & = & \frac{\delta s_{\mathrm{int}}[\chi]}{\delta \chi_{xy } } \\
\chi_{xy } & = & \psi_{xy } + g[\psi]_{xy } \\
g[\psi]_{xy } & = & \left ( c^{-1 } + 2 \frac{\delta \gamma^0[\psi]}{\delta \psi } \right)^{-1}_{xy } \ \end{aligned}\ ] ] @xmath746 is another ( set of ) bilocal fields .
in this appendix we compute the third and fourth renormalized cumulants of the disorder .
one uses the parameterization : @xmath747 we need the matrix @xmath748 up to the fourth cumulant : @xmath749 \nonumber \\ & & + \frac{1}{t^3 } [ \delta_{ab } \sum_{c g h } \tilde{q}'_{1,acgh } - \sum_{gh } \tilde{q}'_{1,abgh } ] \
.\qquad \end{aligned}\ ] ] the equality of ( [ chipower1 ] ) ( pushed to the fourth cumulant , i.e. the above formula ) and of ( [ power1 ] ) using ( [ simpli ] ) , implies : @xmath750 thus we need compute the terms with one and two free replica sums , @xmath751 and @xmath752 . because of ( [ p ] ) it means that we need to compute @xmath753 , @xmath754 , @xmath755 and @xmath756 . to compute them we use the definitions ( [ chidef ] ) , ( [ defoff ] ) .
we thus need powers of the matrix @xmath111 , but only terms with zero , one or two replica sums .
the expression of @xmath757 given in ( [ power2 ] ) is thus sufficient , and we also need : @xmath758\ ] ] dropping all terms with three or more sums .
one then finds : @xmath759\!\
! \nonumber \\ & & + \frac{8}{t^2 } i_4 \sum_{ef } ( \tilde{b}'_{ae } \tilde{b}'_{be } ( \tilde{b}'_{af } + \tilde{b}'_{bf } + \tilde{b}'_{ef } ) - \tilde{b}'_{ae } \tilde{b}'_{bf } \tilde{b}'_{ef } - \tilde{b}'_{ab } ( \tilde{b}'_{af } \tilde{b}'_{ae } + \tilde{b}'_{bf } \tilde{b}'_{be } + \tilde{b}'_{ae } \tilde{b}'_{bf } ) \ .\end{aligned}\ ] ] which yields @xmath751 and @xmath752 using ( [ p ] ) . to obtain the third cumulant we now insert @xmath751 in ( [ eqcum2 ] ) .
one can rewrite @xmath760 , indeed taking the derivative of ( [ saddlepointforb ] ) with respect to @xmath761 shows that : @xmath762 this becomes , regrouping the terms in @xmath763 and dividing by the common denominator @xmath764 we obtain : @xmath765 \ .\nonumber \\ & & \end{aligned}\ ] ] this first yields : @xmath766inserting this back yields : @xmath767\ .\end{aligned}\ ] ] in terms of functions it gives : @xmath768\ .\end{aligned}\ ] ] integrating once , this yields the simple expression for the third cumulant given in the text .
note that , up to terms which vanish at @xmath769 , it can be expressed only in terms of the function @xmath770 . ) , in the same order.,title="fig:",scaledwidth=50.0% ] from ( [ eqcum2 ] ) one has : @xmath771 \ , \qquad\end{aligned}\ ] ] where we have used @xmath772 obtained by further differentiation of ( [ saddlepointforb ] ) with respect to @xmath761 .
we also define @xmath773 + 2 i_2 t ( \tilde{b}'_{ae } + \tilde{b}'_{be } ) { \nonumber}\\ & & + 8 i_3 \big [ - \frac{1}{2 } ( \tilde{b}'_{aa } + \tilde{b}'_{bb } ) ( \tilde{b}'_{ae } + \tilde{b}'_{be } ) + \frac{1}{2 } ( \tilde{b}'_{ae } \tilde{b}'_{ae } + \tilde{b}'_{be } \tilde{b}'_{be } ) + \tilde{b}'_{ab } ( \tilde{b}'_{ae } + \tilde{b}'_{be } ) - \tilde{b}'_{ae } \tilde{b}'_{be } \big]\ , \qquad \end{aligned}\ ] ] and @xmath774 + 2 i_2 t ( \tilde{s}'_{1,aef } + \tilde{s}'_{1,bef } ) + 4 i_3 t ( \tilde{b}'_{ae } \tilde{b}'_{af } + \tilde{b}'_{be } \tilde{b}'_{bf } ) { \nonumber}\\ & & + 8 i_3 \big [ - ( \tilde{b}'_{aa } \tilde{s}'_{1,aef } + \tilde{b}'_{bb } \tilde{s}'_{1,bef } ) - ( \tilde{b}'_{ae } \tilde{s}'_{1,aaf } + \tilde{b}'_{be } \tilde{s}'_{1,bbf } ) + ( \tilde{b}'_{af } \tilde{s}'_{1,fae } + \tilde{b}'_{bf } \tilde{s}'_{1,fbe } ) { \nonumber}\\ & & \qquad\quad + \tilde{b}'_{ab } ( \tilde{s}'_{1,bef } + \tilde{s}'_{1,aef } ) + ( \tilde{b}'_{ae } \tilde{s}'_{1,abf } + \tilde{b}'_{be } \tilde{s}'_{1,abf } ) - ( \tilde{b}'_{af } \tilde{s}'_{1,fbe } + \tilde{b}'_{bf } \tilde{s}'_{1,fae})\big ] { \nonumber}\\ & & + 8 i_4 \big [ \tilde{b}'_{ae } \tilde{b}'_{ae } ( 2 \tilde{b}'_{af } + \tilde{b}'_{ef } ) + \tilde{b}'_{be } \tilde{b}'_{be } ( 2 \tilde{b}'_{bf } + \tilde{b}'_{ef } ) - ( \tilde{b}'_{ae } \tilde{b}'_{af } + \tilde{b}'_{be } \tilde{b}'_{bf } ) \tilde{b}'_{ef } - 3 \tilde{b}'_{aa } ( \tilde{b}'_{af } \tilde{b}'_{ae } + \tilde{b}'_{bf } \tilde{b}'_{be } ) { \nonumber}\\ & & \qquad \quad - 2 \tilde{b}'_{ae } \tilde{b}'_{be } ( \tilde{b}'_{af } + \tilde{b}'_{bf } + \tilde{b}'_{ef } ) + 2 \tilde{b}'_{ae } \tilde{b}'_{bf } \tilde{b}'_{ef } + 2 \tilde{b}'_{ab } ( \tilde{b}'_{af } \tilde{b}'_{ae } + \tilde{b}'_{bf } \tilde{b}'_{be } + \tilde{b}'_{ae } \tilde{b}'_{bf } ) \big ] { \nonumber}\ , \end{aligned}\ ] ] which should be further symmetrized over @xmath775 for later use ( indicated by @xmath776 below ) . the fourth cumulant equation yields , regrouping and simplifying the denominators ( and using also the third cumulant equation ) : @xmath777 { \nonumber}\\ & & + 4 i_3 t \tilde{b}''_{ab } ( \tilde{b}'_{ae } \tilde{b}'_{af } + \tilde{b}'_{be } \tilde{b}'_{bf } ) + 8 i_3 \tilde{b}''_{ab } \text{sym}_{ef } \big [ - ( \tilde{b}'_{aa } \tilde{s}'_{1,aef } + \tilde{b}'_{bb } \tilde{s}'_{1,bef } ) - ( \tilde{b}'_{ae } \tilde{s}'_{1,aaf } + \tilde{b}'_{be } \tilde{s}'_{1,bbf } )
{ \nonumber}\\ & & + ( \tilde{b}'_{af } \tilde{s}'_{1,fae } + \tilde{b}'_{bf } \tilde{s}'_{1,fbe } ) + \tilde{b}'_{ab } ( \tilde{s}'_{1,bef } + \tilde{s}'_{1,aef } ) + ( \tilde{b}'_{ae } \tilde{s}'_{1,abf } + \tilde{b}'_{be } \tilde{s}'_{1,abf } ) - ( \tilde{b}'_{af } \tilde{s}'_{1,fbe } + \tilde{b}'_{bf } \tilde{s}'_{1,fae } ) \big ] { \nonumber}\\ & & + 8 i_4 \tilde{b}''_{ab } \text{sym}_{ef } \big [ \tilde{b}'_{ae } \tilde{b}'_{ae } ( 2 \tilde{b}'_{af } + \tilde{b}'_{ef } ) + \tilde{b}'_{be } \tilde{b}'_{be } ( 2 \tilde{b}'_{bf } + \tilde{b}'_{ef } ) - ( \tilde{b}'_{ae } \tilde{b}'_{af } + \tilde{b}'_{be } \tilde{b}'_{bf } ) \tilde{b}'_{ef } - 3 \tilde{b}'_{aa } ( \tilde{b}'_{af } \tilde{b}'_{ae } + \tilde{b}'_{bf } \tilde{b}'_{be } ) { \nonumber}\\ & & - 2 \tilde{b}'_{ae } \tilde{b}'_{be } ( \tilde{b}'_{af } + \tilde{b}'_{bf } + \tilde{b}'_{ef } ) + 2 \tilde{b}'_{ae } \tilde{b}'_{bf } \tilde{b}'_{ef } + 2 \tilde{b}'_{ab } ( \tilde{b}'_{af } \tilde{b}'_{ae } + \tilde{b}'_{bf } \tilde{b}'_{be } + \tilde{b}'_{ae } \tilde{b}'_{bf } ) \big ] .\end{aligned}\ ] ] setting @xmath778 and solving one finds : @xmath779 \tilde b'_{ae } \tilde b'_{af}\ ] ] this gives the final result for the fourth cumulant : @xmath780 i_{4}\nonumber \\
+ 192\!\!&\big[&\ ! 4(\tilde b'_0)^2\tilde b'_{ab}\tilde b'_{cd}{\tilde b''}_{ac}-4\tilde b'_0(\tilde b'_{ab})^2\tilde b'_{cd}{\tilde b''}_{ac}-8\tilde b'_0\tilde b'_{ab}\tilde b'_{ac}\tilde b'_{cd}{\tilde b''}_{ac}+4(\tilde b'_{ab})^2\tilde b'_{ac}\tilde b'_{cd}{\tilde b''}_{ac}+4\tilde b'_{ab}(\tilde b'_{ac})^2\tilde b'_{cd}{\tilde b''}_{ac}\nonumber \\ & & + 8\tilde b'_0\tilde b'_{ab}\tilde b'_{bc}\tilde b'_{cd}{\tilde b''}_{ac}-8\tilde b'_0\tilde b'_{ac}\tilde b'_{bc}\tilde b'_{cd}{\tilde b''}_{ac}-8\tilde b'_{ab}\tilde b'_{ac}\tilde b'_{bc}\tilde b'_{cd}{\tilde b''}_{ac}+4(\tilde b'_{ac})^2\tilde b'_{bc}\tilde b'_{cd}{\tilde b''}_{ac}+4(\tilde b'_{ab})^2\tilde b'_{ac}\tilde b'_{ad}{\tilde b''}_{ad}\nonumber \\ & & -4(\tilde b'_{ab})^2\tilde b'_{ac}\tilde b'_{cd}{\tilde b''}_{ad}+2\tilde b'_{ab}\tilde b'_{ac}\tilde b'_{bd}\tilde b'_{cd}{\tilde b''}_{ad}+(\tilde b'_{ab})^2(\tilde b'_{cd})^2{\tilde b''}_{ad}+4(\tilde b'_0)^2\tilde b'_{ac}\tilde b'_{cd}{\tilde b''}_{bc}-4\tilde b'_0(\tilde b'_{ac})^2\tilde b'_{cd}{\tilde b''}_{bc}\nonumber \\ & & + ( \tilde b'_{ad})^2(\tilde b'_{cd})^2{\tilde b''}_{bd}\big ] i_{3}^{2}\nonumber \\ + 192 t \!\!&\big[&\
! 2\tilde b'_{ab}\tilde b'_{ac}\tilde b'_{ad}\tilde b''_{ab}+2\tilde b'_{ab}\tilde b'_{ad}\tilde b'_{bc}\tilde b''_{ab}-2\tilde b'_0\tilde b'_{ab}\tilde b'_{ad}\tilde b''_{ac}+(\tilde b'_{ab})^2\tilde b'_{ad}\tilde b''_{ac}-2\tilde b'_{ab}\tilde b'_{ad}\tilde b'_{bc}\tilde b''_{ac}-2\tilde b'_0\tilde b'_{ab}\tilde b'_{cd}\tilde b''_{ac}\nonumber \\ & & + ( \tilde b'_{ab})^2\tilde b'_{cd}\tilde b''_{ac } -2\tilde b'_{ab}\tilde b'_{ac}\tilde b'_{ad}\tilde b''_{0}\big]\frac { i_{2}i_{3}}{1 + 4 i_{2}\tilde b''_{0 } } \nonumber \\ + 32 t \!\!&\big[\ ! & \tilde b'_{ab}\tilde b'_{ac}\tilde b'_{ad}\big ] \frac{i_{3}}{1 + 4 i_{2}\tilde b''_{0 } } \nonumber
\\ + 48 t^{2 } \!\!&\big[&\ ! \tilde
b'_{ab}\tilde b'_{ad}\tilde b''_{ac}+\tilde b'_{ab}\tilde b'_{cd}\tilde b''_{ac } \big ] \frac{i_{2}^{2}}{(1 + 4 i_{2}\tilde b''_{0})^{2}}\nonumber \\ -128 t^{2 } \!\ ! & \big [ & \ ! \tilde b'_{ac}\tilde b'_{bc}\tilde b'_{cd}\tilde b'''_{0}\big ] \frac{i_2 ^ 3}{(1 + 4 i_{2}\tilde b''_{0})^{3 } } \bigg \ } \label{final4th}\end{aligned}\ ] ] where @xmath781 denotes @xmath782 times the sum of all 24 permutations of the indices @xmath783 , and we note @xmath784 , @xmath785 and @xmath786 .
note that all terms containing @xmath638 can be eliminated by the redefinition @xmath787 .
the cumulant expansion can be generalized to study the effective action for non - uniform configurations .
the functional @xmath788 $ ] is a functional of the field @xmath789 and can be expanded as : @xmath790 = \frac{- 1}{2 t^2 } \sum_{ab } \tilde{b}[v_{ab}^2 ] - \frac{1}{3 !
t^3 } \sum_{abc } \tilde{s}[v_{ab}^2,v_{bc}^2,v_{ca}^2 ] + \dots \ .\ ] ] the self - consistent equation ( [ scfunctional ] ) then yields , by a similar expansion in numbers of sums : @xmath791}{\delta ( v_a(x ) \cdot v_b(x ) ) } = b'\bigg ( v_{ab}^2(x ) + 2 t i_1 \nonumber \\ & \qquad \quad~ + 4 \int_y c_{xy}^2 \big[\frac{\delta \tilde b[v\cdot v]}{\delta ( v_a(y ) \cdot v_b(y ) ) } - \frac{\delta \tilde b[v\cdot v]}{\delta ( v_a(y ) \cdot v_a(y ) ) } \big]\bigg)\end{aligned}\ ] ]
@xmath792@xmath793 @xmath794 to refine the analysis and study the behavior for all values of @xmath45 , let us rewrite from ( [ secder0 ] ) the condition that @xmath348 remains analytic for all @xmath45 .
it reads : @xmath795 where we have also inserted the value of the inverse function , with @xmath796 , for the power law models .
this condition is equivalent to the vanishing of the replicon , i.e. it is the line where the rs solution of the gvm becomes unstable to rsb .
one can then plot ( see fig .
[ f : supercurve ] ) the function @xmath797 for the three cases defined in the text , lr @xmath792 , marginal @xmath793 and sr @xmath794 where @xmath798 . the lr and marginal cases , which correspond to continuous and one step marginal rsb solutions , have been discussed in the text and there the frg gives back exactly @xmath614 of the mp solution .
we defer the detailed study of the sr case to further work , and give here a few general remarks .
first one notices on fig .
[ f : supercurve ] that solving the frg equation decreasing the mass from infinity one first has the analytic solution which coincides with the rs one .
for @xmath799 it remains analytic down to @xmath486 . for @xmath800 a cusp
arises when the left branch of the line is reached .
thus despite the reentrance of the analytic solution at small @xmath45 , freezing of the frg solution has already occurred and it is clearly important to understand how to extend the frg in the shaded region . on the other hand , a 1-step solution of the mp saddle point equations exists , obtained by varying the free energy w.r.t .
its precise boundary depends on the model , but it is generally contained within the shaded region ( for details see @xcite ) .
an intriguing property of the gvm is the simultaneous existence , within the rightmost portion of the shaded region near the axis @xmath486 , of two locally stable solutions , one rs and the other 1-step rsb .
thus , although the line in fig .
[ f : supercurve ] is the locus of a continuous transition from rs to rsb , in this rightmost portion of the @xmath802 diagram , the 1-step non - marginal rsb solution appears discontinuously , before the line is reached as @xmath44 is lowered .
work is in progress to make contact between frg and rsb in this sr case , and in particular to understand whether there are also two branches of solutions of the frg equation in that region .
from the rg point of view it is interesting to search for finite temperature fixed points ( fp ) of the frg equation , especially in view of future extensions to finite @xmath1 ( since we know at least in some cases , that these persist at finite @xmath1 ) .
it is convenient to use eq.([linear ] ) .
these fps exist only in the marginal case @xmath361 , i.e. @xmath720 for @xmath327 or @xmath416 for @xmath31 , so that @xmath803 does not flow .
this is the case studied here .
following the same steps as in section [ frgapproach ] , the general solution of the fixed - point condition @xmath804 in ( [ linear ] ) for a fixed value of @xmath805 , imposing @xmath394 is @xmath806\ .\end{aligned}\ ] ] taking a derivative at @xmath408 , we obtain a self - consistency condition for @xmath807 .
one solution is @xmath808 , the `` zero - temperature '' fixed point discussed in section [ frgapproach ] .
the other one is @xmath809 with the condition that it must be positive .
reinserting in ( [ fp1 ] ) we obtain the final form for the finite - temperature fixed point : @xmath810 \left[\left(\frac{y_0}{y}\right)^{\!\frac{2 \zeta}{\epsilon - 2 \zeta } } -1\right]\qquad \nonumber \\ & = & \frac{y - y_{0}}{\epsilon } + \frac{t_{m}}{2\zeta } \left[\left(\frac{y_{0}}{y } \right)^{\!\frac{2 \zeta}{\epsilon - 2 \zeta}}-1 \right]\end{aligned}\ ] ] the term in the second line of ( [ lf95 ] ) was not present in the `` zero - temperature '' fixed - point solution ( [ fixedpoints ] ) .
note that it does not vanish at @xmath16 but at the higher temperature @xmath483 such that @xmath811 . at this point ,
also @xmath807 vanishes and the solution @xmath812 becomes non - analytic .
the fixed - point analysis alone does neither fix the value of @xmath387 , nor @xmath335 .
however we can now explicitly check that this fixed - point solution identifies with the analytic solution ( [ solu1 ] ) , ( [ solu2 ] ) when setting @xmath117 , using @xmath813 and @xmath814 .
this identification works only for @xmath815 , and since @xmath387 is now fixed by ( [ solu2 ] ) , we can compute @xmath335 and find that it is given by ( [ lf52 ] ) . below @xmath335
the solution freezes at @xmath459 at the zero temperature fixed point .
let us now solve the fixed point condition @xmath817 in ( [ linear ] ) for @xmath818 , imposing @xmath394 .
one finds : @xmath819 determining @xmath807 again one finds , in addition to the solution @xmath808 @xmath820 reinserting one finds finally the finite - temperature fixed points @xmath821 with @xmath498 and @xmath822 . there is thus a line of fixed points with @xmath416 in @xmath31 , parameterized by temperature , @xmath387 being again undetermined . to compare with the solution of the flow equation
, we obviously need to consider a broader class of sr models with @xmath823 .
the solution is then : @xmath824 with @xmath498 . for small disorder @xmath825 , and @xmath826 , @xmath583 flows to zero as @xmath117 and
the solution remains analytic . for @xmath487
the solution develops a cusp when @xmath387 reaches @xmath827 , i.e. at the larkin mass : @xmath828 thus only for @xmath483 the solution reaches for @xmath117 an analytic finite-@xmath44 fixed point associated with @xmath416 , of the form ( [ lf99 ] ) .
thus in @xmath31 the line of finite - temperature fixed points with @xmath416 corresponds to the line of critical fixed points as the parameter @xmath210 is varied .
in this appendix , we derive simple rg equations for the mp solution in the full rsb case .
this gives a more direct derivation of the key - equations ( [ rgrsb1 ] ) and ( [ rgrsb2 ] ) .
we start from ( see ( [ sadsig ] ) and ( [ sadsig2 ] ) , equivalent to ( 5.4 ) of mp ) : @xmath829 taking a derivative with respect to @xmath10 yields @xmath830(u))^2}\ ] ] for all @xmath10 such that @xmath831 . using that due to ( see ( [ sadsig2 ] ) , equivalent to ( 5.2 ) of mp ) : @xmath832 ( u))}\nonumber \\ & & - \int_{u}^{1}\frac{{{\mathrm{d}}}v}{v^2 } \frac{1}{k^{2}+{m^2 } + [ \sigma ] ( v)}\end{aligned}\ ] ] we have @xmath833(u))^2 } \ .\ ] ] on the other hand , we can take a derivative of ( [ mp(5.4 ) ] ) with respect to @xmath834 : @xmath835 eliminating @xmath836 one finds : @xmath837(u))^2 } \nonumber \\ & & \qquad = -\int_{k } \partial_{m^2 } ( \tilde{g}(k ) - g(k , u ) ) \ .\qquad \end{aligned}\ ] ] taking another derivative with respect to @xmath10 after using ( [ 6.15 ] ) gives @xmath838 ( u))^{2}}= \partial_{{m^2}}\ ! \int_{k } \frac{\sigma ' ( u)}{(k^{2}+{m^2 } + [ \sigma ] ( u))^{2 } } \ .\ ] ] noting that the derivatives of the numerator cancel , we get @xmath839 ( u))^{3}}= \int_{k } \frac{\sigma ' ( u ) ( 1+\partial _ { { m^2 } } [ \sigma ] ( u ) ) } { ( k^{2}+{m^2 } + [ \sigma ] ( u))^{3 } } \ .\ ] ] since for all @xmath10 , @xmath840 ( u))^{-3}}\neq 0 $ ] and by assumption @xmath841 , rsb reveals its universality in the simple relation : @xmath842(u ) \ , \ ] ] which , upon another derivation , yields the two `` rg equations '' : @xmath843 ( u ) \right ) & = & 0 \\ { m^2 } \frac{{{\mathrm{d}}}}{{{\mathrm{d}}}{m^2 } } \sigma ( u)&=&0 \ .\end{aligned}\ ] ]
) , describing the approach to the fixed point in presence of an additional bare power law tail.,title="fig : " ] since we have found the solution of the frg equation for arbitrary disorder correlations , it is instructive to study the convergence to the frg fixed point in the case where the inital disorder is not of the simple form ( [ corrlr ] ) on an explicit example .
we start from a disorder correlator which is the superposition of two power laws : @xmath844 with @xmath845 , s.t . for large @xmath846
the first term dominates and fixes the exponent @xmath847 .
we will determine the inverse function @xmath848 only for @xmath849 small .
one finds @xmath850 with @xmath851 , and we have defined @xmath852 we have chosen @xmath853 to simplify all prefactors . inserting into ( [ cuspcond ] )
, we obtain : @xmath854 by multiplying with @xmath855 and @xmath856 , we obtain the equivalent formula @xmath857 the left - hand side is plotted on figure [ nonleadingcurve ] ; note that always the first exponent is negative , and the second and third are positive .
thus the solution for @xmath858 or @xmath859 is simply @xmath860 . for non - vanishing @xmath861 and @xmath45
, the solution can be obtained graphically as the intersection of @xmath862 with @xmath863 .
note that there is a solution for any @xmath861 . for @xmath864 ,
the approach to @xmath865 is obtained by linearizing @xmath866 , with the result @xmath867
in this section we recall the corresponding result for the effective action of the generic pure @xmath49 model at large @xmath1 .
one mechanism by which the effective action may becomes non - analytic is given on standard example of @xmath50 .
the generic @xmath49 model in dimension @xmath0 is defined by the action : @xmath868\ ] ] here @xmath45 is used as a parameter , the bare mass being @xmath869 . for a uniform mode
one has @xmath24= l^{d } \tilde{\gamma}[v]$ ] in terms of the rescaled field @xmath870 .
one defines : @xmath871 = \frac{1}{2 } m^2 v^2 + \tilde v(v^2 ) = \tilde w(v^2)\ .\ ] ] similarly , one defines @xmath872 whether absorbing or not the mass into the ( bare or renormalized ) potential . again , for @xmath99 , one has @xmath873 . the same method as in section ( [ sec : calculgamma ] ) yields the saddle point equations for infinite @xmath1 : @xmath874 more details , a graphical derivation , and the @xmath12 expansion are given in @xcite .
a condition for the stability of the theory is that : @xmath875 for the @xmath50 theory @xmath882 this reads : @xmath883 for @xmath884 with @xmath394 . at @xmath16 , as @xmath45 is decreased , there is a transition in any @xmath0 when @xmath885 vanishes . for @xmath363 ,
the transition persists for @xmath495 , and occurs when @xmath886 vanishes , with the standard result : @xmath887 which depends strongly on the uv - cutoff @xmath46 .
@xmath387 vanishes at the transition , and in the ordered phase the effective action has a non - analytic form .
in addition of the branch ( [ phi4 ] ) for @xmath888 , @xmath889 , the function @xmath375 has a branch @xmath890 for @xmath891 , where @xmath892 is the order parameter .
exactly at @xmath483 we should recover that the effective action exhibits the standard power - law non - analyticity @xmath893 = m^{1 + \delta}$ ] .
indeed , from the self - consistent solution ( [ phi4 ] ) , subtracting the same with @xmath894 and @xmath890 one gets @xmath895 for @xmath896 .
this corresponds to @xmath897 .
it can be recovered by solving the frg equation .
one can look for fixed - point solutions of the form if one wants the two first terms to dominate and to scale in the same way , one needs @xmath899 and @xmath900 .
for all three terms to scale the same way one needs @xmath901 , @xmath902 . inserting ( [ lf103 ] ) into ( [ lf102 ] ) yields @xmath903 again this can be transformed into a linear rg equation for @xmath904 : @xmath905 the solutions of the above equation with @xmath901 , @xmath902 are : @xmath906 a particular solution is @xmath907 in the limit of zero mass this yields @xmath908 .
although the mechanism for the disordered systems studied in the main text seems to be different from @xmath50-models , it raises the question of the meaning of the non - analyticity in the disordered problem .
is it the signature that we are dealing with a glass phase , where a symmetry has been broken ?
we know that for infinite @xmath1 , this is also accompanied by rsb , but this does not have to be so in general , i.e. the cusp can arise without rsb , just from localization ( single ground state dominance ) . | we study the replica field theory which describes the pinning of elastic manifolds of arbitrary internal dimension @xmath0 in a random potential , with the aim of bridging the gap between mean field and renormalization theory .
the full effective action is computed exactly in the limit of large embedding space dimension @xmath1 .
the second cumulant of the renormalized disorder obeys a closed self - consistent equation .
it is used to derive a functional renormalization group ( frg ) equation valid in any dimension @xmath0 , which correctly matches the balents fisher result to first order in @xmath2 .
we analyze in detail the solutions of the large-@xmath1 frg for both long - range and short - range disorder , at zero and finite temperature .
we find consistent agreement with the results of mezard parisi ( mp ) from the gaussian variational method ( gvm ) in the case where full replica symmetry breaking ( rsb ) holds there .
we prove that the cusplike non - analyticity in the large @xmath1 frg appears at a finite scale , corresponding to the instability of the replica symmetric solution of mp .
we show that the frg exactly reproduces , for any disorder correlator and with no need to invoke parisi s spontaneous rsb , the non - trivial result of the gvm for small overlap .
a formula is found yielding the complete rsb solution for all overlaps .
since our saddle - point equations for the effective action contain both the mp equations and the frg , it can be used to describe the crossover from frg to rsb .
a qualitative analysis of this crossover is given , as well as a comparison with previous attempts to relate frg to gvm .
finally , we discuss applications to other problems and new perspectives . |
the consideration of relativistic effects is essential to the proper understanding of the chemistry of heavy - element containing molecules @xcite .
the dirac equation provides the relativistic quantum mechanical description of a single electron in the presence of external electromagnetic potentials .
as the dirac operator comprises four - dimensional operators , its eigenfunctions possess four components .
the coulomb interaction turned out to be sufficiently accurate for the description of chemical phenomena and the corresponding many - electron hamiltonian is called the dirac coulomb hamiltonian . a relativistic method based on the dirac coulomb hamiltonian
is called a four - component method .
it yields negative - energy solutions , which are pathologic and of no use in chemistry .
also , a large number of basis functions is required to properly describe the negative - energy states .
these are two drawbacks of four - component methods that motivated the development of two - component methods , which remove the negative energy solutions and can , in principle , exactly reproduce the results of four - component calculations @xcite .
several relativistic two - component methods were developed in the past decades .
one of the widely used approaches is the second - order douglas
hess method ( dkh2 ) @xcite .
it employs the free - particle foldy wouthuysen ( fw ) @xcite transformation as well as sequential douglas
kroll @xcite transformations to decouple the four - component operator .
higher - order @xcite and even arbitrary - order @xcite dkh methods have been developed .
the zeroth - order regular approximation ( zora ) @xcite is another highly successful relativistic two - component method . within the zora framework
, it is particularly easy to implement the calculation of molecular properties ; see refs .
@xcite for very recent examples and ref .
@xcite for a review . in recent years
it was shown that , if the goal is to exactly decouple the four - component hamiltonian matrix instead of the hamiltonian operator , the formulae of exact decoupling can be directly obtained .
the barysz sadlej
snijders ( bss ) @xcite method aims at exact decoupling of the free - particle fw - transformed four - component hamiltonian matrix by a matrix operator of the form derived in ref .
the key matrix operator , which is used to construct the decoupling transformation matrix , is obtained by solving an iterative equation . however , invoking the free - particle fw transformation turns out to be not necessary for the construction of the exact decoupling transformation in a one - step protocol .
the pioneering work of this one - step transformation was provided by dyall @xcite in the form of the so - called normalized elimination of the small component ( nesc ) approach .
later it was generalized to the so - called exact two - component ( x2c ) method by several groups @xcite . since the dkh method also employs matrix techniques to evaluate the douglas
kroll transformation , it is also able to exactly decouple the four - component hamiltonian matrix .
this has been shown within the arbitrary - order approach @xcite . for reviews of these developments
@xcite . even within a scalar relativistic approximation , where the spin - dependent operators are neglected
, the construction of the relativistic hamiltonian matrix dominates the step of the one - electron integral evaluation .
this is the case because the relativistic transformation is done within an un - contracted basis set and involves many matrix manipulations such as multiplication and diagonalization . for a large molecule including only one or a few heavy atoms , this procedure would waste computational resources since the relativistic effects are highly local and only significant for heavy atoms .
it is therefore clear that a local relativistic method is desirable .
even for a molecule containing only heavy atoms , such a local relativistic approach would be preferable as we shall see .
the relativistic hamiltonian operator is universal in coordinate space and a local approach is unfeasible at the operator level .
however , since most quantum chemical calculations apply a linear combination of atom - centered basis functions , we may employ these basis functions to construct atomic projectors as it is done in charge and spin population analysis @xcite .
for those hamiltonian matrix blocks for which relativistic corrections are important , we shall derive a local relativistic hamiltonian matrix to evaluate it .
the locality is then exploited in the basis - function space instead of coordinate space .
a crucial aspect to determine will be which block of the hamiltonian matrix requires a relativistic description and how to evaluate it .
undoubtedly , the heavy atom diagonal blocks will require a relativistic description , while the relativistic description of heavy atom off - diagonal blocks depends on their contribution to physical observables and on the feasibility of applying the relativistic description . for relativistic transformation techniques , such as dkh and x2c ,
it is impossible to apply the relativistic transformation to an off - diagonal block alone without information about other blocks .
it requires the information of a full square matrix to construct the relativistic transformation .
such a difficulty disappears if one employs an operator - based relativistic method such as zora
. the off - diagonal block of the hamiltonian matrix can be directly computed from relativistic integrals .
however , the zora method provides a poor description of atomic inner - shell orbitals , which carry the largest part of the relativistic effect . in this sense ,
the exact decoupling methods are the best candidate for applying a local relativistic scheme since they do not neglect important relativistic effects .
all current exact decoupling methods employ the transformation technique .
it is therefore not trivial to apply the relativistic description to heavy - atom off - diagonal hamiltonian matrix elements .
however , in this work we propose that if the local approximation is applied to the decoupling transformation instead of the hamiltonian matrix , one could obtain results which include the relativistic description to heavy atom off - diagonal blocks .
we also note that the main effort for the construction of the relativistic hamiltonian matrix with exact - decoupling methods is due to the evaluation of the decoupling transformation matrices .
therefore , a local approximation to the decoupling transformation would also lead to a significant reduction of the computational cost .
the following notation will be used throughout this paper .
upper case labels @xmath1 denote heavy atoms which require a relativistic description , while lower case labels @xmath2 denote light atoms for which a ( standard ) nonrelativistic description can be assumed to be accurate enough based on numerical evidence compiled within computational chemistry in the past decades . to be more precise ,
the former labels refer to nuclei with high nuclear charge numbers , while the latter refer to those with low nuclear charge numbers ( typically with nuclear charges of less than two dozen protons ) .
of course , the distinction whether a nucleus is considered heavy or light bears some sort of arbitrariness and it will be made on the basis of the accuracy required in a quantum chemical calculation . in general , there are then five different types of hamiltonian matrix blocks @xmath3 , @xmath4 , @xmath5 , @xmath6 , and @xmath7 that need to be considered . the evaluation of the relativistic hamiltonian matrix with exact - decoupling approaches requires a set of square matrices .
they are the nonrelativistic kinetic energy matrix @xmath8 , @xmath9 the external potential matrix @xmath10 , @xmath11 and the overlap matrix @xmath12 , @xmath13 as well as an additional relativistic matrix @xmath14 @xmath15 denotes the @xmath16-th 2-spinor atom - centered basis function , @xmath17 the external potential , @xmath18 the speed of light , @xmath19 the vector of pauli spin matrices . the basis functions @xmath20 may be grouped according to the atomic nucleus to which they are assigned , which is key to the local approaches discussed in the following .
the matrix representation of the relativistic hamiltonian is then evaluated as a function of the above - mentioned matrices @xmath21 a straightforward idea for a trivial local approximation is to apply the evaluation of the relativistic hamiltonian matrix only to atomic ( i.e. , diagonal ) blocks @xmath22 ( instead of to the full matrix , but where @xmath23 may also represent a group of atoms ) and to ignore the relativistic correction to all off - diagonal blocks @xmath24 adding more atoms to a group @xmath23 will improve the accuracy , but the computational advances will be lost when a group @xmath23 becomes very large .
this approach was employed in ref .
@xcite and explored in detail in ref .
@xcite for the low - order dkh method .
we refer to it as the _ diagonal local approximation to the hamiltonian _ ( dlh ) matrix .
it is clear that the dlh approximation can also be applied to relativistic exact - decoupling approaches .
the dlh approximation works well at large interatomic distances , but the difference to a reference energy increases with shorter distance .
however , relativistic corrections to off - diagonal blocks are also important especially when inter - atomic distances are short .
the necessity of applying the relativistic description to off - diagonal blocks has already been discussed in ref .
@xcite with a so - called two - center approximation . in ref .
@xcite , the dkh transformation was applied to pairs of atoms ( @xmath25 ) @xmath26 with @xmath27 to obtain a relativistically corrected off - diagonal hamiltonian matrix @xmath4 .
unfortunately , such a two - center approximation introduces an inconsistent treatment to diagonal blocks .
for instance , @xmath3 can be obtained from either the pair ( @xmath25 ) or from another pair ( @xmath28 ) .
so far , we have only discussed the @xmath29 and @xmath30 blocks which , by definition , require a relativistic description for both atoms . since relativistic corrections for light atoms
are very small and may be neglected , @xmath6 and @xmath7 can be approximated by the nonrelativistic hamiltonian matrices @xmath31 a relativistic scheme for the heavy
light hybrid off - diagonal blocks @xmath5 should , however , be considered explicitly .
the general expression of exact decoupling transformations can be written as @xmath32 @xmath33 and @xmath34 are two - component unitary operators , and only @xmath33 is required for the evaluation of the electrons - only hamiltonian .
the @xmath35-operator generates the electronic small - component functions @xmath36 from the large - component functions @xmath37 @xmath38 two - dimensional hamiltonians are then obtained by applying the unitary transformation of eq .
( [ general_ur2c ] ) to blockdiagonalize ( @xmath39 ) the four - dimensional dirac - based hamiltonian @xmath40 @xmath41 it yields both the electrons - only hamiltonian @xmath42 ( as the upper left block in @xmath43 ) and the one for negative - energy solutions , although the latter will be discarded . as discussed by dyall in ref .
@xcite for the nesc method , the breit - pauli approximation to the @xmath35 matrix , which is used to construct the relativistic transformation matrix , is @xmath44 where @xmath45 denotes the identity matrix .
such an approximation is not variationally stable and thus can not be used for variational calculations .
however , one could employ this approximation for light atoms only @xmath46 since the corrections to light atoms are small and may not affect variational calculations .
this idea was suggested in ref .
@xcite for the x2c method but no results were presented .
although the lowest level approximation to @xmath35 eq . ( [ xeqiaprox ] ) did not provide useful results , the atomic approximation to the @xmath35 matrix gave results with very small errors .
as discussed by dyall in ref .
@xcite , the @xmath35 matrix can be approximated as the direct sum of atomic blocks @xmath47 this approximation resembles the linear combination of atomic four - spinors ansatz in four - component calculations .
the same approach was also employed in ref .
@xcite in the sense of an atomic approximation to the projection on electronic states .
if the electronic states are further transformed to a two - component picture by a renormalization matrix , this gives rise to a local x2c method .
as discussed in refs .
@xcite , the local approximation to the @xmath35 matrix eq .
( [ localxmat ] ) works well for spectroscopic constants of diatomic molecules
. however , there are some drawbacks of the local-@xmath35 matrix approximation to the x2c method .
firstly , it only reduces the cost for the evaluation of @xmath35 matrix , while the exact - decoupling transformation matrices are still of molecular dimension as the renormalization matrix is not atomic block diagonal even if the @xmath35 matrix is . the computational demands for the evaluation of the relativistic approximation are thus still tremendous for large molecules .
secondly , the @xmath48 approximation did not provide a satisfactory treatment of the @xmath49 blocks , since it may suffer from variational collapse and eqs .
( [ nrhaa ] ) and ( [ nrhab ] ) can not be fulfilled within this approximation .
in this article , we suggest an local approximation to the exact - decoupling transformation .
we may approximate the unitary transformation by taking only the atomic diagonal blocks ( all off - diagonal blocks are then set to zero ) @xmath50 where the atomic unitary transformations @xmath51 are obtained from the diagonalization of the corresponding @xmath52 blocks of matrix operator @xmath40 .
we denote this approximation to @xmath53 as the _ diagonal local approximation to the unitary decoupling transformation _ ( dlu ) . if we now substitute the approximate unitary transformation of eq .
( [ localu ] ) into eq .
( [ transu ] ) , the diagonal blocks , @xmath54 turn out to be the same as in the dlh approach , while the off - diagonal blocks then read @xmath55 which is to be compared with the expression @xmath56 where @xmath53 has not been approximated ( here , @xmath45 and @xmath57 run over all atomic blocks ) . hence , the dlu approach also introduces a relativistic description to off - diagonal interatomic blocks when compared with the dlh approximation .
the cost for the assembly of the unitary transformation @xmath53 within the dlu approach is then of order @xmath58 , where @xmath58 measures the system size , namely the number of atoms .
it is linear scaling since the atomic approximation is directly applied to the unitary transformation .
however , the next step of applying the unitary decoupling transformation to obtain the relativistic hamiltonian matrices is no longer linear scaling . if no local approximation was applied for the unitary decoupling transformation , according to eq .
( [ utranfull ] ) where the summation indices @xmath59 run over all atomic blocks , the calculation of @xmath4 matrix will require @xmath60 matrix multiplications .
since the number of matrices to be calculated is of order @xmath61 , the total cost of the relativistic transformation without local approximation is therefore of order @xmath62 .
if the dlu approximation is applied , no summation will be needed and only two matrix multiplications will be required for each heavy heavy block @xmath4 .
the total cost is then of order @xmath61 .
if the distance of two atoms @xmath23 and @xmath63 is sufficiently large , the relativistic description of @xmath4 can be neglected .
thus , we may define neighboring atomic pairs according to their distances .
then , only the hamiltonian matrix @xmath64 of neighboring pairs requires the transformation @xmath65 whereas all other pairs are simply taken in their nonrelativistic form . since the number of neighboring pairs is usually a linear function of system size .
the total cost is then reduced to order @xmath58 . to investigate the relativistic correction to hybrid @xmath5 blocks of the electrons - only hamiltonian @xmath42
, we need to consider the explicit structure of the unitary decoupling transformation , @xmath66 where only the electronic part of the exact - decoupling transformation represented by @xmath67 and @xmath68 [ which are the upper part of eq .
( [ general_ur2c ] ) ] is employed . from eq .
( [ utrans ] ) it is easy to see that the electrons - only hamiltonian reads @xmath69 in the nonrelativistic ( nr ) limit , where the speed of light approaches infinity , we have @xmath70 if we insert them into the relativistic electrons - only hamiltonian ( [ eonlyham ] ) , we arrive at the nonrelativistic hamiltonian matrix , @xmath71 this suggests a solution for the relativistic treatment of the hybrid heavy light blocks @xmath5 , because we may set all light - atom diagonal blocks in the unitary transformation to identity matrices @xmath72 to reproduce the nonrelativistic hamiltonian matrices of light - atom - only blocks , we must set the @xmath14 matrix of light atoms to zero , @xmath73 the @xmath74 matrices are also set to zero @xmath75 we denote this approximation for the treatment of hybrid blocks as the dlu(nr ) approach .
if we do not introduce any approximations to the @xmath14 matrices , the transformation will yield a hamiltonian that differs from the nonrelativistic hamiltonian matrix on the order of @xmath76 , which corresponds to the order achieved in the breit
pauli approximation .
we therefore denote it as the dlu(bp ) approach .
the final local approximation to the unitary matrices is then @xmath77 where the atomic unitary transformation matrices of heavy atoms are evaluated by the relativistic exact decoupling approaches .
the explicit expressions of hamiltonian matrix blocks @xmath3 and @xmath4 are @xmath78 if the dlu(nr ) approach is employed , the unitary transformation matrices of light atoms are simply set to identity matrices and the explicit form of @xmath5 , @xmath6 , and @xmath7 is @xmath79 in the dlu(bp ) approach , however , they read @xmath80 the relativistic transformation of a property operator @xmath81 @xcite requires an additional matrix @xmath82 and its matrix elements are given by @xmath83
the x2c method employs @xmath84 as the exact - decoupling unitary transformation .
the unitary operators for the electronic states are then @xmath85 for an actual implementation , we need their matrix representations in a finite non - orthogonal space of the atom - centered basis functions .
the @xmath35-matrix is evaluated by diagonalizing the following generalized eigenvalue equation @xmath86 where @xmath87 and @xmath88 denote the large- and small - component molecular - orbital coefficients , respectively .
we obtain the @xmath35 matrix by the following equation @xmath89 the matrix representation of @xmath67 and @xmath68 is then given by @xmath90 with @xmath91 defined as @xmath92 in our atomic approximation to the unitary transformation @xmath53 , we need to evaluate the above equation for each atomic block of the original hamiltonian in order to obtain the diagonal blocks @xmath93 and @xmath94 . the decoupling transformations are then assembled according to eqs .
( [ localul ] ) and ( [ localus ] ) and the relativistic hamiltonian matrix follows from eq .
( [ utrans ] ) . in the bss approach
, all matrices are transformed to an orthonormal basis - function space first .
it is convenient to calculate the transformation matrix @xmath95 by diagonalizing the nonrelativistic kinetic energy matrix @xmath96 since the eigenvalues @xmath97 can be used for later evaluation of the free - particle fw transformation .
the eigenvector matrix @xmath95 has the following properties : @xmath98 the free - particle fw ( fpfw ) transformation features four diagonal block matrices @xmath99 with @xmath100 and @xmath101 .
it is then applied to yield a transformed four - dimensional hamiltonian matrix @xmath102 @xmath103 next , the exact - decoupling bss transformation , @xmath104 which has the same structure as the exact - decoupling transformation @xmath105 , is applied .
the @xmath106 matrix is obtained by diagonalizing the free - particle fw transformed four - component hamiltonian matrix @xmath102 and employing a similar relation to the one in eq .
( [ x2cstep ] ) . after the exact - decoupling bss transformation has been carried out
, the hamiltonian matrix is back - transformed to the original ( non - orthogonal ) basis @xmath107 consequently , we obtain the matrix form of @xmath108 and @xmath109 for the bss approach as @xmath110 the expressions for the relativistic hamiltonian matrix ( and property matrices ) are then as given above .
the dkh approach features the same initial transformation as the bss approach to obtain the transformed hamiltonian matrix @xmath111 .
subsequent decoupling transformations are expressed as @xmath112 with the generalized parametrization of the @xmath113 @xcite , @xmath114}a_{k , i}\mathbf{w}_k^{i},\ ] ] where @xmath115 are anti - hermitian matrix operators and @xmath116 is related to the order of the dkh expansion . since the expression of @xmath115 is too lengthy to be presented here , we refer the reader to ref .
@xcite for details .
if @xmath116 is large ( strictly , if it approaches infinity ) , exact decoupling is achieved . usually , a low value for @xmath116 is often sufficient for calculations of relative energies and valence - shell properties ( i.e. , @xmath116=1 for the original dkh2 approach ) . with the complete decoupling dkh transformation written as @xmath117 the matrix forms of @xmath118 and @xmath119 are @xmath120
we should note that the finite - order dkh hamiltonian matrices are not the result of directly applying the decoupling transformation . to give consistent results , the transformation in eq .
( [ utrans ] ) is only used to obtain the off - diagonal blocks while the diagonal blocks are replaced by the traditional finite - order dkh hamiltonian . however ,
for high - order calculations , the differences of with / without replacing the diagonal blocks are very small .
as we have shown , there exists a systematic hierarchy of several levels for the local relativistic approximation to hamiltonian matrix elements . in this section ,
we study the accuracy of the different levels on a test molecule set and introduce the following notation to distinguish the results : 1 . * full * : full molecular relativistic transformation , no approximation .
* dlu * : diagonal local approximation to decoupling transformations , @xmath67 and @xmath68 are evaluated for each atomic ( diagonal ) block .
however , three variants are possible : * * dlu(all ) * : calculate relativistic transformation for all atoms ( all atoms are treated as heavy elements ) .
i.e. , apply the atomic unitary transformation to the diagonal and to all off - diagonal blocks of the hamiltonian * * dlu(nb ) * : calculate relativistic transformation for all atoms ( all atoms are treated as heavy elements ) , but apply the relativistic transformation only to blocks of the hamiltonian assigned to neighboring atoms ( and to the diagonal blocks , of course ) to achieve linear scaling .
* * dlu(@xmath121,bp ) * : distinguish heavy and light atoms , where @xmath121 denote the heavy atoms and apply the bp approximation to the light atoms . * * dlu(@xmath121,nr ) * : distinguish heavy and light atoms , where @xmath121 denote the heavy atoms and apply the nr approximation to the light atoms . 3 .
* dlh : * diagonal local approximation to hamiltonian matrix blocks .
the relativistic correction for off - diagonal interatomic blocks are neglected . * * dlh(all ) :* calculate a relativistic description for all diagonal atomic blocks . *
* dlh(@xmath121,nr / bp ) :* only heavy - atom blocks are considered for the relativistic description . note
that this classification system does not account for all possible local relativistic approximations : i ) @xmath23 could also represent a group of atoms and ii ) additional off - diagonal approximations may also be included .
we have implemented the local relativistic schemes discussed so far for the x2c , bss , and dkh exact - decoupling approaches into the molcas programme package @xcite , into which we have recently implemented x2c , bss , and the polynomial - cost dkh schemes @xcite . in our scalar - relativistic calculations the definition of the @xmath14 matrix was @xmath122 , wh@xmath123(ph@xmath124)@xmath124 , w(ch@xmath125)@xmath126 and pb@xmath127 .
moreover , two reactions @xmath128 were chosen to study reaction energies .
their structures ( see fig . [ structures ] ) have been optimized with the turbomole program package @xcite employing the bp86 density functional @xcite with a valence triple - zeta basis set with polarization functions on all atoms @xcite in combination with stuttgart effective core potentials for w , pb , and i as implemented in turbomole .
all molecules are closed - shell and for our analysis _ all - electron _ hartree fock calculations were then performed with the molcas programme package . since our aim is to test the validity of local relativistic approximations instead of obtaining accurate results which can be compared to experiments , we do not need to consider electron - correlation effects in our calculations .
the all - electron atomic natural orbital ( ano ) basis sets @xcite were employed at a double - zeta level for all atoms .
the exact - decoupling dkh calculations were performed at 35-th order , since higher order results did not improve the accuracy for our purpose so that this high dkh order can be considered exact .
results of finite - order ( standard ) dkh2 calculations are also given for the sake of comparison .
.computation times ( in seconds ) for the relativistic one - electron hamiltonian in various local schemes ( for pb@xmath127 ) .
[ cols= " < , > , > , > , > " , ] however , the nr approximation is much better .
the errors on the reaction energy within the dlu(nr ) scheme , which are in the range of 0.1 to 0.3 mh and thus less than 1 kj / mol , are perfectly acceptable .
this success is due to the correct nonrelativistic limit of the nr approximation within the dlu scheme .
not unexpectedly , the dlh(nr ) results have quite larger errors , because dlh(all ) already does not provide good results for the phosphine dissociation energy from the wh@xmath123(ph@xmath124)@xmath124 complex .
if we study the difference between dlh(nr ) and dlh(all ) , we find that the nr approximation to actually give small errors on top of dlh(all ) . however , the dlh scheme is not recommended since its accuracy is not guaranteed at all .
only in those cases for which the dlh scheme may work , we can further add the nr approximation on top of it for light atoms .
as can be seen from both table [ tab_rei5 ] and table [ tab_rewp3 ] , different relativistic approaches behave basically identical with respect to the discussion of the nb and nr approximations .
therefore , the approximations for the reduction of computational cost discussed in this article can be applied to all relativistic transformation approaches .
in this work , we aimed at a rigorous local approximation to relativistic transformation schemes , which makes them applicable to and efficient for calculations on large molecules .
we developed a systematic hierarchy of approximations that is based on the assembly of the unitary transformation from atomic contributions ( dlu ) rather than on a local approximation directly applied to the matrix representation of the hamiltonian ( dlh ) .
the straightforward dlh scheme , which has evolved in the development of the dkh method ( see references given above ) , turns out to be not a very accurate local approximation since it only covers the relativistic treatment of the atom same - atom diagonal blocks in the hamiltonian
. it may fail if inter - atomic distances become short so that a relativistic treatment of the off - diagonal interatomic blocks becomes important . as a consequence ,
the dlh approximation is not suitable , for example , in studies that crucially depend on accurate electronic energy surfaces .
even if the molecular geometry is fixed , the dlh approach does not provide a balanced description for different molecular properties , since the contributions of the off - diagonal blocks to total observables are quite different for different properties .
by contrast , the dlu scheme proposed by us in this paper overcomes the drawbacks of the dlh scheme .
it is an excellent approach to take into account the atom other - atom ( i.e. , interatomic ) off - diagonal relativistic transformations .
it does not show an instability for varying molecular structures ( i.e , for electronic energy surfaces ) and properties .
the size of the error introduced by dlu is much smaller than that of dlh ; it is typically only 1% of the latter .
the errors of the dlu approach when compared to the full relativistic transformation without any local approximation turn out to be tiny .
they are even smaller than the difference among the different exact - decoupling approaches , which are already very small .
the dlu scheme is valid for all exact - decoupling approaches , while the local @xmath35-matrix scheme only works for the x2c approach .
furthermore , the dlu scheme can also be applied to finite - order dkh approaches such as dkh2 , which has been very successful in computational chemistry and whose computational costs are less than those of any exact - decoupling approach . if one has to use the dlh scheme for some reason , such as the lack of a dlu implementation or because of its linear - scaling behavior , the selection of a relativistic approaches is not decisive as the errors introduced by the dlh scheme are already higher than the difference of different relativistic approaches .
hence , the approximate , but fast dkh2 method may be safely combined with the dlh approximation , while this can not be recommended for any exact - decoupling approach . however , the linear - scaling behavior of dlh is no good reason since the dlu scheme can also be made linear - scaling within the neighboring - atomic - blocks approximation , which produces negligible errors .
one can further reduce the computational costs by employing the bp or nr approximation to all light atoms . however , the bp approximation turns out to be not suitable for variational calculations since it introduces large errors . only the nr approximation yields acceptable results
. the errors of the nr approximation will be larger if the nuclear charge of the atom for which the nr approximation is invoked increases .
therefore , the application of the nr approximation will depend on the balance of accuracy and computational cost in an actual calculation . while this work focused on the accurate approximation of exact - decoupling methods at the self - consistent - field level with respect to total electronic energies and first - order properties of a small , but well - selected test molecule set , further investigations into the dlu scheme concerning , for example , higher - order properties , spin orbit and electron - correlation effects are under way in our laboratory . | we present a systematic hierarchy of approximations for _ local _ exact - decoupling of four - component quantum chemical hamiltonians based on the dirac equation .
our ansatz reaches beyond the trivial local approximation that is based on a unitary transformation of only the atomic block - diagonal part of the hamiltonian .
systematically , off - diagonal hamiltonian matrix blocks can be subjected to a unitary transformation to yield relativistically corrected matrix elements .
the full hierarchy is investigated with respect to the accuracy reached for the electronic energy and for selected molecular properties on a balanced test molecule set that comprises molecules with heavy elements in different bonding situations .
our atomic ( local ) assembly of the unitary exact - decoupling transformation called local approximation to the unitary decoupling transformation ( dlu ) provides an excellent local approximation for any relativistic exact - decoupling approach .
its order-@xmath0 scaling can be further reduced to linear scaling by employing a neighboring - atomic - blocks approximation .
therefore , dlu is an efficient relativistic method well suited for relativistic calculations on large molecules .
if a large molecule contains many light atoms ( typically hydrogen atoms ) , the computational costs can be further reduced by employing a well - defined nonrelativistic approximation for these light atoms without significant loss of accuracy .
we also demonstrate that the standard and straightforward transformation of only the atomic block - diagonal entries in the hamiltonian denoted dlh in this paper introduces an error that is on the order of the error of second - order douglas
hess ( i.e. , dkh2 ) when compared with exact - decoupling results .
hence , the local dlh approximation would be pointless in an exact - decoupling framework , but can be efficiently employed in combination with the fast to evaluate dkh2 hamiltonian in order to speed up calculations for which ultimate accuracy is not the major concern .
* date : * june 14 , 2012 + * status : * printed in j. chem . phys .
_ 136 _ * 2012 * 244108 |
the increasing quantity of data from hera has given a great boost to the testing of qcd via determinations of the strong coupling constant and the possibility of observing , over a wide kinematic range , its running behaviour .
+ however in any comparison between theory and experimental data proper account has to be taken of the non - perturbative behaviour inherent in any observable qcd process . while event shape variables are considered useful observables for @xmath1 determinations they can suffer from significant power suppressed effects which are non - perturbative in origin .
these @xmath0 corrections have already been calculated for a number of @xmath2 event shapes [ 1 ] and similar results are obtained here for dis . to make the analogy with @xmath2 shape variables clearer
it proves useful to work in the breit frame and to define current jet event shapes as done in the next section . the dispersive approach [ 2 ]
is used to estimate the power corrections , which is equivalent to a calculation of the leading renormalon ambiguity .
one can define several infrared safe quantities that characterize the shape of the event in the current hemisphere @xmath3 of the breit frame .
perturbative results are available in [ 3 ] .
the event shapes considered below are the thrust , the jet mass , the jet broadening and the @xmath4 parameter with the following definitions in the above order where @xmath5 denotes a unit 3-vector along the current direction and the summation index @xmath6 runs over only those partons that are in the current hemisphere .
@xmath7 @xmath8 @xmath9 @xmath10 where the @xmath11 denote the eigenvalues of the symmetric matrix @xmath12 defined as @xmath13 the subscript @xmath14 indicates the normalization factor @xmath15 .
one can also use a normalization factor given by the sum of energies of partons in the breit frame , but the normalization does not affect the power corrections we compute .
our estimate of the leading power corrections to the perturbative results are based on the approach of ref.[2 ] .
non - perturbative effects at long distances are assumed to give rise to a modification @xmath16 in the qcd effective coupling at low values of the scale @xmath17 .
the effect on some observable @xmath18 is then given by a characteristic function @xmath19 as follows : @xmath20 where @xmath21 the characteristic function is obtained by computing the relevant one - loop graphs with a non - zero gluon mass @xmath22 [ 4 ] . + arbitrary finite modifications of the effective coupling at low scales would generally introduce power corrections of the form @xmath23 into the ultraviolet behaviour of the running coupling itself . as discussed in ref.[2 ] , such a modification would destroy the basis of the operator product expansion [ 5 ] .
this leads to the constraint that only terms in the small-@xmath24 behaviour of the characteristic function that are non - analytic at @xmath25 will lead to power behaved non - perturbative contributions [ 2 ] .
we find that in dis as for @xmath26 the leading non - analytic behaviour takes the form of @xmath27 with an extra @xmath28 in the jet broadening .
the details of the calculation can be found in [ 3 ] .
the results for the power corrections are presented below .
@xmath29 is an phenomenological parameter first suggested in ref.[2 ] which is expected to be approximately universal .
@xmath30 data are consistent with a value for @xmath31 of about @xmath32 gev . in the case of the current jet thrust
we get a non - perturbative contribution @xmath33 for the current jet mass one gets a correction given by @xmath34 for the @xmath4 parameter we get @xmath35 in the case of the jet broadening a slightly different behaviour is obtained @xmath36 where @xmath37 is an undetermined scale .
the plots for the combined leading order prediction and power correction are presented for all of these event shape variables in figures 1 - 2 .
the mrs@xmath38 [ 8 ] parton distribution was used in every case .
it should be pointed out that as explained in ref.[3 ] the predictions for the numerical coefficients of @xmath39 for the @xmath4 parameter and the jet broadening are somewhat uncertain .
this is due to an inherent limitation of the massive gluon technique with regard to event shapes .
the use of a gluon mass in perturbative calculations is equivalent to summing inclusively over all bubbles in the renormalon chain . in the case of less inclusive variables like
event shapes the phase space is dependent on the bubble structure which is ignored by the massive gluon formulation .
this has been shown for event shapes in [ 6 ] and for gluon fragmentation functions in ref.[7 ] .
it is believed that this limitation does not affect the form of the power corrections to any of the shape variables and in fact appears not to make any significant difference to the predictions for the thrust and the jet mass where our results are not sensitive to the inclusion of a gluon mass in the definition of the shape variable .
the power corrections here arise from using a massive gluon phase space alone . in the case of the @xmath4 parameter and the jet
broadening the predictions for the numerical coefficients depend on whether or not a gluon mass is included throughout .
a glance at the relevant figures also reveals that there are enormously large power corrections in the case of @xmath40 and @xmath41 even at @xmath42 gev .
this fact coupled with the uncertainty about the coefficient would mean that these variables are unsuitable for an extraction of the strong coupling . on the other hand the predictions for power corrections to the thrust and the jet mass are more reliable making these , in our opinion , good observables for determination of @xmath43 .
the relevant experimental results have been presented at this workshop by the h1 collaboration [ 9,10 ]
all of the work presented here was done in collaboration with b.r.webber .
i am also grateful to trinity college , cambridge and the committee of vice - chancellors and principals of the universities of the united kingdom for financial support .
p.nason and m.h .
seymour , _ nucl .
b 454_(1995 ) 291 .
m.beneke , v.m .
braun and l.magnea , _ hep - ph/9701309_. a.d.martin , r.g roberts and w.j stirling , _
( 1994)6734 .
k.rabbertz , these proceedings .
h1 collaboration , c.adloff et al .
, _ preprint desy 97 - 098 _ , submitted to _ | results are presented for the leading power - suppressed ( @xmath0 ) contributions to shape variables in deep inelastic lepton scattering .
these are then combined with available leading order perturbative estimates .
cavendish - hep-97/07 + hep - ph/9706263 |
the ads / cft correspondence has revealed deep relation between string theory and gauge theory , in particular the correspondence between iib strings on @xmath2 and @xmath3 super yang - mills theory ( sym ) at the thooft large @xmath4 limit @xcite . while the correspondence was mostly tested for those bps states where partial supersymmetry protects it against quantum correction @xcite , semiclassical analysis on the near - bps sector is also considered in @xcite ; the integrability of this berenstein - madalcena - nastase ( bmn ) limit allows for quantitative tests of correspondence beyond bps states , where the energy of classical string solutions is compared to the anomalous dimension of sym operators with large r - charge . on the other hand , intimate relation between sym dynamics and integrable spin chain
was realized in @xcite , where the planar limit of the dilatation operator was identified with the hamiltonian of integrable spin chains .
the correspondence between rotating string and spin chain was further explored in many papers , see @xcite for a sample of it .
recently , a giant magnon solution carrying momentum @xmath5 was first found on @xmath6 with dispersion relation @xcite : @xmath7 where both energy @xmath8 and r - charge @xmath9 becomes infinite as @xmath10 stays finite but large .
in addition , an infinitely winding solution with spiky configuration was also found with dispersion relation @xcite : @xmath11 for angular difference between two spikes @xmath12 and spike height @xmath13 .
in fact , both solutions can be obtained from the same rotating string at different limits . since
then many papers have been devoted to this study @xcite .
in particular , solutions with nontrivial ns - ns @xmath1-field were also investigated in various context , such as in the melvin twisted background @xcite , lunin - maldacena background @xcite , and @xmath6 with ns - ns @xmath1-field @xcite . in this paper
, we consider a rotating string on @xmath0 with background ns - ns @xmath1-field .
this background can be obtained in the type iib strings with appropriate embedding ansatz .
we obtain both solutions of spiky string and giant magnon as two limiting cases , providing appropriate choice of integral constants . as an integrable system , we find it is straightforward to write down the corresponding sine - gordon model upon a redefinition of mass of sine - gordon field , and the scattering and bound state of magnons are ready to be translated into solutions of multi - kinks . in the spirit of the ads
/ cft correspondence , we also expect the existence of a dual description given by a sym deformed by such @xmath1-field .
this conjecture is later supported by providing a dictionary established between dispersion relation of magnon solution in the gravity side and excitation of spin chain formed by the long trace operator in the proposed @xmath1-deformed sym .
the outline of this paper is as follows : first in the section ii and iii , we present a rotating string solution on different coordinate systems of @xmath0 with ns - ns @xmath1-field using the nambu - goto string action . in the two limiting cases
, we obtain the solution for spiky string and giant magnon . in the section iv , we take advantage of the polyakov action in order to make a connection to the sine - gordon model . in the section
v , we investigate the obtained solutions within the description of spin chain , and conjecture the existence of a @xmath1-deformed super yang - mills theory , which is dual to our string background with ns - ns @xmath1-field . at last
, we have a discussion in the section vi .
we consider a rigid string rotating on @xmath0 with background ns - ns @xmath1-field .
the relevant metric reads , @xmath14 with magnetic ns - ns @xmath1-field @xmath15 this background can be obtained from the type iib strings with appropriate embedding ansatz .
one possible construction is to consider a near - horizon geometry of intersecting ns1 and ns5 branes @xcite .
the magnetic component of ns - ns flux supported by ns5 branes fills the @xmath0 with constant field strength , while the electric component is decoupled from the string worldsheet . the corresponding nambu - goto ( ng ) lagrangian is @xmath16^{\frac12 } \nonumber\\ & & \qquad - b \sin^2\vartheta ( \partial_\tau \varphi_1 \partial_\sigma \varphi_2 - \partial_\sigma \varphi_1 \partial_\tau \varphi_2 ) \biggr\},\end{aligned}\ ] ] where @xmath17 is the string tension .
we assume the string rotates in @xmath18 and @xmath19 directions , with the following embedding ansatz : @xmath20 the solution of @xmath21 , in terms of @xmath22 , can be obtained by solving the equation for @xmath23 : @xmath24 and the @xmath22 can be solved by the ratio of equations for @xmath23 and @xmath18 : @xmath25}{\nu_2 \cos^2\vartheta ( \kappa c_2 - \nu_1 c_1 \sin^2\vartheta - b \kappa \nu_2 \sin^2\vartheta)}.\ ] ] the general expression of @xmath21 is rather complicated , to be explicit as follows : @xmath26 , \nonumber\\ % \alpha_2 & = & - c_2 ^ 2 \nu_2 ^ 2 + 3 b^2 \nu_1 ^ 2 \nu_2 ^ 2 - 4 c_2 b \nu_1 ^ 2 \nu_2 + 2 c_2 b \nu_2 ^ 3 + c_1 ^ 2 \nu_2 ^ 2 - b^2 \nu_2 ^ 4 + c_1 ^ 2 \kappa^2 + c_2 ^ 2 \nu_1 ^ 2 + \kappa^2 \nu_2 ^ 2 - 2 c_1 c_2 \kappa \nu_1 - \nu_1 ^ 2 \nu_2
^ 2 + 4 c_1 b \kappa \nu_1 \nu_2 , \alpha_2 & = & - \alpha_0 + \nu_2 \left [ \nu_2 ( c_1 ^ 2 - c_2 ^ 2 - \nu_1 ^ 2 + \kappa^2 ) + 2 b ( c_2 \nu_2 ^ 2 - c_2 \nu_1 ^ 2 + c_1 \kappa
\nu_1 ) + b^2 \nu_2 ( 2 \nu_1 ^ 2 - \nu_2 ^ 2 ) \right ] , \nonumber\\ \alpha_0 & = & - ( c_1 \kappa - c_2 \nu_1 + b \nu_1 \nu_2)^2 . \nonumber\end{aligned}\ ] ] the desired boundary condition @xmath27 at @xmath28 requires @xmath29 which give two possible constraints : @xmath30 corresponding to spiky string and giant magnon respectively .
the first choice of parameters effectively imposes the following conditions at @xmath31 : ( i ) @xmath22 is regular by @xmath32 and ( ii ) @xmath33 by @xmath34 .
then the solutions of @xmath22 and @xmath21 reduce to @xmath35 where @xmath36 the conserved quantities , energy and two angular momenta , are given by @xmath37}{\kappa \sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \ , \cos\vartheta \sqrt{\sin^2\vartheta - \sin^2\vartheta_0 } } \nonumber\\ = 2 t \left [ \frac{\nu_1 ( 1 - b^2 ) \cos\vartheta_0}{\sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2)(1 - b^2 ) } } + \frac{b \nu_1 ( \nu_2 + b \kappa ) \ln\frac{2 \cos\vartheta_0}{\cos\frac{\pi}2}}{\kappa \sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \cos\vartheta_0 } \right ] , \label{j1spikys3 } \\
j_2 & = & 2 \int_{\vartheta_0}^{\frac{\pi}2 } \ , \frac{d\vartheta}{\vartheta ' } \ , \frac{\partial { \cal l}_{\rm ng}^{s^3}}{\partial \dot \varphi_2 } % = - 2 t \int_{\vartheta_0}^{\frac{\pi}2 } d\vartheta \frac{\sin\vartheta [ \kappa \nu_2 ( 1 - b^2 ) \cos^2\vartheta + b ( \nu_1 ^ 2 - \kappa^2 ) ] } { \kappa \sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \ , \cos\vartheta \sqrt{\sin^2\vartheta - \sin^2\vartheta_0 } } \nonumber\\ = - 2 t \left [ \frac{\nu_2 ( 1 - b^2 ) \cos\vartheta_0}{\sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2)(1 - b^2 ) } } + \frac{b ( \nu_1 ^ 2 - \kappa^2 ) \ln\frac{2
\cos\vartheta_0}{\cos\frac{\pi}2}}{\kappa \sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \cos\vartheta_0 } \right ] . \label{j2spikys3}\end{aligned}\ ] ] the terms with @xmath38 represent the divergent part of the physical quantities , indicating the energy diverges in general .
the longitude angle between two spikes , @xmath39 is also a divergent quantity . their combination
, however , renders a finite value : @xmath40 the spiky string can be visualized as a string wrapped around the equator an infinite number of times , as shown in the fig .
[ fig : spike ] . the divergent quantities
@xmath8 and @xmath41 are due to the infinite length of string . however , the latitude angle from peak of spike to the north pole is nothing but @xmath42 , which stays finite .
this relation holds independently from the @xmath1-field
. moreover , the angular momenta of string , @xmath43 and @xmath44 , are also divergent in general , except for two special values of parameters . in the first case ,
@xmath1-field is absent , i.e. @xmath45 , and we have the dispersion relation analog with that of single spiky string @xcite @xmath46 however , in this analog , @xmath43 is related to the energy of spiky string and the momentum of spiky string can also be identified as @xmath47 . in the other case
, we have @xmath48 and @xmath49 , which , however , reduces to a trivial configuration , namely @xmath50 and @xmath51 . in general
, one can use @xmath8 to _ regularize _ @xmath43 and @xmath44 as follows : @xmath52 and to derive the dispersion relation : @xmath53 in order to picture the explicit profile of spiky string , we can integrate ( [ eqn : vartheta ] ) and obtain a relation between @xmath54 and @xmath55 @xmath56 where there is no bound for @xmath54 .
in fact , one can easily see that @xmath57 as @xmath58 .
the relation for @xmath59 is complicated but can be obtained by iteration .
section @xmath60 with @xmath45 .
it shows the string wraps around the equator an infinite number of times.,scaledwidth=45.0% ] the other choice of parameter , corresponding to giant magnon , is for @xmath61 and @xmath62 .
we then obtain @xmath63 where @xmath64 the conserved quantities , energy and two angular momenta , are given as follows : @xmath65}{\nu_1 \sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \ , \cos\vartheta \sqrt{\sin^2\vartheta - \sin^2\vartheta_1}}\nonumber\\ \!=\ ! 2 t \left [ \frac{- \nu_1 ( 1 - b^2 ) \cos\vartheta_1}{\sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2)(1 - b^2 ) } } \!+\ ! \frac{(\nu_1 ^ 2 \!-\ ! c_1 ^ 2 \!-\ ! c_1 b \nu_2 \!-\
! b^2 \nu_1 ^ 2 ) \ln\frac{2 \cos\vartheta_1}{\cos\frac{\pi}2}}{\nu_1 \sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \cos\vartheta_1 } \right ] , \label{j1magnons3 } \\
j_2 & = & 2 \int_{\vartheta_1}^{\frac{\pi}2 } \ , \frac{d\vartheta}{\vartheta ' } \ , \frac{\partial { \cal l}_{\rm ng}}{\partial \dot \varphi_2 } % = 2 t \int_{\vartheta_1}^{\frac{\pi}2 } d\vartheta \frac{\nu_2 ( 1 - b^2 ) \sin\vartheta \cos\vartheta}{\sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \ , \sqrt{\sin^2\vartheta - \sin^2\vartheta_1}}\nonumber\\ = 2 t \frac{\nu_2 ( 1 - b^2 ) \cos\vartheta_1}{\sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2)(1 - b^2)}}. \label{j2magnons3}\end{aligned}\ ] ] we have string configuration plotted in the fig .
[ fig : magnon ] for various values of @xmath1-field .
the longitude angle spanned between two ends of string is @xmath66 where the momentum of magnon is identified as @xmath67 .
this relation again holds independently of the @xmath1-field .
the energy , @xmath8 , and one angular momentum , @xmath43 , are divergent .
however , we can rescale energy as @xmath68 and then obtain the following dispersion relation : @xmath69 similarly , for the explicit profile of giant magnon we can integrate ( [ mag : varphi ] ) and ( [ mag : vartheta ] ) , then we have simple relations : @xmath70 this implies the finite range of @xmath54 , i.e. @xmath71 . section of @xmath72 with various @xmath1-field , where @xmath45 ( red ) , @xmath73 ( green ) , @xmath74 ( blue ) . to the left , string configuration in @xmath75 . to the right , string configuration in @xmath76.,title="fig:",scaledwidth=45.0% ]
section of @xmath72 with various @xmath1-field , where @xmath45 ( red ) , @xmath73 ( green ) , @xmath74 ( blue ) . to the left ,
string configuration in @xmath75 . to the right ,
string configuration in @xmath76.,title="fig:",scaledwidth=45.0% ]
in this section , we study a rotating string in a different coordinate system .
let us transform @xmath0 to a new angular coordinate system : @xmath78 such that the @xmath0 metric is written as the hopf fibration 3-sphere ] , i.e. @xmath79 @xmath80,\ ] ] and the b field becomes @xmath81 straightforwardly , the solution is transformed as follows : @xmath82 instead of repeating the same calculation , we now consider a more general background @xmath83 , \nonumber\\ b & = & ( b_0 + b_1 \cos\theta + b_2 \cos\phi ) \ , d\phi \wedge d\psi,\end{aligned}\ ] ] which recovers the @xmath0 case for the special values @xmath84 .
then the ng action becomes @xmath85^{\frac12 } \nonumber\\ & & \qquad - ( b_0 + b_1 \cos\theta + b_2 \cos\phi ) ( \partial_\tau \phi \partial_\sigma \psi - \partial_\tau \psi \partial_\sigma \phi ) \biggr\}.\end{aligned}\ ] ] assuming a typical form of solution @xmath86 the general solution for @xmath87 is do not appear in the solution since they effectively are total derivative terms in the ng action . ]
@xmath88 ^ 2}{c_1 ^ 2 [ \kappa^2 - k^2 \mu_1
^ 2 \sin^2\theta - k^2 ( \mu_2 + \mu_1 \cos\theta)^2 ] } - \sin^2\theta - ( \psi ' + \cos\theta)^2 , % \\ % ? ?
\psi ' = - \frac{c_1 ( \kappa^2 - k^2 \mu_2 ^ 2 \sin^2\theta ) - \kappa ( c_2 - b_1 \mu_2 \cos\theta ) ( \mu_1 + \mu_2 \cos\theta)}{c_1 ( \kappa^2 \cos\theta + k^2 \mu_1 \mu_2 \sin^2\theta ) - \kappa ( c_2 - b_1 \mu_2 \cos\theta ) ( \mu_2 + \mu_1 \cos\theta)},\ ] ] which is solved from the equation of @xmath89 .
: @xmath90 which is inverse of the solution from equation of @xmath89 together with changing parameters @xmath91 . after imposing @xmath92
, one obtains the desired @xmath87 .
this duality can be understood as exchange of coordinates @xmath89 and @xmath93 , while @xmath1-field has to flip the sign simultaneously . ]
the expression of @xmath94 is @xmath95 , \nonumber\\ % & = & \mu_2 ^ 2 ( - b_1 ^ 2 \mu_1 ^ 2 + k^4 \mu_2 ^ 2 - c_1 ^ 2 k^2 - 2 c_1 b_1 \kappa + k^4 \mu_1 ^ 2 + 4 c_2 b_1 \mu_1 - k^2 \kappa^2 - b_1 ^ 2 \mu_2 ^ 2 ) , % \nonumber\\ \alpha_1 & = & - 2 \mu_2 \left [ c_2 ^ 2 \mu_1 - c_1 c_2 \kappa + k^4 \mu_1 \mu_2 ^ 2 + b_1 ( c_1 \kappa \mu_1 - c_2 \mu_1 ^ 2 - c_2 \mu_2 ^ 2 ) \right ] , \nonumber\\ % & = & - 2 \mu_2 ( - c_2 b_1 \mu_1 ^ 2 + c_1 b_1 \kappa \mu_1 + c_2 ^ 2 \mu_1 - c_1 c_2 \kappa + k^4 \mu_1 \mu_2 ^ 2 - c_2 b_1 \mu_2 ^ 2 ) , % \nonumber\\ \alpha_0 & = & - ( c_1 \kappa - c_2 \mu_1)^2 - \mu_2 ^ 2 \left [ ( \mu_1 ^ 2 + \mu_2 ^ 2 ) k^4 - ( c_1 ^ 2 + \kappa^2 ) k^2 + c_2 ^ 2 \right ] .
\nonumber\end{aligned}\ ] ] let us analyze the boundary conditions at the equator ; at @xmath96 , we generally have a _ fixed _
values for @xmath97 and @xmath98 , regardless of the parameters . in order to relax this restriction
, we should impose @xmath99 and then @xmath87 and @xmath94 become @xmath100 } , \\
\psi ' & = & - \frac{\kappa \mu_2 ( 1 + \cos\theta ) [ c_2 - b_1 \mu_1 + b_1 \mu_2 ( 1 - \cos\theta ) ] + c_1 k^2 \mu_2 ^ 2 \sin^2\theta}{\mu_1 ( 1 + \cos\theta ) [ c_2 \kappa - c_1 k^2 \mu_2 + \mu_2 ( c_1 k^2 - b_1 \kappa ) \cos\theta]}.\end{aligned}\ ] ] now we can avoid @xmath94 to be vanishing at the equator if one can factor out @xmath101 from the square root .
this requires @xmath102 which give two possible constraints @xmath103 together with the constraint ( [ s1s2c1 ] ) , we arrive at two choices of parameters corresponding to spiky string and giant magnon respectively : @xmath104 the spiky string is given by the first choice in ( [ s1s2constraint ] ) , @xmath105 and @xmath106 , and the @xmath94 and @xmath87 are simplified as @xmath107 where @xmath108 the energy of spiky string is @xmath109 ( 1 - \cos\theta)}{\kappa \sqrt{2 \mu_1 \mu_2 ( b_1 ^ 2 - k^4 ) } \sin\theta \sqrt{\cos\theta_0 - \cos\theta } } , \nonumber\\ = 2 t \frac{\sqrt2 k^2 [ \kappa^2 - k^2 ( \mu_1 - \mu_2)^2)]}{\kappa \sqrt{2 \mu_1 \mu_2 ( b_1 ^ 2 - k^4 ) } \cos\frac{\theta_0}2 } \ln \frac{2 \cos\frac{\theta_0}2}{\cos\frac{\pi}2}.\end{aligned}\ ] ] however , for the particular values : @xmath110 , then we have @xmath111 which is consistent with ( [ espikys3 ] ) with opposite sign due to the different branch solution of @xmath94 .
unlike the energy , the angular momenta do include contribution from @xmath112 and @xmath113 , therefore be more complicated .
their expressions are @xmath114 and @xmath115 for the 3-sphere case : @xmath116 , the coefficients reduce to @xmath117 and the angular momenta are @xmath118 moreover , the following combinations of angular momenta @xmath119 are consistent with ( [ j1spikys3 ] ) and ( [ j2spikys3 ] ) respectively , as we can expect them from the coordinate transformation ( [ hopf ] ) .
the second choice of parameters , @xmath120 , gives giant magnon which simplifies @xmath94 and @xmath87 as @xmath121 where @xmath122 the energy of the magnon reads @xmath123 ( 1 - \cos\theta)}{(\mu_1 - \mu_2 ) \sqrt{2
\mu_1 \mu_2 ( b_1 ^ 2 - k^4 ) } \sin\theta \sqrt{\cos\theta_1 - \cos\theta}},\nonumber\\ = - 2 t \frac{\sqrt2 k [ c_1 ^ 2 - k^2 ( \mu_1 - \mu_2)^2)]}{(\mu_1 - \mu_2 ) \sqrt{2
\mu_1 \mu_2 ( b_1 ^ 2 - k^4 ) } \cos\frac{\theta_1}2 } \ln \frac{2 \cos\frac{\theta_1}2}{\cos\frac{\pi}2}.\end{aligned}\ ] ] for the particular values : @xmath110 , then we have @xmath124 which is identical with ( [ emagnons3 ] ) .
the angular momenta again include contribution from @xmath112 and @xmath113 .
their expressions are @xmath125 and @xmath126 for the 3-sphere case : @xmath116 , we have @xmath127 and the angular momenta become @xmath128 similarly , the following combinations of angular momenta @xmath129}{\nu_1 \sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2 ) ( 1 - b^2 ) } \cos\frac{\theta_1}2 } \ln \frac{2 \cos\frac{\theta_1}2}{\cos\frac{\pi}2 } , \\
j_+ & = & j_1 + j_2 % = \frac{t(\mu_1 + \mu_2 ) ( 1 - b^2)}{\sqrt{2 \mu_1 \mu_2 ( b^2 - 1 ) } } \sqrt{\cos\theta_1 + 1 } = \frac{2 t \nu_2 ( 1 - b^2 ) \cos\frac{\theta_1}2}{\sqrt{(\nu_1 ^ 2 - \nu_2 ^ 2)(1 - b^2)}},\end{aligned}\ ] ] are identical to ( [ j1magnons3 ] ) and ( [ j2magnons3 ] ) respectively .
in order to make connection to the sine - gordon model @xcite , it is convenient to use the polyakov action @xmath130 \nonumber\\ & & \qquad + \cos^2\vartheta \left [ ( \partial_\tau \varphi_2)^2 - ( \partial_\sigma \varphi_2)^2 \right ] + 2 b \sin^2\vartheta ( \partial_\tau \varphi_1 \partial_\sigma \varphi_2 - \partial_\sigma \varphi_1 \partial_\tau \varphi_2 ) \bigr\},\end{aligned}\ ] ] supplemented by the virasoro constraint : @xmath131 & & \nonumber\\ + \cos^2\vartheta [ ( \partial_\tau \varphi_2)^2 + ( \partial_\sigma \varphi_2)^2 ] & = & 0 .
% g_{\mu\nu } \partial_\tau x^\mu \partial_\sigma x^\nu = 0 , \qquad g_{\mu\nu } ( \partial_\tau x^\mu \partial_\tau x^\nu + \partial_\sigma x^\mu \partial_\sigma x^\nu ) = 0.\end{aligned}\ ] ] now we will work with a similar , but slightly different embedding ansatz : @xmath132 from the virasoro constraint , one obtains the relation @xmath133 and the equation of motion @xmath134.\end{aligned}\ ] ] the rest of equations of motion read @xmath135 in order to have desired boundary condition , i.e. @xmath27 at @xmath28 , one should impose the following constraints on the parameters @xmath136 which , together with the constraint ( [ vc1 ] ) , leads to two possible solutions @xmath137 corresponding to spiky string and giant magnon respectively .
the value of @xmath1 in both cases is chosen to avoid divergence of @xmath138 at the equator .
we also explicit the evaluation of worldsheet action for later use , @xmath139 with the choice of @xmath140 and a turning point at @xmath28 , we claim to have a spiky solution : @xmath141 we may find solution of @xmath55 , @xmath142 the worldsheet action reduces to @xmath143 if we impose @xmath144 and define the field @xmath145 then we recover the sine - gordon equation , @xmath146 the conserved quantities can be computed by the following formula @xmath147 which give exactly the same results as in ng formulation .
now we are ready to identify a spiky string as a kink or soliton of mass parameter @xmath148 and velocity @xmath149 in the corresponding sine - gordon equation ( [ sg_spike ] ) , provided @xmath150 where a plus ( minus ) sign is assigned to ( anti-)kink solution . with the choice of @xmath151 and a turning point at @xmath28
, we claim to have a magnon solution : @xmath152 we may find solution of @xmath55 , @xmath153 the worldsheet action reduces to @xmath154 therefore , if we define the field @xmath155 then the sine - gordon equation is recovered , @xmath156 once again , we will have the same physical quantities with @xmath157 .
now we are ready to identify a magnon as a kink of mass parameter @xmath148 and velocity @xmath158 in the corresponding sine - gordon equation ( [ sg_magnon ] ) , provided @xmath159 where a plus ( minus ) sign is assigned to ( anti-)kink solution .
solutions of n ( anti-)kinks can be easily constructed and correspond to configuration of multiple ( anti-)magnons . in particular , a solution corresponds to scattering between two magnons , in the center of mass frame , reads @xmath160 and scattering between magnon and anti - magnon is @xmath161 moreover , a breather solution ( bound state of magnon and anti - magnon ) is @xmath162
if we carefully inspect the conserved quantities and dispersion relation , we may find that a simple redefinition , @xmath163 can bring us back to the ordinary description without @xmath1-field .
that is in the giant magnon solution .
] , @xmath164 therefore , one is tempted to claim a dual super yang - mills theory if @xmath0 is part of @xmath165 or @xmath166 if we are dealing with ns1-ns5 system .
] deformed by such a @xmath1-field , denoting @xmath1-deformed sym , can be realized by a simple replacement ( [ b - relation ] ) . in particular
, the spin - chain description in the @xmath167 sector of @xmath1-deformed sym consists of operators of the form , @xmath168 + \cdots,\ ] ] where @xmath169 and @xmath170 are two complex adjoint scalars of the theory and the dots denote all possible orderings of the fields . with no deformation , @xmath43 and
@xmath44 are two integer - valued charges corresponding to a @xmath171 subgroup of the r - symmetry group .
with deformation , they are rescaled by a factor @xmath172 and no need to be integers .
therefore we obtain a _ fractional _
chain of non - integer length @xmath173 . to be precise , at one - loop the dilatation operator in the @xmath167 sector can be mapped onto the hamiltonian of the heisenberg spin chain @xcite .
the scaling dimension @xmath174 is related to the energy @xmath175 @xmath176 where each magnon carries individual energy @xmath177 and momentum @xmath178 . at the limit @xmath179 , equation ( [ dispersion_magnon ] ) agrees with ( [ spinchain ] ) , up to @xmath180 , for single magnon
if we identify @xmath181 this provides a strong evidence to support our conjecture of existence of a @xmath1-deformed sym dual to our gravity background .
we have studied a rotating string on @xmath0 with background ns - ns @xmath1-field and found solutions of spiky string and giant magnon as two limiting cases . the dispersion relation between regularized conserved charges is still in a simple expression but refined by ( [ b - relation ] ) .
it is interesting to compare our results with those from other deformed backgrounds with nontrivial ns - ns @xmath1-field . in the lunin - maldacena background , for instance
, @xmath182-deformation causes a shift in the magnon momentum @xcite , and the dual field theory is the @xmath183 sym , which is marginally deformed from @xmath3 . in our background , @xmath1-field has the effect of recaling string tension , and we have suspect a @xmath1-deformed sym to be the dual description , justified by matching the dispersion relation on both sides . as a future direction , it might be interesting to investigate the finite size effect to our solutions @xcite , where one expects to set one of turning points to be located at some other @xmath185 instead of the equator .
moreover , by rewriting @xmath0 as a hopf fibration , we are able to carry out our results to the study of rotating string on other backgrounds such as deformed @xmath186 @xcite or @xmath187 @xcite .
this work is in progress .
the authors would like thank hsien - chung kao , pei - ming ho and takeo inami for useful discussion .
the authors are supported by the taiwan s national science council and national center for theoretical sciences under grant no .
nsc 96 - 2112-m-008 - 006-my3 ( cmc , jht ) , nsc97 - 2112-m-002 - 015-my3 ( wyw ) and nsc97 - 2119-m-002 - 001 .
j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv . theor . math .
* 2 * , 231 ( 1998 ) [ int .
j. theor .
phys . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] .
s. s. gubser , i. r. klebanov and a. m. polyakov , `` gauge theory correlators from non - critical string theory , '' phys .
b * 428 * , 105 ( 1998 ) [ arxiv : hep - th/9802109 ] .
e. witten , `` anti - de sitter space and holography , '' adv .
* 2 * , 253 ( 1998 ) [ arxiv : hep - th/9802150 ] .
o. aharony , s. s. gubser , j. m. maldacena , h. ooguri and y. oz , `` large n field theories , string theory and gravity , '' phys . rept .
* 323 * , 183 ( 2000 ) [ arxiv : hep - th/9905111 ] .
d. e. berenstein , j. m. maldacena and h. s. nastase , `` strings in flat space and pp waves from n = 4 super yang mills , '' jhep * 0204 * , 013 ( 2002 ) [ arxiv : hep - th/0202021 ] .
n. beisert , s. frolov , m. staudacher and a. a. tseytlin , `` precision spectroscopy of ads / cft , '' jhep * 0310 * , 037 ( 2003 ) [ arxiv : hep - th/0308117 ] . m. kruczenski , `` spin chains and string theory , '' phys .
lett . * 93 * , 161602 ( 2004 ) [ arxiv : hep - th/0311203 ] .
s. frolov and a. a. tseytlin , `` rotating string solutions : ads / cft duality in non - supersymmetric sectors , '' phys .
b * 570 * , 96 ( 2003 ) [ arxiv : hep - th/0306143 ] .
s. bellucci , p. y. casteill , j. f. morales and c. sochichiu , `` sl(2 ) spin chain and spinning strings on ads(5 ) x s**5 , '' nucl .
b * 707 * , 303 ( 2005 ) [ arxiv : hep - th/0409086 ] .
o. lunin and j. m. maldacena , `` deforming field theories with u(1 ) x u(1 ) global symmetry and their gravity duals , '' jhep * 0505 * , 033 ( 2005 ) [ arxiv : hep - th/0502086 ]
. s. a. frolov , r. roiban and a. a. tseytlin , `` gauge - string duality for superconformal deformations of n = 4 super yang - mills theory , '' jhep * 0507 * , 045 ( 2005 ) [ arxiv : hep - th/0503192 ] .
w. h. huang , `` spin chain with magnetic field and spinning string with b field , '' phys .
d * 74 * , 027901 ( 2006 ) [ arxiv : hep - th/0605242 ] .
w. y. wen , `` spin chain from marginally deformed ads(3 ) x s**3 , '' phys .
d * 75 * , 067901 ( 2007 ) [ arxiv : hep - th/0610147 ] .
w. y. wen , `` 2d heisenberg model from rotating membrane , '' nucl .
b * 791 * , 164 ( 2008 ) [ arxiv:0705.2634 [ hep - th ] ] .
d. bak , d. gang and s. j. rey , `` integrable spin chain of superconformal u(m)xu(n ) chern - simons theory , '' arxiv:0808.0170 [ hep - th ] .
d. m. hofman and j. m. maldacena , `` giant magnons , '' j. phys .
a * 39 * , 13095 ( 2006 ) [ arxiv : hep - th/0604135 ] .
r. ishizeki and m. kruczenski , `` single spike solutions for strings on s2 and s3 , '' phys .
d * 76 * , 126006 ( 2007 ) [ arxiv:0705.2429 [ hep - th ] ] .
a. e. mosaffa and b. safarzadeh , `` dual spikes : new spiky string solutions , '' jhep * 0708 * , 017 ( 2007 ) [ arxiv:0705.3131 [ hep - th ] ] .
h. y. chen , n. dorey and k. okamura , `` dyonic giant magnons , '' jhep * 0609 * , 024 ( 2006 ) [ arxiv : hep - th/0605155 ] .
g. arutyunov , s. frolov and m. zamaklar , `` finite - size effects from giant magnons , '' arxiv : hep - th/0606126
. d. astolfi , v. forini , g. grignani and g. w. semenoff , `` gauge invariant finite size spectrum of the giant magnon , '' phys . lett .
b * 651 * , 329 ( 2007 ) [ arxiv : hep - th/0702043 ] .
j. a. minahan , a. tirziu and a. a. tseytlin , `` infinite spin limit of semiclassical string states , '' jhep * 0608 * , 049 ( 2006 ) [ arxiv : hep - th/0606145 ] . c. s. chu , g. georgiou and v. v. khoze , `` magnons , classical strings and beta - deformations , '' jhep * 0611 * , 093 ( 2006 ) [ arxiv : hep - th/0606220 ] . m. spradlin and a. volovich , `` dressing the giant magnon , '' jhep * 0610 * , 012 ( 2006 ) [ arxiv : hep - th/0607009 ] . c. kalousios , m. spradlin and a. volovich , `` dressing the giant magnon .
ii , '' jhep * 0703 * , 020 ( 2007 ) [ arxiv : hep - th/0611033 ] .
n. p. bobev and r. c. rashkov , `` multispin giant magnons , '' phys .
d * 74 * , 046011 ( 2006 ) [ arxiv : hep - th/0607018 ] .
w. h. huang , `` giant magnons under ns - ns and melvin fields , '' jhep * 0612 * , 040 ( 2006 ) [ arxiv : hep - th/0607161 ] .
j. plefka , `` spinning strings and integrable spin chains in the ads / cft correspondence , '' arxiv : hep - th/0507136 .
b. h. lee , r. r. nayak , k. l. panigrahi and c. park , `` on the giant magnon and spike solutions for strings on ads@xmath188 s@xmath189 , '' jhep * 0806 * , 065 ( 2008 ) [ arxiv:0804.2923 [ hep - th ] ] .
j. r. david and b. sahoo , `` giant magnons in the d1-d5 system , '' jhep * 0807 * , 033 ( 2008 ) [ arxiv:0804.3267 [ hep - th ] ] .
p. bozhilov , `` spin chain from membrane and the neumann - rosochatius integrable system , '' phys .
d * 76 * , 106003 ( 2007 ) [ arxiv:0706.1443 [ hep - th ] ] .
h. dimov and r. c. rashkov , `` on the anatomy of multi - spin magnon and single spike string solutions , '' nucl .
b * 799 * , 255 ( 2008 ) [ arxiv:0709.4231 [ hep - th ] ] .
j. kluson and k. l. panigrahi , `` d1-brane in beta - deformed background , '' jhep * 0711 * , 011 ( 2007 ) [ arxiv:0710.0148 [ hep - th ] ] .
r. ishizeki , m. kruczenski , m. spradlin and a. volovich , `` scattering of single spikes , '' jhep * 0802 * , 009 ( 2008 ) [ arxiv:0710.2300 [ hep - th ] ] .
j. kluson , `` fundamental string and d1-brane in i - brane background , '' jhep * 0801 * , 045 ( 2008 ) [ arxiv:0711.4219 [ hep - th ] ] .
j. kluson , `` note about string with euclidean world - sheet in ads(5)xs(5 ) , '' arxiv:0802.0935 [ hep - th ] .
m. c. abbott and i. v. aniceto , `` vibrating giant spikes and the large - winding sector , '' jhep * 0806 * , 088 ( 2008 ) [ arxiv:0803.4222 [ hep - th ] ] .
c. ahn and p. bozhilov , `` finite - size effects for single spike , '' jhep * 0807 * , 105 ( 2008 ) [ arxiv:0806.1085 [ hep - th ] ] .
b. h. lee , k. l. panigrahi and c. park , `` spiky strings on ads@xmath190 , '' [ arxiv:0807.2559 [ hep - th ] ] . c. ahn ,
p. bozhilov and r. c. rashkov , `` neumann - rosochatius integrable system for strings on @xmath191 , '' jhep * 0809 * , 017 ( 2008 ) [ arxiv:0807.3134 [ hep - th ] ] . c. gomez and r. hernandez , `` the magnon kinematics of the ads / cft correspondence , '' jhep * 0611 * , 021 ( 2006 ) [ arxiv : hep - th/0608029 ] .
g. t. horowitz and a. strominger , `` black strings and p - branes , '' nucl .
b * 360 * , 197 ( 1991 ) .
j. p. gauntlett , `` intersecting branes , '' arxiv : hep - th/9705011
. k. pohlmeyer , `` integrable hamiltonian systems and interactions through quadratic constraints , '' commun .
phys . * 46 * , 207 ( 1976 ) .
j. a. minahan and k. zarembo , `` the bethe - ansatz for n = 4 super yang - mills , '' jhep * 0303 * , 013 ( 2003 ) [ arxiv : hep - th/0212208 ] .
d. israel , c. kounnas , d. orlando and p. m. petropoulos , `` electric / magnetic deformations of @xmath0 and @xmath192 , and geometric cosets , '' fortsch .
phys . * 53 * , 73 ( 2005 ) [ arxiv : hep - th/0405213 ] .
o. aharony , o. bergman , d. l. jafferis and j. maldacena , `` n=6 superconformal chern - simons - matter theories , m2-branes and their gravity duals , '' arxiv:0806.1218 [ hep - th ] . c. ahn ,
`` squashing gravity dual of n=6 superconformal chern - simons gauge theory , '' arxiv:0809.3684 [ hep - th ] . | we study solutions for a rotating string on @xmath0 with a background ns - ns @xmath1-field and show the existence of spiky string and giant magnon as two limiting solutions .
we make a connection to the sine - gordon model via the polyakov worldsheet action and study the effect of @xmath1-field . in particular
, we find the magnon solution can be mapped to the excitation of a fractional spin chain .
we conjecture a @xmath1-deformed sym to be the gauge theory dual to this background . |
the smith cloud is perhaps the best known and most studied member of the high - velocity clouds ( hvcs ) .
while covering over a third of the sky , little is known about the nature of these clouds , with many suggestions as to their origin . while we now know the distances to a number of hvcs , their orbits are more difficult to determine . since its discovery in @xcite , the smith cloud was originally believed to be an extension of the galactic disk @xcite but it was later shown to be separated kinematically from the galactic disk @xcite leading to its subsequent classification as an hvc @xcite . while the origin of the cloud is still unknown , with suggestions of gas belonging to an infalling dwarf galaxy @xcite or gas lifted from the galactic disk @xcite , the distance , mass and velocity of the cloud
are now well established .
distance indicators based on stellar absorption - line bracketing , h@xmath7 flux , and the kinematic distance of gas interacting with the cloud , provide a consistent estimate of @xmath8 kpc @xcite .
the hi mass of the cloud is found to be @xmath9 m@xmath10 @xcite , with wham data indicating a comparable or greater amount in ionized hydrogen @xcite .
the velocity components have also been determined @xcite allowing the orbit of the cloud to be calculated for the first time .
the smith cloud s importance is that it provides insight into gas at a more advanced stage of accretion than the magellanic stream .
earlier studies have shown how easily gas disrupts as it passes through a hot corona @xcite .
it is therefore remarkable that a massive , high - velocity gaseous system has survived so close to the disk . in light of recent developments ,
we consider the prospect of a confining dark halo surrounding at least some hvcs to be plausible , as first proposed by @xcite , and therefore worthy of further investigation .
a number of ultra - faint dwarf galaxies have now been discovered in the galactic halo @xcite .
what is particularly striking is that the inferred dark halo mass exceeds @xmath11
m@xmath2 even for systems with very few stars @xcite .
these observations hint at the prospect of dark matter dominated galaxies containing hi that have not experienced star formation .
complex h is one such ` dark galaxy ' candidate , with a mass dominated by dark matter if it is self - gravitating @xcite .
here we suggest that the smith cloud is another candidate for a dark galaxy , illustrated in fig .
[ fig : sc ] , with the gas partly stabilised by a dark matter component .
we calculate the subhalo required for the survival of a gas cloud with properties similar to those observed in the smith cloud today . in [ sec : dm ] we discuss the properties of three possible dark matter profiles that may encompass the smith cloud . in [ sec : ms ] we discuss the orbit of the smith cloud within a realistic galactic potential . in [ sec : sac ] we present the equations that are used to calculate the ram pressure stripping of the subhalos along their orbit . in [ sec : mr ] we present the results of the semi - analytic models and discuss the implications of these results for the smith cloud . in [ sec : conc ] we summarize these results and discuss the limitations of our model . .
the tidally stripped mass today is @xmath12 m@xmath10 in dark matter .
the core size assumes an initial gas mass at apogalacticon of @xmath13 m@xmath10 , although most of this gas is distributed along the orbit of the smith cloud today .
we note that the burkert profile has no surviving core , and the nfw core is displayed in white for visibility.,scaledwidth=45.0% ]
numerical simulations of dark matter halos indicate that they tend to be characterized by two parameters , typically a radial scale length and a density .
one of the first of these dark matter profiles was inspired by rotation curves in dwarf spiral galaxies @xcite , the ` burkert ' profile , which incorporates a core of uniform density followed by a steady decline with increasing radius .
early dissipationless n - body simulations however revealed that the dark matter core is more cusp - like .
the most widely used of the cuspy profiles is the ` nfw ' profile introduced by @xcite .
but more recently , it seems that neither of these profiles are sufficiently flexible to describe what is now being seen in simulations involving 10@xmath14 particles , with a more appropriate model provided by the einasto profile @xcite .
all of these profiles may be described by a density function @xmath15 with the density at any point given by @xmath16 , where @xmath17 is the characteristic density .
similarly the mass may be described by @xmath18 and the potential by @xmath19 , where @xmath20 is the characteristic velocity of the halo .
the gas density for a hydrostatic cloud is given by @xmath21 $ ] , where @xmath22 is the isothermal sound speed of the gas , and @xmath23 is the nuclear gas density .
these profiles are defined in terms of a scaled radius @xmath24 , where @xmath25 is a scale radius .
these parameters are defined by @xcite : @xmath26^{1/2 } , \label{eq : vs}\\ r_s & = & \left(\frac{3}{4\pi\delta\rho_u}\right)^{1/3}m_{vir}^{1/3}\frac{1}{x_{vir}},\end{aligned}\ ] ] where @xmath27 is the overdensity of the dark matter @xmath28 at @xmath29 , @xmath30 g @xmath31 is the average density of the universe , @xmath32 is the virial mass , and @xmath33 is the scaled virial radius .
@xmath33 is correlated , with a standard deviation @xmath34 , to the virial mass in the virial mass range of @xmath35 m@xmath10 by @xcite : @xmath36 convenient forms for the nfw and burkert profiles are given in @xcite . in table 5 of that paper , we note that there is a typographical error in the nfw @xmath37 equation which should read @xmath38 . in table
[ table : prof ] , we give our formalism for the einasto profile , and include the burkert and ( corrected ) nfw profiles for completeness .
the three density and mass profiles are shown in fig .
[ fig : fm ] for median subhalos of virial mass @xmath39 m@xmath10 , with corresponding scale radius @xmath40 kpc and profile dependent characteristics density determined by ( [ eq : rhods ] ) .
nfw & burkert & einasto + @xmath41 & @xmath42 & @xmath43/4 $ ] + @xmath44 $ ] & @xmath45$]&@xmath46 + @xmath47 $ ] & @xmath48&@xmath49 + & @xmath50 $ ] & @xmath51 $ ] + @xmath52 & @xmath53 & @xmath54 + & @xmath55^{(3/2)(v_s / c_g)^2}$ ] & + & & @xmath56 + for dark matter subhalos moving within the galactic environment , the radial profile will be altered in two competing ways .
the condensation of baryons contracts the inner core of the subhalo while ram - pressure stripping leads to an expansion of the subhalo .
the overall impact of adiabatic contraction can be calculated iteratively using @xcite : @xmath57 = r_im_i(r_i),\ ] ] where @xmath58 is the final baryon mass distribution , @xmath59 is the final dark matter distribution and @xmath60 is the initial total mass distribution at an initial orbital radius @xmath61 . by assuming the orbits of baryonic material
do not cross , then @xmath62 where @xmath63 is the baryon to dark matter fraction .
we have investigated the impact of adiabatic contraction and find that it has a negligible influence in modelling the smith cloud .
we note that the virialization time of the dark matter halo can also be important . here
the overdensity of the dark matter compared to the average density is given by @xcite : @xmath64 where @xmath65 , and for a flat universe @xmath66 the physical density of a halo is directly proportional to this overdensity factor such that @xmath67 where @xmath68 is the central density of a halo that virialized at @xmath29 .
this factor also contributes to other halo properties such as the scale radius @xmath69 and the scale velocity @xmath70 .
we consider two models of evolution , one in which the smith cloud is infalling for the first time , hereafter the infalling orbit models , and a second model where the smith cloud has already been maximally stripped due to previous orbits , hereafter the repeated orbit models .
these both share common features : ( i ) they have the same trajectory today , ( ii ) the dark matter halo has been tidally stripped down from some larger initial mass ( @xmath71 ) in an identical fashion before our calculations commence at apogalacticon .
the important distinction is tidal stripping of the gas is possible in the infalling orbit models but not in the repeated orbit models ; in both cases , ram pressure stripping by the hot halo is important . for each case , the evolution of the smith cloud is considered for the nfw , einasto and burkert models .
the evolution of the model clouds was calculated as a function of three variables : the initial virial mass at the time of formation ( i.e. before the dark - matter halo fell into the galaxy ) , the dark matter profile at this time , and the initial hydrogen gas mass at apogalacticon .
for both the repeated orbit model and infalling orbit models , the evolution of @xmath72 model clouds were calculated , corresponding to @xmath73 logarithmically spaced virial masses in the range @xmath74 m@xmath10 and @xmath75 logarithmically spaced gas masses in the range @xmath76 m@xmath10 .
the orbit of the smith cloud was calculated using the position and velocity data from @xcite for the tip of the smith cloud : @xmath77 kpc , @xmath78 km s@xmath5 .
the form of the galactic potential is given by @xcite normalized by a circular velocity of @xmath79 km s@xmath5 at the solar circle .
fig : orb ] , we show the predicted orbit of the cloud system . in agreement with lockman et al ( 2008 ) , we find that the smith cloud has intersected the disk @xmath80 myr ago and will pass through the disk again in @xmath81 myr . for all subhalo models ,
we investigate the effects of dynamical friction on the orbit trajectory .
the formalism used is described by @xcite : we point out that the value for the circular velocity in their table 1 should be @xmath82 km s@xmath5 ( not 181 km s@xmath5 as quoted ) to be consistent with their analysis . but over the past few hundred million years , dynamical friction is found to have only minimal effect , even in the high mass limit .
this is because , once again , the impact of gas loss from the subhalo close to the disk is found to dominate the evolution of the subhalo .
we assume that any drag between the model clouds and the galactic corona is negligible and does not affect the orbit .
each model cloud is considered to be a dark matter potential well filled with gas in isothermal hydrostatic equilibrium .
we assume a primordial helium abundance @xmath83 and metallicity of @xmath84 .
we also assume the gas has a temperature of @xmath85 k and adopt an ionization fraction of @xmath86 for the smith cloud , slightly below the newly updated h@xmath87/h@xmath88 ratio in @xcite .
this temperature and ionization fraction then give a sound speed of @xmath89 km s@xmath5 .
the gas is distributed in the potential well according to the gas density profile @xmath90 , where @xmath24 is the scale radius , @xmath20 is the halo circular velocity given by ( [ eq : vs ] ) and @xmath91 is given in table [ table : prof ] . for the initial dimension of the model clouds , the sound crossing
time is @xmath92 myr , falling to about @xmath93 myr at the disk .
we therefore begin each orbit at the apogalacticon , approximately @xmath94 kpc from the galactic centre , which took place @xmath95 myr ago , and @xmath96 myr after the last transit of the disk .
we hence assume the cloud has regained hydrostatic equilibrium by this point and is the basis for our initial assumption of hydrostatic equilibrium . due to the comparatively large sound crossing time
we assume that the radial gas density profile does not change throughout the calculation , only being truncated , however , due to the lower sound crossing time at the disk we use a more restrictive condition as the cloud transits the disk ( [ sec : sac ] ) . in order to trace the long - term impact of ram - pressure stripping
, we must define the density profile of the hot coronal halo and the galactic gas disk .
the acceleration due to the stellar density profile is already included in the wolfire potential . the galactic corona is assumed to be a mix of h@xmath97he at @xmath98 k with the density calculated from the gravitational potential of @xcite , normalized to a plasma density of @xmath99 @xmath31 at @xmath100 kpc from the galactic centre @xcite .
the density profile of the hi disk was calculated using an exponential drop - off in density both in the radial and vertical direction .
we use the new prescription in @xcite which includes the vertical flaring of the outer disk .
the smith cloud was evolved through a procedure that iteratively calculates the competing processes on the dark matter confined cloud .
the initial radius of the gas comes from integrating the gas density profile over space giving ( modified from * ? ? ?
* ) : @xmath101 where @xmath25 is the scale radius of the halo , @xmath102 , @xmath103 is the gas distribution profile for constant scale velocity and sound speed .
here we have replaced @xmath23 with @xmath104 by assuming that initially the outer edge of the halo is bound by thermal pressure from the galactic halo . at each step
we calculate whether the gas at the cloud edge is prone to stripping by examining if the force applied by the external medium exceeds the maximum gravitational force in the subhalo at a given radius . for an isothermal spherical cloud ,
gas will be stripped if @xcite : @xmath105with @xmath106 and @xmath107 being the density of the galactic gas and the outer subhalo gas respectively , @xmath108 the relative velocity between these two gases , and @xmath109 the amount of mass within a radius @xmath110 . the mass within
a radius @xmath110 comprises the dark matter subhalo @xmath111 and the gas within that radius @xmath112 .
if this stripping criterion is satisfied , then a shock passes through the gas , at a speed given by @xcite : @xmath113 this gas assumed to be no longer bound and is removed from the cloud . by assuming that the shock removes a shell of gas , the outer radius of the gas changes at a rate given by @xmath114 .
as @xmath115 tends to be much larger than the gas sound speed , @xmath22 , we assume that the density profile of the gas does not change , becoming truncated at the edge of the shock . for infalling models , the tidal stripping must also be considered . at each step , the tidal radius and tidal mass of the subhalo are also calculated .
the tidal radius is given by @xcite : @xmath116\frac{m_{gal}(r)}{r^3},\ ] ] where @xmath110 is the distance between the centre of the subhalo and the galactic centre .
any gas outside of the tidal radius @xmath117 will also be removed , with the new gas radius being the minimum of the tidal radius or the radius resulting from ram pressure stripping .
the total mass can then be calculated by modifying ( [ eq : mass ] ) such that @xmath118 where @xmath119 .
the central column density is calculated using @xmath120 we iterate through the orbit in @xmath121 myr steps to determine the evolution of radius , mass and column density for the cloud / halo system .
convergence was tested by iterating through on @xmath122 myr steps for some surviving clouds , with the smaller step size producing less than @xmath123 difference in mass calculations throughout the orbit .
a special step , however , is considered when the smith cloud is moving through the galactic disk .
consider a cloud with a gas density @xmath124 , radius @xmath125 moving at speed @xmath126 through the galactic halo .
consider a gaseous disk where , at some radius , it has a local gas density @xmath127 and a half - scale height @xmath128 .
will the cloud survive its transit through the disk ?
we write down two shock conditions : one for the shock driven into the cloud by the disk , and one for the shock driven by the cloud into the disk .
the cloud survives if the shock through the cloud takes longer to propagate than the shock through the disk .
the shock propagates into the disk gas at speed @xmath129 and another into the cloud , at speed @xmath130 the shock needs to be fast enough to heat the disk gas significantly .
the key to punching a hole is that the shock must go right through the disk gas before passing through the cloud . in other words
we need roughly @xmath131 or equivalently @xmath132 .
this leads to @xmath133 it follows that , to first order , the cloud will survive if @xmath134 where @xmath135 and @xmath136 are the column densities of the cloud and the disk along the orbit respectively .
this is the kinetic energy argument presented by @xcite .
we note that the dark matter subhalo itself has negligible impact on the galactic gas .
we therefore expect that all gas in the smith cloud at a column density less than @xmath137 is stripped when the cloud moves through the disk .
as the passage through the disk occurs on a short timescale @xmath138 myr , we treat this removal of gas as instantaneous , and in addition to the normal ram pressure stripping that occurs during that time step .
we also set our model disk to rotate with equal angular velocity to the smith cloud as it passes through at @xmath139 kpc , and a static disk otherwise .
although this choice is unphysical and in both cases contradict , the observed relative velocity between the smith cloud and the ism , the cloud spends only a short time near the disk , and hence any additional gas stripped via ram pressure will contribute only a small correction to the mass lost from the cloud near the disk .
in our simulations , a large number of model clouds survive to the present day , although not all of these are similar to the smith cloud as observed today .
we interpret our results graphically through contour plots of properties derived from the mass , radius and column density of the model clouds .
the simplest properties @xmath140 mass , column density @xmath140 are displayed for the repeated orbit model nfw profile in fig . [ fig : nfwcont ] . by removing model results which are greatly dissimilar to those observed for the smith cloud
, we can find an allowed parameter space for the initial cloud properties ( fig .
[ fig : smith ] ) . in figs . 4 - 7
, we present our results in terms of the mass of the tidally - stripped subhalo rather than the mass of the initial infalling halo .
the lower limit of the allowed parameter space is found by requiring the amount of gas in the model clouds to match the observations .
the smith cloud has an hi mass of at least @xmath141 m@xmath10 @xcite , and at least as much again in hii @xcite , giving a final gas mass of at least @xmath142 m@xmath10 .
the excluded region corresponding to lower cloud gas masses is shown in green ( fig .
[ fig : smith ] ) bounded by a solid black line .
we include several upper bounds to the allowed parameter space .
the universal baryonic to dark matter mass fraction is @xmath143 from the latest microwave background observations.cdm cosmology with parameters ( @xmath144 while there may be galaxies that exceed this limit ( e.g. dwarf galaxies that form in tidal tails of strong mergers ) , it is reasonable to assume that on average most dark matter subhalos do not accrete more than this fraction .
thus , we can probably exclude all clouds with baryon to dark matter mass fractions above this limit , displayed by the red shaded region bounded by a dot - dash line in fig .
[ fig : smith ]
. in dwarf galaxies , the highest hi column densities seen outside of star - forming regions are always less than @xmath145 @xmath146 .
this critical column density is found in regions with @xmath147 of optical emission ( b. ekta 2009 , private communication ) .
since no stars are observed in the smith cloud , we include this upper limit to the hi column density , shown by the light blue region bounded by a dotted line in fig .
[ fig : smith ] .
an extensive literature demonstrates that star formation is intimately associated with the presence of molecular gas ( e.g. * ? ? ?
for a canonical hvc metallicity of about @xmath148 z@xmath10 , it is unlikely that molecular gas can be measured directly from the smith cloud with existing submillimetre telescopes @xcite .
this sets a stronger , but presumably less rigorous constraint on the allowed regions of the smith cloud ; the excluded regions are shown in purple bounded by a dot - dash line in fig . [ fig : smith ] .
the detection of molecular gas in the smith cloud would alter this limit from an upper limit , to a lower limit and hence alter the allowed parameter space to exist only in the purple region .
all of the upper limit bounds shown in fig .
[ fig : smith ] have a strong dependence on the gas distribution within the model . here
we have assumed that the gas was initially in hydrostatic equilibrium with the profile only being truncated and not altered further ; shocks that substantially modify the profile are associated with gas that gets stripped .
the nfw and einasto profiles restrict the allowed regions of the cloud to a narrow range of initial gas masses and subhalo masses , with the subhalo mass range restricted to @xmath149 m@xmath10 and gas mass initially @xmath150 m@xmath10 , although we note that the allowed gas mass range is much larger at smaller subhalo masses than at the higher mass end . by comparison to the gas mass present today
, the smith cloud will have lost between @xmath86 and @xmath151 of the gas present at apogalacticon .
a large fraction of this gas loss in all model clouds is due to the transit of the disk suggesting , suggesting that the smith cloud may not survive the next passage through the disk .
the tidal stripping of dark matter does not substantially affect the gas contained in the model clouds , with the gas contained within the protected core of the potential well . in the infalling orbit models
, the smith cloud will also have lost approximately @xmath86 of the dark matter mass as well to tidal stripping , although most of this mass loss will have occurred before the semi - analytic calculations were begun .
this scenario of an hvc infalling for the first time , possibly after interacting with a third body , such as a dwarf galaxy , has the potential to increase the baryonic to dark matter ratio above the universal fraction , particularly if the orbit avoided dense sections of the disk .
the burkert profile retains a lower column of gas , due to a lower central density , with the effect that the region is not as well constrained as in the case of a nfw or a einasto profile .
it is this lower column that leads to a higher mass requirement in order to survive traversal of the disk , which is dictated by the column density of the cloud .
we display the mass history of three infalling orbit model clouds corresponding to the same input parameters for each dark - matter profile ( m@xmath152 m@xmath10 [ m@xmath153 m@xmath10 ] , m@xmath155=@xmath13 m@xmath10 ) in fig .
[ fig : massloss ] .
the hydrostatic pressure bound gas radius is located within the tidally protected core of the dark matter subhalo , and therefore the mass history of the repeated orbit model with the same initial conditions will be similar .
as the subhalo loses dark matter to tidal stripping and gas to ram pressure stripping at different rates , the baryon to dark matter fraction will alter over the course of the orbit .
this change , displayed in fig .
[ fig : bdmratio ] for the infalling orbit model clouds corresponding to the input parameters ( m@xmath152 m@xmath10 [ m@xmath153 m@xmath10 ] , m@xmath155=@xmath13 m@xmath10 ) is particularly apparent at the transit of the disk where a large amount of gas is lost but the dark matter component remains unchanged . noting that the dark matter halo will have already been tidally stripped at the beginning of the evolutionary model suggests that hvcs may undergo radical changes in their baryon to dark matter fraction along their orbits .
we have modelled the smith cloud by assuming that it is supported against total disruption by a dark - matter subhalo . by considering the evolution of this system over the past @xmath95 myr
, we are able to compare our predicted properties to the known properties of the smith cloud today . under the assumption of dark - matter confinement
, the subhalo encompassing the smith cloud has a tidal mass of @xmath156 m@xmath10 at the present time .
even before this subhalo arrived at its present orbit , it will have lost @xmath157% of its dark matter and a comparable amount of progenitor gas due to tidal and ram pressure stripping by the galaxy . in this respect , the likelihood of an infalling ( dark - matter confined ) `` smith cloud '' event is comparable to the rate of dwarf galaxy infall , i.e. roughly one event per gyr . if the nfw and einasto profiles are appropriate descriptions of dark matter halos , we find that the smith cloud can only exist in a narrow strip of the available parameter space defined by the initial dark matter subhalo mass and the initial gas mass .
this remains true with and without dark matter stripping .
this narrow strip also appears , albeit shifted with respect to the present axes , if the galactic halo is considered to be @xmath158 over or underdense from our model setup , or if the dark matter containing subhalo was virialized at an earlier redshift and correspondingly is denser than a subhalo virialized today .
consistent with our picture , the observed wake in fig .
1 presumably resulted from ablation processes due to the impulsive shock of the disk transit .
we note with interest that the wake flares ( i.e. expands transverse to the direction of the wake ) beyond the putative dark matter core ( fig .
1 ) . in our picture , this can be understood in the context of a transsonic flow .
the smith cloud is moving at up to 300 km s@xmath5 with respect to a hot halo with adiabatic sound speed @xmath159200 km s@xmath5 making the cloud - halo interface mildly supersonic .
once the stripped gas is left behind , it must undergo mixing with the hot halo through kelvin - helmholtz instabilities , causing the stream to expand perpendicular to the flow .
the temperature of the hi gas is predicted to rise along the wake . unless the confining core - dominated
, dark matter halo is very massive , we do not expect the declining potential seen by the wake material to influence the degree of flaring .
these issues are being addressed in new hydrodynamical simulations of the smith cloud that we will present elsewhere . with respect to further constraints on shock interactions
, we expect that shock signatures at uv to x - ray wavelengths will have largely faded away , and the hi `` hole '' at the crossing point will have been substantially stretched by differential shear . for cloudlets smaller
than 100 pc , thermal conduction due to the halo corona @xcite and the galactic radiation field convert the ablated gas to a clumpy plasma .
we therefore strongly encourage more extensive mapping of the hi and the ionized gas .
as we discuss elsewhere @xcite , it is very difficult to see how the smith cloud , like several other large hvcs , could have come in from , say , the distance of the magellanic stream .
this was our motivation for considering a cloud stabilised against disruption by a confining dark matter halo .
in essence , we have considered the prospect that the smith cloud constitutes a ` dark galaxy ' where star formation never took place .
any evidence of a co - moving stellar population would have profound consequences .
an essential requirement for star formation to occur appears to be the presence of molecular gas ( e.g. * ? ? ?
we envisage much deeper searches for the presence of molecular gas in compact hvcs , in particular , with the newly commissioned _ cosmic origins spectrograph _ on the _ hubble space telescope_. our model is not unique .
the cloud may have been dislodged from the outer disk or confining potential of an infalling dwarf galaxy .
a cloud metallicity of [ fe / h]@xmath4 - 1 may be appropriate in either scenario but there are no obvious candidates at the present time .
the interloper must be on a prograde orbit which rules out some infalling dwarfs ( e.g. @xmath160cen ; * ? ? ? * ) but conceivably implicates disrupting dwarfs like canis major , assuming these were still losing gas in the recent past .
mn is supported by an australian postgraduate award .
jbh is supported by a federation fellowship from the australian research council .
we are indebted to the referee for a very constructive and well considered report .
in addition we are grateful to ralph sutherland for his helpful comments on the paper .
bekki , k. , & freeman , k. c. 2003 , , 346 , l11 bland - hawthorn , j. , veilleux , s. , cecil , g. n. , putman , m. e. , gibson , b. k. & maloney , p. r. 1998 , , 299 , 611 bland - hawthorn , j. , sutherland , r. , agertz , o. , & moore , b. 2007 , , 670 , l109 bland - hawthorn , j. in iau symp .
254 , the galaxy disk in cosmological context , ed .
j. andersen , j. bland - hawthorn , & b. nordstrm ( cambridge : cambridge univ .
press ) , 241 blitz , l. , spergel , d. n. , teuben , p. j. , hartmann , d. & burton , w. b. 1999 , , 514 , 818 blumenthal , g. r. , faber , s. m. , flores , r. , & primack , j. r. 1986 , , 301 , 27 bryan , g. , & norman , m. 1998 , , 495 , 80 burkert , a. 1995 , , 447 , l25 calzetti , d. , & kennicutt , r. c. 2009 , arxiv:0907.0203 ekta , chengalur , j. n. , & pustilnick , s. a. 2008 , , 391 , 881 geha , m. , willman , b. , simon , j. d. , strigari , l. e. , kirby , e. n. , law , d. r. , & strader , j. 2009 , , 692 , 1464 hayashi , e. , navarro , j. f. , taylor , j. e. , stadel , j. , & quinn , t. 2003 , 584 , 541 heitsch , f. , & putman , m. e. 2009 , , 698 , 1485 hill , a. s. , haffner , l. m. , & reynolds , r. j. 2009 , , 703 , 1832 jiang , i. , binney , j. 2000 , , 314 , 468 kalberla , p. m. w. , & dedes , l. 2008 , , 487 , 951 kirby , e. n. , simon , j. d. , geha , m. , guhathakurta , p. , & frebel , a. 2008 , , 685 , l43 lockman , f. j. 1984 , , 283 , 90 lockman , f. j. , benjamin , r. a. , heroux , a. j. , & langston , g. i. 2008 , , 679 , l21 mazure , a. , & capelato , h. v. 2002 , , 383 , 384 mccarthy , i. g. , frenk , c. s. , font , a. s. , lacey , c. g. , bower , r. g. , mitchell , n. l. , balogh , m. l. , & theuns , t. 2008 , , 383 , 593 mckee , c. f. , & cowie , l. l. 1977 , , 215 , 213 moore , b. , & davis , m. 1994 , , 27 , 209 navarro , j. f. , frenk , c. s. , & white , s. d. m 1997 , , 528 , 607 putman , m. e. , bland - hawthorn , j. , veillux , s. , gibson , b. k. , freeman , k. c. , & maloney , p.
r. 2003 , , 597 , 948 quilis .
v. , & moore , b. 2001 , , 555 , l95 richter , p. and sembach , k. r. and wakker , b. p. and savage , b. d. 2001 , , 562 , l181 simon , j. d. , blitz , l. , cole , a. a. weinberg , m. d. , cohen , m. 2006 , , 640 , 270 simon , j. d. , & geha , m. 2007 , , 670 , 313 smith , g. p. 1963
netherlands , 17 , 203 springel , v. , wang , j. , vogelsberger , m. , ludlow , a. , jenkins , a. , helmi , a. , navarro , j. f. , frenk , c. s. , & white , s. d. m. 2008 , , 391 , 1685 sternberg , a. , mckee , c. f. & wolfire , m. g. 2002 , , 143 , 419 strigari , l. e. , bullock , j. s. , kaplinghat , m. , simon , j. d. , geha , m. , willman , b. , & walker , m. g. 2008 , , 454 , 1096 sofue , y. , kudoh , t. , kawamura , a. , shibata , k. , & fujimoto , m. 2004 , , 56 , 633 wakker , b. p. , & van woerdan , h. 1997 , , 35 , 217 wakker , b. p. , york , d. g. , wilhelm , r. , barentine , j. c. , richter , p. , beers , t. c. , ivezi , vz , & howk , j. c. 2008 , , 672 , 298 wolfire , m. g. , mckee , c. f. , hollenbach , d. , & tielens , a. g. g. m. 1995 , , 453 , 673 zait , a. , hoffman , y. , & shlosman , i. 2008 , 682 , 835 | the smith cloud is a massive system of metal - poor neutral and ionized gas ( m@xmath0 @xmath1 m@xmath2 ) that is presently moving at high velocity ( v@xmath3 @xmath4 300 km s@xmath5 ) with respect to the galaxy at a distance of 12 kpc from the sun .
the kinematics of the cloud s cometary tail indicates that the gas is in the process of accretion onto the galaxy , as first discussed by @xcite . here
, we re - investigate the cloud s orbit by considering the possibility that the cloud is confined by a dark matter halo .
this is required for the cloud to survive its passage through the galactic corona .
we consider three possible models for the dark matter halo ( nfw , einasto , burkert ) including the effects of tidal disruption and ram - pressure stripping during the cloud s infall onto and passage through the galactic disk .
for the nfw and einasto dark - matter models , we are able to determine reasonable initial conditions for the smith cloud , although this is only marginally possible with the burkert model .
for all three models , the progenitor had an initial ( gas+dark matter ) mass that was an order of magnitude higher than inferred today . in agreement with @xcite ,
the cloud appears to have punched through the disk @xmath4 70 myr ago . for our most successful models ,
the baryon to dark matter ratio is fairly constant during an orbital period but drops by a factor of @xmath6 after transiting the disk .
the cloud appears to have only marginally survived its transit , and is unlikely to retain its integrity during the next transit @xmath4 30 myr from now . |
the giant region ngc 3603 , located at a distance of about 6 kpc , shows the densest concentration and largest collection of visible massive stars known in our galaxy .
recent hst images of @xmath602 angular resolution @xcite have shown that ngc 3603 consists of three wolf - rayet ( wr ) stars and @xmath670 o - type stars , with an estimated 4050 of these stars located in the central @xmath630@xmath730 ( 1 pc @xmath7 1 pc ) region .
the three h - rich wr - stars of subtype wnl are also located within @xmath61 of each other in the central core and are the brightest members of the cluster .
they are believed to be massive main sequence stars which , like the bright stars in r136a , drive very strong winds @xcite .
ngc 3603 has also been shown to be a seat of active star formation outside the main cluster region ( e.g. @xcite ) .
in particular , three distinct dense cometary nebulae were detected there using hst and vlt : are these ultra compact regions ( uchr ) or proplyds ?
uchrs are small ( @xmath80.1 pc ) nebulae , _ internally _ photoionized by a deeply embedded , massive star .
even though the embedded stars are very luminous , they are invisible at optical wavelengths because of the surrounding dust ; instead they show strong far - infrared emission .
their radio continuum emission is often associated with oh and h@xmath9o masers .
the morphology of uchrs in high resolution radio continuum images ranges from spherical , cometary , core - halo , shell , to irregular @xcite .
in contrast , a proto - planetary disk ( proplyd ) is a phenomenon describing a low - mass , young stellar object ( yso ) with a circumstellar disk , embedded in a dense ( neutral and ionized ) envelope that is being _ externally _ photo - evaporated by the ultraviolet radiation from one or more massive stars .
these low - mass stars are not able to ionize their surroundings significantly . the disks in these systems are either observed directly ( as in orion where the disks are seen directly or in silhouette against the bright background nebula @xcite ) , or are inferred because in almost all of the proplyds ( young ) low - mass stars are visible at optical or near - infrared wavelengths ( see e.g. @xcite ) . at low angular resolution proplyds
look like uchrs , but at high resolution , optical and near - infrared images reveal their different nature .
so most proplyds would have been classified as uchrs prior to those observations . in some cases
the distinction between the two classes remains difficult .
given the fact that there are about four times more uchrs than expected @xcite from star formation rates and the potential problems identifying host stars for uchrs ( e.g. some may not have enough fir flux to account for an internal ob star ) , it may well be that a large fraction of catalogued uchrs are mis - identified proplyds .
interferometric radio continuum observations of proplyds are needed to determine their structure , spectral indices and thus the nature of their emission , as well as mass loss rates and extinctions when compared to measured fluxes .
proplyds were first identified in the orion nebula , where over 150 of them are known @xcite .
two proplyds have been identified in more distant nebulae : one in ngc 2024 by @xcite and one , g5.971.17 , in the lagoon nebula by @xcite .
the three objects , hereafter referred to p1 , p2 and p3 , in ngc 3603 @xcite are the biggest , youngest and most massive ones found _
so far_. they are also the most distant known .
these emission nebulae are clearly resolved in the hst / wfpc2 observations , and share the overall morphology of the proplyds in orion .
all three nebulae are rim - brightened and tear - drop shaped with the tails pointing away from the central ionizing cluster . according to @xcite the brightest object ( p1 ) , which has a projected distance of 1.3 pc from the cluster , has the spectral ( excitation ) characteristics of a uchr .
optical spectra reveal the presence of an underlying , heavily reddened continuum source , which is also confirmed by near - infrared vlt / isaac observations .
the wfpc2 observations show that only the outermost layer is ionized whereas the interior is neutral .
the morphology of p1 is described as a heart - shaped head with a collimated structure in between , which can be understood as the superposition of two individual proplyds .
in contrast ,
p2 and p3 , located at projected distances of 2.2 pc and 2.0 pc from the stellar cluster , respectively , show approximately axisymmetric morphologies .
no embedded disk or central star has been detected so far in any of these nebulae , preventing a clear identification as proto - planetary disks . the optical point source ( see fig .
2c ) close to p3 is probably not physically linked to the nebula @xcite .
the structures in ngc 3603 are about two orders of magnitude fainter than typical uchrs , but have a typical extent of 9000 au with tails extending to 21000 au , much larger in size than the proplyds in orion .
recent 3.4 cm radio continuum and recombination line measurements of ngc 3603 by @xcite , which were focussed on abundance measurements and the bright continuum emission from the ionized gas in this region , have an angular resolution of @xmath67 and a @xmath10 sensitivity of 55 mjy . by obtaining high sensitivity and high angular resolution ( @xmath612 ) observations we primarily aimed to study the radio emission of the winds from many of the early - type stars in the cluster , as well as to detect and resolve the s and other gaseous regions in the cluster periphery .
this paper will focus on the sources only , which have been detected and are shown to be clearly resolved with the . a subsequent paper containing a detailed study of the whole ngc
3603 region based on our observations will follow .
radio continuum observations of ngc 3603 were made with the australia telecope compact array ( ) in the 6a , b , c and d configurations in five observing runs in february , april , june , september and november 2000 with a total of 60 hours assigned observing time .
the observing frequencies were 4.8 ghz ( 6 cm ) and 8.64 ghz ( 3 cm ) with a bandwidth in each case of 128 mhz .
detailed observing parameters are given in table 1 .
the three data sets were combined , reduced and analysed in miriad using standard procedures .
the full stokes parameters were measured .
the flux density scale was calibrated using observations of the primary calibrator , 1934638 , assuming flux densities of 2.84 and 5.83 jy at 3 and 6 cm , respectively .
a primary beam correction was carried out . to obtain high angular resolution and filter the extended emission to emphasize the small scale structure we used uniform weighting of the _ uv _ data while omitting the shortest baselines , up to @xmath11 , respectively .
we find that for a @xmath11 cutoff all bright extended nebula emission is removed from the s even at 6 cm .
1 shows an example of a combined 3 cm radio map where a @xmath12 cut of @xmath13 was chosen . a zoom on the three s
is shown in fig . 2 .
for the 6 cm maps a @xmath12 cut of @xmath11 was used . the average r.m.s . in these combined radio maps
was approximately 0.2 mjy at 3 cm and 0.4 mjy at 6 cm , but varies throughout the map .
the r.m.s .
levels near the s , which lie in rather empty regions , were @xmath14 mjy at 3 cm and @xmath15 mjy at 6 cm .
the dynamic range in the 3 and 6 cm images is about 1:100 . fitting a 2d - gaussian to the dirty beam
gives fwhm values of 094 @xmath7 077 to 106 @xmath7 087 at 3 cm and 153 @xmath7 123 to 183 @xmath7 149 at 6 cm depending on the @xmath12 cut employed . in fig .
2 we have restored the cleaned maps with a circular gaussian beam of width 1 at 3 cm and 2 at 6 cm . to create spectral index maps between 3 and 6 cm we convolved the 3 cm map with a gaussian appropriate to achieve the resolution of the beam fwhm at 6 cm .
the spectral index maps were then derived from the 6 cm and convolved 3 cm maps using a clip value of 1.0 mjy beam@xmath3 , corresponding to approximately @xmath16 at 3 cm and @xmath10 at 6 cm ( statistical errors ) .
error maps for the spectral index maps were calculated using error propagation. we note , however , that spectral indices of extended sources as obtained from interferometric data have to be regarded with caution .
the three sources found with the at 3 and 6 cm are clearly resolved , showing a head - tail extent of @xmath64 ( see fig . 2 ) .
p3 shows the most pronounced head - tail structure with a 3 cm flux density ratio between head and tail of about 10:1 .
the tail is very well defined and at least 2 long , pointing away from the central star cluster .
unfortunately , p3 is rather faint in the low - sensitivity hst broad band image shown by @xcite ; it is located outside the region of their high - sensitivity image .
the three s have a cometary or head - tail shape at cm - radio wavelengths , matching their appearance in optical images ( see fig . 2 )
. integrated radio flux densities and peak fluxes are derived by summing over the whole s , and by fitting circular gaussians to each object .
statistical flux uncertainties are estimated as map r.m.s .
times number of beamsizes covered by the source .
dependent on the method of flux determination , beam size and @xmath12 cut employed to the maps ( 5 , 10 , 15 , 20 , 25 k@xmath17 ) the resulting flux densities vary .
the standard deviation of this variation is considered as systematic uncertainty .
we found that the systematic uncertainties are always larger than the statistical error . in table 2
we report the average fluxes and their uncertainties of all 3 nebulae , together with their average spectral indices derived from the integrated fluxes .
circular gaussians are used to at least attempt disentangling the head and tail structures in these cometary - shaped objects .
the radio heads of the s are typically 23 in diameter while the total head - tail extent is @xmath64 .
p3 shows the most pronounced head - tail structure of the three ( most likely due to the viewing angle ) and the smallest head ( diameter @xmath62 ) among all three s , with the largest peak flux at 3 cm while the integrated head flux is comparable . its head - to - tail flux ratio is about 10:1 , whereas in the other two s part of the tail may be confused with the head flux density .
both , p1 and p2 , appear inclined with respect to the plane of the sky , with p1 inclined at a larger angle than p2 .
this can be seen in fig . 2 where we have indicated the projected direction to the ionizing source .
3 shows the residual radio continuum image of p1 after removing a single circular gaussian component .
the residual map of p1 shows two excess emission sites , one corresponding to the tail , and a second one north of it .
recent images indicate that p1 is composed of two separate cometary - shaped objects @xcite .
we interpret the northern excess as the second head .
a two component gaussian fit to p1 at 3 cm gives peak fluxes of 4.6 mjy beam@xmath3 and 1.4 mjy beam@xmath3 , and integrated flux densities of 9.9 mjy and 5.7 mjy for the two heads , with a separation of the two peaks of 13 .
a two component model could not be applied to the lower resolution 6-cm data .
the tails of all three s are directed away from the star cluster , approximately coincident with the tails in within the positional uncertainty of the radio / optical pointings ( see fig . 2 ;
note the shift of 05 in declination and right ascension of the hst - image which we have employed in these figures ) .
the uncertainty of the radio position is @xmath6015 while hst has a pointing uncertainty of @xmath18 .
the present data were obtained in all four stokes parameters .
no polarization signal above 3@xmath19 was detected .
possible flux variability on timescales @xmath20 days expected from e.g. flaring pre - main sequence ( pms ) stars with non - thermal radio spectra , can not be tested with the present data .
the radio continuum fluxes ( see table 2 ) are a factor of 9 to 21 larger than those predicted by @xcite assuming optically - thin thermal bremsstrahlung as the radio emission mechanism @xcite , which scales with the ( extinction corrected ) flux . the neutral part of the envelope of the may attenuate photons from the far side of the object depending on the viewing angle .
however , @xcite have estimated this effect to be less than 25% of the total light for an isotropic ensemble of proplyds .
furthermore , the reasonable coincidence of the sizes of the s in the radio and optical band suggests that this effect is not important here . for p1 and p3 the integrated fluxes at 6 cm are higher than that at 3 cm , whereas similar flux densities were predicted .
the average spectral index of each object , shown in table 2 , is calculated on the basis of the integral fluxes at 3 cm and 6 cm in this table .
modeling the uv distribution of the 6 cm map to match the uv distribution at 3 cm gives similar results .
the average spectral indices of p1 and p3 are negative ( p3 only marginally ) , between @xmath21 and @xmath22 ( @xmath23 ) , indicating non - thermal emission , while the spectrum of p2 ( @xmath24 ) is in agreement with the expectation from optically thin thermal bremsstrahlung .
hydrodynamic simulations predict spectral indices of @xmath25 near the ionization fronts , which can increase up to @xmath26 when considering in addition a hydrodynamically collimated jet ( s. richling , private communication ) .
negative spectral indices can not be explained by this model .
because of the high resolution and sensitivity in the radio maps , we were able to generate spectral index maps for all three s. fig .
4 shows examples of the spectral index maps .
here we used the radio maps shown in fig . 2 and constrain ourselves to brightness densities greater than 1.0 mjybeam@xmath3 , which corresponds to roughly 5 @xmath19 at 6 cm and 10 @xmath19 at 3 cm ( based on statistical error estimates ) .
the radio emission zone within each appears to be inhomogeneous .
the spectra in at least two of the sources appear steeper in the tail than in the head , with a tendency of positive spectral indices located towards the region facing the ionizing star cluster .
@xcite have derived the reddening towards ngc 3603 from ubv photometry , and found @xmath27 mag for the cluster core . however , as can be seen in their data , the visual extinction is not uniform across the whole star forming region , but increases significantly with distance from the star cluster , mainly towards the north and the south .
this is confirmed by recent vlt / isaac data on the basis of 4750 stars , showing that the extinction can be as high as a@xmath28 mag ( brandner , private communication ) .
for example , star mtt 68 , close to p3 , has a@xmath29 mag , whereas star mtt 81 near p1 has a@xmath30 mag , and star mtt 79 close to p2 has extinction a@xmath31 mag .
here we use @xmath32 as used by melnick et al .
( 1989 ) for ngc 3603 .
thus , it appears likely that p2 and p3 suffer from stronger foreground extinction than the stars in the cluster center which have on average a@xmath33 mag . using the ratio between radio flux density and flux as given by mccullough et al . , ( 1995 ; their equation 5 ) and assuming that the radio flux density at 3 cm is dominated by thermal bremsstrahlung , we predict the flux using our 3 cm radio continuum data .
the predicted fluxes are then compared to the fluxes measured by @xcite and used to derive lower limits for the extinction , @xmath34 .
( if ( gyro-)synchrotron radiation dominates the radio emission , which comes from an emitting region as specified in sect . 4.4 , the derived extinction @xmath34 would be higher , with the exact value depending on the magnetic field . )
table 3 shows the calculated fluxes ( assuming @xmath35 k ) and extinctions .
note the uncertainty in comparing different areas for which the fluxes and the radio continuum fluxes have been obtained .
@xcite give the surface brightness of the heads , measured for a 05 aperture radius , whereas we quote peak fluxes per beam where the beam size is @xmath610@xmath710 .
for @xmath36 , we find a@xmath37 for p1 and a@xmath38 for p2 .
we find thus significant excess extinction for both s , which may be intrinsic to the s themselves , otherwise the extinction estimate derived from the stars is underestimated .
estimates for the electron / ion density inside the heads can be obtained from our radio data if thermal bremsstrahlung dominates the radio emission .
this is likely to be the case for p2 and maybe p3 , while for p1 the derived values should be considered as upper limits .
the bremsstrahlung intensity in the radio domain for a thermal , ionized h - gas of constant density in a sphere is @xcite @xmath39 where @xmath40 is the source distance in kpc , @xmath41 is the apparent radius of the source in arcsec , @xmath42 and @xmath43 are the densities of the thermal electrons and ions , respectively , in @xmath44 @xmath5 , @xmath45 is the observing frequency in ghz and @xmath46 is the electron temperature of the ionized gas in @xmath44 k. @xmath47 is the ratio of thermally emitting to total volume of the source , and is introduced to account for the possibility of smaller emitting volumes than the apparent extent of the source .
this is likely because only the outermost layers in proplyds are believed to be ionized .
we estimate the proplyd heads @xmath6 2 - 3 in diameter . for @xmath48 and @xmath49
we estimate the electron density by comparing the expected radio flux with the observed 3 cm proplyd head flux densities , and find @xmath50(9 , 7 and 13)@xmath51 @xmath5 for p1 , p2 and p3 , respectively .
( the corresponding emission measures are em = 5.9 , 3.5 and 9.3 @xmath52 @xmath53 pc . ) if the emitting volume is smaller than the total source volume , the electron density and emission measure increase accordingly . from spectral considerations
the emission must be optically thin , i.e. free - free absorption may set in at frequencies lower than 4.8 ghz .
thus values of em can not exceed @xmath6 several @xmath54 @xmath53 pc , in order to avoid free - free absorption at frequencies above 4.8 ghz .
this constrains the size of the thermally emitting volume of the 3 s to at least 4% , 2% and 9% of the total source volumes , and consequently puts limits on the thermal electron densities in the sources : 4.5 , 4.9 and 4.3 @xmath55 @xmath5 for p1 , p2 and p3 , respectively .
note that these numbers should be considered as upper limits if the observed radio flux is dominantly of non - thermal origin .
these numbers are in reasonable agreement with @xmath56 @xmath5 derived from [ sii ] line ratios by @xcite . for the orion proplyds densities of order @xmath57 @xmath5
have been found @xcite .
note that our derived numbers suffer from uncertainties due to the unknown geometrical projection angle of the s , and hence the true source diameters .
mass loss from the s occurs through external heating of the envelope by the far ultraviolet ( fuv ) radiation field from hot stars , especially from the core of ngc 3603 , and subsequent evaporation . knowing the electron density , the mass - loss rate in a wind is given by @xmath58 , with @xmath59 the system radius , @xmath60 the ion mass and @xmath61 the evaporation velocity . assuming a pure hydrogen gas , @xmath62
, we find @xmath63year@xmath3 for the heads of all three s , where @xmath64 is the wind velocity in units of 20 km s@xmath3 .
@xcite found evaporation flow velocities of order 1025 km s@xmath3 .
the derived upper limits for the mass - loss rates are of the same order as estimated by @xcite , and about two orders of magnitude higher than the typical values for the orion proplyds . a rough estimate of the mass reservoir of the ngc 3603 s
can then be derived by multiplying the mass loss rate with an estimated evaporation time scale of @xmath65 years @xcite .
this gives a mass for the objects of order 110 , 101000 times larger than for the orion proplyds .
the radius of the putative disks in these massive objects can be estimated using the evaporation model of @xcite which considers the mass reservoir in a disk - like shape .
we find an approximate disk radius of 2000 au ( @xmath66 ) for all three ngc 3603 s : @xcite found @xmath63400 au from hydrodynamical simulations .
the s in ngc 3603 are 2030 times larger and much more spectacular than those in orion , which were detected at 2 and 20 cm with the vla ( see @xcite and references therein ; @xcite ) .
orion contains numerous dense clumps , originally classified as uchrs , but now known to be proplyds .
most orion proplyds have central disks , sometimes seen only in silhouette against the background light . on recent hst - images
dozens of jets powered by young stars embedded in proplyds have been found @xcite .
all orion proplyds detected at radio wavelengths are non - variable thermal emitters with spectral indices between 0.1 and 0.2 @xcite , and located within 0.04 pc of the dominant ionizing o - star ( @xmath67 ori c ) .
the non - thermal orion radio sources show strong variability , and are mostly associated with visible pre - main sequence ( pms ) stars .
their non - thermal spectra are usually explained as due to stellar flaring activity .
g5.971.17 ( @xmath68 = 1.8 kpc ) was originally classified as an ultracompact region , but is now thought to be a proplyd at @xmath69 au projected distance from the o7 star herschel 36 in the center of m8 , the lagoon nebula @xcite .
eso near - infrared , hst optical and vla radio continuum observations indicate that g5.971.17 is a young star surrounded by a circumstellar disk that is likely being photo - evaporated by her 36 , similar to the proplyds in orion .
the previous hypothesis suggested that g5.971.17 is a uchr intrinsically ionized by an embedded b0 star .
however , @xcite showed that g5.971.17 is predominantly externally ionized , and thus the spectral type of the embedded star should be later than b5 .
the flux over 06 is consistent with the @xmath70 cm appearance .
optical and near infrared ( nir ) continuum images show the central star is displaced from the peak of the bow shock by 0125 .
nir photometry of g5.971.17 revealed that its central star is extremely red , which can not be explained by extinction laws that use spherical matter distributions .
this supports the idea of the circumstellar matter surrounding the central star of g5.971.17 being arranged as a disk .
the cm - radio spectrum of g5.971.17 appears flat , which is interpreted as optically thin free - free emission @xcite .
no oh or h@xmath9o masers , which are often associated with uchrs , are known in g5.971.17 . in comparison ,
the ngc 3603 s appear rather peculiar and spectacular .
not only are they the largest and most massive ones found so far . exposed to an extremely strong fuv radiation field
, they also suffer the largest mass losses , though their distances from the ionizing source are larger than for the orion or lagoon nebula proplyds .
no embedded source nor disk has been found for any of the ngc 3603 s. at cm - wavelengths they are so far the only s showing non - thermal radio emission . the spatial distribution of the ngc 3603 s is also peculiar .
they are not only distributed apparently along a straight line ( see sect . 4.3 for further discussion ) , but all three also lie clearly to the north of the ionized gas region and avoid the region north - east and south - west from the stellar cluster .
this is in contrast to the orion proplyds which do not appear to be located in preferred regions of the nebula ( e.g. @xcite ) .
since sequential star formation proceeding from north to south seems evident in ngc 3603 ( e.g. @xcite ) , this suggests that the structures in this region might have emerged just recently .
a possible way to prove that the s have indeed been ionized _ externally _ by the star cluster is to show that the observed radio fluxes for the s are consistent with the number of ionizing photons they receive from the cluster stars . for simplicity
we assume thermal radio emission , and note that this leads to an upper limit for the number of ionizing photons needed to produce the observed radio flux density . in this case
the brightness temperature is directly proportional to the number of ionizing photons . because ngc 3603 contains a large number of stars that are extremely hot and luminous they provide an ionizing flux that is several orders of magnitude larger than that in orion or m 8 ( see table 4 ) . the expected brightness temperatures at the projected distances of 1.3 pc , 2.2 pc and 2.0 pc of the s from the ionizing source due to a lyman flux of @xmath71photons s@xmath3 @xcite are 200 k , 70 k and 90 k at 3 cm assuming @xmath72
this is only slightly higher than our estimates derived from the 3 cm maps of 40 k , 30 k and 70 k for the three s , respectively .
we can also compare the number of ionizing photons required to deliver the observed flux densities at the location of the s , and compare this number with the number of lyman photons they receive from the star cluster .
noting that the s subtend a fraction of ( 6 , 3 and 2)@xmath73 , of the total solid angle seen by the cluster stars , they thus receive ( 6 , 3 and 2)@xmath74 photons s@xmath3 , respectively . on the other hand , the required number of ionizing photons at the location of each to deliver the observed radio flux density is about @xmath75 photons s@xmath3 in all three cases .
these comparisons show that the number of required lyman photons is slightly lower than provided by the star cluster , and thus would even allow an additional dust attenuation and/or , more likely , longer linear distances between the s and the ionizing source in the external ionizing scenario
. significant ionizing power from within the s appears therefore unlikely , which puts the spectral type of any central stars to later than @xmath6b1 , and hence mass to below @xmath76 .
the present radio data as well as the recently published hst / vlt - images do not provide any direct evidence that the observed objects in ngc 3603 host any ( proto-)stellar objects or disks . to a limiting @xmath77-magnitude of 18 , no circumstellar
disk or central stellar object in any of the three s has been detected by hst+vlt observations .
instead the cometary - shaped nebulae may rather be massive emission regions photo - ionized by the radiation from the star cluster . @xcite
have presented model calculations for such clumps , called fliers or ansea which are commonly located along the major axis towards the ionizing source .
indeed , all three s apparently lie approximately along one straight line , with the outflow source , sher 25 , located roughly midway between p1 and p3 . although the bipolar outflow from sher 25 appears not to point in the direction of any of the three sources , the ring - like structure around sher 25 , roughly perpendicular to its jet , has been shown to expand with a velocity of @xmath78 km s@xmath3 @xcite , and could in principle be related to the origin of the clumps .
in addition , models such as the flier model which are not based on disk - like mass distributions typically give short evaporation times , inconsistent with the distance of the s from sher 25 ( w. brandner , private communication ) and the approximate date of the mass - loss event about 6630 years ago in sher 25 @xcite .
thus , the existence of such features at large distances from the cluster centre appears to favor models where most of the material is concentrated in compact structures such as disks .
the actual presence of disk - like structures is however not confirmed . the significantly non - thermal spectrum ( @xmath79 ) from at least one of the three ngc 3603 s ( p1 ; p3 seems to be marginally non - thermal ) appears to be steepest ( @xmath80 ) in the tail , slightly flatter ( @xmath81 to 0.1 ) in the head , and have a tendency of positive spectral indices ( @xmath82 ) located towards the region facing the ionizing star cluster rather than away from it .
this is not expected from non - thermal radio emission from a wind wind collision region .
the non - thermal radiation must originate rather from inside the s , either from their putative central stars hidden behind thick material , the surrounding disks or the envelopes . a stellar origin analogous to active star coronal emission seems to be unlikely because of the extended nature of the observed source .
non - thermal radio spectra could be produced by a population of energetic particles emitting synchrotron , gyrosynchrotron or gyroresonance radiation in a magnetic field .
for gyrosynchrotron radiation the emission is concentrated at harmonic numbers @xmath83 ( where @xmath84 ghz is the gyrofrequency , @xmath85 is the magnetic field strength in gauss ) , and thus magnetic fields strengths of order @xmath86 g are required .
these high field values can not be present in extended sources , which leaves synchrotron radiation as the only plausible emission mechanism .
disk or envelope , whatever the ultimate source might be , the negative spectral indices suggest optically thin non - thermal processes as the dominant radiation mechanism in p1 , and maybe p3
. one could argue that the shock created by the evaporation flow predicted in proplyd models could accelerate charged particles within the objects .
other possibilities may be inflow - outflow activity causing shocks during the early stage of star formation , compression and reconnection of magnetic fields in the collapsing envelope , or magnetic reconnection in star disk interactions in young stellar objects leading to production of energetic particles .
magnetic fields are known to exist in molecular clouds @xcite , and therefore it appears plausible that gas clumps still contain `` fossil '' fields . in a magnetized plasma ,
acceleration due to particle collisions and subsequent bremsstrahlung can often be negligible in comparison with acceleration due to gyration around the field lines . in place of free - free emission , there is then synchrotron emission from energetic particles .
the negative radio spectral indices in p1 and p3 are conveniently explained by non - thermal particle spectra .
because thermal bremsstrahlung is proportional to @xmath87 and synchrotron emission is proportional to @xmath88 ( @xmath89 ) , the former dominates if the density is high enough , or if the temperature or field strength is low enough . indeed the electron densities found in the ngc 3603 s are @xmath6two orders of magnitude smaller than in the orion proplyds ( see table 4 ) .
the different energy dependencies of the two kinds of emission may lead to one dominating at low frequencies and the other at high frequencies .
a lower limit for a possible magnetic field is given by the razin - tsytovich effect . in order to avoid a razin - suppression of the radio spectrum above @xmath90 cm the magnetic field
can not be smaller than @xmath2 g for an electron density of @xmath44 @xmath5 in the emission region .
on the other hand , to avoid gyroresonance absorption above @xmath90 cm the field strength can not be higher than @xmath91 g. the equipartition magnetic field , the minimum possible value in flare loops , for a @xmath92 k particle distribution with density @xmath44 @xmath5 , is @xmath93 g. note that synchrotron photons of energy 4.8 ghz in a @xmath94 g field must have been produced by @xmath76 gev electrons .
efficient particle acceleration may occur , e.g. , at the proplyd surface , from flow amplification of even very weak ambient magnetic fields followed by local reconnection events ( e.g. @xcite ) . the flux density from a homogeneous sphere with diameter @xmath95 due to synchrotron radiation in a uniform field and a low energy cutoff @xmath96 in units mev of a @xmath97 particle distribution with @xmath98 spectral index power law , which we use as a reasonable assumption here .
] is given by ( e.g. dulk 1985 ) @xmath99 the non - thermal electrons might have been energized by e.g. shock acceleration or magnetic reconnection events , out of the pool of thermal particles , or they may be galactic cosmic rays penetrating the starforming region . in the latter case @xmath100@xmath5 and @xmath101 , and thus @xmath102 g ( for @xmath103 ) . for the former case , because energy is conserved , the production efficiency of the high energy particles is related to the properties of the thermal background particles , and vice versa .
realistic non - linear diffusive particle acceleration scenarios typically yield @xmath104 ( i.e. the fraction of total electrons which end up with high energies ; e.g. @xcite ) , while magnetic reconnection might be as efficient as @xmath105 for the present densities and temperature @xcite .
since the thermal electron density @xmath106 can not be significantly smaller than @xmath107 @xmath44@xmath5 ( derived from [ sii ] line ratios ; see sect.3.4 ) , and @xmath108 if thermal and non - thermal electrons have the same source , we find that the size of the actual magnetized volume emitting non - thermal radiation must be smaller than the appearance of the on the sky for this case . if particle trapping is at work , @xmath109 is possible , and consequently the ratio of emitting to total volume of the becomes even smaller .
the present data , however , do not allow to quantitatively constrain the actual size of the source responsible for the non - thermal radiation .
we note that the recent discovery of variable x - ray emission from some proplyds in orion by @xcite support the idea of a non - thermal emission volume smaller than the appearance of the on the sky .
the size of the emitting region was estimated on the basis of variability arguments to order 110 au .
the three massive ( @xmath6110 ) nebulae in ngc 3603 , which have recently been discovered by @xcite , have been clearly detected and resolved with the _ atca _ at 3 and 6 cm , with one of them likely to be composed of two cometary - shaped objects .
their flux densities are about 1020 times higher than expected from the measurements , and a non - thermal average spectrum can be associated with at least one of the three sources .
this is the first time that non - thermal radio emission has been detected from sources .
our spectral index maps show that the emission region is rather inhomogeneous , with negative spectral indices in the tail and part of the head whereas positive spectral indices , indicating thermal free - free emission , tend to be detected from a small region facing the star cluster .
we show that synchrotron emission from a magnetized , relativistic , non - thermal particle population may explain the non - thermal spectral regions .
energetic electrons necessary for synchrotron radiation may be produced through shocks or by magnetic reconnection out of the pool of thermal particles , or may be galactic cosmic rays .
a stellar origin of the observed flux densities appears unlikely considering that the sources are extended .
thus magnetic fields , which have been neglected in proplyd models so far , are possibly associated with the proplyd envelopes or disks , and appear to be crucial in understanding the physical processes in these objects .
@xcite have shown that an inhomogeneous dust distribution in ngc 3603 causes increasing extinction with distance from the star cluster .
in particular , the extinction at the location of the s is estimated to lie between @xmath111 mag .
extinction estimates derived from the radio- to -luminosity ratio are 12 magnitudes higher
. the ultimate discovery of the so - far undetected disks , which are thought to be part of all proplyds , would give important hints about the origin of the extremely high and non - thermal radio fluxes and the nature of the cometary - shaped clumps seen in ngc 3603 .
this may be accomplished in future mm observations with an upgraded .
we are grateful to wolfgang brandner for providing the hst image of ngc 3603 to overlay on our radio images , for interesting discussions and carefully reading the manuscript .
we thank jessica chapman , jim caswell , simon johnston , tylor bourke and sergey marchenko for reading the manuscript and many constructive comments which improved our paper significantly .
am acknowledges a postdoctoral bursary from the quebec government .
afjm thanks nserc ( canada ) and fcar ( quebec ) for financial support .
irs acknowledges the receipt of a pparc advanced fellowship .
lccccc date ( local time ) & february 9/10 & april 8/9 & june 19 & sept .
13/14 & nov .
8/9 + configurations & 6a & 6d & 6b & 6a & 6c + time on source & 561 min . & 514 min . & 480 min . & 513 min . &
+ pointing position & & & + ra ( j2000 ) & + dec ( j2000 ) & + total bandwidth & + no . of channels & + frequencies
& + beam size & + & + average r.m.s .
near & + proplyds [ mjybeam@xmath3 ] & + flux calibrator & + phase calibrator & + lcccccc radio position & 11:15:13.042 , & 11:15:13.058 , & 11:15:16.557 , & 11:15:16.579 , & 11:15:01.336 , & 11:15:01.330 , + ra , dec ( j2000 ) & 61:15:50.38 & 61:15:50.53 & 61:16:06.57 & 61:16:06.62 & 61:14:45.69 & 61:14:45.53 + peak flux & @xmath113 & @xmath114 & @xmath115 & @xmath116 & @xmath117 & @xmath118 + ( mjybeam@xmath3 ) & & & & & & + integrated & & & & & & + flux ( mjy ) & @xmath119 & @xmath120 & @xmath121 & @xmath122 & @xmath123 & @xmath124 + spectral index & & & + of whole source & & & & & & + @xmath125 ( @xmath126s@xmath3 ) & 0.56 & 0.56 & 0.20 & 0.20 & & + predicted & & & & & & + radio flux ( mjy ) & @xmath127 & @xmath128 & @xmath129 & @xmath130 & & + llccc & & orion nebula & lagoon nebula & + & & m42 & m8 , ngc 6523 & ngc 3603 + distance & [ kpc ] & 0.45 & 1.8 & 6.1 + central star(s ) & & @xmath131 ori c & herschel 36 & cluster + star type(s ) & & o7 & o7.5v & 3 wns+ @xmath670 o stars + @xmath132 & [ ] & @xmath65 & @xmath65 & 2 @xmath133 + lyman flux & [ photonss@xmath3 ] & @xmath134 & 2 @xmath135 & @xmath71 + extinction & [ mag ] & 46 & @xmath136 & 46 + masers & & h@xmath9o , sio & none known & oh , h@xmath9o , ch@xmath137oh + number of proplyds & & @xmath138 & 1 = g5.971.7 & 3 + head size & [ au ] & 45355 & 1080 & 720010800 + distance to central stars & [ pc ] & 0.010.15 & 0.024 & 1.3 , 2.2 , 2.0 + radio flux & [ mjy ] & 0.31.4 & 17 & 1012 + brightness temp . &
[ k ] & @xmath139 & @xmath140 & 3090 + emission measure & [ pc@xmath53 ] & @xmath142 & @xmath143 & @xmath144 + electron density & [ @xmath5 ] & @xmath57 & @xmath146&@xmath147 + mass loss rate & [ yr@xmath3 ] & @xmath148 & @xmath149 & 10@xmath150 + bow shock distance & [ au ] & @xmath151 & 540 & 44400,17400 , + velocity & [ ] & @xmath152 & @xmath76 & 1025 + evap .
time scale&[years ] & @xmath147 & @xmath65 & @xmath65 + proplyd mass & [ ] & 0.010.1 & @xmath14 & @xmath153 + disk radius & [ au ] & 27175 & 160 & 3400 + disk mass & [ ] & @xmath154 & @xmath155 & not known + | three cometary - shaped objects in the giant region ngc 3603 , originally found and identified as proto - planetary disks ( proplyds ) by @xcite using hst+vlt in the optical and near - infrared , have been detected with the australia telescope compact array ( ) in the radio continuum at 3 and 6 cm .
all three objects are clearly resolved with an extent of a few arcseconds .
the integrated 6 cm fluxes are up to 1.3 times higher than the 3 cm fluxes with spectral indices averaged over the whole clump between @xmath0 and 0.5 ( @xmath1 ) , indicating the likely presence of non - thermal emission in at least some of the sources .
we present spectral index maps , and show that the sites of negative radio spectral indices are predominantly concentrated in the direction of the tails in at least two of the three nebulae while positive spectral indices are found in the region facing the ionizing star cluster .
we propose that thermal bremsstrahlung and non - thermal synchrotron radiation are at work in all three sources . in at least one of the three objects optically thin non - thermal synchrotron emission appears to dominate when averaged over their whole spatial extent , while the spectrum of the second source shows a marginal indication of a non - thermal spectrum .
the average spectrum of the third source is in agreement with thermal bremsstrahlung .
all measured fluxes are at least one order of magnitude higher than those predicted by @xcite .
upper limits for mass loss rates due to photo - evaporation are calculated to be @xmath2 year@xmath3 and for electron densities to be @xmath4 @xmath5 . due to the unexpectedly large radio luminosities of the features and because the radio emission is extended a ( proto-)stellar origin of the non - thermal emission from a dust enshrouded star appears unlikely .
instead we propose that magnetized regions within the envelope of the nebulae exist . |
[ [ intro ] ] on the euclidean unit @xmath3-sphere , the set of points which are at a given distance of the north pole is in general a circle .
when the distance is @xmath4 , this circle becomes trivial : it is reduced to the south pole .
such a focusing phenomenon can not appear on a surface of constant , non - positive curvature . in this case
the image by the exponential map of a unit circle of vectors tangent to the surface at a given point is never reduced to one point . in this paper
, we generalize this remark in the context of contact topology .
our motivation comes from the theory of the orderability of the group of contactomorphisms of eliashberg , kim and polterovitch @xcite .
consider a @xmath5-dimensional manifold @xmath6 endowed with a _ cooriented _ contact structure @xmath7 . at each point of @xmath6 ,
the contact hyperplane then separates the tangent space in _ a positive and a negative side_. [ isopos ] a smooth path @xmath8 $ ] in the space of legendrian embeddings ( resp .
immersions ) of a @xmath9-dimensional compact manifold @xmath10 in @xmath11 is called _ a legendrian isotopy ( resp .
if , in addition , for every @xmath12 and every @xmath13 $ ] , the velocity vector @xmath14 lies in the _ positive _ side of @xmath7 at @xmath15 , then this legendrian isotopy ( resp .
homotopy ) will be called _
positive_. this notion of positivity does not depend on the parametrization of the @xmath16 s .
if the cooriented contact structure @xmath7 is induced by a globaly defined contact form @xmath17 , the above condition can be rephrased as @xmath18 .
in particular , a positive contact hamiltonian induces positive isotopies . a positive isotopy ( resp .
homotopy ) will be also called a _ positive path _ in the space of legendrian embeddings ( resp .
immersions ) .
[ example1 ] the space @xmath19 of one - jets of functions on a @xmath9-dimensional manifold @xmath20 has a natural contact one form @xmath21 , where @xmath22 is the liouville one - form of @xmath23 and @xmath24 is the @xmath25-coordinate .
the corresponding contact structure will be denoted by @xmath26 . given a smooth function @xmath27 , its one - jet extension @xmath28 is a legendrian submanifold .
a path between two functions gives rise to an isotopy of legendrian embeddings between their one - jets extensions .
a path @xmath29}$ ] of functions on @xmath20 such that , for any fixed @xmath30 , @xmath31 is an increasing function of @xmath32 , gives rise to a positive legendrian isotopy @xmath33}$ ] in @xmath34 .
conversely , one can check that a positive isotopy consisting only of one - jets extensions of functions is always of the above type . _ in particular there are no positive loops consisting only of one - jet extensions of functions_. [ ex1 ] consider a riemannian manifold @xmath35 .
its unit tangent bundle @xmath36 has a natural contact one - form : if @xmath24 is a unit tangent vector to @xmath20 , and @xmath37 a vector tangent to @xmath38 at @xmath24 , then @xmath39 the corresponding contact structure will be denoted by @xmath40 .
the constant contact hamiltonian @xmath41 induces the geodesic flow .
any fiber of @xmath36 is legendrian . moving
a fiber by the geodesic flow is a typical example of a positive path .
let @xmath20 be a closed manifold .
[ thm1 ] there is no closed positive path in the component of the space of legendrian embeddings in @xmath42 containing the one - jet extensions of functions . the liouville one - form of @xmath23 induces a contact distribution of the fiber - wise spherization @xmath43 .
this contact structure is contactomorphic to the @xmath40 of example [ ex1 ] .
our generalization of the introductory remark [ intro ] is as follows .
[ thm2 ] there is no positive path of legendrian embeddings between two distinct fibers of @xmath44 , provided that the universal cover of @xmath20 is @xmath2 .
[ thm3 ] * 0 ) . *
any compact legendrian submanifold of @xmath45 belongs to a closed path of legendrian embeddings . *
* there exists a component of the space of legendrian embeddings in @xmath46 whose elements are homotopic to @xmath47 and which contains a closed positive path . *
there exists a closed positive path in the component of the space of legendrian immersions in @xmath46 which contains the one - jet extensions of functions .
* iii ) . * given any connected surface @xmath20 , there exists a positive path of legendrian immersions between any two fibers of @xmath48 .
f. laudenbach @xcite proved recently the following generalization of theorem [ thm3 ] * ii ) * : for any closed @xmath20 , there exists a closed positive path in the component of the space of legendrian immersions in @xmath42 which contains the one - jet extensions of functions .
theorem [ thm3 ] * 0 ) * implies that for any contact manifold @xmath11 , there exists a closed positive path of legendrian embeddings ( just consider a darboux ball and embed the example of theorem [ thm3 ] * 0 * ) .
a legendrian manifold @xmath49 will be called _ positive _ if it is connected by a positive path to the one - jet extension of the zero function .
the one - jet extension of a positive function is a positive legendrian manifold .
but , in general , the value of the @xmath24 coordinate can be negative at some points of a positive legendrian manifold .
consider a closed manifold @xmath20 and fix a function @xmath50 .
assume that @xmath51 is a regular value of @xmath52 .
denote by @xmath53 the union for @xmath54 of the @xmath55 .
smooth embedding _ of @xmath56 in @xmath34 , foliated by the @xmath55 .
we denote by @xmath57 the subset @xmath58 .
consider the manifold @xmath59 .
its boundary @xmath60 is the set @xmath61 .
fix some field @xmath62 and denote by @xmath63 the total dimension of the homology of @xmath1 with coefficients in that field ( @xmath64 ) .
we say that a point @xmath65 is _ above _ some subset of the manifold @xmath20 if its image under the natural projection @xmath66 belongs to this subset .
[ thm4 ] for any positive legendrian manifold @xmath67 in general position with respect to @xmath53 , there exists at least @xmath63 points of intersection of @xmath10 with @xmath57 lying above @xmath68 .
more precisely , for a generic positive legendrian manifold @xmath10 , there exists at least @xmath63 different _ positive _ numbers @xmath69 such that @xmath10 intersects each manifold @xmath70 above @xmath68 .
theorem [ thm4 ] implies the morse estimate for the number of critical points of a morse function @xmath71 on @xmath20 .
this can be seen as follows . by adding a sufficiently large constant to @xmath71
, one can assume that @xmath72 is a positive legendrian manifold .
if @xmath52 is a constant positive function , then @xmath73 , and intersections of @xmath74 with @xmath53 are in one to one correspondence with the critical points of @xmath71 .
furthermore , @xmath71 is morse if and only if @xmath10 is transversal to @xmath53 . in fact , one can prove that theorem [ thm4 ] implies ( a weak form of ) arnold s conjecture for lagrangian intersection in cotangent bundles , proved by chekanov @xcite in its legendrian version .
this is no accident : our proof rely on the main ingredient of chekanov s proof : the technique of generating families ( see theorem [ chgen ] ) .
one can also prove that theorem [ thm4 ] implies theorem [ thm1 ] .
theorem [ thm5 ] below , which in turn implies theorem [ thm2 ] , is also a direct consequence of theorem [ thm4 ] .
[ thm5 ] consider a line in @xmath2 .
denote by @xmath53 the union of all the fibers of @xmath75 above this line .
consider one of these fibers and a positive path starting from this fiber .
the end of this positive path is a legendrian sphere .
this sphere must intersect @xmath53 in at least @xmath3 points . in theorem [ thm4 ] and in theorem
[ thm5 ] , we observe the following feature : the submanifold @xmath53 is foliated by legendrian submanifolds .
we pick one of them , and we conclude that we can not disconnect it from @xmath53 by a positive contact isotopy . in dimension 3 , our @xmath53 is a surface foliated by legendrian curves ( in a non generic way ) .
recall that generically , a closed oriented surface @xmath76 contained in a contact @xmath77-manifold @xmath78 is _ convex _ : there exists a vector field tranversal to @xmath76 and whose flow preserves @xmath7 .
equivalently , a convex surface admits a _ homogeneous neighborhood _ @xmath79 , @xmath80 , where the restriction of @xmath7 is @xmath25-invariant .
given such an homogeneous neighborhood , we obtain a smooth , canonically oriented , multicurve @xmath81 , called the _ dividing curve _ of @xmath76 , made of the points of @xmath76 where @xmath7 is tangent to the @xmath25-direction .
it is automatically transversal to @xmath7 . according to giroux @xcite ,
the dividing curve @xmath82 does not depend on the choice of @xmath83 up to an isotopy amongst the multicurves transversal to @xmath7 in @xmath76 .
characteristic foliation _
@xmath84 of @xmath85 is the integral foliation of the singular line field @xmath86 .
let @xmath76 be a closed oriented surface of genus @xmath87 and @xmath88 , @xmath79 , be an homogeneous neighborhood of @xmath89 .
the surface @xmath76 is @xmath7-convex , and we denote by @xmath82 its dividing multicurve .
we assume that @xmath7 is tight on @xmath83 , which , after giroux , is the same than to say that no component of @xmath82 is contractible in @xmath76 .
[ thm6 ] assume @xmath10 is a legendrian curve in @xmath76 having minimal geometric intersection @xmath90 with @xmath82 . if @xmath91}$ ] is a positive legendrian isotopy of @xmath92 then @xmath93 .
the positivity assumption is essential : if we push @xmath10 in the homogeneous direction , we get an isotopy of legendrian curves which becomes instantaneously disjoint from @xmath76 . if @xmath94 , this is a positive isotopy of @xmath10 that disjoints @xmath10 from @xmath76 . for a small positive isotopy ,
the result is obvious .
indeed , @xmath10 is an integral curve of the characteristic foliation @xmath84 of @xmath76 , which contains at least one singularity in each component of @xmath95 .
for two consecutive components , the singularities have opposite signs .
moreover , when one moves @xmath10 by a small positive isotopy , the positive singularities are pushed in @xmath96 and the negative ones in @xmath97 . between two singularities of opposite signs , we will get one intersection with @xmath76 .
the relationship with the preceeding results is given by the following corollary of theorem [ thm4 ] , applied with @xmath98 and @xmath99 , for some fixed @xmath100 .
in this situation , the surface @xmath53 of theorem [ thm4 ] will be called @xmath101 .
it is an infinite cylinder foliated by legendrian circles .
its characteristic foliation @xmath102 has @xmath103 infinite lines of singularities .
the standard contact space @xmath46 is itself an homogeneous neighborhood of @xmath101 , and the corresponding dividing curve consists in @xmath103 infinite lines , alternating with the lines of singularities .
let @xmath104 .
theorem [ thm4 ] gives : [ ex : fund ] let @xmath105 be a generic positive deformation of @xmath106 . then @xmath107 .
indeed , there are @xmath108 intersections with @xmath109 , and @xmath108 other intersections which are obtained in a similar way with the function @xmath110
. this corollary will be the building block to prove theorem [ thm6 ] .
this paper is organized as follows .
the proof of theorem @xmath77 , which in a sense shows that the hypothesis of theorems [ thm1 ] and [ thm2 ] are optimal , consists essentially in a collection of explicit constructions .
it is done in the next section ( [ proof3 ] ) and it might serve as an introduction to the main notions and objects discussed in this paper .
the rest of the paper is essentially devoted to the proof of theorems 1 , 2 , 4 , 5 and 6 , but contains a few statements which are more general than the theorems mentioned in this introduction .
this work was motivated by a question of yasha eliashberg .
this paper is an based on an unpublished preprint of 2006 @xcite .
since then , chernov and nemirovski proved a statement which generalises our theorems 1 and 2 @xcite , and they found new applications of this to causality problems in space - time .
all this is also related to the work of bhupal @xcite and to the work of sheila sandon @xcite , who reproved some results of @xcite using the generating families techniques .
vincent colin is partially supported by the anr symplexe , the anr floer power and the institut universitaire de france .
petya pushkar is partially supported by rfbr grant 08 - 01 - 00388 .
in order to prove statement * i ) * of theorem [ thm3 ] , we begin by the description of a positive loop in the space of legendrian embeddings in @xmath111 . due to theorem [ thm1 ] , this can not happen in the component of the zero section @xmath47 .
take @xmath112 and consider a legendrian submanifold @xmath10 homotopic to @xmath47 and embedded in the half - space @xmath113 .
consider the contact flow @xmath114 .
the corresponding contact hamiltonian @xmath115 is positive near @xmath116 , for all @xmath117 , and hence , @xmath116 is a positive path . on the other hand , one can go from @xmath118 back to @xmath10 just by increasing the @xmath24 coordinate , which is also a positive path .
this proves statement * i ) * of theorem [ thm3 ] .
[ fig2 ] , wxhich is homotopic to @xmath47 through legendrian immersions . ]
[ [ thm3a ] ] we now consider statement * ii)*. take @xmath119 as above , but assume in addition that @xmath10 is homotopic to @xmath47 through legendrian immersions .
such a @xmath10 exists ( one can show that the @xmath10 whose front projection is depicted in fig .
[ fig2 ] is such an example ) , but can not be legendrian isotopic to @xmath47 , since , by @xcite , it would intersect @xmath120 ) .
[ [ step-1 . ] ] * step 1 .
* + + + + + + + + + the homotopy between @xmath47 and @xmath10 can be transformed into a positive path of legendrian immersions between @xmath47 and a vertical translate @xmath121 of @xmath10 , by combining it with an upwards translation with respect to the @xmath24 coordinate .
[ [ step-2 . ] ] * step 2 .
* + + + + + + + + + then , using the flow @xmath122 ( defined in [ posloop ] ) for @xmath123 $ ] with @xmath108 big enough , one can reach reach another translate @xmath124 of @xmath10 , on which the @xmath24 coordinate can be arbitrarily low .
[ [ step-3 . ] ] * step 3 .
* + + + + + + + + + consider now a path of legendrian immersions from @xmath10 to @xmath47 .
it can be modified into a positive path between @xmath124 and @xmath47 , like in step 1 .
this proves statement * ii ) * of theorem [ thm3 ] . [ [ section ] ] the proof of statement * 0 ) * uses again the same idea .
given any compact legendrian submanifold @xmath125 , there exists @xmath121 , which is legendrian isotopic to @xmath10 and which is contained into the half - space @xmath126 , for some system @xmath127 of canonical coordinates on @xmath128 .
it is possible to find a positive path between @xmath10 and a sufficiently high vertical translate @xmath124 of @xmath121 . because @xmath129 , one can now slide down @xmath124 by a positive path as low as we want with respect to the @xmath24 coordinate , as above .
hence we can assume that @xmath10 is connected by a positive path of embeddings to some @xmath130 , which is a vertical translate of @xmath121 , on which the @xmath24 coordinate is very negative .
so we can close this path back to @xmath10 in a positive way .
[ [ section-1 ] ] we now prove statement * iii)*. consider two points @xmath131 and @xmath132 on the surface @xmath20 , an embedded path from @xmath131 to @xmath132 , and an open neighborhood @xmath83 of this path , diffeomorphic to @xmath133 .
hence it is enough to consider the particular case @xmath134 .
we consider this case below .
we now recall the classical `` hodograph '' contactomorphism @xcite which identifies @xmath135 and @xmath46 , and more generally @xmath136 and ( @xmath137 . the same trick will be used later to prove theorems 2 and 5 ( sections [ proof2 ] and [ proof4 ] ) .
fix a scalar product @xmath138 on @xmath2 and identify the sphere @xmath139 with the standard unit sphere in @xmath2 .
identify a covector at a point @xmath140 with a vector in the hyperplane tangent to the sphere at @xmath141 ( perpendicular to @xmath141 ) . then to a point @xmath142
we associate the cooriented contact element at the point @xmath143 , which is parallel to @xmath144 , and cooriented by @xmath141 .
one can check that the fiber of @xmath145 over some point @xmath146 is the image by this contactomorphism of @xmath147 , where @xmath148 , @xmath149 .
one can assume that @xmath150 .
the case when @xmath151 follows directely from theorem [ thm3 ] * ii ) * via the contactomorphism described above .
the fiber @xmath152 corresponds to @xmath47 .
suppose now that @xmath153 .
we need to find a positive path of legendrian immersions in @xmath46 between @xmath47 and @xmath154 . to achieve this , it is enough to construct a positive path of legendrian immersions between @xmath47 and a translate of @xmath47 that would be entirely below @xmath154 , with respect to the @xmath24 coordinate .
this can be done as in [ thm3a ] , just by decreasing even more the @xmath24 coordinate like in step 2 .
this finishes the proof of theorem [ thm3 ]
. we will in fact prove theorem [ thm7 ] below , which is more general than theorem [ thm1 ] .
fix a closed ( compact , without boundary ) manifold @xmath20 and a smooth fibration @xmath188 such that @xmath182 is compact .
a legendrian manifold @xmath175 will be called a _ @xmath182-quasifunction _ if it is legendrian isotopic to a manifold given by some generating family @xmath182-quadratic at infinity .
we say that a connected component @xmath208 of the space of legendrian submanifolds in @xmath42 is _
if @xmath208 contains an @xmath182-quasifunction .
for example , the component @xmath208 containing the one jets extensions of the smooth functions on @xmath1 is @xmath182-quasifunctional , with @xmath182 coinciding with @xmath20 ( the fiber is just a point ) .
[ thm7 ] [ noqloops ] an @xmath182-quasifunctional component contains no closed positive path .
[ [ section-2 ] ] the proof of theorem 7 will be given in [ pfthm7 ] .
it will use the following generalization of chekanov s theorem ( see @xcite ) , and proposition [ prop ] below .
[ chgen ] consider a legendrian isotopy @xmath209}$ ] such that @xmath106 is an @xmath182-quasifunction .
then there exist a number @xmath210 and a smooth family of functions @xmath182-quadratic at infinity @xmath211 , such that for any @xmath212 $ ] , @xmath213 is a generating family of @xmath16 .
note that it follows from theorem [ chgen ] that _ any _ legendrian manifold in some @xmath182-quasifunctional component is in fact an @xmath182-quasifunction .
consider a positive path @xmath214}$ ] given by a family @xmath215}$ ] of e - quadratic at infinity generating families .
we are going to prove the following inequality : [ prop ] the viterbo numbers of the family @xmath213 are monotone increasing functions with respect to @xmath32 : @xmath216 for any @xmath217 .
assume that the inequality is proved for a generic family .
this , together with continuity of viterbo numbers , gives us a weak inequality @xmath218 , for any family .
but positivity is a @xmath219open condition , so we can perturb the initial family @xmath213 into some family @xmath220 coinciding with @xmath213 when @xmath32 is sufficiently close to @xmath221 , such that @xmath220 still generates a positive path of legendrian manifolds and such that the family @xmath222}$ ] is generic .
we have @xmath223 and hence inequality is strong for all families .
we now prove the inequality for generic families .
excellent morse functions form an open dense set in the space of all e - quadratic at infinity functions on @xmath224 .
the complement of the set of excellent morse functions forms a discriminant , which is a singular hypersurface . a generic one - parameter family of e - quadratic at infinity functions @xmath213 on @xmath225 has only a finite number of transverse intersections with the discriminant in its smooth points , and for every @xmath32 except possibly finitely many , the hessian @xmath226 is non - degenerate at every critical point of the function @xmath213 . we will use the notion of _ cerf diagram _ of a family of functions @xmath227}$ ] on a smooth manifold .
the cerf diagram is a subset in @xmath228\times \rm$ ] consisting of all the pairs of type @xmath229 , where @xmath230 is a critical value of @xmath231 . in the case of a generic family of functions on a closed
manifold , the cerf diagram is a curve with non - vertical tangents everywhere , with a finite number of transversal self - intersections and cuspidal points as singularities .
the graph of the viterbo number @xmath232 is a subset of the cerf diagram of the family @xmath213 . to prove the monotonicity of the viterbo numbers , it is sufficient to show that the cerf diagram of @xmath213 has a positive slope at every point except finite set .
the rest of the proof of proposition [ prop ] is devoted to that .
we say that a point @xmath131 on a legendrian manifold @xmath233 is non - vertical if the differential of the natural projection @xmath234 is non - degenerate at @xmath131 .
let @xmath16 be a smooth family of legendrian manifolds and @xmath235 a non - vertical point .
by the implicit function theorem , there exists a unique family @xmath236 , defined for @xmath32 sufficiently close to @xmath237 , such that @xmath238 and @xmath239 .
we call the number @xmath240 _ vertical speed _ of the point @xmath241 .
[ l1 ] for a positive path of legendrian manifolds , the vertical speed of every non - vertical point is positive .
consider a path @xmath16 in the space of legendrian manifolds given by a generating family @xmath213 .
consider the point @xmath242 and the point @xmath243 such that @xmath244 then @xmath131 is non - vertical if and only if the hessian @xmath245 is non - degenerate .
for such a point @xmath131 , the following lemma holds : [ l2 ] the vertical speed at @xmath131 is equal to @xmath246 .
let @xmath247 be a family of smooth functions and assume that the point @xmath248 is a morse critical point for @xmath249 . by the implicit function theorem , for each @xmath32 sufficiently close to @xmath237 ,
the function @xmath247 has a unique critical point @xmath250 close to @xmath251 , and @xmath250 is a smooth path .
[ l3 ] the speed of the critical value @xmath252 is equal to @xmath253 .
indeed , @xmath254 . at almost every point on the cerf diagram ,
the slope of the cerf diagram at this point is the speed of a critical value of the function @xmath213 . by lemma [ l3 ] and lemma [ l2 ] , it is the vertical speed at some non - vertical point . by lemma [ l1 ]
it is positive .
this finishes the proof of proposition [ prop ] .
suppose now that there is a closed positive loop @xmath209}$ ] in some @xmath182-quasifunctional component @xmath208 .
the condition of positivity is open .
we slightly perturb the loop @xmath209 } $ ] such that @xmath255 is a finite set of cardinality @xmath256 .
note that @xmath257 , since @xmath106 is a @xmath182-quasifunction .
consider the @xmath256-th multiple of the loop @xmath209}$ ] . by theorem [ chgen ]
, @xmath16 has a generating family @xmath213 , for all @xmath258 $ ] . by proposition [ prop ] ,
we have that @xmath259 all these @xmath260 numbers belong to the set @xmath255 .
this is impossible due to the cardinality of this set .
this finishes the proof of theorem [ thm7 ] and hence of theorem @xmath261 .
theorem [ thm2 ] is a corollary of theorem [ thm1 ] , via the the contactomorphism between @xmath262 and @xmath263 we have seen in [ hodograph ] .
consider the fiber @xmath152 of the fibration @xmath264 .
it corresponds to a legendrian manifold @xmath265 , where @xmath266 is the function @xmath267 .
it is a morse function for @xmath268 , and has only two critical points and two critical values @xmath269 .
the critical points of @xmath266 are non - degenerate if @xmath268 .
it follows from proposition [ morseprop ] that @xmath270 , @xmath271 . indeed , any small generic morse perturbation of @xmath71 has two critical points with critical values close to @xmath269 .
viterbo numbers for this perturbation should be different . hence by continuity @xmath270 , @xmath271 .
viterbo numbers for @xmath272 are equal to zero , because @xmath273 .
the existence of a positive path would contradict the monotonicity ( proposition [ prop ] ) of viterbo numbers .
we briefly recall the construction of a generating family for a legendrian manifold ( the details can be found in @xcite ) .
let @xmath155 be a smooth fibration over a smooth manifold @xmath20 , with fiber @xmath156 .
let @xmath157 be a smooth function .
for a point @xmath141 in @xmath20 we consider the set @xmath158 whose points are the critical points of the restriction of @xmath71 to the fiber @xmath159 .
denote @xmath160 the set @xmath161 .
assume that the rank of the matrix @xmath162 ( @xmath163 , @xmath141 are local coordinates on the fiber and base respectively ) formed by second derivatives is maximal ( that is , equal to the dimension of @xmath20 ) at each point of @xmath160 .
this condition holds for a generic @xmath71 and does not depend on the choice of the local coordinates @xmath164 .
the set @xmath165 is then a smooth submanifold of the same dimension as @xmath20 , and the restriction of the map @xmath166 where @xmath167 denotes the differential along @xmath20 , to @xmath160 defines a legendrian immersion of @xmath168 into @xmath169 .
if this is an embedding ( this is generically the case ) , then @xmath71 is called a _ generating family _ of the legendrian submanifold @xmath170 .
a point @xmath65 is by definition a triple consisting in a point @xmath171 in the manifold @xmath20 , a covector @xmath172 and a real number @xmath173 .
a point @xmath174 will be called a _ critical point _ of the legendrian submanifold @xmath175 if @xmath176 .
the value of the @xmath24 coordinate at a critical point of a legendrian manifold @xmath10 will be called a
_ critical value _ of @xmath10 .
the set of all critical values will be denoted by @xmath177 .
observe that , for a manifold @xmath178 given by a generating family @xmath71 the set @xmath179 coincides with the set of critical values of the generating family @xmath71 .
we call a critical point @xmath174 _ nondegenerate _ if @xmath10 intersects the manifold given by the equation @xmath180 transversally at @xmath131 .
if an embedded legendrian submanifold @xmath181 is given by a generating family @xmath71 , then the non - degenerate critical points of @xmath71 are in one to one correspondance with the non - degenerate critical points of @xmath181 .
we describe now the class of generating families we will be working with . pick a closed manifold @xmath182 which is a fibration over some closed manifold @xmath20 .
a function @xmath183 is called _
@xmath182-quadratic at infinity _ if it is a sum of a non - degenerate quadratic form @xmath184 on @xmath185 and a function on @xmath186 with bounded differential ( i.e. the norm of the differential is uniformly bounded for some riemannian metric which is a product of a riemannian metric on @xmath182 and the euclidean metric on @xmath185 ) .
this definition does not depend on the choice of the metrics .
if a function which is @xmath182-quadratic at infinity is a generating family ( with respect to the fibration @xmath187 ) , then we call it a _ generating family @xmath182-quadratic at infinity_. we gather here some results from morse theory which will be needed later .
let @xmath188 be a fibration , @xmath182 is a closed manifold . consider a function @xmath71 , @xmath182-quadratic at infinity
. denote by @xmath189 the set @xmath190 .
for sufficiently big positive numbers @xmath191 , the set @xmath192 is a deformation retract of @xmath193 .
hence the homology groups @xmath194 depend only on @xmath195 .
we will denote them by @xmath196 .
it is known ( see @xcite ) that for sufficiently big @xmath195 , @xmath196 is isomorphic to @xmath197 .
for any function @xmath71 which is @xmath182-quadratic at infinity , and any integer + @xmath198 , we define a _ viterbo number _ @xmath199 by @xmath200 where @xmath201 is the map induced by the natural inclusion @xmath202 , when @xmath195 is a sufficiently big number .
our definition is similar to viterbo s construction @xcite in the symplectic setting .
the following proposition is an adaptation of @xcite : [ morseprop ] * i. * each number @xmath199 , @xmath203 is a critical value of @xmath71 , and if @xmath71 is an excellent morse function ( i.e all its critical points are non - degenerate and all critical values are different ) then the numbers @xmath199 are different . *
* consider a family @xmath204}$ ] of functions which are all @xmath182-quadratic at infinity .
for any @xmath203 the number @xmath205 depends on @xmath32 continuously .
if the family @xmath204}$ ] is generic ( i.e. intersects the discriminant consisting of non excellent morse functions transversally at its smooth points ) then @xmath205 is a continuous piecewise smooth function with a finite number of singular points . at this moment
, it is unknown wether @xmath206 depends on @xmath71 for a given legendrian manifold @xmath178 .
conjecturally there should be a definition of some analogue of @xmath207 in terms of augmentations on relative contact homology .
in this section we prove theorem [ thm4 ] and deduce theorem [ thm5 ] from it .
we need first to generalize some of the previous constructions and results to the case of manifolds with boundary .
let @xmath20 be a compact closed manifold .
fix a function @xmath274 such that @xmath51 is a regular value of @xmath52 .
denote by @xmath1 the set @xmath275 .
denote by @xmath276 the dimension @xmath277 ( all the homologies here and below are counted with coefficients in a fixed field @xmath62 ) .
the definition of the viterbo numbers for a function quadratic at infinity on a manifold with boundary is the same as in the case of closed manifold .
we repeat it briefly . given a function @xmath71 which is quadratic at infinity
, we define the viterbo numbers @xmath278 as follows .
a _ generalized critical value _ of @xmath71 is a real number which is a critical for @xmath71 or for the restriction @xmath279 .
denote by @xmath189 the set @xmath280 .
the homotopy type of the set @xmath189 is changed only if @xmath195 passes through a generalized critical value .
one can show that , for sufficiently big @xmath281 , the homology of the pair @xmath282 is independent of @xmath283 , and naturally isomorphic , by the thom isomorphism , to @xmath284 .
so , for any @xmath285 and sufficiently large @xmath286 the projection @xmath287 is well defined and independent of @xmath286 . denote the image of this projection by @xmath288 .
the viterbo numbers are @xmath289 any viterbo number @xmath290 is a generalized critical value of the function @xmath71 .
obviously , @xmath291 .
for any continuous family @xmath213 of quadratic at infinity functions , @xmath292 depends continuously on @xmath32 .
consider a @xmath293-parameter family of quadratic at infinity functions @xmath294}\colon
n\times\rm^{k}\to \rm$ ] , such that @xmath213 is a generating family for the legendrian manifold @xmath16 and such that the path @xmath295}$ ] is positive .
we will consider the restriction of the function @xmath213 to @xmath296 and denote it by @xmath213 also .
the following proposition generalizes proposition [ prop ] .
[ prop1 ] the viterbo numbers of the family @xmath213 are monotone increasing : @xmath297 for any @xmath298 .
the difference with proposition [ prop ] is that the cerf diagram of a generic family has one more possible singularity .
this singularity corresponds to the case when a morse critical point meets the boundary of the manifold . in this case , the cerf diagram is locally diffeomorphic to a parabola with a tangent half - line . we now prove theorem [ thm4 ] .
consider the 1-parameter family of functions @xmath299 , @xmath300 on @xmath301 .
the manifold @xmath10
intersects @xmath302 at some point above @xmath1 if and only if the function @xmath303 has @xmath51 as an ordinary critical value ( not a critical value of the restriction to the boundary ) .
consider the numbers @xmath304 . by proposition [ prop1 ] , @xmath305 .
for a sufficiently big value of @xmath22 , each of them is negative . to show that , consider a sufficiently small @xmath306 belonging to the component of the regular values of @xmath52 which contains @xmath51 .
denote by @xmath307 the set @xmath308 .
the manifold @xmath309 is diffeomorphic to the manifold @xmath1 , and the inclusion map is an homotopy equivalence .
denote by @xmath310 the restriction of @xmath311 to the @xmath312 .
consider the following commutative diagram : @xmath313 where @xmath283 are sufficiently big numbers , @xmath314 denote thom isomorphisms and @xmath315 are the maps induced by the natural inclusions .
it follows from the commutativity of the diagram and from the fact that @xmath316 is an isomorphism that @xmath317 for every @xmath108 . for sufficiently big @xmath22 and for every @xmath318 ,
the critical values of the function @xmath310 restricted to @xmath319 are negative .
hence all generalized critical values of @xmath310 are negative .
it follows that all the numbers @xmath320 are negative , and the same holds for @xmath304 .
we fix @xmath321 such that @xmath322 for every @xmath323 .
consider now @xmath304 as a function of @xmath324 $ ] .
we are going to show that its zeroes correspond to the intersections above @xmath68 . for a manifold @xmath105 in general position ,
all the generalized critical values of @xmath325 are non - zero .
in particular all the critical values of the function the @xmath326 are non - zero .
the function @xmath326 coincides with @xmath327 since @xmath328 on @xmath60 .
hence , if zero is a critical value for @xmath311 , then it is an ordinary critical value at some inner point .
this finishes the proof of theorem [ thm4 ] .
the function @xmath329 can be constant on some sub - intervals in @xmath3300,\lambda_0[$ ] , even for a generic function @xmath325 .
indeed , the critical values of the restriction of @xmath311 to @xmath331 do not depend on @xmath22 .
it is possible that @xmath332 is equal to such a critical value for some @xmath22 s .
the following proposition concerns the case of a general ( non necessarily generic ) positive legendrian manifold . we suppose again that @xmath52 is a function having @xmath51 as regular value and that @xmath10 is a positive manifold .
[ common ] for any connected component of the set @xmath333 there exists a positive @xmath22 such that @xmath10 intersects with @xmath334 above this component .
consider a connected component @xmath335 of the manifold @xmath1 .
it is possible to replace @xmath52 by some function @xmath336 such that @xmath51 is a regular value for @xmath336 , @xmath336 coincides with @xmath52 on @xmath335 and @xmath336 is negative on @xmath337 .
we consider @xmath338 as a function of @xmath22 .
it is a continuous function , positive in some neighborhood of zero , and negative for the big values of @xmath22 .
fix some @xmath17 and @xmath339 such that @xmath340 and @xmath341 .
assume that for any @xmath342 $ ] , @xmath10 does not intersect @xmath343 above @xmath335 . then this is also true for any small enough generic perturbation @xmath121of @xmath10 .
denote by @xmath344 a generating family for @xmath121 .
each zero @xmath321 of @xmath345 corresponds to an intersection of @xmath121 with @xmath346 above @xmath335 .
such a @xmath321 exists by theorem [ thm4 ] .
this is a contradiction .
we can suppose that the origin of @xmath2 belongs to the line considered in the statement of theorem [ thm5 ] .
consider now again the contactomorphism of [ hodograph ] @xmath347 .
for such a choice of the origin , the union of all the fibers above the points on the line forms a manifold of type @xmath348 , where @xmath52 is the restriction of linear function to the sphere @xmath139 .
the manifold @xmath333 has one connected component ( it is an hemisphere ) . by proposition
[ common ] there is at least one intersection of the considered positive legendrian sphere with @xmath349 .
another point of intersection comes from @xmath350 .
these two points are different because @xmath350 does not intersect with @xmath349
the strategy for proving theorem [ thm6 ] is to link the general case to the case of @xmath351 .
let @xmath352 .
we first consider the infinite cyclic cover @xmath353 of @xmath76 associated with @xmath354\in \pi_1 ( s)$ ] .
the surface @xmath353 is an infinite cylinder .
we call @xmath355 the corresponding cover of @xmath83 endowed with the pullback @xmath356 of @xmath7 . by construction
, @xmath355 is @xmath356-homogeneous .
we also call @xmath357 a continuous compact lift of @xmath358 in @xmath355 .
[ lemmaa ] if we denote by @xmath365 the projection forgetting the @xmath25-factor , we can find a lift @xmath366 of a @xmath367-small deformation of @xmath364 in @xmath353 which contains @xmath368 , whose geometric intersection with @xmath369 is @xmath370 and whose boundary is legendrian . to prove this
, we only have to find a legendrian lift @xmath371 of a small deformation of @xmath372 , and to make a suitable slide of @xmath364 near its boundary along the @xmath25-factor to connect @xmath371 to a small retraction of @xmath373 .
the plane field @xmath356 defines a connection for the fibration @xmath374 outside any small neigbourhood @xmath375 of @xmath376 .
we thus can pick any @xmath377-horizontal lift of @xmath378 . we still have to connect the endpoints of these legendrian arcs in @xmath379 .
these endpoints lie at different @xmath25-coordinates , however this is possible to adjust since @xmath356 is almost vertical in @xmath379 ( and vertical along @xmath380 ) . to make it more precise ,
we first slightly modify @xmath364 so that @xmath372 is tangent to @xmath381 near @xmath376 .
let @xmath382 be the metric closure of a component of @xmath383 contained in the metric closure @xmath384 of a component of @xmath385 .
on @xmath386 , the contact structure @xmath356 is given by an equation of the form @xmath387 where @xmath230 denotes the @xmath25-coordinate and @xmath339 is a @xmath293-form on @xmath388 , such that @xmath389 is an area form that goes to @xmath390 as we approach @xmath391 .
now , let @xmath392 be another arc properly embedded in @xmath384 and which coincides with @xmath382 near its endpoints .
if we take two lifts of @xmath382 and @xmath392 by @xmath4 starting at the same point ( these two lifts are compact curves , since they coincide with the characteristic foliation near their endpoints , and thus lift to horizontal curves near @xmath376 where @xmath339 goes to infinity ) , the difference of altitude between the lifts of the two terminal points is given by the area enclosed between @xmath382 and @xmath392 , measured with @xmath389 .
as @xmath389 is going to infinity near @xmath393 , taking @xmath392 to be a small deformation of @xmath382 sufficiently close to @xmath394 , we can give this difference any value we want .
this proves lemma [ lemmaa ] the surface @xmath366 is @xmath356-convex and its dividing set has exactly @xmath103 components going from one boundary curve to the other .
all the other components of @xmath399 are boundary parallel . moreover , the curve @xmath368 intersects by assumption exactly once every non boundary parallel component and avoids the others
. then one can easily embed @xmath366 in a larger annulus @xmath400 and extend the system of arcs @xmath401 outside of @xmath366 by gluing small arcs , in order to obtain a system @xmath402 of @xmath103 non boundary parallel arcs on @xmath400 simultaneously , we extend the contact structure @xmath356 from @xmath403 , considered as an homogeneous neighborhood of @xmath366 , to a neigbourhood @xmath404 of @xmath400 . to achieve
this one only has to extend the characteristic foliation , in a way compatible with @xmath402 , and such that the boundary of @xmath400 is also legendrian .
note that the @xmath25-factor is not changed above @xmath366 . to summarize
, @xmath405 is an homogeneous neigborhood of @xmath400 for the extension @xmath406 , and @xmath400 has legendrian boundary with dividing curve @xmath407 . by genericity
, we can assume that the characteristic foliation of @xmath400 is morse - smale .
then , using giroux s realization lemma @xcite , one can perform a @xmath367-small modification of @xmath400 relative to @xmath408 , leading to a surface @xmath409 , through annuli transversal to the @xmath25-direction , and whose support is contained in an arbitrary small neighborhood of saddle separatrices of @xmath410 , so that the characteristic foliation of @xmath409 for @xmath406 is conjugated to @xmath411 .
if this support is small enough and if we are in the generic case ( which can always been achieved ) where @xmath412 does nt meet the separatrices of singularities of @xmath410 , we get that @xmath413 .
as we are dealing with homogeneous neighborhoods , we see that @xmath414 is conjugated with @xmath397 .
this proves lamme [ lemma : embedd ] .
let @xmath88 be a @xmath7-homogeneous neighborhood of a sphere @xmath76 .
if @xmath7 is tight ( i.e. @xmath82 is connected ) , then any legendrian curve @xmath416 can be made disjoint from @xmath76 by a positive isotopy .
consider @xmath417 with coordinates @xmath418 endowed with the contact structure @xmath419 .
the radial vector field @xmath420 is contact . due to giroux s realization lemma ,
the germ of @xmath7 near @xmath76 is isomorphic to the germ given by @xmath26 near a sphere @xmath421 transversal to @xmath384 .
let @xmath106 be the image of @xmath10 in @xmath421 by this map .
by genericity , we can assume that @xmath106 avoids the vertical axis @xmath422 .
now , if we push @xmath106 enough by the flow of @xmath423 , we have a positive isotopy of @xmath106 whose endpoint @xmath105 avoids @xmath421 .
this isotopy takes place in a @xmath26-homogeneous collar containing @xmath421 and obtained by flowing back and forth @xmath421 by the flow of @xmath384 .
this collar embeds in @xmath83 by an embedding sending @xmath421 to @xmath76 and the @xmath384-direction to the @xmath25-direction .
arnold , v. i. ; gusein - zade , s. m. ; varchenko , a. n. singularities of differentiable maps .
i. the classification of critical points , caustics and wave fronts .
monographs in mathematics 82 . , birkhauser , 1985 . | we show that there is no positive loop inside the component of a fiber in the space of legendrian embeddings in the contact manifold @xmath0 , provided that the universal cover of @xmath1 is @xmath2 .
we consider some related results in the space of one - jets of functions on a compact manifold .
we give an application to the positive isotopies in homogeneous neighborhoods of surfaces in a tight contact 3-manifold . |
in recent times there has been a mounting interest in the spectral theory of non - selfadjoint operators .
as opposed to self - adjoint operators , the norm of the resolvent can be very large even far away from the spectrum . as a consequence ,
the spectrum of such an operator can change a lot , even under small perturbations . a major tool to quantify these zones of spectral instability
is given by the @xmath9-pseudospectrum .
we follow the work of l.n .
trefethen and m. embree @xcite and define the @xmath9-pseudospectrum of a closed linear operator @xmath10 on a banach space @xmath11 by @xmath12 where @xmath13 denotes the spectrum of @xmath10 .
equivalently , @xmath14 emphasized by the works of l.n .
trefethen , m. embree , e.b .
davies , m. zworski , j. sjstrand , cf .
@xcite , and many others , spectral instability of non - selfadjoint operators has become a prominent and important subject of research . in view of
is natural to study the spectrum of such operators under small random perturbations .
one line of recent research interest has focused on the case of elliptic ( pseudo-)differential operators subject to small random perturbations , see for example @xcite and @xcite , as we did ourselves in the present paper .
the central object of this work is the following semiclassical operator : let @xmath15 and consider @xmath16 , given by @xmath17 where we assume that @xmath18 @xmath19 is such that @xmath20 has exactly two critical points , one minimum and one maximum , say at @xmath21 , with @xmath22 for all @xmath23 .
+ the operator @xmath1 is a model for a non - selfadjoint semiclassical differential operator and it has first been introduced by m. hager in @xcite and has already been subject to further study in @xcite . as the natural domain of @xmath1
we take the semiclassical sobolev space @xmath24 where @xmath25 denotes the @xmath26-norm on @xmath27 if nothing else is specified .
we will use the standard scalar products on @xmath28 and @xmath29 defined by @xmath30 and @xmath31 we denote the semi - classical principal symbol of @xmath1 by @xmath32 the poisson bracket of @xmath33 and @xmath34 is given by @xmath35 let @xmath36 be the semiclassical principal symbol of the operator @xmath1 as in ( [ eqn : semclprinsymmodlop ] ) .
then , we define @xmath37 in the case of ( [ eqn : defnmodeloperator ] ) and ( [ eqn : semclprinsymmodlop ] ) @xmath38 is already closed due to the ellipticity of @xmath1 .
the spectrum of @xmath1 is discrete with simple eigenvalues , given by @xmath39 where @xmath40 .
next , for @xmath41 , consider the equation @xmath42 .
it has precisely two solutions @xmath43 where @xmath44 are given by @xmath45 and @xmath46 .
it has been shown in @xcite that for all @xmath47 and all @xmath48 @xmath49 with @xmath50 constants that only depend on @xmath51 .
this implies that such an @xmath51 is inside the @xmath52-pseudospectrum of @xmath1 .
alternatively , this fact follows as well from the work of n. dencker , j. sjstrand and m. zworski @xcite : since we have that @xmath53 for all @xmath54 it follows from @xcite that we can construct @xmath52-quasimodes @xmath55 of @xmath1 with the semiclassical wave front set @xmath56 .
we recall that for @xmath57 , @xmath58 , for some fixed @xmath59 , the semiclassical wave front set of @xmath60 is defined by @xmath61 where @xmath62 denotes the weyl quantization of @xmath63 .
+ next , by the natural projection @xmath64 and a slight abuse of notation we identify the points @xmath65 with points @xmath66 such that @xmath67 and @xmath68 .
furthermore , we will identify @xmath27 with the interval @xmath69 .
let us consider a random perturbation of @xmath1 given by @xmath70 where @xmath71 and @xmath72 is an integral operator from @xmath28 to @xmath73 defined as follows : let @xmath74 , for @xmath75 , and define @xmath76 with @xmath77 big enough and @xmath78 , @xmath79 .
moreover , @xmath80 are complex valued independent random variables with complex gaussian distribution law @xmath81 .
it has been observed by w. bordeaux - montrieux in @xcite that the hilbert - schmidth norm of @xmath72 is bounded with probability close to @xmath82 , indeed @xmath83 hence , we restrict our probability space to a open ball @xmath84 , with @xmath85 , of radius @xmath86 and centered at @xmath87 .
there , @xmath72 is a compact operator , so the spectrum of the perturbed operator @xmath8 is discrete .
+ in the present work we are interested in the eigenvalues of @xmath8 in the interior of the pseudospectrum , indeed in the sequel we will always assume that there exists a @xmath88 such that @xmath89 it will be very useful to give bounds on the coupling constant @xmath3 in terms of the imaginary part of the action between @xmath90 and @xmath91 , defined by : [ defn : zeroorderdensity ] let @xmath51 be as in , let @xmath33 denote the semiclassical principal symbol of @xmath1 in ( [ eqn : semclprinsymmodlop ] ) and let @xmath92 be as above .
define @xmath93 we suppose that the coupling constant @xmath71 in ( [ eqn : defpertp ] ) satisfies @xmath94 with @xmath95 for some @xmath96 and @xmath97 large and where the last inequality is uniform in @xmath98 .
thus , @xmath3 satisfies the inequality @xmath99 in @xcite m. hager showed that , with a probability close to @xmath82 , the eigenvalues of the perturbed operator @xmath8 contained in @xmath51 follow a weyl law , i.e. @xmath100 this result was extended to a much more general class of semiclassical pseudo - differential operators by m. hager and j. sjstrand in @xcite and w. bordeaux - montrieux gave in @xcite an extension in the @xmath82-dimensional case to sets @xmath51 which satisfy @xmath101)^{2/3}$ ] .
furthermore , a similar result has been obtained by t. christiansen and m. zworski for randomly perturbed toeplitz operators in @xcite . to the best of the author
s knowledge there has never been a study of the interaction between eigenvalues of the class of non - selfadjoint operators in which is crucial for the understanding of the spectral behavior of such operators .
the principal aim of this work is to study the correlation between two points in the spectrum of the perturbed operator @xmath8 .
therefore , following the approach used in @xcite , we study the moments of linear statistics of the random point process of eigenvalues of the operator @xmath8 , defined as following : let @xmath8 be as in ( [ eqn : defpertp ] ) , then we define the point process @xmath102 where the eigenvalues are counted according to their multiplicities and @xmath103 denotes the dirac - measure at @xmath104 .
such an approach is more classical in the study of zeros of random polynomials and gaussian analytic functions ; we refer the reader to the works of b. shiffman and s. zelditch @xcite , m. sodin @xcite an the book @xcite by j. hough , m. krishnapur , y. peres and b. virg . in @xcite
, we studied the first moment of the point process @xmath105 with the random variables @xmath106 restricted to a ball @xmath107 with @xmath86 , i.e. the measure @xmath108 defined by @xmath109 = \int_{{\mathds{c}}}\varphi(z ) d\mu_1(z)\ ] ] for all @xmath110 with @xmath111 as in .
in particular , showed that @xmath108 is absolutely continuous with respect to the lebesgue measure and we gave a precise @xmath0-asymptotic formula for the lebesgue density of @xmath108 . the above
described approached gives a detailed description of the average density of eigenvalues and in comparison to the works of m. hager in @xcite and of w. bordeaux - montrieux in @xcite , * we can allow for a smaller coupling constant , i.e. @xmath112 ; * we can study the distribution of eigenvalues of @xmath8 in all of @xmath6 . for more details , we refer to @xcite .
in this work , we are interested in the second moment of the point process @xmath105 .
we recall some facts about second moments of point processes from @xcite , using the example of @xmath105 .
the second moment of @xmath105 is defined by the positive linear functional on @xmath113 , @xmath114 , defined by @xmath115 = \int_{{\mathds{c}}^2}\varphi(z , w ) d\mu_2(z , w)\ ] ] for all @xmath116 .
note that we have the splitting @xmath117 + { \mathds{e}}\left [ \sum_{\substack{z , w\in\sigma(p_h^{\delta } )
\\ z\neq w } } \varphi(z , w)\mathds{1}_{b(0,r)}\right ] \notag \\ & = \int_{{\mathds{c}}^2}\varphi(z , z ) d\widetilde{\mu}_2(z , z ) + \int_{{\mathds{c}}^2}\varphi(z , w ) d\nu(z , w).\end{aligned}\ ] ] both terms are positive linear functionals on @xmath113 , and thus the above representation by the two measures @xmath118 and @xmath119 is well - defined .
the measure @xmath118 is supported on the diagonal @xmath120 and is given by the push - forward of @xmath108 under the diagonal map @xmath121 , i.e. @xmath122 . the second measure , @xmath119 , is called the two - point intensity measure of @xmath105 and it is supported on @xmath123 .
their sum naturally yields @xmath124 , i.e. @xmath125 .
hence , @xmath124 has atoms on the diagonal @xmath126 and we see that @xmath124 is not absolutely continuous with respect to the lebesgue measure on @xmath127
. however , this may be the case for the measure @xmath119 .
+ since @xmath108 has already been treated in @xcite , the aim of this work is to study the two - point intensity measure @xmath119 , given by @xmath128 = \int_{{\mathds{c}}^2}\varphi(z , w ) d\nu(z , w),\end{aligned}\ ] ] and to give an asymptotic formula for its lebesgue density which will give us the correlated behavior of two points of the spectrum of @xmath8 . throughout this work
we shall denote the lebesgue measure on @xmath129 by @xmath130 ; denote @xmath131 ; work with the convention that when we write @xmath132 then we mean implicitly that @xmath133 ; denote by @xmath134 that there exists a constant @xmath97 such that @xmath135 ; write @xmath136 , with @xmath137 , if @xmath138 .
from the assumptions @xmath18 we see that the direct image @xmath139 of the symplectic volume form @xmath140 on @xmath141 is absolutely continuous and let @xmath142 denote its lebesgue density , i.e. @xmath143 we give an @xmath0-asymptotic formula for the lebesgue density of the two - point intensity measure @xmath119 valid up to a distance @xmath144 to the diagonal . for @xmath51 as in and @xmath145 , we define the set @xmath146 and we prove the following result : [ thm:2pt_density ] let @xmath111 be as in .
let @xmath71 be as in with @xmath147 .
let @xmath119 be the measure defined in and let @xmath142 be as in .
then , for @xmath148 with @xmath88 large enough , there exist smooth functions * @xmath149 , * @xmath150 , * @xmath151 , with @xmath152 and there exists a constant @xmath153 such that for all @xmath154 with @xmath155 we prove this theorem in section [ sec : correlationformula ] .
we will prove that two eigenvalues of @xmath8 exhibit the following interaction : [ prop:2pt_dens_prop ] under the hypothesis of theorem [ thm:2pt_density ] , we have that * for @xmath156 @xmath157 * for @xmath158 @xmath159 let us give some comments on this result : that the we can not analyze the eigenvalue interaction completely up to the diagonal is due to some technical difficulties . in the above proposition , two eigenvalues of the perturbed operator @xmath8
show the following types of interaction : * _ short range repulsion _ :
the two - point density decays quadratically in @xmath160 if two eigenvalues are too close . *
_ long range decoupling _ : if the distance between two eigenvalues is @xmath161 the two - point density is given by the product of two one - point densities .
we have shown in @xcite ( see as well @xcite ) that the one - point density of eigenvalues in @xmath51 as in , is given by @xmath162 = \int \varphi(z)d(z;h)l(dz ) \quad \varphi\in\mathcal{c}_0(\omega),\ ] ] where @xmath163 it follows from the assumptions @xmath18 that @xmath164 for all @xmath48 as in . hence , under the assumptions of theorem [ thm:2pt_density ] , the conditional average density of eigenvalues of @xmath8 given that @xmath165 is well defined and given by @xmath166 we have the following asymptotic behavior of conditional average density @xmath167 : [ prop:2pt_dens_cond ] under the hypothesis of theorem [ thm:2pt_density ] , we have that for @xmath168 * for @xmath169 @xmath170 * for @xmath171 @xmath172 in the above proposition we see that , given an eigenvalue @xmath165 , the density of finding another eigenvalue in the vicinity of @xmath173 shows the following behavior : * _ short range repulsion _ : the density @xmath167 decays quadratically in @xmath174 if the distance between @xmath104 and @xmath173 is smaller than a term of order @xmath175 .
it follows from @xmath18 that @xmath142 grows towards the boundary of @xmath6 , hence the short range repulsion is weaker for @xmath51 close to the boundary of @xmath6 , as we expected from the numerical simulations presented in @xcite . *
_ long range decoupling _ : if the distance between @xmath104 and @xmath173 is larger than a term of order @xmath176 , the density @xmath167 is given up to a small error by the @xmath82-point density @xmath177 ( see ) .
hence , we see that at these distances two eigenvalues of @xmath8 are up to a small error uncorrelated . to illustrate proposition [ prop:2pt_dens_cond ] , figure [ fig : cond_dens ] shows a plot of of the principal terms of the conditional density @xmath178 as a function of @xmath179 , for @xmath180 and @xmath181 , assuming for simplicity that @xmath182 . on the left hand side of the graph
we see the quadratic decay , whereas on the right hand side the density is given by @xmath183 . for @xmath180.,scaledwidth=50.0% ]
in this section we will give a short reminder of a well - posed grushin problem for the perturbed operator @xmath8 which has already been used in @xcite .
we will then employ the resulting effective hamiltonians to derive a formula for the two - point intensity measure defined in .
we begin by giving a short refresher on grushin problems .
they have become an essential tool in microlocal analysis and are employed with great success in a vast number of works . as reviewed in @xcite , the central idea is to set up an auxiliary problem of the form @xmath184 where @xmath185 is the operator under investigation and @xmath186 are suitably chosen .
we say that the grushin problem is well - posed if this matrix of operators is bijective .
if @xmath187 , on typically writes @xmath188 the key observation goes back to the shur complement formula or , equivalently , the lyapunov - schmidt bifurcation method , i.e. the operator @xmath189 is invertible if and only if the finite dimensional matrix @xmath190 is invertible and when @xmath190 is invertible , we have @xmath191 @xmath190 is sometimes called effective hamiltonian .
+ next , we give a short reminder of the grushin problem used to study @xmath8 .
first , we introduce the following auxiliary operators which have already been used by m. hager j. sjstrand in @xcite . for @xmath192
we consider @xmath193 and @xmath194 , two @xmath104-dependent elliptic self - adjoint operators from @xmath28 to @xmath28 , defined by @xmath195 with natural domains given by @xmath196 . since @xmath27 is compact and these are elliptic , non - negative , self - adjoint operators their spectra are discrete and contained in the interval @xmath197 . since @xmath198 it follows that @xmath199 and @xmath200 .
furthermore , if @xmath201 is an eigenvalue of @xmath193 with corresponding eigenvector @xmath202 we see that @xmath203 is an eigenvector of @xmath204 with the eigenvalue @xmath205 .
similarly , every non - vanishing eigenvalue of @xmath204 is an eigenvalue of @xmath193 and moreover , since @xmath206 , @xmath207 are fredholm operators of index @xmath87 we see that @xmath208 .
hence the spectra of @xmath193 and @xmath204 are equal @xmath209 now consider the orthonormal basis of @xmath28 @xmath210 consisting of the eigenfunctions of @xmath193 . by the previous observations we have @xmath211 thus defining @xmath212 to be the normalized eigenvector of @xmath213 corresponding to the eigenvalue @xmath214 and the vectors @xmath215 , for @xmath216 , as the normalization of @xmath217 such that @xmath218 yields an orthonormal basis of @xmath28 @xmath219 consisting of the eigenfunctions of @xmath204 . since @xmath220
we can conclude that @xmath221 .
following sjstrand in @xcite , we us the eigenfunctions of the operators @xmath222 and @xmath213 to create a well - posed grushin problem . the sequel is taken from @xcite , but it originates in the works of hager @xcite , bordeaux - montrieux @xcite and sjstrand @xcite .
[ prop : grushunpertop ] let @xmath223 with @xmath224 and let @xmath225 and @xmath212 be as in ( [ eqn : relationejfj ] ) . define @xmath226 then @xmath227 is bijective with the bounded inverse @xmath228 where @xmath229 , @xmath230 and @xmath231 and @xmath232 .
furthermore , we have the estimates for @xmath48 @xmath233 [ def : x ] for @xmath75 we denote the integer part of @xmath234 by @xmath235 as in section [ sec : introduction ] .
let @xmath77 be big enough as above and define @xmath236 .
let @xmath237 and @xmath212 be as in ( [ eqn : relationejfj ] ) , let @xmath54 and let @xmath238 and @xmath239 denote the fourier coefficients of @xmath237 and @xmath212 .
we define the vector @xmath240 to be given by @xmath241 [ prop : grushpertop ] let @xmath223 .
let @xmath59 be as in definition [ def : x ] and let @xmath242 be the ball of radius @xmath243 , @xmath97 large , centered at @xmath87 .
let @xmath244 be as in proposition [ prop : grushunpertop ]
. then @xmath245 is bijective with the bounded inverse @xmath246 where @xmath247 and @xmath248 with @xmath249 , @xmath250 , and @xmath251 here , the dot - product @xmath252 is the natural bilinear one .
[ rem : et_estim ] the effective hamiltonian @xmath253 depends smoothly on @xmath48 and holomorphically on @xmath254 . as in @xmath255 and
proposition @xmath256 in @xcite , we have the following estimates : for all @xmath48 , all @xmath250 and all @xmath257 @xmath258 where @xmath259 is as in definition [ defn : zeroorderdensity ] .
moreover , as remarked in @xcite the effective hamiltonian @xmath253 satisfies a @xmath260-equation , i.e. there exists a smooth function @xmath261 such that @xmath262 this implies that the zeros of @xmath253 are isolated and countable and we may use the same notion of multiplicity as for holomorphic functions .
by of the above well - posed grushin problem for the perturbed operator @xmath8 we have that @xmath263 .
hence , to study the the two - point intensity measure @xmath119 defined in ( [ eqn : nu ] ) , we investigate the integral @xmath264 with @xmath265 . using remark [ rem : et_estim ]
, we see that the integral is finite since the number of pairs of zeros of @xmath266 in @xmath267 is uniformly bounded for @xmath250 .
recall @xmath105 defined in . using ( * ? ? ?
* lemma 7.1 ) , we get the following regularization of the @xmath7-fold counting measure @xmath268 @xmath269 where @xmath270 such that @xmath271 . assuming that @xmath265 is such that @xmath272 , we see by the lebesgue dominated convergence theorem that the two - point intensity measure of the point process @xmath105 is given by @xmath273 with @xmath274 { \mathrm{e}}^{-\alpha^*\alpha } l(d\alpha).\ ] ] using , we see that the main object of interest is the random vector @xmath275 the first order term ( with respect to @xmath3 ) the correlation matrix of @xmath276 is given by the gramian matrix @xmath277 defined by @xmath278 with @xmath279 notice that the matrices @xmath280 depend on @xmath0
; see definition [ def : x ] .
next , we will state a formula for the lebesgue density of the two - point intensity measure @xmath119 in terms of the permanent of the shur complement of @xmath277 , i.e @xmath281 . for the definition of the permanent of a matrix we refer the reader to @xcite .
it is defined as follows : [ defn : permanent ] let @xmath282 be a square matrix and let @xmath283 denote the symmetric group of order @xmath284 .
the permanent of @xmath285 is defined by @xmath286 although the definition of the permanent resembles closely that of the determinant , the two object are quite different .
many properties known to hold true for determinants , fail to be true for permanents . for our purposes it is enough to note that it is multi - linear and symmetric . for more details concerning permanents and their properties we refer the reader to @xcite
. we will prove the following result : [ prop:2ptcorrelation ] let @xmath111 be as in .
let @xmath71 be as in hypothesis and let @xmath287
. moreover , let @xmath288 be as in .
then , there exists a smooth function @xmath289 and there exists a constant @xmath290 such that for all @xmath291 @xmath292 the proof of proposition [ prop:2ptcorrelation ] will take up most of the rest of this paper .
therefore we give a short overview on how we will proceed : in section [ sec : stph ] , we give a formula for the scalar product @xmath293 by constructing quasimodes for the operators @xmath294 and @xmath207 to approximate the eigenfunction @xmath237 and @xmath212 , and by using the method of stationary phase . in section [ susec : gramian ] , we will use this formula to study the invertibility of the matrices @xmath295 and @xmath296 .
furthermore , we will study the permanent of @xmath296 . in section [ sec : correlationformula ]
, we give a proof of proposition [ prop:2ptcorrelation ] .
in this section we are interested in the scalar product @xmath293 . recall from definition [ def : x ] that the vector @xmath297 , @xmath48 , is given by @xmath298 , where @xmath237 and @xmath212 are the eigenfunctions of the operators @xmath193 and @xmath194 , respectively , associated to their first eigenvalue @xmath214 .
the fourier coefficients @xmath299 and their @xmath104- and @xmath300-derivatives are of order @xmath301 , @xmath302 , for @xmath303 with @xmath97 large enough ( cf .
* propositions 5.5 and 5.6 ) ) .
the parseval identity implies that for @xmath304 @xmath305 the aim of this of this section is to prove the following result : [ prop : sp_xx ] let @xmath111 be as in hypothesis and let @xmath306 be as in .
furthermore , for @xmath48 let @xmath142 denote the lebesgue density of the direct image of the symplectic volume form on @xmath141 under the principal symbol @xmath33 , i.e. @xmath307 .
then , there exists a constant @xmath97 such that for all @xmath308 @xmath309 where : * @xmath310 is a family of smooth functions depending only on @xmath311 , which satisfy @xmath312 + { \mathcal{o}}(h^2 ) .
\end{aligned}\ ] ] and @xmath313 * @xmath314 is a family of smooth functions which are almost @xmath104-holomorphic and almost @xmath315-anti - holomorphic extensions from the diagonal @xmath316 of @xmath317 , i.e. @xmath318 moreover , we have that @xmath319 and for @xmath320 with @xmath321 , @xmath322 and @xmath323 * the function @xmath324 has the following symmetries : @xmath325 let us give some remarks on the above results : note that the formula for @xmath326 stated above is simply a special case of the more general taylor expansion @xmath327 with @xmath328 and @xmath329 .
+ next , we define for @xmath330 , as in proposition [ prop : sp_xx ] , @xmath331 from the above proposition we can immediately deduce some growth properties of certain quantities that will be become important in the sequel . [ cor : detper_a ] under the assumptions of proposition [ prop : sp_xx ] , * @xmath332 * @xmath333 @xmath334 * @xmath335 @xmath336 to prove proposition [ prop : sp_xx ] , we shall study the scalar products @xmath337 and @xmath338
. we will prove [ prop : sp_ee ] let @xmath111 be as in hypothesis and let @xmath339 be as in .
then , there exists a constant @xmath97 such that for all @xmath308 @xmath340 where : * @xmath341 is a family of smooth functions depending only on @xmath311 , which satisfy @xmath342 * @xmath343 is a family of smooth functions which are almost @xmath104-holomorphic and almost @xmath315-anti - holomorphic extensions from the diagonal @xmath316 of @xmath344 , i.e. @xmath345 moreover , for @xmath320 with @xmath346 , one has that @xmath347 and that @xmath348 * the function @xmath349 has the following symmetries : @xmath350 to prove proposition [ prop : sp_ee ] , we begin by constructing an oscillating function to approximate @xmath351 .
let us recall from section [ sec : introduction ] that the points @xmath21 denote the minimum and the maximum of @xmath352 and that for @xmath48 the points @xmath353 are the unique solutions to the equation @xmath354 .
furthermore , we will identify frequently @xmath27 with the interval @xmath69 .
moreover , let us recall that by the natural projection @xmath64 we identify the points @xmath65 with points @xmath66 such that @xmath355 .
+ let @xmath356b-2\pi , a[$ ] be an open interval such that @xmath357 for all @xmath48 .
let @xmath358b-2\pi , a[)$ ] and define for @xmath359 @xmath360 where , for a fixed @xmath361 , @xmath362 [ rem : holo ] note that the function @xmath363 is solution to @xmath364 on @xmath365 , since the phase function @xmath366 satisfies the eikonal equation @xmath367 furthermore , let us remark that @xmath368 depends holomorphically on @xmath104 .
next , we are interested in the @xmath26-norm of @xmath369 .
[ lem : norm_qm_e ] let @xmath51 be as in .
then , there exists a family of smooth functions @xmath341 , such that @xmath370 and @xmath371 in view of the definition of @xmath372 , see and , one gets that @xmath373 the critical point for @xmath374 is given by the equation @xmath375 the critical point is unique and is given by @xmath376 and satisfies @xmath377 , see .
this implies in particular that the critical point is non - degenerate .
more precisely , @xmath378 the critical value of @xmath379 is given by @xmath380 using the method of stationary phase , one gets @xmath381 where @xmath382 is smooth in @xmath104 . using
, one gets that @xmath383 recall from that the function @xmath237 is an eigenfunction of the operator @xmath193 ( c.f section [ sec : auxope ] ) corresponding to its first eigenvalue @xmath214 .
we set @xmath384 where @xmath385 denotes the spectral projection for @xmath193 onto the eigenspace associated with @xmath214 .
next , we prove that up to an exponentially small error in @xmath386 , @xmath237 is given by the normalization of @xmath369 .
let @xmath51 be as in .
then , there exists a constant @xmath97 such that for all @xmath48 and all @xmath387 @xmath388 the proof of the lemma is similar to the proof of ( * ? ? ?
* proposition 3.11 ) .
this result implies that @xmath389 by remark [ rem : holo ] , @xmath390 is holomorphic in @xmath104 and anti - holomorphic in @xmath315 . we can study this scalar product by the method of stationary phase : in view of
, it remains to study the oscillatory integral @xmath391 where @xmath368 is given in and @xmath392 is defined by @xmath393 using , we see that @xmath394 dy.\ ] ] since the imaginary part of @xmath392 can be negative , we shift the phase function by the + * minimum of @xmath395 . *
the critical points of the function @xmath396 are given by the equation @xmath397 .
since @xmath51 is convex , this equation has , for @xmath160 small enough , on the support of @xmath398 the unique solution @xmath399 and it satisfies @xmath400 ( cf . ) .
moreover , it depends smoothly on @xmath104 and @xmath315 since @xmath401 is smooth .
therefore , @xmath402 which implies that @xmath403 is a minimum point , and that @xmath404 dy \leq 0 .
\end{aligned}\ ] ] we define @xmath405 , and notice that @xmath406 .
hence , we can write as follows : @xmath407 to study @xmath408 by the method of stationary phase , we are interested in the + * critical points of @xmath409 . * clearly they are the same as for @xmath410 .
note that for @xmath411 one has that @xmath412 which has , on the support of @xmath398 , the unique critical point @xmath339 and it satisfies @xmath413 ( cf . ) .
therefore , @xmath414 which implies that @xmath415 is a non - degenerate critical point . in the case where @xmath416 the situation is more complicated . by we see that if @xmath417 , for @xmath160 small enough , the critical point is real and given by @xmath403 , i.e. the minimum point of @xmath395 . however , if @xmath418 , we need to consider an almost @xmath234-analytic extension of @xmath392 , which we shall denote by @xmath419 .
as described in @xcite , the `` critical point '' of @xmath419 is then given by @xmath420 and we will see , by the following result , that it `` moves '' to the complex plane . [
lem : cp_ee ] let @xmath111 be as in .
let @xmath398 be as in ( [ eqn : holoquasmod ] ) and let @xmath33 be the principal symbol of @xmath1 ( cf .
section [ sec : introduction ] ) .
let @xmath421 be as in .
furthermore , let @xmath422 denote an almost analytic extension of @xmath366 to a small complex neighborhood of the support of @xmath398 , and define @xmath423 .
then , the there exists a @xmath97 such that for @xmath330 the function @xmath424 has exactly one zero , @xmath425 , and : * it depends almost holomorphically on @xmath104 and almost anti - holomorphically @xmath315 at the diagonal @xmath426 , i.e. @xmath427 * it is non - degenerate in the sense that @xmath428 * for @xmath304 with @xmath429 , @xmath88 large enough , one has @xmath430 the proof of lemma [ lem : cp_ee ] will be given the proof of proposition [ prop : sp_ee ] .
let @xmath419 denote an almost @xmath234-analytic extension of @xmath392 . using the method of stationary phase for complex - valued phase functions ( cf .
theorem 2.3 in @xcite ) and lemma [ lem : cp_ee ] , one gets that @xmath431 using that lemma [ lem : norm_qm_e ] and imply @xmath432 , we obtain from the above and .
+ in , @xmath433 is given by the critical value of @xmath434 and by the logarithm of the amplitude @xmath435 , given by the stationary phase method , i.e. @xmath436 and @xmath437 which depends smoothly on @xmath104 and @xmath315 in the sense that all @xmath104-,@xmath438-,@xmath315- and @xmath439-derivatives remain bounded as @xmath2 .
@xmath440 is by definition @xmath104-holomorphic , @xmath315-anti - holomorphic and smooth in @xmath234 . by lemma [ lem : cp_ee ] , we know that the critical point @xmath441 is almost @xmath104-holomorphic and almost @xmath315-anti - holomorphic in @xmath442 , a small neighborhood of the diagonal @xmath411 . hence , @xmath326 is almost @xmath104-holomorphic and almost @xmath315-anti - holomorphic in @xmath442 .
equivalently , @xmath326 is an almost @xmath104-holomorphic and almost @xmath315-anti - holomorphic extension from the diagonal of @xmath443 . since @xmath444
, we obtain by taylor expansion up to order @xmath7 of @xmath326 at @xmath445 , that @xmath446 for @xmath160 small enough .
similarly , @xmath447 which implies that @xmath448 concluding the proof of the second point of the proposition .
+ finally , let us give a give a proof of the stated symmetries .
the fact that @xmath449 follows directly from the fact that @xmath450 .
one then computes that @xmath451 which concludes the proof of the proposition .
we are interested in the solutions of the following equation : @xmath452 where @xmath453 denotes an almost analytic extension of @xmath401 .
since @xmath454 , it follows from the assumptions on @xmath401 that @xmath455 for all @xmath456 .
since @xmath401 depends smoothly on @xmath234 , there exists a small complex open neighborhood @xmath457 of @xmath458 such that @xmath459 and such that for all @xmath460 @xmath461 thus , it follows by the implicit function theorem , that for @xmath330 , with @xmath97 large enough , there exists a unique solution @xmath425 to ( [ eqn : cp_psi+ ] ) and it depends smoothly on @xmath330 .
furthermore , we have that @xmath462 .
taking the @xmath104- and @xmath300- derivative of ( [ eqn : cp_psi+ ] ) at the critical point @xmath463 yields that @xmath464 and similarly that @xmath465 using that @xmath466 , one calculates that for @xmath411 we have that @xmath467 where @xmath468 taylor s theorem implies that @xmath469 recall that the principal symbol of the operator @xmath1 is given by @xmath470 ( cf .
section [ sec : introduction ] ) , which implies that @xmath471 . to conclude the symmetric form of the taylor expansion stated in the lemma , we expand around the point @xmath445 , for @xmath160 small enough , with @xmath472 and @xmath473 , which is possible since @xmath51 is by assumed to be convex .
finally , by taking the imaginary part of the taylor expansion of @xmath463 , we conclude by ( [ eqn : cp_zder ] ) and ( [ eqn : cp_wder ] ) that @xmath474 we have , as in section [ susec : sp_ee ] , [ prop : sp_ff ] let @xmath111 be as in hypothesis and let @xmath475 be as in .
then , there exists a constant @xmath97 such that for all @xmath308 @xmath476 where : * @xmath477 is a family of smooth functions depending only on @xmath478 , which satisfy @xmath479 * @xmath480 is a family of smooth functions which are almost @xmath104-holomorphic and almost @xmath315-anti - holomorphic extensions from the diagonal @xmath316 of @xmath481 , i.e. @xmath482 moreover , for @xmath320 with @xmath346 , one has that @xmath483 and that @xmath484 * the function @xmath485 has the following symmetries : @xmath486 before the proof of proposition [ prop : sp_xx ] , let us give a short description of the connection between the functions @xmath344 , @xmath481 in proposition [ prop : sp_ee ] , [ prop : sp_ff ] , and the symplectic volume form on the phase space @xmath141 .
[ prop : hess_phas_symplvol ] let @xmath487 be as in and let @xmath382 and @xmath488 be as in propositions [ prop : sp_ee ] and [ prop : sp_ff ] .
furthermore , let @xmath33 be the principal symbol of @xmath1 , let @xmath489 be the two solutions to @xmath490 , see section [ sec : introduction ] .
then , @xmath491 \notag \\ & = \frac{1}{4}\left ( \frac{1}{\frac{1}{2i}\{\overline{p},p\}(\rho_{-}(z ) ) } + \frac{1}{\frac{1}{2i}\{p,\overline{p}\}(\rho_{+}(z ) ) } \right ) + { \mathcal{o}}(h ) \end{aligned}\ ] ] is , up to an error of order @xmath0 , the lebesgue density of the direct image , under the principal symbol @xmath33 , of the symplectic volume form @xmath492 on @xmath141 , i.e. @xmath493 using that @xmath494 , with @xmath495 , is the solution to @xmath496 with @xmath497 ( cf .
section [ sec : introduction ] ) , we get that @xmath498 using propositions [ prop : sp_ee ] and [ prop : sp_ff ] , one then computes that @xmath499 since @xmath500 , we conclude by the result of ( * ? ? ?
* proposition 6.2 ) that @xmath501 l(dz ) = \frac{1}{4 } p_*(d\xi\wedge dx ) + { \mathcal{o}}(h)l(dz ) .
\qedhere\ ] ] the results follow immediately from and the propositions [ prop : sp_ee ] , [ prop : sp_ff ] and [ prop : hess_phas_symplvol ] .
the aim of this section is to study the gramian matrix @xmath277 defined in by @xmath502 where @xmath503 the invertibility of the matrix @xmath277 will be essential to the proof of proposition [ prop:2ptcorrelation ] .
indeed , we prove the following [ prop : graminvert ] let @xmath111 be as in and let @xmath304
. then , @xmath504 to prove proposition [ prop : graminvert ] we will first study the matrices @xmath10 and , if @xmath505 exists , the matrix @xmath296 given by the shur complement formula applied to @xmath277 , i.e. @xmath506 we begin by studying the determinant of @xmath10 .
it is non - zero if and only if the vectors @xmath297 and @xmath507 are not co - linear .
in particular we are interested in a lower bound of this determinant for @xmath104 and @xmath315 close .
[ prop : deta ] let @xmath111 be as in hypothesis . for @xmath508 , with @xmath88 large enough ( cf . proposition [ prop : sp_xx ] ) , we have @xmath509 where @xmath510 is as in .
moreover , * for @xmath511 @xmath512 * for @xmath513 @xmath514 * let @xmath515 and let @xmath88 be large enough , then for @xmath516 , @xmath517 since the matrix @xmath10 is self - adjoint , we have a lower bound on the matrix norm of @xmath10 by its smallest eigenvalue . using proposition [ prop : sp_xx ]
we see that @xmath518 and on calculates that for a fixed @xmath515 and for @xmath519 the two eigenvalues of @xmath10 are given by @xmath520 and by taylor expansion we conclude the following result : [ cor : lb_mn_a ] under the assumptions of proposition [ prop : deta ] , we have that for @xmath521 and @xmath519 @xmath522 by corollary [ cor : detper_a ] , one has that @xmath523 with @xmath524 the first two estimates are then an immediate consequence of the above formula . in the case where @xmath525 , one computes , using taylor s formula , that @xmath526 which implies that @xmath527 we prove the following result . [ prop : mat_gamma ] let @xmath111 be as in , and let @xmath528 and @xmath324 for @xmath529 be as in proposition [ prop : sp_xx ] .
for @xmath529 let @xmath510 be as in and define @xmath530 then , for @xmath515 and @xmath531 , with @xmath88 large enough , we have that @xmath532 we will give a proof of this result further below .
first , we state formulae for the trace , the determinant and the permanent of @xmath296 .
[ cor : det_gamma ] under the assumptions of proposition [ prop : mat_gamma ] , we have that @xmath533,\end{aligned}\ ] ] @xmath534 \notag \\ & + \frac{4}{h^2 } \left ( ( \partial_{z\overline{w}}^2\psi)(z , z;h ) ( \partial_{z\overline{w}}^2\psi)(w , w;h ) - ( \partial_{z\overline{w}}^2\psi)(z , w;h ) ( \partial_{z\overline{w}}^2\psi)(w , z;h ) { \mathrm{e}}^{-\frac{2}{h}k(z , w ) } \right ) \notag \\ & + { \mathcal{o}}(h^{\infty})\end{aligned}\ ] ] and that @xmath535 the result follows from a direct computation using proposition [ prop : mat_gamma ] ; for the definition of the permanent of a matrix see definition [ defn : permanent ] .
we have the following bound on the trace of @xmath296 : [ prop : trace_est_gam ] under the assumptions of proposition [ prop : mat_gamma ] , we have that for @xmath536 @xmath537 let us turn to the proofs of the above propositions .
we begin by considering a very helpful congruency transformation . in view of proposition
[ prop : sp_xx ] , we prove [ lem : conj ] let @xmath111 be as in , and let @xmath528 , @xmath317 and @xmath324 be as in proposition [ prop : sp_xx ] , for @xmath529 .
define the matrices @xmath538 @xmath539 and @xmath540 with @xmath541 .
then , for @xmath542 , we have that @xmath543 to abbreviate the notation , we define for @xmath529 the following function @xmath544 by proposition [ prop : sp_xx ] , we see that @xmath276 is bounded by @xmath82 and that all its derivatives are bounded polynomially in @xmath545 .
furthermore , the matrices @xmath546 and @xmath547 are given by @xmath548 where @xmath529 and @xmath549 and @xmath550 and @xmath551 one computes that @xmath552{\mathrm{e}}^{-\frac{1}{h}\phi(z;h)-\frac{1}{h}\phi(w;h ) } { \mathrm{e}}^{\frac{2}{h}\psi(z , w ) } \notag \\ & \phantom{= } + { \mathcal{o}}_{\mathcal{c}^{\infty}}\!\left(h^{\infty}\right ) , \end{aligned}\ ] ] and that @xmath553 \left[2(\partial_{\overline{w}}\psi)(z , w;h ) -(\partial_{\overline{w}}\phi)(w;h)\right ] + \notag \\ & \phantom{= \frac{1}{h^2}- } 2h(\partial_{z\overline{w}}^2\psi)(z , w;h ) \big]{\mathrm{e}}^{-\frac{1}{h}\phi(z_1;h)-\frac{1}{h}\phi(z_2;h ) } { \mathrm{e}}^{\frac{2}{h}\psi(z_1,z_2 ) } + { \mathcal{o}}_{\mathcal{c}^{\infty}}\!\left(h^{\infty}\right ) .
\end{aligned}\ ] ] using that @xmath554 and that @xmath555 for @xmath542 ( cf .
proposition [ prop : deta ] ) , we see that @xmath556 defining , @xmath557 we see that @xmath558 a direct computation then yields that @xmath559 in view of lemma [ lem : conj ] , it remains to consider the matrix @xmath560 in the sequel we will suppress the @xmath0-dependency of the function @xmath326 to abbreviate our notation . recall the definition of @xmath561 from lemma [ lem : conj ] and note that @xmath562 for @xmath563 , proposition [ prop : sp_xx ] implies that @xmath564 is positive .
hence , the inverse of @xmath561 exists and is given by @xmath565 to calculate @xmath566 , we use lemma [ lem : conj ] and the symmetries of the function @xmath567 given in proposition [ prop : sp_xx ] .
indeed , one gets that @xmath568 and one computes that @xmath569 is given by @xmath570 with @xmath571 { \mathrm{e}}^{\frac{1}{h } ( 2\psi(z , z ) + 4 { \mathrm{re\,}}\psi(z , w))},\end{aligned}\ ] ] @xmath572 { \mathrm{e}}^{\frac{2}{h } ( \psi(z , z ) + \psi(z , w)+ \psi(w , w))},\end{aligned}\ ] ] and @xmath573 { \mathrm{e}}^{\frac{1}{h } ( 2\psi(w , w ) + 4 { \mathrm{re\,}}\psi(z , w))}.\end{aligned}\ ] ] since the matrix @xmath285 is clearly self - adjoint , one has that @xmath574 . comparing the coefficients of @xmath285 with with those of @xmath575 ( cf .
lemma [ lem : conj ] ) and using the symmetries of @xmath326 ( cf .
proposition [ prop : sp_xx ] ) , we see that @xmath576 with @xmath577 as in the hypothesis of proposition [ prop : mat_gamma ] .
recall from that the function @xmath510 is defined by @xmath578 where @xmath579 .
using , we find that the first matrix in is equal to @xmath580 it follows by lemma [ lem : conj ] that @xmath581 in the last equality we used that @xmath582 is bounded from below by a power of @xmath0 ; see lemma [ lem : conj ]
. carrying out the matrix multiplication @xmath583 implies the statement of the proposition .
the shur complement formula yields that the determinant of the gramian matrix @xmath277 is given by @xmath584 .
hence , using proposition [ prop : deta ] and corollary [ cor : det_gamma ] , we see that @xmath585 \notag \\ & + \frac{4}{h^2 } \left ( ( \partial_{z\overline{w}}^2\psi)(z , z;h ) ( \partial_{z\overline{w}}^2\psi)(w , w;h ) - ( \partial_{z\overline{w}}^2\psi)(z , w;h ) ( \partial_{z\overline{w}}^2\psi)(w , z;h ) { \mathrm{e}}^{-\frac{2}{h}k(z , w ) } \right ) \notag \\ & ~~\cdot\left(1 - { \mathrm{e}}^{-\frac{2}{h } k(z , w ) } + { \mathcal{o}}(h^{\infty})\right ) + { \mathcal{o}}(h^{\infty}).\end{aligned}\ ] ] next we consider the taylor expansion of the terms @xmath586 and @xmath587 up to first order .
similarly as in proposition [ prop : sp_xx ] , we develop around the point @xmath588 and get that @xmath589 and @xmath590 moreover , one has that for @xmath591 @xmath592 since we suppose that @xmath593 , the above error term is equal to @xmath594 .
since @xmath595 is evaluated at a point on the diagonal , it follows from proposition [ prop : sp_xx ] , that @xmath596 plugging the above taylor expansion into , one gets that @xmath597 \left(1 - { \mathrm{e}}^{-\frac{2}{h } k(z , w ) } + { \mathcal{o}}(h^{\infty})\right ) \notag \\ & \phantom{=}\left . -4 { \mathrm{e}}^{-\frac{2}{h } k(z , w ) } \left ( \left ( \frac{\sigma_{h}(z , w)|z - w|^2}{4h } \right)^2(1+{\mathcal{o}}(|z - w| ) ) + \frac{\sigma_{h}(z , w)|z - w|^2}{4h } { \mathcal{o}}(|z - w| ) \right ) \right\ } \notag \\ & \phantom{= } + { \mathcal{o}}(h^{\infty } ) \notag \\ & = \frac { \sigma_{h}(z , w)^2}{4h^2 } \bigg\ { \left ( 1 - { \mathrm{e}}^{-\frac{2}{h } k(z , w ) } \right)^2 + { \mathcal{o}}(|z - w| ) \left(1 - { \mathrm{e}}^{-\frac{2}{h } k(z , w)}\right ) + { \mathcal{o}}(h^{\infty } ) \notag \\ & \phantom{=}\left . -4 { \mathrm{e}}^{-\frac{2}{h } k(z , w ) } \left [ \left ( \frac{\sigma_{h}(z , w)|z - w|^2}{4h } \right)^2 + { \mathcal{o}}\left(\frac{|z - w|^5}{h^2}\right ) + { \mathcal{o}}\left(\frac{|z - w|^3}{h}\right ) \right ] \right\}.\end{aligned}\ ] ] recall from that @xmath598 ,
wherefore we see that @xmath599 is positive for @xmath600 .
next , we suppose that @xmath601 .
hence , one gets that @xmath602 \right\}.\end{aligned}\ ] ] using the taylor expansion of the @xmath603 and , one gets that @xmath604 note that the principal term on the right hand side dominates the error terms . the same holds true for the other error terms in .
next let us suppose that @xmath605 . since @xmath606 it follows by and that @xmath607 is positive for @xmath593 . using , and , one gets that @xmath608 .
\end{aligned}\ ] ] since @xmath609 it follows that for @xmath536 the trace of @xmath296 is positive .
furthermore , the above inequality applied to , implies the upper bound stated in the proposition .
the permanent of the matrix @xmath296 is vital to the @xmath7-point density of eigenvalues and therefore , we shall give a more detailed description of it than the one given in corollary [ cor : det_gamma ] .
[ prop : perm_gamma ] let @xmath610 be as in theorem [ thm:2pt_density ] and let @xmath510 be as in . under the assumptions of proposition [ prop : mat_gamma ] , we have that for @xmath515 and @xmath531 , @xmath611 .
\end{aligned}\ ] ] applying , and to the formula for @xmath612 given in proposition [ prop : trace_est_gam ] and using the notation introduced in , one gets that @xmath613 thus , one computes that @xmath614 \notag \\ & + \frac{1}{4h^2 } \left ( \sigma_h(z , z ) \sigma_h(w , w ) + \sigma_h(z , w;h)^2(1 + { \mathcal{o}}(|z - w| ) { \mathrm{e}}^{-\frac{2}{h}k(z , w ) } \right ) \notag \\ & + { \mathcal{o}}(h^{\infty})\end{aligned}\ ] ] and we conclude the statement of the proposition .
we begin by proving the results of theorem [ thm:2pt_density ] , proposition [ prop:2pt_dens_prop ] and of proposition [ prop:2pt_dens_cond ] . first ,
let us treat the case of the long range interaction : we suppose that @xmath615 . here , we have that for any power @xmath515 the term
@xmath616 remains bounded . using that @xmath617 with @xmath618 and using that @xmath619 , it follows that @xmath620 next , we consider the case where @xmath156 .
recall from theorem [ thm:2pt_density ] that @xmath621 with @xmath623 equal to @xmath624 similarly to , we have that @xmath625 .
we start by considering the first term in : @xmath626 set @xmath627 . using the taylor expansions of the functions @xmath603 , @xmath628 and @xmath629 , one computes , that is equal to @xmath630 \end{aligned}\ ] ] which simplifies to @xmath631 hence ,
@xmath632 which concludes the proof .
[ lem : implfunthm ] let @xmath634 , let @xmath635 , with @xmath636 , and let @xmath637 denote the complex open ball of radius @xmath638 centered at @xmath87 . for @xmath639 , define @xmath640 .
we consider a holomorphic function @xmath641 such that * for all @xmath642 the jacobian of @xmath276 with respect to @xmath315 is given by @xmath643 where @xmath644 is a matrix - valued holomorphic function and * @xmath645 such that @xmath646 for all @xmath642 .
then , for all @xmath639 and for all @xmath647 , with @xmath648 , the equation @xmath649 has exactly one solution @xmath650 , it satisfies @xmath651 and it depends holomorphically on @xmath104 and on @xmath652 .
observe that the choice of @xmath653 yields a uniform lower bound on @xmath654 and so we can choose the radius of the ball @xmath655 uniformly in @xmath104 .
this will become important in the proof of proposition [ prop:2ptcorrelation ] .
let @xmath639 and set @xmath656 we begin by observing that @xmath657 is invertible for all @xmath658 and the norm of the inverse is bounded ( uniformly in @xmath104 ) .
indeed , for one has that @xmath659 + _ claim @xmath660 _ : @xmath661 is injective . let @xmath662 and define @xmath663 .
hence , with @xmath664 , we have that @xmath665 thus , @xmath666 where @xmath667 .
therefore , @xmath668 , and we see that @xmath669 is invertible and the norm of its inverse is @xmath670 ( uniformly in @xmath104 ) . hence , @xmath671 and we conclude that @xmath661 is injective .
in particular , we have proven the uniqueness of the solution to the equation .
+ for @xmath676 , we take @xmath677 .
using the fact that @xmath678 is invertible everywhere , the implicit function theorem implies that for all @xmath679 there exists a solution @xmath674 , if @xmath680 is small enough ( cf . ) .
let @xmath673 , and define @xmath681 .
let @xmath682 $ ] be the supremum of @xmath683 $ ] such that there exists a solution to @xmath684 for all @xmath685 .
we have already proven that @xmath686 . as @xmath687
we have that @xmath688 .
since @xmath689 is relatively compact in @xmath690 , there exists a sequence @xmath691 such that @xmath692 with @xmath693 .
thus , @xmath694 and we see by that @xmath695 . if @xmath696 , we get by the implicit function theorem , that for all @xmath697 , with @xmath71 small enough , there exists a solution @xmath674 .
therefore , we can solve @xmath698 for all @xmath699 , which is a contradiction .
hence , @xmath700 , which concludes the proof of the existence of a solution . + finally , note that for all @xmath642 the jacobian @xmath701 is invertible and the norm of its inverse is uniformly bounded , indeed @xmath702 in particular , we have that the determinant of the jacobian is never equal to @xmath87 , and we conclude by the holomorphic implicit function theorem that the solution @xmath703 to the equation depends holomorphically on @xmath104 and @xmath652 . in view of , it remains to study the integral @xmath704 with @xmath705 for @xmath706 .
we begin by performing a change of variables in the @xmath106-space .
+ + * change of variables : * for @xmath707 as in definition [ def : x ] , define the matrix @xmath708 and note that the gramian matrix @xmath277 ( cf . ) satisfies @xmath709 moreover , @xmath277 is invertible by virtue of proposition [ prop : graminvert ] , since @xmath710 .
next , we define the matrix @xmath711 by @xmath712 @xmath713 is invertible and thus satisfies that @xmath714 . define the matrix @xmath715 and notice that @xmath716 we see that @xmath717 .
next , we define the matrix @xmath718 @xmath719 is an isometry since @xmath720 .
thus , its columns form an orthonormal family in @xmath721 .
it follows from that the kernel of @xmath722 and of @xmath723 are equal , i.e. @xmath724 . the same holds true for the range of @xmath723 and of @xmath722 , i.e. @xmath725 . next , we choose an orthonormal basis , @xmath726 , of the space of random variables @xmath106 such that @xmath727 , the column vectors of the matrix @xmath719 , are among them . in particular ,
let @xmath728 for @xmath729 , and let @xmath730 be in the orthogonal complement of the space spanned by @xmath731 .
hence , we write for @xmath732 @xmath733 where @xmath734 .
moreover , note that @xmath735 next , we apply this change of variables to the vector @xmath276 given in and we get @xmath737 furthermore , one computes that @xmath738 and we get that @xmath739 next , to simplify our notation , we call the @xmath740 variables again @xmath106 . also , to abbreviate our notation , define @xmath741 and @xmath742 hence , @xmath748 as noted in remark [ rem : et_estim ] , @xmath749 and @xmath750 are smooth in @xmath104 , and @xmath750 is holomorphic in @xmath106 .
moreover , @xmath750 satisfies the estimates @xmath751 and @xmath749 satisfies the estimates @xmath752 with @xmath259 as in definition [ defn : zeroorderdensity ] .
finally , we perform the above described change of variables in the integral , and , using the fact that we chose an orthonormal basis of the @xmath106-space , we get that @xmath753 next , let @xmath754 and split the ball @xmath755 , @xmath756 , into two pieces : pick @xmath757 such that @xmath758 , and define @xmath759 .
then , we perform the splitting : @xmath760 with @xmath761 and @xmath762 * the integral @xmath763 * first , we perform a new change of variables in the @xmath106-space .
let @xmath764 such that @xmath765 we use the following notation : @xmath766 .
it is sufficient to check that we can express @xmath767 as a function of @xmath768 .
therefore , we apply lemma [ lem : implfunthm ] to the function @xmath769 where @xmath106 plays the role of @xmath770 in the lemma
. in particular , @xmath740 plays the role of @xmath315 .
let us check that the assumptions of lemma [ lem : implfunthm ] are satisfied : @xmath771 is by definition holomorphic in @xmath106 . using and
, we see that its jacobian , with respect to the variables @xmath740 , is given by @xmath772 the cauchy inequalities and ( [ eqn : tauest ] ) imply that @xmath773 this estimate is uniform in @xmath250 and @xmath774 .
expansion of the determinant yields that @xmath775 using that @xmath10 is self - adjoint , we see by corollary [ cor : lb_mn_a ] that for @xmath774 @xmath776 by the hypothesis , we have that @xmath777 .
hence , one gets that for all @xmath250 @xmath778 hence @xmath771 satisfies the assumptions of lemma [ lem : implfunthm ] . in the integral @xmath763 we restricted @xmath779 to the open ball @xmath780 .
it follows by lemma [ lem : implfunthm ] that for all @xmath781 with @xmath782 the equation @xmath783 has exactly one solution @xmath784 in the ball @xmath785 .
moreover , the solution satisfies @xmath786 , and it depends holomorphically on @xmath787 and @xmath779 and is smooth in @xmath104 . using , we see that the solution is implicitly given by @xmath788 with @xmath789 where @xmath750 satisfies the estimate ( [ eqn : tauest ] ) . since the support of @xmath398 is compact ( cf .
section [ sec : countzero2 ] ) , we can restrict our attention to @xmath787 and @xmath790 in a small poly - disc of radius @xmath791 centered at @xmath87 , with @xmath792 large enough such that @xmath793 . by choosing @xmath794 ,
@xmath97 large enough , we see that @xmath795 implies . from , and
, it follows that @xmath796 with @xmath797 where @xmath798 satisfies the estimate given in ( [ eqn : tauest ] ) .
furthermore , and imply that @xmath799 by performing this change of variables in the integral @xmath763 and by picking @xmath800 small enough as above , we get that @xmath763 is equal to @xmath801 where @xmath802 the integrand of @xmath763 depends continuously on @xmath787 .
hence , by performing the limit @xmath803 , we get @xmath804 with @xmath805 using , one computes that @xmath806 and , using , we get @xmath807 where @xmath808 . using , and one computes that @xmath809,\end{aligned}\ ] ] where the constant @xmath97 comes from the upper bound of @xmath810 given in . by the hypothesis , we conclude that @xmath811 which implies that @xmath812 for all @xmath779 with @xmath813 .
hence , @xmath814 next , we want to apply a multi - dimensional version of the mean value theorem for integrals to ( [ eqn : int1_2 ] ) .
indeed , let @xmath815 be open , relatively compact and path - connected , it then holds true that for a continuous function @xmath816 and a positive integrable function @xmath817 , there exists a @xmath818 such that @xmath819 hence , the mean value theorem applied to ( [ eqn : int1_2 ] ) yields that @xmath820 here , @xmath821 denotes the evaluation of the jacobian @xmath822 ( cf . ) at the intermediate point for @xmath779 given by mean value theorem .
note that @xmath821 depends smoothly on @xmath746 and @xmath747 because @xmath750 and @xmath10 do .
similarly , @xmath823 above denotes the evaluation of the function @xmath824 at the intermediate point for @xmath779 given by mean value theorem .
it depends smoothly on @xmath746 and @xmath747 because @xmath749 and @xmath750 do . moreover , using ( [ eqn : tauest ] ) , we see that it satisfies @xmath825 in remains to study the integral @xmath826 define the linear forms @xmath827_{11}\alpha_3 + [ \gamma^{\frac{1}{2}}]_{12}\alpha_4 , \quad l_2(\alpha ' ) = [ \gamma^{\frac{1}{2}}]_{21}\alpha_3 + [ \gamma^{\frac{1}{2}}]_{22}\alpha_4.\ ] ] using , we get that @xmath828 in the last equation we used , , and the fact that the hilbert - schmidt norm of @xmath829 is @xmath830 which follows from the fact that elements of the matrix @xmath829 are bounded by a term of order @xmath545 . by proposition [ prop : trace_est_gam ]
, one gets that the hilbert - schmidt norm of @xmath831 is bounded , indeed one has that @xmath832 since @xmath813 , one gets @xmath833 where the error estimate is uniform in @xmath779 .
here we used as well that by the hypothesis , we have that @xmath834 .
hence , @xmath835 extend the function @xmath836 to the whole of @xmath837 by a function that satisfies the same bounds , i.e. bounded by a term of order @xmath838 , and note that @xmath839 integration by parts yields that @xmath840 note that for any permutation @xmath841 , where @xmath283 is the symmetric group , we have that @xmath842 .
thus , in view of definition [ defn : permanent ] , we have that @xmath843 we conclude that @xmath844 where we used the fact that @xmath845 for @xmath846 , see proposition [ prop : mat_gamma ] , to obtain the last equality .
+ + * the integral @xmath847 * in this step we will estimate the second integral of equation ( [ eqn : i2 ] ) . therefore , we will increase the space of integration @xmath848 it is easy to see that lemma [ lem : implfunthm ] holds true for the set @xmath849 .
therefore , we can proceed as for the integral @xmath763 : perform the same change of variables and perform the limit of @xmath850 . as for @xmath763 ,
the integrand remains bounded by at most a finite power of @xmath545 which then yields that @xmath851 where the exponential decay comes from the fact that @xmath852 .
therefore , @xmath853 with @xmath854 acknowledgments .
i would like to thank very warmly my thesis advisor johannes sjstrand for reading the first draft of this work and for his kind and enthusiastic manner in supporting me along the way .
.5 cm a. melin and j. sjstrand , _ fourier integral operators with complex - valued phase functions _ , fourier integral operators and partial differential equations ( jacques chazarain , ed . ) , lecture notes in mathematics , vol .
459 , springer berlin / heidelberg , 1975 , pp .
120223 .
b. shiffman and s. zelditch , _ distribution of zeros of random and quantum chaotic sections of positive line bundles _ , communications in mathematical physics * 200 * ( 1999 ) , 661683 , 10.1007/s002200050544 . | we consider a non - selfadjoint @xmath0-differential model operator @xmath1 in the semiclassical limit ( @xmath2 ) subject to random perturbations with a small coupling constant @xmath3 .
assume that @xmath4 for constants @xmath5 suitably large .
let @xmath6 be the closure of the range of the principal symbol .
we study the @xmath7-point intensity measure of the random point process of eigenvalues of the randomly perturbed operator @xmath8 and prove an @xmath0-asymptotic formula for the average @xmath7-point density of eigenvalues . with this
we show that two eigenvalues of @xmath8 in the interior of @xmath6 exhibit close range repulsion and long range decoupling .
.5 cm rsum .
nous considrons un oprateur diffrentiel non - autoadjoint @xmath1 dans la limite semiclassique ( @xmath2 ) soumis de petites perturbations alatoires . de plus , nous imposons que la constant de couplage @xmath3 vrifie @xmath4 pour certaines constantes @xmath5 choisies assez grandes .
soit @xmath6 ladhrence de limage du symbole principal de @xmath1 .
dans cet article , nous donnons une formule @xmath0-asymptotique pour la @xmath7-points densit des valeurs propres en tudiant la mesure de comptage alatoire des valeurs propres lintrieur de @xmath6 . en tudiant cette densit ,
nous prouvons que deux valeurs propres sont rpulsives
distance courte et indpendantes long distance . |
the spectral energy distributions ( seds ) of active galactic nuclei ( agns ) carry important information on the physical processes of the accretion process
. many aspects of the agn phenomenon , including the sed , have been successfully interpreted within the accretion disk framework , specifically one in which the disk is assumed to be optically thick and physically thin ( blandford & rees 1992 , and references therein ) .
previous work , however , has concentrated nearly exclusively on high - luminosity agns mainly bright seyfert nuclei and qsos .
very little data exist on the spectral properties of low - luminosity agns , such as those commonly found in nearby galaxies ( ho , filippenko , & sargent 1997a ) , because they are difficult to study . yet ,
knowledge of the seds of agns in the low - luminosity regime is fundamental to understanding the physical nature of these objects and their relation to their more luminous counterparts .
the intrinsic weakness of low - luminosity nuclei poses practical challenges on obtaining the data . aside from the issue of sensitivity ,
often the main limitation stems from insufficient angular resolution necessary to separate the faint central source from the galaxy background . at virtually all wavelengths of interest
the core emission constitutes only a small fraction of the total light , and hence contamination from the host galaxy is severe . to date ,
only a handful of objects have been adequately studied with multiwavelength observations , and even for these the wavelength coverage is sometimes highly incomplete and the data only approximate ( ngc 1316 and ngc 3998 : fabbiano , fassnacht , & trinchieri 1994 ; sgr a@xmath1 : narayan , yi , & mahadevan 1995 ; m81 : ho , filippenko , & sargent 1996 ; ngc 4258 : lasota 1996 ; m87 : reynolds 1996 ; ngc 4594 : fabbiano & juda 1998 , nicholson 1998 ) .
nonetheless , these studies already suggest that the seds of low - luminosity agns look markedly different compared to the seds normally seen in luminous agns .
the spectral peculiarities seen in low - luminosity agns hint at possibly significant departures in these objects from the standard agn accretion disk model .
this paper presents seds for a small sample of low - luminosity agns for which reasonably secure black hole masses have been determined by dynamical measurements .
data are presented for a total of seven objects ( see table 1 for a summary ) , four for the first time ( ngc 4261 , ngc 4579 , ngc 6251 , and m84 ) ; the seds for the remaining three ( m81 , m87 , and ngc 4594 ) have been substantially updated and improved compared to previous publications .
the sample consists of four objects spectroscopically classified as low - ionization nuclear emission - line regions or liners ( heckman 1980 ; see ho 1999a for a review ) , one seyfert , and two objects that border on the definition of liners and seyferts .
it is worth remarking that , although the sample is admittedly small and heterogeneous , it contains _ every _ known low - luminosity object that has both a black hole mass determination and sufficient multiwavelength data to establish the sed .
the only two not included , namely sgr a@xmath1 and ngc 4258 , have already been extensively discussed in the literature .
because of the many complexities ( 2 ) involved in defining the nuclear seds of these weak sources , because so few data of this kind exist in the literature , and because a companion paper ( ho 1999b ) relies critically on the details presented here , i devote considerable attention in 2 to describing the data selection for each object .
section 3 summarizes the most noteworthy trends observed collectively in the sample , and section 4 discusses possible complications introduced by dust extinction ; readers not interested in the details of each object may wish to skip directly to sections 3 and 4 for an overview of the main results .
ho et al . ( 1999b ) will present detailed modeling of these data based on accretion - disk calculations .
as mentioned in 1 , measuring the weak signal from low - luminosity agns requires observations that minimize the contamination from the bright background of the host galaxy .
high angular resolution , therefore , is indispensable at all wavelengths .
this study takes the following strategy for data selection .
( 1 ) at radio wavelengths , only interferometric data will be used , preferably measured through vlbi techniques . the radio jet component ,
if present , potentially can contaminate the core emission even in sub - arcsecond resolution .
( 2 ) data in the infrared ( ir ) window are at the moment most poorly constrained .
there are no useful far - ir data because nearly all the existing measurements have been obtained using the _ infrared astronomical satellite _ , which has a beam 1 .
ground - based measurements in the mid - ir ( 1020 ) and near - ir ( 13 ) are widely available in the literature , but these data should be regarded strictly as upper limits because of the relatively large apertures employed ( @xmath2310 ) .
the energy distribution of normal stellar populations implies that the contamination from starlight in the near - ir should greatly exceed that in the mid - ir ; thus , the mid - ir points should be more representative of the direct emission from the nucleus , although emission from hot ( @xmath2100 k ) dust grains can substantially boost the luminosity in this band ( e.g. , willner 1985 ) .
( 3 ) all of the optical and ultraviolet ( uv ) data are derived from observations made with the _ hubble space telescope ( hst)_. photometry points generally pertain to an aperture of @xmath201 , while apertures 1 have been used for spectroscopic measurements .
( 4 ) finally , i quote only x - ray fluxes that originate from a nuclear source known to be compact at soft x - ray energies ( 0.52.5 kev ) .
the criterion for compactness , unless otherwise noted , is currently limited to the resolution of the high - resolution imager ( hri ) , approximately @xmath25 , on either the _ einstein _ or the _ rosat _ satellite .
several objects have hard x - ray ( 210 kev ) spectra acquired with _ asca _ , which has an angular resolution of @xmath25 ; hri images show that the soft x - ray emission is compact in all these cases .
the following subsections describe the data chosen for each object .
additional details can be found in tables 28 , and the individual seds are shown in figures 17 . the sed presented in ho ( 1996 ) has been updated ( table 2 ; fig .
1 ) with new optical and uv photometry points from _ hst _ , an additional high - frequency radio point , and a more up - to - date hard x - ray spectrum from _ asca_. the radio continuum between @xmath21 and 15 ghz shows an inverted spectrum ( @xmath3 0.3 to 0.6 , where @xmath4 ) .
it is difficult to assess the reality of the apparent slight turnover between 5 and 15 ghz because the two data points were not observed simultaneously ; the radio core of m81 is highly variable on many timescales ( ho 1999a ) .
the spectroscopically measured optical
uv slope of the featureless continuum is surprisingly steep . even after accounting for an estimated reddening of @xmath5 = 0.094 mag , the slope is still @xmath22 ( ho 1996 ) .
devereux , ford , & jacoby ( 1997 ) recently obtained an _ hst _
wfpc2 image of the nucleus of m81 using a filter centered near 1500 .
they obtained a flux @xmath24 times higher than that reported in the faint object spectrograph ( fos ) uv spectrum of ho ( 1996 ) .
the two new points based on optical images ( bower 1996 ; devereux 1997 ) , on the other hand , are consistent with the spectroscopic measurements .
maoz ( 1998 ) have reanalyzed the fos uv data and concluded that the nucleus may have been miscentered in the aperture during the observations .
this may account for the discrepancy between the fos and wfpc2 fluxes , although it can not be excluded that the nucleus varied in the uv between the two observing epochs . adopting the wfpc2 uv flux ,
the optical - uv slope is now @xmath3 1.31.4 .
the nucleus of m81 emits a nonthermal continuum in the hard x - ray band . the spectrum between 210 kev is well described by a single power law with a slope of @xmath6 = [email protected] ( ishisaki 1996 ; see also serlemitsos , ptak , & yaqoob 1996 ) , similar to that seen in luminous seyfert 1 nuclei ( turner & pounds 1989 ; nandra 1997 ) ; the luminosity in this case , however , is just 2 .
it is noteworthy that m81 emits substantially more energy in the x - rays relative to the uv than in luminous agns . the two - point spectral index between 2500 and 2 kev , @xmath8 , is 1.08 , smaller than the average value in quasars ( 1.4 ) or in luminous seyfert 1 nuclei ( 1.2 ) ( mushotzky & wandel 1989 ) .
figure 2 ( table 3 ) displays the sed of the nucleus .
the radio core is resolved even on a scale of several milliarcseconds .
no obvious point source can be seen in the 1.6 and 8.4 ghz maps of jones & wehrle ( 1997 ) , so i take as a reasonable approximation the peak intensities at these frequencies .
the resulting radio spectrum has @xmath6 = 0 .
the three optical points derived from wfpc2 images ( ferrarese , ford , & jaffe 1996 ) define a slope of @xmath3 2.2 . the fos spectrum of ferrarese et al . , taken with a similar - sized aperture ( 01 ) , suggests that most of the continuum is nonstellar .
the very steep optical continuum probably results mainly from reddening by dust internal to ngc 4261 and most likely associated with the nuclear disk .
patchy extinction can be seen throughout the disk and in the immediate vicinity of the nucleus ( ferrarese et al .
dust is also the most likely culprit for the complete absence of uv emission : the upper limit to the flux at 2300 is 10@xmath9 ( zirbel & baum 1998 ) , a factor of 80 lower than expected from a simple power - law extrapolation of the observed optical continuum .
estimating the intrinsic optical and uv luminosity of the nucleus , however , is difficult without prior knowledge of the absorbing column , the source geometry , and the extinction law ( see 4 ) .
the extinction due to the galaxy is small .
the foreground hydrogen column density is only @xmath10 = 1.6 2 , which , for the conversions @xmath11/(5.8 @xmath12 ) mag and @xmath13 = 3.1 ( bohlin , savage , & drake 1978 ) , translates to @xmath14 = 0.084 mag ( table 1 ) .
a rough estimate of the magnitude of the internal extinction can be obtained from the balmer decrement observed through a larger aperture ( 2@xmath154 ) by ho ( 1997b ) , h / h = 4.9.2 .
ferrarese et al .
could not remove the underlying stellar absorption lines from their spectra because of the limited signal - to - noise ratio of their data .
this effect preferentially biases the balmer line intensities , h more so than h , to low values . ] after removing the galactic contribution using the extinction law of cardelli , clayton , & mathis ( 1989 ) , h / h(internal ) = 4.8 ; assuming the galactic extinction law and a case b@xmath16 intrinsic h / h of 3.1 , which is thought to be appropriate for the conditions in the narrow - line regions of agns ( e.g. , gaskell & ferland 1984 ) , @xmath14(internal ) = 1.4 mag . this value is much higher than that inferred from the soft x - ray observations of worral & birkinshaw ( 1994 ) , who placed a limit of @xmath17 3.9 2 , or @xmath18 0.2 mag for a normal dust - to - gas ratio .
worral & birkinshaw ( 1994 ) observed ngc 4261 with the _ rosat _ position - sensitive proportional counter ( pspc ) . despite the coarse angular resolution of the pspc ( @xmath225 )
, they found that @xmath250% of the flux comes from a spatially unresolved component whose spectrum is well fitted by a power - law function .
the nonthermal component emits @xmath19(0.21.9 kev ) = 4.7 . inspection of an archival _ rosat _ hri image confirms that most of the soft x - ray emission indeed does stem from a compact component .
the uncertain impact of dust extinction on the optical - uv continuum renders estimates of @xmath8 highly unreliable . if @xmath14(internal ) = 0 mag ( correcting only for the galactic contribution ) , the optical slope is @xmath6 = 2.06 , which , when extrapolated to 2500 yields @xmath8 = 0.44 .
choosing @xmath14(internal ) = 1.4 mag and the galactic extinction law , the optical slope becomes @xmath6 = 0.67 , and @xmath8 = 0.84 .
these two extremes probably bracket the true value .
jones , terzian , & sramek ( 1981 ) mapped the nucleus with a resolution of 4 mas at 1.6 ghz .
the morphology of the radio core can be modeled by an elliptical gaussian with dimensions 1.4 mas @xmath15 6.0
mas , presumably because a parsec - scale jet still contributes to the emission on these scales .
the radio flux from the accretion flow , therefore , ought to be less than indicated in figure 3 ( table 4 ) . at optical wavelengths the nucleus appears unresolved in wfpc2 images .
bower ( 1997 ) measure @xmath20 = 19.9 mag and @xmath21 = 1.6 mag .
the stellar contribution to the pointlike nucleus is unclear , but judging from the complete absence of stellar absorption features between 6300 and 6800 in spectra acquired through a 02 slit ( bower 1998 ) , most of the light is probably nonstellar .
the ground - based spectrum in ho ( 1995 ) , for instance , shows a very noticeable ( equivalent width @xmath21 ) absorption line at = 6495 due to + .
the @xmath21 color of the nucleus , which corresponds to @xmath6 = 3.5 , again suggests substantial reddening by dust , as does the nondetection of uv emission ( zirbel & baum 1998 ) ; the images in bower ( 1997 ) , in fact , clearly show dust patches projected on the front of the nucleus .
these authors used the @xmath21 color map and an assumed intrinsic @xmath21 color for ellipticals to estimate a mean internal extinction of @xmath14 = 0.54 mag within a relatively large region of 145@xmath1584 .
close to the nucleus the extinction could be higher than this value .
the intrinsic balmer decrement measured through a 2@xmath154 aperture around the nucleus , for instance , is h / h = 4.7 ( ho 1997b ) , or @xmath14 = 1.3 mag .
the optical and uv points were dereddened using this @xmath14 .
the _ einstein _ hri image of m84 reveals a relatively complex central morphology with significant extended emission ( fabbiano , kim , & trinchieri 1992 ) ; the extended structure is also clearly evident in an archival _ rosat _ hri image . the 0.54 kev luminosity given by fabbiano ( 1992 ) ,
5.3 ( assuming a `` raymond - smith '' thermal spectrum with @xmath22 = 1 kev and a line - of - sight @xmath10 = 1.7 2 ) , should be taken only as an upper limit to the luminosity of a pointlike nucleus .
extrapolating the @xmath3 3.1 optical slope to the uv gives an estimate of @xmath23 0.75 .
reynolds ( 1996 ) presented a sparse sed for the nucleus .
figure 4 ( table 5 ) gives an updated version that includes additional data from the ir to uv region . as in the objects discusses so far , the radio spectrum is either flat or slightly inverted ; the vlbi points ( filled symbols ) have a mean spectral index @xmath24 . the apparent change in the spectral index from one point to another reflects the nonuniform resolution of the four experiments , which results in varying degrees of contamination from the jet , the nonsimultaneity of the observations , or both .
it is instructive to note that the jet in this case introduces very substantial contamination to the radio core emission even on a scale of @xmath21 .
the open symbols show the three vla points from biretta , stern , & harris ( 1991 ) ; clearly they lie systematically higher by about a factor of 10 compared to the vlbi flux densities .
the optical ( 6800 ) to uv ( 1200 ) spectrum , observed nearly simultaneously with the _ hst_/fos through the same ( 02 ) aperture , traces a smooth , continuous curve that can be well fitted with a double power law ( @xmath6 = 1.75 for 4500 , @xmath6 = 1.41 for 4500 ; tsvetanov 1998 ) .
the two @xmath25-band points derived from archival _ hst _ images agree very well with the fos spectrum ( within @xmath710% ) . on the other hand ,
the point taken from the faint object camera ( foc ) measurement of maoz ( 1995 ) is about 50% higher than predicted from the fos spectrum ; variability is a plausible explanation for this discrepancy .
the fluxes in the @xmath26 ( 1.25 ) and @xmath27 ( 2.2 ) bands ( stiavelli , peletier , & carollo 1997 ) , both acquired under sub - arcsecond seeing conditions , also appear to follow the extrapolated optical power law .
taken at face value , the 10 point deviates quite strongly from the power law at shorter wavelengths , but the relatively large aperture of the observations ( 6 ) allows room for substantial contamination . the soft x - ray flux recorded by the _ einstein _
hri is @xmath250% lower than that seen in the _ rosat _ hri [ @xmath19(1 kev ) = 5.4 , assuming a power - law spectrum with @xmath6 = 0.7 and a line - of - sight @xmath10 = 2.5 2 ; see reynolds 1996 ] .
this level of long term x - ray variability is not uncommon in low - luminosity agns ( serlemitsos 1996 ) .
spectrum of m87 is available , but the nucleus could not be clearly detected in the relatively short exposure ( reynolds 1996 ) . combining the uv spectrum with the average of the two hri fluxes , @xmath8 = 1.06 .
m87 belongs to the minority of liners ( @xmath225% ; maoz 1995 ; barth 1998 ) that shows prominent uv emission .
the optical - uv continuum of m87 evidently suffers little or no internal extinction .
the best - fitting double power - law model shown in figure 4 requires only @xmath14 = 0.12 mag for a milky way extinction law ( tsvetanov 1998 ) , very similar to the galactic contribution of @xmath14 = 0.078 mag .
note that the balmer decrement measured by ho ( 1997b ) in a 2@xmath154 aperture yields a substantially larger @xmath14 of 1.0 mag .
this example underscores the dangers of comparing observations taken with markedly different apertures .
the 2.3-ghz and 8.4-ghz interferometric observations of sadler ( 1995 ) , both made with a beam of 003 , define a rising radio spectrum with @xmath6 = 0.19 ( fig . 5 ; table 6 ) .
the _ hst_/fos uv spectrum of barth ( 1996 ) extends from @xmath23300 to 1150 ; excluding a broad feature near 3000 due to balmer continuum and emission , the spectrum can be described by a nearly featureless power - law function , @xmath28 .
there is tentative evidence that the nucleus varies in the uv .
the 2300 flux in the fos spectrum is a factor of 3 lower than that reported by maoz ( 1995 ) based on an foc image . in either case
, the uv output , again , is quite low with respect to the x - rays : @xmath8 = 0.78 and 1.02 for the fos and foc uv flux , respectively .
a compact , nonthermal component dominates the _ asca _ spectrum of ngc 4579 .
terashima ( 1998 ) find that the 210 kev continuum can be modeled as a moderately absorbed ( intrinsic @xmath10 = 4 2 ) power law with @xmath6 = 0.72 and luminosity 1.5 .
the value of @xmath10 derived from the x - ray spectrum roughly matches the intrinsic extinction estimated by barth ( 1996 ) based on the shape of the observed uv continuum .
fabbiano & juda ( 1997 ) presented an approximate sed for the nucleus of ngc 4594 .
the optical and uv points used in that study , however , were highly uncertain .
the @xmath29-band data , based on the pre - refurbishment foc image of crane ( 1993 ) , may have been underestimated because of the nonlinear behavior of the foc ( see crane & vernet 1997 ) . on the other hand ,
the uv flux taken with the very large aperture ( 10@xmath1520 ) of the _ international ultraviolet explorer _ is undoubtedly far too high . here
the sed of ngc 4594 is reassessed , paying close attention to aperture effects in the difficult optical - uv region ; a preliminary version of these data appears in nicholson ( 1998 ) .
the high - resolution data between 0.6 and 15 ghz define a compact flat - spectrum core with @xmath6 ranging from 0.2 to 0.4 ( fig . 6 ;
table 7 ) .
the nonstellar component of the optical and uv continuum , on the other hand , is somewhat difficult to specify .
the optical fos spectrum of kormendy ( 1997 ) , taken through a 021 aperture , has a very red continuum ( @xmath30 ) .
the steep optical slope is most likely caused by an increasing contamination of starlight toward longer wavelengths ( even within such a small aperture ) rather than by dust reddening of a purely featureless continuum .
the dilution of the depth of the stellar absorption lines indicated to kormendy et al .
that approximately 50% of the light at @xmath29 within their 021 aperture comes from the nonstellar continuum .
this roughly agrees with the strength of the point source extracted from an archival wfpc2 @xmath20-band image ( table 7 ) .
the point - source luminosity similarly measured from an @xmath25-band image , however , falls significantly below ( by a factor @xmath24 ) the extrapolated fos spectrum if one assumes the degree of stellar contamination to be constant between @xmath29 and @xmath25 .
this suggests that the nonstellar component contributes much less to the red end of the fos spectrum ( @xmath26800 ) , probably on the order of 25% or so .
the @xmath25-band flux is therefore adopted for the red end of the fos spectrum .
if one considers the two photometry points to be a reliable measure of the nonstellar component , the optical slope decreases to @xmath6 = 1.3 .
the uv spectrum , taken with an even larger aperture of 086 , likewise suffers from wavelength - dependent starlight contamination ( nicholson 1998 ) .
i excluded from the sed the portion of the spectrum longward of = 2200 , where incipient stellar absorption features and a sharply rising continuum suggest a sizable contribution from stars .
the final adopted optical and uv points ( shown as filled symbols ) are well joined by a power law with @xmath3 1.5 .
the _ asca _ 210 kev spectrum is predominantly nonthermal .
nicholson ( 1998 ) obtained a best fit with a power law with @xmath6 = 0.63 and @xmath19(210 kev ) = 1.1 .
the morphology of the central region as seen in the _ rosat _ hri image ( fabbiano & juda 1997 ) indicates that most of the hard x - ray emission is likely to originate from the nucleus .
the x - ray band energetically dominates over the uv band , with @xmath8 = 0.89 .
the average extinction inferred from the balmer decrement on arcsecond scales appears modest ( @xmath14 = 0.25 mag ; ho 1997b ) and roughly agrees with the intrinsic hydrogen column derived from the _ asca _ spectrum ( @xmath10 = 5.3 2 or @xmath14 = 0.28 mag ) .
good quality vlbi maps are available to isolate the core radio emission ( cohen & readhead 1979 ; jones 1986 ) , but the nonsimultaneous nature of the observations makes definition of the intrinsic radio spectrum ambiguous ( fig . 7 ; table 8) .
the two low - frequency points , for instance , yield a much steeper spectral index than the two high - frequency points ( @xmath6 = 0.4 compared to @xmath6 = 1.2 ) .
the only conclusion that can be drawn at this stage is that the radio spectrum , as in all the sources studied here , is consistent with that of a self - absorbed synchrotron source .
four _ hst _ photometry points have been measured for the pointlike nucleus , corresponding roughly to the @xmath31 , and @xmath25 bands ( crane & vernet 1997 ) .
taking into consideration the nonlinear effects of the foc that affected the @xmath32 and @xmath29 fluxes , the optical to near - uv slope is consistent with @xmath33 . combining the extrapolated uv flux with the soft x - ray ( pspc ) flux reported by worral & birkinshaw ( 1994 ) , @xmath8 = 0.83 .
birkinshaw & worral ( 1993 ) performed a detailed analysis of the pspc data and concluded that nearly all ( @xmath290% ) of the emission arises from a spatially unresolved , power - law component with a diameter 4 .
more detailed spectral information comes from turner et al.s ( 1997 ) statistical study of the _ asca _ spectra of a large sample of seyfert 2 nuclei , which included ngc 6251 .
the best - fitting model found by turner et al . requires a thermal ( raymond - smith ) plasma with a temperature of @xmath22 = 0.85 kev added to a power - law component characterized by @xmath6 = 1.11@xmath34 , @xmath19(210 kev ) = 1.3 , and an intrinsic @xmath10 = 1.4 2 .
the absorbing column obtained from the x - rays predicts a sizable internal extinction of @xmath14 = 0.75 mag . the balmer decrement given in shuder & osterbrock ( 1981 )
requires a much larger @xmath35 5 mag , although the reported h intensity appears to be rather uncertain .
luminous agns generally display a fairly `` universal '' sed ( see , e.g. , sanders 1989 and elvis 1994 ) .
the continuum from the ir to the x - rays , roughly flat in log @xmath36log @xmath37 space , can be represented by an underlying power law with @xmath3 1 superposed with several distinct components , the most prominent of which is a broad uv excess .
this so - called big blue bump is conventionally interpreted as thermal emission arising from an optically thick , geometrically thin accretion disk ( shields 1978 ; malkan & sargent 1982 ) . the largest spectral difference among agns manifests itself in their brightness in the radio band a factor of nearly @xmath38@xmath39 in radio power distinguishes `` radio - loud '' from `` radio - quiet '' objects .
the broad - band spectra of the seven low - luminosity agns presented here share a number of common traits , and yet they differ markedly from the seds of luminous agns .
to illustrate this point , figure 8 compares the seds of the present sample with the median sed of radio - loud and radio - quiet luminous agns taken from elvis ( 1994 ) ; all the curves have been normalized at 1 kev .
several features of the low - luminosity seds are noteworthy : \(1 ) the optical - uv slope is quite steep .
the power - law indices for the seven objects average @xmath40 ( range 1.03.1 ; see table 9 ) , or 1.5 if the possibly highly reddened objects ngc 4261 and m84 are excluded , whereas in luminous agns @xmath41 0.51.0 .
\(2 ) the uv band is exceptionally dim relative to the optical and x - ray bands .
there is no evidence for a big blue bump component in any of the objects .
indeed , the sed reaches a local minimum somewhere in the far - uv or extreme - uv region .
this leads to the above - mentioned steep optical - uv slope and to systematically low values of @xmath8 .
table 9 gives @xmath42 , to be compared with @xmath43 = 1.21.4 for luminous seyferts and qsos ( mushotzky & wandel 1989 ) . in other words , the low - luminosity agns in the present sample ,
most of which are liners , are systematically `` x - ray loud '' ( relative to the uv ) compared to agns of higher luminosity .
this modification of the sed from uv to x - ray energies leads to a harder ionizing spectrum , and it offers an explanation , at least in part , for the characteristically lower ionization state of the emission - line regions ( ho 1999b ) .
\(3 ) the hard x - ray ( 210 kev ) spectra , where available , are well fitted with a power - law function with @xmath3 0.60.8 , very similar to spectra observed in high - luminosity sources .
\(4 ) there is tentative evidence for a maximum in the sed at mid - ir or longer wavelengths . despite the relatively large apertures employed in the mid - ir observations , the 10- point should be largely uncontaminated by starlight ,
although dust emission could contribute significantly in this band .
\(5 ) the nuclei have radio spectra that are either flat or inverted .
the radio brightness temperature , where available , reach at least 10@xmath4410@xmath45 k. the radio cores , therefore , are self - absorbed synchrotron sources .
\(6 ) one usually gauges the degree of radio dominance in agns by the ratio of the specific luminosities in the radio to the optical band .
for instance , kellermann ( 1989 ) classify the radio strength of qsos by the parameter @xmath46 ; radio - quiet members have @xmath47 0.11 , and radio - loud members are distinguished by @xmath48 100 . adopting the same criterion , _ all _ of the objects , including the three spiral galaxies in the sample ( m81 , ngc 4579 , and ngc 4594 ) , qualify as being _ radio - loud_. m81 has the smallest radio - to - optical ratio , but @xmath48 is still @xmath250 .
this finding runs counter to the usual notion that only elliptical galaxies host radio - loud agns .
note that if the total ( host galaxy + nucleus ) @xmath29 luminosity had been used , which in these sources significantly exceeds the nuclear value alone , all the objects , with the possible exception of ngc 6251 , would have been considered radio quiet .
\(7 ) the sample of objects studied here is intrinsically extremely faint . to cast this statement in a more familiar context and to fully appreciate the enormous challenge in detecting these objects
, we note that agns that occupy the upper end of the luminosity distribution , namely qsos , typically have nonstellar continua with absolute magnitudes 30
@xmath49 @xmath50 @xmath49 23 .
classical seyfert nuclei , such as those from the markarian survey , are characterized by 23 @xmath49 @xmath50 @xmath49 18 .
by contrast , the nonstellar nuclear magnitudes listed in table 9 lie in the range 14.7 @xmath49 @xmath51 @xmath49 8.9 .
excluding the two extreme cases , @xmath52 = 11.5 mag .
the host galaxies themselves , on the other hand , are luminous @xmath53 systems ( @xmath54 21.1 mag ; table 1 ) , and hence the nuclei comprise merely @xmath20.01% of the total optical light of the host galaxies .
\(8 ) the bolometric luminosities of the sources ( table 9 ) , obtained by integrating the power - law segments shown on figure 8 , range from @xmath55 = 2 to 8 , or @xmath0 times the eddington rate for the black hole masses listed in table 1 .
the bolometric luminosities would be lower if the mid - ir peak has been overestimated , but this would not affect the conclusion that the eddington ratios are very low .
the x - ray band , arbitrarily defined here as the region from 0.5 to 10 kev , carries 6%33% of @xmath55 .
important aspects of the interpretation of the data depend on the intrinsic luminosity of the uv region , as this impacts conclusions concerning the optical - uv continuum slope , the presence of the big blue bump , and the strength and shape of the ionizing spectrum .
the uv bandpass , unfortunately , is strongly affected by dust extinction .
although the effects of extinction can , in principle , be corrected , in practice such a procedure is fraught with a number of uncertainties , which i briefly mention here .
first , it is unclear how to measure the amount of dust affecting the uv continuum source .
two measures of extinction are traditionally used .
one is based on comparison of the observed ratio of a pair of emission lines with their intrinsic ratio , after adopting a form for the extinction law .
the intrinsic spectrum of the hydrogen recombination lines is well known for conditions prevailing in the low - density narrow - line regions of agns , and the derived optical extinction does not depend sensitively on the exact form of the extinction law .
a second method uses the neutral hydrogen column density derived from x - ray spectra to calculate the absorbing column under the assumption that the dust - to - gas ratio and the grain properties are the same as those in the galaxy .
the latter assumptions , however , need not hold , especially in the vicinity of an agn where grains can be destroyed by the harsh radiation field ( voit 1991 ) .
low values of @xmath10 derived from x - ray observations also do not necessarily imply low extinction because some dust can be associated with the ionized medium . in either case ,
a major source of uncertainty is whether the uv continuum source traverses through the same absorbing medium as probed by the x - rays or by the optical emission lines .
the extinction obtained from the x - ray absorbing column in agn spectra , for instance , often greatly exceeds the extinction inferred from the balmer decrement ( e.g. , reichert 1985 ) .
one can attempt to estimate directly the extinction of the continuum by searching for spectral signatures imprinted by the assumed extinction law .
for a galactic extinction law , the most noticeable feature in the mid - uv is a broad depression centered near 2200 .
the uv spectra of agns , on the other hand , usually do not show this feature ( mckee & petrosian 1974 ; neugebauer 1980 ; malkan & oke 1983 ; tripp , bechtold , & green 1994 ) despite independent evidence for dust from other indicators ( e.g. , emission - line ratios ) .
this result implies one or more of the following possibilities : ( 1 ) the uv continuum and the line - emitting gas experience different amounts of extinction because the distribution of dust along the line - of - sight is patchy ; ( 2 ) the strong radiation field destroys dust grains close to the central continuum source ; and ( 3 ) the extinction law in extragalactic environments differs from that of the galaxy , specifically in having a much weaker 2200- bump .
support for the latter possibility comes from observations of starburst galaxies ( calzetti , kinney , & storchi - bergmann 1994 ) and of the small magellanic cloud ( smc ; bouchet 1985 ) whose uv spectra generally show a very weak , if any , 2200- bump .
gordon , calzetti , & witt ( 1997 ) argue that the absence of the 2200- feature in these systems is inherent in their extinction law and not merely a consequence of geometry effects .
the properties of dust grains , specifically the small graphite grains responsible for the 2200- bump ( mathis 1994 ) , evidently can be greatly affected by environmental conditions such as star - formation activity and/or metallicity . it is not difficult to imagine that the same may be the case in the vicinity of an agn , where the extinction law can be modified dramatically by the intense radiation field ( laor & draine 1993 ; czerny 1995 ) . is the apparent faintness of the uv band intrinsic to the seds or is it instead simply a consequence of dust extinction ?
luminous agns have an optical - uv slope of @xmath56 0.51.0 , whereas in the present sample @xmath56 1.52.0 .
let us estimate how much uv extinction is required to redden the `` typical '' agn spectrum to that seen here .
following maoz ( 1998 ) , i consider the extinction curve of the smc ( bouchet 1985 ) , as parameterized by pei ( 1992 ) , and the empirical starburst `` attenuation curve '' of calzetti ( 1994 ) , as parameterized by calzetti ( 1997 ) .
both curves lack a 2200- bump , but the calzetti et al .
curve is greyer . adopting a fiducial optical and uv wavelength of 6500 and 2500 , respectively
, reddening @xmath57 by a slope of 1 requires @xmath14 = 0.6 mag and @xmath58 = 1.5 mag for the smc curve .
the shallower calzetti et al .
curve gives @xmath14 = 1.1 mag and @xmath58 = 2.0 mag . for a more extreme case of @xmath59 = 2 , @xmath14 = 1.3 mag and @xmath58 = 3.0 mag for the smc curve , and @xmath14 = 2.3 mag and @xmath58 = 4.1 mag for the calzetti et al . curve .
these cases are illustrated in figure 9 .
given the limited accuracy of the available data and the unknown form of the extinction law in agns , it is difficult to rule out dust reddening as the principal cause of the observed weakness of the uv band , although one can probably exclude galactic - type dust grains from the absence of the predicted strong 2200- feature ( thin solid lines in fig .
in fact , the necessary amount of visual extinction , @xmath35 12 mag , is not unreasonable compared to those obtained from the balmer decrements ( table 1 ) . note that an average correction of @xmath58 = 2.0 mag will change @xmath8 from 0.9 to 1.2 , as seen in luminous seyfert 1s . nonetheless , in order to account for
the _ systematic _ differences between the seds of the two luminosity classes , one would have to postulate that low - luminosity agns _ systematically _ have either greater levels of extinction or a different dust extinction law compared to high - luminosity agns .
the first explanation is not supported by the data , at least insofar as balmer decrements can be used to gauge the continuum extinction ; the internal extinctions listed in table 1 do not appear anomalous compared to those found in more luminous seyfert galaxies ( e.g. , shuder & osterbrock 1981 ; dahari & de robertis 1988 ) .
the second hypothesis is ad hoc and difficult to test experimentally .
moreover , if exposure to the agn environment indeed does modify the extinction law , one would navely expect the effect to be greatest on the most powerful agns , exactly the opposite of what is observed .
thus , although dust extinction can in principle be responsible for the spectral peculiarities seen in low - luminosity objects , the alternative view that the nonstandard seds are intrinsic to the sources is also tenable and probably more favorable .
broad - band spectra are presented for seven low - luminosity agns .
although the sample is still limited and heterogeneous , this is the most extensive effort so far to systematically investigate the seds of these weak sources .
contamination by emission from the host galaxy can severely corrupt the faint signal from the nucleus , and careful attention has been paid to select only the highest quality nuclear fluxes in assembling the seds
. a comparative study of the gross features of the seds reveals that the seds of low - luminosity agns , as a group , look strikingly different compared to the standard energy spectrum of luminous seyfert galaxies and qsos .
most of the differences stem from the exceptional faintness of the uv continuum in the low - luminosity objects .
the so - called big blue bump is very weak or altogether absent , thereby making the continuum shape between optical and uv wavelengths steeper than normal and the x - ray band energetically more important .
extinction by dust grains with properties different from those of galactic composition may be responsible for suppressing the uv emission in low - luminosity agns , but an alternative , more intriguing possibility is that the absence of the big blue bump is a property intrinsic to these sources .
another notable property of the seds is the prominence of the compact , flat - spectrum radio component relative to the emission in other energy bands .
all seven nuclei in the sample , including the three hosted by spiral galaxies , can be considered `` radio - loud . ''
the agns investigated here indeed are very quiescent objects .
their bolometric luminosities range from 2 to 8 , which correspond to an eddington ratio of @xmath0 .
l. c. h. acknowledges financial support from a harvard - smithsonian center for astrophysics postdoctoral fellowship , from nasa grant nag 5 - 3556 , and from nasa grants go-06837.01 - 95a and ar-07527.02 - 96a from the space telescope science institute ( operated by aura , inc .
, under nasa contract nas5 - 26555 ) .
i thank chien peng and edward moran for assistance in analyzing some of the _ hst _ and _ rosat _ hri images , respectively , and i am grateful to marianne vestergaard for valuable comments on an earlier version of the paper .
i thank ramesh narayan for discussions on the theoretical interpretation of the data .
this work made extensive use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with nasa . | as a step toward elucidating the physical conditions in nearby active galaxies , this paper presents spectral energy distributions ( seds ) of a sample of seven low - luminosity active galactic nuclei ( agns ) .
seds for four objects are presented for the first time ( ngc 4261 , ngc 4579 , ngc 6251 , and m84 ) ; the data for the remaining three ( m81 , m87 , and ngc 4594 ) have been substantially updated compared to previous studies .
the nuclear fluxes were carefully selected so as to avoid contamination by emission from the host galaxy , which can be substantial for very weak nuclei . the present sample of low - luminosity nuclei all exhibit seds that look markedly different from the canonical broad - band continuum spectrum of luminous agns .
the most striking difference is that the low - luminosity objects lack an ultraviolet excess ( the `` big blue bump '' ) , a feature normally associated with emission from a standard optically thick , geometrically thin accretion disk .
the weakness of the ultraviolet band leads to an unusually steep optical - ultraviolet continuum shape and a more pronounced contribution from the x - rays to the ionizing spectrum .
it is argued that the absence of the big blue bump is a property intrinsic to the seds and not an artifact of strong dust extinction .
another notable property of the seds is the prominence of the compact , flat - spectrum radio component relative to the emission in other energy bands .
all seven nuclei in the sample , including three hosted by spiral galaxies , technically qualify as `` radio - loud '' objects according to conventional criteria .
finally , the integrated spectra confirm the exceptional weakness of the nuclei : the bolometric luminosities range from 2 to 8 , or @xmath0 times the eddington rate for the black hole masses previously reported for these galaxies . |
one of the strongest predictions from our present cosmological paradigm , cold dark matter ( cdm ) , refers to the universal density profile shared by dark matter haloes on all mass scales ( navarro , frenk & white 1996 , 1997 ; hereinafter nfw ) . these authors suggested a simple formula to describe the spherically averaged profile of dark matter haloes @xmath3
^ 2}. \label{eq : rhonfw}\ ] ] this `` cuspy '' profile diverges towards the centre of the halo as @xmath4 , although the enclosed mass is finite .
subsequent work has confirmed the absence of central - density cores in dark matter simulations ( moore et al .
1999b , ghigna et al . 2000 ; fukushige & makino 2001 ; gao et al . 2004 ; navarro 2004 ; diemand et al .
2005 ; springel et al .
2008a ) .
despite this consensus , there have been conflicting claims about the slope of the inner halo profile , @xmath5 .
for example , diemand et al . ( 2004 , 2005 ) claims a central slope of @xmath6 , whereas other authors argued in favour of steeper , @xmath7 , cusps ( moore et al .
1999b ; gighna et al .
2000 ) . recently ,
navarro et al . ( 2010 ) have found small , but significant deviations from an nfw profile in the inner - most regions of different milky way - size halo realizations ( the aquarius project ; springel et al .
this result may suggest that the slope of the inner halo profile may not strictly be `` universal '' ( see also gao et al .
2008 ) , which might potentially explain the discrepancy found in the literature about the value of @xmath8 .
the existence of cores or cusps in the centre of cdm haloes has important observational consequences for the discovery and interpretation of possible signals of dark matter annihilation in gamma - ray surveys ( stoehr et al .
2003 ; diemand et al .
2007 ; kuhlen et al . 2008 ; springel et al .
2008b ) , since the strength of the signal goes with the square of the dark matter density .
interestingly , the presence of cusps in dark matter haloes has been challenged by measurements of the rotation curves of low surface brightness galaxies ( lsbs ) .
observations of lsbs , which are supposed to be embedded in a dark matter - dominated potential , seem to favour constant - density cores , @xmath9 , for most observed systems ( e.g. flores & primack 1994 ; salucci & burkert 2000 ; swaters et al .
2003 ; simon et al . 2003 ; trachternach et al .
2008 ; oh et al .
2008 ; de block 2008 , 2010 ) .
dwarf spheroidal galaxies ( dsphs ) are another example of dark matter - dominated galaxies where significant constraints on @xmath8 can potentially be derived from the kinematics of their stellar constituents .
two types of analysis can be found in the literature : the first solves the jeans equations to estimate the mass enclosed within the stellar radius .
unfortunately , in pressure supported systems mass and orbital anisotropy are degenerate , making any deductions of the inner mass profile being cored or cuspy in general model - dependent ( koch et al .
2007a , b ; gilmore et al .
2007 , evans et al . 2009 ; walker et al . 2009 ) . a second analysis which aims at breaking the above degeneracy appeals to full multi - component distribution - function models , which fit the intrinsic and projected moments from hundreds of discrete radial velocities .
investigations using this technique seem to favour the presence of cores in the dark matter haloes of dsphs ( wilkinson et al . 2002 ; gilmore et al .
2007 ) . the apparent difficulties of cdm to reproduce observations on small scales using pure - dark matter simulations has not deterred a rich body of work from postulating alternative explanations within the present paradigm .
several authors have argued that baryons may play a fundamental role in sculpting the inner profiles of dark matter haloes , even in galaxies that at present are heavily dark matter - dominated .
for example , mass outflows driven by supernova explosions may under plausible assumptions remove dark matter cusps ( e.g. navarro , vincent & frenk 1996 ; read & gilmore 2005 ; governato et al . 2010 ) .
also , massive clusters that sink into the inner - most regions of dark matter haloes via dynamical friction will exchange momentum with central particles , ejecting them from the cusp and creating a core in the process ( goerdt et al .
2008 ) .
it is however debatable whether these processes are relevant on dsph scales , given that only a couple of mw dwarfs ( sagittarius and fornax ) hold a bound cluster population , and that the large majority of dwarfs with suitable kinematics show mass - to - light ratios @xmath10 .
clues to these questions may be gathered from examining the overall properties of the satellite galaxy _ population _ , rather than from individual systems .
this paper is aimed at studying the influence of dark matter cusps and cores on the evolution of satellite galaxies driven by tides , as well as exploring whether the satellite population around spirals may retain information on the inner structure of dark matter haloes .
the arrangement of the paper is as follows : [ sec : models ] presents the models assumed to describe the host and satellite galaxies , as well as the numerical set - up .
[ sec : tidev ] presents the results of our n - body experiments regarding the influence of cusps and cores on the dynamical evolution of dsphs .
appendix[sec : sa_code ] uses these results to construct a semi - analytic algorithm that simulates the formation of spiral galaxies through the accretion of individual subhaloes .
[ sec : sa ] shows the results of applying this code to milky way - like galaxies and analyzes the properties of the surviving satellite population for a range of inner halo slopes and disc masses .
we end with a brief summary in
[ sec : summary ] .
the virial radius is defined so that the mean over - density relative to the critical density is @xmath17 , @xmath18 where @xmath19 , and @xmath20 km / s / mpc is the present day value of hubble s constant .
the choice of @xmath17 varies in the literature , with some authors using a fixed value , such as nfw , who adopted @xmath21 , and others who choose a value motivated by the spherical collapse model , where ( for a flat universe ) @xmath22 ( lahav et al 1991 , eke et al 1996 ) .
the latter gives @xmath23 at @xmath24 in the concordance @xmath25cdm cosmogony , which adopts the following cosmological parameters : @xmath26 , @xmath27 , @xmath28 , consistent with constraints from cmb measurements and galaxy clustering ( see , e.g. , spergel et al 2007 and references therein ) . the density profile given by eq .
[ [ eq : rho ] ] may diverge as @xmath29 if the inner slope @xmath30 . in the literature such dark matter haloes are called `` cuspy '' , and a prototypical example is the nfw profile , which corresponds to @xmath31 . in contrast , `` cored '' haloes have a converging density profile at small radii , i.e. @xmath9 . also , density profiles with outer slopes @xmath32 lead to finite mass models , but for @xmath33 the cumulative mass profiles diverges as @xmath34 . to avoid that problem , kazantzidis et al .
( 2004 ) impose a truncation at @xmath35 that can be written as @xmath36 where @xmath37 is a small quantity , whose value we choose as @xmath38 . to obtain a continuous logarithmic slope @xmath39
is defined as @xmath40 our models are purely gravitational and therefore scale - free . when convenient for interpretive purposes , we shall scale our numerical units to physical parameters assuming that the host galaxy potential has @xmath41 , @xmath42 kpc and @xmath43 kpc , assuming @xmath44 .
these parameters are thus similar to those estimated by klypin et al .
( 2003 ) for the mw dark matter halo . to explore the effects that the presence of a baryonic disc in the host galaxy induces on the satellite galaxy population , we add a new force component @xmath45 in the form of a miyamoto - nagai ( 1975 ) disc model @xmath46 where @xmath47 is the disc mass , @xmath48 kpc and @xmath49 kpc are the radial and vertical scale lengths . , and same virial mass and concentration .
_ cuspy _ and _ cored _ dark matter haloes are those with @xmath30 and @xmath50 , respectively .
dotted lines correspond to initial n - body realizations .
symbols show the same realizations integrated for 14 gyr in isolation .
the good agreement between lines and dots implies that our numerical setup is adequate and free from numerical artifacts induced by numerical limitations down to scales comparable with the grid - size of the highest - resolution zone used by superbox , @xmath51 kpc.,width=317 ] our satellite galaxy models are constructed by generating 5-million particle realizations of the profile outlined in eq . [
[ eq : rho ] ] , using the code kindly made available by s. kazantzidis ( kazantzidis 2004 , 2006 ) .
these models are composed purely of dark matter , thus neglecting the contribution from baryons to the overall potential of satellite galaxies .
this approximation is valid for the large majority of satellite galaxies , i.e. dwarf galaxies , which typically show very high mass - to - light ratios ( aaronson 1983 , aaronson & olszewski 1987 ; mateo 1998 ; walker et al . 2007 ; simon & geha 2007 ; strigari et al . 2008 ; pearrubia et al .
2008a ) .
we have chosen parameters for the dwarf consistent with those expected for dwarf galaxy halos ( see , e.g. , pearrubia et al .
2008a ) . using the same physical scaling adopted for the host galaxy ( [ sec : prof ] )
we consider halo models with @xmath52 , @xmath53 kpc , and @xmath54 .
in these models the scale radius is independent of the inner slope of the profile , @xmath55 kpc .
we emphasize , however , that our models are scale free , so that the only relevant parameters in the calculations are the _ relative _ masses and sizes of the host and dwarf . in what follows , capital and small letters will be used to denote quantities corresponding to the `` host '' or the `` satellite '' models , respectively .
[ fig : prof ] shows the resulting density profiles of the satellite for four different values of the slope of the inner density profile , @xmath8 .
note that differences arise mostly within @xmath56 .
although in those regions the density of dark matter can differ by several orders of magnitude depending on @xmath8 , in fact the differences in enclosed mass are considerably smaller , @xmath57 and @xmath58 for @xmath59 and 1.5 , respectively . in the following sections we will see how these relatively small differences in central mass play a capital role in the overall evolution of satellite galaxies orbiting around spirals .
we follow the evolution of the dsph n - body model in the host potential using superbox , a highly efficient particle - mesh gravity code ( see fellhauer et al .
2000 for details ) .
superbox uses a combination of different spatial grids in order to enhance the numerical resolution of the calculation in the regions of interest . in our case
, superbox uses three nested grid zones centered on the highest - density particle cell of the dwarf .
this center is updated at every time step , so that all grids follow the satellite galaxy along its orbit .
each grid has @xmath60 cubic cells : ( i ) the inner grid has a spacing of @xmath61 and is meant to resolve the innermost region of the dwarf , where the effects of varying @xmath8 are strongest .
( ii ) the middle grid extends to cover the whole dwarf , with spacing @xmath62 .
( iii ) the outermost grid extends out to @xmath63 and is meant to follow particles that are stripped from the dwarf and that orbit within the main galaxy .
superbox uses a leap - frog scheme with a constant time - step to integrate the equations of motion for each particle .
we select the time - step according to the criterion of power et al ( 2003 ) ; applied to our dwarf galaxy models , this yields @xmath64 myr for the physical unit scaling adopted in sec .
[ sec : prof ] .
we have checked explicitly that our procedure for building the dsph model leads to a system that is in equilibrium and that does not evolve in isolation away from the prescribed configuration .
this is shown in fig .
[ fig : prof ] , where we plot the initial density profiles of four dark matter models with different inner slopes ( symbols ) .
dotted lines show the same profiles , but after evolving the system in isolation for 14 gyr with superbox .
the good agreement between lines and symbols shows that our numerical choices are appropriate , and that the evolution of the n - body model , on the scales of interest , is free from artifacts induced by the finite spatial and time resolution of the calculation .
estimated at different galactocentric distances for host galaxies without ( _ upper panel _ ) and with a baryonic disc component with a fiducial mass @xmath65 ( _ lower panel _ ) , respectively .
the `` mixed '' models refer to satellites with cored @xmath66 profiles orbiting about a host galaxy with a @xmath67 halo profile .
all radii are normalized by the scale radius @xmath68 of the satellite model .
arrows mark the position of the host scale radius , i.e. @xmath69 , for this particular satellite model.,width=317 ] satellite galaxies may lose some fraction of their initial mass to tides if their orbits bring them close to the host galaxy centre .
we may quantify the expected strength of the tides by computing the `` tidal radius '' ( @xmath70 ) of the satellite galaxy if this was placed on a circular orbit at pericentre @xmath71 here @xmath72 denotes mean enclosed density within a radius @xmath73 .
pearrubia et al .
( 2008b ) have shown that the total amount of mass , @xmath74 , provides an accurate estimate of the amount of mass that the satellite may expect to retain after completing several orbits in the host potential . in order to explore the resilience to tides of galaxies with different @xmath8-values we have calculated the mass enclosed within the tidal radius of the models shown in fig .
[ fig : prof ] as a function of pericentric distance using eq .
[ [ eq : rho ] ] and [ [ eq : rtid ] ] and imposing a universal density profile ( but different scaling ) for _ both _ host and satellite models . the results are shown in fig . [
fig : rtrp ] . in absence of a baryonic disc embedded in the host halo ( i.e. @xmath75 ) , cuspy and
cored satellite models show similar resilience to tides , although cored satellites are slightly more prone to tidal stripping than cuspy ones .
satellites with divergent profiles show a monotonic decline of @xmath76 as the pericentric radius decreases .
interestingly , this is not the case for satellites with a finite central density , which can retain a small bound mass fraction @xmath77const even if the orbits are fairly eccentric @xmath78 .
motivated by recent claims that star formation / feedback processes in the inner regions of dark matter haloes can turn cusps into cores ( e.g. governato et al .
2010 , and references therein ) we have also explored a `` mixed '' case wherein our satellite models are cored and orbit a host halo with a @xmath67 profile ( dotted - dashed lines ) . fig . [ fig : rtrp ] shows that in this configuration satellite galaxies would undergone tidal disruption if their orbits bring them at galactocentric distances @xmath79 . in models where the parent galaxy hosts a baryonic disc with a fiducial mass @xmath80 ( _ lower panel _ ) the influence of the disc component increases with decreasing @xmath8 at all pericentric radii
this is because at small orbital pericentres , @xmath81 , the mean density of spiral galaxies embedded in _
cored _ haloes scales as @xmath82 r_{\rm peri}^{-1}.\ ] ] thus , in models with @xmath9 , the disc component dominates the host tidal field if @xmath83^{-1 } m_d /\rho_0)$ ] .
complete tidal disruption is assured if the mean density of the host is higher than the central density of the satellite .
this occurs at @xmath84 r_s^3 y(c_{\rm vir})$ ] , where @xmath85 $ ] . using the model parameters adopted in
[ sec : models ] , disruption is expected for orbits with @xmath86 . in contrast , at small radii the mean density of spiral galaxies with cuspy , @xmath87 , haloes scales as @xmath88 r_{\rm peri}^{-1}.\ ] ] thus in these models the disc tidal field does _ not _ dominate at small radii . furthermore , in the limit @xmath89 one can readily show that eq . [ [ eq : rtid ] ] has an analytical solution , which can be written as @xmath90^{1/2}r_{\rm peri}$ ] , where @xmath91 .
since at small radii @xmath92 , this implies that @xmath93 if and only if @xmath94 , which suggests that only radial orbits may lead to full tidal disruption of satellites orbiting in such a galaxy potential .
a similar prediction was made by goerdt et al .
( 2007 ) using different analytical arguments . as we shall see below , the apparent `` indestructibility '' of cuspy subhaloes
is confirmed by our n - body simulations .
it is clear from inspecting this figure that baryonic discs may represent an important factor in the survival of satellite galaxies around spiral galaxies , and that the disc influence strongly varies depending on the slope of the inner density profile . to explore self - consistently the tidal evolution of satellite galaxies we place the n - body models outlined in
[ sec : satmod ] in orbit within the host galaxy assuming different combinations of apocentric / pericentric ratio .
all orbits have apocentres @xmath95 kpc . we consider a large range of pericentric radii @xmath96 and @xmath97 of @xmath98 .
note that the smallest pericentre is comparable to the scale radius of the satellite , and that this type of close - range encounter is rare in self - consistent cosmological simulations . in this contribution
, this series of orbits can be regarded as one of increasing tidal strength for eccentric orbits of similar orbital period . in all cases
we follow the evolution of the satellite for @xmath99 gyr in the scale units introduced in [ sec : prof ] , which corresponds to 35 orbital periods .
tides are strongest at pericentre and trigger mass loss episodes that recur with every pericentric passage .
the progressive stripping of satellites is shown in fig .
[ fig : mass ] , where dots represent the fraction of mass that remains bound to the satellite as a function of time .
pericentric passages are easily recognizable as the times when sudden drops in bound mass occur . between pericentric passages
the bound mass fraction remains nearly constant , suggesting , as indicated above , that tidal forces act in an impulsive regime and have little influence on the main body of the satellite at times other than pericentre . as expected from fig .
[ fig : rtrp ] , in the absence of a disc component in the host galaxy ( left panel ) the difference in mass loss that results from varying the slope of the inner density profile of our satellite models is in general small , and mostly arises at the latest stages of the orbital evolution .
an exception can be gleaned from the most - eccentric modes plotted in fig .
[ fig : mass ] ( @xmath100 ) , which show that after the first pericentric interaction the bound mass fraction of satellites with prominent cusps ( _ filled circles _ ) is a factor 23 lower than that of the cored models ( _ open squares _ ) .
however , this situation reverses as the number of pericentric passages increases , which means that , when cored haloes begin to disrupt , they do so more rapidly than cuspy haloes . the right panel of fig .
[ fig : mass ] shows that the presence of a disc component in the host introduces remarkable differences between the mass evolution of cored and cuspy satellites . whereas the rate at which satellites with prominent cusps , @xmath7 , lose mass
is barely altered , the _ full _ tidal disruption of models with a @xmath66 profile is assured if their orbits bring them into the vicinity of the disc .
indeed , dark matter haloes with @xmath101 show an extreme resilience to tides in all models explored here . as fig .
[ fig : mass ] shows , the first pericentric passage results in a significant decrease of the bound mass fraction .
however , the mass loss rate progressively slows down as the number of consecutive pericentric passages increases , and the bound mass tends to converge to a finite value .
this result implies that _ dwarf galaxies embedded in haloes with strong cusps , @xmath101 , can not be fully disrupted by the tidal field of the host galaxy _ , as expected from the analytical arguments given in
[ sec : tidstr ] , that would result in the total disruption of the remnant due to the positive binding energy of the nfw in that region .
however , in [ sec : dens_evol ] we show that tidal stripping does _ not _ truncate the density profile of dark matter haloes ( see also pearrubia et al . 2008b , 2009 ) ] .
this has interesting consequences for the properties of the satellite galaxy population , an issue to which we return below .
the tidal mass stripping of dark matter haloes proceeds gradually from the outside in ( e.g. stoehr et al .
2002 ; kazantzidis et al . 2004 ; kampakoglou & benson 2007 ; pearrubia et al .
this is illustrated in fig .
[ fig : xyz ] , where we plot , at consecutive apocentres of the orbit , the position of the satellite particles in the satellite frame .
this clearly shows how a halo is gradually stripped from the outside in , following an `` onion - peel '' process in which mass loss affects predominantly the outer , low - density regions of the satellite galaxy . by colour - coding the particles according to the value of the projected surface density at their location we can appreciate fundamental differences in the evolution of dark matter haloes depending on whether they are cuspy ( left panel ) or cored ( right panel ) , even if both satellite models have lost the same fraction of their original mass , 97% , after three pericentric passages .
the first interesting difference is that the spatial sizes of the cuspy satellite models suffer a dramatic decrease as a result of mass loss . yet
the bound remnants show densities that have been barely altered by the action of tides .
the response of cored satellite models to tidal mass loss is practically the opposite . in these systems
the change in size if much less pronounced , and the density drops more strongly than in the cuspy models at all radii , an issue that we explore in more detail in the following section .
[ fig : rhoevol ] shows the evolution of the density profile of cored ( @xmath66 , _ open circles _ ) and cuspy ( @xmath67 , _ filled circles _ ) satellite models for the most extreme orbit ( @xmath100 ) probed in our series .
the systems are shown at three consecutive orbital apocentres , when the satellites have had some time to find dynamical equilibrium after being perturbed by the host tidal field at pericentre .
the dotted / dashed lines are the same in each panel and show the initial density profiles for ease of reference .
a few interesting points can be gleaned from this figure .
one is that it is possible to identify two distinct regions on the basis of the shape of the density profile : an inner region where the profile varies smoothly with radius , and an outer region where an `` excess '' of material is present with respect to a naive extrapolation of the inner profile .
several authors ( e.g oh , lin & aarseth 1995 ; mayer et al .
2002 ; johnston et al . 2002 ) have shown that this type of `` breaks '' in the density profile are transient features that arise from the presence of unbound particles in the outskirts of the system as they escape from the satellite .
more recently , pearrubia et al . (
2009 ) have shown that the position of the `` break '' , @xmath102 , marks the radius at which the local crossing time exceeds the time elapsed since pericentre , @xmath103 , so that @xmath104 , where @xmath105 is the mean velocity dispersion of the satellite .
particles inside the break radius are thus within the region where enough time has elapsed for equilibrium to be reached .
a second interesting point of note is that the evolution of the density profile is remarkably sensitive to the shape of the inner profile .
tides acting on cuspy haloes tend to induce strong drops of density in the outer regions , as expected from fig .
[ fig : xyz ] , whilst keeping the shape of the inner profile barely altered , even after the satellite has been stripped of 97% of its initial mass .
the bottom - right panel shows that , for a similar mass loss , tides tend to lower the central density of cored haloes more severely than in the case of haloes with a centrally - concentrated profile .
the evolution of the satellite s circular velocity curve has a different amplitude depending on the slope of the inner halo profile .
this is shown in fig .
[ fig : rhoevol ] , in which we mark the position where the circular velocity reaches its peak,@xmath106 , at consecutive orbital apocentres .
as this figures shows , mass loss results in a progressive shift of @xmath106 towards the satellite centre , a process that is more pronounced in the case of cuspy satellite models .
we inspect this issue in some detail in the following section . . for comparison
we also plot the initial density profiles in each panel , where dotted and dashed lines denote haloes with @xmath67 and 0.0 , respectively .
the quantity @xmath107 refers to the bound mass fraction .
arrows indicate the position of @xmath106 , i.e. the radius where the circular velocity curve peaks .
note that the density profile in the inner regions of dark matter haloes is preserved even when large fractions of the initial mass are lost to tides.,width=317 ] as shown by hayashi et al .
( 2003 ) and pearrubia et al .
( 2008b ) , the evolution of the structural parameters of nfw haloes as they are tidally stripped does solely depend on the _ total amount of mass lost _ , and not on the details of _ how _ the mass is stripped .
[ fig : vmrm ] shows that this result is generic to dark matter haloes that follow the density profile outlined by eq .
[ [ eq : rho ] ] .
the left and right panels of the figure show respectively the fractional change of the peak velocity , @xmath108 , and its location , @xmath106 , as a function the satellite bound mass fraction @xmath109 for three of the four orbits surveyed in our study .
different symbols correspond to different orbits and are shown at consecutive orbital apocentres when the main body of the satellite galaxy is close to dynamical equilibrium .
open and filled symbols denote models in which a disc component has been embedded in the parent galaxy .
this figure illustrates a few interesting differences in the evolution of satellite galaxies depending on how distended the halo profile is in the inner - most regions .
the first is that , at a given mass loss fraction , the peak velocity declines more significantly as we decrease the value of @xmath8 .
the second difference is in the evolution of the satellite size , which is clearly more substantial for satellite galaxies with cusps , as expected from fig .
[ fig : xyz ] and [ fig : rhoevol ] .
the right panel of fig .
[ fig : vmrm ] shows that @xmath106 is severely shortened by the action of tides if the slope of the inner halo profile is @xmath110 .
interestingly , the location of the peak velocity of cored models can not drop further than a factor @xmath111 , regardless of the amount of mass lost to tides .
this is in marked contrast with cuspy haloes , which may shrink into oblivion if tidal stripping carries on indefinitely .
irrespective of pericentre , number of completed orbits , host potential , or slope of the inner density profile , the evolution of the structural parameters depends solely on how much the satellite mass has varied .
this is a remarkable result , especially because our series of models includes extreme cases in which the satellite models lose more than 99.99% of their original mass .
following pearrubia et al .
( 2008b ) , we use a simple empirical formula to fit the evolution of the structural parameters @xmath112 where @xmath113 and @xmath114 represents either @xmath108 or @xmath106 .
the best - fit values of @xmath115 and @xmath116 have been plotted in each of the panels of fig .
[ fig : vmrm ] , and the results are shown with dashed lines .
the fact that the evolution of satellite galaxies as they are stripped by tides follows well - defined _ tracks _ will be used in the following sections to build a simple semi - analytical algorithm that follows the disruption of dark matter haloes with different inner profiles with reasonable accuracy ( see appendix a ) . and its location , measured at the apocentres of the various orbits in our simulation series .
open and closed symbols denote simulations with and without a baryonic disc component ( @xmath65 ) , respectively .
the dashed lines are our fits to the evolutionary tracks using eq .
[ [ eq : g ] ] . ,
the results presented in [ sec : tidev ] show that the response of galaxies to tides strongly depends on the slope of the inner density profile of dark matter haloes , @xmath8 , as well as to the presence of a baryonic disc embedded in the host galaxy .
this raises the question of whether the satellite populations surrounding spiral galaxies may provide insights into the inner structure of dark matter haloes . in order to explore this issue , simulations that follow the hierarchical formation of spiral galaxies through the accretion of individual satellites
are required .
we begin by outlining a method to construct merger trees of spiral galaxies , and then discuss the effects of varying @xmath8 and the disc - to - parent halo mass ratio on the present population of satellite galaxies . at @xmath24
is shown in the upper panels . here
we plot the virial mass and concentration in the upper - left and -right panels , respectively .
the disc component embedded in the host galaxy also grows with time . for this particular plot
we show the evolution of a disc with @xmath65 and @xmath48 kpc at @xmath24 .
the disc mass and size are plotted in the lower - left and -right panels , respectively .
, width=317 ] we construct merger trees using monte carlo methods .
specifically we employ the merger tree algorithm described by parkinson et al .
( 2008 ) which is itself an empirical modification of that described by cole et al .
we adopt the parameters @xmath117 that are found to provide the best fit to the statistics of halo progenitor masses measured from the millennium simulation by cole et al .
we use a mass resolution ( i.e. the lowest - mass halo which we trace in our trees ) of @xmath118 , which is sufficient to achieve resolved galaxy properties for all of the calculations considered in this work . within this evolving hierarchy of dark matter
, the evolution of the host galaxy is followed using the galform semi - analytic model of galaxy formation ( cole et al .
in particular , we use the implementation of bower et al .
( 2006 ) which includes feedback from active galactic nuclei and which produces a good match to the evolving stellar mass function of galaxies .
for the purposes of this work the key results extracted from the model are the evolving properties ( mass and size ) of the disc forming at the centre of the dark matter halo .
these are computed in galform by determining , at each time step , the rate at which hot diffuse gas in the halo is able to cool and sink towards the halo centre , where it becomes rotationally supported , forming a disc .
simple prescriptions for star formation and feedback from supernovae then allow the gaseous and stellar mass of the galaxy to be determined .
knowledge of the angular momentum of the disc , assumed to have been conserved throughout the process of cooling , collapse and disc formation , allows for an estimate of the size to be found , following the methods of mo , mao & white ( 1998 ) .
we run ten different realizations of a mw - like galaxy with a mass of @xmath119 at @xmath24 .
although all models end up with the same mass and virial concentration , the assembly of these haloes is remarkably distinct , a direct manifestation of _ cosmic variance_. a useful distinction between different halo models can be made by deriving the time at which a significant mass fraction has been accreted .
for example , fig .
[ fig : hostevol ] includes examples of _ late - forming _ galaxies , which do not accrete 10% of their present mass until redshit @xmath120 ( e.g solid lines ) , as well as _
early - forming _ galaxies , which have this mass fraction in place by @xmath121 ( e.g. long - dashed and dotted - dashed lines ) .
the disc mass is set by choosing a value for the present disc - to - halo mass ratio @xmath122 .
since the baryonic content of galaxies can not exceed the universal baryon fraction , @xmath123 ( spergel et al . 2007 ) , we consider disc masses within the range @xmath124 . at any given redshift , the disc mass is re - scaled as @xmath125 . in general the disc component forms at @xmath126 ( e.g. _ solid lines _ ) , although in some extreme cases the formation process can start as early as @xmath121 , ( e.g. _ long - dashed lines _ ) .
the disc mass suffers strong fluctuations with time , as shown in the bottom - left panel of fig .
[ fig : hostevol ] .
the formation time of discs plays a key role in the mass evolution of satellites , as discussed in the following sections . as a function of redshift ( _ upper panel _ ) extracted from the five merger - tree realizations shown in fig .
[ fig : hostevol ] . distance from the host galaxy centre at the time of accretion , @xmath127 , as a function of redshift ( _ lower panel _ ) .
note that if orbital decay did not take place , @xmath128 throughout the orbital evolution of satellite galaxies .
, width=317 ] our merger tree realizations follow the evolution only of subhaloes with masses above @xmath129 when they first cross the host virial radius .
this choice is motivated by the fact that primordial gas is unable to cool by lyman-@xmath13 emission and initiate star formation in haloes with a temperature below @xmath130 k , which approximately corresponds to the above mass threshold .
it is typically assumed that lighter subhaloes remain _ dark _
( e.g. haiman et al . 2000 ) , and thus undetectable through photometric surveys .
unfortunately , determining the number of stars that a given subhalo can form from primordial gas is still a topic of hot discussion ( see tassis et al . 2008 ; ricotti 2009 and references therein ) .
our calculations do not include a baryonic component in the accreted subhaloes and may therefore be regarded as _ pure dark matter models_. despite the uncertainties in the properties of the stellar components embedded in our satellite models , the following sections will show that a number of general inferences may be made in reference to the structure of satellite galaxies .
the cumulative number of satellite galaxies accreted at a given redshift , @xmath131 , is shown in the upper panel fig .
[ fig : num ] . since the host halo grows by accreting dark matter subhaloes , the number of accreted satellites and the virial mass are closely correlated at any given epoch
. an interesting point of note is that a host galaxy with @xmath132 is expected to accrete around @xmath133 satellites that may potentially host a stellar component .
this is in remarkable contrast with the number of known satellite galaxies in the mw , which to date adds up to a mere couple of dozens ( e.g. belokurov et al .
2007 ; koposov et al . 2009 ) .
the apparent mismatch between theoretical predictions and observations is typically referred as the `` missing satellite problem '' ( moore et al .
1999a ; klypin et al .
1999 ) .
when a subhalo first crosses the host virial radius , it becomes a _ satellite galaxy_. the inner satellite structure is fixed by applying the virial mass - concentration relation found by nfw .
its orbital parameters are drawn from the distribution of benson ( 2005 ) , which was measured from n - body simulations .
this distribution gives the radial and tangential velocity components of the orbit . by definition ,
the distance from the host galaxy centre at accretion time , @xmath127 , equals the host virial radius , and closely correlates with orbital apocentre ( ludlow et al .
thus , a natural result of the inside - out growth of galaxies is a strong correlation between the accretion time of satellites and their location within the host .
indeed , satellites accreted at early times can only be found in the inner - most regions of their parent galaxies .
this is illustrated in the lower panel fig .
[ fig : num ] , where we show the infall distance as a function of redshift for the merger - tree realizations shown in fig .
[ fig : hostevol ] .
two distinct epochs can be distinguished on the basis of the effects introduced by cosmic variance : a late epoch , i.e. during the last 8 gyr ( @xmath134 ) , when the orbits of accreting satellites have a similar distribution of orbital parameters , and an early epoch ( @xmath135 ) , when the accretion history of the host galaxy is dominated by cosmic variance .
this result suggests that the properties of the surviving satellite galaxy population at @xmath24 will show some scatter between different merger trees , and that that scatter will be more evident for satellites that were accreted at early , @xmath135 , epochs .
the destructive action of tides is illustrated in fig .
[ fig : radmass ] , where we show the present mass and galactocentric distance of cuspy ( _ left panels _ ) and cored ( _ right panels _ ) satellite models extracted from the _ ten _ host galaxy realizations outlined in [ sec : mtree ] . for ease of reference
, we mark with dashed lines the minimum mass that satellites must carry as they are accreted into the host in order to be included in our merger trees .
thus in this figure all satellites with @xmath136 have lost some fraction of their original mass to tides .
[ fig : radmass ] illustrates a number of interesting points .
the most obvious is that satellites accreted at early times tend to lose a large fraction of their original mass to tides , which is highlighted in this plot by marking the masses and positions of galaxies that fell into the host galaxy at @xmath137 ( solid dots ) .
this redshift roughly corresponds to the epoch at which the universe was completely re - ionized ( e.g. fan et al .
2002 ) , and thus we refer to these galaxies as _ re - ionization fossils_. it is typically believed that such systems went through a short phase of star formation before they crossed the virial radius of the parent galaxy and lost their primordial gaseous component through ram pressure stripping , a process captured by the n - body simulations of mayer et al .
the stellar compositions of these systems are thus expected to be distinct from those of the satellite galaxies that continued forming stars after re - ionization and were accreted at a later time . among the known mw satellites
belokurov et al .
( 2009 ) identifies four possible candidates for being fossils of re - ionization : segue i and ii , bootes ii and coma .
future spectroscopic surveys of mw dsphs will possibly extend this census .
because the initial orbital apocentres of our satellite models roughly correspond to the virial radius at accretion time , which scales as @xmath138 , re - ionization fossils are expected to populate the inner - most regions of the host , @xmath139 kpc ( see also gao et al .
2009 ) . whether or not re - ionization fossils survive tidal disruption to the present day
strongly depends on the inner profile of dark matter haloes , as well as on the mass of the host disc .
[ fig : radmass ] shows that the disc component strongly depletes the inner regions of the host galaxy from satellites . due to their small apocentres ,
re - ionization fossils count among the satellites most prone to undergo full tidal disruption in @xmath66 models . as a result ,
the number of surviving re - ionization fossils is close to negligible if the inner profile of dark matter haloes is cored .
although cuspy satellite models also shed large fractions of their mass to tides , a bound remnant always survives the shredding process ( see fig .
[ fig : mass ] ) . in these models
the main effect of the disc component is to lower the present mass ( and luminosity ) of these systems .
a second point of interest is that the satellite galaxies that have have been acted on by tides extend all the way out to the host virial radius .
this is because in cdm most satellites move on highly eccentric orbits ( e.g. benson 2005 ) .
thus , a significant number of galaxies with relatively large apocentres are likely to reach the inner - most regions of the host and experience severe stripping during a hubble time , as shown in fig .
[ fig : radmass ] . in fig .
[ fig : mrprof ] we show the mean cumulative number of satellites per realization ( _ upper panels _ ) and bound mass ( _ lower panels _ ) as a function of galactocentric distance , as derived from the above models .
thin and thick lines distinguish between fossils of re - ionization and the whole satellite sample , respectively .
as anticipated above , the cumulative number of cuspy satellites barely depends on disc mass , again illustrating the fact that cusps can not be completely destroyed by tides .
cored satellites , in contrast , can be fully disrupted by tides , which translates into a progressive decrease of the total number of satellites as more massive discs are assumed .
this is particularly obvious for the satellites accreted prior to re - ionization ( i.e. @xmath140 ) . adopting a fiducial disc mass @xmath65 and a cored halo profile , we expect of the order 23 fossils of re - ionization in a mw - like galaxy
this number nearly quadruples in halo models with @xmath67 and is barely dependent on disc mass .
a result that may have important observational consequences for the detection of dark matter annihilation signals in gamma - ray surveys refers to the large abundance of satellite galaxies at small galactocentric radii .
cuspy models predict the presence of the order of @xmath141 satellites within 10 kpc from the host centre with masses that range between @xmath142@xmath143 , a number that barely depends on the disc mass .
this number drops by a factor @xmath144 in models that assume a cored dark matter profile and a fiducial disc mass of @xmath65 .
since the gamma - ray flux from a particular source is inversely proportional to the distance squared , these galaxies may provide interesting targets for dark matter annihilation surveys , an issue that we will explore in a separate contribution .
the lower panels of fig .
[ fig : mrprof ] show that the variation of the mean satellite _ bound _ mass as a function of galactocentric distance could potentially constrain the inner profile of satellite galaxies . in particular ,
our models show that the satellite mass is expected to drop at distances @xmath145 kpc if haloes have a cuspy profile .
this is in a marked contrast with cored models , where the effect is the opposite . here
the negative mass gradient arises simply from the depletion of satellites with small orbital apocentres .
satellites that at @xmath24 are located at distances comparable to the disc size happen to be observed by chance near pericentre .
an interesting point concerns the extreme stripping suffered by a large number of satellites in these models .
pearrubia et al .
( 2008a , b ) estimate that satellite galaxies must lose @xmath14690% of the dark matter envelope before stars begin to be shed to tides . in our models
this implies that all satellites in fig .
[ fig : radmass ] with masses @xmath147 have lost a fraction of their stellar components to tides and have associated stellar streams .
the fact that tidal stripping is very sensitive to the inner profile of dark matter haloes suggests that surveys of stellar streams in spiral galaxies may put strong constraints on the inner profile of dark matter haloes .
unfortunately , the absence of a stellar component in our satellite models prevents us from exploring this issue in depth . .
we measure these quantities for two satellite samples : those that were accreted at @xmath137 , to which we refer as _ fossils of re - ionization _ ( thin lines ) , and the whole sample , which includes all satellites with masses above @xmath143 at accretion ( thick lines ) .
solid , dotted and dashed lines denote host galaxies with @xmath148 and 0.2 , respectively .
note the present mass and spatial distribution of satellite galaxies strongly depends on the slope of the inner halo profile.,width=317 ] the results shown in previous sections suggest that by simply counting the number of satellites around spiral galaxies one could make strong inferences on the inner structure of cdm haloes .
however , this requires theoretical models to be able to predict properties of the _ stellar populations _ hosted by dark matter subhaloes and to make comparisons that involve observable quantities , such as galaxy luminosities and surface brightnesses . in practice , this is difficult because , in spite of strong theoretical efforts ( e.g. macci et al .
2009 ; cooper et al .
2009 ) , the processes that shape the baryonic properties of satellite galaxies during their formation and later accretion are still rather uncertain .
a more promising way forward might arguably be to compare the masses of satellites derived from the kinematics of the stellar tracers against theoretical predictions . in mw - size galaxies ,
the vast majority of satellite galaxies correspond to dwarf spheroidals ( dsphs ) .
although dsph mass estimates are often dominated by large uncertainties ( e.g. kravtsov 2010 and references therein ) , recent contributions have shown that the satellite mass enclosed within both the projected and tri - dimensional half - light radii of these systems is well constrained and robust to a wide range of halo models and velocity anisotropies ( walker et al .
2009 , wolf et al .
these data exist for most of the known mw satellites and a handful of m31 ones , which altogether encompass half - light radii that span nearly two orders of magnitude in size . in fig .
[ fig : mrh ] we show the mass estimates for mw ( walker et al .
2009 ) and m31 ( collins et al . 2009 ; kalirai et al .
2009 ) dwarfs with closed and open symbols , respectively .
lines are derived from the sample of models shown in fig .
[ fig : radmass ] by simply measuring the mass of individual satellites enclosed within the range of radii spanned by the observational data , @xmath149 .
error bars attached to the curves denote deviations around the mean , whereas those associated with the symbols represent observational uncertainties .
some of the surviving satellite galaxies in our models have smaller sizes than the largest half - light radii in the sample . to remove this bias , only haloes with @xmath150 are considered in the calculation of @xmath151 .
this figure shows two remarkable results .
the first one is that cuspy satellite models ( @xmath67 ) provide a better description of the masses of local group dwarfs than cuspy ( @xmath66 ) ones .
this is particularly obvious at small radii , i.e. the region populated by ultra - faint dsphs , where the difference between both halo models are largest ( see fig .
[ fig : prof ] ) .
although not shown here , we have also explicitely confirmed that this mismatch do not arise from biases between the satellite masses and their galactocentric distance or their time of accretion .
it is also worth noticing that the deviations around the mean are relatively small taking into accout the large range of bound masses encompassed by our satellite models , and provide a good representation of the scatter shown by the observational data in models where the disc component represents a small mass fraction of the host halo .
this result naturally accounts for the appearance of a `` universal mass profile '' for dwarf galaxies ( walker et al .
2009 ) .
the second interesting point concerns the impact of the disc component on the masses of satellite galaxies .
clearly , the mean enclosed mass at a given radius tends to decrease as more massive disc models are adopted . at a given radius ,
the mean satellite mass can drop as much as a factor @xmath152 for the most massive disc components considered here , @xmath153 .
interestingly , the systematically low masses shown by the handful of m31 satellites with available kinematics may be the telltale evidence of a relatively heavy disc component . in this scenario ,
the mw and m31 may be embedded in similar - size haloes , but the disc - to - halo mass ratio of m31 would be a factor @xmath154 higher than that of the mw .
interestingly , by converting k - band absolute magnitudes into stellar masses hammer et al .
( 2007 ) have recently estimated that the disc of m31 may approximately be a factor @xmath155 more massive than that of the mw , although these measurement still have large uncertainties .
this result , if confirmed , would lend support to the scenario that explains the relatively low mass of m31 satellites as tidal in origin .
further clues to the mass of m31 disc may be gathered from the faint - end of the dsph population .
our modeling indicates that the mass difference between m31 and mw dsph populations should also be present in dwarfs with half - light radii smaller than those shown here .
this cdm - motivated prediction may be verified with present observational capabilities . .
lines denote average values and error bars deviations around the mean . filled and open symbols show , respectively , the masses of the mw and m31 dwarf galaxies with known velocity dispersion estimated at the half - light radius , @xmath156 , using walker et al .
( 2009 ) formula . to allow a meaningful comparison against observational constraints ,
only satellite models with @xmath150 are considered in the calculation ( see text ) .
, width=317 ] in previous sections a mass threshold at @xmath157 was imposed to account for the fact that only satellites with masses @xmath158 can form stars ( e.g. heinman et al .
there is , however , some controversy about the value of @xmath159 that defines the minimum galaxy mass ( see e.g. tassis et al . 2008 and ricotti 2009 for disparate results that place @xmath159 above and below the adopted @xmath143 , respectively ) .
because the satellite mass function diverges as @xmath160 , imposing a mass cut onto the satellite population implies that satellites with @xmath161 dominate in number . as a result , increasing ( decreasing ) the value of the mass threshold shifts upward ( downward ) the curves shown in fig .
[ fig : mrh ] at all radii .
this suggests that the masses of local group dwarf galaxies can effectively be used to place constraints on @xmath159 .
[ fig : mthr ] shows a crude attempt to estimate the minimum satellite mass above which star formation is expected to occur . to estimate the accuracy of our models in describing the masses of the mw dwarf galaxies shown in fig .
[ fig : mrh ] , we apply a @xmath162 fit , where @xmath163 and @xmath164 .
the observational variance is @xmath165 , where @xmath166 denotes dwarf masses estimated at the half - light radius with uncertainties @xmath167 .
the number of mw dwarfs in walker et al .
( 2009 ) catalogue is @xmath168 , where @xmath169 is the number of satellite galaxies _ per realization _ with masses above @xmath159 .
this number strongly depends on the mass threshold , and for a mw - like halo it ranges between @xmath170 and @xmath171 for @xmath172 and @xmath173 , respectively .
interestingly , the masses of mw dsphs strongly constrain the mass threshold for star formation in dark matter subhaloes .
we find that only those subhalos that were accreted with masses above @xmath1 and @xmath173 have average masses compatible with those in the sample of mw dwarfs with suitable kinematic data .
subhaloes that obey this condition may be identified with `` luminous '' satellites while the rest remain dark , in agreement with a broad number of theoretical expectations ( bullock et al .
2009 ; kravtsov 2010 and references therein ) . comparing both panels we find that , independently of the host disc mass @xmath47 , satellite models with a cuspy inner profile provide a better match the mw dwarf galaxy masses than cored models , as expected from fig .
[ fig : mrh ] . however , there are a number of issues in this comparision that call for caution .
for example , the census of mw dwarfs is still incomplete and limited by the surface brightness that observational surveys can reach .
it is thus unclear the mass - size relation shown in fig .
[ fig : mrh ] holds for the entire satellite population .
also , we have implicitely assumed a lack of correlation between the mass and the stellar properties of dwarf galaxies . although this is motivated by observational data ( e.g. mateo 1999 ; walker et al .
2007 ; pearrubia et al .
2008a ) , self - consistent cosmological models that follow star formation / feedback in subhaloes are required to carry out a more quantitative analysis , as well as to explore in depth the contraints on galaxy formation models provided by the dsph population in the local group .
, as a function of host disc mass .
points are grey - coded according to the accuracy of cuspy ( _ upper panel _ ) and cored ( _ lower panel _ ) satellite models in describing the mass - size relation shown by mw dwarfs ( see fig .
[ fig : mrh ] ) .
note that ( i ) cuspy models provide a considerably better match to observations and ( ii ) a mass threshold at @xmath174@xmath173 is clearly preferred for a large range of disc masses.,width=317 ]
we have used n - body simulations to explore the differences that dark matter cusps and cores introduce in tidal mass stripping of satellite galaxies .
our models assume cosmologically - motivated initial conditions , where satellites are dark - matter dominated systems moving on eccentric orbits within a much more massive host galaxy , which is composed of a `` baryonic '' disc embedded in an extended dark matter halo , whose density profile is assumed to be the same as that of satellite galaxies .
our findings can be summarized as follows * our models show that the resilience of satellite galaxies to tides increases extraordinarily with the slope of the inner halo profile , @xmath8
. in particular , satellite models with @xmath101 can not be fully disrupted by the tidal field of the parent galaxy , and always retain a bound remnant , even after losing more that 99.99% of mass to tides .
in contrast , satellites with cored haloes @xmath66 may undergo full tidal disruption if their orbits bring them in the vicinity of the disc .
* tides do not affect the inner structure of dark matter haloes , so that @xmath8 stays constant during the evolution of satellite galaxies . remarkably , the structural parameters of the satellite halo , e.g. the peak velocity @xmath108 and its location @xmath106 , evolve in a manner that varies depending on the value of @xmath8 , but that is solely controlled by the total amount of mass lost to tides . to examine the differences that dark matter cores and cusps may introduce in the present satellite population surrounding spiral galaxies , we carry out ten different realizations of merger trees that describe the hierarchical formation of a spiral galaxy with properties similar to those of the mw .
these realizations are re - run adopting different values of @xmath8 as well as disc - to - halo mass ratios , @xmath175
. we account for the fact that only a small fraction of subhaloes may be `` luminous '' by only considering those that fall into the host with masses above @xmath129 .
our main conclusions are the following * the net effect of tides is to decrease the average mass of satellite galaxies at all galactocentric radii .
increasing the mass of the host disc clearly magnifies this effect : the average bound mass drops by a factor @xmath144 after injecting a disc component in the host with @xmath176 , and a factor @xmath177 for @xmath178 ( see also donghia et al .
* the present number of satellite galaxies that are accreted before re - ionization puts strong constraints on the inner profile of dark matter haloes .
our models predict that a mw - like galaxy may contain @xmath179 of these galaxies orbiting within the central 20 kpc if @xmath67 , in very good agreement with recent results from the aquarius team ( gao et al .
the number of these systems , however , quickly drops to zero in models that assume a cored halo profile and host disc component with @xmath180 .
these systems may be singled out by anomalous metallicity patterns , and thus represent an interesting observational case to unravel the inner structure of haloes . * in models that adopt a cored dark matter halo profile , discs efficiently deplete the inner regions of the host from satellites . in contrast , cuspy satellites can survive in the inner - most regions of spiral galaxies even after losing large fractions of their original mass to tides , which may have important observational consequences for the detection of dark matter annihilation signals in gamma - ray surveys . * the relationship between half - light radius and enclosed dynamic mass shown by most local group dwarf galaxies , which spans approximately two orders of magnitude in size and three in mass , is successfully reproduced by our models if adopting a @xmath67 halo model .
in contrast , cored halo models systematically underestimate the masses of the ultra - faint dsphs recently discovered in the mw . * in order to explain the size - mass relationship exhibited by the mw dsph population , a subhalo mass threshold for star formation is needed .
we find that only those subhalos that were accreted with masses _ above _ @xmath1@xmath173 may have hosted luminous satellites that agree with the observational constraints , while the rest remain dark or beyond the present detection limits . *
our models show that high disc - to - halo mass ratios tend to lower the overall masses of satellite galaxies .
a relatively massive m31 disc may explain why many of its dsphs with suitable kinematic data fall below the size - mass relationship established from mw dwarfs .
recent estimates suggest that the disc of m31 may be a factor @xmath155 more massive than that of the mw ( hammer et al .
2007 ) , lending support to this scenario .
this explanation is likely to be validated ( or challenged ) by the results of many ongoing kinematic surveys of m31 dsphs , which will soon determine whether the difference in the size - mass relationship is systematic to the majority of the dsph population surrounding m31 .
jp thanks stelios kazantzidis for kindly providing the code used to generate the spherical equilibrium n - body models .
ajb acknowledges the support of the gordon and betty moore foundation .
we thank the anonymous referee for his / her insightful comments .
aaronson , m. 1983 , , 266 , l11 aaronson , m. , & olszewski , e.
w. 1987 , , 94 , 657 arena , s. e. , & bertin , g. 2007 , , 463 , 921 belokurov , v. , et al .
2007 , , 654 , 897 belokurov , v. , et al .
2009 , , 397 , 1748 benson , a. j. , lacey , c. g. , frenk , c. s. , baugh , c. m. , & cole , s. 2004 , , 351 , 1215 benson , a. j. 2005 , , 358 , 551 benson , a. j. 2008 , , 388 , 1361 bower , r. g. , benson , a. j. , malbon , r. , helly , j. c. , frenk , c. s. , baugh , c. m. , cole , s. , & lacey , c. g. 2006 , , 370 , 645 bullock , j. s. , stewart , k. r. , kaplinghat , m. , & tollerud , e. j. 2009 , arxiv:0912.1873 cole , s. , lacey , c. g. , baugh , c. m. , & frenk , c. s. 2000 , , 319 , 168 cole , s. , helly , j. , frenk , c. s. , & parkinson , h. 2008 , , 383 , 546 collins , m. l. m. , et al .
2009 , arxiv:0911.1365 colpi , m. , mayer , l. , & governato , f. 1999 , , 525 , 720 cooper , a. p. , et al .
2009 , arxiv:0910.3211 de blok , w. j. g. , walter , f. , brinks , e. , trachternach , c. , oh , s .- h .
, & kennicutt , r. c. 2008 , , 136 , 2648 de blok , w. j. g. 2010 , advances in astronomy , 2010 , 5 diemand , j. , moore , b. , & stadel , j. 2004 , , 353 , 624 diemand , j. , zemp , m. , moore , b. , stadel , j. , & carollo , c. m. 2005 , , 364 , 665 diemand , j. , kuhlen , m. , & madau , p. 2007
, , 657 , 262 donghia , e. , springel , v. , hernquist , l. , & keres , d. 2009 , arxiv:0907.3482 eke , v. r. , navarro , j. f. , & steinmetz , m. 2001 , , 554 , 114 evans , n. w. , an , j. , & walker , m. g. 2009 , , 393 , l50 fan , x. , narayanan , v. k. , strauss , m. a. , white , r. l. , becker , r. h. , pentericci , l. , & rix , h .- w .
2002 , , 123 , 1247 fellhauer , m. , kroupa , p. , baumgardt , h. , bien , r. , boily , c. m. , spurzem , r. , & wassmer , n. 2000 , new astronomy , 5 , 305 flores , r. a. , & primack , j. r. 1994 , , 427 , l1 fukushige , t. , & makino , j. 2001 , , 557 , 533 gao , l. , white , s. d. m. , jenkins , a. , stoehr , f. , & springel , v. 2004 , , 355 , 819 gao , l. , navarro , j. f. , cole , s. , frenk , c. s. , white , s. d. m. , springel , v. , jenkins , a. , & neto , a. f. 2008 , , 387 , 536 gao , l. , theuns , t. , frenk , c. s. , jenkins , a. , navarro , j. , springel , v. , & white , s. d. m. 2009 , arxiv:0909.1593 ghigna , s. , moore , b. , governato , f. , lake , g. , quinn , t. , & stadel , j. 2000 , , 544 , 616 goerdt , t. , gnedin , o. y. , moore , b. , diemand , j. , & stadel , j. 2007 , , 375 , 191 goerdt , t. , read , j. i. , moore , b. , & stadel , j. 2008 , arxiv:0806.1951 governato , f. , et al .
2010 , , 463 , 203 haiman , z. , abel , t. , & rees , m. j. 2000 , , 534 , 11 hammer , f. , puech , m. , chemin , l. , flores , h. , & lehnert , m. d. 2007 , , 662 , 322 hashimoto , y. , funato , y. , & makino , j. 2003 , , 582 , 196 hayashi , e. , navarro , j. f. , taylor , j. e. , stadel , j. , & quinn , t. 2003 , , 584 , 541 johnston , k. v. , choi , p. i. , & guhathakurta , p. 2002
, , 124 , 127 just , a. , & pearrubia , j. 2005 , , 431 , 861 kalirai , j. s. , et al .
2009 , arxiv:0911.1998 kampakoglou , m. , & benson , a. j. 2007 , , 374 , 775 kazantzidis , s. , magorrian , j. , & moore , b. 2004 , , 601 , 37 kazantzidis , s. , zentner , a. r. , & kravtsov , a. v. 2006 , , 641 , 647 king , i. 1962 , , 67 , 471 klypin , a. , kravtsov , a. v. , valenzuela , o. , & prada , f. 1999 , , 522 , 82 klypin , a. , zhao , h. , & somerville , r. s. 2002 , , 573 , 597 koch , a. , wilkinson , m. i. , kleyna , j. t. , gilmore , g. f. , grebel , e. k. , mackey , a. d. , evans , n. w. , & wyse , r. f. g. 2007a , , 657 , 241 koch , a. , kleyna , j. t. , wilkinson , m. i. , grebel , e. k. , gilmore , g. f. , evans , n. w. , wyse , r. f. g. , & harbeck , d. r. 2007b , , 134 , 566 koposov , s. , et al .
2008 , , 686 , 279 koposov , s. e. , yoo , j. , rix , h .- w . , weinberg , d. h. , macci , a. v. , & escud , j. m. 2009 , , 696 , 2179 kravtsov , a. v. , klypin , a. a. , bullock , j. s. , & primack , j. r. 1998 , , 502 , 48 kravtsov , a. 2010 , advances in astronomy , 2010 , 8 kuhlen , m. , diemand , j. , & madau , p. 2008
, , 686 , 262 lahav , o. , lilje , p. b. , primack , j. r. , & rees , m. j. 1991 , , 251 , 128 ludlow , a. d. , navarro , j. f. , springel , v. , jenkins , a. , frenk , c. s. , & helmi , a. 2009 , , 692 , 931 macci , a. v. , kang , x. , fontanot , f. , somerville , r. s. , koposov , s. , & monaco , p. 2009
, , 1907 mateo , m. l. 1998 , , 36 , 435 mayer , l. , moore , b. , quinn , t. , governato , f. , & stadel , j. 2002 , , 336 , 119 mayer , l. , kazantzidis , s. , mastropietro , c. , & wadsley , j. 2007 , , 445 , 738 miyamoto , m. , & nagai , r. 1975 , , 27 , 533 mo , h. j. , mao , s. , & white , s. d. m. 1998 , , 295 , 319 moore , b. , ghigna , s. , governato , f. , lake , g. , quinn , t. , stadel , j. , & tozzi , p. 1999a , , 524 , l19 moore , b. , quinn , t. , governato , f. , stadel , j. , & lake , g. 1999b , , 310 , 1147 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1996 , , 462 , 563 navarro , j. f. , eke , v. r. , & frenk , c. s. 1996 , , 283 , l72 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1997 , , 490 , 493 ( nfw ) navarro , j. f. , et al .
2004 , , 349 , 1039 navarro , j. f. , et al .
2010 , , 402 , 21 oh , k. s. , lin , d. n. c. , & aarseth , s. j. 1995 , , 442 , 142 oh , s .- h . , de blok , w. j. g. , walter , f. , brinks , e. , & kennicutt , r. c. 2008 , , 136 , 2761 parkinson , h. , cole , s. , & helly , j. 2008 , , 383 , 557 pearrubia , j. , just , a. , & kroupa , p. 2004
, , 349 , 747 pearrubia , j. , & benson , a. j. 2005 , , 364 , 977 pearrubia , j. , mcconnachie , a. w. , & navarro , j. f. 2008a , , 672 , 904 pearrubia , j. , navarro , j. f. , & mcconnachie , a. w. 2008b , , 673 , 226 pearrubia , j. , navarro , j. f. , mcconnachie , a. w. , & martin , n. f. 2009 , , 698 , 222 power , c. , navarro , j. f. , jenkins , a. , frenk , c. s. , white , s. d. m. , springel , v. , stadel , j. , & quinn , t. 2003 , , 338 , 14 prugniel , p. , & combes , f. 1992 , , 259 , 25 read , j. i. , & gilmore , g. 2005 , , 356 , 107 read , j. i. , goerdt , t. , moore , b. , pontzen , a. p. , stadel , j. , & lake , g. 2006 , , 373 , 1451 ricotti , m. 2009 , arxiv:0911.2792 salucci , p. , & burkert , a. 2000 , , 537 , l9 simon , j. d. , bolatto , a. d. , leroy , a. , & blitz , l. 2003 , , 596 , 957 simon , j. d. , & geha , m. 2007 , , 670 , 313 spergel , d. n. , et al .
2007 , , 170 , 377 springel , v. , et al . 2008a , , 391 , 1685 springel , v. , et al . 2008b , , 456 , 73 stoehr , f. , white , s. d. m. , tormen , g. , & springel , v. 2002 , , 335 , l84 stoehr , f. , white , s. d. m. , springel , v. , tormen , g. , & yoshida , n. 2003 , , 345 , 1313 strigari , l. e. , bullock , j. s. , kaplinghat , m. , simon , j. d. , geha , m. , willman , b. , & walker , m. g. 2008 , , 454 , 1096 swaters , r. a. , verheijen , m. a. w. , bershady , m. a. , & andersen , d. r. 2003 , , 587 , l19 tassis , k. , kravtsov , a. v. , & gnedin , n. y. 2008 , , 672 , 888 taylor , j. e. , & babul , a. 2001 , , 559 , 716 tollerud , e. j. , bullock , j. s. , strigari , l. e. , & willman , b. 2008 , , 688 , 277 trachternach , c. , de blok , w. j. g. , walter , f. , brinks , e. , & kennicutt , r. c. 2008 , , 136 , 2720 van den bosch , f. c. , lewis , g. f. , lake , g. , & stadel , j. 1999 , , 515 , 50 velazquez , h. , & white , s. d. m. 1999 , , 304 , 254 wahde , m. , & donner , k. j. 1996 , , 312 , 431 walker , m. g. , mateo , m. , olszewski , e. w. , gnedin , o. y. , wang , x. , sen , b. , & woodroofe , m. 2007 , , 667 , l53 walker , m. g. , mateo , m. , olszewski , e. w. , pearrubia , j. , wyn evans , n. , & gilmore , g. 2009 , , 704 , 1274 wilkinson , m. i. , kleyna , j. , evans , n. w. , & gilmore , g. 2002 , , 330 , 778 wolf , j. , martinez , g. d. , bullock , j. s. , kaplinghat , m. , geha , m. , munoz , r. r. , simon , j. d. , & avedo , f. f. 2009 , arxiv:0908.2995 zentner , a. r. , & bullock , j. s. 2003 , , 598 , 49 zhao , h. 1996 , , 278 , 488
to follow the dynamical evolution of dark matter subhaloes , we have constructed a simple semi - analytic code that computes their orbit , mass and structural evolution since they cross for the first time the virial radius of the host galaxy to the present .
this code permits us to explore in a flexible manner aspects of galaxy formation that , if simulated with the aid of large n - body algorithms , would be notably _ expensive _ in terms of cpu requirements , such as the importance of cosmic variance in making observational predictions , or the effects of a varying the baryonic disc mass on the satellite galaxy population . to calculate their orbits , we treat dark matter subhaloes as point masses that move in a host halo potential whose parameters evolve as shown in [ sec : mtree ] . at each time step ,
our code solves the following equation of motion where the first term on the right side of the equation is the host halo force @xmath182 and @xmath183 where @xmath184 has been defined in eq . [
[ eq : rho ] ] and [ [ eq : rhotrun ] ] . the second term on the right side of the equation refers to the force induced by the baryonic disc , which we approximate as a miyamoto - nagai ( 1975 ) model ( see [ sec : mtree ] ) .
the third term on the right side of the equation is the dynamical friction force exerted by the diffuse sea of dark matter particles on massive satellites orbiting around their host .
several studies of satellite orbital decay have shown that , in spherical systems , chandrasekhar s formula for dynamical friction ( chandrasekhar 1943 ) is sufficiently accurate if the coulomb logarithm is treated as a free parameter to fit to n - body orbits ( e.g. colpi , mayer & governato 1999 ; van den bosch et al .
semi - analytic methods that include chandrasekhar s dynamical friction have been demonstrated to reproduce accurately the overall dynamical evolution of satellite galaxies in a cosmological context ( e.g. velzquez & white 1999 ; taylor & babul 2001 ; zentner & bullock 2003 ; benson et al .
2004 ; pearrubia & benson 2005 ) and therefore represent a useful tool for conducting extensive studies of a large parameter space . dividing the background potential into a disc and dark matter halo components , chandrasekhar s formula can be written as @xmath185 where @xmath186 and @xmath187 are the coulomb logarithms of the disc and halo components , respectively .
the relative velocity of a satellite with respect to disc and dark matter halo particles is simply @xmath188 , where @xmath189 is the disc circular velocity on the galaxy plane ( @xmath190 ) ; and @xmath191 , where @xmath192 is the satellite velocity vector . for simplicity
we assume here that the velocity distributions @xmath193 of both disc and host halo are maxwellian and isotropic @xmath194 ; \label{eq : fv}\ ] ] where @xmath195 and @xmath196 is the one - dimensional velocity dispersion , which is defined as @xmath197dr'$ ] .
kazantzidis et al . (
2004 ) have examined the inadequacies of the local maxwellian approximation when applied to nfw halos and , although it may introduce problems to generate halos in perfect equilibrium , the deviations from gaussianity of the exact distribution function have a negligible impact on the friction term ( see their fig .
4 ) . pearrubia et al . (
2004 ) find that the best fit to self - consistent n - body orbits is obtained for a coulomb logarithm @xmath198 , in good agreement with the recent results of arena & bertin ( 2007 ) .
following taylor & babul ( 2001 ) , we use @xmath199 .
we must emphasize , however , that these quantities are fairly sensitive to numerical aspects , such as particle number and spatial resolution ( e.g. prugniel & combes 1992 , wahde & donner 1996 ) as well as to the host s density profile ( hashimoto et al .
2003 , just & pearrubia 2005 , read et al .
2006 , arena & bertin 2007 ) .
it is therefore necessary to note that the decay times of satellites orbiting in dark matter halos are uncertain to a significant degree , since orbital decay rates depend on the value assumed for @xmath200 and @xmath201 .
satellite galaxies may lose a fraction of their original mass to tides if their orbits bring them into the inner regions of the parent galaxy . a simple way to estimate
the bound mass fraction of a satellite galaxy is to define the tidal radius as the distance from the satellite centre at which the force due to the satellite s self - gravity and that from the external tidal force balance . for satellites moving on circular orbits , one can estimate that radius as ( king 1962 ) @xmath202 where @xmath203 is the angular frequency of the orbit and @xmath204 is the potential of the host halo . in calculating @xmath205
, we will average over the asphericity of the potential due to the disc component , so that @xmath206 ) at different orbital apocentres ( fig .
[ fig : rhoevol ] ) . profiles are arbitrarily normalized by the quantity @xmath207 , which is kept fixed in each panel to emphasize the density drop resulting from tidal mass stripping .
dotted lines correspond to our analytical model predictions ( see text ) .
, width=317 ] as fig .
[ fig : rhoevol ] illustrates , tidal mass stripping lowers the underlying density of dark matter subhaloes at all radii , but more effectively in the outer - most regions .
this process has been studied in some detail by pearrubia et al .
( 2008b , 2009 ) , who showed that the outer profile of dark matter haloes that lose a fraction ( even as small as 10% ) of its mass to tides can be well fitted by a power - law profile that scales as @xmath208 as @xmath209 . in adopting the density profile outlined by eq . [
[ eq : rho ] ] , this means that @xmath210 .
it is interesting to note that stripped haloes have @xmath211 , so that the resulting cumulative mass profile does _ not _ diverge at large radii .
remarkably , in the inner - most regions the dark matter halo profile is barely altered by the action of tides . in practical terms
this means that @xmath8 as well as @xmath13 can be assumed to stay nearly constant throughout the dynamical evolution of our subhalo models .
furthermore , the evolution of the structural parameters of a subhalo as it is stripped can be accurately described by _ tidal tracks _
, i.e. mono - parametric functions that only depend on the fraction of mass lost to tides ( see fig . [ fig : vmrm ] ) .
we can use the above information to describe the changes induced by tidal mass stripping on the density profiles of individual dark matter subhaloes as a function of their bound mass fraction .
this is done in a fairly simple way : ( i ) for @xmath212 the indices of eq . [ [ eq : rho ] ] are modified as @xmath213 .
( ii ) for a given value of @xmath214 we calculate @xmath215 using eq . [
[ eq : g ] ] .
( iii ) the new scale radius @xmath216 is then calculated from the condition @xmath217 , ( iv ) whereas @xmath218 is obtained using the definition @xmath219 .
this simple method permits us to follow the evolution of the density profile of satellite galaxies as they orbit within their host with an appropiate degree of confidence . for example , symbols in fig .
[ fig : rhotest ] illustrate the density profiles shown in fig .
[ fig : rhoevol ] , whereas dotted lines show the profiles derived from the above recipes .
except for the excess of mass visible at large radii in some cored models , which corresponds to unbound material on its way to escape from the subhalo , our semi - analytic algorithm successfully matches the evolved density profiles , even those of subhaloes that have lost a large fraction of their initial mass .
this code will permit us to explore in a very efficient way aspects of galaxy formation that would otherwise require an extensive use of supercomputers , such as accounting for cosmic variance to make theoretical predictions , or the effects that varying the host galaxy disc - to - halo mass ratio would induce on the surviving satellite galaxy population .
we have used the n - body simulations of [ sec : tidev ] to test our mass loss algorithm , as well as the analytical description of the effects that tides cause on the density profiles of dark matter haloes .
the results are shown in fig .
[ fig : inditest ] , where we plot the mass our subhalo models as a function of time as derived from our n - body models ( see fig . [ fig : mass ] ) , as well as from the above semi - analytical method ( dotted lines ) .
the agreement between both techniques is reasonable , taking into account the extreme mass - loss events considered here and the large range of orbits and structural parameters adopted in this study . more importantly ,
our simple semi - analytical description is able to capture the strong effect that discs have on the mass evolution of dark matter subhaloes depending on the inner shape of the density profile , which will permit us to explore the effects that baryonic discs have on the satellite galaxy population depending on whether haloes are cuspy or cored . | we use n - body simulations to study the effects that a divergent ( i.e. `` cuspy '' ) dark matter profile introduces on the tidal evolution of dwarf spheroidal galaxies ( dsphs ) .
our models assume cosmologically - motivated initial conditions where dsphs are dark - matter dominated systems on eccentric orbits about a host galaxy composed of a dark halo and a baryonic disc .
we find that the resilience of dsphs to tidal stripping is extremely sensitive to the cuspiness of the inner halo profile ; whereas dwarfs with a cored profile can be easily destroyed by the disc component , those with cusps always retain a bound remnant , even after losing more than 99.99% of the original mass .
for a given halo profile the evolution of the structural parameters as driven by tides is controlled solely by the total amount of mass lost .
this information is used to construct a semi - analytic code that follows the tidal evolution of individual satellites as they fall into a more massive host , which allows us to simulate the hierarchical build - up of spiral galaxies assuming different halo profiles and disc masses .
we find that tidal encounters with discs tend to decrease the average mass of satellite galaxies at all galactocentric radii . of all satellites , those accreted before re - ionization ( @xmath0 ) , which may be singled out by anomalous metallicity patterns ,
provide the strongest constraints on the inner profile of dark haloes .
these galaxies move on orbits that penetrate the disc repeatedly and survive to the present day _ only _ if haloes have an inner density cusp .
we show that the size - mass relationship established from milky way ( mw ) dwarfs strongly supports the presence of cusps in the majority of these systems , as cored models systematically underestimate the masses of the known ultra - faint dsphs .
our models also indicate that a massive m31 disc may explain why many of its dsphs with suitable kinematic data fall below the size - mass relationship derived from mw dsphs .
we also examine whether our modelling can constrain the mass threshold below which star formation is suppressed in dark matter haloes .
we find that luminous satellites must be accreted with masses above @xmath1@xmath2 in order to explain the size - mass relation observed in mw dwarfs .
galaxies : halos galaxy : evolution galaxy : formation galaxy : kinematics and dynamics |
low surface brightness ( lsb ) galaxies are characterized by low star formation rate @xcite and low disk surface density @xcite .
the spiral structure in lsbs is often incipient or fragmentary and usually faint and difficult to trace @xcite , and they generally do not host any strong large - scale spiral structure , the kind we see in case of normal hsb galaxies like our milky way .
we note that , we are interested only in small lsbs , which are more abundant , and do not include the giant lsbs like malin 1 . some of the giant lsbs show fairly strong , large - scale spiral structure as in ugc 6614 @xcite and in the inner regions they are dynamically similar to their high surface brightness ( hsb ) counterparts @xcite .
@xcite has applied the technique of density - wave theory to put constraint on the decomposition of the rotation curve in lsbs .
however we note that in the sample considered by @xcite contains giant lsb ( e.g. , ugc 6614 ) which often show large - scale spiral structure ( e.g. , see * ? ? ?
the lsbs are dark matter dominated from the very inner regions @xcite . within the optical disk ,
the dark matter constitutes about @xmath1 per cent of the total mass of lsbs , whereas for the hsbs the contributions of the dark matter halo mass and stellar mass are comparable @xcite .
thus , the lsbs constitute a natural laboratory to study the effect of dark matter halo on different aspects of galactic dynamics .
several past studies have shown the effect of dominant dark matter halo in the suppression of global non - axisymmetric bar modes @xcite , in making the galactic disks superthin @xcite and in prohibiting the swing amplification mechanism from operating , thus explaining the lack of small - scale spiral structure as noted observationally @xcite . according to the density wave theory ,
the grand - design spiral arms are the high density regions of a rigidly rotating spiral density wave , with a well defined pattern speed , that are self - consistently generated by the combined gravity of the unperturbed disk and the density wave @xcite .
for a recent review on this see @xcite . in a recent work @xcite ( hereafter paper
1 ) investigated the role of a dominant dark matter halo on the global spiral modes within the framework of the density wave theory by treating the galactic disk as a fluid . using the input parameters of a superthin lsb galaxy ugc 7321 and the galaxy , they found that for ugc 7321 , the dark matter halo has a negligible effect on arresting the global spiral modes when the disk is modeled as a fluid .
this is in contrast to the results for small - scale spiral features where the dark matter was shown to suppress the small - scale , swing - amplified spiral structures almost completely @xcite .
ghosh et al .
( paper 1 ) argued that that since lsbs are relatively isolated , tidal interactions are less likely to occur compared to those for the high surface brightness ( hsb ) galaxies .
thus even though the global spiral modes are formally permitted in the fluid disk plus dark matter halo model , it was argued that it is the lack of tidal interaction that makes it difficult for the global spiral structure to develop in these galaxies . in this paper
we address the effect of dark matter halo on global @xmath0 modes by modeling the galactic disk more realistically as a @xmath2 system .
a tidal encounter is likely to give rise to global modes , as has been seen in simulations , as in m51 ( see e.g. , * ? ? ?
we use the dispersion relation for a @xmath2 disk to construct global standing - wave like solutions by invoking the bohr - sommerfeld quantization condition ( for details see paper 1 ) .
note that fluid disks allow wavelike solutions at small wavelengths , since fluid pressure provides the restoring force ; but collisionless disks suppress wavelike modes for very small wavelengths ( e.g. see * ? ? ?
the 2 contains formulation of the problem and the input parameters . in 3 we present the wkb analysis and the relevant quantization rule . 4 * 5 * contain the results and discussion , respectively while 6 contains the conclusions .
we model the galactic disk as a collisionless system characterized by an exponential surface density @xmath3 , and one - dimensional velocity dispersion @xmath4 . for simplicity ,
the galactic disk is taken to be infinitesimally thin .
in other words , we are interested in perturbations that are confined to the mid - plane ( @xmath5 ) . the dark matter halo is assumed to be non - responsive to the gravitational perturbations of the disk .
we have used cylindrical coordinates @xmath6 in our analysis . in this subsection
, we describe the models that we have used for the study of effect of dark matter halo on the global spiral modes .
the dynamics of the disk is calculated first only under the gravity of the disk ( referred to as disk - alone case ) and then under the joint gravity of disk and the dark matter halo ( referred to as disk plus halo case ) .
we took an exponential stellar disk with central surface density @xmath7 and the disk scalelength @xmath8 , which is embedded in a concentric dark matter halo whose density follows a pseudo - isothermal profile characterized by core density @xmath9 and core radius @xmath10 . the net angular frequency , @xmath11 and the net epicyclic frequency , @xmath12 for a galactic disk embedded in a dark matter halo , concentric to the galactic disk , are given as : + @xmath13 the expressions for @xmath14 , @xmath15 in the mid - plane ( @xmath5 ) for an exponential disk and @xmath16 , @xmath17 for a pseudo - isothermal halo in the mid - plane ( @xmath5 ) have been calculated earlier ( see paper 1 for details ) . in the early - type galaxies ,
the bulge component dominates in the inner regions , and hence exclusion of the bulge component from the models for early - type galaxies will underestimate the rotation curve in the inner regions . in paper 1 , it was shown that for the galaxy , inclusion of bulge yielded more realistic results ( for details see 5.1 in paper 1 ) .
also note that ugc 7321 has no discernible bulge @xcite .
therefore , only for our galaxy we have included bulge in both the disk - alone and disk plus dark matter halo models .
we adopt a plummer - kuzmin bulge model for our galaxy which is characterized by total bulge mass @xmath18 and bulge scalelength @xmath19 .
the expressions for @xmath20 , @xmath21 in the mid - plane ( @xmath5 ) for such a bulge having a plummer - kuzmin profile are given in paper 1 .
therefore , for the galaxy , these @xmath20 and @xmath21 terms will be added in quadrature to the corresponding terms due to disk and dark matter halo .
the input parameters for different components of ugc 7321 @xcite and the galaxy @xcite are summarized in table 1 . [
cols="^,^,^,^,^,^,^ " , ] shows that the modes do not extend up to the resonance points , consistent with the fact that @xmath22 is not equal to zero anywhere for modes in fig .
[ fig3].,title="fig:",width=326,height=230 ] shows that the modes do not extend up to the resonance points , consistent with the fact that @xmath22 is not equal to zero anywhere for modes in fig .
[ fig3].,title="fig:",width=336,height=230 ] from table 2 , it is evident that the dark matter halo has a negligible effect on the global spiral modes for the galaxy , the difference can be seen only in the change of the specific values of pattern speed @xmath23 .
this result is at par with our expectation , since the dark matter halo is known to be not important in the inner regions of the galaxy ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . these results are consistent with the results of paper 1 where the galactic disk was modeled as a fluid disk .
the key difference with results from paper 1 is that the specific values of pattern speed @xmath23 which give the global modes are different ( for comparison , see table 3 in paper 1 ) .
@xcite found a range of @xmath24 km s@xmath25 kpc@xmath25 for the pattern speed of spiral arms for the galaxy . on comparison
the pattern speed values obtained in this paper lie outside the observed range of pattern speed .
the shape of the closed contours in this work ( figs .
1 - 3 ) is different from that in paper 1 .
a curve of constant @xmath23 had nearly identical @xmath22 values at the turning points of a closed curve of _ type b _ in paper 1 , whereas the contours look slanted in the present paper , meaning that the @xmath22 values at the turning points are substantially different .
this is due to the difference in the dispersion relation for the two cases .
a more detailed interpretation is not possible in a simple analytical form , since the dispersion relation for the collisionless case ( equation ( [ disp - equation ] ) ) has a transcendental form . in this paper for simplicity
we have not included the low velocity dispersion component , namely , gas in the system which could further modify the obtained range of pattern speeds .
note that any late - type spiral galaxy contains non - negligible amount of gas , and it has been shown that the gas plays a significant role in various dynamical issues ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the amount of gas present in ugc 7321 is quite small in comparison to that in hsb galaxies @xcite .
therefore , the results for ugc 7321 obtained in this paper - in particular the suppression of global modes by the dominant dark matter halo is unlikely to change significantly if gas is included in the calculation .
the effect of gas on the global spiral modes in a gravitationally coupled two - component ( stars and gas ) system , as applicable to the gas - rich hsb galaxies , will be taken up in a future paper .
the results presented here are based on the assumption that the grand - design spiral structure seen in disk galaxies is due to the density waves , as proposed by @xcite .
however this hypothesis has not yet been fully confirmed observationally .
for example , the angular off - set in age of the stellar population , as predicted by the classical density wave theory has not been found in studies by @xcite in their sample of late - type disk galaxies .
also the study of @xmath26 observation of ngc 1068 using generalized tremaine - weinberg method by @xcite has revealed that the pattern speed of the spiral structure of this galaxy varies rapidly with radius , indicative of short - lived spiral features .
a study of the gas content and the @xmath27 line of sight velocity distribution in ngc 6754 by @xcite revealed a different sense of streaming motion in the trailing and leading edge of this galaxy , indicating that the spiral arms of this galaxy is likely to be a transient . on the other hand , it is also worth noting that @xcite have recently presented a study of measuring pitch angle from the images taken in different wavelengths for a large sample of disk galaxies and found that the pitch angle varies from one wavelength to another as predicted by the classical density wave theory , and thus furnishes strong observational evidence for the validity of the density - wave theory of spiral structure in disk galaxies
. generally the various @xmath28-body studies so far do not show the evidence of long - lived spiral structure .
several @xmath28-body simulations have shown that the spiral arms are transient and get wound up quickly @xcite , which goes against the classical density wave picture .
interestingly an opposite trend is shown in a recent work by @xcite who reported long - lived spiral structure in a @xmath28-body study of models for galaxies with intermediate bulges .
we have investigated the effect of a dominant dark matter halo on the possible existence of global spiral arms while modeling the galactic disk as a _ collisionless _ system .
the wkb dispersion relation and the bohr - sommerfeld quantization condition were used to obtain discrete global spiral modes present in any model .
we have analysed a superthin lsb galaxy ugc 7321 , and the galaxy , for this work .
we found that for ugc 7321 both the disk - alone and the disk plus dark matter halo cases did not yield any discrete global spiral modes .
even an increase in the stellar central surface density by a factor of few failed to produce any global spiral modes .
thus our findings provide a natural explanation for the observed dearth of strong large - scale spiral structure in the lsbs .
our results differ from those obtained in paper 1 where the galactic disk was treated as a fluid , where it was found that the dominant dark matter halo had a negligible constraining effect on the existence of global spiral modes in these galaxies .
this difference is due to the different dispersion relation for fluid and collisionless systems , which played a pivotal role in determining the discrete global mode(s ) present in a model .
thus the correct collisionless treatment for stars as done here has led us to a result for the non - existence of global spiral modes which is in agreement with the observations . as a check , for the galaxy we carried out a similar modal analysis and found that the dark matter halo has a negligible effect on the global spiral modes , as expected since the galaxy is not dark matter dominated in the inner regions , in contrast to ugc 7321 .
thus , the dark matter halo that dominates from the innermost regions is shown to suppress the growth of local , swing - amplified non - axisymmetric features @xcite as well as the global spiral modes as shown here .
* acknowledgements : * we thank the anonymous referee for the constructive comments that have helped to improve the paper .
cj would like to thank the dst , government of india for support via j.c .
bose fellowship ( sb / s2/jcb-31/2014 ) .
banerjee a. , jog c.j .
2013 , mnras , 431 , 582b | low surface brightness ( lsb ) galaxies are dominated by dark matter halo from the innermost radii ; hence they are ideal candidates to investigate the influence of dark matter on different dynamical aspects of spiral galaxies . here
, we study the effect of dark matter halo on grand - design , @xmath0 , spiral modes in a galactic disk , treated as a collisionless system , by carrying out a global modal analysis within the wkb approximation .
first , we study a superthin , lsb galaxy ugc 7321 and show that it does not support discrete global spiral modes when modeled as a disk - alone system or as a disk plus dark matter system .
even a moderate increase in the stellar central surface density does not yield any global spiral modes .
this naturally explains the observed lack of strong large - scale spiral structure in lsbs .
an earlier work @xcite where the galactic disk was treated as a fluid system for simplicity had shown that the dominant halo could not arrest global modes .
we found that this difference arises due to the different dispersion relation used in the two cases and which plays a crucial role in the search for global spiral modes .
thus the correct treatment of stars as a collisionless system as done here results in the suppression of global spiral modes , in agreement with the observations .
we performed a similar modal analysis for the galaxy , and found that the dark matter halo has a negligible effect on large - scale spiral structure .
galaxies : kinematics and dynamics , galaxies : spiral , galaxies : structure , galaxies : individual : ugc 7321 , galaxies : halos , instabilities |
[ [ section ] ] the generation of entanglement is an essential tool for the realization of scalable quantum computing and long distance quantum communication .
the ability to entangle a photon for reliable long range quantum communication and another particle with a long - term quantum memory would allow for the construction of such a scalable quantum network .
large strides have been made towards this in many different physical media .
entanglement and the means for communication have been demonstrated most recently in quantum dots @xcite , nitrogen vacancy centers in diamond @xcite , neutral atoms @xcite , atomic ensembles @xcite , superconducting qubits @xcite , and ions @xcite .
here we report on entanglement and bell inequality violation in a system of a photon and a ^138^ba^+^ ion .
this system is particularly well suited for quantum computation @xcite and communication because its transitions have relatively long wavelengths making them more suitable for long range fiber optic transmission .
[ [ section-1 ] ] in this experiment , entanglement was produced by spontaneous emission of a photon from an excited ^138^ba^+^ ion with multiple decay channels similar to @xcite .
the two entangled qubits were the zeeman sublevels of the @xmath1 ground state of the ion and the emitted photon s polarization .
entanglement was verified by performing state detection on each qubit in two non - orthogonal bases .
[ [ section-2 ] ] the famous 1935 paper by einstein , podolsky , and rosen concluded that the description of reality given by the wave function in quantum mechanics was not complete @xcite .
this was based on the observation that entangled quantum systems require nature to be nonlocal or non - real . however in 1964 , john bell demonstrated that the principles of locality and realism assumed by einstein , podolsky , and rosen in their paper were inconsistent with the predictions of quantum mechanics @xcite .
the mathematical formalism he presented gave an inequality that must be obeyed by any theory satisfying local realism .
many experimental tests have shown violations of some form of the bell inequality @xcite , however none have simultaneously closed both of the major loopholes ( the detection , or fair sampling , " loophole and the locality loophole ) associated with such a measurement .
[ [ section-3 ] ] in our experiment , we demonstrate violation of the bell inequality of the form suggested by clauser , horne , shimony , and holt ( chsh ) @xcite . experimental tests of the chsh inequality have three requirements : ( 1 ) repeated creation of entangled pairs of qubits , ( 2 ) the ability to rotate each qubit independently through a polar angle @xmath2 , @xmath3 on the bloch sphere , and ( 3 ) state measurement of each qubit . chsh show that all theories satisfying local realism must obey the following inequality : @xmath4 where @xmath5 is a correlation function measurement , and @xmath6 is the fraction of the total events in which qubits a and b are found to be in states denoted by @xmath7 and @xmath8 respectively following rotations of the qubits by polar angles @xmath2 and @xmath3 respectively on the bloch sphere . [ [ section-4 ] ] quantum mechanics predicts that the chsh inequality can be violated for particular entangled states and measurements .
for example , given the bell state @xmath9 , one calculates the corresponding correlation function @xmath10 .
this results in a violation of eq .
( [ chsh ] ) for certain angles and maximum violation @xmath11 for @xmath12 , @xmath13 and @xmath14 , @xmath15 .
[ [ section-5 ] ] the bell signal measured in an experiment can be reduced , however , by infidelity in the state that is produced .
a measure of entanglement fidelity can be found by calculating a state s overlap with the appropriate maximally entangled bell state @xcite .
thus with respect to the particular bell state @xmath9 , the fidelity is given by the expectation value of the density matrix ( @xmath16 ) in the bell state : @xmath17 where @xmath16 is given in the computational basis , @xmath18 .
the first two density matrix elements are just the correlation probabilities of detecting the states @xmath19 with @xmath20 and @xmath21 with @xmath22 .
these correlation probabilities ( @xmath23 ) are given by the probability of detecting qubit @xmath24 in the state corresponding to @xmath25 multiplied by the conditional probability of detecting qubit @xmath26 in the state corresponding to @xmath25 given that the state of qubit @xmath24 is the state corresponding to @xmath25 .
the second two elements can be determined by rotating each qubit s measurement basis by the polar angle @xmath27 on the bloch sphere .
then @xmath28 , where @xmath29 is the @xmath27 polar rotation operator with relative phase @xmath30 . with @xmath30
set to zero , one finds that @xmath31 thus we find a lower bound for the fidelity expressed in terms of the correlation probabilities in the original and rotated bases : @xmath32 one also finds the expected bell signal for the angles of maximal violation from the density matrix in the unrotated and rotated bases ( with the relative phase of the rotation @xmath33 ) of : @xmath34
[ [ section-6 ] ] the ion was confined in a linear paul trap and doppler cooled on the @xmath1 to @xmath35 transition with a 493 nm laser .
a 650 nm laser was used to repump from the long - lived @xmath36 state ( lifetime @xmath37 s @xcite ) to which the @xmath35 state may decay .
an applied magnetic field @xmath38 gauss provided a quantization axis , and a circularly polarized beam of 493 nm light aligned parallel to the quantization axis was used for optical pumping to either of the two zeeman sublevels of the @xmath1 ground state .
[ [ section-7 ] ] ionic qubit state detection was accomplished by the use of a stabilized 1762 nm fiber laser to drive the shelving " transition from the @xmath1 ( @xmath39 ) ground state sublevel ( defined as @xmath40 ) to the @xmath41 ( @xmath42 ) metastable sublevel ( lifetime @xmath43 s ) using adiabatic rapid passage sweeps @xcite .
once excited to the @xmath41 shelved state , the ion is removed from the cooling cycle and will appear dark " while the cooling lasers are incident on the ion , whereas an unshelved ion will be bright . "
[ [ section-8 ] ] a dedicated wire loop in the vicinity of the trap inside the vacuum chamber was used to generate a tunable radio frequency ( rf ) magnetic field in order to drive magnetic dipole transitions between the zeeman sublevels for the necessary ion qubit rotations .
an rf synthesizer ( based on a direct digital synthesis ( dds9910 ) chip ) generated a stable sinusoidal voltage which was amplified and dropped over a 50 ohm rf resistor for impedance matching , producing an oscillating current in the loop .
this setup allowed for ion qubit rotations with a tunable phase .
an example of ground state rabi oscillations is shown in fig . [ zeemanflops ] .
the reduction in contrast is due to imperfect optical pumping ( @xmath44 ) and shelving efficiency ( @xmath45 ) .
efficient transfer of the @xmath46 ground state to the @xmath47 level .
thus , a shelving efficiency near 1 corresponds to the ion in the @xmath39 ground state , while a shelving efficiency near 0 corresponds to the ion in the @xmath48 ground state .
( color online . ) ] [ [ section-9 ] ] spontaneously emitted photons on the @xmath35 to @xmath1 transtion were collected through an @xmath49 lens while other wavelengths were extinguished by an interference filter .
photon collection occurred along a direction perpendicular to the quantization axis .
the two transitions of note in this experiment , @xmath50 and @xmath51 , are emitted into the dipole radiation patterns @xmath52 and @xmath53 .
thus , perpendicular to the quantization axis , i.e. @xmath54 , the two polarizations are linear and orthogonal , with their relative intensity differing by a factor of @xmath55 .
a @xmath56 waveplate along the imaging axis allowed for rotations of the photonic qubit . after passing through a polarizing beamsplitter ( pbs ) , where @xmath50 and @xmath51 correspond to s and p polarizations , respectively , in the unrotated photon basis , the photons were detected at one of two photomultiplier tubes ( pmts ) each with quantumefficiency @xmath57 .
[ [ section-10 ] ] the experimental cycle , resulting states , and the energy level diagram for ^138^ba^+^ are detailed in fig .
[ dutycycle ] . the experimental cycle ( @xmath58 rep rate ) consisted of four major steps .
first the ion was doppler cooled for @xmath59s .
next the ion was pumped to the @xmath1 ( @xmath39 ) ground state sublevel with @xmath60s of the circularly polarized optical pumping beam .
state initialization was completed by following the optical pumping with a resonant rf @xmath51-pulse to transfer the ion state to @xmath1 ( @xmath61 ) .
finally the ion was weakly excited ( excitation probability @xmath62 ) to the @xmath35 ( @xmath39 ) sublevel ( lifetime @xmath63 @xcite ) with a 20 ns exposure to the optical pumping beam .
this was followed by photon- and ion - state detection .
the photons were detected by the pmts inside a detection window of 20 ns , delayed somewhat from the excitation pulse .
the excitation pulse length , detection window , and its delay were optimized to reduce background counts and double excitations while maintaining a reasonable experimental rate . following a single photon detection ,
the ion state was determined by a shelving pulse followed by doppler cooling . dark " state detection efficiency was @xmath45 limited by the shelving efficiency , and bright " state detection efficiency was @xmath64 .
s by 493 nm and 650 nm beams ; ( ii ) state initialization is performed with 10 @xmath65s of optical pumping by a @xmath50-polarized 493 nm beam followed by ( iii ) an 18 @xmath65s rf @xmath51-pulse resonant with the @xmath66 mhz ground state separation ; ( iv ) the ion is weakly excited with the @xmath50-polarized beam ; ( v ) the excited ion decays emitting a @xmath50-polarized or @xmath51-polarized photon , resulting in the corresponding ground state sublevel , and ( vi ) ion detection is performed using the 1762 nm shelving beam , followed by turning on the cooling lasers .
( color online . ) ] [ [ section-11 ] ] the resultant entangled state from the spontaneous decay of the ion is @xmath67 as it decays to either the @xmath1 ( @xmath68 ) or @xmath1 ( @xmath39 ) sublevel while emitting either a @xmath50 or @xmath51 photon respectively with equal probability .
however , along the imaging axis , the radiation patterns of the two polarizations differ by a factor of two .
thus the ideal measured state is @xmath69 whose overlap with a bell state is @xmath70 .
[ [ section-12 ] ] we first measured the correlation between the ionic and photonic qubits by varying the photonic measurement basis with the @xmath56 waveplate . without rotating the ionic qubit ,
the state probabilities are proportional to @xmath71 or @xmath72 , thus correlation fringes are produced with varying @xmath73 . at each waveplate setting , 100 runs were taken , and the conditional probabilities were plotted in fig . [ correlations ] ( i ) . at the rotation angle corresponding to maximum correlation
, 1000 runs were taken for better statistics in the fidelity calculation .
[ [ section-13 ] ] then to verify entanglement , the ionic and photonic qubits were both rotated by @xmath27 on the bloch sphere , and the correlation was measured while varying the phase of the ionic qubit rotation , @xmath74 .
the relative phase of the ionic versus photonic qubit rotations is given by @xmath75 , where @xmath76 is the ground state zeeman splitting and @xmath77 is the time delay between emission of the photon and application of the ionic qubit rotation .
if one assumes that @xmath78 is constant , then varying @xmath74 will produce correlation fringes since the state probabilities are proportional to @xmath79 . at each @xmath74 setting ,
500 runs were taken , and the conditional probabilities were plotted in fig . [ correlations ] ( ii ) . at the maximum correlation , 2000
runs were taken for better statistics in the fidelity calculation . following these measurements ,
a chsh measurement was made with a relative phase @xmath80 using those rotations resulting in a maximal violation of eq .
[ chsh ] . [ [ section-14 ] ]
the entanglement generation rate depends on the probability of success @xmath81 of a single experimental run which relies on the excitation probability and the probability of detecting the spontaneously emitted photon .
given an excitation probability ( @xmath82 ) along with the branching ratio to the ground state ( @xmath83 ) , the quantum efficiency of the pmts ( @xmath57 ) , fraction of solid angle of light collection @xmath84 ) , fraction of photons within pmt detection window ( @xmath85 ) , and the transmission losses ( @xmath86 ) through the imaging optics including an interference filter with @xmath87 transmission , we expect the experimental entanglement generation rate to be @xmath88 which agrees well with our results .
[ [ section-15 ] ] the correlation fringes in the measured conditional probabilities of the photonic and ionic qubit states are plotted in fig .
[ correlations ] .
we find a maximal correlation of @xmath89 , @xmath90 , @xmath91 , and @xmath92 for the unrotated bases and @xmath93 , @xmath94 , @xmath95 , and @xmath96 for the rotated bases where the errors given are statistical for 1000 and 2000 successful trials , respectively .
the correlations in the rotated basis are weaker largely due to timing jitter in the application of the ion state rotation . from this
we use eq .
[ fidelity2 ] to calculate the entanglement fidelity to be @xmath97 , which clearly exceeds the limit of @xmath98 for entanglement @xcite . and @xmath99 ( i ) as a function of the photon qubit measurement basis angle by rotating the half - waveplate while leaving the ion qubit unrotated , and ( ii ) versus the ion - quibt phase , @xmath74 , between the rotations for rotation of each qubit by a polar angle of @xmath27 .
( color online . ) ] and @xmath99 ( i ) as a function of the photon qubit measurement basis angle by rotating the half - waveplate while leaving the ion qubit unrotated , and ( ii ) versus the ion - quibt phase , @xmath74 , between the rotations for rotation of each qubit by a polar angle of @xmath27 .
( color online . ) ] .[tab : bellsig ] correlation function measurements and calculated bell signals . [
cols="^,^,^ " , ] [ [ section-16 ] ] the chsh measurement was made using the maximally violating qubit rotation angles @xmath12 , @xmath27 and @xmath100 , @xmath101 .
the four correlations were measured in two ways : first using the @xmath102 rotations for the ion qubit , and second using the @xmath102 rotations for the photonic qubit . @xmath103
successful entanglement events make up each correlation measurement . from these measurements ,
we find a bell signal of @xmath104 for the first case and @xmath105 for the second case .
the errors are statistical and the bell signal exceeds @xmath55 by over eight standard deviations .
a summary of these results is displayed in table [ tab : bellsig ] .
note that based on the fidelity of the entangled state , we use eq .
[ bellsignal ] to calculate an expected bell signal to be @xmath106 .
[ [ section-17 ] ] the fidelity and bell signal are reduced by several factors .
multiple excitations ( @xmath107 ) , imperfect rotation of the ionic qubit due to different arrival times of photons within the 20 ns pmt detection window ( @xmath108 ) , mixing of photon polarizations ( @xmath109 ) , background and dark counts ( @xmath110 ) , imperfect pbs ( @xmath111 ) , and magnetic field fluctuations ( @xmath112 ) .
[ [ section-18 ] ] the ion - photon entanglement demonstrated here allows for the construction of a quantum network over a large distance .
especially interesting would be a loophole free bell inequality test using photon mediated entanglement swapping between distant ion - photon entangled pairs .
^138^ba^+^ is especially suited for this due to the high fidelity of ion state detection and the relatively low attenuation of 493 nm light over fiber optic cable . in this case
high efficiency of the collection of photons from one entangled ion - photon pair would significantly increase the experimental rate of ion - ion entanglement generation . using a parabolic mirror trap ,
our group has recently attained as much as @xmath113 light collection efficiency from a single trapped ion @xcite .
future work includes improvement of entanglement fidelity by the use of ultrafast pulses in order to approach unit excitation with a single pulse , thus eliminating multiple excitations and background counts .
the authors wish to thank john wright , matt hoffman , tomasz sakrejda , spencer williams , richard graham , zichao zhou , and anupriya jayakumar for useful conversations .
the authors would also like to acknowledge the preceding work of matt dietrich , nathan kurz , paul pham , and ryan bowler that contributed to this experiment .
this research was supported by national science foundation grant no . 0904004 .
32ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1038/nature11577 [ * * , ( ) ] link:\doibase 10.1038/nature11573 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2013.128 [ * * , ( ) ] link:\doibase 10.1038/nature09256 [ * * , ( ) ] link:\doibase 10.1126/science.1157233 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.043604 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.010502 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.260403 [ * * , ( ) ] link:\doibase 10.1038/nature12227 [ * * , ( ) ] link:\doibase 10.1038/nature04353 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.240501 [ * * , ( ) ] link:\doibase 10.1126/science.1130886 [ * * , ( ) ] link:\doibase 10.1126/science.1084528 [ * * , ( ) ] link:\doibase 10.1038/nature06118 [ * * , ( ) ] link:\doibase 10.1038/nature02377 [ * * , ( ) ] http://arxiv.org/abs/1208.0391 [ ( ) ] link:\doibase 10.1103/physrev.47.777 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1038/nature12012 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.150404 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.090401 [ * * , ( ) ] link:\doibase 10.1038/nature08363 [ * * , ( ) ] link:\doibase 10.1038/nature01881 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.95.020403 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.23.880 [ * * , ( ) ] link:\doibase 10.1103/physreva.54.3824 [ * * , ( ) ] link:\doibase 10.1103/physreva.75.052506 [ * * , ( ) ] link:\doibase 10.1103/physreva.85.023401 [ * * , ( ) ] link:\doibase 10.1080/09500340600741082 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/28/11/009 [ * * , ( ) ] link:\doibase 10.1038/35005011 [ * * , ( ) ] @noop * * , ( ) | we report on the demonstration of ion - photon entanglement and bell inequality violation in a system of trapped ^138^ba^+^ ions .
entanglement between the zeeman sublevels of the ground state of a single ^138^ba^+^ ion and the polarization state of a single 493 nm photon emitted by the ion with a fidelity of @xmath0 was achieved , along with a bell signal of 2.3 , exceeding the classical limit of 2 by over eight standard deviations .
this system is a promising candidate for a loophole - free bell inequality violation test as the wavelengths of the transitions of ^138^ba^+^ are in the visible region and thus suitable for long range transmission over fiber optic cable . |
this paper focuses on the average complexity of solving random 3-sat instances using backtrack algorithms .
being an np - complete problem , 3-sat is not thought to be solvable in an efficient way , _
i.e. _ in time growing at most polynomially with @xmath3 .
in practice , one therefore resorts to methods that need , _ a priori _ , exponentially large computational resources .
one of these algorithms is the ubiquitous davis putnam loveland logemann ( dpll ) solving procedure@xcite .
dpll is a complete search algorithm based on backtracking ; its operation is briefly recalled in figure 1 .
the sequence of assignments of variables made by dpll in the course of instance solving can be represented as a search tree , whose size @xmath7 ( number of nodes ) is a convenient measure of the hardness of resolution .
some examples of search trees are presented in figure [ trees ] . in the past few years
, much experimental and theoretical progress has been made on the probabilistic analysis of 3-sat @xcite .
distributions of random instances controlled by few parameters are particularly useful in shedding light on the onset of complexity .
an example that has attracted a lot of attention over the past years is random 3-sat : all clauses are drawn randomly and each variable negated or left unchanged with equal probabilities .
experiments @xcite and theory @xcite indicate that clauses can almost surely always ( respectively never ) be simultaneously satisfied if @xmath1 is smaller ( resp .
larger ) than a critical threshold @xmath8 as soon as the numbers @xmath9 of clauses and @xmath3 of variables go to infinity at a fixed ratio @xmath1 .
this phase transition @xcite is accompanied by a drastic peak in hardness at threshold @xcite .
the emerging pattern of complexity is as follows . at small ratios @xmath10 , where @xmath11 depends on the heuristic used by dpll , instances are almost surely satisfiable ( sat ) ,
see @xcite and @xcite for recent reviews .
the size @xmath7 of the associated search tree scales , with high probability , linearly with the number @xmath3 of variables , and almost no backtracking is present @xcite ( figure [ trees]a ) . above the critical ratio , that is when @xmath12 , instances are a.s .
unsatisfiable ( unsat ) and proofs of refutation are obtained through massive backtracking ( figure [ trees]b ) , leading to an exponential hardness : @xmath13 with @xmath14 @xcite . in the intermediate range , @xmath15 ,
finding a solution a.s . requires exponential effort ( @xmath14 ) @xcite .
the aim of this article is two - fold .
first , we propose a simple and intuitive framework to unify the above findings .
this framework is presented in section 2 .
it is based on the statistical physics notions of dynamical trajectories and phase diagram , and was , to some extent , implicitly contained in the pioneering analysis of search heuristics by @xcite . secondly , we present in section 3 a quantitative study of the growth of the search tree in the unsat regime .
such a study has been lacking so far due to the formidable difficulty in taking into account the effect of massive backtracking on the operation of dpll .
we first establish an exact relationship between the average size of the search tree and the powers of the evolution operator encoding the elementary steps of the search heuristic .
this equivalence is then used ( in a non rigorous way ) to accurately estimate the logarithm @xmath6 of the average complexity @xmath7 as a function of @xmath1 , @xmath16 \quad , \ ] ] where @xmath17 denotes the expectation value for given @xmath3 and @xmath1 .
the approach emphasizes the relevance of partial differential equations to analyse algorithms in presence of massive backtracking , as opposed to ordinary differential equations in the absence of the latter @xcite . in section 4
, we focus upon the upper sat regime _
i.e. _ upon ratios @xmath18 . combining the framework of section 2 and the analysis of section 3 we unveil the structure of the search tree ( figure [ trees]c ) and calculate @xmath6 as a function of the ratio @xmath1 of the 3-sat instance to be solved . for the sake of clarity and since the style of our approach may look unusual to the computer scientist reader , the status of the different calculations and results ( experimental , exact , conjectured , approximate , ... ) are made explicit throughout the article .
.5 cm .5 cm
the action of dpll on an instance of 3-sat causes changes to the overall numbers of variables and clauses , and thus of the ratio @xmath1 .
furthermore , dpll reduces some 3-clauses to 2-clauses . a mixed 2+p - sat distribution , where @xmath2 is the fraction of 3-clauses , can be used to model what remains of the input instance at a node of the search tree .
using experiments and methods from statistical mechanics @xcite , the threshold line @xmath19 , separating sat from unsat phases , may be estimated with the results shown in figure [ diag ] . for @xmath20 , _
i.e. _ to the left of point t , the threshold line is given by @xmath21 , as rigorously confirmed by @xcite , and saturates the upper bound for the satisfaction of 2-clauses .
above @xmath22 , no exact value for @xmath19 is known .
note that @xmath8 corresponds to @xmath23 .
the phase diagram of 2+p - sat is the natural space in which dpll dynamic takes place .
an input 3-sat instance with ratio @xmath1 shows up on the right vertical boundary of figure [ diag ] as a point of coordinates @xmath24 . under the action of dpll , the representative point moves aside from the 3-sat axis and follows a trajectory
this trajectory obviously depends on the heuristic of split followed by dpll ( figure 1 ) .
possible simple heuristics are @xcite , * _ unit - clause ( uc ) : _ randomly pick up a literal among a unit clause if any , or any unset variable otherwise . * _ generalized unit - clause ( guc ) : _ randomly pick up a literal among the shortest avalaible clauses . * _ short clause with majority ( sc@xmath25 ) : _ randomly pick up a literal among unit clauses if any ; otherwise randomly pick up an unset variable @xmath26 , count the numbers of occurences @xmath27 of @xmath26 , @xmath28 in 3-clauses , and choose @xmath26 ( respectively @xmath28 ) if @xmath29 ( resp .
@xmath30 ) . when @xmath31 , @xmath26 and @xmath28 are equally likely to be chosen .
rigorous mathematical analysis , undertaken to provide rigorous bounds to the critical threshold @xmath32 , have so far been restricted to the action of dpll prior to any backtracking , that is , to the first descent of the algorithm in the search tree .
the corresponding search branch is drawn on figure [ trees]a .
these studies rely on the two following facts : first , the representative point of the instance treated by dpll does not `` leave '' the 2+p - sat phase diagram .
in other words , the instance is , at any stage of the search process , uniformly distributed from the 2+p - sat distribution conditioned to its clause per variable ratio @xmath1 and fraction of 3-clauses @xmath2 .
this assumption is not true for all heuristics of split , but holds for the above examples ( @xmath33 , @xmath34 , @xmath35 ) @xcite .
analysis of more sophisticated heuristics require to handle more complex instance distributions @xcite .
secondly , the trajectory followed by an instance in the course of resolution is a stochastic object , due to the randomness of the instance and of the assignments done by dpll . in the large size limit ( @xmath36 ) , this trajectory gets concentrated around its average locus in the 2+p - sat phase diagram .
this concentration phenomenon results from general properties of markov chains @xcite .
let us briefly recall @xcite analysis of the average trajectory corresponding to the action of dpll prior to backtracking .
the ratio of clauses per variable of the 3-sat instance to be solved will be denoted by @xmath37 .
the numbers of 2 and 3-clauses are initially equal to @xmath38 respectively . under the action of dpll , @xmath39 and
@xmath40 follow a markovian stochastic evolution process , as the depth @xmath41 along the branch ( number of assigned variables ) increases .
both @xmath39 and @xmath40 are concentrated around their expectation values , the densities @xmath42 $ ] ( @xmath43 ) of which obey a set of coupled ordinary differential equations ( ode ) @xcite , @xmath44 where @xmath45 is the probability that dpll fixes a variable at depth @xmath46 ( fraction of assigned variables ) through unit - propagation .
function @xmath47 depends upon the heuristic : @xmath48 , @xmath49 ( if @xmath50 ; for @xmath51 , see @xcite ) , @xmath52 where @xmath53 and @xmath54 is the @xmath55 modified bessel function . to obtain the single branch trajectory in the phase diagram of figure [ diag ]
, we solve odes ( [ ode ] ) with initial conditions @xmath56 , and perform the change of variables @xmath57 results are shown for the guc heuristics and starting ratios @xmath58 and 2.8 in figure [ diag ] .
the trajectory , indicated by a light dashed line , first heads to the left and then reverses to the right until reaching a point on the 3-sat axis at a small ratio .
further action of dpll leads to a rapid elimination of the remaining clauses and the trajectory ends up at the right lower corner s , where a solution is found .
@xcite have shown that , for ratios @xmath59 ( for the guc heuristics ) , the full search tree essentially reduces to a single branch , and is thus entirely described by the odes ( [ ode ] ) .
the amount of backtracking necessary to reach a solution is bounded from above by a power of @xmath60 .
the average size of the branch , @xmath7 , scales linearly with @xmath3 with a multiplicative factor @xmath61 that can be calculated @xcite .
the boundary @xmath11 of this easy sat region can be defined as the largest initial ratio @xmath37 such that the branch trajectory @xmath62 issued from @xmath63 never leaves the sat phase during dpll action . in other words ,
the instance essentially keeps being sat throughout the resolution process .
we shall see in section 4 this does not hold for sat instances with ratios @xmath64 .
in this section , we present an analysis of search trees corresponding to unsat instances , that is , in presence of massive backtracking .
we first report results from numerical experiments , then expose our analytical approach to compute the complexity of resolution ( size of search tree ) . for ratios above threshold ( @xmath65 ) ,
instances almost never have a solution but a considerable amount of backtracking is necessary before proving that clauses are incompatible .
figure [ trees]b shows a generic unsat , or refutation , tree .
in contrast to the previous section , the sequence of points @xmath66 attached to the nodes of the search tree do not arrange along a line any longer , but rather form a cloud with a finite extension in the phase diagram of figure [ diag ] .
examples of clouds are provided on figure [ patch ] .
1.5 cm @xmath67 .3 cm the number of points in a cloud _ i.e. _ the size @xmath7 of its associated search tree grows exponentially with @xmath3 @xcite .
it is thus convenient to define its logarithm @xmath6 through @xmath68 .
we experimentally measured @xmath7 , and averaged its logarithm @xmath6 over a large number of instances .
results have then be extrapolated to the @xmath69 limit @xcite and are reported in table 1 .
@xmath6 is a decreasing function of @xmath37 @xcite : the larger @xmath70 , the larger the number of clauses affected by a split , and the earlier a contradiction is detected
. we will use the vocable `` branch '' to denote a path in the refutation tree which joins the top node ( root ) to a contradiction ( leaf ) .
the number of branches , @xmath71 , is related to the number of nodes , @xmath7 , through the relation @xmath72 valid for any complete binary tree . as far as exponential ( in @xmath3 ) scalings are concerned , the logarithm of @xmath71 ( divided by @xmath3 ) equals @xmath6 . in the following paragraph
, we show how @xmath71 can be estimated through the use of a matrix formalism .
1.5 cm denotes the depth in the tree , that is the number of variables assigned by dpll along each branch . at depth @xmath41 , one literal is chosen on each branch among 1-clauses ( unit - propagation , grey circles not represented on figure 2 ) , or 2,3-clauses ( splitting , black circles as in figure 2 ) .
if a contradiction occurs as a result of unit - propagation , the branch gets marked with c and dies out .
the growth of the tree proceeds until all branches carry c leaves .
the resulting tree is identical to the one built through the usual , sequential operation of dpll .
, title="fig : " ] .5 cm the probabilistic analysis of dpll in the unsat regime appears to be a formidable task since the search tree of figure [ trees]b is the output of a complex , sequential process : nodes and edges are added by dpll through successive descents and backtrackings ( depth - first search ) .
we have imagined a different building up of the refutation tree , which results in the same complete tree but can be mathematically analyzed . in our imaginary process ( figure [ struct ] ) , the tree grows in parallel , layer after layer ( breadth - first search ) . at time @xmath73
, the tree reduces to a root node , to which is attached the 3-sat instance to be solved , and an attached outgoing edge . at time @xmath41 , that is , after having assigned @xmath41 variables in the instance attached to each branch , the tree is made of @xmath74 branches , each one carrying a partial assignment of variables . at next time step @xmath75
, a new layer is added by assigning , according to dpll heuristic , one more variable along every branch . as a result
, a branch may keep growing through unitary propagation , get hit by a contradiction and die out , or split if the partial assignment does not induce unit clauses .
this parallel growth process is markovian , and can be encoded in an instance dependent matrix we now construct .
to do so , we need some preliminary definitions : partial state of variables .
the partial state @xmath76 of a boolean variable @xmath77 is one of the three following possibilities : undetermined ( @xmath78 ) if the variable has not been assigned by the search heuristic yet , true ( @xmath46 ) if the variable is partially assigned to true , false ( @xmath79 ) if the variable is partially assigned to false .
the partial state @xmath80 of a set of boolean variables @xmath81 is the collection of the states of its elements , @xmath82 .
let @xmath83 be an instance of the sat problem , defined over a set of boolean variables @xmath84 with partial state @xmath80 .
a clause of @xmath83 is said to be * satisfied if at least one of its literals is true according to @xmath80 ; * unsatisfied , or violated if all its literals are false according to @xmath80 ; * undetermined otherwise ; then its ` type ' is the number @xmath85 of undetermined variables it includes .
the instance @xmath83 is said to be satisfied if all its clauses are satisfied , unsatisfied if one ( at least ) of its clauses is violated , undetermined otherwise .
the set of partial states that violate @xmath86 is denoted by @xmath87 .
vector space attached to a variable .
to each boolean variable @xmath77 is associated a three dimensional vector space @xmath88 with spanning basis @xmath89 , @xmath90 , @xmath91 , orthonormal with respect to the dot ( inner ) product denoted by @xmath92 , @xmath93 the partial state attached to a basis vector @xmath94 is @xmath76 ( = @xmath95 .
letters @xmath78 , @xmath46 , @xmath79 stand for the different partial states the variable may acquire in the course of the search process .
note that the coefficients of the decomposition of any vector @xmath96 over the spanning basis , @xmath97 can be obtained through use of the dot product : @xmath98 with @xmath99 . by extension
, @xmath100 denotes the transposed of vector @xmath101 .
vector space attached to a set of variables .
we associate to the set @xmath81 of @xmath3 boolean variables the @xmath102dimensional vector space @xmath103 .
the spanning basis of @xmath104 is the tensor product of the spanning basis of the @xmath105 s . to lighten notations ,
we shall write @xmath106 for @xmath107 . the partial state attached to a basis vector @xmath108 is @xmath109 .
the dot product naturally extends over @xmath104 : @xmath110 if @xmath111 , 0 otherwise .
any element @xmath112 can be uniquely decomposed as a linear combination of vectors from the spanning basis .
two examples of vectors are @xmath113 and @xmath114 , respectively the sum of all vectors in the spanning basis and the fully undetermined vector , @xmath115 where @xmath116 is the identity operator on @xmath104 . to establish identity ( [ closure ] ) , apply the left hand side operator to any vector @xmath117 and take advantage of the orthonormality of the spanning basis .
( heuristic - induced ) transition probabilities let @xmath118 be a partial state which does not violate instance @xmath86 .
call @xmath119 , with @xmath120 and @xmath121 , the partial state obtained from @xmath80 by replacing @xmath122 with @xmath77 .
the probability that the heuristic under consideration ( uc , guc , ... ) chooses to assign variable @xmath123 when presented partial state @xmath80 is denoted by @xmath124 .
the probability that the heuristic under consideration then fixes variable @xmath123 to @xmath125 is denoted by @xmath126 .
a few elementary facts about transition probabilities are : 1 .
@xmath127 if @xmath128 .
2 . @xmath129 .
3 . assume that the number @xmath130 of undetermined clauses of type 1 ( unit clauses ) is larger or equal to unity .
call @xmath131 the number of unit clauses containing variable @xmath123 , and @xmath132 the number of unit clauses satisfied if @xmath123 equals @xmath133 .
clearly @xmath134 . then , as a result of unit propagation , @xmath135 4 .
in the absence of unitary clause ( @xmath136 ) , transition probabilities depend on the details of the heuristic .
for instance , in the case of the uc heuristic , 1 .
if @xmath137 , @xmath138 and @xmath139 , 2 .
if @xmath128 , @xmath127 , + where @xmath140 is the number of undetermined variables in partial state @xmath80 .
the sum of transition probabilities from a partial state @xmath80 is equal to unity , @xmath141 = 1 \quad .\ ] ] it is important to stress that the definition of the transition probabilities does not make any reference to any type of backtracking .
it relies on the notion of variable assignement through the heuristic of search only .
let us now introduce the ( heuristic - induced ) evolution operator .
the evolution operator is a linear operator @xmath142 acting on @xmath143 encoding the action of dpll for a given unsatisfiable instance @xmath86 .
its matrix elements in the spanning basis are 1 . if @xmath80 violates @xmath83 , @xmath144 2 .
if @xmath80 does not violate @xmath83 , @xmath145 where @xmath146 are the attached partial states to @xmath147 , and @xmath148 is the number of undetermined clauses of type 1 ( unitary clauses ) for partial state @xmath80 .
notice that we use the same notation , @xmath142 , for the operator and its matrix in the spanning basis .
the different cases encountered in the above definition of @xmath142 are symbolized in figure [ thm ] .
we may now conclude : branch function and average size of refutation tree call branch function the function @xmath71 with integer - valued argument @xmath41 , @xmath149 where @xmath142 is the evolution operator associated to the unsatisfiable instance @xmath83 , @xmath150 denotes the @xmath151 ( matricial ) power of @xmath152 , and vectors @xmath153 are defined in ( [ sigvec],[uvec ] ) .
then , there exist two instance
dependent integers @xmath154 and @xmath155 such that , @xmath156 furthermore , @xmath157 is the expectation value over the random assignments of variables of the size ( number of leaves ) of the search tree produced by dpll to refute @xmath86 .
the smallest non zero @xmath158 for which ( [ statq ] ) holds is the largest number of variables that the heuristic needs to assign to reach a contradiction .
proof of theorem 6 let @xmath80 be a partial state .
we call refutation tree built from @xmath80 a complete search tree that proves the unsatisfiability of @xmath86 conditioned to the fact that dpll is allowed to assign only variables which are undetermined in @xmath80 .
the height of the search tree is the maximal number of assignments leading from the root node ( attached to partial state @xmath80 ) to a contradictory leaf .
let @xmath41 be a positive integer .
we call @xmath159 the average size ( number of leaves ) of refutation trees of height @xmath160 that can be built from partial state @xmath80 . clearly , @xmath161 for all @xmath162 , and @xmath163 if @xmath164 .
recall @xmath87 is the set of violating partial states from definition 1 .
assume now @xmath41 is an integer larger or equal to 1 , @xmath80 a partial state with @xmath148 unitary clauses .
our parallel representation of dpll allows us to write simple recursion relations : 1 . if @xmath165 , @xmath166 .
2 . if @xmath167 and @xmath168 , @xmath169 3 .
if @xmath167 and @xmath170 , @xmath171 \quad .\ ] ] these three different cases are symbolized on figure [ thm]a , b and c respectively . from definitions ( [ defmu],[defms ] ) , these recursion relations are equivalent to @xmath172 for any partial state @xmath80 .
let @xmath173 be the vector of @xmath143 whose coefficients on the spanning basis @xmath174 are the @xmath175 s . in particular , @xmath176 then identity ( [ transpo ] ) can be written as @xmath177 where @xmath178 is the transposed of the evolution operator .
note that the branch function ( [ funcq ] ) is simply @xmath179 .
we deduce @xmath180 where the second sum runs over all @xmath181 sequences @xmath182 of @xmath41 partial states with associated weight @xmath183 the length of a sequence is the number of partial states it includes .
we call @xmath184genuine a sequence of partial states @xmath185 with non zero weight ( [ weightseq ] )
. the second sum on the right hand side of equation ( [ funcqprim ] ) may be rewritten as a sum over all @xmath184genuine sequences @xmath185 of length @xmath41 only .
take @xmath186 .
any @xmath184-genuine sequence @xmath187 of length @xmath188 includes at least one partial state belonging to @xmath87 .
suppose this is not true .
there exists a genuine sequence @xmath189 with @xmath190 , @xmath191 .
call @xmath192 the number of undetermined variables in partial state @xmath193 . since the sequence is genuine , @xmath194 for every @xmath41 comprised between 1 and @xmath188 . from the evolution operator definition ( [ defms ] ) , @xmath193 contains exactly one more undetermined variable than @xmath195 , and @xmath196 for all @xmath197 .
hence @xmath198 . but @xmath199 and @xmath200 are , by definition , integer numbers comprised between 0 and @xmath3 . from lemma 7 , the index @xmath201 of a @xmath184genuine sequence @xmath187 of length @xmath188 , @xmath202 exists and is larger , or equal , to @xmath203 .
let us define @xmath204 from definition ( [ defmu ] ) , @xmath187 is simply @xmath184 repeated @xmath201 times followed by @xmath205 , and @xmath206 .
call @xmath207 the smallest index of all @xmath184-genuine sequences of length @xmath188 , and @xmath208 the minimum of @xmath209 over @xmath186 . then , from equation ( [ funcqprim ] ) , @xmath210 where @xmath211 .
thus @xmath212 is a right eigenvector of @xmath178 with eigenvalue unity , and @xmath213 for all @xmath214 .
@xmath158 , which depends upon instance @xmath86 , is the length of the longest genuine sequence without repetition .
it is the maximal number of ( undetermined ) variables to be fixed before a contradiction is found .
take @xmath164 .
then there is no @xmath80-genuine sequence of length @xmath158 .
suppose this is not true .
there exist @xmath164 and a @xmath80genuine sequence @xmath215 of length @xmath158 .
as @xmath80 does not violate @xmath86 , and @xmath86 is not satisfiable , there are still some undetermined variables in partial state @xmath80 .
a certain number of them , say @xmath216 , must be assigned to some @xmath217 values to reach a contradiction , that is , a partial state @xmath186 .
therefore there exists a @xmath184genuine sequence , @xmath218 , of length @xmath216 ending with @xmath80 and with no repeated partial state .
concatenating @xmath218 and @xmath215 , we obtain a @xmath184genuine sequence of length @xmath219 and without repetition , in contradiction with the above result . using lemma 8
, we may replace @xmath220 in equation ( [ funcqprim ] ) with @xmath113 , and find @xmath221 for all @xmath214 . hence
, @xmath222 is the average size ( over the random assignments made by the heuristic ) of the refutation tree to instance @xmath83 generated from the fully undetermined partial state . .
* a*. if the partial state @xmath80 already violates the instance @xmath86 , it is left unchanged . *
b*. if the partial state does not violate @xmath83 and there is at least one unitary clause , a variable is fixed through unit propagation ( grey node ) e.g. @xmath223 .
the output partial state is @xmath224 . *
c*. if the partial state does not violate @xmath83 and there is no unitary clause , a variable @xmath123 is fixed through splitting ( black node ) .
two partial states are generated , @xmath225 and @xmath226.,title="fig : " ] .5 cm we illustrate the above definitions and results with three explicit examples of instances involving few variables : instance over @xmath227 variable consider the following unsat instance built from a single variable , @xmath228 the 3dimensional vector space @xmath229 is spanned by vectors latexmath:[$|u\rangle , |t\rangle , latexmath:[\[{\bf h } = \left ( \begin{array}{c c c } 0 & 0 & 0 \\ \frac 12 & 1 & 0 \\ \frac 12 & 0 & 1 \end{array }
\right ) \ \hbox{\rm with } \quad entries can be interpreted as follows .
starting from the @xmath78 state , variable @xmath231 will be set through unit - propagation to @xmath46 or @xmath79 with equal probabilities : @xmath232 . once the variable has reached this state , the instance
is violated : @xmath233 .
all other entries are null . in particular ,
state @xmath78 can never be reached from any state , so the first line of the matrix is filled in with zeroes : @xmath234 .
function ( [ funcq ] ) is easily calculated @xmath235 therefore , @xmath236 .
indeed , refutation is obtained without any split , and the search tree involves a unique branch of length 1 ( figure [ tree - exemple]a ) . our next example is a 2-sat instance whose refutation requires to split one variable .
instance over @xmath237 variables , with a unique refutation tree .
@xmath238 the evolution matrix @xmath142 is a @xmath239 matrix with 16 non zero entries , @xmath240 we now explain how these matrix elements were obtained . from the undetermined state @xmath241 ,
any of the four clause can be chosen by the heuristic .
thus , any of the two literals @xmath231 , @xmath242 has a probability @xmath243 to be chosen : @xmath244 .
next , unit - propagation will set the unassigned variable to true , or false with equal probabilities @xmath243 ( [ m22 ] ) .
finally , entries corresponding to violating states in eqn ( [ m23 ] ) are calculated according to rule ( [ defmu ] ) .
the branch function @xmath245 equals 1 for @xmath73 , 2 for any @xmath246 ; thus , @xmath247 and @xmath248 , in agreement with the associated search tree symbolized in figure [ tree - exemple]b .
we now introduce an instance with a non unique refutation tree .
instance with @xmath249 variables , and two refutation trees .
@xmath250 notice the presence of a ( trivial ) clause containing opposite literals , which allows us to obtain a variety in the search trees without considering more than three variables .
the evolution matrix @xmath142 is a @xmath251 matrix with 56 non zero entries ( for the guc heuristic ) , @xmath252 the first split variable is @xmath253 if the last clause is chosen ( probability @xmath254 ) , or @xmath231 or @xmath242 otherwise ( with probability @xmath255 each ) , leading to expressions ( [ m31 ] ) and ( [ m32 ] ) .
the remaining entries of @xmath142 are obtained in the same way as explained in example 10 .
we obtain @xmath256 , @xmath257 and @xmath258 .
therefore , @xmath259 and @xmath260 where the different contributions to @xmath157 and their probabilities are explicitely written down , see figures [ tree - exemple]b and [ tree - exemple]c . ,
@xmath261 and @xmath262 .
grey and black nodes correspond to variables assigned through unit - propagation and split respectively , as in figure [ struct ] .
* a*. example 9 : refutation of instance @xmath263 is obtained as a result of unit - propagation .
the size ( number of leaves ) of the search tree is @xmath264 . * b*. example 10 : search tree generated by dpll on instance @xmath261 .
the black and grey node correspond to the split of @xmath231 and unit - propagation over @xmath242 , or vice - versa .
the size of the tree is @xmath265 . * c*. example 11 : search tree corresponding to the instance @xmath262 when dpll first splits variable @xmath253 .
the size of the tree is @xmath266 .
if the first split variable is @xmath231 or @xmath242 , the refutation search tree of instance @xmath262 corresponds to case * b*. , title="fig : " ] .5 cm let us denotes by @xmath267 the expectation value of a function @xmath268 of the instance @xmath86 over the random 3-sat distribution , at given numbers of variable , @xmath3 , and clauses , @xmath269 . from theorem 6 , the expectation value of the size of the refutation tree is @xmath270 calculation of the expectation value of the @xmath271 power of @xmath152 is a hard task that we were unable to perform for large sizes @xmath3 .
we therefore turned to a simplifying approximation , hereafter called dynamical annealing .
this approximation is not thought to be justified in general , but may be asymptotically exact in some limiting cases we will expose later on . a first temptation is to approximate the expectation of the @xmath271 power of @xmath152 with the @xmath271 power of the expectation of @xmath152 .
this is however too a brutal approximation to be meaningful , and a more refined scheme is needed .
clause projection operator consider an instance @xmath86 of the 3-sat problem .
the clause vector @xmath272 of a partial state @xmath80 is a three dimensional vector @xmath273 where @xmath274 is the number of undetermined clauses of @xmath86 of type @xmath275 .
the clause projection operator , @xmath276 , is the operator acting on @xmath143 and projecting onto the subspace of partial state vectors with clause vectors @xmath277 , @xmath278
\ function .
the sum of all state vectors in the spanning basis with clause vector @xmath277 is denoted by @xmath279 .
the sum of all state vectors in the spanning basis with clause vector @xmath277 and @xmath280 undetermined variables is denoted by @xmath281 .
it is an easy check that @xmath282 is indeed a projection operator : @xmath283 .
as the set of partial states can be partitioned according to their clause vectors , @xmath284 we now introduce the clause vector - dependent branch function @xmath285 summation of the @xmath71 s over all @xmath277 gives back function ( [ funcq ] ) from identity ( [ identriv ] ) .
the evolution equation for @xmath286 is , @xmath287 where we have made use of identities ( [ closure ] ) and ( [ identriv ] ) .
we are now ready to do the two following approximation steps : dynamical annealing ( step a ) substitute in equation ( [ ap2 ] ) the partial state vector @xmath288 and @xmath289 undetermined variables .
following step a , equation ( [ ap2 ] ) becomes an approximated evolution equation for @xmath71 , @xmath290 \ ; b ( \vec c , t ) \quad , \ ] ] where the new evolution matrix @xmath291 , not to be confused with @xmath152 , is @xmath292 = \frac { \langle \sigma ( \vec c)| { \bf h } | \sigma _ { n - t}(\vec c ' ) \rangle } { \langle \sigma | \sigma _ { n - t}(\vec c ' ) \rangle } \quad .\ ] ] then , dynamical annealing ( step b ) substitute in equation ( [ ap3 ] ) the evolution matrix @xmath291 with @xmath293 = \frac { \overline{\langle \sigma ( \vec c)| { \bf h } | \sigma _ { n - t}(\vec c ' ) \rangle}}{\overline{\langle \sigma | \sigma _ { n - t}(\vec c ' ) \rangle } } \ ] ] that is , consider the instance @xmath86 is redrawn at each time step @xmath75 , keeping information about clause vectors at time @xmath41 only .
let us interpret what we have done so far .
the quantity we focus on is @xmath294 , the expectation number of branches at depth @xmath41 in the search tree ( figure [ struct ] ) carrying partial states with clause vector @xmath295 . within the dynamical annealing approximation , the evolution of the @xmath296 s is markovian , @xmath297\ ; \bar b(\vec c';t ) \ .\ ] ] the entries of the evolution matrix @xmath298 $ ] can be interpreted as the average number of branches with clause vector @xmath277 that dpll will generate through the assignment of one variable from a partial assignment ( partial state ) of variables with clause vector @xmath299 . for the guc heuristic
, we find @xcite , @xmath300 = { c_3 ' \choose c_3'-c_3 } \ ; \left ( \frac 3{n - t}\right)^{c_3'-c_3 } \ ; \left(1-\frac 3{n - t}\right)^{c_3 } \times \nonumber \\ & & \qquad \qquad \qquad \qquad \sum_{w_2=0}^{c_3'-c_3 } \left ( \frac 1 2 \right)^{c_3'-c_3 } { c_3'-c_3\choose w_2}\times \nonumber \\ & & \left \{(1 - \delta _ { c'_1 } ) \ ; \left ( 1 - \frac1{2(n - t ) } \right)^{c'_1 - 1 } \sum_{z_2=0}^{c_2 ' } { c_2 ' \choose z_2 } \left ( \frac 2{n - t}\right)^{z_2 } \right .
\times \nonumber \\ & & \left ( 1- \frac 2{n - t}\right)^{c_2'-z_2 } \sum_{w_1=0}^{z_2 } \left ( \frac 1 2 \right)^{z_2 } { z_2\choose w_1}\ ; \delta_{c_2-c_2'-w_2+z_2 } \;\delta_{c_1-c_1'-w_1 + 1 } + \nonumber \\ & & \delta_{c_1 ' } \sum_{z_2=0}^{c_2'-1 } { c_2'-1 \choose z_2 } \left ( \frac 2{n - t}\right)^{z_2}\ , \left ( 1- \frac 2{n - t}\right)^{c_2'-1-z_2 } \times \nonumber \\ & & \left .
\sum_{w_1=0}^{z_2 } \left(\frac 1 2 \right)^{z_2 } { z_2\choose w_1}\ ; \delta_{c_2-c_2'-w_2+z_2 + 1 } \ ; [ \delta_{c_1-w_1 } + \delta_{c_1 - 1-w_1 } ] \right\ } \ , \end{aligned}\ ] ] where @xmath301 denotes the kronecker delta function over integers @xmath84 : @xmath302 if @xmath303 , @xmath304 otherwise .
expression ( [ bbra ] ) is easy to obtain from the interpretation following equation ( [ bradp ] ) .
let us introduce the generating function @xmath305 of the average number of branches @xmath306 where @xmath307 , through @xmath308 evolution equation ( [ ap3 ] ) for the @xmath296 s can be rewritten in term of the generating function @xmath309 , @xmath310 where @xmath311 is a vectorial function of argument @xmath312 whose components read @xmath313 \quad , \nonumber\\ \gamma_2({\vec y})&=&y_2+\ln\left[1+\frac 2 { n - t } \left ( \frac{e^{-y_2}}{2 } \left(1 + e^{y_1}\right ) -1\right)\right ] \quad , \nonumber \\ \gamma_3 ( { \vec y})&=&y_3+\ln\left[1+\frac 3 { n - t } \left ( \frac{e^{-y_3}}{2 } \left(1 + e^{y_2}\right ) -1\right)\right ] \quad .\end{aligned}\ ] ] to solve equation ( [ eqev ] ) , we infer the large @xmath3 behaviour of @xmath309 from the following remarks : 1 . each time dpll assigns variables through splitting or unit - propagation , the numbers @xmath274 of clauses of length @xmath275 undergo @xmath314 changes .
it is thus sensible to assume that , when the number of assigned variables increases from @xmath315 to @xmath316 with @xmath317 very large but @xmath318 e.g. @xmath319 , the densities @xmath320 and @xmath321 of 2- and 3-clauses have been modified by @xmath322 .
2 . on the same time interval @xmath323
, we expect the number of unit - clauses @xmath324 to vary at each time step .
but its distribution @xmath325 , conditioned to the densities @xmath326 , @xmath327 and the reduced time @xmath46 , should reach some well defined limit distribution .
this claim is a generalization of the result obtained by @xcite for the analysis of the guc heuristic in the absence of backtracking .
as long as a partial state does not violate the instance , very few unit - clauses are generated , and splitting frequently occurs . in other words , the probability that @xmath328 is strictly positive as @xmath3 gets large .
the above arguments entice us to make the following asymptotic expression for the generating function @xmath309 for large @xmath329 at fixed ratio @xmath330 , the generating function ( [ gener ] ) of the average numbers @xmath296 of branches is expected to behave as @xmath331 \quad .\ ] ] hypothesis ( [ scalinghyp2 ] ) expresses in a concise way some important information on the distribution of clause populations during the search process that we now extract .
call @xmath6 the legendre transform of @xmath332 , @xmath333 \qquad .
\label{inversion}\ ] ] then , combining equations ( [ gener ] ) , ( [ scalinghyp2 ] ) and ( [ inversion ] ) , we obtain @xmath334 \quad , \ ] ] up to non exponential in @xmath3 corrections . in other words ,
the expectation value of the number of branches carrying partial states with @xmath335 undetermined variables and @xmath336 @xmath275-clauses ( @xmath43 ) scales exponentially with @xmath3 , with a growth function @xmath337 related to @xmath338 through identity ( [ inversion ] ) .
moreover , @xmath339 is the logarithm of the number of branches ( divided by @xmath3 ) after a fraction @xmath46 of variables have been assigned .
the most probable values of the densities @xmath340 of @xmath275-clauses are then obtained from the partial derivatives of @xmath332 : @xmath341 for @xmath43 .
let us emphasize that @xmath332 in equation ( [ scalinghyp2 ] ) does not depend on @xmath342 .
this hypothesis simply expresses that , as far as non violating partial states are concerned , both terms on the right hand side of ( [ eqev ] ) are of the same order , and that the density of unit - clauses , @xmath343 , identically vanishes . similarly , function @xmath344 is related to the generating function of distribution @xmath325 , @xmath345 where @xmath346 ( @xmath43 ) on the left hand side of the above formula .
inserting expression ( [ scalinghyp2 ] ) into the evolution equation ( [ eqev ] ) , we find @xmath347 \frac{\partial \varphi } { \partial y_2 } ( y_2,y_3;t ) \nonumber \\ & + & \frac 3{1-t } \left [ e^{-y_3 } \left ( \frac{1+e^{y_2}}2 \right ) -1 \right ] \frac{\partial \varphi } { \partial y_3 } ( y_2,y_3;t ) \nonumber \\ & + & \ln \left [ 1 + k(y_1,y_2 ) \ ; e^{\psi ( -\infty , y_2,y_3;t ) - \psi(y_1,y_2,y_3;t ) } \right]\end{aligned}\ ] ] where @xmath348 .
as @xmath332 does not depend upon @xmath349 , the latter may be chosen at our convenience e.g. to cancel @xmath350 and the contribution from the last term in equation ( [ mdar ] ) , @xmath351 such a procedure , sometimes called kernel method and , to our knowledge , first proposed by @xcite , is correct in the major part of the @xmath352 space and , in particular , in the vicinity of @xmath353 we focus on in this paper space ; a complete analysis of this case was carried out by @xcite . ] .
we end up with the following partial differential equation ( pde ) for @xmath332 , @xmath354 \quad , \ ] ] where @xmath355 incorporates the details of the splitting heuristic+ \frac{c_2}{1-t } \ ; \left ( \frac 32 e^{-y_2 } -2 \right ) \qquad .\ ] ] ] , @xmath356 & = & -y_1(y_2 ) + \frac { 3\ , c_3}{1-t}\ ; \left [ e^{-y_3}\;\left ( \frac{1+e^{y_2}}{2}\right ) -1 \right]\nonumber \\ & + & \frac{c_2}{1-t } \ ; \left ( e^ { -y_1(y_2 ) } -2 \right ) \qquad .\end{aligned}\ ] ] we must therefore solve the partial differential equation ( pde ) ( [ croi2 ] ) with the initial condition , @xmath357 obtained through inverse legendre transform ( [ inversion ] ) of the initial condition over @xmath296 , or equivalently over @xmath6 , @xmath358 @xmath359 .5 cm we can interpret the dynamical annealing approximation made in the previous paragraphs , and the resulting pde ( [ croi2 ] ) as a description of the growth process of the search tree resulting from dpll operation .
using legendre transform ( [ inversion ] ) , pde ( [ croi2 ] ) can be written as an evolution equation for the logarithm @xmath360 of the average number of branches with parameters @xmath361 as the depth @xmath330 increases , @xmath362 \qquad .
\label{croi}\ ] ] partial differential equation ( pde ) ( [ croi ] ) is analogous to growth processes encountered in statistical physics @xcite .
the surface @xmath6 , growing with `` time '' @xmath46 above the plane @xmath361 , or equivalently from ( [ change ] ) , above the plane @xmath363 ( figure [ dome ] ) , describes the whole distribution of branches .
the average number of branches at depth @xmath46 in the tree equals @xmath364 where @xmath365 is the maximum over @xmath363 of @xmath366 reached in @xmath367 . in other words ,
the exponentially dominant contribution to @xmath368 comes from branches carrying 2+p - sat instances with parameters @xmath367 , that is clause densities @xmath369 , @xmath370 .
parametric plot of @xmath371 as a function of @xmath46 defines the tree trajectories on figure [ diag ] .
the hyperbolic line in figure [ diag ] indicates the halt points , where contradictions prevent dominant branches from further growing .
each time dpll assigns a variable through unit - propagation , an average number @xmath372 of new 1-clauses is produced , resulting in a net rate of @xmath373 additional 1-clauses .
as long as @xmath374 , 1-clauses are quickly eliminated and do not accumulate .
conversely , if @xmath375 , 1-clauses tend to accumulate . opposite 1-clauses @xmath77 and @xmath376 are likely to appear , leading to a contradiction @xcite .
the halt line is defined through @xmath377 , and reads @xcite , @xmath378\;\frac 1{1-p } \qquad .\ ] ] it differs from the halt line @xmath379 corresponding to a single branch @xcite .
as far as dominant branches are concerned , an alternative and simpler way of obtaining the halt criterion is through calculation of the probability @xmath380 that a split occurs when a variable is assigned by dpll , @xmath381 from equations ( [ psigen],[mdar ] ) .
the probability of split vanishes , and unit - clauses accumulate till a contradiction is obtained , when the tree stops growing . along the tree trajectory
, @xmath382 grows thus from 0 , on the right vertical axis , up to some final positive value , @xmath383 , on the halt line .
@xmath384 is our theoretical prediction for the logarithm of the complexity ( divided by @xmath3 ) to match the definition used for numerical experiments ; this is done in table 1 ] .
equation ( [ croi ] ) was solved using the method of characteristics .
using eqn .
( [ change ] ) , we have plotted the surface @xmath6 at different times , with the results shown in figure [ dome ] for @xmath385 .
values of @xmath384 , obtained for @xmath386 by solving equation ( [ croi ] ) compare very well with numerical results ( table 1 ) .
we stress that , though our calculation is not rigorous , it provides a very good quantitative estimate of the complexity .
it is therefore expected that our dynamical annealing approximation be quantitavely accurate .
it is a reasonable conjecture that it becomes exact at large ratios @xmath387 , where pde ( [ croi2 ] ) can be exactly solved : asymptotic equivalent of @xmath6 for large ratios resolution of pde ( [ croi ] ) in the large ratio @xmath387 limit gives ( for the guc heuristic ) , @xmath388 ^ 2 \ ; \frac 1{\alpha_0 } \quad .\ ] ] this result exhibits the @xmath389 scaling proven by @xcite , and is conjectured to be exact . as @xmath37 increases , search trees become smaller and smaller , and correlations between branches , weaker and weaker , making dynamical annealing more and more accurate . ) .
dpll starts with a satisfiable 3-sat instance and transforms it into a sequence of 2+p - sat instances .
the leftmost branch in the tree symbolizes the first descent made by dpll . above node
@xmath390 , instances are satisfiable while below @xmath391 , instances have no solutions . a grey triangle accounts for the ( exponentially ) large refutation subtree that dpll has to go through before backtracking above @xmath391 and reaching @xmath390 .
by definition , the highest node reached back by dpll is @xmath390 .
further backtracking , below @xmath390 , will be necessary but a solution will be eventually found ( right subtree ) , see figure [ trees]c.,title="fig : " ] .3 cm
the interest of the trajectory framework proposed in this paper is best seen in the upper sat phase , that is , for ratios @xmath37 ranging from @xmath11 to @xmath32 .
this intermediate region juxtaposes branch and tree behaviors , see search tree in figures [ trees]c and [ treeinter ] .
the branch trajectory , started from the point @xmath392 corresponding to the initial 3-sat instance , hits the critical line @xmath393 at some point g with coordinates ( @xmath394 ) after @xmath395 variables have been assigned by dpll , see figure [ diag ] .
the algorithm then enters the unsat phase and , with high probability , generates a 2+p - sat instance with no solution . a dense subtree that dpll has to go through entirely , forms beyond g till the halt line ( left subtree in figure [ treeinter ] ) .
the size of this subtree can be analytically predicted from the theory exposed in section 3 .
all calculations are identical , except initial condition ( [ initphi ] ) which has to be changed into @xmath396 as a result we obtain the size @xmath397 of the unsatisfiable subtree to be backtracked ( leftmost subtree in figure [ treeinter ] ) .
@xmath398 denotes the number of undetermined variables at point @xmath309 .
@xmath309 is the highest backtracking node in the tree ( figures [ trees]c and [ treeinter ] ) reached back by dpll , since nodes above g are located in the sat phase and carry 2+p - sat instances with solutions .
dpll will eventually reach a solution .
the corresponding branch ( rightmost path in figure [ trees]c ) is highly non typical and does not contribute to the complexity , since almost all branches in the search tree are described by the tree trajectory issued from g ( figure [ diag ] ) .
we expect that the computational effort dpll requires to find a solution will , to exponential order in @xmath3 , be given by the size of the left unsatisfiable subtree of figure [ treeinter ] .
in other words , massive backtracking will certainly be present in the right subtree ( the one leading to the solution ) , and no significant statistical difference is expected between both subtrees .
we have experimentally checked this scenario for @xmath399 .
the average coordinates of the highest backtracking node , @xmath400 ) , coincide with the computed intersection of the single branch trajectory ( section 2.2 ) and the estimated critical line @xmath393 @xcite . as for complexity ,
experimental measures of @xmath6 from 3-sat instances at @xmath401 , and of @xmath402 from 2 + 0.78-sat instances at @xmath403 , obey the expected identity @xmath404 and are in very good agreement with theory ( table 1 ) .
therefore , the structure of search trees corresponding to instances of 3-sat in the upper sat regime reflects the existence of a critical line for 2+p - sat instances .
in this paper , we have exposed a procedure to understand the complexity pattern of the backtrack resolution of the random satisfiability problem ( figure [ sche ] ) .
main steps are : 1 .
identify the space of parameters in which the dynamical evolution takes place ; this space will be generally larger than the initial parameter space since the algorithm modifies the instance structure .
while the distribution of 3-sat instances is characterized by the clause per variable ratio @xmath1 only , another parameter @xmath2 accounting for the emergence of 2-clauses has to be considered .
2 . divide the parameter space into different regions ( phases ) depending on the output of the resolution e.g. sat / unsat phases for 2+p - sat .
3 . represent the action of the algorithm as trajectories in this phase diagram .
intersection of trajectories with the phase boundaries allow to distinguish hard from easy regimes ( figure [ sche ] ) .
in addition , we have also presented a non rigorous study of the search tree growth , which allows us to accurately estimate the complexity of resolution in presence of massive backtracking . from a mathematical point of view , it is worth noticing that monitoring the growth of the search tree requires a pde , while odes are sufficient to account for the evolution of a single branch @xcite .
an interesting question raised by this picture is the robustness of the polynomial / exponential crossover point t ( figure 3 ) .
while the ratio @xmath405 separating easy ( polynomial ) from hard ( exponential ) resolutions depends on the heuristics used by dpll ( @xmath406 , @xmath407 ) , t appears to be located at the same coordinates @xmath408 for all three uc , guc , and sc@xmath25 heuristics . from a technical point of view
, the robustness of t comes from the structure of the odes ( [ ode ] ) .
the coordinates of t , and the time @xmath409 at which the branch trajectory issued from @xmath410 hits the critical line @xmath411 tangentially , obey the equations @xmath412 with @xmath413 .
the set of odes ( [ ode ] ) , combined with the previous conditions , gives @xmath414 @xcite .
this robustness explains why the polynomial / exponential crossover location of critically constrained 2+p - sat instances , which should _ a priori _
depend on the algorithm used , was found by @xcite to coincide roughly with the algorithm
independent , tricritical point on the @xmath415 line .
our approach has already been extended to other decision problems , e.g. the vertex covering of random graphs @xcite or the coloring of random graphs @xcite ( see @xcite for recent rigorous results on backtracking in this case ) .
it is important to stress that it is not limited to the determination of the average solving time , but may also be used to capture its distribution @xcite and to understand the efficiency of restarts techniques @xcite .
finally , we emphasize that theorem 6 relates the computational effort to the evolution operator representing the elementary steps of the search heuristic for a _ given _ instance .
it is expected that this approach will be useful to obtain results on the average - case complexity of dpll at fixed instance , where the average is performed over the random choices done by the algorithm only @xcite .
we thank j. franco for his constant support during the completion of this work .
r. monasson was in part supported by the aci jeunes chercheurs `` algorithmes doptimisation et systmes dsordonns quantiques '' from the french ministry of research .
coarfa , c. , dernopoulos , d.d .
, san miguel aguirre , a. , subramanian , d. and vardi , m.y .
random 3-sat : the plot thickens . in r. dechter , editor , _ proc .
principles and practice of constraint programming ( cp2000 ) _ , lecture notes in computer science 1894 , 143 - 159 ( 2000 ) .
cocco , s. and monasson , r. trajectories in phase diagrams , growth processes and computational complexity : how search algorithms solve the 3-satisfiability problem , _ phys . rev .
* 86 * , 1654 ( 2001 ) ; analysis of the computational complexity of solving random satisfiability problems using branch and bound search algorithms , _ eur .
j. b _ * 22 * , 505 ( 2001 ) . cocco , s. and monasson r. exponentially hard problems are sometimes polynomial , a large deviation analysis of search algorithms for the random satisfiability problem , and its application to stop - and - restart resolutions , _ phys . rev .
e _ * 66 * , 037101 ( 2002 ) .
crawford , j. and auton , l. experimental results on the cross - over point in satisfiability problems , _ proc .
11th natl .
conference on artificial intelligence ( aaai-93 ) , _ 2127 , the aaai press / mit press , cambridge , ma ( 1993 ) ; _ artificial intelligence _ * 81 * ( 1996 ) .
gent , i. , van maaren , h. and walsh , t. ( eds ) .
sat2000 : highlights of satisfiability research in the year 2000 , _ frontiers in artificial intelligence and applications _
63 , ios press , amsterdam ( 2000 ) .
gu , j. , purdom , p.w . , franco , j. and wah , b.w .
algorithms for satisfiability ( sat ) problem : a survey .
_ dimacs series on discrete mathematics and theoretical computer science _ * 35 * , 19 - 151 , american mathematical society ( 1997 ) .
mitchell , d. , selman , b. and levesque , h. hard and easy distributions of sat problems , _ proc . of the tenth natl .
conf . on artificial intelligence ( aaai-92 )
_ , 440 - 446 , the aaai press / mit press , cambridge , ma ( 1992 ) .
monasson , r. , zecchina , r. , kirkpatrick , s. , selman , b. and troyansky , l. determining computational complexity from characteristic phase transitions. _ nature _ * 400 * , 133137 ( 1999 ) ; 2+p - sat : relation of typical - case complexity to the nature of the phase transition , _ random structure and algorithms _ * 15 * , 414 ( 1999 ) . | an analysis of the average - case complexity of solving random 3-satisfiability ( sat ) instances with backtrack algorithms is presented .
we first interpret previous rigorous works in a unifying framework based on the statistical physics notions of dynamical trajectories , phase diagram and growth process .
it is argued that , under the action of the davis putnam loveland logemann ( dpll ) algorithm , 3-sat instances are turned into @xmath0-sat instances whose characteristic parameters ( ratio @xmath1 of clauses per variable , fraction @xmath2 of 3-clauses ) can be followed during the operation , and define resolution trajectories . depending on the location of trajectories in the phase diagram of the 2+p - sat model , easy ( polynomial ) or hard (
exponential ) resolutions are generated .
three regimes are identified , depending on the ratio @xmath1 of the 3-sat instance to be solved .
lower sat phase : for small ratios , dpll almost surely finds a solution in a time growing linearly with the number @xmath3 of variables .
upper sat phase : for intermediate ratios , instances are almost surely satisfiable but finding a solution requires exponential time ( @xmath4 with @xmath5 ) with high probability .
unsat phase : for large ratios , there is almost always no solution and proofs of refutation are exponential .
an analysis of the growth of the search tree in both upper sat and unsat regimes is presented , and allows us to estimate @xmath6 as a function of @xmath1 .
this analysis is based on an exact relationship between the average size of the search tree and the powers of the evolution operator encoding the elementary steps of the search heuristic .
satisfiability , analysis of algorithms , backtrack . |
we consider only simple , finite graphs @xmath0 and denote the vertex set and edge set of @xmath0 by @xmath1 and @xmath2 , respectively .
the _ open neighbourhood _ of a vertex @xmath3 in @xmath0 is the set @xmath4 .
if @xmath5 is a nonempty subset of @xmath1 then @xmath6 denotes the subgraph of @xmath0 induced by @xmath5 .
a graph @xmath0 is _ hamiltonian _ if it has a _ hamiltonian cycle _ ( a cycle containing all the vertices of @xmath0 ) , and _
traceable _ if it has a _
hamiltonian path _
( a path containing all the vertices of @xmath0 ) .
a graph@xmath7 is _ maximal nonhamiltonian _ ( mnh ) if @xmath0 is not hamiltonian , but @xmath8 is hamiltonian for each @xmath9 @xmath10 , where @xmath11 denotes the complement of @xmath12 a graph @xmath0 is _ maximal nontraceable _ ( mnt ) if @xmath0 is not traceable , but @xmath13 is traceable for each @xmath14 a graph @xmath0 is _ hypohamiltonian _ if @xmath0 is not hamiltonian , but every vertex - deleted subgraph @xmath15 of @xmath0 is hamiltonian .
we say that a graph @xmath0 is _ maximal hypohamiltonian _
( mhh ) if it is mnh and hypohamiltonian . in 1978
bollobs @xcite posed the problem of finding the least number of edges , @xmath16 , in a mnh graph of order @xmath17 .
bondy @xcite had already shown that a mnh graph with order @xmath18 that contained @xmath19 vertices of degree @xmath20 had at least @xmath21 edges , and hence @xmath22 for @xmath18 .
combined results of clark , entringer and shapiro @xcite , @xcite and lin , jiang , zhang and yang @xcite show that @xmath23 for @xmath24 and for @xmath25 .
the values of @xmath16 for the remaining values of @xmath17 are also given in @xcite .
let @xmath26 be the minimum size of a mnt graph of order @xmath17 .
dudek , katona and wojda @xcite showed that @xmath27 for all @xmath17 and , by means of a recursive construction , they found mnt graphs of order @xmath17 and size @xmath28 . to date
, no cubic mnt graphs have been reported .
we construct an infinite family of cubic mnt graphs , thus showing that @xmath29 for infinitely many @xmath17 .
now let @xmath30 be the minimum size of a 2-connected mnt graph of order @xmath31 .
we prove that @xmath32 for @xmath18 .
it then follows from our constructions that @xmath33 for @xmath34 for @xmath35 , @xmath36 for @xmath37 , @xmath38 for @xmath39 and @xmath40 , and @xmath41 for @xmath42 .
bondy @xcite proved that if @xmath0 is a 2-connected mnh graph and @xmath43 with degree @xmath44 , then each neighbour of @xmath3 has degree at least 4 .
he also showed that the neighbours of such a vertex are in fact adjacent . in order to prove a corresponding result for 2-connected mnt graphs we need the following result .
[ subgraph]let @xmath45 be a path in a mnt graph @xmath0 .
if @xmath46 is not complete , then some internal vertex of @xmath45 has a neighbour in @xmath47 .
let @xmath48 and @xmath3 be two nonadjacent vertices of @xmath46
. then @xmath49 has a hamiltonian path @xmath50 .
let @xmath51 and @xmath52 be the two endvertices of @xmath45 and suppose no internal vertex of @xmath45 has a neighbour in @xmath47 .
then @xmath50 has a subpath @xmath53 in @xmath54 and @xmath53 has either one or both endvertices in @xmath55 .
if @xmath53 has only one endvertex in @xmath55 , then @xmath50 has an endvertex in @xmath45 . in either case
the path obtained from @xmath50 by replacing @xmath53 with @xmath45 is a hamiltonian path of @xmath0 .
[ mntdeg2]if @xmath0 is a mnt graph and @xmath56 with @xmath57 , then the neighbours of @xmath3 are adjacent .
if in addition @xmath0 is @xmath20-connected , then each neighbour of @xmath3 has degree at least @xmath58 .
let @xmath59 and let @xmath45 be the path @xmath60 . since @xmath61 , it follows from lemma [ subgraph ] that @xmath46 is a complete graph ; hence @xmath62 and @xmath63 are adjacent .
now assume that @xmath0 is @xmath20-connected . since @xmath0 is not traceable we assume @xmath64
. then also @xmath65 otherwise @xmath66 would be a cut vertex of @xmath0 .
let @xmath67 be a neighbour of @xmath62 and let @xmath45 be the path @xmath68 .
since @xmath69 the graph @xmath46 is not complete , and hence it follows from lemma [ subgraph ] that @xmath62 has a neighbour in @xmath47 .
thus @xmath70 .
similarly @xmath71 .
we also have the following two lemmas concerning mnt graphs that have vertices of degree 2 .
[ mnt1deg2]suppose @xmath0 is a @xmath20-connected mnt graph .
suppose @xmath72 such that @xmath73 and @xmath74 and @xmath75 have exactly one common neighbour @xmath51 .
then @xmath76 the vertices @xmath74 and @xmath77 can not be adjacent otherwise @xmath51 would be a cut vertex .
let @xmath78 ; @xmath79 .
it follows from lemma [ mntdeg2 ] that @xmath51 is adjacent to @xmath80 ; @xmath79 .
let @xmath45 be the path @xmath81 .
since @xmath46 is not complete , it follows from lemma [ subgraph ] that @xmath51 has a neighbour in @xmath47 .
hence @xmath82 .
[ mnt2deg2]suppose @xmath0 is a mnt graph .
suppose @xmath83 such that @xmath73 and @xmath74 and @xmath77 have the same two neighbours @xmath66 and @xmath84 .
then @xmath85 . also @xmath86 from lemma [ mntdeg2 ] it follows that @xmath66 and @xmath84 are adjacent .
let @xmath45 be the path @xmath87 .
@xmath46 is not complete since @xmath88 and @xmath89 are not adjacent .
thus it follows from lemma [ subgraph ] that @xmath62 has a neighbour in @xmath47 .
now suppose @xmath90 and @xmath91 .
then a hamiltonian path @xmath50 in @xmath92 contains a subpath of either of the forms given in the first column of table 1 .
note that @xmath93 ; @xmath94 and that @xmath95 represents a subpath of @xmath50 in @xmath96 .
if each of the subpaths is replaced by the corresponding subpath in the second column of the table we obtain a hamiltonian path @xmath97 in @xmath0 , which leads to a contradiction .
@xmath98 @xmath99 hence @xmath100 .
thus @xmath101 .
similarly @xmath102 . thus @xmath103 and hence @xmath104 .
now let @xmath45 be the path @xmath105 .
since @xmath46 is not complete , it follows from lemma [ subgraph ] that @xmath62 or @xmath63 has a neighbour in @xmath47 . hence @xmath106 .
we now consider the size of a 2-connected mnt graph .
[ mnt3deg2]suppose @xmath0 is a mnt graph of order @xmath107 and that @xmath108 and @xmath109 are vertices of degree @xmath20 in @xmath0 having the same neighbours , @xmath66 and @xmath84 .
then @xmath110 is complete and hence @xmath111 .
suppose @xmath110 is not complete
. then there exist @xmath112 which are not adjacent .
however , since @xmath113 is an independent set , no path in @xmath114 having @xmath115 as an edge can contain all three of @xmath108 and @xmath116 .
thus @xmath114 is not traceable .
thus @xmath117 hence @xmath118 suppose @xmath0 is a @xmath20-connected mnt graph . if @xmath0 has order @xmath18 and @xmath19 vertices of degree @xmath20 , then @xmath119 .
if @xmath0 has three vertices of degree 2 having the same two neighbours then , by lemma [ mnt3deg2 ] @xmath120 since @xmath121 we now assume that @xmath0 does not have three vertices of degree 2 that have the same two neighbours .
let @xmath122 be the vertices of degree 2 in @xmath0 and let @xmath123 then by lemmas [ mntdeg2 ] , [ mnt1deg2 ] and [ mnt2deg2 ] the minimum degree , @xmath124 of @xmath125 is at least 3 .
hence @xmath126 thus @xmath127 for @xmath128 . for @xmath129
this bound is realized for @xmath130 ( a zelinka type i graph @xcite ) and @xmath131 ( a graph constructed in @xcite ) .
these graphs are depicted in fig . 1 .
@xmath132{cubicfig1.eps}\ ] ] @xmath133 we now construct an infinite family of 2-connected cubic mnt graphs , showing that @xmath134 for infinitely many @xmath17 .
in order to construct cubic mnt graphs we require the following lemmas concerning mhh graphs . 1 .
@xmath137 has a hamiltonian path ending at any of its vertices .
2 . there is no hamiltonian path in @xmath137 with both endvertices in @xmath138 .
3 . for any @xmath139 there exists a hamiltonian path in @xmath140 with the other two vertices of @xmath138 being the endvertices . 1
. @xmath137 is hamiltonian .
if a hamiltonian path exists in @xmath137 having both endvertices in @xmath138 , then @xmath125 has a hamiltonian cycle , which is a contradiction .
3 . since @xmath141 is hamiltonian
there is a hamiltonian cycle in @xmath141 containing the path @xmath142 , where @xmath143 .
thus there is a hamiltonian path in @xmath140 with endvertices @xmath3 and @xmath144 .
[ mnt1]suppose @xmath125 is a mnh graph having a vertex @xmath67 of degree @xmath135 . put @xmath136 .
if @xmath145 and @xmath146 are nonadjacent vertices in @xmath137 , then @xmath147 has a hamiltonian path with both endvertices in @xmath138 .
let @xmath164 $ ] be an inflated @xmath165 obtained from @xmath160 ; @xmath151 and a vertex @xmath51 by joining @xmath51 to the semi - edge incident with @xmath154 for @xmath151 and joining the remaining semi - edges as depicted in fig .
let @xmath166 denote @xmath167 ; @xmath151 .
we call @xmath154 , @xmath168 and @xmath156 the _ exit vertices _ of @xmath166 .
we introduce the following notation which we use in the theorem below : @xmath171 denotes a hamiltonian path in a graph @xmath0 from @xmath3 to @xmath144 ; @xmath172 denotes a hamiltonian path in @xmath0 ending at @xmath144;@xmath173 denotes a hamiltonian path in @xmath0 beginning at @xmath3 ; and@xmath174 denotes a hamiltonian path in @xmath0 .
we now show that @xmath0 is nontraceable .
suppose @xmath50 is a hamiltonian path of @xmath176 .
then at least one of the @xmath177 s , say @xmath178 , does not contain an endvertex of @xmath50 .
thus @xmath50 passes through @xmath178 , using two of the exit vertices of @xmath178 .
however , by lemma [ hypo2](b ) such a path can not contain all the vertices of @xmath178 .
* @xmath179 ; @xmath180.without loss of generality consider @xmath181 . by lemma [ mnt1 ]
there is a hamiltonian path in @xmath182 with endvertices two of @xmath183 , @xmath184 and @xmath185._subcase ( i ) .
_ suppose the endvertices are @xmath183 and @xmath185 .
( a similar proof holds for @xmath183 and @xmath184 . ) by using lemma [ hypo2](a ) we obtain the hamiltonian path @xmath186 _ subcase ( ii ) .
_ suppose the endvertices are @xmath184 and @xmath185 . by using lemma [ hypo2](c )
we obtain the hamiltonian path @xmath187 * case 2 .
* @xmath188 and @xmath189 ; @xmath190 ; @xmath191.without loss of generality we choose @xmath181 and @xmath192 . by using lemmas [ hypo2](a ) and ( c )
we find a hamiltonian path @xmath193 in @xmath49 .
all subcases can be reduced to the following:_subcase ( i ) .
_ @xmath194 . @xmath195 _ subcase ( ii ) .
_ @xmath196 .
@xmath197 _ subcase ( iii ) .
_ @xmath198 .
@xmath199 _ subcase ( iv ) .
_ @xmath200 . @xmath201 _ subcase ( v ) .
_ @xmath202 .
@xmath203 * case 3 .
* @xmath204 and @xmath205 ; @xmath190 ; @xmath94.without loss of generality we choose @xmath181 and @xmath192 .
let @xmath206 be the graph obtained from @xmath0 by contracting @xmath207 to a single vertex @xmath208
. then @xmath206 is isomorphic to @xmath209 and hence , by condition * ( c ) * , @xmath210 has a hamiltonian cycle containg the path @xmath211 .
thus @xmath212 has a hamiltonian path with endvertices @xmath48 and @xmath183 . using this fact and lemma [ hypo2](a ) we construct the hamiltonian path @xmath213 * case 4 . *
@xmath214 and @xmath215 ; @xmath180.without loss of generality we choose @xmath181._subcase ( i ) .
_ @xmath216.consider @xmath217 ( the case @xmath218 follows similarly . ) by using lemmas [ hypo2](a ) and ( c ) we obtain the hamiltonian path @xmath219 _ subcase ( ii ) .
_ @xmath220.according to condition * ( c ) * and an argument similar to that in case 3 , there is a hamiltonian path in @xmath178 with endvertices @xmath3 and @xmath221 , where @xmath222 .
suppose @xmath223 .
( a similar proof holds for @xmath224 . ) using this fact and lemma [ hypo2](a ) we construct the hamiltonian path @xmath225 the petersen graph ( @xmath226 ) , the coxeter graph ( @xmath227 ) and the isaacs snarks @xmath228 ( @xmath229 ) for odd @xmath230 are all cubic mhh graphs ( see @xcite , @xcite ) .
we determined , by using the graph manipulation package developed by siqinfu and sheng bau * , that a snark of order @xmath231 , reported by chisala @xcite , is also mhh .
the petersen graph , the snark of order 22 and the coxeter graph are shown in fig . 4 . @xmath232{petersen1.eps } \ \ \ \ \
\includegraphics[scale=0.5]{snark22.eps } \ \ \ \ \ \includegraphics[scale=0.4]{cox2.eps}\ ] ] @xmath233 all the cubic mhh graphs mentioned above satisfy condition * ( c)*. in fact , it follows from theorem 10 in @xcite that the isaac s snarks satisfy a stronger condition , namely that if @xmath234 then , for _ every _ @xmath235 there exists a hamiltonian cycle in @xmath236 containing @xmath237 , and this condition holds for all @xmath238 .
we determined , again by using the graph manipulation package , that each of the graphs shown in fig .
4 also satisfy this extended condition for the specified vertex @xmath67 .
the programme allows one to sketch a graph @xmath0 on the computer screen by placing vertices and adding edges . on request
the programme will either draw in a hamiltonian cycle or state that the graph is non - hamiltonian .
we tested to see if each of the above mentioned mnh graphs @xmath0 satisfied the condition * ( c ) * by considering symmetry and adding appropriate edges and noting the structure of the hamiltonian cycle drawn in .
* acknowledgements * we wish to thank sheng bau for allowing us the use of the programme , graph manipulation package version 1.0 ( 1996 ) , siqinfu and sheng bau , inner mongolia institute of finance and economics , huhhot , cn-010051 , people s republic of china . *
we wish to thank the referee who brought to our attention the infinite family @xmath241 $ ] of mnt graphs , where @xmath242 is the snark of order @xmath231 and @xmath228 and @xmath243 are isaac s snarks , which gives @xmath38 for @xmath244 . | we determine a lower bound for the number of edges of a 2-connected maximal nontraceable graph , and present a construction of an infinite family of maximal nontraceable graphs that realize this bound .
* keywords : * maximal nontraceable , maximal nonhamiltonian , hypohamiltonian , hamiltonian path , hamiltonian cycle , traceable , nontraceable , hamiltonian , nonhamiltonian , maximal hypohamiltonian * 2000 mathematics subject classification : * 05c38 |
probing into cp violation in the lepton sector is one of the most challenging goals in particle physics .
long baseline neutrino oscillation experiments are the prime candidates for observational means for detecting such effect .
however , it has been known since sometime ago that the earth matter effect acts as a contamination to the measurement of genuine cp violating effects due to the leptonic kobayashi - maskawa phase in such experiments @xcite .
therefore , it is of crucial importance to achieve a complete understanding of the features of interplay between the genuine cp and the matter effects to have a realistic design of experiments for measuring the cp violating phase .
we describe in this article a new powerful tool which we call cp trajectory diagram in bi - probability space " @xcite .
it enables us to represent pictorially the three effects , the effects of ( a ) genuine cp violation due to the @xmath2 term , ( b ) cp conserving @xmath3 term , and ( c ) fake cp violation due to earth matter , separately in a single diagram . by using the cp trajectory diagram
we observe that there is a two - fold ambiguity in the determination of @xmath0 which is related with the sign of @xmath1 . as described in detail in ref .
@xcite , this is a remnant of the approximate degeneracy in the vacuum oscillation probability under the transformations ( @xmath4 ) and ( @xmath5 ) .
we then discuss principles of tuning beam energy for a given baseline distance to resolve the ambiguity , and to maximize the cp - violating effect . finally , we point out that the ambiguity may be resolved with 50 % chance in the super - jhf experiment @xcite with a megaton class water cherenkov detector .
it is quite contrary to the conventional belief that the sign of @xmath1 can not be determined with such a short baseline as @xmath6 km . in a companion article , which is a report for the proceedings of taup2001 @xcite
, we discuss further physical implications of our results .
in particular , we explore the possibility of an _ in situ _ simultaneous measurement of @xmath0 and the sign of @xmath1 in a single experiment .
we now introduce the cp trajectory diagram in bi - probability space spanned by @xmath7 and @xmath8 .
suppose that we compute the oscillation probability @xmath9 and @xmath10 with a given set of oscillation and experimental parameters .
then , we draw a dot on the two - dimensional plane spanned by @xmath9 and @xmath10 . when @xmath0 is varied we have a set of dots which forms a closed trajectory , closed because the probability must be a periodic function of @xmath0 , a phase variable . in fig
1 plotted is the contours of oscillation probabilities @xmath9 and @xmath10 which is drawn by varying the cp violating phase @xmath0 from 0 to @xmath11 .
they are averaged over the gaussian neutrino energy distribution with peak energy of 500 mev and width 100 mev .
as you might have guessed , these diagrams are elliptic .
it is so exactly in vacuum and in a good approximation even in matter at relatively short baseline and as far as the mass hierarchy @xmath12 is the case @xcite .
what does cp trajectory diagram actually represent ?
there is a very simple answer to this question .
the lengths of major and minor axes represent the coefficients of cp violating @xmath13 and cp conserving @xmath14 terms , respectively , in the neutrino oscillation probability @xmath15 . whereas the distance between two ( positive and negative @xmath1 ) ellipses
gives the size of the matter effect .
the last point can be explicitly verified by running the same calculation with twice larger matter effects , as done in ref .
therefore , you can see the size of these three terms just by eye when it was projected onto the cp trajectory diagram . 0.3 cm -2.0 cm fig .
1 : cp trajectory in the bi - probability ( given in % ) plane for the baseline @xmath16 km and energy @xmath17 mev . as indicated in the figures , the solid and the dashed lines are for @xmath18 and @xmath19 cases , respectively , and the dotted and the dash - dotted lines correspond to the same signs of @xmath1 as above but with matter effect switched off .
the mixing parameters are fixed as @xmath20 ev@xmath21 , @xmath22 , @xmath23 ev@xmath21 , @xmath24 , @xmath25 .
we take @xmath26 g/@xmath27 where @xmath28 is the matter density and @xmath29 is the electron fraction . 0.5
cm it is obvious from fig .
1 that the approximate degeneracy ( @xmath4 ) and ( @xmath5 ) which exists in the vacuum case is lifted by the matter effect ; matter helps !
however , it is also clearly seen in fig .
1 that there are remaining degeneracies if the sign of @xmath1 is not known _
a priori_. the above discussion assumes that the value of @xmath30 is known prior to the experiment .
if the value is not known in advance there is an another ambiguity , the ( @xmath31 ) ambiguity , as discussed in detail in ref .
therefore , there exits a combined ambiguity , ( @xmath0 @xmath32 sign of @xmath1 ) times ( @xmath31 ) , which in the worst case can be 4-fold . the ambiguity was referred to as the `` clover - leaf ambiguity '' in taup2001 @xcite .
if one wants to resolve the two - fold ambiguity ( @xmath0 @xmath32 sign of @xmath1 ) in a single experiment one can not tune the experimental parameters so that the @xmath13 term is maximal because then the trajectory shrinks to a straight line , closing the possibility of resolution of the ambiguity . then , there should be a compromise .
first , one can obtain the conditions to maximize lengths of major and minor axes , which lead to @xcite @xmath33 it is shown in ref .
@xcite that one can compromize these two requirements in the energy region of 0.5 to 2 gev .
we now address our final subject , probably the most important one , the possibility of simultaneous determination of @xmath0 and @xmath1 in the super - jhf experiment .
we assume that @xmath30 is known in a reasonable accuracy prior to this experiment . in fig .
2 plotted is the cp trajectory diagram in the number of appearance events plane for @xmath34 and @xmath35 channels .
we assume a water cherenkov detector of fiducial volume 0.9 mton , and 4 mw of proton beam power which is planned in the jhf experiment in its phase ii @xcite .
we use the off - axis ( oa ) 3 degree beam whose neutrino energy peaks at @xmath36 gev .
we refer ref .
@xcite for a detailed explanation of how the computation of number of events is done and the results for alternative choice of beams .
one can clearly see from fig .
2 that at least a half of the parameter space fulfilling the condition @xmath37 does not suffer from the ambiguity problem . then ,
with the 50 % chance ( thanks to nature s kind setting ! ) they will be able to determine the cp violating angle @xmath0 and the sign of @xmath1 simultaneously in the super - jhf experiment .
0.2 cm -2.0 cm fig .
2 : cp trajectory in event number plane @xmath38 for oa beam 3 degree and the baseline @xmath16 km .
the dotted circles correspond to 3 @xmath39 statistical uncertainty .
this work was supported by the brazilian funding agency fundao de amparo pesquisa do estado de so paulo ( fapesp ) , and by the grant - in - aid for scientific research in priority areas no . 12047222 , japan ministry of education , culture , sports , science and technology .
j. arafune and j. sato , phys . rev .
* d55 * ( 1997 ) 1653 ; j. arafune , m. koike and j. sato , phys . rev . * d56 * ( 1997 ) 3093 [ erratum _ ibid . _ * d 60 * ( 1999 ) 119905 ] ; h. minakata and h. nunokawa , phys . rev .
* d57 * ( 1998 ) 4403 ; phys .
b413 * ( 1997 ) 369 ; _ ibid . _
* b495 * ( 2000 ) 369 ; k. dick , m. freund , m. lindner and a. romanino , nucl .
b * 562 * ( 1999 ) 29 ; o. yasuda , acta .
* b 30 * ( 1999 ) 3089 ; m. koike and j. sato , phys . rev . * d61 * ( 2000 ) 073012 ; erratum _ ibid . _ * d 62 * ( 2000 ) 079903 ; a. cervera , a. donini , m. b. gavela , j. j. gomez cadenas , p. hernandez , o. mena and s. rigolin , nucl . phys .
* b579 * ( 2000 ) 17 [ _ erratum - ibid .
_ * b593 * ( 2000 ) 731 ] ; s. j. parke and t. j. weiler , phys .
* b501 * ( 2001 ) 106 . | we introduce cp trajectory diagram in bi - probability space " as a powerful tool for pictorial representation of the genuine cp and the matter effects in neutrino oscillations .
existence of the correlated ambiguity in a determination of cp violating phase @xmath0 and the sign of @xmath1 is uncovered .
principles of tuning beam energy for a given baseline distance are proposed to resolve the ambiguity and to maximize the cp - odd effect .
we finally point out , quite contrary to what is usually believed , that the ambiguity may be resolved with 50 % chance in the super - jhf experiment despite its relatively short baseline of 300 km .
= 1.cm and |
solar - like pre - main sequence ( pms ) stars and , in general , the low mass pms stars ( m@xmath6 ) , or t tauri stars ( ttss ) , are complex dynamical systems made of two basic components , star and accretion disk , as well as a dynamical interface , the stellar magnetosphere ( see gmez de castro 2013 , for a recent review ) .
magnetic fields of few kg have been detected in the surface of the ttss ( guenther et al 1999 ; johns - krull et al 1999 ; johns - krull 2007 ) .
the surface field is not bipolar but it has a rather complex structure as the sun s field has ( johns - krull et al 2004 ) .
hence , though higher order , multi - polar components fall off more rapidly with radius than the dipolar field , it is difficult to track the final path followed by matter from the inner disk border to the stellar surface .
in addition , the magnetosphere has its own dynamics and forcing due to the interaction with the disk .
unfortunately , the characteristics of the ttss extended atmosphere and magnetospheres are still escaping the diagnosis ( see hartmann 2009 ) .
little is known about them , apart from having a density of @xmath7 @xmath8 and an electron temperature between some few thousand kelvin and 100,000 k ( gmez de castro & verdugo 2007 , hereafter gdcv2007 ) .
recent attempts to derive the magnetospheric properties from optical lines have shown that , for instance , the h@xmath9 profile is strongly dependent on densities and temperatures assumed inside the magnetosphere and in the disk wind reagion ( lima et al .
the line widths of typical atmospheric / magnetospheric tracers are about 200 - 300 km / s that exceed by far what expected from thermal or rotational broadening even if the lines are assumed to be formed in a magnetosphere that extends to some 4 - 5 stellar radii and corotates with the star . as today
, it is still unclear whether the broadening is produced by unresolved macroscopic flows or by magnetic waves propagating on the magnetospheric field ( hartmann et al .
in general , the broadening of magnetospheric tracers does not vary significantly in time , pointing out that the average motions are rather stable .
the combined effect of funnel flows and inclined magnetic rotators as simulated by romanova et al ( 2004,2012 ) is the current baseline for the numerical simulation of ttss magnetospheres .
the stellar field is assumed to be anchored in the inner part of the disk creating a sheared layer between the rigid body rotation of the star and the keplerian rotation of the disk .
this interaction has a profound influence on the star and the accretion flow but also acts as a dynamo that transform part of the angular momentum excess in the inner disk into magnetic field amplification that self - regulates through quiescent periods of field building - up and eruptions when the energy excess is released ( see gmez de castro & von rekowsky 2011 , for an evaluation of the uv output from such interface ) .
however , magnetospheric heating processes are poorly known and accretion shock models are unable to predict the observed line fluxes / broadening ( johns - krull 2009 ) .
the atmospheric and magnetospheric energy output is released mainly in the ultraviolet .
the richness of spectral tracers for a broad range of magnetospheric temperatures and densities is unmatched by any other range .
there are several studies of the atmospheric / magnetospheric properties of the ttss based on low dispersion uv data ( lemmens et al .
1992 , hulamo et al .
1998 , johns - krull et al .
2000 , yang et al .
2012 , gmez de castro & marcos - arenal , 2012 , hereafter gdcma2012 ) .
however , only the cosmic origins spectrograph ( cos ) ( see green et al 2012 for a description of the instrument ) , on board the hubble space telescope ( hst ) , has been sensitive enough to gather high signal - to - noise ratio ( snr ) profiles of hot plasma tracers such as the n v [ uv1 ] resonance multiplet or the he ii h@xmath9 transition with a resolution above 10,000 in late k and m - type ttss ( see penston & lago 1984 for c iv profiles obtained with the iue , ardila et al 2002 , for hst profiles obtained with the goddard high resolution spectrograph and ayres 2005 , for the coolcat set based on hst data obtained with the space telescope imaging spectrograph and compare them with the presented in sect . 2 ) . following a previous work where the magnetospheric properties of ttss were examined based on low dispersion hst data ( gdcma2012 ) , the high resolution profiles of the n v [ uv1 ] , o iii ] and he ii(1640 ) transitions are analysed in this work with two main objectives .
firstly , it is intended to determine whether it is feasible to discriminate the contribution to the high energy radiation flux from the high density atmospheric plasma and from the accretion flow . moreover ,
the unexpected correlation between x - ray flux and the high energy uv tracers found by gdcma2012 is re - examined on the light of the kinematic information contained in the high resolution hst / cos profiles . in sect .
2 the observations are described .
the results are presented in sect .
two components are found to contribute to the he ii flux a low density component ( ldc ) associated with the accretion flow and a high density component ( hdc ) of more uncertain origin .
the discussion on the possible source of the hdc , its association with accretion shocks and the properties of the ldc are addressed in sect .
the article concludes with a brief summary on the relevance of these results .
the he ii profiles of ttss in the taurus - aurigae star forming region have been retrieved from the hst archive .
only high resolution observations have been considered .
most of them have been obtained with the cosmic origins spectrograph ( cos ) and the gratings g130 m and g160 m .
the resolution is @xmath1024,000 and each target has been observed three times with slight offsets in the wavelength range to guarantee that the 18.1 gap between the two segments in the fuv detector is covered ( see cos handbook ) .
moreover , some few ttss have been observed with the space telescope imaging spectrograph ( stis ) , namely , t tau , dr tau , and df tau .
only the t tau profile had a high enough signal - to - noise ratio ( snr ) in the he ii line to be considered for this work .
the log of the observations is provided in table 1 ; additional information on the hst programs that obtained these observations ( programs ids . 11533 , 11616 and 8627 ) can also be found in the table .
the observational strategy allowed to search for variations in the profiles in time scales of @xmath11 minutes but neither flux , nor morphology were found to display significant changes .
the one - dimensional spectra produced by the cos calibration pipeline ( calcos v2.17.3 ) were aligned and co - added .
cos targets must be centred to within @xmath12 arcsec to achieve the nominal wavelength accuracy of @xmath13 km s@xmath2 .
the ( r(3 ) 1 - 7 ) 1489.636 and ( p(5 ) 1 - 7 ) 1504.845 h@xmath14 lines have been used to set the zero of the wavelength scale for the targets in table 1 .
h@xmath14 emission is dominated by the molecular disk around the stars in most sources ( france et al .
the lines have been selected to be strong ( from herczeg et al .
2002 ) and detectable in most of the sources .
the profiles of the 1489.636 line are plotted in figure 1 ( see also figure 3 in france et al . 2012 ) and the he ii profiles are plotted in figure 2 .
no shifts have been applied to dn tau and ip tau observations because the h@xmath14 emission was too weak to be used for this purpose .
neither shifts have been applied to hbc 427 , lkca 19 and lkca 4 because h@xmath14 emission was not detected .
the h@xmath14 profiles are sometimes asymmetric with respect to the rest wavelength , very especially in rw aur . in this case
, the original zero from the calcos pipeline has been left .
only a subset of aa tau observations has been used since guide star acquisition failed ( see fig .
3 ) . as a result
, the first two exposures produced similar profiles while the last one produces a slightly broader and more red - shifted ( by @xmath15 or 18 km s@xmath2 at 1640 ) profile . for this star , the last observation was rejected and only the two first observations were averaged to produce the profiles in figs . 1 and 2 . the he ii profiles can be generically described as composed of a bright and narrow emission feature and a broad , weaker component that differs from one star to another .
notice that the he ii lines are very strong ; this fact , together with the strong h@xmath14 emission , contributes to the continuum jump in the low resolution advanced camera system on hst reported by gdcma2012 .
close to the he ii line , there are the o iii]@xmath16 intercombination lines ; a doublet with components at @xmath17 and 1666.156 , which originate under transitions from the level 2s2p@xmath18 @xmath19s@xmath14 to the term 2s@xmath202p@xmath20 @xmath18p with j=1 and j=2 , respectively .
the components should have an intensity ratio @xmath211:3 ; equal to the ratio of their transition probabilities ( 145 s@xmath2 and 426 s@xmath2 ) .
the transition is optically thin to a critical density of @xmath22 @xmath8 .
the o iii ] profiles are represented in fig .
the two lines of the multiplet are observed in gm aur , df tau , hn tau , dr tau , su aur , rw aur and de tau ; however the flux of the weakest , 1660.802 line , has only been measured for strong sources . in all cases ,
the flux ratio between the two lines of the multiplet is 1:3 ( within the error bars ) . to complete the view on the distribution of hot plasma in the ttss environment , also the profiles of the resonance uv 1 multiplet of the n v
have been retrieved from the hst archive ( see table 2 ) . in the blue edge of the 1238.8 line , the strongest in the doublet , there are some narrow emission lines produced by molecular hydrogen ( lines : @xmath23 1 - 2 p(8 ) , @xmath24 2 - 2 r(11 ) ) that somewhat blur the profile .
the zero of the wavelength scale has been set again , resourcing to h@xmath14 emission lines .
the ( p(2 ) 0 - 4 ) 1338.63 line has been used for this purpose , since it is strong in most of the stars and it is not blended with other features ( see figure 5 with the h@xmath14 profiles ) . as mentioned above ,
the original zero of the wavelength scale has not been shifted for dn tau , ip tau , hbc 427 , lkca 19 and lkca 4 because the h@xmath14 lines were either very weak or absent .
the n v profiles can be most generally described by a single component that ranges from being narrow in stars like hbc 427 , lkca 19 and lkca 4 to be broad and asymmetric in aa tau or gm aur ( see fig .
. some relevant properties of the ttss to be used in sect .
3 are gathered in table 3 .
notice that there are wide variations in published values of important parameters such as the stellar luminosity or the extinction ( see also comments in gdcma2012 ) ; data in table 3 are gathered for reference for other researchers .
the x - ray fluxes have been retrieved from the xmm - newton extended survey of the taurus molecular cloud ( xest ) ( guedel et al .
the he ii , the 1238.82 n v and the o iii ] lines fluxes have been measured after subtracting the local continuum .
also the fluxes for the 1489.636 and 1338.63 h@xmath14 lines have measured ( see sect .
5).the lines fluxes are provided in table 4 ; they are not extinction corrected .
for some sources , the measurement of the 1238.82 n v flux has required subtracting the nearby h@xmath14 features . in such cases , the h@xmath14 flux has been subtracted by linear interpolation in the n v profile
. however , there are some few profiles where this linear interpolation was uncertain ( see quality flags in table 4 ) .
from figures 1 to 6 a generic trend can be inferred : * the _ weak - line ttss ( wttss ) _ in the sample , _
i.e. _ evolved ttss with no evidence of mass infall ( lkca 19 , lkca 4 , hbc 427 ) , do not display h@xmath14 emission , neither nebular o iii ] emission .
they have only rather narrow he ii and n v lines . * _ intermediate objects ( ttss ) _ like ip tau , dn tau and v836 tau have weak h@xmath14 and no nebular o iii ] emission .
both he ii and n v lines have narrow emission profiles . *
_ the classical ttss ( cttss ) _ emit in all these tracers ( see below ) .
cttss cover a broad range of profiles morphologies .
strong h@xmath14 emission is detected in all of them and the lines are narrow unless in rw aur a ( see france et al 2012 ) .
nebular o iii ] emission is detected in all of them unless in dm tau , ux tau a and aa tau .
o iii ] is especially strong in t tau , df tau , hn tau , dr tau and rw aur .
su aur , and de tau seem to be intermediate objects .
hints of possible o iii ] emission at 1666 are seen in gm aur , unfortunately the s / n is low and the weakest component of the multiplet is not detected . a low s / n
feature is detected at 1666 aa in dn tau , ip tau , aa tau and ux tau spectra .
observations of the c iii ] intercombination transition line are only available for three of the stars in the sample , namely , de tau , t tau and rw aur and no significant differences have been found between the o iii ] and the c iii ] profiles obtained either with the goddard high resolution spectrograph ( ghrs ) ( compare fig . 4 with fig.1 in gmez de castro et al 2001 ) and stis ( compare fig . 4 with fig.1 in gmez de castro et al 2003 ) . in general , the he ii profiles can be described by a narrow emission component superimposed on a broader contribution that mimics the observed in the o iii ] nebular lines , when high enough s / n data profiles are observed .
this effect is clearly apparent in hn tau and rw aur .
note that in dr tau and df tau , the overlap is lost at bluewards shifted velocities ; there is bluewards shifted o iii ] emission with no he ii counterpart .
this bluewards shifted excess could be caused by the contribution of an unresolved jet to the line emission .
as shown for ry tau by gdcv2003 , the semiforbidden emission has two components : one associated with the jet and another with the accretion flow that were disentangled for this source due to their variability .
the n v profiles however , do not follow the same morphological trend .
a rather narrow symmetric profile is observed in de tau that becomes wider in dm tau , df tau , ux tau a , gm aur and su aur , all of them sources with absent or weak o iii ] nebular emission .
dr tau , hn tau , rw aur and aa tau display very peculiar profiles that clearly indicate that n v emission is not produced in the stellar atmosphere but in another dynamical component .
in particular , the n v profiles of rw aur , hn tau and aa tau are asymmetric extending from red - wards shifted velocities to peak at blue - wards shifted velocities ; moreover , the rw aur profile peaks at the velocity of the optical jet ( hirth et al 1997 ) that was also detected in the c iii ] intercombination line by gmez de castro & verdugo , 2003 .
this type of profile asymmetry has also been detected in the c iii ] profile of ry tau ( gdcv2007 ) , who pointed out that the line could be formed in a pre - main sequence analogue of the solar wind . in this context
, it is worth remarking that the h@xmath14 profile of rw aur has a completely different asymmetry , the flux peaks to the red of the line , suggesting the presence of infalling cold molecular gas similar to that observed in the pre - main sequence close binary ak sco ( see also gmez de castro et al 2013 ) .
these groups can be cleanly recognised in velocity dispersion diagrams .
the characterisation of the underlying velocity field and thermal properties of the line emission region is complex in the ttss environment . instead of using the standard fitting to gaussian or voight profiles , it is preferable to characterise the profile in terms of the standard pearson statistics moments and measures , i.e. , mean or centroid , dispersion , kurtosis and skewness .
they provide a quantitative measurement of the deviation of the profile from the expected for a thermal plasmas ; a normal distribution convolved with the line spread function ( lsf ) of cos ( kriss 2011 ) .
note that in this approach , the profile is assumed to be formed by the contribution to the line flux of independent gas parcels , _
i.e. _ the ( background subtracted ) profile is treated as a histogram of the flux emitted per parcel in the wavelength ( velocity ) space ; a similar approach was followed by gdcv2007 to compare the observed c iii ] and si iii ] profiles of ry tau with the theoretical predictions .
this treatment permits to characterise the profiles that are formed in complex velocity fields . in table 5
, the dispersions of the he ii and n v lines are provided together with that obtained for two control lines , the h@xmath14 transitions at 1339 and 1489 ; rw aur , hn tau and aa tau are not included because their n v profiles are peculiar . as shown in the bottom panel of fig .
8 , the average dispersion of the h@xmath14 profiles is @xmath25 km s@xmath2 and @xmath26 km s@xmath2 for the 1339 and 1489 lines , respectively .
there is not any correlation between the broadening of these two h@xmath14 lines , pointing out that the scattering of the dispersions in the diagram is related with random effects associated with the measurement process .
a different trend is drawn from the n v and he ii dispersions .
the dispersions of the wttss and dn tau are comparable to those measured in the h@xmath14 line . in intermediate objects , like v836tau and ip tau ,
the dispersions in the he ii line are comparable to those measured in the h@xmath14 lines but the n v lines are significantly broader .
finally in the cttss both @xmath27 and @xmath28 are larger than @xmath29 .
note that stellar rotation may contribute to the line dispersion ; su aur , the fastest rotator in the sample has also the largest dispersion .
however , a large dispersion can also be produced by the profile asymmetry .
for instance , dm tau has dispersions comparable to those measured in su aur and it is one of the slowest rotators in the sample ( see table 3 ) .
two objects do not follow this trend : dr tau and de tau , both have intermediate dispersions in the n v lines and large dispersions in the he ii lines caused by the broad emission component . a quick inspection in the summary of the ttss properties ( see table 3 ) indicates that the only possible cause of this discrepancy is the high accretion rate , as otherwise , expected . from figures 2 , 4 and 6 ,
it is clearly inferred that there are , at least , two different plasmas contributing to the spectral lines under study : * _ a low density component ( ldc ) _ that it is most conspicuously traced by the o iii ] line .
the ldc also produces the he ii broad emission component observed in rw aur , hn tau , dr tau and df tau .
the critical density of the o iii ] sets an upper limit to the electron density of the plasma in the ldc of @xmath30@xmath8 , ( see also sect .
the ldc profiles display a non - thermal broadening and draw a complex velocity field around the stars , i.e. , it is not associated to a simple standing atmospheric structure whose kinematics is dominated by stellar rotation .
in fact , the ldc profiles seem to rather trace some kind of complex magnetospheric infalling pattern , high above the stellar surface . also , in some cases , could be associated to unresolved wind structures ( gmez de castro & ferro - fontn 2005 ) , as the reported for ry tau ( gdcv2007 ) .
notice that the co - existence of o iii ] and he ii radiation from the same kinematic structure would point out an unrealistically high electron temperatures for the line emission region ( @xmath31 ) , if collisional equilibrium at a single temperature is assumed and electron densities below the o iii ] critical density are considered .
in fact , its uv spectrum is reminiscent of that observed in photoionized nebulae ( see also sect .
to the current sensitivity , the contribution of this component to the n v flux is negligible . * _ a high density / temperature component ( hdc ) _ that dominates the n v emission .
o iii ] profiles are very different from n v profiles suggesting that the density is of the n v formation region is higher than the o iii ] critical density .
though the kinematics of the n v emission region is clearly different from that traced by the ldc , the n v flux is correlated with the he ii flux as shown in fig .
9 ; the spearman rank correlation coefficient is @xmath32 ( with significance level , @xmath33 , see sachs 1982 for details ) and , + @xmath34 with rms = 0.29 ( see bottom panel in fig . 9 ) .
also the fluxes normalised to the stellar surface are correlated with @xmath35 , with @xmath36 ( see top panel in fig .
9 ) and , + @xmath37 with an rms=0.31 .
the normalised flux is defined as the rate @xmath38 or @xmath39 and was introduced by gdcma2021 to provide a measure of the line emissivity weighted over an unknown thickness but corrected from stellar radii and surface temperature . in this manner , the normalised fluxes compensate for scaling effects associated with the broad range of mass , luminosity and stellar radius covered by the ttss .
stars whose he ii flux has a significant contribution from the ldc are marked in the plot .
notice that they are evenly distributed in the figure suggesting that he ii and n v fluxes are correlated , independently of whether the he ii flux is dominated by the narrow emission component .
based on low resolution observations , gdcma2012 pointed out that the normalised he ii flux anti correlates with the strength of the x - ray flux as derived in the xest survey ( guedel et al 2007 ) carried out with the xmm - newton telescope in the 0.3 - 10.0 kev band . in fig .
10 , the normalised he ii fluxes from the high resolution cos / hst observation ( see table 3 ) are represented against the normalised x - ray luminosities as derived from the xest survey .
also the chandra / acis observations of lkca 4 , de tau , gm aur from yang et al .
( 2012 ) are used ; they are integrated x - ray luminosities in the 0.3 - 10 kev range .
the he ii flux has been extinction corrected according to valencic et al ( 2004 ) assuming r=3.1 and extinctions in table 3 .
the low dispersion , gdcma2012 data , have been also plotted for those sources with no available high dispersion data .
there is significant contribution to the he ii flux from the ldc in some few sources : su aur , hn tau , de tau and dm tau , that are marked in the figure .
a first inspection of the fig 10 , shows three groups .
the wttss ( hbc 427 , lkca 4 , hd283472 ) , ux tau and the fast rotator su aur are very close to the main sequence stars regression line .
the cttss follow the generic trend pointed out by gdcma2012 , the he ii flux increases as the x - ray flux decreases .
notice that the sources with strong ldcs are far from this trend , excluding the fast rotator su aur .
if hn tau , de tau and dm tau are excluded , the spearman rank correlation coefficient is @xmath40 , with @xmath41 , and the least square fit is , @xmath42 with rms = 0.15 .
uncertainties in the a@xmath43 values may produce a shift towards large @xmath44 in the plot reinforcing the trend ; note that the regression line runs nearly parallel to the extinction direction . to examine its effect ,
the same diagram has been plotted in the bottom panel of fig 10 using , stellar luminosities , x - ray luminosities and a@xmath43 values from the recent compilation by yang et al ( 2012 ) . as shown , the three groups are no so cleanly separated and the regression line is more clear ; only de tau is far from the trend and has been excluded from the calculation .
the results are similar though the negative slope is softer : @xmath45 with rms = 0.15 and @xmath46 , with @xmath47 . in summary , though the clean separation between wttss and accreting ttss may be an extinction associated effect , the trend to release preferentially the high energy excess in the uv rather than in the x - ray channel in accreting objects holds with the only possible exception of the sources with a strong contribution the he ii flux from the ldc . from the current data sets and observations , it can not be ascertain whether there is a statistically meaningful deviation of the sources with strong ldcs from the main trend .
unfortunately , there are not x - ray measurements of dr tau or df tau and rw aur was only detected to have a low soft x - ray flux with the einstein satellite ( damiani et al 1995 ) . if present , such a trend could indicate that x - ray radiation is dominated by different components in sources with strong ldcs ( strong nebular component ) and in sources with weak or absent ldcs .
the x - rays energy distribution of the ttss is often modelled by two components : a soft component at @xmath48 k and a hard component at @xmath49 k ( see i.e. glassgold et al , 2000 ) .
the hard x - ray component is thought to be associated with magnetic energy release in the stellar coronae .
the nature of the soft x - ray component is more uncertain and often , it has been hypothesised that could be formed in accretion shocks ( lamzin 1998 , gullbring et al 1998 ) . unfortunately
, only three stars in our sample namely , t tau , su aur and hbc 427 have a high enough count rate to allow a spectral fitting to two different optically thin plasmas and non - conclusive results could be derived from the fits ( see table 6 with the two - components fit to the x - ray spectrum of these sources ( from table 6 in guedel et al .
( 2007 ) ) .
the x - ray spectrum of su aur , a ctts , is dominated by the low temperature component with t=5.22 mk however , both soft and hard x - ray component , have similar emission measures in hbc 427,a non - accreting wttss . moreover , the hard x - ray component dominates the x - ray spectrum of the ctts , t tau .
ttss are complex objects ; they are convective pms stars where a solar - like dynamo begins to set in , while still the fossil magnetic field is diffusing .
ttss are surrounded by an external dynamo that powers and makes rise the stellar magnetosphere to the inner border of the molecular disk ( see romanova et al 2012 for recent simulations ) .
matter from the disk , slides down onto the star along the magnetospheric field lines to end , free - falling onto the open holes of the magnetic configuration . in this environment
, hot plasma radiating in the uv tracers studied in this work can be located in the magnetosphere , in the atmosphere , in the accretion shocks and also in the outflow ( either solar - like or driven from the star - disk magnetic interface or the disk ) .
both , magnetosphere and outflow , have significantly lower densities than the stellar atmosphere or the accretion shocks ; as a result , spectral lines radiation is dominated by radiative de - excitation processes and forbidden and semiforbidden transitions are strong from this plasma . in the dense atmosphere , collisional de - excitation is relevant and forbidden transitions are quenched .
uv semiforbidden transitions can not be observed from the accretion shock itself , because it is too hot and dense however , the soft x - ray radiation produced in the shock front photoionizes the preshock gas , which has a density similar to that of the stellar magnetosphere and may produce forbidden lines radiation ( gmez de castro & lamzin , 1999 ) .
unfortunately , the high column density prevents the uv radiation from the photoionization cascade to escape easily from the accretion column .
also , the profiles of some tracers , from the infrared he i transition ( beristain et al . 2001 , fisher et al . 2008 ) to the uv lines do not agree with the predictions of accretion shock models ( johns - krull 2009 ) . within this context , the data presented in sect .
3 provide some amazing results : *
the he ii , o iii ] and n v fluxes do not depend on the spectral type .
this confirms that _ the line emission is not dominated by main sequence like atmospheric magnetic activity _ , _ i.e. _ with the release of the magnetic energy produced by the stellar dynamo , since this one depends on the spectral type ( see e.g. ayres et al 1995 and gdcm2012 ) .
* the he ii and n v fluxes correlate well independently of whether the line profiles are very different , _
e.g. _ , independently of whether the line emission is produced in the same physical structure .
this confirms that all processes ( accretion , atmospheric emission and outflow ) are coupled , as otherwise expected ( see gomez de castro 2013 for a recent review ) . in turn
, it makes difficult to get specific tracers of individual processes without the kinematical information , _
i.e. _ without high resolution spectroscopy .
* the high resolution profiles of the n v line show a symmetric line broadening that increases from non accreting to accreting stars , being significantly suprathermal in these last sources ( see table 5 ) . the profile shape and
the density of the line formation region suggests an atmospheric origin ( with the exceptions already mentioned in sect .
the connection between line broadening and accretion suggests that the density and extent of the high atmospheric layers depends on the accretion rate , _
i.e. _ on the evolutionary state , as otherwise predicted from the theoretical models ( dantona & mazzitelli 1997 , siess et al 2000 )
. transport of magnetic energy from the stellar interior to the surface is expected to occur at a different pace in accreting sources than in wttss .
moreover , the extended magnetosphere powered by the disk - star magnetic locking , must affect the stellar atmosphere introducing new sources of stirring and turbulence ( see e.g. kivelson & russell , 1995 ) .
* however , the physical source of the narrow emission component in the he ii profile keeps being uncertain .
it could either be associated with accretion shocks or with atmospheric features . in this section ,
the possible source of the he ii narrow emission component is analysed as well as some constraints on the extent of the magnetosphere inferred from the semiforbidden line radiation .
the kinematics of the region where the narrow component forms , is clearly distinct from that of the n v or the o iii ] lines formation region ( see fig
the dispersion of the narrow emission component of the he ii ranges from @xmath11 km s@xmath2 to @xmath50 km s@xmath2 while the dispersion of the n v line varies varies from @xmath51 km s@xmath2 to @xmath52 km s@xmath2 for the same sources ( see fig .
the dispersion of the narrow emission component in the he ii profile has been evaluated as above ( see sect.3 ) but setting an upper wavelength cut - off to reject the contribution of the ldc .
note that even setting this cut - off there is an unknown contribution from the ldc to the flux .
moreover , the narrow component seems to be slightly redshifted in _ all _ sources , from 24 km s@xmath2 in ux tau to 37 km s@xmath2 in de tau ( see appendix 1 and table 7 ) .
hence , it would be tempting to suggest that the line is produced in accretion shocks .
accretion shocks are produced by the impact of the free - falling material from the disk onto the stellar surface .
the kinetic energy of the infalling matter , with typical free - fall speeds of @xmath53 km s@xmath2 is involved into gas heating at the shock front that reaches temperatures of about 1mk .
the soft x - ray radiation from the shock front is expected to photoionize ( pre - ionize ) the infalling gas column ( see i.e. lamzin 1998 , gmez de castro & lamzin 1999 , muzerolle et al 2001 , orlando et al 2009 ) . also , the shock front could back illuminate the stellar surface , becoming a source of atmospheric photoionization to be added to the coronal x - ray radiation . in this context , the narrow component of the he ii line could be produced very close to the shock front .
the slight , but small redshift , could be interpreted as an indication of the line being formed in postshock material , and the small width could be caused by thermal broadening ; the thermal velocity of fully ionized , solar abundance plasma at 50,000 k is 23 km s@xmath2 .
however , this can not be concluded from these data alone .
firstly , the apparent redshift could be caused by the blending of the narrow component with the broad component which is asymmetric in most sources , _
e. g. _ , with very low or absent flux at bluewards shifted velocities .
the line broadening is affected by the same problem though no
so dramatically given the relative strengths of the narrow and broad components . unfortunately ,
unless very high s / n profiles ( @xmath54 ) of the semiforbidden and the he ii lines are obtained , any fitting is hampered by these uncertainties . finally , it is also , intriguing that the broadening of the he ii emission line in non - accreting ttss , such as lkca 4 , lkca 19 and hbc 427 , is comparable to the observed in ux tau or in v836 tau that are ttss with low accretion rates . in this respect , it is worth noticing that the correlation between the he ii flux and the accretion luminosity , as derived from the u - band excess ( ingleby et al 2009 , gullbring et al 1998 ) is mild , as shown in fig . 12 ( see also gdcma2012 ) .
taking into account all these facts , as well as the good correlation between the he ii and the n v fluxes ( fig .
9 ) and the convergence of the he ii and n v lines broadenings towards the non - accreting wttss , a contribution to the narrow component of the he ii line from the stellar atmosphere can not be neglected . in summary ,
the data analysed in this work are non - conclusive concerning the source of the narrow component of the he ii line .
for all the reasons mentioned above , it is most probable that both physical components , accretion shocks and stellar atmosphere , contribute to the flux .
the profiles produced in the ldc can not be ascribed to a simple kinematics shared by all sources , neither the radiating plasma can be modelled by simple collisional plasma models .
however , some constraints on its overall physical properties can be derived from the ratios of the intercombination lines of o iii ] , si iii ] and c iii ] .
the plasma density can be constrained from the si iii]@xmath55c iii ] ratio ( gmez de castro & verdugo 2001 ) .
the detection of o iii]@xmath56 can be used to fix up the temperature since the o iii]@xmath55si iii ] ratio is very sensitive to it .
the si iii ] and c iii ] lines have been observed with stis for three sources in the sample : rw aur ( gmez de castro & verdugo 2003 , hereafter gdcv2003 ) , de tau and t tau ( gmez de castro et al 2003 ) .
t tau profiles display a non - negligible contribution from the large scale jet .
de tau si iii ] and c iii ] profiles are rather narrow and similar to the o iii ] line . from rw
aur study , gdcv2003 pointed out that the emitting volume is clumpy with a rather small filling factor , as otherwise expected if magnetospheric radiation is produced in plasma filaments and clumps .
this clumpy nature together with the broad temperature range covered by the various spectral tracers suggest that the excitation mechanism could be photoionization instead of collisional excitation .
gdcv2003 produced two grids of photonionization models to explore possible regimes for line excitation making use of cloudy ( ferland 1996 ) , a code designed to simulate emission line regions in astrophysical environments .
the first set assumed that ldc had a belt like geometry , being illuminated by the ambient x - ray radiation field : soft and hard components , at @xmath57 k and 2.8@xmath58 k respectively , with a total x - rays luminosity of @xmath59 erg s@xmath2 .
this model turned out to be unable to reproduce comparable strengths of the three spectral tracers .
however , if the o iii ] , si iii ] and c iii ] emission is assumed to be produced in dense gas around small x - ray sources , such as reconnecting loops , the lines ratios could be reproduced for electron densities of @xmath60 @xmath8 . for soft x - ray sources , with t@xmath61 k ,
luminosities of @xmath62 erg s@xmath2 and radii of 10@xmath63 @xmath8 , the inferred o iii ] emissivity is @xmath64 erg s @xmath65 . using this one as a fiducial value ,
an estimate of the ldc volume , @xmath66 could be derived from the line strength as , @xmath67 ) 4 \pi d^2}{\epsilon _ { o iii]}}$ ] where @xmath68)$ ] is the reddening corrected line flux and @xmath69 is the filling factor of the hot plasma . for filling factors of 10 , and assuming that the emission is concentrated in a spherical shell of radius r@xmath70 and thickness 0.01r@xmath70 , ldc radii from 4 to 9 r@xmath71 are inferred .
these values are within a factor of 1.5 of the magnetospheric radii derived from accretion luminosities for stars with known magnetic fields ( johns - krull 2007 , gdcma2012 ) .
unfortunately , the uncertainties in the plasma distribution prevent more detailed evaluations .
the uv radiation from n v , he ii and o iii ] is dominated by the contribution of three main components of the ttss atmospheric / magnetospheric environment : the magnetospheric flow of infalling matter , the disturbed upper atmospheric layers and the accretion shock .
the diffuse magnetospheric plasma ( ldc ) is best traced by the o iii ] line : it produces asymmetric profiles , preferentially red - shifted .
he ii emission is observed from the same kinematical structures that radiate in o iii ] .
this plasma is not excited by thermal collisions at a single electron temperature .
in fact , there are indications that photoionization processes could be significant .
the hot , dense layers of the stellar atmosphere are best traced by the n v line . however , the line dispersion increases steadily with the accretion rate suggesting a connection between the disturbances in the upper atmospheric layers and the accretion flow .
the he ii flux is dominated by a narrow emission component of uncertain origin . in this work ,
we have presented evidences indicating that it may be formed in hot postshock material in accretion shocks but also , evidences of its connection with atmospheric tracers .
both accretion shocks and the upper atmospheric layers seem to contribute to the narrow component of the he ii profile .
the anticorrelation between x - ray and uv flux found but gdcma2012 has been confirmed , suggesting that the dissipation of magnetic energy proceeds in the ttss differently than in main sequence stars .
the denser environment produced by mass accretion ( see i.e. petrov et al .
2011 ) seems to favour the ultraviolet channel for the dissipation of the magnetic energy excess .
all the observations indicate that the uv radiation field during pms evolution is much harder than the usually implemented in the modelling of protostellar disks evolution .
an example of its effect in the dust grains charging and charging profile can be found in pedersen & gmez de castro 2011 .
theoretical modelling of protostellar disks chemistry and life generation environments should take this fact into account .
kevin france brought to my attention the hst / cos data corresponding to aa tau .
the analysis of the data pointed out that the 1640 jump in the hst / acs data was dominated by an unexpectedly strong he ii line .
this article has grown from extending this analysis to the ttss observed with hst / cos .
i would like to thank an anonymous referee for suggesting the use of the h@xmath14 lines to set the zero of the wavelength scale .
this work has been partly funded by the ministerio de economia y competitividad of spain through grant aya2011 - 29754-c03-c01 .
the narrow emission component of the he ii line seems to be systematically red - shifted with respect to the h@xmath14 lines .
the shifts are small ( @xmath72 or 29 km s@xmath2 ) as shown in table 8 .
the zero of the wavelength scale has been set with the ( r(3 ) 1 - 7 ) 1489.636 and ( p(5 ) 1 - 7 ) 1504.845 h@xmath14 lines because they were observed in most of the stars .
these lines are far from the 1640 he ii line and , in principle , small uncertainties in the wavelength calibration could drive to these shifts . however , as shown in figures 14 and 15 , this is not the case .
the p(17)3 - 9 line is plotted for the whole sample in figure 14 though only is clearly observed in : dm tau , ux tau a , aa tau , gm aur and dr tau ( the h@xmath14 emission from rw aur can not be used for this purpose ) . in all cases ,
the h@xmath14 transition is at rest . in figure 15 , the c i[uv1 ]
multiplet is plotted .
it is observed as narrow emission lines in dn tau , dm tau , ux tau a and df tau . in all cases ,
the lines are at rest wavelength .
this red - shift is observed both in cttss and intermediate objects ( it can not be measured in wttss because of the lack of h@xmath14 emission ) . thus , unless the p(14 ) 3 - 10 h@xmath14 transition , that it is blended with the he ii line , is unusually strong , it must be concluded that the narrow emission component of the heii line is red - shifted in most sources .
note however , that the blending with the broad component could produce an apparent redshift if this broad component is asymmetric .
ardila , d. r. , basri , g. , walter , f. m. , et al .
2002 , , 566 , 1100 ayres , t. r. , fleming , t. a. , simon , t. , et al . , 1995 , , 96 , 223 ayres , t.r . , 2005 , proceedings of the 13th cambridge workshop on cool stars , stellar systems and the sun , held 5 - 9 july , 2004 in hamburg , germany . edited by f. favata , g.a.j .
hussain , and b. battrick .
esa sp-560 , european space agency , p.419 beristain , g. , edwards , s. , kwan , j. , 2001 , , 551 , 1037 bertout , c. , siess , l. and cabrit , s. , 2007 , astronomy & astrophysics , 473 , l21l24 brown , a. de m. ferraz , m. c. and jordan , c. , 1984 , royal astronomical society , 207 , 831 clarke , c.j . , bouvier , j. , 2000 , m.n.r.a.s . , 319 , 457 dantona , f. , & mazzitelli , i. 1997 , , 68 , 807 damiani , f. , micela , g. , sciortino , s. et al .
1995 , , 446 331 ferland , g. , j. , 1996 , _ hazy , a brief introduction to cloudy _
, university of kentucky , department of physics and astronomy internal report .
fischer , w. , kwan , j. , edwards , s. et al . , 2008 , , 687 , 1117 france , k. , schindhelm , e. , herczeg , g. j. et al . 2012 , , 756 , 171 glassgold , a. e. , feigelson , e. d. , montmerle , t. , 2000 , protostars and planets iv ( book - tucson : university of arizona press ; eds mannings , v. , boss , a.p . , russell , s. s. ) , p. 429
gmez de castro , a.i .
& lamzin , s. , 1999 , monthly notices of the r.a.s .
, 304 , l41 gmez de castro , a. i. , & verdugo , e. 2001 , , 548 , 976 gmez de castro , a. i. , & verdugo , e. 2003 , , 597 , 443 gmez de castro , a.i . ,
verdugo , e. , ferro - fontn , c. , 2003 , the future of cool - star astrophysics : 12th cambridge workshop on cool stars , stellar systems , and the sun ( 2001 july 30 - august 3 ) , eds .
a. brown , g.m .
harper , and t.r .
ayres , ( university of colorado ) , p. 40 - 49 .
gmez de castro , a.i . ,
ferro - fontn , c. , 2005 , , 362 , 569 gmez de castro , a. i. , & verdugo , e. 2007 , , 654 , l91 gmez de castro , a. i. 2009 , , 698 , l108 gmez de castro , a.i .
, von rekowsky , b. , 2011 , , 411 , 849 gmez de castro , a.i . , marcos - arenal , p. , 2012
, , 749 , 190 gmez de castro , a.i .
2013 , planets , stars and stellar systems , editor - in - chief : oswalt , t. d. mclean , i.s . ; bond , h.e . ; french , l. ; kalas , p. ; barstow , m.a . ; gilmore , g.f . ; keel , w.c .
( eds . ) , springer , isbn 978 - 90 - 481 - 8817 - 8 green , j. c. , froning , c. s. , osterman , s. et al .
2012 , , 744 , 60 gdel , m. , briggs , k. r. , arzner , k. , et al .
2007 , , 468 , 353 guenther e.w . , lehman h. , emerson j.p .
, , 341 , 768 , 1999 hartmann , l. , 2009 , protostellar jets in context , by kanaris tsinganos , tom ray , matthias stute .
astrophysics and space science proceedings series .
berlin : springer , pp.23 - 32 hartmann , l. , avrett , e. edwards , s. , 1982 , , 261 , 279 herczeg , g. j. , linsky , j. l. , valenti , j. a. , 2002 , , 572 , 310 hirth , g.a .
, mundt , r. , solf , j. , 1997 , astronomy & astrophysics sup .
, 126 , 437 hulamo , n. , gmez de castro , a.i . , franqueira , m. , 1998 , ultraviolet astrophysics beyond the iue final archive , esa publications division , esa sp , 413 , 121 ingleby , l. calvet , n. , bergin , e. et al . , 2009 , , 703 , l137 johns - krull , c. m. , valenti , j. a. , koresko , c. , 1999 , , 516 , 900 johns - krull , c. m. , valenti , j. a. , & linsky , j. l. 2000 , , 539 , 815 johns - krull , c. m. , valenti , j. a. , & saar , s. h. 2004 , , 617 , 1204 johns - krull , c. m. , 2007 , , 664 975 johns - krull , c.m . , _ protostellar jets in context , astrophysics and space science proceedings series _ , 33 , 2009 kenyon , s. j. and hartmann , l. , 1995 , , 101 , 117 kivelson , m.g . , russel , c.t . , 1995 , _ introduction to space physics _
, cambridge univ . press ,
p. 330 kriss , g.a .
2011 , improved medium resolution line spread functions for cos fuv spectra , technical report lamzin , s. , 1998 , astronomy reports , 42 , 322 lemmens , a. f. p. , rutten , r. g. m. , zwaan , c. 1992 , , 257 , 671 lima , g.h.r.a .
, alencar , s.h.p . ,
calvet , n. et al .
, 2010 , , 522 , 104l linsky , j. l. , bornmann , p. l. , carpenter , k. g. et al . 1982 , , 260 , 670 muzerolle , j. , calvet , n. , hartmann , l. , 1998 , , 492 , 743 muzerolle , j. , calvet , n. , hartmann , l. , 2001 , , 550 , 994 neuhauser , r. , sterzik , m.f . ,
schmidt , j.h.m.m .
et al 1995 , , 297 , 391 orlando , s. , sacco , g. g. , argiroffi , c. et al . , 2009 , ,
510 , 71 pedersen , a. , gmez de castro , a.i .
, 2011 , , 740 , 77 penston , m. & lago , m.t.v.t . ,
1983 , monthly notices of the r.a.s .
, 202 , 77 petrov , p.p .
, gahm , g. , f. , stempels , h.c .
et al . , 2011 , , 535 , 6 romanova , m. m. , ustyugova , g. v. , koldoba , a. v. et al . , 2004 , , 610 , 920 romanova m. m. , ustyugova , g. v. , koldoba , a. v. et al . 2012 , monthly notices of the r.a.s .
, 421 , 63 sachs , l. , 1982 , applied statistics , a handbook of techniques , springer series in statistics , springer - verlag , ny ( isbn:0 - 387 - 90976 - 1 ) siess , l. , dufour , e. , & forestini , m. 2000 , , 358 , 593 valencic , l. a. , clayton , g. c. , & gordon , k. d. 2004 , , 616 , 912 yang , h. , herczeg , g. j. , linsky , j. l. , et al , 2012 , , 744 , 121 line of the ttss studied in this work .
the zero of the wavelength scale has been set using h@xmath14 lines as reference ( see text ) .
notice the asymmetry and broadening of rw aur profile that prevents its use for this purpose .
fluxes are given in units of 10@xmath73 erg s@xmath2 @xmath65 @xmath2 ( f@xmath74),scaledwidth=100.0% ] profiles of the ttss in taurus observed with hst .
the rest wavelength of the two lines in the multiplet is marked for reference ( see also table 1 ) .
fluxes are given in units of 10@xmath73 erg s@xmath2 @xmath65 @xmath2 ( f@xmath74).,scaledwidth=100.0% ] line of the ttss studied in this work .
the zero of the wavelength scale has been set using h@xmath14 lines as reference ( see text ) .
notice the asymmetry and broadening of the rw aur profile that prevents its use for this purpose .
fluxes are given in units of 10@xmath73 erg s@xmath2 @xmath65 @xmath2 ( f@xmath74).,scaledwidth=100.0% ] profiles ( black ) over - plotted on the respective he ii profiles ( blue ) .
the o iii ] profiles have been re - scaled for comparison .
the scaling factor is indicated in the figure for every profile.,scaledwidth=80.0% ] 1489 and 1389 lines are plotted in the bottom panel .
the dispersion of the n v line is larger than the measured in the h@xmath14 1389 line in all sources .
however the dispersion of h@xmath14 1489 is comparable to the observed in the he ii line in wttss and transitional objects . the average @xmath75(h@xmath14(1489 ) )
= 31@xmath765 km s@xmath2 is represented in the top panel ; the solid line represents the average and the error band is marked with dashed lines .
circled sources have a significant contribution from the ldc to the he ii flux.,scaledwidth=70.0% ] , f@xmath77 and l@xmath78 as in table 3 .
circled sources have a significant contribution from the ldc to the he ii flux .
_ bottom : _
a@xmath43 , f@xmath77 and l@xmath78 as in table 2 of hao yang et al .
, title="fig:",scaledwidth=40.0% ] + , f@xmath77 and l@xmath78 as in table 3 .
circled sources have a significant contribution from the ldc to the he ii flux .
_ bottom : _
a@xmath43 , f@xmath77 and l@xmath78 as in table 2 of hao yang et al .
title="fig:",scaledwidth=40.0% ] profiles of the ttss in taurus observed with hst .
the rest wavelength of the lines in the multiplet is marked for reference .
fluxes are given in units of 10@xmath73 erg s@xmath2 @xmath65 @xmath2 ( f@xmath74).,scaledwidth=100.0% ] llllllll object & spectral & l@xmath79 & a@xmath43 & vsin(i ) & age & period & l@xmath78 + & type & ( l@xmath71 ) & ( mag . ) & ( km / s ) & @xmath80 ( yr ) & ( d ) & 10@xmath81 erg s@xmath2 + su aur & g1 & 7.8 & 0.90 & 66.2 & 6.80@xmath76 0.08 & 3.5 & 11.641 + t tau & k0 & 7.8 & 1.8 & 20.1 & & 2.8 & 9.395 + lkca 19 & k0 & 1.56 & 0 & 19 & 7.17@xmath76 0.01 & 2.24 & + gm aur & k3 & 1.2 & 1.1 & 12.4 & 6.9@xmath76 0.2 & 12 & 0.69 + rw aur & k4 & 1.72 & 0.32 & 17.2 & 7.20@xmath76 0.11 & 5.4 & + hn tau & k5 & 0.7 & 0.52 & 52.8 & & & 0.18 + ux tau & k5 & 1.29 & 0.0 & 26 & [email protected] & & + hbc 427 & k5bin & 2 & 0.6 & & & & 6.3 + aa tau & k7 & 1.1 & 1.4 & 11 & & 8.2 & 1.039 + dr tau & k7 & 1.7 & 1.0 & @xmath82 & 5.92@xmath76 0.2 & 7.3 + lkca 4 & k7 & 0.73 & 0.69 & 26.1 & 6.43@xmath76 0.25 & 3.38 & 1.99 + v836 tau & k7 & 0.8 & 1.1 & @xmath83 & & 6.76 + de tau & m0 & 1.2 & 1.1 & 10.0 & & 7.6 & 0.025 + ip tau & m0 & 0.7 & 0.9 & 11.0 & 6.6@xmath76 0.2 & 3.3 & + df tau & m1,m3.5&0.47,0.53 & 0.04 & 16.1 & 6.28@xmath76 0.17 & 8.5 & + dm tau & m1 & 0.35 & 0.6 & 10 & 6.87@xmath76 0.34 & & 1.181 + dn tau & m1 & 1.2 & 0.8 & 8.1 & 6.15@xmath76 0.11 & 6.6 & 1.072 + uz tau & m1 & 0.31 & 1.0 & 15.9 & & & 0.736 + llllll object & f(he ii ) & f@xmath84(o iii ] ) & f ( n v ) & f@xmath85 ( h@xmath14 ) & f@xmath86 ( h@xmath14 ) + & + su aur & 2.05@xmath76 0.13 & 0.46@xmath76 0.18 @xmath87 & 0.53@xmath76 0.05 @xmath88 & 2.82 @xmath76 0.14 & 0.41 @xmath76 0.04 + t tau & 4.87@xmath76 0.96 & 1.14@xmath76 0.41 @xmath87 & & & + lkca 19 & 1.43@xmath76 0.10 & & 0.22@xmath76 0.03 @xmath89 & & + gm aur & 6.22@xmath76 0.19 & & 2.17@xmath76 0.22 @xmath89 & 8.76 @xmath76 0.67 & 1.08 @xmath76 0.06 + rw aur & 0.66@xmath76 0.12 & 2.10@xmath76 0.43 @xmath90 & 0.66@xmath76 0.11 @xmath91 & 6.49 @xmath76 0.91 & 3.26 @xmath76 0.15 + hn tau & 1.33@xmath76 0.26 & 1.24 @xmath76 0.26 @xmath92 & 0.27@xmath76 0.03 @xmath88 & 4.45 @xmath76 0.23 & 1.04 @xmath76 0.07 + ux tau & 2.25@xmath76 0.12 & & 0.88@xmath76 0.08 @xmath88 & 4.29 @xmath76 0.21 & 0.85 @xmath76 0.03 + hbc 427 & 0.77@xmath76 0.10 & & 0.08@xmath76 0.02 @xmath89 & & + aa tau & 2.93@xmath76 0.06 & & 0.14 @xmath76 0.01@xmath88 & 9.12 @xmath76 0.24 & 1.16 @xmath76 0.05 + dr tau & 0.93@xmath76 0.15 & 2.14 @xmath76 0.49 @xmath90 & 0.25 @xmath76 0.05@xmath88 & 2.33 @xmath76 0.34 & 0.52 @xmath76 0.09 + lkca 4 & 0.54@xmath76 0.06 & & 0.07 @xmath76 0.01@xmath89 & & + v836 tau & 0.44@xmath76 0.03 & & 0.07 @xmath76 0.01@xmath88 & 0.45 @xmath76 0.05 & 0.15 @xmath76 0.01 + de tau & 0.87@xmath76 0.16 & 0.27 @xmath76 0.11 @xmath92 & 0.35 @xmath76 0.05@xmath91 & 2.76 @xmath76 0.28 & 0.45 @xmath76 0.05 + ip tau & 1.33@xmath76 0.04 & & 0.08 @xmath76 0.02@xmath88 & 0.17 @xmath76 0.05 & 0.13
@xmath76 0.02 + df tau & 7.32@xmath76 0.20 & 1.42 @xmath76 0.47 @xmath90 & 1.07
@xmath76 0.07@xmath91 & 15.9@xmath76 0.6 & 0.94 @xmath76 0.07 + dm tau & 7.28@xmath76 0.18 & & 0.68 @xmath76 0.10@xmath88 & 6.72 @xmath76 0.31 & 0.61 @xmath76 0.03 + dn tau & 6.53 @xmath760.06 & & 0.88 @xmath76 0.08@xmath89 & 0.88 @xmath76 0.13 & 0.15
@xmath76 0.04 + uz tau & 1.06@xmath76 0.34 & 0.43 @xmath76 0.33 & & & 0.55 @xmath76 0.02 + lllll stars & @xmath75(h@xmath14 1339 ) & @xmath75(h@xmath14 1489 ) & @xmath75(n v 1238 ) & @xmath75(he ii ) + & ( km s@xmath2 ) & ( km s@xmath2 ) & ( km s@xmath2 ) & ( km s@xmath2 ) + aatau & 52.38 & 33.32 & 89.66 & 40.19 + detau & 39.61 & 26.72 & 79.97 & 68.05 + dftau & 52.38 & 36.71 & 116.3 & 48.66 + dmtau & 48.51 & 21.8 & 138.1 & 88.85 + dntau & 48.51 & 39.81 & 36.35 & 39.7 + drtau & 34.3 & 34.49 & 70.27 & 70.34 + gmaur & 65.69 & 30.85 & 123.6 & 62.25 + hntau & 56.01 & 29.54 & 130.9 & 87.04 + iptau & 19.81 & 33.32 & 72.7 & 32.58 + suaur & 28.01 & 26.72 & 143 & 63.9 + uxtau & 34.3 & 23.55 & 99.35 & 58.41 + v836tau & 34.3 & 32.1 & 67.85 & 33.26 + lkca19 & & & 33.93 & 41.47 + lkca4 & & & 38.77 & 36.05 + llllll star & t@xmath93 & t@xmath14 & em@xmath93 & em@xmath14 & em@xmath93/em@xmath14 + & ( mk ) & ( mk ) & 10@xmath94 & 10@xmath94 & + t tau & 4.52 & 23.65 & 33.86 & 57.85 & 0.585 + su aur & 5.22 & 23.30 & 87.01 & 32.45 & 2.681 + hbc 427 & 9.04 & 28.06 & 14.98 & 15.07 & 0.994 + | the heii(1640 ) line and the resonance doublet of nv(uv1 ) provide a good diagnostic tool to constrain the excitation mechanism of hot ( @xmath0k ) atmospheric / magnetospheric plasmas in t tauri stars . making use of the data available in the hubble space telescope archive
, this work shows that there are , at least , two distinct physical components contributing to the radiation in these tracers : the accretion flow sliding on the magnetosphere and the atmosphere .
the nv profiles are symmetric and at rest with respect to the star , in most sources . the velocity dispersion of the profile increases from non - accreting ( @xmath1 km s@xmath2 ) to accreting ( @xmath3 km s@xmath2 ) ttss suggesting that the macroturbulence field in the line formation region decreases as the stars approach the main sequence .
evidence of the nv line being formed in a hot solar - like wind has been found in rw aur , hn tau and aa tau . the heii profile has a strong narrow component that dominates the line flux ; the dispersion of this component ranges from 20 to 60 km s@xmath2 .
the source of this radiation is uncertain .
current data suggest that both accretion shocks and atmospheric emission might contribute to the line flux . in some sources
, the he ii line shows a broad and redwards shifted emission component often accompanied by semiforbidden o iii ] emission that has a critical electron density of @xmath4 @xmath5 . in spite of the different origins ( inferred from the kinematics of the line formation region ) , nv and heii fluxes
are strongly correlated , with the only possible exception of some of the heaviest accretors . |
the important task of quantum mechanics is to find the exact bound - states solution of the schrdinger equation for certain potentials of physical interest . generally speaking
, there are a few main traditional methods to study the exact solutions of quantum systems like the coulomb , the harmonic oscillator [ 1,2 ] , the pseudoharmonic [ 3,4 ] and the kratzer [ 4,5 ] potentials . additionally , in order to obtain the bound - states solutions of central potentials , one has to resort to numerical techniques or approximation schemes .
for many of the quantum mechanical systems , most popular approximation methods such as shifted @xmath2 expansion [ 6 ] , perturbation theory [ 7 ] , path integral solution [ 8 ] , algebraic methods with the susyquantum mechanics method and the idea of shape invariance , further closely with the factorization mrthod [ 9 ] , exact quantization rule [ 10,11 ] , the conventional nikiforov and uvarov ( nu ) method [ 12 - 25 ] .
some of these methods have drawbacks in applications . although some other methods give simple relations for the eigenvalues , however , they lead to a very complicated relations for the eigenfunctions .
the study of exact solutions of the schrdinger equation for a class of non - central potentials with a vector potential and a non - central scalar potential is of considerable interest in quantum chemistry [ 26 - 35 ] . in recent years
, numerous studies [ 36 - 40 ] have been made in analyzing the bound states of an electron in a coulomb field with simultaneous presence of aharanov - bohm ( ab ) [ 41 ] field , and/or a magnetic dirac monopole [ 42 ] , and aharanov - bohm plus oscillator ( abo ) systems . in most of these studies ,
the eigenvalues and eigenfunctions are obtained by means of seperation of variables in spherical or other orthogonal curvilinear coordinate systems .
the path integral for particles moving in non - central potentials is evaluated to derive the energy spectrum of this system analytically [ 43 ] .
in addition , the idea of susy and shape invariance is also used to obtain exact solutions of such non - central but seperable potentials [ 44 ] .
very recently , the conventional nu method has been used to give a clear recipe of how to obtain an explicit exact bound - states solutions for the energy eigenvalues and their corresponding wave functions in terms of orthogonal polynomials for a class of non - central potentials [ 45 ] .
recently , chen and dong [ 46 ] found a new ring - shaped potential and obtained the exact solution of the schrdinger equation for the coulomb potential plus this new ring - shaped potential which has possible applications to ring - shaped organic molecules like cyclic polyenes and benzene .
recently , cheng and dai [ 47 ] , proposed a new potential consisting from the modified kratzer s potential [ 48 ] plus the new proposed ring - shaped potential in [ 46 ] .
they have presented the energy eigenvalues for this proposed exactly - solvable non - central potential in three dimensional @xmath3i.e .
, @xmath4-schrdinger equation through the nu method .
the two quantum systems solved by refs [ 46,47 ] are closely relevant to each other as they deal with a coulombic field interaction except for a slight additional centrifugal potential barrier acts as a repulsive core which is for any arbitrary angular momentum @xmath5 prevents collapse of the system in any dimensional space due to the slight perturbation to the original angular momentum barrier . in a very recent work [ 49 ] ,
we have given a clear recipe of how to obtain analytically the exact energy eigenvalues and the corresponding wave functions of the @xmath0-dimensional space schrdinger equation with the proposed potential in [ 47 ] by means of the conventional nu method .
the conventional nikiforov - uvarov ( @xmath6 ) method [ 12 ] , which received much interest in recent years , has been introduced for solving schrdinger equation [ 13 - 21 ] , klein - gordon [ 22,23 ] , dirac [ 24 ] and salpeter [ 25 ] equations .
the purpose of this work is to follow parallel solution to [ 47 ] and give exact complete solutions to the @xmath0-dimensional schrdinger equation with pseudo - coulomb potential plus ring - shaped potential , a coulombic - like potential with an additional centrifugal barrier , for any arbitrary @xmath7-states using the conventional nikiforov - uvarov method .
for some appropriate choices of parameters , the obtained expressions of energy eigenvalues and their corresponding wave functions for this alternative form of the potential coincide with the @xmath0-@xmath8ensional results given in [ 49 ] and with the three - dimensions results given in [ 47 ] in the limiting case of @xmath9 this work is organized as follows : in section [ bc ] , we shall briefly introduce the basic concepts of the nu method .
section [ es ] is mainly devoted to the exact solution of the schrdinger equation in @xmath0-dimensions for this quantum system by means of the @xmath6 method .
finally , the relevant results are discussed in section [ rac ] .
the nu method is based on reducing the second - order differential equation to a generalized equation of hypergeometric type [ 12 ] . in this sense , the schrdinger equation , after employing an appropriate coordinate transformation @xmath10 transforms to the following form : @xmath11 where @xmath12 and @xmath13 are polynomials , at most of second - degree , and @xmath14 is a first - degree polynomial . using a wave function , @xmath15 of the simple ansatz : @xmath16 reduces ( 1 ) into an equation of a hypergeometric type @xmath17 where @xmath18 @xmath19 and @xmath20 is a parameter defined as @xmath21 the polynomial @xmath22 with the parameter @xmath23 and prime factors show the differentials at first degree be negative .
it is worthwhile to note that @xmath20 or @xmath24 are obtained from a particular solution of the form @xmath25 which is a polynomial of degree @xmath26 further , the other part @xmath27 of the wave function ( 2 ) is the hypergeometric - type function whose polynomial solutions are given by rodrigues relation @xmath28 , \ ] ] where @xmath29 is the normalization constant and the weight function @xmath30 must satisfy the condition [ 12 ] @xmath31 the function @xmath32 and the parameter @xmath20 are defined as @xmath33 @xmath34 in principle , since @xmath35 has to be a polynomial of degree at most one , the expression under the square root sign in ( 9 ) can be arranged to be the square of a polynomial of first degree [ 12 ] .
this is possible only if its discriminant is zero . in this case ,
an equation for @xmath36 is obtained . after solving this equation , the obtained values of @xmath36
are substituted in ( 9 ) .
in addition , by comparing equations ( 6 ) and ( 10 ) , we obtain the energy eigenvalues .
the pseudo - coulomb potential plus ring - shaped potential in spherical coordinates is defined as [ 47 ] @xmath37 where @xmath38 @xmath39 and @xmath40 are positive real constants .
the potential in ( 11 ) coincides with the potential introduced by cheng - dai [ 47 ] if the parameters take their proper values and also turns to be the modified kratzer potential in the limiting case of @xmath41 [ 48]@xmath42 in fact the energy spectrum for this potential can be obtained directly by considering it as special case of the general non - central seperable potentials [ 45 ] .
our aim is to derive analytically the energy spectrum for a moving particle in the presence of a potential ( 11 ) in a very simple way .
the @xmath0-dimensional space schrdinger equation in spherical polar coordinates written for potential ( 11 ) takes the form [ 1,6 ] @xmath43 \psi ( r,\theta , \varphi ) \ ] ] @xmath44 \psi ( r,\theta , \varphi ) = 0,\ ] ] where @xmath45 being the reduced mass of the two particles and @xmath46 being the total wave function separated as follows @xmath47 on substituting equation ( 13 ) into ( 12 ) leads to a set of second - order differential equations : @xmath48 @xmath49 h(\theta ) = 0,\ ] ] @xmath50 r(r)+\frac{2\mu } { \hbar ^{2}}\left [ e+\frac{a}{r}-\frac{b}{r^{2}}-c\right ] r(r)=0.\ ] ] the solution in ( 16 ) is periodic and must satisfy the period boundary condition @xmath51 from which we obtain @xmath52 further , equation ( 15 ) representing the angular wave equation takes the simple form @xmath53 h(\theta ) = 0,\ ] ] which has already been solved before in our previous work [ 49 ] .
we present the final results for the angular part of the wave function as @xmath54 where @xmath55 with @xmath56 and @xmath57 on the other hand , we are left to solve equation ( 16 ) .
after lengthy , but straightforward , calculations , equation ( 16 ) , representing the radial wave equation can be rewritten as [ 6 ] : @xmath58 g(r)=0,\ ] ] where @xmath59 the two particles in equation ( 18 ) interacting via coulombic - like field have a slight change in the angular momentum barrier acts as a repulsive core which for any arbitrary @xmath5 prevents collapse of the system in any space dimension due to the additional centrifugal potential barrier .
thus , equation ( 23 ) has to be solved by using the nu method in the next subsection .
the aim of this subsection is to solve the problem with a different radial separation function @xmath60 in any arbitrary dimensions .
we now study the bound - states ( real ) solution @xmath61 of equation ( 23 ) . letting @xmath62 and substituting these expressions in equation ( 23 ) ,
one gets @xmath63 to apply the conventional nu - method , equation ( 26 ) is compared with ( 1 ) and the following expressions are obtained @xmath64 substituting the above expressions into equation ( 9 ) gives @xmath65 according to this conventional method , the expression in the square root be the square of a polynomial .
thus , the two roots @xmath36 can be readily obtained as @xmath66 in view of that , we arrive at the following four possible functions of @xmath67@xmath68 & \text{\ for } k_{1}=\alpha + \varepsilon \sqrt{4\gamma + 1 } , \\ \frac{1}{2}-\left [ \varepsilon r+\frac{1}{2}\sqrt{4\gamma + 1}\right ] & \text{\ for } k_{1}=\alpha + \varepsilon \sqrt{4\gamma + 1 } , \\
\frac{1}{2}+\left [ \varepsilon r-\frac{1}{2}\sqrt{4\gamma + 1}\right ] & \text{\ for } k_{2}=\alpha -\varepsilon \sqrt{4\gamma + 1 } , \\
\frac{1}{2}-\left [ \varepsilon r-\frac{1}{2}\sqrt{4\gamma + 1}\right ] & \text{\ for } k_{2}=\alpha -\varepsilon \sqrt{4\gamma + 1}. \end{array } \right.\ ] ] the correct value of @xmath69 is chosen such that the function @xmath70 given by equation ( 5 ) will have negative derivative [ 12 ] .
so we can select the physical values to be @xmath71 , \ ] ] which yield @xmath72 using equations ( 6 ) and ( 10 ) , the following expressions for @xmath20 are obtained , respectively , @xmath73 @xmath74 so we can obtain the energy eigenvalues as @xmath75 where @xmath76 with @xmath7 defined in ( 21 ) .
therefore , the final energy spectra in equation ( 35 ) take the following coulombic - like form [ 7 ] @xmath77 where @xmath78 , \ ] ] is simply obtained by means of substituting equation ( 36 ) into ( 35 ) .
\(i ) if @xmath79 @xmath80 @xmath81 and @xmath82 ( modified kratzer potential - plus - ring - shaped potential ) , the equation ( 35 ) , with the help of equation ( 36 ) , is transformed into the following form @xmath83 and it is consistent with ref .
[ 47 ] .
\(ii ) if @xmath79 @xmath80 @xmath84 and @xmath41 ( modified kratzer potential)@xmath85 the equation ( 35 ) is transformed into the form @xmath86 and it is consistent with ref . [ 48 ] .
\(iii ) if @xmath79 @xmath87 and @xmath88 the equation ( 35 ) is transformed into the form @xmath89 and it is consistent with ref . [ 49 ] .
\(iv ) if @xmath90 @xmath91 and @xmath92 the equation ( 35 ) is transformed into the form @xmath93 } \right ) ^{2}},\ ] ] where @xmath94 and it is consistent with ref . [ 46 ] .
let us now turn attention to find the radial wavefunctions for this potential . using @xmath95 @xmath69 and @xmath96 in equations ( 4 ) and ( 8)
, we find @xmath97 @xmath98 then from equation ( 7 ) , we obtain @xmath99 and the wave function @xmath60 can be written in the form of the generalized laguerre polynomials as @xmath100 where @xmath101 .\ ] ] finally , the radial wave functions of the schrdinger equation are obtained @xmath102 where @xmath103 with @xmath104 is given in equation ( 38 ) and @xmath105 is the normalization constant to be determined below .
using the normalization condition , @xmath106 and the orthogonality relation of the generalized laguerre polynomials , @xmath107 ^{2}dz=\frac{(2n+\eta + 1)(n+\eta ) ! } { n!},$ ] we have @xmath108 letting @xmath82 and @xmath109 the equation ( 48 ) transforms into equation ( 20 ) in ref.[46 ] .
therefore , we may express the normalized total wave functions as @xmath110 @xmath111
in this paper , the schrdinger equation in any arbitrary dimensions has been solved for its exact bound - states with a a pseudo - coulomb potential plus ring - shaped potential by means of a simple conventional nu method .
the analytical expressions for the total energy levels of this system is found to be different from the results obtained for the modified kratzer s potential in [ 48 ] and also more general than the one obtained recently in @xmath0-dimensions [ 49 ] .
hence , the noncentral potentials treated in [ 45 ] can be introduced as perturbation to the central potentials by adjusting the strength of the coupling constant @xmath40 in the non - central part in terms of the coupling constants of the central part .
in addition , the angular part , the radial part and then the total wave functions are also found .
thus , the schrdinger equation with a new non - central but seperable potential has also been studied ( cf .
[ 45,49]and the references therein ) .
this method is very simple and useful in solving other complicated systems analytically without given a restiction conditions on the solution of some quantum systems as the case in the other models . finally , we point out that these exact results obtained for this new proposed form of the potential ( 11 ) may have some interesting applications in the study of different quantum mechanical systems , atomic and molecular physics .
ikhdair and r. sever _ , _ z. phys .
c 56 ( 1992 ) 155 ; c 58 ( 1993 ) 153 ; d 28 ( 1993 ) 1 ; hadronic j. 15 ( 1992 ) 389 ; int
phys . a 18 ( 2003 )
4215 ; a 19 ( 2004 ) 1771 ; a 20 ( 2005 ) 4035 ; a 20 ( 2005 ) 6509 ; a 21 ( 2006 ) 2191 ; a 21 ( 2006 ) 3989 ; a 21 ( 2006 ) 6699 ; int .
e ( in press ) ( preprint hep - ph/0504176 ) ; s. ikhdair _
et al , _ tr . j. phys . 16 ( 1992 ) 510 ; 17 ( 1993 ) 474 .
ikhdair and r. sever , int .
phys . a 21 ( 2006 ) 6465 ; j. math .
( doi 10.1007/s10910 - 006 - 9080 - 2 ) ; j. math .
( preprint quant - ph/0604078 ) ; j. mol .
structure : theochem ( doi:10.1016/j.theochem.2007.01.019 ) . | we present analytically the exact energy bound - states solutions of the schrdinger equation in @xmath0-dimensions for an alternative ( often used ) pseudo - coulomb potential - plus- ring - shaped potential of the form @xmath1 by means of the conventional nikiforov - uvarov method .
we give a clear recipe of how to obtain an explicit solution to the radial and angular parts of the wave functions in terms of orthogonal polynomials .
the total energy of the system is different from the pseudo - coulomb potential because of the contribution of the angular part .
the general results obtained in this work can be reduced to the standard forms given in literature .
keywords : energy eigenvalues and eigenfunctions , coulomb potential , ring - shaped potential , non - central potentials , nikiforov and uvarov method .
pacs numbers : 03.65.-w ; 03.65.fd ; 03.65.ge . |
as the study of galaxy formation comes more and more to require a reliable description of star - gas interactions and feedback on relatively small spatial scales , it is forced towards questions that overlap with fundamental issues in star formation theory .
conversely , as evidence accumulates that most stars today in environments ranging from the disk of the milky way to circumnuclear starbursts are born not in isolation but in groups , the study of star formation in general seems increasingly to demand a focus on larger scales than might previously have been considered .
although these two disciplines have not yet converged
the smallest scales currently of importance for galaxy formation can still be one or two orders of magnitude greater than the `` large '' scales in star formation they will necessarily continue to approach one another as further progress is made .
it is therefore important to identify and understand those areas of truly common interest , on the largest stellar scales and the smallest galactic ones . in many ways ,
the systems of old globular clusters in galaxy halos are positioned just at this interface ; they can potentially be exploited to constrain aspects of both local and global star - formation processes , in both protogalactic and present - day settings .
it has , indeed , long been recognized that the global properties of globular cluster systems ( gcss ) might be used to good effect as probes of galaxy formation and evolution .
( this is especially true of bright early - type galaxies , whose gcss are more populous and better studied than those in typical spirals like the milky way . ) for example , following the suggestion that large fractions of the gcss in ellipticals may have formed in major merger events ( e.g. , schweizer 1987 ; ashman & zepf 1992 ; zepf & ashman 1993 ) , and the related discovery of young super star clusters in systems like the antennae galaxies ( whitmore & schweizer 1995 ) , much recent discussion has centered on the interpretation of the distribution of metallicities ( or broad - band colors ) among e - galaxy globulars as clues to their hosts dynamical histories ( see geisler , lee , & kim 1996 ; forbes , brodie & grillmair 1997 ; ct , marzke , & west 1998 ; kissler - patig , forbes , & minniti 1998 ) .
a somewhat older approach has used just the total population of gcss , as a function of parent galaxy luminosity , hubble type , and local environment , to make arguments relating to issues such as biasing in a cold dark matter cosmology ( west 1993 ) ; the merger history of ellipticals ( harris 1981 ; van den bergh 1984 ; ashman & zepf 1992 ; forbes et al .
1997 ) ; the formation of cd galaxies and their envelopes ( mclaughlin , harris , & hanes 1993 , 1994 ) ; star formation in galaxies at the centers of cluster cooling flows ( harris , pritchet , & mcclure 1995 ) ; the influence of early galactic winds on the evolution of giant ellipticals ( harris , harris , & mclaughlin 1998 ) ; and even larger - scale aspects of galaxy clusters such as populations of intergalactic stars ( e.g. , west et al . 1995 ; harris et al .
yet a third characteristic of gcss their radial distributions ( the projected number density of globulars around a galaxy as a function of galactocentric radius)has also served as the basis for discussions of elliptical galaxy formation ( e.g. , harris 1986 ; ashman & zepf 1992 ; kissler - patig 1997 ; van den bergh 1998 ) .
rather less developed , however , is a clear sense of the relevance of gcss to _ local _ star formation on smaller scales : what relation , if any , do the fundamental properties of globular cluster systems bear to the way stars form now , in the dense clumps within galactic giant molecular clouds ?
can the study of one of these phenomena possibly shed light on the other ?
larson ( 1988 , 1993 , 1996 ) , has emphasized the possible parallels between the basic pattern of local star formation and the birth of globular clusters within large protogalactic clouds in a scenario for the collapse of galaxies from clumpy initial conditions ( as in , e.g. , the classic picture of searle & zinn 1978 , and current models of hierarchical galaxy formation ) . a more quantitative discussion along these lines , including an exposition of links to several of the gross characteristics of gcss , is given by harris & pudritz ( 1994 ) .
most recently , mclaughlin & pudritz ( 1996 ) and elmegreen & efremov ( 1997 ) have focussed in detail on the globular cluster luminosity function ( the number of clusters per unit magnitude , from which may be derived the mass function @xmath11 for the entire system of globulars in a given galaxy ) in attempts to model this property of gcss using our current understanding of present - day star formation in molecular clouds and their clumps .
the work presented in this paper further explores the connections between the formation of stars , globular clusters , and galaxies , by concentrating on two of the primary attributes of gcss : their specific frequencies ( or total populations ) and radial distributions .
the next section discusses the significance of these gcs properties as they relate to general questions on the formation of stellar clusters . in 3 below , data are compiled from the literature , and a geometrical deprojection algorithm applied , to construct _ volume _ density profiles for the stars , the x - ray gas , and the globular clusters in three galaxies with well studied gcss : m87 ( the cd at the center of the virgo cluster ) , m49 ( giant elliptical in virgo ) , and ngc 1399 ( cd galaxy in the fornax cluster ) .
an intercomparison of the density profiles of the three halo components yields a point - by - point measure of the gcs contribution to the total luminous mass in each system , as a function of _ three - dimensional _ galactocentric radius , @xmath12 .
these gcs mass ratios , which are constant throughout each galaxy ( at large @xmath12 ) and the same in all three , are interpreted as an indication of the formation efficiency of globular clusters . in
4 , these results are supplemented with broader considerations of the gcss in other early - type galaxies and in the milky way halo , and with a direct estimate for the relative formation rate of open clusters in the galactic disk , to argue for the existence of a _ universal efficiency , or probability , for the formation of a bound stellar cluster from a dense cloud of gas_. section 5 discusses the implications of this for a few specific issues in galaxy and gcs formation and evolution , and 6 concludes with a summary .
the possibility that globular cluster systems can be connected not only to galaxy formation , but to ongoing star formation as well , is suggested by the fact that this latter process operates largely in a _ clustered mode _ : by mass , most new stars in the milky way appear in groups within the largest clumps in molecular clouds , and only rarely in true isolation .
this is , by now , an established empirical fact ( e.g. , lada et al .
1991 ; zinnecker et al .
1993 ) , and it is easily seen ( patel & pudritz 1994 ) to be a direct consequence of the different power - law slopes in the mass function of molecular clumps ( @xmath13 , so that the largest clumps , which weigh in at @xmath14@xmath15 , contain most of the star - forming gas mass in any molecular cloud ) and the stellar initial mass function ( @xmath16 , putting most of the mass in young stars into @xmath17 objects ) .
similarly , young super star clusters , which tend to have sizes and masses that are generally comparable to the old globulars in the milky way , can account for as much as @xmath1820% of the uv light in starburst galaxies ( meurer et al .
1995 ) .
this is , however , _ not _ to say that all , or even most , stars are born into bona fide clusters that exist as coherent dynamical units for any length of time ; on the contrary , it would appear that very few multiple - star systems actually form as gravitationally bound clusters .
( a quantification of `` very few '' is one of the goals of this paper . ) at some point during the collapse and fragmentation of a cluster - sized cloud of gas , the massive stars which it has formed will expel any remaining gas by the combined action of their stellar winds , photoionization , and supernova explosions . if the _ cumulative _ star formation efficiency of the cloud , @xmath19 , is below a critical threshold when the gas is lost , then the blow - out carries away sufficient energy that the stellar group remaining is unbound , and disperses into the field .
the precise value of this threshold sfe depends on details of the dynamics and magnetic field in the gas cloud before its self - destruction , and on the timescale over which the massive stars dispel the gas ; but various estimates place it in the range @xmath200.5 ( e.g. , hills 1980 ; mathieu 1983 ; elmegreen & clemens 1985 ; verschueren 1990 ; lada , margulis , & dearborn 1984 )
. a complete theory of star formation must therefore be able to anticipate the final cumulative sfe in any single piece of gas with ( say ) a given mass and density , and thereby predict whether or not it will form a bound cluster .
no such theory yet exists .
more within reach are considerations of a statistical nature , in which a _ distribution _ of cumulative sfes ( and , thus , the fraction of clouds that achieve @xmath21 before self - destruction ) is calculated for an ensemble of star - forming clouds , in either a protogalactic or a present - day context ( see , e.g. , elmegreen 1983 ; elmegreen & clemens 1985 ; elmegreen & efremov 1997 ) .
but even here , the systematic effects of a range in the initial masses of the gaseous clouds , or of changes in the assumed star formation law ( which is probably the most outstanding `` free '' parameter here ) have not been explored fully .
it is therefore not clear that the dominant factor governing the formation of bound stellar clusters vs. unbound associations has been identified .
it is nonetheless possible to constrain any forthcoming theories of star and cluster formation with empirical evaluations of the probability that a cluster - sized cloud of gas is able to achieve a cumulative sfe of at least the critical 20%50% .
this probability which can equally well be interpreted as that fraction of an ensemble of massive star - forming clouds which manages to produce bound stellar systems is referred to here as the efficiency of cluster formation . as was indicated above , an estimate of this
will be taken first from a detailed investigation of the old globular cluster populations in three giant elliptical galaxies ; general arguments will then be used to show that the result also applies to globular cluster formation in other galaxies , and to the formation of disk clusters today .
a first ( and very rough ) assessment of the formation efficiency of globular clusters appeals to the most basic and most easily measured observable of a gcs : namely , its total population , @xmath22 , summed over all cluster luminosities and galactocentric positions .
this is usually compared to the integrated @xmath23-band luminosity of the parent galaxy s halo ( as a measure of its unclustered field star population ) through the so - called specific frequency ( harris & van den bergh 1981 ) : @xmath24 even though any observational derivation of @xmath3 necessarily involves ( possibly large ) extrapolations of data that directly count only the brightest clusters in some spatially limited subset of a gcs , these estimates tend usually to be surprisingly robust . to date , specific frequencies have been obtained for of order 100 galaxies , some 80% of which are early - type systems ( see the compilations of harris 1991 ; durrell et al . 1996 ; kissler - patig 1997 ; blakeslee 1997 ; blakeslee , tonry , & metzger 1997 ; harris et al .
1998 ) . given this bias in the observational database ,
some of what follows will make use of results that are really specific to elliptical galaxies ; spirals are brought more explicitly into the discussion in 4 below . specific frequency
is easily related to the total masses of a galaxy s gcs and of its unclustered halo stars , if both the mean mass of the individual globulars in the system and the mass - to - light ratio of the overall _ stellar _ component are known . as is discussed further in 3.1 below
, the basic distribution of globular cluster masses is quite similar from galaxy to galaxy , so that the mean cluster mass in the milky way gcs@xmath25serves as a reliable estimate for other systems as well .
meanwhile , the stellar mass - to - light ratios of large galaxies are essentially equal to the dynamical @xmath26 in their _ cores _ , which are baryon - dominated . as is also discussed below ( 4.1 ) ,
observations show that the core @xmath26 of hot galaxies increases with luminosity ; but , even though this turns out to be rather an important consideration in detail , it will suffice for now to use a single value of @xmath27 as a rough characterization of large ellipticals ( van der marel 1991 ; cf .
faber & gallagher 1979 ; binney & tremaine 1987 ) . the definition of @xmath3 in equation ( [ eq:21 ] ) may then be rewritten as ( cf .
zepf & ashman 1993 ; harris & pudritz 1994 ; harris et al .
1998 ) @xmath28 to be clear , it is stressed again that the mass - to - light ratio applied here refers to measurements in the cores of ellipticals , and is not meant to include any dark matter other than stellar remnants . in any event , an inversion of this last expression yields @xmath29 where @xmath30 is suggested by the studies cited above as a representative value for early - type galaxies ranging in luminosity from @xmath31 to @xmath32 .
of course , there is some uncertainty in this `` mean '' @xmath3 , and some intrinsic scatter about it .
moreover , the data suggest that @xmath3 varies systematically with @xmath9the simple scaling @xmath33 that corresponds to a constant specific frequency is not strictly obeyed particularly among dwarf ellipticals and between brightest cluster galaxies .
these issues have been discussed in detail in the literature , and they are revisited here , in 2.2 . with this caveat in mind , however , @xmath34 is useful as a crude , order - of - magnitude guide to a `` typical '' global gcs mass ratio in galaxy halos .
the advantage of working in terms of this _ mass _ ratio is that , whatever its precise value in any one system at the current epoch , it is expected to be a fairly well conserved quantity that has not changed drastically over the course of gcs and galaxy evolution in a hubble time .
_ this is true even though the total population @xmath35 of any gcs is bound to decrease over time _
, as individual globulars are disrupted by the cumulative effects of external gravitational shocks and internal two - body relaxation ( see , e.g. , aguilar , hut , & ostriker 1988 ; ostriker , binney , & saha 1989 ; capuzzo - dolcetta 1993 ; murali & weinberg 1997a , 1997b , 1997c ; gnedin & ostriker 1997 ; vesperini & heggie 1997 ; but cf .
portegies zwart et al .
these dynamical processes discriminate preferentially against low - mass and low - density globular clusters , and it turns out that even if large _ numbers _ of such objects are lost , no great change is effected in the total _ mass _ of the remaining gcs .
this conclusion rests on the fact that the mass functions of gcss particularly those in early - type galaxies are shallow enough that most of the total @xmath36 is contained in the most massive members of the system ; but these are the least susceptible to dynamical destruction .
the implication is that the total mass of a gcs may not change much over its lifetime ; thus , _ if _ the current stellar mass of a galaxy were a reliable indication of the total amount of gas that was initially available to form stars , then the presently observed ratio @xmath37 would serve as a rough , _ global _ estimate of the efficiency of globular cluster formation .
note that this reasoning also shows that galactic halos can not be built up from disrupted globulars ; see , e.g. , ashman & zepf ( 1992 ) and harris & pudritz ( 1994 ) .
( one of the main conclusions of this paper will be that , in fact , @xmath38 is _ not _ always an adequate measure of an initial gas supply , and that @xmath39 therefore can not always be equated directly to a cluster formation efficiency .
however , since most previous discussions of gcs specific frequencies do draw on this simple assumption at some level , it is of interest to develop some of its implications . )
a simple numerical example should emphasize the robustness of @xmath36 .
thus , consider a newly formed system of globulars whose masses are distributed according to a power - law mass function between some lower and upper limits : @xmath40 where @xmath41 and reasonable choices for @xmath42 and @xmath43 may be @xmath44 and @xmath45 .
( the precise values are not critical here . )
now let this gcs be eroded over @xmath2 yr , keeping in mind that each cluster in the system has a definite , mass - dependent lifetime against the dynamical mechanisms mentioned above . in a rough approximation to a combination of complicated processes , there is then a mass scale ( say @xmath46 ) above which few or no globulars will have disappeared by the current epoch .
thus , the gcs mass function at @xmath47 is effectively unchanged , even after a hubble time , from the initial distribution .
however , clusters with initial masses @xmath48 _ can _ have been removed from the system , and the gcs mass function below @xmath46 may differ significantly from its initial form .
if this low - mass end of the evolved @xmath11 can also be described by a power - law , then the full , present - day gcs mass function is just @xmath49 where @xmath50 for consistency .
observed gcss do , in fact , have mass functions that conform to roughly this sort of double power law ( harris & pudritz 1994 ; mclaughlin 1994 ; mclaughlin & pudritz 1996)although there is somewhat more structure than this at the highest cluster masses in @xmath11 , it can be safely ignored for the purposes of the present argument .
empirically , the mass scale @xmath46 is _ always _ about @xmath51 ; at this mass , the logarithmic slope of @xmath52 changes rather abruptly from @xmath53 to @xmath54 , with precise values that can vary slightly from galaxy to galaxy .
there are strong arguments to support the notion that observed gcs mass functions at @xmath47 are indeed accurate reflections of the formation distributions ( namely , the high - mass sides of @xmath55 in observed gcss show no significant variation with galactocentric radius inside a single system , and are quite similar though not identical from galaxy to galaxy ; see harris & pudritz 1994 ; mclaughlin & pudritz 1996 ; elmegreen & efremov 1997 ) .
in addition , some theoretical studies ( e.g. , surdin 1979 ; okazaki & tosa 1995 ; vesperini 1997 ; elmegreen & efremov 1997 ) have suggested that the shallower slopes at lower masses might be caused entirely by dynamical evolution from an initial @xmath56 , roughly as described above . in reality , this view could very well be too extreme : empirically , the system of young super star clusters in the antennae , which is not likely to have been severely affected by dynamical destruction , may have a mass function of the form ( [ eq:25 ] ) rather than ( [ eq:24 ] ) ( fritze - von alvensleben 1999 ) ; and theoretically , such an initial shape to @xmath55 may in fact be _ preserved _ during a hubble time of dynamical evolution ( vesperini 1997 ) .
thus , the intent here is not necessarily to argue for an explanation of present - day gcs mass spectra in purely evolutionary terms , but to show that even if such is adopted as an extreme possibility , the implied time dependence of total gcs masses is quite weak
. given equations ( [ eq:24 ] ) and ( [ eq:25 ] ) , then , straightforward integrations allow for comparisons of the present gcs population to the initial one : @xmath57[1-(m_\ell / m_*)^ { 1-\gamma_1}]}\over{(m_u / m_\ell)^{\gamma_2 - 1}-1}}\right\}\ , \label{eq:26}\ ] ] and of the current total mass to its initial value : @xmath58[1-(m_\ell / m_*)^ { 2-\gamma_1}]}\over{1-(m_\ell / m_u)^{2-\gamma_2}}}\right\}\ .
\label{eq:27}\ ] ] values of @xmath59 and @xmath60 may be taken as representative of the typical situation in elliptical galaxies ( harris & pudritz 1994 ; mclaughlin 1994 , 1995 ; mclaughlin & pudritz 1996 ) .
with @xmath61 , @xmath62 , and @xmath63 , it is then easily seen that @xmath64 in this simplistic scenario .
thus , _ even though dynamical evolution reduces the total population of this hypothetical gcs by more than an order of magnitude , it diminishes the total mass by only _
( the same line of argument also shows that the often appreciable uncertainties in the exact form of @xmath11 at cluster masses
@xmath48or even blind extrapolations from observations of only brighter clusters do not seriously bias observational estimates of @xmath36 in any galaxy . )
this result is best interpreted as an estimate for the possible depletion of a gcs _ globally _
( i.e. , spatially averaged over the whole of a parent galaxy ) , because the values of @xmath65 , @xmath66 , and @xmath46 used are taken from observations of @xmath11 in projection , often over fairly large areas on the sky .
thus , the normalization @xmath67 in equations ( [ eq:24 ] ) and ( [ eq:25 ] ) can be taken to include an implicit integration over galactocentric radius , and the exponents @xmath68 and mass scale @xmath46 to be appropriate averages .
the importance of these considerations is not actually clear , since radial dependences are not expected theoretically in current models for the initial gcs mass function ( mclaughlin & pudritz 1996 ; elmegreen & efremov 1997 ) and observational evidence for such trends at the current epoch is slight to non - existent ( e.g. , mclaughlin et al .
1994 and harris et al . 1998
; though see also kavelaars & hanes 1997 and murali & weinberg 1997b ) .
nevertheless , it should be noted that dynamical friction , gravitational shocks , and even evaporation couple to galactic tidal fields in such a way that if a gcs s present - day mass function _ were _ sculpted to any great extent by these processes , then the _ local _ ratios @xmath69 and @xmath70 would be expected to increase outwards from the center of the system , from values less than equation ( [ eq:28 ] ) at small galactocentric radii to near unity at large distances ( e.g. , beyond a stellar effective radius ; see especially murali & weinberg 1997b ) .
returning now to the main issue , the global efficiency of globular cluster formation in a galaxy may be defined as @xmath71 where @xmath5 refers to the total gas supply that was available to the protogalaxy , whether in a rapid monolithic collapse or , more likely , in a slower assembly ( at high redshift ) of many distinct , subgalactic clouds . from the preceding discussion ,
it is clear that if a newly formed gcs were able to evolve `` passively''that is , if its mass were potentially vulnerable only to reductions caused by forces internal to an isolated parent galaxy then a present - day observation of @xmath36 would suffice , to at least a @xmath1825% level of accuracy globally , as a measure of @xmath72 .
passive evolution also implies mass conservation for the galaxy as a whole , so that a reliable value for @xmath5 could be obtained , at any epoch , from an inventory of the total mass in gas plus unclustered stars and stellar remnants .
thus , a good _ observational estimate _ of the global cluster formation efficiency in this simple situation is provided by @xmath73 which is expected to be nearly a time - independent quantity .
realistically , of course , galaxies are not such pristine entities ; interactions between systems may effect significant changes in both @xmath36 and @xmath74 over time , making it difficult to evaluate ( or even clearly define ) initial gcs and gas masses .
for example , the formation history of an elliptical may include major mergers with gas - rich spiral systems , and it has been proposed that these might involve the formation of new globular clusters and substantial increases in @xmath36 ( e.g. , ashman & zepf 1992 ; kumai , basu , & fujimoto 1993a ) . alternatively , additional globulars might be captured by a large galaxy that accretes small , gas - poor systems ( e.g. , ct et al . 1998 ) or accumulates tidal debris from other members in a cluster of galaxies ( see muzzio 1987 ) .
however , processes such as these also increase the total masses of gas and/or stars in the final system . in _
large _ galaxies ( i.e. , in those that do not suffer significant gas loss from strong galactic winds ; cf .
4.2 ) , any changes in the ratio @xmath75 are bound to be smaller than changes in either of these quantities individually .
the presently observed @xmath76 in any large galaxy with a complex dynamical history is then a mass - weighted average of the true @xmath4 in some number of discrete star - formation events and individual accreted systems ; and in the limit that the underlying efficiency of cluster formation is independent of environment and epoch ( as the results of 3 and 4 will suggest is nearly correct ) , it is still true that @xmath77just as in the case of passive evolution , and subject only to the small error associated there with the effects of gcs erosion by cluster evaporation and tidal disruption . in gas - poor elliptical galaxies and the halos of spirals , equation ( [ eq:210 ] ) reduces to @xmath78 , so that as was mentioned above the cluster formation efficiency can be obtained directly from obervations of the global gcs specific frequency ( cf .
eq . [ [ eq:23 ] ] )
. more generally , however , ellipticals can contain large amounts of hot , x - ray emitting gas ( e.g. , forman , jones , & tucker 1985 ) , and when it happens that @xmath79 is non - negligible relative to @xmath38 , measurements of @xmath3 are only low - order approximations to @xmath80 ( which is itself only an estimate of the real efficiency @xmath4 ) . keeping this in mind , the observed systematics of both global _ and local _ gcs specific frequencies provide a useful starting point for an examination of the extent to which cluster formation efficiencies might vary from galaxy to galaxy , or from place to place within a single system . as was intimated in the discussion around equation ( [ eq:23 ] ) ,
not all galaxies have the same globular cluster specific frequency .
there are several aspects to this : ( 1 ) spiral galaxies tend to have lower @xmath3at least when their gcs populations are normalized to their _ total _
( bulge plus disk ) light
than do average ellipticals ( harris 1991 ) .
( 2 ) although a constant @xmath30 is a fair _ approximation _ for many normal giant ellipticals , fits to the data suggest that , in fact , @xmath81 ( kissler - patig 1997 ; cf .
santiago & djorgovski 1993 ) . at the extremes of the early - type galaxy sequence ,
moreover , ( 3 ) the @xmath3 values for centrally dominant or first - ranked members of galaxy clusters show a much steeper dependence on galaxy luminosity@xmath82 or so is implied by the correlations of blakeslee ( 1997 ) and harris et al .
( 1998)and ( 4 ) specific frequencies actually appear to increase towards _ lower _ luminosities in dwarf ellipticals : @xmath83 ( miller et al .
1998 ; see also durrell et al .
thus , some brightest cluster galaxies , and some of the faintest dwarfs , have global specific frequencies that exceed the `` average '' by factors of 3 or 4 .
all in all , observed @xmath3 values span a range from @xmath181 to @xmath1820 .
as will be shown in 4 below , the first two of these points can be understood fairly easily . here
, it suffices to note that ( 1 ) when only the _ halo _ stars and globular clusters in the milky way are considered , and allowances made for the lower pop .
ii mass - to - light ratio in our galaxy vs. large ellipticals , @xmath80 here is no different from that in `` normal '' early - type systems .
it is reasonable to expect that this result which is contrary to the claim made by zepf & ashman ( 1993)should hold in other spirals as well .
also , ( 2 ) the observed dependence of @xmath3 on @xmath9 among non - brightest cluster ellipticals is fully consistent with that expected from equation ( [ eq:22 ] ) for gas - poor systems with a uniform efficiency of cluster formation ( @xmath84 ) and a mass - to - light ratio ( @xmath85 ) that scales with luminosity in the way observed for galaxies in the fundamental plane .
however , if taken at face value , items ( 3 ) and ( 4 ) just above appear to imply that gcs formation efficiencies can vary by as much as an order of magnitude from galaxy to galaxy , in a non - monotonic fashion and for reasons that have never been understood .
these issues are not so readily dismissed .
the high @xmath3 `` phenomenon '' in many first - ranked cluster ellipticals , especially , has been discussed at length in the literature ( e.g. , mclaughlin et al .
1994 ; harris et al . 1995 ; west et al . 1995 ; forbes et al . 1997 ; blakeslee 1997 ; blakeslee et al . 1997 ; harris et al .
1998 ) , and is referred to here as the _ first specific frequency problem_. equally non - trivial , though only more recently recognized , is the _ inverse _ variation of @xmath3 with @xmath9 among dwarf ellipticals . local specific frequencies , which are obtained for a single gcs by comparing its projected radial distribution to its galaxy s surface brightness profile ( normalized as in eq .
[ [ eq:21 ] ] ) , are also of interest here ; they are related in the obvious way to the efficiency of cluster formation as a function of ( projected ) galactocentric radius , @xmath86 . of particular importance is the fact that @xmath3 is seen to be an _ increasing _ function of @xmath86 in many galaxies .
this is equivalent to the well known fact that gcss are often less centrally concentrated than the stellar halos of their parent galaxies .
this statement has two contexts , which must be separated .
first , it seems that the gcs of most any galaxy spiral or elliptical , small or large has an apparent core radius that is significantly larger than that of the unclustered field stars ( e.g. , djorgovski & meylan 1994 ; lauer & kormendy 1986 ; mclaughlin 1995 ; forbes et al . 1996
; minniti , meylan , & kissler - patig 1996 ; ct et al .
1998 ) ; that is , the _ central regions _ of gcss show projected radial distributions with slopes that are significantly shallower than those of the light profiles on kpc scales , so that the density ratio of the two halo components is an increasing function of radius .
second , the gcss of some large galaxies are spatially more extended than the stellar halos on _ larger scales _
, @xmath87100 kpc ( harris & racine 1979 ; harris 1986 ) . however , this is a less general result than that relating to the large core radii of globular cluster systems .
it is true of many dominant cluster galaxies like m87 in virgo ( see the discussion in mclaughlin et al .
1995 ) , and also of the m31 halo ( pritchet & van den bergh 1994 ) ; but many elliptical systems ( e.g. , durrell et al . 1996 ; kissler - patig et al .
1997 ) , and the milky way itself ( harris 1976 ; zinn 1985 ) , show gcs and halo - star distributions that are in quite close agreement ( and , thus , a constant local @xmath3 ) outside of any core regions .
the physical factors that govern whether the gcs density falls off at the same rate as or more slowly than the unclustered halo density in a given system have not yet been identified .
( no case has ever been found in which a gcs radial distribution is _ steeper _ than a galaxy s light profile . )
the first of these points is not easily related to the radial dependence of cluster formation efficiencies because , as was suggested in 2.1 , the dynamical mechanisms that act to destroy globulars work most efficiently at small @xmath86 , where the stellar densities are highest . although the cores of gcss could quite plausibly have always been somewhat larger than those of halos or bulges generally ( e.g. , harris & pudritz 1994 ; ct et al .
1998 ) , it seems inevitable that any initial discrepancies must have been augmented by an inside - out erosion of the cluster system from tidal shocking , dynamical friction , and evaporation ( ostriker et al . 1989 ; capuzzo - dolcetta 1993 ; murali & weinberg 1997b ) .
detailed modelling is therefore required to disentangle the effects of formation and evolutionary processes in this instance . at galactocentric
radii greater than a few kpc , however , the timescales on which the evolutionary mechanisms operate can exceed a hubble time , and the studies cited here show that dynamics will not significantly affect gcs radial distributions beyond about one effective radius in the stellar halo ( see also aguilar et al .
thus , to the extent that the assumption @xmath78 is valid , the distension of gcss relative to stellar halos and the concomitant systematic increase in local @xmath3 in the outer parts of _ some _ galaxies would seem to say that bound clusters were sometimes ( but not always , and for unknown reasons ) _ more _ likely to form at larger @xmath86 , in gas that was presumably at lower ambient densities and pressures . this situation will be referred to as the _ second specific frequency problem_. it is clearly of interest , from the points of view of both galaxy and star formation , to verify whether these global and local @xmath3 variations really do reflect differences in @xmath4and , if so , to understand their origins .
accordingly , much attention has been paid to the first @xmath3 problem in particular , usually in efforts to empirically correlate the @xmath3 measurements for brightest cluster galaxies ( bcgs ) with global properties of the galaxies themselves or of the clusters they inhabit .
one such correlation does appear to be of fundamental importance , as will be discussed momentarily , but it suggests that @xmath3 varies in the face of _ constant _ @xmath88 .
indeed , no proposed solution of the first specific frequency problem has identified a physical mechanism that could possibly lead to the superefficient production of bound star clusters in some galaxies ( see the references given above ) .
in fact , one suggestion that high global specific frequencies ( meaning @xmath89 ) in centrally located bcgs result from the acquisition of globulars and stars which are tidally stripped from the outer , high-@xmath3 regions of other galaxies in a cluster actually appears to rely on the existence of the _ second _ specific frequency problem ( see forbes et al . 1997 ; ct et al .
1998 ) ; but this is itself essentially unexplained . nor , to date , has any explanation ( or prediction ) been given for the high @xmath3 that have now been observed in many faint dwarf galaxies .
the view is taken here that _
the second specific frequency problem is inseparable from the first _
, in the specific sense that both relate to the possible influence of environment on the probability of bound cluster formation .
any real insight into either one of these issues must therefore include a better understanding of the other .
this view is in keeping with recent evidence ( forbes et al .
1997 ; kissler - patig 1997 ; van den bergh 1998 ) for at least a rough correlation between global @xmath3 values and gcs density - profile slopes in elliptical galaxies : those gcss with the highest specific frequencies tend also to have the shallowest radial distributions .
such a correspondence hints strongly at ( although it does not directly show ) an intimate connection between an increasing local @xmath3 and a high global @xmath3 .
a significant development in this field has come with the recent recognition of a correlation between the global specific frequencies of bcgs and the x - ray luminosity of the hot gas in their parent clusters ( west et al . 1995 ;
blakeslee 1997 ; blakeslee et al .
a relatively large and homogeneous sample of bcgs shows @xmath3 values that increase from about 4 , in clusters with x - ray luminosities ( on 500-kpc scales ) @xmath90 erg s@xmath91 at 0.54.5 kev , to @xmath92 for @xmath93 erg s@xmath91 .
three different interpretations have already been attached to this result : west et al .
( 1995 ) propose that it points to the existence of intergalactic globulars which collect naturally at the centers of galaxy clusters and are superimposed in projection on any galaxy there .
this would certainly boost the apparent total population @xmath22 of a central bcg , so that globulars need not have formed in situ with abnormally high specific frequency ; but they would still have had to do so elsewhere in the cluster ( see also ct et al .
1998 ; harris et al . 1998 ) . in a different approach , blakeslee ( 1997 ) and blakeslee et al .
( 1997 ) consider @xmath94 to be mainly a reflection of the temperature of intracluster gas , and thus a measure of a cluster s mass and the depth of its potential .
they then claim that the observed @xmath3@xmath94 correlation is consistent with the globulars in bcgs having formed in numbers directly proportional to the _ total _ mass , including dark matter , of their _ entire _ galaxy clusters ( see kavelaars 1998 for a similar suggestion on the scale of individual galaxies ) . in this scenario ,
cluster formation would obey a universal efficiency of sorts although one which is quite different from that discussed here and the first @xmath3 problem would result from some process that caused the bcgs in more massive clusters to be systematically underluminous ( i.e. , underefficient in the production of unclustered field stars ) . here , however , the role that the dark matter content of a galaxy cluster might play in the formation of globular clusters is not clear .
moreover , any rule such as @xmath95 clearly can apply over only a limited range of conditions .
thus , harris et al . (
1998 ) instead pursue the hypothesis that gcss formed with a constant @xmath4 defined essentially as in equation ( [ eq:29 ] ) above .
this is adopted as a working _ assumption _ by harris et al .
, who construct plausibility arguments for a universal value of @xmath96 ( as could be expected ) and suggest that the higher global @xmath3 in more x - ray luminous bcgs might have resulted from stronger ( partial ) galactic winds which heated larger fractions of their initial gas supplies to kev temperatures after an early bout of star and globular cluster formation
. such hot gas would be largely removed from further star formation but still confined to the vicinity of these central cluster galaxies and the final bcgs would be optically underluminous as a result . it is this basic idea that is closest to the discussion in this paper ( although many aspects of the quantitative analysis here differ significantly from those in harris et al . 1998 ) .
that is , the fundamental point of the global @xmath3@xmath94 correlation for bcgs is the associated implication that the first specific frequency problem may stem , not from intrinsic variations in the global @xmath4 of these systems , but from the fact that they can contain large amounts of gas .
high gas mass ratios invalidate the assumption that @xmath84 , and therefore cause @xmath3 to overestimate the true globular cluster formation efficiency .
if it is correct to infer from this that the `` excess '' globulars in systems with high global @xmath3 are associated with hot gas , then the fact that the distribution of gas is generally more extended than that of the stars in ellipticals ( e.g. , trinchieri , fabbiano , & canizares 1986 ; fabbiano 1989 ) suggests a possible explanation for the second specific frequency problem as well as the first .
these two issues would then be closely related possibly even just different aspects of a single phenomenon as was suggested above ; and both might be settled simply by measuring the full estimator @xmath97 both locally and globally in individual systems . in the next section , this universal-@xmath4 hypothesis is _ tested _ , for the first time , by evaluating the observable @xmath80 as a function of galactocentric radius in each of the galaxies m87 , m49 , and ngc 1399 .
it turns out that the inclusion of the x - ray gas in m87 is indeed key to the alleviation of both the first and the second @xmath3 problems there : once this is done , @xmath80 is constant locally , and essentially the same globally , in all three of these systems .
the discussion is expanded to include other giant ellipticals and bcgs , as well as dwarf ellipticals , in 4 .
it is shown directly that all available gcs data are consistent with a single efficiency of cluster formation .
notably , the increase of @xmath3 towards low @xmath9 in the dwarfs ( miller et al .
1998 ; durrell et al .
1996 ) can be explained in terms of this universal @xmath4 once allowances are made for the effects of the violent gas outflows ( a more extreme version of the feedback envisioned by harris et al .
1998 for bcgs ) that were driven by early bursts of star formation in these low - mass systems .
m87 , m49 , and ngc 1399 are ideally suited to a closer examination of these issues , for three reasons : ( 1 ) they are relatively nearby , making it possible to observe internal trends within the systems in the first place .
m87 and m49 both are assumed here to lie at a distance of @xmath98 mpc , appropriate for the core of the virgo cluster ( pierce et al . 1994 ; freedman et al .
. a relative distance modulus of @xmath99=0.2 $ ] ( e.g. , mcmillan , ciardullo , & jacoby 1993 ; tonry et al .
1997 ) then puts ngc 1399 at @xmath100 mpc .
( 2 ) each is a giant elliptical ( m87 and ngc 1399 are actually type cd ) , and is therefore accompanied by a large gcs : @xmath101 for m87 ; @xmath102 for m49 ; and @xmath103 in ngc 1399 ( e.g. , harris 1991 ; see also 5 below ) . and
, most importantly , ( 3 ) even this small sample of three galaxies suffices to bring out all aspects of the two specific frequency problems summarized in 2.2 : m87 , as is well known , suffers from both ; m49 does not show the first , but has long been thought to exhibit the second ; and ngc 1399 has been claimed as an example of the first problem , but not the second .
because of their proximity , these systems have been well studied in all respects of importance here .
@xmath104-band surface photometry extends in m87 to a projected galactocentric radius of about @xmath105 kpc ( de vaucouleurs & nieto 1978 , 1979 ) ; in m49 , to a similar distance ( king 1978 ; caon , capaccioli , & donofrio 1994 ) ; and in ngc 1399 , to @xmath106 kpc ( killeen & bicknell 1988 ; caon et al .
these measurements are easily converted to projected stellar mass density profiles @xmath107 , as mass - to - light ratios have been obtained from a spectroscopic analysis applied uniformly to all three systems ( van der marel 1991 ) .
x - ray observations , which have been made with several different satellites , yield gas densities over still larger areas around each galaxy . here , use will be made of the _ rosat _ data for m87 ( taken from nulsen & bhringer 1995 ) ; of _ rosat _ data for m49 ( irwin & sarazin 1996 ) ; and of _ einstein _ observations in ngc 1399 ( killeen & bicknell 1988 ) .
these studies all give the _ volume _ electron density @xmath108 , as a function of _ three - dimensional _ galactocentric radius , derived either from fits of a parametric function to the projected x - ray brightness profile or from a direct deprojection of the same .
finally , integrated optical magnitudes have been measured for globular clusters in these galaxies at radii @xmath109 kpc generally , over most of the areas covered by the surface photometry of unresolved halo stars .
these are used to construct projected gcs number densities , @xmath110 , which are trivially converted to mass profiles , @xmath111 , once a mean globular cluster mass is specified . for m87 ,
relevant investigations are those of harris & smith ( 1976 ) , harris ( 1986 ) , grillmair , pritchet , & van den bergh ( 1986 ) , lauer & kormendy ( 1986 ) , cohen ( 1988 ) , mclaughlin et al .
( 1993 , 1994 ) , mclaughlin ( 1995 ) , whitmore et al .
( 1995 ) , and harris et al .
data on m49 can be found in harris & petrie ( 1978 ) , harris & van den bergh ( 1981 ) , harris ( 1986 ) , harris et al .
( 1991 ) , and lee , kim , & geisler ( 1998 ) .
the gcs of ngc 1399 has been observed by hanes & harris ( 1986 ) , bridges , hanes , & harris ( 1991 ) , wagner et al .
( 1991 ) , and kissler - patig et al .
( 1997 ) . while surface brightness profiles and x - ray gas densities often can be obtained for most of an entire galaxy with a single set of observations as is the case in the studies cited above
this is generally _ not _ so for gcs radial distributions .
older , photographic surveys , which did have large ( @xmath112 kpc ) fields of view , were hampered by rather bright limiting magnitudes that became systematically more so towards smaller galactocentric radii .
newer , ccd photometry achieves deeper limiting magnitudes , and digital data reduction techniques can quantify and correct for any spatial variations ; but until very recently , ccd fields of view were quite small , and best suited to analyses of the central regions ( @xmath113 kpc ) of gcss .
thus , the first step in comparing the stellar , gaseous , and globular cluster contents of any large galaxy is to combine existing photographic and ccd gcs data into a single , composite radial distribution that gives the projected density @xmath114 , to a uniform limiting magnitude , over as large a range of galactocentric radius as possible .
this is done for each of m87 , m49 , and ngc 1399 in 3.1 .
but then , the surface densities of stars and gcss are related to their volume densities by an integral of the form @xmath115 , while the observed x - ray brightness along some line of sight is proportional to @xmath116 .
a direct intercomparison of the _ raw _ gcs , optical , and x - ray imaging data in a galaxy therefore makes little sense .
thus , to discuss these three components simultaneously in m87 , m49 , and ngc 1399 , either ( 1 ) their gcs and stellar surface densities can be deprojected and compared to their derived gas volume densities , or ( 2 ) the gas surface densities , @xmath117 , can be obtained from the published volume density profiles and compared to the directly observed @xmath107 and @xmath118 .
the results of these two separate exercises are presented in
3.2 and 3.3 .
when joining gcs radial distributions from two or more separate studies of the same galaxy , two important factors must be taken into account .
first , the gcs of a distant galaxy ( essentially , any with @xmath119 mpc ) is usually identified as a centrally concentrated excess of point - like sources , above a more uniform distribution of unresolved background galaxies and foreground stars .
the surface density of this `` background , '' @xmath120 , is subtracted from the total density of point sources at any @xmath86 to form the surface density profile @xmath121 of the gcs alone ; but @xmath120 is a function of seeing conditions , detector characteristics , and limiting magnitude .
background corrections must therefore be applied separately to each of the datasets to be combined .
second , differences in limiting magnitude ( call this @xmath122 ) are of further importance because a deep survey will clearly see more globulars , at a given @xmath86 , than one with a bright limit . before they can be simply matched onto one another , then , distinct background - corrected profiles @xmath110 must be normalized to the same effective @xmath122 . the published gcs observations that will be used here for m87 , m49 , and ngc 1399 are summarized in table [ tab1 ] .
these papers were selected from the more numerous references cited above , simply because they tabulate all of the relevant data in full .
the third column of table [ tab1 ] gives the minimum and maximum @xmath86 contained in each study s field of view ( note that @xmath123 kpc for a distance of @xmath98 mpc to m87 and m49 , while @xmath124 mpc for ngc 1399 implies @xmath125 kpc ) ; column 4 gives the background surface density to the limiting magnitude in column 5 ; and column 6 lists the scalings applied to each dataset to bring them all to a common @xmath122 .
llcccl m87 & mclaughlin ( 1995 ) & @xmath126 & @xmath127 & 23.9 & 1.000 @xmath128 & mclaughlin et al .
( 1993 ) & @xmath129 & @xmath130 & 24.0 & 0.954 & harris ( 1986 ) & @xmath131 & @xmath132 & 23.4 & 1.35 & & & & & & & & & & m49 & harris & petrie ( 1978 ) & @xmath133 & @xmath134 & 23.3 & 1.000 @xmath135 & harris & van den bergh ( 1981 ) & @xmath136 & @xmath137 & @xmath138 & 0.86 & harris et al .
( 1991 ) & @xmath139 & @xmath140 & 23.8 & 0.707 & lee et al .
( 1998 ) & @xmath141 & @xmath142 & 23.45 & 0.892 & & & & & & & & & & ngc 1399 & kissler - patig et al .
( 1997 ) & @xmath143 & @xmath144 & 24.0 & 1.000 @xmath145 & hanes & harris ( 1986 ) & @xmath146 & @xmath147 & 23.4 & 1.575 generally , the background levels @xmath120 have been determined and applied by the original authors , either from observations of a blank field near the observed gcs , or from `` on - frame '' extrapolations of counts at the edges of the gcs field itself ( see harris 1986 ) ; these values are reproduced in table [ tab1 ] only for completeness .
the single exception to this is the m49 study of lee et al .
( 1998 ) , for which @xmath120 has been derived here , using the method of harris ( 1986 ) .
( it is possible , particularly with on - frame estimates , to overestimate @xmath120 and produce spurious curvature at large radii in a subtracted @xmath121 profile .
as is discussed below , however , this is not a major concern in the composite gcs radial distributions obtained here . )
the limiting magnitudes @xmath122 in table [ tab1 ] are also those quoted in the original papers .
these have been used to renormalize the individual @xmath121 profiles , by adopting for computational convenience a standard gaussian approximation to the globular cluster luminosity function ( gclf , @xmath148 ; cf .
that is , with the number of globulars at a given apparent magnitude given roughly by @xmath149 $ ] ( see harris 1991 ) , then the density of clusters to some limiting magnitude , say @xmath150 , is referred to the density up to some other limit , i.e. , @xmath151 , through @xmath152}\over { 1+{\rm erf}\left[(v_{{\rm lim},2}-v^0)/\sqrt{2}\,\sigma_v\right]}}\ . \label{eq:31}\ ] ] the right - hand side of this expression is just the scaling factor given in column 6 of table [ tab1 ] , where @xmath153 for each gcs has been taken as the observed limiting magnitude in the first of each set of references .
thus , the published radial distributions for m87 have all been normalized to an effective @xmath154 ; for m49 , to @xmath155 ; and to @xmath156 in ngc 1399 .
the gaussian peaks @xmath157 and dispersions @xmath158 used in equation ( [ eq:31 ] ) are given in the first column of table [ tab1 ] .
these parameters are taken from the fits of whitmore et al .
( 1995 ) and harris et al .
( 1998 ) to the m87 gclf , from lee et al .
( 1998 ) for m49 , and from kohle et al .
( 1996 ) for ngc 1399 .
note that all of the scale factors derived here are rather modest , and thus do not introduce significant uncertainty into the final , combined @xmath121 profiles .
= 4.0truein a few practical notes should be made regarding this scaling procedure .
first , it is applied to the gcs densities at every observed galactocentric radius , and it therefore assumes that there is no intrinsic radial variation in the gclf or the underlying gcs mass function . as was also mentioned in 2.1 above ,
this is a reasonable assumption that is directly supported by the observations themselves .
second , the limiting magnitudes for the photographic studies of harris & petrie ( 1978 ) and harris ( 1986 ) are rather poorly determined ; harris quotes only approximate ranges for @xmath159 .
the values adopted here , which are fully consistent with these ranges , have been chosen to give scaled @xmath110 that agree well with more recent ccd studies with accurately known @xmath122 .
and third , the scaling applied to the m49 gcs counts of harris & van den bergh ( 1981 ) has not been calculated in this way at all ; rather , the factor of 0.86 is derived , by harris ( 1986 ) , by requiring a good overall agreement with the @xmath121 profile of harris & petrie ( 1978 ) .
the results of this exercise are illustrated in fig .
clearly , the different projected density profiles in each gcs corroborate each other .
it remains only to extract a single @xmath121 at each @xmath86 , in each system , from the overlapping data that have been collected .
it is not generally possible to simply average different densities at a given @xmath86 , since profiles such as these are constructed by counting clusters in a series of wide , circular annuli that need not coincide from study to study . instead , fig .
[ fig1 ] shows that a perfectly acceptable approach is to simply adopt one of the ( scaled ) profiles over some range in @xmath86 , and then switch to any other that might extend beyond that range . at this point
, it must be kept in mind that ultimately the data will be deprojected to yield volume gcs densities , @xmath160 .
the method developed in the appendix to do this requires that the inner and outer radii be known for the annuli used to define the composite @xmath121 .
this further restricts the useful published datasets to those which contain this information on the radial bins as well ( that is , _ in addition _ to just a density and a mean radius for each annulus ) . in m87 ,
then , a composite gcs profile is derived by using the data of mclaughlin ( 1995 ) from @xmath161 to @xmath162 ; those of mclaughlin et al .
( 1993 ) from @xmath163 to @xmath164 ; and those of harris ( 1986 ) from @xmath164 to @xmath165 .
( in order to make annuli that do not overlap in the final @xmath121 , some interpolation is applied to the datasets around the matching points at @xmath163 and @xmath164 . ) for m49 , the scaled profile of harris & van den bergh ( 1981 ) is used over @xmath166 , and that of harris & petrie ( 1978 ) for @xmath167 . and in ngc 1399 , the densities of kissler - patig et al .
( 1997 ) are adopted for @xmath168 , with the data of hanes and harris ( 1986 ) continuing on out to @xmath169 . in all cases ,
ccd data have usually been chosen over photographic data in any regions of overlap . however , superseding this preference is a concern over possible errors in the background estimation @xmath120 ; where this seems too high for example , when the background - subtracted @xmath121 at the largest radii in a smaller - scale ccd study drop significantly below the densities indicated by a more extensive photographic survey ( as in the outermost data points of kissler - patig et al .
1997 for ngc 1399)the photographic profiles are given precedence .
enforcing consistency in this way between independent gcs datasets ensures that the question of potentially overestimated @xmath120 values remains an issue only at the largest radii ( beyond the regime of any overlap ) in the widest - field observations of each galaxy .
the composite radial distributions thus obtained are listed in the third columns of tables [ tab2 ] through [ tab4 ] , as average surface densities in series of concentric circular annuli ( with inner , outer , and average projected radii given in the first and second columns of the tables ) .
they are also drawn as the bold lines in the three panels of fig .
there are two details of particular note .
first , the densities for the innermost two annuli in m49 have been scaled up from the ( already renormalized ) gcs profile of harris & van den bergh ( 1981 ) , by a further factor of 1.1 , to put the composite @xmath121 at small radii into better agreement with the ccd data of harris et al .
( 1991 ) and lee et al .
this ad hoc correction most likely just reflects some residual incompleteness in the photographic gcs counts at small @xmath170 , and in any event does not significantly affect the results which follow .
second , all of the @xmath121 profiles have been smoothed somewhat by rebinning , in order to allow for a more stable deprojection in 3.2 below : the outermost six annuli of harris ( 1986 ) are combined into one for the m87 profile ; the outermost 11 rings of harris & petrie ( 1978 ) are blended into 4 for m49 ; and the last 8 radial bins of hanes & harris ( 1986 ) become 4 in the ngc 1399 gcs .
( as an added benefit of this smoothing , any possible errors in the subtracted @xmath120 are expected to affect only the outermost _ one or two points _ in the final @xmath121 profiles . ) similarly , counts in the `` northeast '' and `` northwest '' sectors of ngc 1399 in the analysis of kissler - patig et al .
( 1997 ; see their table 8) have been added together in every annulus they define ; and these annuli have been combined in pairs to give the coarser profile in table [ tab4 ] here .
ccccc 0.153 0.328 & 0.244 & [email protected] & 0.245 & [email protected] 0.328 0.655 & 0.494 & [email protected] & 0.505 & [email protected] 0.655 0.983 & 0.808 & [email protected] & 0.827 & [email protected] 0.983 1.310 & 1.157 & [email protected] & 1.152 & [email protected] 1.310 1.870 & 1.500 & [email protected] & 1.602 & [email protected] 1.870 2.160 & 2.010 & [email protected] & 2.018 & [email protected] 2.160 2.490 & 2.320 & [email protected] & 2.328 & [email protected] 2.490 2.880 & 2.680 & [email protected] & 2.689 & [email protected] 2.880 3.320 & 3.090 & [email protected] & 3.104 & [email protected] 3.320 3.830 & 3.570 & [email protected] & 3.580 & [email protected] 3.830 4.430 & 4.120 & [email protected] & 4.135 & [email protected] 4.430 5.110 & 4.760 & [email protected] & 4.776 & [email protected] 5.110 5.900 & 5.490 & [email protected] & 5.512 & [email protected] 5.900 6.820 & 6.350 & [email protected] & 6.368 & [email protected] 6.820 7.870 & 7.330 & [email protected] & 7.354 & [email protected] 7.870 8.973 & 8.403 & [email protected] & 8.431 & [email protected] 8.973 9.962 & 9.454 & [email protected] & 9.474 & [email protected] 9.962 10.95 & 10.44 & [email protected] & 10.46 & [email protected] 10.95 11.93 & 11.43 & [email protected] & 11.45 & [email protected] 11.93 12.92 & 12.42 & [email protected] & 12.43 & [email protected] 12.92 13.90 & 13.40 & [email protected] & 13.41 & [email protected] 13.90 14.87 & 14.38 & [email protected] & 14.39 & [email protected] 14.87 15.84 & 15.35 & [email protected] & 15.36 & [email protected] 15.84 16.81 & 16.32 & [email protected] & 16.33 & [email protected] 16.81 22.52 & 19.46 & [email protected] & 19.77 & [email protected] ccccc 0.732 1.464 & 1.035 & [email protected] & 1.129 & [email protected] 1.464 2.196 & 1.793 & [email protected] & 1.848 & [email protected] 2.196 2.928 & 2.536 & [email protected] & 2.575 & [email protected] 2.928 3.660 & 3.274 & [email protected] & 3.304 & [email protected] 3.660 4.392 & 4.009 & [email protected] & 4.034 & [email protected] 4.392 4.996 & 4.684 & [email protected] & 4.699 & [email protected] 4.996 5.992 & 5.471 & [email protected] & 5.505 & [email protected] 5.992 6.988 & 6.471 & [email protected] & 6.500 & [email protected] 6.988 7.981 & 7.467 & [email protected] & 7.493 &
[email protected] 7.981 8.973 & 8.462 & [email protected] & 8.484 & [email protected] 8.973 10.95 & 9.912 & [email protected] & 9.985 & [email protected] 10.95 13.90 & 12.33 & [email protected] & 12.47 & [email protected] 13.90 16.81 & 15.28 & [email protected] & 15.39 & [email protected] 16.81 19.69 & 18.19 & [email protected] & 18.28 & [email protected] ccccc 0.761 1.518 & 1.075 & [email protected] & 1.171 & [email protected] 1.518 2.274 & 1.858 & [email protected] & 1.915 & [email protected] 2.274 3.031 & 2.625 & [email protected] & 2.666 & [email protected] 3.031 3.788 & 3.388 & [email protected] & 3.420 & [email protected] 3.788 4.544 & 4.149 & [email protected] & 4.175 & [email protected] 4.544 5.301 & 4.908 & [email protected] & 4.930 & [email protected] 5.301 6.058 & 5.667 & [email protected] & 5.686 & [email protected] 6.058 6.814 & 6.425 & [email protected] & 6.442 & [email protected] 6.814 7.571 & 7.183 & [email protected] & 7.197 & [email protected] 7.571 8.104 & 7.833 & [email protected] & 7.840 & [email protected] 8.104 9.573 & 8.808 & [email protected] & 8.854 & [email protected] 9.573 11.04 & 10.21 & [email protected] & 10.32 & [email protected] 11.04 12.50 & 11.75 & [email protected] & 11.78 & [email protected] 12.50 13.97 & 13.21 & [email protected] & 13.24 & [email protected] given these radial distributions , it is a simple matter to derive the total _ mass _
surface density profiles of the gcss .
this requires , first , another multiplicative scaling of @xmath110 in each case , so that the number density of _ all _ globulars is properly reflected .
( note that the observations used here reach limiting magnitudes which are just at , or brighter than , the peak of the symmetric gaussian that best fits the gclf .
this implies that most of the faint globulars are lost in the noise in these studies , and that half or fewer have been directly observed at any @xmath86 . )
these final correction factors are obtained from equation ( [ eq:31 ] ) above , by setting @xmath172 and putting @xmath173 at the effective limit for each composite @xmath121 . in this way
, it is found that the m87 distribution in table [ tab2 ] should be multiplied by 1.795 ; the m49 densities in table [ tab3 ] , by 2.743 ; and the ngc 1399 numbers , by 1.875 .
once this is done , the total number densities are multiplied by a mean globular cluster mass to obtain the mass densities @xmath174 .
the mean cluster mass adopted here is @xmath175 , which ( as was mentioned in 2.1 above ) is the value for the milky way gcs . in both its shape _ and _ its mass scale , the galactic gclf , or mass function , bears enough similarity to those of the cluster systems being discussed here ( and , indeed , to most others : harris 1991 ) that the same @xmath176 may safely be applied to them as well .
the only issue is whether or not there is any systematic radial trend in mean globular cluster masses .
it was also pointed out in 2.1 that there are no strong radial dependences observed in the gclf of m87 ( mclaughlin et al . 1994 ; harris et al .
1998 ) , suggesting that @xmath176 is constant with @xmath12 in this galaxy , and thus likely also in similar systems like m49 and ngc 1399 . the question can be addressed in more detail only for the milky way , which although clearly an imperfect substitute for these giant ellipticals is the only galaxy where accurate luminosities and three - dimensional galactocentric distances both can be assigned to globulars on an individual basis .
figure [ fig2 ] shows that there is no dependence on radius in the observed @xmath176 of the milky way .
the top panel plots the volume number density of globulars in the galactic halo , derived from the database of harris ( 1996 ) ; drawn as the solid line is a least - squares fit of a parametric density distribution in the family discussed by dehnen ( 1993 ) and tremaine et al .
( only metal - poor clusters have been counted , as these are the ones that display true halo kinematics : zinn 1985 ; minniti 1995 . ) in the bottom panel , the total @xmath23-band luminosities of the clusters ( taken again from harris 1996 ) have been summed separately in each radial bin , and converted to total cluster masses by applying a mass - to - light ratio of @xmath177 ( mandushev , spassova , & staneva 1991 ; pryor & meylan 1993 ) .
the mass densities @xmath178 that result are shown as the circular points .
the solid line now corresponds to the same @xmath179 fit from the top panel , but multiplied by the average mass calculated for all globulars with @xmath180 .
this shows that @xmath181 to a good approximation at every @xmath12 , and therefore that the mean globular cluster mass is not a strong function of galactocentric radius .
the other curves in the bottom panel of this figure are discussed in 4.3 . for now , the point is simply to lend some circumstantial support to the claim that @xmath182 should not vary significantly with galactocentric radius in m87 , m49 , and ngc 1399 .
= 4.0truein the total projected mass densities for these three gcss are shown as the filled circles in fig .
[ fig3 ] , where angular distances and areas have been converted to linear units using @xmath183 mpc and @xmath184 mpc .
the final @xmath111 have also been scaled upwards by additional factors of @xmath185 for this figure , to facilitate direct comparisons with the mass surface densities of the galaxies stars ( drawn as lines ) .
these latter distributions are derived from published @xmath104-band surface brightness profiles ( m87 : de vaucouleurs & nieto 1978 , 1979 ; m49 : king 1978 , caon et al .
1994 ; ngc 1399 : caon et al .
1994 ) , according to the standard relation @xmath186 for @xmath187 in units of mag arcsec@xmath188 .
( extinctions of @xmath189 mag , 0 , and 0 are adopted for m87 , m49 , and ngc 1399 ; see burstein & heiles 1984 . )
the stellar mass - to - light ratios @xmath190 for all three galaxies come from van der marel s ( 1991 ) kinematic analysis of their cores ; his values have been corrected for the distances @xmath191 assumed here , and are taken to be independent of @xmath86 .
= 4.0truein the comparisons of @xmath111 and @xmath192 in fig . [ fig3 ]
illustrate the @xmath3 `` problems '' in m87 and show that , contrary to previous claims , _ they do not exist _ in m49 and ngc 1399 . as was indicated above
, the global specific frequency of m49 is quite normal ; the most recent estimate places it at @xmath193 ( lee et al .
1998 ) . given a stellar mass - to - light ratio of @xmath194 [ which follows from @xmath195 and @xmath196 , equation ( [ eq:23 ] ) then implies that @xmath197 globally in this galaxy . in keeping with this , the middle panel of fig .
[ fig3 ] shows directly that @xmath198 at essentially any @xmath199 kpc .
this is slightly higher than the global gcs mass fraction because the globular cluster surface density profile is shallower than the stellar distribution at _ small _ galactocentric radii .
as was discussed in 2.2 , such a feature might result , at least in part , from the dynamical destruction of globulars in the densest regions of the galaxy .
[ again , the theoretical work of murali & weinberg 1997b , or of aguilar et al .
1988 , suggests that the important spatial scale is generally of order a stellar effective radius , @xmath200 .
these are marked by vertical lines for the galaxies in fig .
[ fig3 ] ; note that @xmath201 kpc . ] this aside , m49 has been put forward as an example of the second specific frequency problem ( beginning with harris 1986 , and continuing on to lee et al .
1998 ) on the basis of comparisons at _ large _ @xmath202 between @xmath111 and the @xmath192 _ derived from the surface photometry of king ( 1978)_. king s data are shown as the dashed line in the middle of fig .
[ fig3 ] , and they clearly do suggest that the projected stellar density falls off more steeply with radius than the gcs density in m49 . however , the solid line in this panel represents the more recent @xmath187 data of caon et al .
( 1994 ) , which include many more measurements at @xmath203 kpc and do _ not _ show any gcs - halo discrepancy .
the local specific frequency , which is proportional to @xmath204 , then takes on a roughly constant value throughout the outer halo of m49 , and the second @xmath3 problem does not appear there . the differences between the light profiles of king ( 1978 ) and caon et al .
( 1994 ) illustrate the difficulty of accurate sky subtraction at faint surface brightness levels , and make the point that background corrections are an important consideration in any discussion of galaxy ( or gcs ) structure on large spatial scales .
the surface photometry of caon et al .
( 1994 ) is adopted here for m49 , and thus @xmath205 at large @xmath86 ; but this important caveat must be kept in mind . in 3.2 , it is shown that the gas in m49 adds up to a small fraction of its stellar mass , in which case the local cluster formation efficiency can be estimated by @xmath206 . the fact that @xmath207 therefore implies that @xmath4 was essentially independent of galactocentric radius beyond about @xmath208 kpc .
an extrapolation of this result to smaller distances requires that the systematic decrease in @xmath204 there has resulted entirely from the gradual destruction of ( initially bound ) globulars over a hubble time . as _
possible _ anecdotal support for this interpretation , recall that globally @xmath209 for m49 , while locally @xmath210 at large radii .
if the latter number is assumed to have held all the way to the center of the galaxy initially , then a global @xmath211 is implied .
in such a scenario , if @xmath212 were roughly conserved over a hubble time while globulars were dynamically disrupted ( their remains adding nothing significant to the total stellar mass ) , the global ratio of current to initial gcs masses would be @xmath213 .
this is surprisingly close to the result of the general gcs mass - function arguments in 2.1 ( see eq .
[ [ eq:28 ] ] )
. however , this could be just a coincidence ; it remains possible that @xmath4 really was lower at small galactocentric radii , i.e. , that the gcs _ formed _ with a `` core '' which was somewhat larger than that of the unclustered stellar distribution , and which was only enhanced over time .
the current data can not settle the question either way .
the bottom panel of fig .
[ fig3 ] shows that the second @xmath3 problem does not exist beyond a stellar @xmath200 in ngc 1399 , either .
every study of this gcs has appreciated this point , and it is not a new result ( see the references cited above ) .
( note also that the surface brightness profile used here for ngc 1399 , from caon et al .
1994 , is consistent with that of killeen & bicknell 1988 at large @xmath86 . ) on the other hand , the global @xmath3 in this galaxy and therefore the local @xmath204 at any single @xmath86has always been thought to be significantly higher than the normal , m49-like value of @xmath185 : the latest calculation claims @xmath214 ( kissler - patig et al .
1997 ) . however , it is immediately apparent from fig .
[ fig3 ] that this is _ not _ the case : the bulk scaling of @xmath215 required to match the gcs radial distribution to the stellar mass profile in ngc 1399 is identical , within the observational uncertainties , to the @xmath216 needed in m49 .
since the gas mass is negligible compared to the stellar mass on @xmath217-kpc scales in ngc 1399 , as it is in m49 , @xmath218 holds for both systems , and their cluster formation efficiencies are inferred to have been essentially identical .
the spuriously high global @xmath3 values typically quoted for ngc 1399 can be traced back to the practice of assigning an absolute magnitude of @xmath219 to the galaxy ( from the third reference catalogue : de vaucouleurs et al .
1991 ) ; but , in fact , @xmath220 is indicated by direct integration ( to @xmath221 kpc , assuming @xmath124 mpc ) of the surface brightness profiles of either caon et al .
( 1994 ) or killeen & bicknell ( 1988 ) . given this brighter magnitude , and an estimated gcs population of @xmath222 projected to within 70 kpc of ngc 1399 ( see 5 ) , equation ( [ eq:21 ] ) yields @xmath223 , much closer to the value for m49
( see also ostrov et al .
this is consistent with the alternate representation of the data in fig .
[ fig3 ] , and confirms that the _ first _ specific frequency problem is not an issue in ngc 1399 .
it appears , therefore , that there are no @xmath3 problems in either m49 or ngc 1399 .
there is no evidence for any fundamental variation in @xmath88 from the total populations of these two gcss , and the local @xmath80 in each is independent of galactocentric radius beyond @xmath224 .
these results lend some weight to the claim ( 2.2 ) that the first and second specific frequency problems seem likely to appear , and should be solved , only together .
both of these problems are genuine in m87 , as the top panel of fig .
[ fig3 ] attests .
( the surface photometry of de vaucouleurs & nieto 1978 , which is the standard , is well sampled and is supported by the independent studies of oemler 1976 and king 1978 ; and the total @xmath225 used to calculate the galaxy s @xmath3 is taken from integration of its @xmath187 profile . ) the second specific frequency problem is evidenced by the shallower decline of @xmath111 compared to @xmath192 at @xmath226 kpc .
this also causes the first problem , since scaling the total gcs density upwards by the factors appropriate for m49 or ngc 1399 obviously puts @xmath111 well in excess of @xmath192 throughout most of the galaxy .
indeed , the most recent estimate for the global @xmath3 of m87 is @xmath227 ( harris et al .
1998 ) , which is fully a factor of three higher than that of m49 .
here it is necessary to consider the distribution of hot gas which is known to exist in great quantities around m87in evaluating @xmath80 from equation ( [ eq:210 ] ) .
the volume density profiles of circumgalactic and intracluster gas are routinely derived in the course of x - ray studies of elliptical galaxies . to include such a halo component in this discussion , it is first necessary either to deproject observed gcs and stellar surface densities , or to compute the gas surface density via a standard projection integral . in general , the first of these options potentially has the most insight to offer , since projected quantities always mix information from a wide range of physical radii .
an appendix therefore develops a nonparametric deprojection algorithm which is explicitly designed to handle gcs data that have been binned in projected radius , @xmath86 ( i.e. , counted in circular annuli ) .
this procedure returns a coarse volume density profile@xmath179 vs. three - dimensional @xmath12expressed as a series of average densities in a number of concentric spherical shells .
the algorithm is preferably applied to surface density profiles that extend to fairly large radii ( so that corrections for cluster populations beyond the observed field of view can be estimated with confidence , and are as small as possible ) , and it makes most sense for @xmath121 distributions that are monotonically decreasing functions of @xmath86 ( otherwise , the derived @xmath179 can be zero or negative in some shells ) . after re - binning to smooth some of the original data , as described above , the composite gcs radial distributions in tables [ tab2 ] through [ tab4 ] satisfy each of these criteria .
equation ( [ eq:37 ] ) of the appendix has therefore been applied to each of these @xmath121 profiles , with @xmath228 ( a sort of `` residual background '' correction ) calculated by assuming that @xmath229 beyond the last observed radius of each gcs .
( this reflects the fact that the projected @xmath121 shows roughly an @xmath230 behavior at large @xmath86 in each case . in any event
, only the outermost one or two points in each deprojected profile are affected and then not greatly by this choice . )
the resulting @xmath179 are listed in the fifth columns of tables [ tab2 ] , [ tab3 ] , and [ tab4 ] .
the radii in the first columns , which delimit projected annuli in reference to the surface density data , are identified with the inner and outer three - dimensional radii of spherical shells in reference to the volume densities .
= 4.0truein the deprojection algorithm is an iterative one : the volume density in a spherical shell with some mean radius is computed by first subtracting , from the projected density at that radius , the contribution of any globulars that actually lie at larger physical @xmath12 .
thus , the uncertainties of the derived @xmath179 at any two radii in one system are not independent .
the ( 1-@xmath231 ) errorbars quoted in tables [ tab2 ] , [ tab3 ] , and [ tab4 ] are the standard deviations of results from 1000 different deprojections in each case .
each trial deprojection in each gcs begins with a different @xmath232 profile , in which the surface density for any given annulus is drawn at random from a gaussian distribution that is centered on the value given in column 3 of the appropriate table , and that has a dispersion equal to the observational uncertainty there .
= 4.0truein the tabulated number densities @xmath179 have the same limiting magnitudes as the corresponding surface densities , and are therefore multiplied by the same two factors ( so effectively scaling to @xmath233 , and applying the mean globular cluster mass @xmath234 ) to find the total gcs mass densities @xmath235 .
these distributions are plotted as the filled circles in figs .
[ fig4 ] , [ fig5 ] , and [ fig6 ] , where the density in each spherical shell has been put at a radius @xmath236 , given in columns 4 of tables [ tab2][tab4 ] , calculated according to equation ( [ eq:39 ] ) . as in fig .
[ fig3 ] , the @xmath178 profiles have been scaled up by factors of a few hundred in order to compare them directly with the stellar mass densities @xmath237 ( thin solid lines in figs .
[ fig4 ] to [ fig6 ] ) .
these have been obtained by applying the algorithm of the appendix to the projected profiles @xmath107 derived above from @xmath104-band surface photometry ( see eq .
[ [ eq:32 ] ] ) .
it is immediately apparent that the good agreement seen in fig .
[ fig3 ] between the gcs and stellar densities in ngc 1399 and m49 ( for the data of caon et al .
1994 ) persists after deprojection , as should be the case .
more noteworthy is the fact that the discrepancy between @xmath238 and @xmath239 in m87 , while certainly still present , is somewhat less pronounced than that between @xmath111 and @xmath240 . for example , the truly local picture given by fig .
[ fig4 ] shows that the second @xmath3 problem here is confined to galactocentric radii @xmath24140 kpc , and does not extend inward to the @xmath242 kpc scales suggested by fig .
[ fig3 ] . this just reflects the fact that the projection of any two volume densities inevitably amplifies any small differences that may actually exist between them , and artificially associates disparities at large @xmath12 with smaller projected radii .
a similar comment applies to the comparison of @xmath178 with @xmath239 in m49 : it is now clear , from fig .
[ fig5 ] , that any argument for the second @xmath3 problem in this galaxy could only be made for three - dimensional radii greater than about 50 kpc , and then ( as in 3.1 ) only if king s ( 1978 ) @xmath187 profile which includes just _ a single data point _ at such large @xmath12were preferred over that of caon et al .
( 1994 ) .
= 4.0truein the gas densities in these systems have been obtained from published _ electron _ number densities by way of the relation @xmath243 where a mean particle mass @xmath244 ( in units of @xmath245 ) is equally appropriate for hot ( @xmath246@xmath247 k ) and highly ionized plasmas of any composition from primordial to solar .
the open squares for the m87 gas in fig .
[ fig4 ] come from scaling the electron densities of nulsen & bhringer ( 1995 ) for the different virgo distance assumed here ( 15 mpc vs. their 20 ) , according to @xmath248 .
( these data were published as volume densities after nulsen & bhringer applied a deprojection algorithm , similar to that derived in the appendix , to their _ rosat _ spectra . ) the dotted line in fig .
[ fig4 ] is a fit taken from mclaughlin ( 1999 ) : @xmath249 so that , essentially , @xmath250 over the area of interest here .
this fit has been used to compute the total @xmath251 at every radius with a measured stellar density ; that sum is drawn as the bold solid line in the figure .
similarly , the open squares in fig . [ fig5 ] here use the @xmath252 read from irwin & sarazin s ( 1996 ) figure 6 , after scaling to @xmath98 mpc from their assumed m49 distance of 25.8 mpc .
the dotted line in this figure is a scaled fit ( from brighenti & mathews 1998 ) to these data plus those of trinchieri et al .
( 1986 ) : @xmath253 \
\ m_\odot\,{\rm pc}^{-3}\ . \label{eq:312}\ ] ] the heavy black line again represents the sum of @xmath254 .
the broken line for @xmath255 in ngc 1399 ( fig . [ fig6 ] ) comes from tsai s ( 1993 ) fit to the x - ray observations of killeen & bicknell ( 1988 ) , corrected for @xmath184 mpc : @xmath256^{-0.615}\ . \label{eq:313}\ ] ] it is clear that stars outweigh the gas everywhere in m49 and ngc 1399 : locally , @xmath257 even at the largest radii in figs .
[ fig5 ] and [ fig6 ] ; globally , @xmath258 within @xmath259 kpc and 70 kpc respectively . in both systems , then , @xmath260 and ( albeit to slightly lower accuracy at very large @xmath261 ) @xmath262 .
it is indeed valid to use standard specific frequencies as direct measures of globular cluster formation efficiencies in these two cases , as was done at the end of 3.1 above . to be complete , however ,
all of the stars , gas , and globular clusters can now be used to estimate @xmath88 more precisely , according to the basic equation ( [ eq:210 ] ) . by summing the data at large @xmath12 in each of m49 and ngc 1399 , then ,
average @xmath80 values are obtained : @xmath263 and @xmath264 which account for the scalings applied to @xmath111 and @xmath178 in figs .
[ fig3 ] , [ fig5 ] , and [ fig6 ] .
as has already been discussed , these are equally well interpreted as either local or global formation efficiencies , since the ratio @xmath265 shows no variations with @xmath12 beyond a stellar effective radius ( or even inside this , in ngc 1399 ) , i.e. , outside those regions of the gcss which could be significantly dynamically evolved .
( as in fig
. [ fig3 ] , the stellar @xmath200 are marked with vertical broken lines in figs .
[ fig4 ] , [ fig5 ] , and [ fig6 ] . )
the situation is somewhat different for m87 : referring to fig .
[ fig4 ] , @xmath266 at all radii @xmath267 kpc ( which is where the second @xmath3 problem begins in the deprojected data ) , and @xmath268@xmath269 globally ( i.e. , integrated over scales @xmath270 kpc ) .
the hot gas around m87 is therefore an important factor in calculating its cluster formation efficiency . as before ,
outside of the central regions of the galaxy it is found that @xmath271 explaining the scaling factor applied to the m87 gcs densities in figs .
[ fig3 ] and [ fig4 ] . figure [ fig4 ] shows that this average value is valid also as the local @xmath272 at any galactocentric position beyond about @xmath273 kpc ; as in m49 and ngc 1399 , the inferred globular cluster formation efficiency in m87 was independent of @xmath12 at large radii . in this case , however , if the gas were not included in the calculation , much larger and radially varying formation efficiencies would have been indicated : for example , @xmath274 , and @xmath275 .
note that the first of these ratios , @xmath276 for @xmath277 kpc , is a factor of 2.4 higher than the corresponding quantity in m49 . in combination with the different stellar mass - to - light ratios in these two galaxies ( @xmath278 vs. 10.4
; see van der marel 1991 , or fig .
[ fig3 ] above ) , this just corresponds to the factor of 3 difference observed in their global @xmath3 values .
the high local and global specific frequencies in m87 are thus seen to result from the presence of a population of globular clusters that are intimately associated with the hot gas there . as was mentioned earlier ,
an alternate comparison of a galaxy s gas distribution with its stars and globular clusters could make use of only surface densities@xmath192 , @xmath111 , and a derived @xmath279as functions of projected @xmath86 .
the advantage of such an approach is that @xmath279 is on an equal physical footing with @xmath192 and @xmath111 , and the latter are basically unmanipulated data with smaller relative uncertainties than the derived @xmath239 and @xmath178 profiles .
the cost , aside from possible uncertainties in any extrapolations required to obtain @xmath279 from @xmath255 , is that differences between projected distributions are not easily related to local effects at some true , three - dimensional radius ( e.g. , recall the comparison between projected @xmath111 vs. @xmath192 and deprojected @xmath178 vs. @xmath239 in m87 ) .
more to the point , if the ratio @xmath280 varies as a function of @xmath86 in a galaxy , then variations in the physical quantity of real interest , the local @xmath281 , are clearly indicated , but _ can not be properly quantified _ without resorting to a deprojection of some sort . for this reason
, the procedure of 3.2 must be favored as the best way , generally speaking , to measure local @xmath80 .
that said , however , in the galaxies studied here , @xmath80 is found to be independent of radius at large @xmath12 , i.e. , @xmath282 . in this specific instance
, it should then also happen that @xmath283 at large @xmath86 , with a constant of proportionality that is exactly @xmath80 .
it is therefore useful to compute @xmath279 and the projected analogue of @xmath80 in m87 , m49 , and ngc 1399 , in order both to verify the results of 3.2 ( which depend in good part on the behavior of @xmath178 at large @xmath12 , where the deprojected data are noisiest ) and to obtain potentially higher - precision estimates of the cluster formation efficiencies in these systems .
first , then , the projected mass densities of gas are obtained from the volume density profiles in equations ( [ eq:311 ] ) , ( [ eq:312 ] ) , and ( [ eq:313 ] ) , by way of the standard integral @xmath284 where the coordinate @xmath285 measures distance along the line of sight , and @xmath286 represents the total physical extent of the x - ray corona . for m87 ,
this is taken to be the virial radius of the virgo cluster : @xmath287 mpc ( see mclaughlin 1999 ) . for m49 , which is not at the center of virgo , @xmath288 kpc is adopted .
ngc 1399 lies at the center of the fornax cluster , which is somewhat less massive than virgo ( e.g. , jones et al .
1997 ) , and @xmath286 is set at 1 mpc there .
there is a certain arbitrariness to each of these choices for @xmath286 , but any of them can be changed by a factor of a few without affecting the conclusions which follow .
= 4.0truein the resulting @xmath279 profiles are shown in fig .
[ fig7 ] , along with the same @xmath192 and @xmath111 profiles plotted in fig .
[ fig3 ] above ( although the @xmath111 have been scaled up by slightly different factors in the current figure ) .
it is easily seen that @xmath283 at large @xmath86 in each galaxy , just as expected following the analysis in 3.2 .
moreover , the constants of proportionality are in reasonable agreement with the @xmath80 derived from the deprojected data : by taking the direct mean of the surface - density ratios beyond the galaxies effective radii ( so as once again to avoid a zone of potential gcs erosion ) , it is found that @xmath289 the result for m49 is indistinguishable from that obtained in either 3.1 or 3.2 , owing to the truly low gas densities in this galaxy .
similarly , the projected @xmath80 in ngc 1399 is identical , within the uncertainties , to the deprojected quantity @xmath290 .
the number in equation ( [ eq:318 ] ) is formally 10% smaller because of the apparently large projected gas densities at radii @xmath29150 kpc in ngc 1399 .
however , these values are almost certainly overestimates : in deriving @xmath279 , the shallow volume density profile of equation ( [ eq:312 ] ) , which roughly has @xmath292 , has been assumed to hold even on mpc spatial scales in fornax
well beyond the range of actual x - ray observations and this is likely unrealistic . on the other hand , the projected @xmath293 indicated for m87 in fig .
[ fig7 ] might genuinely be slightly smaller than the @xmath294 inferred from deprojected quantities .
the 20% difference between these two numbers is significant at just the 1-@xmath231 level , however , and although it is interesting that the former is actually closer to the values for m49 and ngc 1399it is not obvious which is more reliable . the x - ray gas density around m87 ( eq . [ [ eq:311 ] ] )
has again been extrapolated to unobserved radii to compute @xmath279 , but in this case the basic profile actually follows the dark matter distribution in virgo ( mclaughlin 1999 ) , and exhibits a realistic , steepening slope at large @xmath12 .
thus , although it is possible that @xmath295 is overestimated at large @xmath86 in m87 , any argument for this is not as compelling as in ngc 1399 . the best measure of @xmath80 in m87 will therefore be taken as the error - weighted mean of the values ( eqs .
[ [ eq:316 ] ] and [ [ eq:318 ] ] ) obtained by analyzing the deprojected and projected data there ; in m49 , the choice is of no consequence , but for definiteness the deprojected @xmath80 will be adopted ; and the deprojected estimate will also be adopted for ngc 1399 , to avoid any concerns that its gas surface density might have been overestimated at large @xmath86 .
thus , @xmath296 once again , these can be seen as estimates of either local or global cluster formation efficiencies if the inner several kpc of m87 and m49 are excluded from the discussion .
the results of the previous section clearly show that _ working in terms of stellar masses rather than luminosities and including the gas in m87 removes both the first and second specific frequency problems there _ : in all three of m87 , m49 , and ngc 1399 , it is found that @xmath282 , in identical proportions within the uncertainties , outside of the central , possibly dynamically depleted regions of the gcss .
( at the most , differences between the constants of proportionality in m87 and m49 might remain at the @xmath1820% level ; but again , this is easily taken up in the observational uncertainties , and in any case is a significant improvement over the 300% difference in their specific frequencies . )
the interpretation put to this is that globular clusters formed with an essentially constant efficiency throughout all of these galaxies , at the level of @xmath297 which is the error - weighted mean of equation ( [ eq:319 ] ) .
it should be kept in mind that these are all large galaxies with complex evolutionary histories .
thus , as was discussed in 2.1 , what has been measured here is necessarily a _ mass - weighted average _ of @xmath88 , possibly over multiple distinct episodes of in situ cluster formation and/or of the accretion of globulars formed elsewhere .
nevertheless , the uniformity of the observational estimates @xmath298 in these systems both as a function of @xmath12 within each and from one entire system to another suggests that these concerns may not be too severe , and that equation ( [ eq:41 ] ) may still serve as a reliable guide to a single , possibly universal efficiency for cluster formation .
this impression can now be checked quantitatively , albeit to rather lower precision than has proved possible with the excellent datasets for m87 , m49 , and ngc 1399 , by appealing to observations of many other gcss .
all indications are that @xmath4 is indeed similar , at least to first order , from galaxy to galaxy and from the protogalactic epoch to the present . given the definition for @xmath80 ( eq . [ [ eq:210 ] ] ) , and denoting by @xmath299 the global ratio @xmath300 in a galaxy , the total population of any gcs may be written as @xmath301 this can , of course , also be derived from equation ( [ eq:22 ] ) above , which explicitly relates @xmath3 to @xmath80 when @xmath302 .
now , in the general discussion of 2 , the stellar mass - to - light ratio was taken to be roughly similar from galaxy to galaxy .
this is not exactly true , however : as was seen in 3 , for example , the recovery of similar @xmath80 in m87 and m49 from their significantly different global and local @xmath3 depends on the @xmath1040% larger @xmath190 of m87 as well as on its greater gas mass .
thus , if other ellipticals are to be brought properly into this discussion , it is necessary first to account for the fact that @xmath303 depends systematically on galaxy luminosity .
this is a well known result of many fundamental - plane analyses of bright ellipticals , and it can be incorporated quantitatively in this analysis by referring to the study of van der marel ( 1991 ; and references therein ) .
specifically , for a sample of 37 early - type systems , van der marel obtains a mean @xmath304-band mass - to - light ratio of @xmath305 , corresponding to @xmath306 , at a mean magnitude of @xmath307 . for a galaxy color of @xmath308 ( typical of large ellipticals ) and @xmath309
km s@xmath91 mpc@xmath91 , this result corresponds to @xmath310 at an absolute magnitude of @xmath311 , or @xmath312 . in addition to this
, van der marel ( 1991 ) finds that @xmath303 increases with @xmath313 as a power law with exponent @xmath314 . since other studies of the fundamental plane
have claimed a slightly shallower exponent of 0.250.30 ( e.g. , faber et al .
1987 ; pahre , djorgovski , & de carvalho 1995 , and references therein ) , @xmath315 is adopted here .
thus , @xmath316 it should be noted again that these mass - to - light ratios refer to the stellar populations in the cores of the galaxies , and are not unduly influenced by any dark matter on larger scales . if equation ( [ eq:43 ] ) is used in ( [ eq:42 ] ) , the dependence @xmath317 , or @xmath318 , emerges naturally for constant @xmath80 and @xmath299 .
this is essentially the scaling found empirically by kissler - patig ( 1997 ) .
the other point of concern is , of course , the hot gas content of ellipticals in general . to get a rough handle on this , notice that the ( soft ) x - ray luminosities of early - type systems scale with their blue luminosities as @xmath319 , with @xmath3203 ( and with significant scatter about the mean trend :
e.g. , forman et al .
1985 ; brown & bregman 1998 ) .
this scaling can be related to the ratio @xmath321 by noting that @xmath322 for the bremsstrahlung emission that dominates @xmath94 in this context .
also , the slope @xmath323 of the @xmath94@xmath324 relation is the same if @xmath325 is used instead , so the implication is that @xmath326 . to go further
, @xmath327 bears some relation to the stellar velocity dispersion : @xmath328 , with @xmath3291.5 ( davis & white 1996 ; brown & bregman 1998 ) ; the @xmath23-band fundamental plane gives for the stellar quantities , @xmath330 and @xmath331 ( e.g. , de carvalho & djorgovski 1989 , 1992 ; pahre et al .
1995 ) ; and , from equation ( [ eq:43 ] ) just above , @xmath332 .
putting all of this together , @xmath333 .
evidently , the uncertainties in @xmath323 and @xmath334 are not crippling to this analysis , as @xmath335 is fairly representative of the range of possibilities
. the scaling can be normalized to apply specifically to brightest cluster galaxies ( bcgs ) by using the observations of m87 , which lies at the spatial and dynamical center of the virgo cluster .
as was mentioned in 3.2 , the global ( @xmath270 kpc ) ratio of gas to stars in m87 is @xmath336 by mass .
since the @xmath23-band luminosity of the galaxy is about @xmath337 ( from de vaucouleurs & nieto 1978 , for @xmath98 mpc ) , it is then reasonable to set @xmath338 putting this and equation ( [ eq:43 ] ) both into equation ( [ eq:42 ] ) , and holding @xmath80 fixed , it can be seen that the mean trend expected for @xmath22 between @xmath339 and @xmath340 ( so between an implied @xmath299 of @xmath100.1 and 1.6 ) is roughly @xmath341 . despite the crudeness of the estimate of @xmath299 used here ,
the power 1.75 is just that inferred from observational scalings presented by harris et al .
( 1998 ) for a large sample of bcgs .
the full relation between @xmath22 and @xmath9 is drawn as the heavy black line in fig .
[ fig8 ] , where the total gcs populations of 65 early - type galaxies are plotted against their @xmath23 luminosities .
these data are the same as those plotted in figure 11 of harris et al .
( 1998 ) , and the original references are given in that paper .
the open circles in fig .
[ fig8 ] here denote bcgs in a large number of rich clusters and a few poor groups ; the filled symbols refer to other ellipticals , some in clusters and some more isolated .
the bold curve has @xmath342 , constant in all galaxies . also shown , as the dashed line ,
is the result of keeping this uniform @xmath80 , but setting @xmath343 in equation ( [ eq:42 ] ) while still having @xmath85 increase with luminosity as in equation ( [ eq:43 ] ) .
the three dotted straight lines are loci of constant specific frequency , @xmath33 , as given by equation ( [ eq:21 ] ) for ( from top to bottom in the figure ) @xmath344 , 5 , and 1.5 . = 4.0truein it is immediately obvious why a roughly constant @xmath345 is often assumed for large ellipticals such as these .
it is equally clear , however , that such a simple characterization tends to overpredict @xmath22 at low @xmath9 , and systematically underestimates it at high luminosities .
accounting for @xmath346 goes some of the way towards explaining this while allowing for a universal formation efficiency @xmath4 ; and referring the gcs populations in bcgs to the galaxies total stellar _ plus gas _ masses seems indeed to remove the rest of the first specific frequency problem in general .
it is encouraging that these rather broad considerations lead to a significantly improved description of the observed @xmath22@xmath347 dependence , in which there are no obvious _
systematic _ deviations from a uniform @xmath348 ( see also 4.4 ) .
there remains a good deal of scatter about the mean line in fig .
[ fig8 ] , but this could well reflect departures of individual galaxies from the fundamental plane , the mean @xmath94@xmath324 ( or @xmath299@xmath9 ) relation , or the assumed @xmath308 , rather than any large variations in the fundamental efficiency of cluster formation .
note , for example , that the observed mass - to - light ratio of m87 is @xmath349 , some 80% higher than that expected from equation ( [ eq:43 ] ) . correcting for this would bring the @xmath350 curve in fig .
[ fig8 ] just up to the m87 data point there . to rather a lesser degree ,
some of the scatter could also arise from gcs - to - gcs differences in the relative magnitude of any dynamical reduction of @xmath351 over a hubble time .
it should be noted that this explanation of the first @xmath3 problem differs somewhat from those previously proposed ( see 2.2 ) . in particular , blakeslee ( 1997 ) and
blakeslee et al . ( 1997 ) showed a correlation between the specific frequencies on 40-kpc scales in bcgs , and the x - ray luminosities over much larger volumes ( @xmath352 kpc ) in their host clusters .
these authors , along with west et al .
( 1995 ) and harris et al .
( 1998 ) , therefore interpret the phenomenon of high specific frequency as one associated in some way with entire clusters of galaxies .
the analysis presented here , however , has attempted to address the problem more locally , by referring to the gas and stellar masses on similar spatial scales ( typically @xmath353 kpc ) to estimate @xmath299 ( eq . [ [ eq:44 ] ] ) in individual systems ; nowhere has any appeal been made to more global properties of galaxy clusters .
this general approach is motivated , of course , by the observed situation in m87 specifically ( 3 ) .
the correlation that does exist between bcg @xmath3 and cluster - wide @xmath94 ( and other such relations ; see the references cited ) may simply stem from the fact that the gas around bcgs in rich clusters blends smoothly into the intracluster medium as a whole .
note that a relation of some kind between `` local '' @xmath94 and @xmath22 is also implied by santiago & djorgovski ( 1993 ) , although they attach no specific interpretation to it .
interestingly , the bold line in fig .
[ fig8 ] departs significantly from linearity ( i.e. , from a simple power - law scaling between @xmath22 and @xmath9 ) for @xmath3540.2 , corresponding to @xmath355@xmath356 or @xmath357 .
this magnitude is similar to that which appears to divide ellipticals into two broad classes in terms of their kinematics , isophote shapes , and light distributions .
( lower - luminosity ellipticals generally show rapid rotation , flattened and disky isophotes , and steep cusps in their central densities , while brighter systems are slowly rotating , with round and boxy isophotes and shallower central density profiles ; see , e.g. , davies et al .
1983 ; bender et al . 1989 ; nieto , bender , & surma 1991 ; tremblay & merritt 1996 ; gebhardt et al . 1996 ; faber et al . 1997 . ) this bears directly on the recent discussion of kissler - patig ( 1997 ; see also van den bergh 1998 ) , who used observed trends in gcs specific frequencies and radial distributions as further evidence for the empirical `` dichotomy '' of ellipticals .
in particular , kissler - patig noted that the cluster systems in early - type galaxies with @xmath358 tend to have lower specific frequencies and steeper @xmath121 profiles than those in brighter galaxies .
although he went on to argue for a discontinuous change in the cluster formation efficiency for galaxies with @xmath325 below and above @xmath359 , fig .
[ fig8 ] now shows that all of the data are well represented by a _ continuously _
increasing @xmath3 that derives from a _ constant _ @xmath4 and a _ smoothly increasing _ gas mass fraction , @xmath299 ; the nonlinearity intrinsic to the resulting @xmath22@xmath9 relation just happens to roughly mimic a break of sorts around @xmath360 .
this interpretation can also account for another correlation suggested by kissler - patig ( 1997 ) , namely , that @xmath3 appears to be lower in ( faint ) ellipticals whose isophotes show disk - like deviations from pure ellipses , and higher in those ( brighter ) galaxies with more box - like shapes .
these isophotal perturbations are themselves correlated with the x - ray luminosities of ellipticals ( bender et al .
1989 ) , in the sense required qualitatively to explain this observation of gcss again in terms of a constant @xmath4 and a systematically increasing gas - to - stellar mass ratio .
this raises the interesting question of which , of gas content or isophote shape , is the more fundamental property of elliptical galaxies ; the present discussion , at least , would appear to favor the first alternative . finally , although the slopes of gcs radial distributions can not be addressed directly by fig .
[ fig8 ] , it seems natural that , given a fixed @xmath4 , ellipticals with large gas contents ( and thus high global @xmath3 ) should generally have shallower gcs density distributions , relative to their stellar profiles , at large galactocentric radii ( simply because @xmath361 usually increases with @xmath12 in early - type galaxies ) .
as was suggested in 2.2 , then , the first and second @xmath3 problems should be closely related , with @xmath362 the controlling factor in both .
again , this is consistent with the observed trend ( noted by forbes et al .
1997 ; kissler - patig 1997 ; and van den bergh 1998 ) towards weaker dependences of @xmath121 on @xmath86 in brighter and higher-@xmath3 galaxies .
this proposal should , of course , be checked on a case - by - case basis , as it was for m87 in 3.2 above , in as many systems as possible . if the bold line in fig .
[ fig8 ] is extrapolated into the regime of dwarf ellipticals , @xmath363 , it clearly implies that global @xmath364 should be observed in these faint galaxies .
this is completely at odds with the data of durrell et al .
( 1996 ) and miller et al .
( 1998 ) , who find an average @xmath30 for a large sample of de s in the local group , in other groups nearby , and in the virgo and fornax clusters .
in addition , miller et al . claim evidence for an _ increasing @xmath3 towards fainter @xmath9_. this can be explained by first recalling that dwarf ellipticals do not fall on the fundamental plane of bright galaxies ; thus , the scalings ( [ eq:43 ] ) and ( [ eq:44 ] ) do not apply to them .
also , these galaxies are bluer than their giant counterparts , so that the conversion to @xmath23-band mass - to - light ratios from measured @xmath365 , as described above , is again inapplicable ; and in any case , even the cores of dwarfs appear to be dominated by dark matter , so that their dynamical @xmath85 are not necessarily representative of the stellar values . most importantly and likely at the heart of these other points
de s present sufficiently shallow potential wells that they should have suffered significant amounts of mass loss during supernova - driven winds in early bouts of star formation .
thus , even though these galaxies are gas - poor at the present , the appropriate observable @xmath366 _ is not an accurate estimator of the true cluster formation efficiency _
@xmath367 ( see durrell et al .
1996 for the first suggestion that this is an important consideration in understanding dwarf gcss ) . even still assuming that the gcss now have @xmath368 , allowance has to be made for an additional factor @xmath369 .
the relation analogous to equation ( [ eq:42 ] ) in this case is therefore @xmath370 where @xmath371 should be a good representation of the stellar populations in systems fainter than about @xmath372 .
= 4.0truein the specific model of dekel & silk ( 1986 ) , which considers supernova - driven winds in dwarf galaxies that are dynamically dominated by dark matter halos , can be used to evaluate the ratio of initial gas to present stellar mass .
that study finds that the winds from systems with velocity dispersions less than about 100 km s@xmath91 , corresponding to @xmath373 at the current epoch , should have expelled whatever gas remained after a single burst of star formation .
moreover , as one moves towards shallower potential wells and fainter @xmath9 , the fractional mass lost , relative to the initial total , is naturally expected to increase . the predicted scaling , in the limit of very large mass loss and for a power spectrum similar to that in a standard cold dark matter cosmology , is roughly @xmath374 . taking the crude step of extrapolating this all the way to the critical velocity dispersion ( i.e. , to @xmath375 ) , and assuming that any gas driven from brighter and more massive galaxies was a negligible fraction of the total
, the implication is that @xmath376 substituting this in equation ( [ eq:45 ] ) , and assuming a constant stellar @xmath177 , yields @xmath377 and @xmath83 for dwarf ellipticals .
although the picture of de s as single - burst populations is now known to be incomplete ( see , e.g. , mateo 1998 ) , this result is nevertheless consistent with the observations of miller et al .
( 1998 ) . figure [ fig9 ] uses the data from durrell et al .
( 1996 ) and miller et al .
( 1998 ) to place 32 nucleated and non - nucleated dwarf ellipticals into the @xmath378 plane with the brighter galaxies from fig .
( of these , 16 are taken from durrell et al . , and 16 from miller et al . only gcss detected at the @xmath379-@xmath231 level
are used . ) the bold dashed line in the lower left of this plot comes from equations ( [ eq:45 ] ) and ( [ eq:46 ] ) , assuming a constant @xmath380 .
the heavy solid line is the same as that in fig .
[ fig8]i.e .
, it also has @xmath381 , but applies to fundamental - plane ellipticals and the light , dotted lines correspond , as before , to constant @xmath344 , 5 , and 1.5 . once again , all the data appear consistent with a universal globular cluster formation efficiency . it will be important to properly confirm this with detailed models for the de s in particular , and to flesh it out with observations of systems at @xmath382 in general .
one object that falls just inside this observational `` gap '' in luminosity is the stellar _
spheroid_disk excluded of the milky way .
the total @xmath23-band luminosity here has been estimated by de vaucouleurs & pence ( 1978 ) to be @xmath383 [ after applying a color of @xmath384 to their @xmath385 . at the same time
, the total number of globular clusters that are genuine halo objects , rather than part of the ( thick ) disk , is @xmath386120 .
( halo globulars are identified on the basis of their metallicity , @xmath387}\leq-0.80 $ ] ; refer back to fig .
[ fig2 ] above . )
this is the point plotted as the large open square in fig .
[ fig9 ] .
in fact , de vaucouleurs & pence ( 1978 ) model the galactic spheroid with an @xmath388 density law with effective radius @xmath389 kpc .
thus , in addition to predicting the absolute magnitude that would be seen by an external observer , it is also possible to parametrize the density profile of the entire stellar halo . to do this in three dimensions , young s ( 1976 ) deprojection of the @xmath388 law
may be used : @xmath390\ , \left({{r_{\rm gc}}\over{r_{\rm
eff}}}\right)^{-7/8}\ , \label{eq:47}\ ] ] which is valid for @xmath391 .
substitutions can be made for @xmath392 and @xmath200 directly from the work of de vaucouleurs & pence .
one way to estimate the @xmath104-band mass - to - light ratio of the galactic population ii is to adopt the same @xmath303@xmath324 scaling used in 4.1 , but to normalize the relation by adopting solar colors rather than the redder ones that apply in ellipticals .
van der marel s ( 1991 ) mean @xmath393 then corresponds to @xmath394 at @xmath395 ( for @xmath309 km s@xmath91 mpc@xmath91 assumed ) , and @xmath396 .
the implied @xmath397 for @xmath398 is consistent with the mean @xmath177 for globular clusters , and with the @xmath399 adopted for the spheroid by bahcall & soneira ( 1980 ) in their model of the galaxy .
equation ( [ eq:47 ] ) is therefore evaluated here with @xmath400 . the dashed line in the bottom panel of fig .
[ fig2 ] in 3.1 above shows this density profile , scaled down by a factor of @xmath401 this value is obtained as the mean of the ratio @xmath402 over all galactocentric radii @xmath403 kpc , and it accounts for the position of the milky way halo in fig .
such a basic correspondence between our galaxy and early hubble types was previously suggested , although developed from a somewhat different argument , by de vaucouleurs ( 1993 ) .
it is also evident from fig .
[ fig2 ] that there is no second @xmath3 problem in the milky way : @xmath404 obtains throughout the entire halo at @xmath405 kpc so again , beyond roughly an effective radius , within which the original gcs might have suffered significant dynamical depletion .
this point has long been appreciated , although it is usually discussed in the context of the fact that the same rough rule @xmath406 is inferred both for the spheroid , from counts of rr lyrae and horizontal - branch stars ( e.g. , preston , schectman , & beers 1991 ; kinman , suntzeff , & kraft 1994 ) , and for the gcs , from simple power - law fits to its density profile ( harris 1976 ; zinn 1985 ; djorgovski & meylan 1994 ) .
this power law is shown as the dotted line in the bottom panel of fig .
[ fig2 ] .
= 4.0truein as a final test of the general viability of a single globular cluster formation efficiency , fig .
[ fig10 ] plots the ratio of observed to predicted gcs @xmath22 as a function of host galaxy luminosity for all of the 98 systems discussed thus far .
the predicted @xmath22 for fundamental - plane ellipticals and the milky way spheroid ( filled circles and open square in the figure ) are given by equation ( [ eq:42 ] ) , with the gas mass fraction @xmath299 set identically to zero ; for bcgs ( open circles ) , by the same relation but with @xmath299 as in equation ( [ eq:44 ] ) ; and for de s ( filled squares ) , by equation ( [ eq:45 ] ) . in all cases , it has been assumed that @xmath381 . the median of the resulting ratios @xmath407 is 1.04 , and the geometric mean is 0.98 . close to 80% ( 77/98 ) of these galaxies
have @xmath408 . as was also mentioned in 4.1 , but can perhaps be seen more clearly here than in fig .
[ fig8 ] or fig . [ fig9 ] , the deviations of observed @xmath35 from the uniform-@xmath4 prediction show no significant dependence on either galaxy luminosity or hubble type .
there are possibly two exceptions to this claim : ( 1 ) the gcs populations in the brightest bcgs , @xmath409 , appear to fall below the values predicted with @xmath381 , by perhaps @xmath410 on average ; and ( 2 ) the two faintest dwarfs , which happen to be the local group spheroidals fornax and sagittarius , both have @xmath22 greater than predicted .
these rough impressions may well not remain , in the mean , after future additions of gcs data from other very high- and low - luminosity galaxies ; but even if they do , they need not imply the existence of fundamental variations in @xmath4 . rather , there is likely enough uncertainty in the adopted relation between gas mass fraction and @xmath9 ( eq .
[ [ eq:44 ] ] ) to account for the first item ; and the second could reflect either a deficiency in the very simple correction ( 4.2 ) for gas blow - out from very faint dwarfs , or the neglect of possibly extensive post - formation dynamical evolution in such low - mass galaxies ( such as tidal stripping of their field stars ; recall , for example , that sagittarius is in the process of being accreted by the milky way ) . even if these issues are ignored completely , a least - squares fit to all of the data shown in fig .
[ fig10 ] reveals an exceedingly weak dependence of @xmath411 on galaxy luminosity : @xmath412 .
fitting to only the fundamental - plane ellipticals and bcgs yields a steeper @xmath413 or so , but this result appears to be driven by just three ge s with anomalously low or high gcs populations ( cf .
[ fig8 ] ) : ngc 3557 , with @xmath414 and @xmath415 ; ngc 5018 , which has @xmath416 and @xmath417 ; and ngc 4278 , with @xmath418 but @xmath419 .
if these systems are excluded , then @xmath420 is indicated instead .
similarly , the slopes of @xmath421 vs. @xmath9 do not differ significantly from 0 for any of the individual hubble - type samples represented in fig .
[ fig10 ] .
thus , current observations of total gcs populations do show some scatter , and it will be important to understand
likely through detailed analyses , along the lines of 3 , of more systems _ individually_whether or not this reflects an intrinsic scatter in the underlying globular cluster formation efficiency ( see also 4.1 ) .
however , there is no clear evidence at present for any systematic departures from a mean global efficiency of @xmath422 .
as was discussed in 2.1 , the formation of stellar clusters continues to be an important element of star formation in general , and the existence of an apparently universal formation efficiency for globular clusters must reflect a generic piece of this process .
however , the picture may still be incomplete : although the discussion to this point has included clusters that formed in a wide variety of environments at galactocentric distances @xmath423100 kpc , in galaxies spanning four orders of magnitude in total luminosity and inhabiting both poor groups and rich clusters all of these formed at essentially a single epoch , and under physical conditions that were possibly much more extreme than those in , say , the disk of the milky way today .
it is important , therefore , to compare the protogalactic @xmath424 to the frequency with which bound star clusters form in a typical quiescent galaxy at the current epoch .
disk clusters in the milky way present one opportunity to do this .
the rate of open cluster formation in the solar neighborhood has been estimated by elmegreen & clemens ( 1985 ) as @xmath425 clusters kpc@xmath188 yr@xmath91 . if the masses of the clusters range between @xmath426 and @xmath427 according to @xmath428 ( from the luminosity spectrum @xmath429 of van den bergh & lafontaine 1984 ) , then a mean mass of @xmath430 is indicated , and the formation rate becomes @xmath431 kpc@xmath188 yr@xmath91 .
this is to be compared with the _ total _ star formation rate in the solar neighborhood , which includes those stars appearing in loose groups and unbound associations ( or even in relative isolation ) as well as in bound clusters .
this rate is @xmath432 kpc@xmath188 yr@xmath91 ( e.g. , mckee 1989 ) . since the mean stellar mass is no different in clusters than in the field , the ratio of these two rates can be taken directly , to assess the local _ number _ fraction of stars born in clusters : @xmath433 .
this is consistent with the classic estimate ( e.g. , roberts 1957 ) of about 10% .
however , the globular cluster @xmath80 has been obtained here as a _
mass _ ratio , since this is the observable quantity that can be related directly to its formation value ( see eq .
[ [ eq:28 ] ] ) . to make a proper comparison with open cluster formation
, it is therefore necessary to refer the mass going into newborn clusters ( per unit area , per unit time ) to the total mass of gas available for any kind of star formation .
this latter figure can be derived from @xmath434 using the mean star formation efficiency , @xmath435 , averaged over many of the dense , star - forming ( not necessarily cluster - forming ) cores in giant molecular clouds .
this number is only poorly known , but statistical arguments ( e.g. , elmegreen 1983 ) suggest that it lies at the level of @xmath436%10% .
thus , @xmath437 which is in reasonably good agreement with the result for globular cluster systems . the numerical value in equation ( [ eq:49 ] ) clearly should not be taken as definitive , given the very rough , order - of - magnitude nature of the argument that has produced it .
that said , however , it is just the order - of - magnitude agreement between equations ( [ eq:49 ] ) and ( [ eq:41 ] ) which suggests that the basic mechanism of cluster formation may not have changed appreciably from protogalactic times to the present .
another obvious place to check the current level of @xmath4 is in starbursts and merging galaxies , which are forming massive young clusters in substantial numbers ( e.g. , whitmore & schweizer 1995 ; meurer et al .
complete inventories of the star - forming gas mass in individual systems , and of the total mass in those of their young clusters that can be shown to be gravitationally bound ( which to date has been done rigorously for only two objects : ho & filippenko 1996a , 1996b ) , would be of considerable interest .
the results should be of particular relevance to the oft - made claim , which is now at the center of some debate ( sternberg 1998 ; brodie et al .
1998 ) , that these super star clusters are the modern - day equivalents of young globulars .
if it turns out , for example , that strongly bound clusters in starbursts are forming in numbers very much in excess of those expected on the basis of @xmath438 ( as , e.g. , the uv observations of meurer et al . 1995 appear to suggest ) , then these spectacular events might _ not _ be representative of a ubiquitous phase in the formation of average elliptical galaxies .
the fact that the @xmath4 values inferred here are everywhere so similar over large ranges in radius inside m87 , m49 , and ngc 1399 ; from dwarf ellipticals to brightest cluster galaxies ; and from massive , old globular clusters to smaller , much younger open clusters should serve as a strong constraint on theories of star formation .
the simplest interpretation is that the probability of attaining a cumulative star formation efficiency of @xmath43950% ( see 2.1 ) in any dense clump of gas more massive than @xmath440@xmath15 ( so from open clusters on up ) can depend only weakly , on average , on local environment ( i.e. , the density and pressure of ambient gas ) or on factors such as global background density ( meaning , e.g. , in the protogalactic context , a 1-@xmath231 vs. a 3-@xmath231 density fluctuation ; or a poor group of galaxies vs. a rich cluster ) .
moreover , implicit in plots like figs .
[ fig8 ] and [ fig9 ] is the fact that cluster formation efficiencies are also rather insensitive to metallicity : the mean abundances of the globulars in the galaxies represented there range over two orders of magnitude , from @xmath441\simeq-2 $ ] to @xmath441\simeq0 $ ] ( see durrell et al . 1996 ; forbes et al .
1997 ) . over a remarkably broad spectrum of physical conditions
, therefore , it appears that similar fractions of star - forming gas always about 0.25% by mass are able to produce bound stellar clusters .
the implicitly mass - averaged nature of the observational estimates of @xmath4 has to be kept in mind , but would seem to be of little concern in most cases .
it is therefore reasonable to place the efficiency of cluster formation alongside the globular cluster mass function @xmath11 ( mclaughlin & pudritz 1996 ) and the mean globular cluster mass @xmath176 ( e.g. , fig . [ fig2 ] above ) as robust and nearly universal physical quantities that must be derived from the most general aspects of the star formation process . as a corollary to this , one implication of figs .
[ fig8 ] and [ fig9 ] above is that the efficiency of _ unclustered _ star formation could _ not _ have been universal .
that is , in both dwarf elliptical and brightest cluster galaxies , where the data are consistent with @xmath442 having been constant in the mean , globular clusters apparently did form in precisely the numbers expected for the total amount of gas initially available to these systems .
but then , the higher than average specific frequencies there are inferred to reflect larger ratios of @xmath443 , so that lower than average amounts of this gas were converted into field stars .
the data for dwarfs agree with a simple scenario in which most of the unused gas supply is driven entirely out of the galaxies by a strong wind following one major burst of star and cluster formation .
if this is basically correct , then the globulars in de s quite possibly were able to form completely before the onset of such a wind , and they could conceivably have been instrumental in driving it ; at the very least , if the protoclusters were still largely gaseous , they must have been already sufficiently dense and well defined that they could survive such a catastrophic event essentially intact .
at the same time , field star formation must have been truncated by the wind , before reaching what would have been a `` normal '' efficiency of @xmath444 .
thus , _ the massive and dense clumps of gas which ultimately formed bound star clusters had to have collapsed more rapidly than those which produced unbound groups and associations_. this conclusion may hold quite generally one possible interpretation being that only those gas clouds in the highest - density tail of some distribution , in any setting , are able to achieve a high cumulative sfe and it may be related to the fact that the globular clusters in dwarfs are generally somehat bluer , and presumably more metal - poor , than the bulk of the field stars there ( durrell et al .
1996 ; miller et al .
1998 ) a similar feedback argument could also apply to bcgs , where it may be that especially strong early bursts of star and cluster formation led to the premature virialization of large amounts of protogalactic gas , and perhaps drove slow , _ partial _
galactic winds , most effectively in the low - density environs at large galactocentric radii ( cf .
harris et al . 1998 ) .
the unused gas would have to remain hot to the present day , and more or less in the vicinity of the parent galaxy , in order to enter the observed @xmath94@xmath324 relation as inferred in 4.1 .
this requirement is consistent with the fact that bcgs are situated at or near the centers of their galaxy clusters , and thus at the bottoms of very deep potential wells . in this scenario , again , globulars would have had to form , in just the numbers expected of them , somewhat before field stars were able to do the same ; and this truncation of unclustered star formation would have had to be more severe at large galactocentric radii ( where the feedback is strongest ) .
this could then account explicitly for the connection between the first and second specific frequency problems , i.e. , between high global @xmath3 and local @xmath204 ratios that increase with @xmath86 .
it also gives a more specific context to the general claim ( harris 1986 ) that the shallowness of gcs radial distributions relative to the stellar light profiles in some ellipticals resulted from the globulars having formed slightly in advance of the stars there .
the main demand on this picture for bcg formation is that it must be able to explain why @xmath300 increases with galaxy luminosity , i.e. , why more massive bcgs apparently had larger fractions of their initial gas mass virialized early on . as harris et al .
( 1998 ) discuss in some detail , this issue is likely related to the fact that brighter bcgs are generally found in more massive galaxy clusters , which may have presented higher - density and more turbulent environments that led to systematically higher star formation rates , and more violent feedback . although speculative , these ideas are suggested directly by the gcs data , and are consistent with all those available . since they are also obviously related to questions on the star - formation histories of de s , on the origin of the @xmath94@xmath324 correlation in the brightest ellipticals , and on the nature of the intracluster medium , they should be addressed with considerably more rigor than has been applied here .
it is particularly striking that the cluster formation efficiency derived in 4.3 for the milky way halo is essentially indistinguishable from that which applied in m87 , m49 , and ngc 1399 , and which appears to have held quite generally in ellipticals of most any description .
this calls into serious question one of the primary motivations for the model of ashman & zepf ( 1992 ) and zepf & ashman ( 1993 ) , who posit that ellipticals are formed primarily by the mergers of gas - rich spirals , and invoke preferential cluster formation during these events to account for what they claim is a typical factor of two difference in the observed ratios @xmath39 of late- and early - type systems .
although it is true that spirals have specific frequencies that are lower than those in bright e galaxies , by factors of perhaps 23 on average ( as is exemplified by the milky way in fig .
[ fig9 ] above ; see also harris 1991 ) , it is now apparent that this is not necessarily a reflection of similar discrepancies in their more fundamental @xmath298 .
if the milky way is typical , it suggests that even though stellar clusters certainly can form in merging galaxies_mergers are not required to explain the relative gcs populations of spirals and ellipticals in general_. conversely , whenever an elliptical does form by the major merger of two spirals , this is allowed to include cluster formation only at the standard efficiency of @xmath445 .
ashman & zepf ( 1992 ) and zepf & ashman ( 1993 ) also attribute the second specific frequency problem to the formation of ellipticals by mergers of spirals with pre - existing gcss . however , it is not clear that this is necessary either , if the extended spatial distribution of globulars in the brightest ellipticals is generally related to their association with hot gas at rather large galactocentric radii .
in fact , a basic expectation of the original ashman - zepf scenario is that the second @xmath3 problem should be _ less _ pronounced in galaxies with higher global @xmath3 ( which they hypothesize to have suffered greater numbers of major mergers ) .
the sense of this predicted trend is _ opposite _ to the one implied here with empirical support directly from the evidence in m87 , and indirectly from the correlations of forbes et al .
( 1997 ) and kissler - patig ( 1997)in which a constant @xmath88 combines with the larger @xmath300 in higher-@xmath3 ellipticals , to produce greater contrasts in their projected @xmath111 and @xmath192 profiles .
although radial - profile and @xmath3 data by themselves may not rule out the merger model for gcss altogether , a revision to take account of the hot gas in large ellipticals is essential if it is to be at all viable .
( see also forbes et al .
1997 and kissler - patig et al .
1998 for critical discussions of this and other aspects of the basic idea . )
lccclcclcc m87 & [email protected] & [email protected] & 2.177@xmath446 & & [email protected] & 1.522@xmath447 & & @xmath138 & @xmath138 & & & & & & & & & m49 & [email protected] & [email protected] & 1.170@xmath448 & & [email protected] & 0.985@xmath449 & & [email protected] & 3260@xmath171475 & & & & & & & & & n1399 & [email protected] & [email protected] & 1.466@xmath450 & & [email protected] & 1.859@xmath451 & & [email protected] & 3630@xmath171460 finally , the gcs density profiles derived specifically for m87 , m49 , and ngc 1399 in 3 also are relevant to issues of the generic spatial structure of gcss , and to the specific matter of the possible existence of intergalactic globulars in clusters of galaxies . to facilitate a discussion of these points , and for reference , table [ tab5 ]
presents the results of least - squares fits of a few different parametric functions to the spatial distributions of the m87 , m49 , and ngc 1399 gcss , as they are given in tables [ tab2 ] , [ tab3 ] , and [ tab4 ] above .
the functions include ( from left to right in table [ tab5 ] ) a power law with a constant - density core region ; the cusped density profile suggested by navarro , frenk , & white ( 1996 , 1997 ) as appropriate for cold dark matter halos ; and the law popularized by hernquist ( 1990 ) and applied widely to the light profiles of elliptical galaxies . in all cases , fits were performed by minimizing the residuals of the functions _ projected _ density profiles relative to the @xmath121 data , which have smaller observational uncertainties than the volume densities do .
the best - fit parameters thus obtained were then used to overlay the original , three - dimensional functions on the deprojected @xmath179 and verify that they provided good descriptions of those as well .
one consequence of this procedure is that no fit is given for the hernquist ( 1990 ) profile , @xmath452 , to the m87 gcs .
although it is possible to fit this function directly to the @xmath179 data , its projection affords a poor description of the full @xmath121 distribution .
this almost certainly reflects the fact that the gcs profile in m87 appears about to become significantly shallower , at least temporarily , beyond the last radius ( @xmath453 kpc ) for which cluster counts exist ( because fig .
[ fig4 ] suggests essentially that @xmath454 there ) .
such a change in slope is obviously not allowed by hernquist s density profile .
thus , the failure of this model in this case makes the point that _ any _ of the density fits in table [ tab5 ] should be taken at face value only over the ranges of @xmath12 that have been directly observed ; extrapolations to much larger radii could potentially introduce appreciable systematic errors .
table [ tab5 ] has three interesting implications : ( 1 ) none of the available data _ require _ the existence of truly constant - density `` cores '' in gcss .
this was also mentioned by mclaughlin ( 1995 ) and it is , of course , plainly seen in figs . [ fig4 ] through [ fig6 ] .
evidence is emerging for a similar situation in elliptical galaxies generally ( crane et al .
1993 ; gebhardt et al .
1996 ) , and apparently also in galaxy clusters ( merritt & tremblay 1994 ; carlberg et al . 1997 ; mclaughlin 1999 ) .
( 2 ) it has been suggested , particularly in the contexts of some dynamical analyses ( e.g. , huchra & brodie 1987 ; weil , bland - hawthorn , & malin 1997 ) that the m87 gcs directly reflects the distribution of dark matter in and around that galaxy . however , this is not true in any simple sense ; neither of the @xmath179 fits given in table [ tab5 ] provides an adequate description of @xmath455 in m87 . instead , it is found there that @xmath456 ( see mclaughlin 1999 ) , in which case fig . [ fig4 ] shows that the gcs becomes a reliable tracer of the dark matter halo _ alone _ only on fairly large spatial scales ( where the number densities become quite low ) .
( 3 ) the analytical fits to @xmath121 and @xmath179 offer a convenient way to extrapolate the observed densities to ( slightly ) smaller and larger radii than those observed , so as to compute the total gcs populations of these three galaxies .
thus , in m87 , it is found that @xmath457 clusters ( over all magnitudes ) are _ projected _ to distances @xmath458 kpc from the galaxy s center ; in m49 , there are @xmath459 globulars within the same projected @xmath86 ; and in ngc 1399 , @xmath222 clusters have @xmath460 kpc .
these total `` projected populations '' are in good agreement with those estimated in many other studies ( see the references cited in 3 ) .
however , it is now possible to determine the three - dimensional radius of the sphere which actually contains all of these clusters in each system . from either of the m87
fits in table [ tab5 ] , it is found that 13600 clusters are contained within @xmath461160 kpc ; in m49 , 6850 globulars come from the sphere @xmath462150 kpc ; and the 4700 globulars in ngc 1399 all have true galactocentric positions @xmath46395 kpc . to put this another way ,
all three gcss are sufficiently centrally concentrated that , in each case , some 85%90% of the globular clusters projected onto the galaxy so again , to within @xmath464 kpc for m87 and m49 , and @xmath465 kpc for ngc 1399really are located at those galactocentric distances in three dimensions .
this last point specifically can clarify and constrain the suggestion ( white 1987 ; west et al .
1995 ; ct et al .
1998 ) that the high specific frequencies in bcgs like m87 are due to the `` contamination '' of their gcss , in projection , by a significant number of globular clusters that are associated with a galaxy cluster as a whole , rather than bound to the central galaxy itself .
harris et al .
( 1998 ) tried to test this idea in some detail for m87 in particular , working from the assumption ( following white 1987 ) that any population of intergalactic globular clusters ( igcs ) should generally be associated with the diffuse stellar light of the cd envelopes that surround many bcgs ( i.e. , these structures were considered to comprise primarily intracluster stars ) .
this hypothesis leads to inconsistencies that prompted harris et al . to reject igcs as the main cause of the first specific frequency problem .
while their arguments still stand within the framework that they constructed , it is now seen that the high @xmath3 of m87 certainly , and likely those of other bcgs as well , can be traced to subsets of their gcss that are associated with hot gas , and not necessarily with stars at all .
this does appear to favor the basic notion of igcs , since the gas around m87 extends to very large spatial scales , and traces out the potential well of the virgo cluster as a whole ( mclaughlin 1999 ) .
if a constant @xmath80 holds all the way out , then , a cluster - wide population of globulars could be expected .
however , since _ all _ of the 13600 globular clusters which give m87 its high specific frequency in projection are physically associated with a volume of radius @xmath18150 kpc around the galaxy , it is not obvious that any of them are truly `` intergalactic '' objects at the current epoch .
projection effects are _ not _ solely responsible for the high @xmath3 of m87 , and although intergalactic globular clusters may well exist , it is not clear that any have yet been observed . of course , since m87 is at the dynamical center of virgo , it can be difficult to distinguish meaningfully between the galaxy and the cluster on 100-kpc scales . in the context of igcs
, the question evidently comes down to one of origin : did the globulars now seen right around m87 actually form there , or were many of them somehow brought in from much further than 150 kpc away ?
any self - consistent answer to this must be able to account for the observed fact that @xmath466 at galactocentric radii @xmath467100 kpc in m87 ; that is , both the very existence and the normalization of this proportionality have to be explained .
although essentially any scenario which is able to do this would be allowed by the current data , none could be _ preferred _ over what remains the simplest option : that the bulk of the m87 gcs formed in situ , with a thoroughly standard and spatially constant @xmath4 . ultimately , only detailed evolutionary models , tailored specifically to the properties of m87 and virgo and considering both stars _ and gas _ , can really decide what fraction of the galaxy s gcs might have come originally from the cluster at large .
stars form mainly in groups , some of which emerge from their natal clouds of gas as gravitationally bound clusters .
an understanding of cluster formation will therefore be integral to any full theory of star formation , and an important element of this is simply the frequency with which it occurs .
that is , what is the likelihood that a star - forming cloud of gas whether this be a dense clump in a galactic giant molecular cloud , or a larger one in a protogalactic fragment will achieve the high cumulative star formation efficiency ( viz .
@xmath43950% ) that allows its stars to remain bound as a cluster after they have cleared away the gas ?
this paper has used observations of the globular cluster systems in galaxy halos to address this question , and to empirically evaluate the efficiency of cluster formation : by mass , @xmath468% .
this is , by all appearances , nearly a universal number , and it should therefore serve as a strong constraint on theories of star and cluster formation in any context . to arrive at this result ,
it was first shown ( 2.1 ) that the total ( global ) _ masses _ of gcss are quite robust , even over a hubble time , against the dynamical processes that work to destroy low - mass globular clusters in the central regions of galaxies . with an observed @xmath36 thus useful as a measure of the initial quantity ,
the specific frequencies of gcss , @xmath469 , serve as _ crude _
first estimates of the basic formation efficiency @xmath1 . to the _
limited _ extent that @xmath30 is an adequate description of early - type galaxies , the implication is that @xmath470 .
the limiting factor in this simplest treatment of gcs data is the assumption that a galaxy s current stellar mass always suffices as an estimate of its total initial gas mass ; if taken too far , this quickly leads to some perplexing conclusions .
as was reviewed in 2.2 , systematic variations in @xmath3 exist along the entire sequence of early - type galaxies , from dwarf ellipticals ( where @xmath471 or so ) through normal giants ( @xmath472 ) and on to the brightest ellipticals , including many of the central galaxies in groups and rich clusters ( @xmath473 ) .
if specific frequency were indeed a good estimator of @xmath4 in general , the implication would seem to be that this also varied , for unknown reasons and in a non - monotonic fashion , as a function of galaxy luminosity .
closely related to this is the fact that the gcss of some , but not all , large ellipticals are spatially more extended than their stellar halos
. that is , local @xmath3 values increase with projected @xmath86 in some systems , which naively would suggest that again , for unknown reasons , and counter to any intuition the efficiency of cluster formation _ sometimes _ increased with galactocentric radius .
an important clue to the true nature of both these `` specific frequency problems''which are most likely two aspects of a single phenomenon lies in the observation that the global @xmath3 in central cluster galaxies increases with the x - ray luminosity of their parent clusters .
some of the globulars in these systems ought then to be associated with the hot gas there ; and @xmath474 is approximated better by the present - day @xmath475 than by @xmath38 alone .
@xmath3 then systematically overestimates the true cluster formation efficiency @xmath4 , which instead is given empirically by @xmath476 .
in addition , since a galaxy s hot gas generally follows a shallower density profile than its stellar halo does , the `` extra '' clusters in very gas - rich ( and high-@xmath3 ) ellipticals might well be expected to also show a radial distribution , @xmath178 or @xmath111 , that is distended relative to @xmath239 or @xmath192 .
these ideas , which extend similar suggestions in the recent literature , have been tested in detail for each of m87 , m49 , and ngc 1399 .
data were collected from the literature on the gcss , surface brightness profiles , and x - ray gas densities of these galaxies , covering spatial scales @xmath477100 kpc . in
3.1 , new comparisons of the projected densities @xmath111 and @xmath192 in m49 and ngc 1399 ( which are gas - poor ) showed that , contrary to previous claims , neither galaxy suffers from either of the two specific frequency problems : the global @xmath3 of m49 has a `` standard '' value of @xmath105 , and the local ratio @xmath204 is constant with radius ( outside of a `` core '' region , @xmath478 , that possibly has been dynamically depleted ) if the recent surface - brightness profile of caon et al .
( 1994 ) is used in the comparison rather than the older @xmath187 measurements of king ( 1978 ) .
the global specific frequency of ngc 1399 is similarly `` average '' , @xmath479 , if its gcs population is normalized to an absolute magnitude for the galaxy ( @xmath220 ) that is significantly brighter than the value more commonly adopted .
the ratio @xmath480 is also constant with radius in this galaxy , and has a value equal to that found in m49 .
m87 , at the center of the virgo cluster , is much richer in gas than either of m49 or ngc 1399 , and both @xmath3 problems are in evidence there . a discrete , geometrical deprojection algorithm was developed to enable a direct comparison between the volume densities @xmath178 , @xmath239 , and @xmath255 at all three - dimensional galactocentric radii in m87 .
it was found in 3.2 that @xmath481 , beyond @xmath482 kpc , and that the constant of proportionality is the same as that in m49 and ngc 1399 . _
the cluster formation efficiency was constant with galactocentric radius inside each of these systems , and the same from one galaxy to another_. specifically , @xmath483 ( observed departures from this ratio in the central regions @xmath484 of m87 and m49 may have resulted , at least in part , from the dynamical destruction of globular clusters there . ) many globulars in m87 are indeed associated with its x - ray gas , and this is in fact responsible for the appearance of both specific frequency problems .
this single case leads to the basic expectation which should be tested in as many individual systems as possible that real discrepancies between @xmath111 and @xmath485 should appear only in gas - rich ellipticals that also have high global specific fequencies .
the obvious generalization of these results is that most galaxies might have been subject to a single , common @xmath4 .
this possibility was explored , and borne out , in 4 .
observations of global @xmath3 ( i.e. , total gcs populations and galaxy luminosities ) in 97 giant ellipticals , brightest cluster galaxies , and dwarf ellipticals _ all _ were shown to be consistent with the predictions of a universal @xmath4 at the same level as that observed directly in m87 , m49 , and ngc 1399 .
the different scalings of @xmath3 with @xmath9 among normal giant ellipticals , central cluster galaxies , and dwarfs stem entirely from fundamentally different relations between @xmath9 and @xmath5 in these different physical regimes .
these relations have been obtained from scaling arguments that employ the optical fundamental plane of bright ellipticals ; the observed correlation between x - ray and optical luminosities in early - type systems ; and the fractional gas loss expected in a specific model ( dekel & silk 1986 ) for winds from a single burst of star formation in dark - matter dominated dwarf ellipticals .
observations of @xmath22 vs. @xmath347 still show some _ scatter _ about the mean prediction of constant @xmath4 , and it will be important to clarify the significance and the origins of this . however , there are no _ systematic _ deviations from a universal cluster formation efficiency , in galaxies that span four orders of magnitude in total luminosity and gcs population .
exactly the same @xmath4 was shown , in 4 , to have obtained for globular clusters in the spheroid , or stellar halo , of the milky way ; and it was argued that it appears also to apply now , to the formation of _ open _ clusters locally in the galactic disk .
this remarkable robustness , in which a single cluster formation efficiency was and is realized in a great variety of local and global environments , demands a very general theoretical explanation . from a more empirical standpoint ,
it bears directly on current questions involving gcs and galaxy formation and evolution on large scales .
some of these were discussed in 5 .
of considerable interest is the fact that , with @xmath4 constant but @xmath39 not , the efficiency of _ unclustered _ star formation could _ not _ have been universal ; field star formation must have been suppressed , in a sense , in dwarf ellipticals and at large galactocentric radii in the brightest , most gas - rich ellipticals .
the possibility that these points could be understood in terms of feedback processes seems promising ( 5 ) , and should be explored in quantitative models .
i am grateful to bill harris for providing the data plotted in figures [ fig8 ] and [ fig9 ] ; to ivan king and chris mckee for helpful comments on an earlier draft of the paper ; and to an anonymous referee for several valuable suggestions , including the addition of figures [ fig7 ] and [ fig10 ] .
this work was supported by nasa through grant number hf-1097.01 - 97a awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa under contract nas5 - 26555 .
the surface density data in tables [ tab2 ] , [ tab3 ] , and [ tab4 ] lend themselves well to a geometrical deprojection of the sort that is often applied to the x - ray emission from ellipticals and galaxy clusters ( e.g. , fabian et al . 1981 ; kriss , cioffi , & canizares 1983 ; nulsen & bhringer 1995 ) .
this technique makes only one assumption that of circular symmetry and is fully nonparametric .
it also avoids the explicit differentiation of @xmath121 that is required in the standard abel - transform approach to deprojection .
such a differentiation can be performed if an ad hoc parametrization is adopted for the global form of the density profile from the outset , but it is clearly desireable to minimize any such a priori restrictions .
although sophisticated , nonparametric algorithms that resort finally to an abel integral have also been developed ( e.g. , merritt & tremblay 1994 ) , they present a greater degree of complexity than is necessary here ; and , in any case , they seem poorly suited to density profiles that are constructed from multiple observations with different fields of view and statistical corrections for different levels of background contamination and photometric incompleteness .
the assumption of spherical symmetry made here is not perfect ( the isopleths of the m87 gcs , for one , are clearly non - circular in projection ; see mclaughlin et al .
1994 ) , but neither is it nonsensical : it takes azimuthally averaged surface densities and returns volume densities that are properly interpreted as angular averages at various galactocentric radii .
a simple example of the basic method , which should serve to define its geometrical set - up and mathematical notation , is illustrated in fig .
[ fig11 ] .
the plane of the sky there is perpendicular to the plane of the page , and the two intersect in the line @xmath486 .
thus , the lines on the right , labelled by projected radius @xmath487 , @xmath488 , and @xmath489 , are lines of sight for an observer situated off the bottom of the page .
the bands @xmath490 and @xmath491 , along with their mirror images on the left side of the figure , are axial cross sections of cylinders that extend to @xmath492 ; in the plane of the sky , of course , the corresponding cross sections are circular annuli .
the circular annuli shown here are not those on the plane of the sky , however ; rather , they are equatorial cross sections of spherical shells with three - dimensional radii @xmath493 and @xmath494 .
evidently , these shells have been chosen such that @xmath495 , and so on .
the origin @xmath496 ( through which passes the axis @xmath497 ) represents the center of a spherical distribution of globular clusters , which are supposed to have been counted up in discrete cylinders @xmath498 and @xmath499 on the sky , to form the surface densities @xmath500 and @xmath501 . from such data ,
it is possible to compute the fraction of clusters in each cylinder that actually reside in the spherical shells @xmath493 and @xmath502 , and thus to form the average volume densities @xmath503 and @xmath504 .
suppose , for the moment ( this will be relaxed just below ) , that @xmath505 marks the outer edge of the gcs in question , i.e. , @xmath506 and @xmath507 .
( it is assumed that the density @xmath120 of any uniform distribution of background objects has already been subtracted to form @xmath121 from the raw data . ) in this case , all of the objects seen in the outermost cylinder actually come from the outer spherical shell , i.e. , from the hatched regions between @xmath488 and @xmath489 in fig .
[ fig11 ] .
since the number of globulars in this cylinder is just @xmath508 all that is required is a calculation of the volume of the shell @xmath509 which is intersected by the cylinder @xmath491 .
denoting this by @xmath510 , the volume density is obviously @xmath511 moving inwards to the cylinder @xmath498 , the total number of clusters observed there includes a contribution from the spherical shell @xmath512 ( hatched regions between @xmath487 and @xmath488 ) , as well as one from the shell of interest , @xmath493 ( solid region ) .
the volume density in the outer shell has already been determined , and the volume of intersection between that shell and the inner cylinder can be calculated [ call this @xmath513 , so the volume density in the inner shell is given by @xmath514 where @xmath515 and @xmath516 have the same meanings as the corresponding quantities in the first cylinder , @xmath517 . clearly , this procedure can be extended and applied iteratively , from the outside in , to the projected number counts in any series of concentric , non - overlapping , circular annuli ; it is only required that there be no gaps between any two consecutive annuli .
thus , let the annuli be defined on the plane of the sky by projected radii @xmath518 , i.e. , set up @xmath519 rings with @xmath498 ; @xmath491 ; and so on .
given average surface densities @xmath121 in each of these rings , the volume densities @xmath179 in the corresponding spherical shells @xmath520 can be obtained once the volume of intersection between some generic shell @xmath521 and a cylinder @xmath522 is known .
a sketch similar to fig .
[ fig11 ] readily shows that this volume is , quite generally , @xmath523 , \label{eq:33}\ ] ] for @xmath524 .
this holds for any set of radii @xmath525 , with the understanding that any term in parentheses which evaluates as negative is to be set to 0 .
an important application is to the case shown in fig .
[ fig11 ] , where , e.g. , @xmath526 and @xmath505 and the volume common to the shell @xmath527 and the cylinder @xmath491 is @xmath528 .
it is most appropriate to define shells such that @xmath529 for all @xmath530 , in which case the results given just above for @xmath531 generalize to @xmath532 } \over{v_{\rm int}(r_{i-1},r_i;r_{i-1},r_i)}}\nonumber \\ & = & { 3\over{4(r_i^2-r_{i-1}^2)^{1/2}}}\left [ n_{\rm cl}(r_{i-1},r_i)-\sum_{j = i+1}^{m } { { n_{\rm cl}(r_{j-1},r_j)\,v_{\rm int } ( r_{j-1},r_j;r_{i-1},r_i)}\over{\pi(r_i^2-r_{i-1}^2)}}\right ] .
\nonumber \\ & & \label{eq:34}\end{aligned}\ ] ] again , this assumes that any uniform background has already been subtracted to obtain the surface densities @xmath121 in every annulus @xmath533 .
this equation is not quite final , however , since its derivation has been facilitated by treating the outermost annulus seen in projection , @xmath534 ( or the outermost spherical shell , @xmath535 ) , as if it ended just at the `` true edge '' of some gcs . of course
, this will never actually be the case if not because it makes little sense to associate sharp boundaries with such distributions in the first place , then because real observations and number counts can cover small fields of view within gcss that are known to extend considerably further .
this means that there is a second type of `` background '' contamination to be dealt with in any dataset , quite apart from that ( already accounted for ) due to a uniform distribution of unassociated galaxies and stars ; corrections have to be made , in every annulus or shell , for the possible presence of bona fide globular clusters at radii @xmath536 , beyond the last annulus for which a surface density can be directly measured .
if the density distribution of such objects is denoted by @xmath537 , their contribution to the projected number density in a cylinder @xmath522 is given by @xmath538 which is just an average of the projected densities along every line of sight contained in the cylinder .
[ as usual , the coordinate @xmath539 here measures distance along these lines of sight ; @xmath486 defines the plane of the sky , as in fig .
[ fig11 ] .
] it is most convenient to express this `` background '' surface density as a fraction of the observed @xmath121 in the outermost annulus @xmath540 : @xmath541 , \label{eq:36}\ ] ] where @xmath542 is defined as in equation ( [ eq:35 ] ) , and the second equality relies on the basic definition of average surface density in a circular annulus of finite width .
the advantage of this formulation is that the form of @xmath543 can not be directly measured for @xmath536 , so the calculation of @xmath544 necessarily depends on an extrapolation of some sort . without any constraints on this extrapolation ,
it is conceivable that an artificial and unphysical situation such as @xmath545 could arise .
this and similar difficulties are avoided , however , by only ever evaluating @xmath544 in terms of @xmath546 ( which is guaranteed always to be @xmath5471 when @xmath548 ) and the known @xmath121 at the largest projected radii observed .
equation ( [ eq:36 ] ) should therefore be used to subtract @xmath549 from the total @xmath121 in every annulus before deprojecting the distribution according to equation ( [ eq:34 ] ) .
the corrected volume densities in the spherical shells @xmath533 may then be written as @xmath550 typically , when measured gcs surface densities extend to sufficiently large radius , an adequate estimate of @xmath546 can be made by assuming a power - law distribution @xmath551 in the outer reaches of the system , @xmath536 ; and in the case that @xmath552 is an integer , equation ( [ eq:36 ] ) can be evaluated analytically . with @xmath553 , for example , @xmath554 .
\label{eq : a1}\ ] ] for @xmath555 , @xmath556 .
\label{eq : a2}\ ] ] and for @xmath557 , @xmath558\ \ \ \ \times \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\ ] ] @xmath559 .
\label{eq : a3}\ ] ] these three power - law exponents are expected to bracket the conditions that hold in real gcss . finally , the volume densities derived in this way are a series of averages of a continuous @xmath537 in spherical shells of finite thickness .
the question therefore arises as to how a single radius @xmath560 should be assigned to each shell so that the discrete distribution , @xmath561 vs. @xmath560 , represents the continuous one as faithfully as possible .
( see harris 1986 or king 1988 for a discussion of this issue in the two - dimensional context . ) clearly , the answer is to choose each @xmath562 $ ] such that @xmath563 if it is assumed that @xmath551 , the second equality is easily solved for @xmath560 in every shell .
however , @xmath552 will generally not be a constant from shell to shell ; that is , a real gcs need not obey a single power - law density distribution over its entire extent ( as is evident , for example , in fig .
[ fig3 ] above ) .
moreover , @xmath552 can not be known a priori , and any attempts to derive it from the data must rely on fits to the average deprojected densities
fits which themselves depend on the values of @xmath560 .
fortunately , a complicated iterative procedure can be avoided by noting that observed radial distributions suggest that @xmath564 anywhere in a typical gcs .
this in turn suggests , and numerical experiments confirm , that simply using @xmath565 everywhere should lead to @xmath560 estimates that are acceptable substitutes for the exact solutions of equation ( [ eq:38 ] ) in any realistic situation .
thus , the formula adopted here is @xmath566^{2/3}\ . \label{eq:39}\ ] ] aguilar , l. , hut , p. , & ostriker , j. p. 1988
, apj , 335 , 720 ashman , k. m. , & zepf , s. e. 1992 , apj , 384 , 50 bahcall , j. n. , & soneira , r. m. 1980 , apjs , 44 , 73 bender , r. , surma , p. , dbereiner , s. , mllenhoff , c. , & madejsky , r. 1989 , a&a , 217 , 35 binney , j. , & tremaine , s. 1987 , galactic dynamics ( princeton : princeton univ . press ) blakeslee , j. p. 1997
, apj , 481 , l59 blakeslee , j. p. , tonry , j. l. , & metzger , m. r. 1997 , aj , 114 , 482 bridges , t. j. , hanes , d. a. , & harris , w. e. 1991 , aj , 101 , 469 brighenti , f. , & mathews , w. g. 1998 , apj , 495 , 239 brodie , j. p. , schroder , l. l. , huchra , j. p. , phillips , a. c. , kissler - patig , m. , & forbes , d. a. 1998 , aj , 116 , 691 brown , b. a. , & bregman , j. n. 1998 , apj , 495 , l75 burstein , d. , & heiles , c. 1984 , apjs , 54 , 33 caon , n. , capaccioli , m. , & donofrio , m. 1994 , a&as , 106 , 199 capuzzo - dolcetta , r. 1993 , apj , 415 , 616 carlberg , r. g. , et al .
1997 , apj , 485 , l13 cohen , j. g. 1988 , aj , 95 , 682 ct , p. , marzke , r. o. , & west , m. j. 1998 , apj , 501 , 554 crane , p. , et al .
1993 , aj , 106 , 1371 davies , r. l. , efstathiou , g. , fall , s. m. , illingworth , g. , & schecter , p. l. 1983 , apj , 266 , 41 davis , d. s. , & white , r. e. 1996 , apj , 470 , l35 de carvalho , r. r. , & djorgovski , s. 1989 , apj , 341 , l37 de carvalho , r. r. , & djorgovski , s. 1992 , apj , 389 , l49 de vaucouleurs , g. 1993 , apj , 415 , 40 de vaucouleurs , g. , & nieto , j .- l . 1978 , apj , 220 , 449 de vaucouleurs , g. , & nieto , j .- l .
1979 , apj , 230 , 697 de vaucouleurs , g. , & pence , w. d. 1978 , aj , 83 , 1163 de vaucouleurs , g. , de vaucouleurs , a. , corwin , jr . , h. g. , buta , r. j. , paturel , g. , & fouqu , p. 1991
, third reference catalogue of bright galaxies ( new york : springer ) dehnen , w. 1993 , mnras , 265 , 250 dekel , a .. , & silk , j. 1986 , apj , 303 , 39 djorgovski , s. , & meylan , g. 1994 , aj , 108 , 1292 durrell , p. r. , harris , w. e. , geisler , d. , & pudritz , r. e. 1996 , aj , 112 , 972 elmegreen , b. g. 1983 , mnras , 203 , 1011 elmegreen , b. g. , & clemens , c. 1985 , apj , 294 , 523 elmegreen , b. g. , & efremov , y. n. 1997 , apj , 480 , 235 fabbiano , g. 1989 , ara&a , 27 , 87 faber , s. m. , & gallagher , j. 1979 , ara&a , 17 , 135 faber , s. m. , dressler , a. , davies , r. l. , burstein , d. , lynden - bell , d. , terlevich , r. , & wegner , g. 1987 , in nearly normal galaxies , ed . s. m. faber ( new york : springer ) , 175 faber , s. m. , et al .
1997 , aj , 114 , 1771 fabian , a. c. , hu , e. m. , cowie , l. l. , & grindlay , j. 1981 , apj , 248 , 47 forbes , d. a. , brodie , j. p. , & grillmair , c. j. 1997 , aj , 113 , 1652 forbes , d. a. , franx , m. , illingworth , g. d. , & carollo , c. m. 1996 , apj , 467 , 126 forman , w. , jones , c. , & tucker , w. 1985 , apj , 293 , 102 freedman , w. , et al .
1994 , nature , 371 , 757 fritze - von alvensleben , u. 1999 , a&a , in press ( astro - ph/9812285 ) gebhardt , k. , et al .
1996 , aj , 112 , 105 geisler , d. , lee , m. g. , & kim , e. 1996 , aj , 111 , 1529 gnedin , o. y. , & ostriker , j. p. 1997 ,
apj , 474 , 223 grillmair , c. , pritchet , c. , & van den bergh , s. 1986 , aj , 91 , 1328 hanes , d. a. , & harris , w. e. 1986 , apj , 309 , 564 harris , w. e. 1976 , aj , 81 , 1095 harris , w. e. 1981 , apj , 251 , 497 harris , w. e. 1986 , aj , 91 , 822 harris , w. e. 1991 , ara&a , 29 , 543 harris , w. e. 1996 , aj , 112 , 1487 harris , w. e. , & petrie , p. l. 1978 , apj , 223 , 88 harris , w. e. , & pudritz , r. e. 1994 , apj , 429 , 177 harris , w. e. , & racine , r. 1979 , ara&a , 17 , 241 harris , w. e. , & smith , m. g. 1976 , apj , 207 , 1036 harris , w. e. , & van den bergh , s. 1981 , aj , 86 , 1627 harris , w. e. , harris , g. l. h. , & mclaughlin , d. e. 1998 , aj , 115 , 1801 harris , w. e. , pritchet , c. j. , & mcclure , r. d. 1995 , apj , 441 , 120 harris , w. e. , allwright , j. w. b. , pritchet , c. j. , & van den bergh , s. 1991 , apjs , 76 , 115 hernquist , l. 1990 , apj , 356 , 359 hills , j. g. 1980 , apj , 225 , 986 ho , l. c. , & filippenko , a. v. 1996a , apj , 466 , l83 ho , l. c. , & filippenko , a. v. 1996b , apj , 472 , 600 huchra , j. p. , & brodie , j. p. 1987 , aj , 93 , 779 irwin , j. a. , & sarazin , c. 1996 , apj , 471 , 683 jacoby , g. h. , branch , d. , ciardullo , r. , davies , r. l. , harris , w. e. , pierce , m. j. , pritchet , c. j. , tonry , j. l. , & welch , d. l. 1992 , pasp , 104 , 599 jones , c. , stern , c. , forman , w. , breen , j. , david , l. , tucker , w. , & franx , m. 1997 , apj , 482 , 143 kavelaars , j. j. , & hanes , d. a. 1997 , mnras , 285 , l31 kavelaars , j. j. 1998 , preprint ( astro - ph/9806094 ) killeen , n. e. b. , & bicknell , g. v. 1988 , apj , 325 , 165 king , i. r. 1978 , apj , 222 , 1 king , i. r. 1988 , pasp , 100 , 999 kinman , t. d. , suntzeff , n. b. , & kraft , r. p. 1994
, aj , 108 , 1722 kissler - patig , m. 1997 , a&a , 319 , 83 kissler - patig , m. , forbes , d. a. , & minniti , d. 1998 , mnras , 298 , 1123 kissler - patig , m. , kohle , s. , hilker , m. , richtler , t. , infante , l. , & quintana , h. 1997 , a&a , 319 , 470 kohle , s. , kissler - patig , m. , hilker , m. , richtler , t. , infante , l. , & quintana , h. 1996 , a&a , 309 , l39 kriss , g. a. , cioffi , d. f. , & canizares , c. r. 1983 , apj , 272 , 439 kumai , y. , basu , b. , & fujimoto , m. 1993a , apj , 404 , 144 kumai , y. , hashi , y. , & fujimoto , m. 1993b , apj , 416 , 576 lada , c. j. , margulis , m. , & dearborn , d. 1984 , apj , 285 , 141 lada , e. a. , depoy , d. l. , evans , n. j. , & gatley , i. 1991 , apj , 371 , 171 larson , r. b. 1988 , in iau symposium no . 126 , globular cluster systems in galaxies , ed . j. e. grindlay and a. g. d. philip ( dordrecht : kluwer ) , 311 larson , r. b. 1993 , in asp conf .
48 , the globular cluster - galaxy connection , ed .
g. h. smith and j. p. brodie ( san francisco : asp ) , 675 larson , r. b. 1996 , in asp conf .
92 , formation of the galactic halo ... inside and out , ed .
morrison and a. sarajedini ( san francisco : asp ) , 241 lauer , t. r. , & kormendy , j. 1986 , apj , 303 , l1 lee , m. g. , kim , e. , & geisler , d. 1998 , aj , 115 , 947 mandushev , g. , spassova , n. , & staneva , a. 1991 , a&a , 252 , 94 mathieu , r. d. 1983 , apj , 267 , l97 mateo , m. l. 1998 , ara&a , 36 , 435 mckee , c. f. 1989 , apj , 345 , 782 mclaughlin , d. e. 1994 , pasp , 106 , 47 mclaughlin , d. e. 1995 , aj , 109 , 2034 mclaughlin , d. e. 1999 , apj letters , 512 , in press ( astro - ph/9812242 ) mclaughlin , d. e. , & pudritz , r. e. 1996 , apj , 457 , 578 mclaughlin , d. e. , harris , w. e. , & hanes , d. a. 1993 , apj , 409 , l45 mclaughlin , d. e. , harris , w. e. , & hanes , d. a. 1994 , apj , 422 , 486 mclaughlin , d. e. , secker , j. , harris , w. e. , & geisler , d. 1995 , aj , 109 , 1033 mcmillan , r. , ciardullo , r. , & jacoby , g. 1993 , apj , 416 , 62 merritt , d. , & tremblay , b. 1994 , aj , 108 , 514 meurer , g. r. , heckman , t. m. , leitherer , c. , kinney , a. , robert , c. , & garnett , d. r. 1995 , aj , 110 , 2665 miller , b. w. , lotz , j. m. , ferguson , h. c. , stiavelli , m. , & whitmore , b. c. 1998 , apj , 508 , l133 minniti , d. 1995 , aj , 109 , 1663 minniti , d. , meylan , g. , & kissler - patig , m. 1996 , a&a , 312 , 49 murali , c. , & weinberg , m. d. 1997a , mnras , 288 , 749 murali , c. , & weinberg , m. d. 1997b , mnras , 288 , 767 murali , c. , & weinberg , m. d. 1997c , mnras , 291 , 717 muzzio , j. c. 1987 , pasp , 99 , 245 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1996 , apj , 462 , 563 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1997 , apj , 490 , 493 nieto , j .- l . , bender , r. , & surma , p. 1991
, a&a , 244 , l37 nulsen , p. e. j. , & bhringer , h. 1995 , mnras , 274 , 1093 oemler , a. 1976 , apj , 209 , 693 okazaki , t. , & tosa , m. 1995 , mnras , 274 , 48 ostriker , j. p. , binney , j. , & saha , p. 1989 , mnras , 241 , 849 ostrov , p. g. , forte , j. c. , & geisler , d. 1998 , aj , 116 , 2854 pahre , m. a. , djorgovski , s. g. , & de carvalho , r. r. 1995 , apj , 453 , l17 patel , k. , & pudritz , r. e. 1994 , apj , 424 , 688 pierce , m. j. , welch , d. l. , mcclure , r. d. , van den bergh , s. , racine , r. , & stetson , p. b. 1994 , nature , 371 , 385 portegies zwart , s. f. , hut , p. , makino , j. , & mcmillan , s. l. w. 1998 , a&a , 337 , 363 preston , g. w. , schectman , s. a. , & beers , t. c. 1991 , apj , 375 , 121 pritchet , c. j. , & van den bergh , s. 1994 , aj , 107 , 1730 pryor , c. , & meylan , g. 1993 , in asp conf .
50 , structure and dynamics of globular clusters , ed .
s. g. djorgovski and g. meylan ( san francisco : asp ) , 357 roberts , m. s. 1957 , pasp , 69 , 59 santiago , b. x. , & djorgovski , s. 1993 , mnras , 261 , 753 schweizer , f. 1987 , in nearly normal galaxies , ed . s. m. faber ( new york : springer ) , 18 searle , l. , & zinn , r. 1978 , apj , 225 , 357 sternberg , a. 1998 , apj , 506 , 721 surdin , v. g. 1979 , sov .
astron . , 23 , 648 tonry , j. l. , blakeslee , j. p. , ajhar , e. a. , & dressler , a. 1997 , apj , 475 , 399 tremaine , s. , richstone , d. o. , byun , y .- i . ,
dressler , a. , faber , s. m. , grillmair , c. , kormendy , j. , & lauer , t. r. 1994 , aj , 107 , 634 tremblay , b. , & merritt , d. 1996 , aj , 111 , 2243 trinchieri , g. , fabbiano , g. , & canizares , c. r. 1986 , apj .
310 , 637 tsai , j. c. 1993 , apj , 413 , l59 van den bergh , s. 1984 , pasp , 96 , 329 van den bergh , s. 1998 , apj , 492 , 41 van den bergh , s. , & lafontaine , a. 1984 , aj , 89 , 1822 van der marel , r. p. 1991 , mnras , 253 , 710 verschueren , w. 1990 , a&a , 234 , 156 vesperini , e. 1997 , mnras , 287 , 915 vesperini , e. , & heggie , d. c. 1997 , mnras , 289 , 898 wagner , s. , richtler , t. , & hopp , u. 1991 , a&a , 241 , 399 weil , m. l. , bland - hawthorn , j. , & malin , d. f. 1997 , apj , 490 , 664 west , m. j. 1993 , mnras , 265 , 755 west , m. j. , ct , p. , jones , c. , forman , w. , & marzke , r. o. 1995 , apj , 453 , l77 white , r. e. 1987 , mnras , 227 , 185 whitmore , b. c. 1997 , in the extragalactic distance scale , ed .
m livio , m. donahue , and n. panagia ( baltimore : stsci ) , 254 whitmore , b. c. , & schweizer , f. 1995 , aj , 109 , 960 whitmore , b. c. , sparks , w. b. , lucas , r. a. , macchetto , f. d. , & biretta , j. a. 1995 , apj , 454 , l73 young , p. j. 1976 , aj , 81 , 807 zepf , s. e. , & ashman , k. m. 1993 , mnras , 264 , 611 zinn , r. 1985 , apj , 293 , 424 zinnecker , h. , mccaughrean , m. j. , & wilking , b. a. 1993 , in protostars and planets iii , ed . e. h. levy and j. i. lunine ( tucson : univ . of arizona press ) , 429 | the specific frequencies of globular cluster systems , @xmath0 , are discussed in terms of their connection to the efficiency of globular cluster formation in galaxy halos , which is claimed to reflect a generic aspect of the star formation process as it operates even at the current epoch .
it is demonstrated that the total _
masses _ of gcss are little affected by the dynamical destruction of low - mass clusters at small galactocentric radii .
this permits direct , empirical estimates of the cluster formation efficiency by mass , @xmath1 , even after @xmath2 years of gcs evolution .
however , the standard practice of using only the stellar luminosities of galaxies as indicators of their initial total gas masses ( and thus relating @xmath3 to @xmath4 in one step ) leads to serious conceptual problems , that are reviewed here . the _ first specific frequency problem _ , which is the well known tendency for many brightest cluster galaxies to have higher than average @xmath3 , is a global one ; the _ second specific frequency problem _ is a local one , in which the more extended spatial distribution of gcss relative to halo stars in some ( not all ) bright ellipticals leads to @xmath3 values that increase with radius inside the galaxies . extending similar suggestions in the recent literature
, it is argued that these trends in @xmath3 do not reflect any such behavior in the underlying @xmath4 ; rather , _ both _ of these problems stem from neglecting the hot , x - ray emitting gas in and around many large ellipticals , and both may be alleviated by including this component in estimates of @xmath5 .
this claim is checked and confirmed in each of m87 , m49 , and ngc 1399 , all of which have been thought to suffer from one or the other of these @xmath3 problems .
existing data are combined to construct gcs surface density profiles that extend over nearly the whole extents of these three galaxies , and a non - parametric , geometrical deprojection algorithm is developed to afford a direct comparison between the _ volume _ density profiles of their gcss , stars , and gas .
it is found , in each case , that @xmath6 at radii beyond roughly a stellar effective radius , inside of which dynamical evolution may have depeleted the initial gcss .
the constant of proportionality is the same in all three galaxies : @xmath7 . taken together , these results suggest that gcss generally should be more spatially extended than stellar halos only in gas - rich galaxies that also have a high global specific frequency . the implication that @xmath4 might have had a universal value is supported by global gcs data for a sample of 97 giant ellipticals , brightest cluster galaxies , and faint dwarfs .
the total globular cluster populations in all of these early - type systems are in excellent agreement with the predictions of a constant @xmath4 at the level observed directly in m87 , m49 , and ngc 1399 ; all _ systematic _ variations in gcs specific frequency between galaxies are shown to result entirely from different relations , in different magnitude ranges , between @xmath8 and the present - day @xmath9 .
an identical @xmath4 is also calculated for the pop .
ii spheroid of the milky way , and is indicated ( although less conclusively ) for the ongoing formation of open clusters . the inferred universal cluster formation efficiency , of @xmath100.25% by mass , should serve as a strong constraint on general theories of star and cluster formation .
the associated inference of a non - universal formation efficiency for unclustered stars is considered , particularly in terms of the suggestion that this might result , both in dwarf galaxies and at large galactocentric radii in the brightest ellipticals , from feedback and galactic winds .
implications for a merger - formation model of early - type gcss , and for the proposed existence of intergalactic globulars in clusters of galaxies , are briefly discussed . |
low - dimensional interacting spins always reveal very interesting magnetic properties , as well as new electronic transport phenomena @xcite . along this line , systems with geometric frustration
have long attracted our attention because the ground - state properties , like degeneracy , are usually responsible for peculiar behaviors at low temperatures .
included are exotic magnetization dependences with external fields in hysteresis curves , for example .
one group of compounds that exhibits such properties is the family cscox@xmath0 , where x stands for cl or br @xcite .
these materials are unidimensional ising - like magnetic systems .
the spin chains are arranged on a triangular lattice , while the intra- and interchain exchange couplings are both antiferromagnetic .
another fascinating spin - chain family of compounds with triangular arrangements has the formula a@xmath5abo@xmath2 ( where a may be ca or sr , whereas a and b are transition metals ) .
one particular system that caught our attention is ca@xmath0co@xmath1o@xmath2 @xcite , which has a rombohedral structure composed of large chains of [ co@xmath1o@xmath2]@xmath4 along the @xmath6-axis of a corresponding hexagonal lattice .
the ca atoms are located among those chains .
the chains are made of alternating face - sharing coo@xmath2 trigonal prisms and coo@xmath2 octahedra .
each co chain is thus surrounded by six equally spaced co chains forming an hexagonal lattice in the @xmath7-plane ( see figure [ hexagono ] ) .
generally , the long co - co interchain distance ( 5.3 ) , in ca@xmath0co@xmath1o@xmath2 , compared to the short co - co intrachain distance ( 2.6 ) , ensures the hierarchy of the magnetic exchange energies , @xmath8 , in modulus .
this fact , together with a strong spin - orbit coupling @xcite , provide a high uniaxial anisotropy along the @xmath6-axis .
neutron diffraction studies @xcite , among others , have revealed a ferromagnetic intrachain interaction ( @xmath9 ) , while an antiferromagnetic interchain interaction ( @xmath10 ) was observed in the @xmath7-plane .
consequently , it is very plausible to describe the magnetism of this compound by considering an ising triangular lattice of large magnetic frozen - moments linked to the chains , each one playing the role of a giant spin , pointing either up or down .
-plane ( blue points ) of the spin - chain compound ca@xmath0co@xmath1o@xmath2 .
each chain is ordered ferromagnetically and is coupled antiferromagnetically to its neighbors . ]
one of the interesting features observed in ca@xmath0co@xmath1o@xmath2 is the appearance of plateaus in the magnetization curve @xmath11 vs @xmath12 , below @xmath13 k , when the magnetic field is applied along the chains ( @xmath6-axis ) . for temperatures in the range , @xmath14 ,
it is observed a rapidly increase of @xmath11 with @xmath12 .
close to zero field , the magnetization reaches a plateau at the value of @xmath15 ( @xmath16 , the saturation magnetization is @xmath17/f.u . ) .
this value remains constant up to @xmath18 t , where the magnetization springs up to its saturation value @xmath16 .
otherwise , when @xmath19 k , called the ferrimagnetic region , the observed plateau at @xmath15 , splits into three small steps , changing at fields @xmath20 t , @xmath21 t , @xmath22 t , respectively .
much effort has been done to understand such behavior , and it is still a controversial matter .
some papers @xcite roused the question that this behavior might be related to experiments involving single - molecule magnets ( smm ) @xcite , where a quantum tunneling of the magnetization ( qtm ) is realized @xcite .
we restrain such statement , at least for temperatures above 4k , as we demonstrate here and leave the question of qtm effects to be searched below 2k , according to the experimental situation .
we follow instead the line of reasoning that metastable states , or peculiar mozaic configurations , might play a role in the splitting of the magnetization of this frustrated material , as first suggested by maignan _
therefore , we started a simulation along the lines given by kudasov @xcite , who predicted an opening of the 1/3 plateau using monte carlo analysis at @xmath23 .
more recently , yao _
@xcite , studied this compound with numerical monte carlo simulations on finite clusters .
the latter works describe the system with an ising hamiltonian on a triangular lattice , using periodic boundary conditions . despite their results
reproduce the steps width , @xmath24 t , in the magnetization curves , they do not find a clear - cut ( _ equilibrium _ ) configuration that explains the splitting of the 1/3 plateau . in this work we report on a similar study using the same model , but
considering instead free boundary conditions on an anisotropic system , which is added for completion with the intrinsic dipolar term at the end .
we apply non - equilibrium techniques on this system .
we detect in this form , by sweeping the field at different rates , the presence of metastable domains - walls that perfectly explains the observed trends of the splitting of the 1/3 plateau at low temperatures , and verify that the latter has long - range order at the equilibrium . to be consistent with the previous existing models , we use the same set of parameters for the hamiltonian @xcite , namely , an interchain antiferromagnetic interaction , @xmath25 @xmath26ev , and a large ising magnetic moment , @xmath27 . disorder or dispersive
effects in the exchange coupling constant were not required to obtain the splitting of the 1/3 plateau , as it will become clear in the following . therefore , our model and methods are different from yao _
@xcite , and we concentrate our efforts mainly to describe the effects of the sweeping rate on the observed metastable states , that generate the sub - steps in the magnetization .
the spin - chain compound ca@xmath0co@xmath1o@xmath2 is formed by close - packed one - dimensional chains along the @xmath6-axis , the chains span a triangular lattice over the @xmath7-plane . due to such geometric configuration and the hierarchy of the exchange interactions ,
the system can be modeled as a two - dimensional ( 2d ) antiferromagnetic ising model with nearest - neighbor exchange interactions @xcite .
the magnetic moments are representative of the [ co@xmath1o@xmath2]@xmath4 chains , which are always perpendicular to the @xmath7-plane .
the choice of this frozen - moment model is a consequence of a time - scale assumption .
we believe that the characteristic time involved in reversing the individual spins of each chain is much shorter than the rearrangement of the triangular lattice by changing the magnetic field .
this hypothesis is based on previous numerical results @xcite on nano - wires , as well as on the large number of sites that are involved in changing the whole triangular lattice as compared to those on a single spin - chain .
we thus performed monte carlo simulations using a triangular ising model , where the magnetic moments are frozen , pointing either up or down , and coupled antiferromagnetically ( @xmath28 ) to their nearest neighbors . in this picture , the hamiltonian @xmath29 of the system is @xmath30 where @xmath31 is the spin of each magnetic moment , @xmath32 is the external magnetic field along the @xmath33-axis , @xmath34 is the electronic land factor , @xmath35 the bohr magneton and @xmath36 indicates summation over all pairs of nearest - neighbors on a triangular lattice .
we used a single - flip monte carlo method together with the standard metropolis algorithm to reject / accept flips in the magnetization with a boltzmann factor @xcite .
magnetization measurement at any given field was done using 1,000 averages over a fixed number of _ accepted _ monte - carlo ( mc ) flips .
we used different 2d finite cluster configurations , including super - hexagons , octagons , dodecagons and other more circularly - shaped geometries to represent the samples .
the magnetization curves found in all these finite clusters were almost the same , the main difference was the number of mc flips needed to reach the plateaus for different sizes . due to space limitations and visualization aspects , a selection of the results for one cluster size with hexagonal geometry was chosen .
extra material showing the evolution of magnetic configurations with the field can be found at the website by the interested reader [ supplementary material ] .
k , and @xmath37 sites .
the splitting of the 1/3 plateau into three successive plateaus is observed below 10 k. they extend up to fields @xmath38 t , 2.4 t and 3.6 t , respectively .
we see that they merge into a unique plateau at @xmath39 as @xmath40 increases ( see arrows ) . ]
in this section we discuss our magnetization results by using monte carlo calculations on finite clusters with open boundary conditions , always increasing the field , starting at @xmath41 t. the initial magnetic state was obtained by thermalizing a random disordered configuration , performing @xmath42 mc flips in a _ training _ period .
afterwards we initiated the increase of the field at a constant sweep rate .
we calculated clusters with hexagonal geometries for @xmath43 , @xmath44 , @xmath45 , @xmath46 and @xmath47 sites .
some other larger clusters were partially calculated , but we verified that the agreement with the experimental results did not improve with size . as the system size increased
the plateaus structure remained unaltered , only small perturbations ( sub - steps ) on each of the three main plateaus tend to disappear . in figure [ tvar ]
we readily observe the formation of three well defined plateaus below @xmath48 k. important to see is that they merge into one , at @xmath15 as @xmath40 increases , indicating a close resemblance with the experimental situation ( see e.g. fig . 3 in @xcite ) .
the presence of extra tiny sub - steps is an unavoidable boundary effect due to the finiteness of the cluster used ( in this case , @xmath37 sites ) .
they are created by border isolated atoms whose contribution is strongly suppressed for larger cluster sizes , therefore , irrelevant for our argument . another important feature to note in the results of figure
[ tvar ] , shared by experiments , is the steeper rise of the magnetization for lower temperatures at the end of the 1/3 plateau ( @xmath49 t ) .
k , and @xmath37 sites .
this is accomplished by letting more or less accepted mc flips ( mcf ) per magnetic field .
notice the tendency of merging into the plateau @xmath15 as the number mcf increases ( see arrows ) , indicating a time decay towards the 1/3 state . this behavior will be further analyzed .
] now , as a way to check how stable are such plateaus , we have performed mc analyses with an increasing number of steps per magnetic field .
in fact , as the mc simulation is done with the accepted configurations , what really counts is mc flips ( mcf ) and not mc steps ( mcs ) .
we clearly see , in figure [ mcfvar ] , the tendency to merge the three plateaus into one as a function of the number of mcf .
this certainly must be related to the experimental sweeping rate @xcite .
this connexion allows us to say that the three plateaus are metastable states and are , therefore , a dynamical effect ; for this very reason the experimental magnetization curves are strongly dependent on the sweeping rate @xcite .
this merging effect can be further explained in the graph of figure [ t4k_5k ] , where we have produced an almost perfect superposition between magnetization curves for different temperatures and , at the same time , different mcf .
in such a case , we tried to draw a parallel between the mcf in our simulation and the sweeping rate in the experimental results , where this behavior is observed .
it is obvious from this result the strong interplay between the effects of time and temperature ( see e.g. fig . 4 in @xcite ) , a feature that was previously used as a fingerprint of qtm .
although mcf are _ not _ related to lab time measurement , we can attribute them to some sweeping rate by comparing the formation of the plateaus .
hence , the three sub - plateaus are made of metastable states , which develop in time to the 1/3 plateau , as in figure [ mcfvar ] .
t , fact which is also seen in the experiments @xcite . ]
dipolar interactions were also considered , as an intrinsic effect always present in nanowires and other low - dimensional systems , but its overall effect here was merely to induce an effective antiferromagnetic field that retarded the field at which the steps occur .
the dipolar term @xmath50 that was added to equation ( [ ising hamiltonian ] ) is given by : @xmath51 where the summation is over all sites , and @xmath52 .
we see such effect in figure [ dipolar ] , with the dipolar interaction playing the role of an effective antiferromagnetic field , @xmath53 added to equation ( [ ising hamiltonian ] ) , that moves the magnetization curves to the right . we noticed that adjusting the exchange constant to @xmath54 , we can almost reproduce the results obtained from our previous 2d ising model , without dipolar interaction ( see figure [ dipolar ] ) . for this reason ,
we think , the magnetization curves are well described using only the exchange and zeeman terms .
k , and @xmath37 sites .
the red curve is for the original @xmath55 value used in figure [ mcfvar ] , without dipolar interaction .
the blue curve is for the same @xmath55 , but now with dipolar interaction , and the green curve is for a reduced exchange , 90% of the original value , and the same dipolar interaction of the blue curve .
the scaling of the red and the green curves is not perfect , but very close . ]
the physics behind these results can be visualized by using raw snapshots of selected configurations . in the next pictures we show different snapshots of the magnetization curves depicted with the red line of figure [ mcfvar ] .
they were taken at three different magnetic fields : @xmath56 t , @xmath57 t , and @xmath58 t , representative of each of the three sub - steps , into which the 1/3 plateau is splitted at low temperatures ( @xmath59 k ) .
our color convention to analyze the snapshots of the subsequent examples is depicted in figure [ apices ] .
following @xcite , we define four triangles and color them differently depending on the magnetization that results by summing the three spins at their vertices : see figure [ apices ] .
is seldom seen for positive magnetic fields in the following snapshots . ]
configurations at zero field are formed by disordered bicontinuous patterns of small compact regions of violet and green phases , with total zero magnetization . for a field close to @xmath60
t , domain walls , as seen in figure [ snap0.81t_4k ] , stabilize their motion to quasi - static domain walls until the field reach values of about 1.2 t. t , at @xmath61 k and @xmath62 mcf , as depicted with a red line in figure [ mcfvar ] .
notice the ordered substrate green zone , with @xmath63 , superimposed by interlinked domain walls , formed mostly by connected triangles in violet color .
these domain walls contribute to a reduction of the @xmath15 value of the green zone to @xmath64 , in this example . ] moving the field across the boundary , @xmath20 t , produces a rapid rearrangement of the linked domain walls positions , by changing simultaneously their internal configuration .
they are now formed by mixed violet and red triangles , as shown in figure [ snap_2.0t_4k ] .
they rate together to zero extra magnetic moment , sometimes oscilating above and below , depending slightly on the simulation parameters . the magnetic configuration in figure [ snap_2.0t_4k ] averages to zero , leading to @xmath63 , the value of the green ordered zone ( formed by the hexagonal unit cells ) . , except at a higher field @xmath57 t , corresponding to a magnetization value of @xmath65 in this case .
observe now the green extended zone superimposed by linked domain walls of mixed violet and red colored triangles , in a proportion that contribute almost zero extra moment .
these linked domain walls , as those before , percolate throughout the system and are rather stable across a finite window of the magnetic field : @xmath66 t @xmath67 t. ] finally , for fields beyond the other boundary , @xmath68 t , the situation is quite different .
now the linked domain walls are formed mostly by red colored triangles ( figure 9 ) with a tendency to coalesce in small groups , or seeds , at the borders of the simulated clusters .
this interesting feature of red seeds at the boundaries , as seen in figure [ snap_2.98t_4k ] , is a peculiar phenomena very similar to pre - nucleation phases , as found in irreversible statistical mechanics studies .
proper nucleation really starts beyond the critical field , @xmath69 t , in this way : first coalescence of small red seeds at the borders and then myriads of red nuclei forming and growing , taking account of the whole cluster , in the bulk , until saturation .
the nucleation process above @xmath70 is a very rapid , disordered one , destroying the linked domain walls structures seen in the former examples .
perfect alignment is only achieved at the end , for fields higher than 4 t. t. other parameters as in figures [ snap0.81t_4k ] and [ snap_2.0t_4k ] .
the magnetic configuration is now formed by the green extended region of @xmath63 superimposed by domain walls in red color , giving a total magnetization of @xmath71 , in this case .
notice also some red seeds distributed at the borders of the cluster , like in pre - nucleation phases of irreversible studies .
nucleation really starts beyond the critical field , @xmath69 t. ] all these non - equilibrium phenomena just described were obtained for a particular sweep rate , of 10@xmath72 mcf .
we tested various sweep rates , observing similar behavior for many of them .
but , none of these metastable interlinked domain walls were present , however , when we go beyond a sweep rate of 10@xmath73 mcf . in that case
, the disordered initial phase goes immediately to the 1/3 plateau , and beyond @xmath69 t , to saturation .
such 1/3 plateau is represented solely by the green zone , without any further domain wall , thus of long - range nature .
therefore , we believe our characterization of such metastable states , sweep - rate dependent , must be related to the multi - step phases of ca@xmath0co@xmath1o@xmath2 , found experimentally at low @xmath40 .
in this section , we show how the 1/3 plateau of the ising model of previous sections shares interesting similarities with the simplest small clusters of triangular geometry .
a generic spin hamiltonian for these clusters is the following heisenberg hamiltonian : @xmath74 ) , and for different temperatures , @xmath75 , as shown in the legend .
we see a tendency to form a 1/3 plateau .
other parameters used are : @xmath76 @xmath26ev , @xmath77 @xmath26ev , and @xmath78 . ]
results for a triangular molecule are shown in figure [ triangulo ] .
@xmath11 ( the thermal average of the total spin z - component ) is zero only at zero field , and it increases rapidly to the 1/3 plateau for any finite @xmath12 , crossing through @xmath79 until @xmath80 , where it rapidly increases again to saturation . such a transition is first - order at @xmath23 only .
increasing the temperature produces a smooth growing up of @xmath11 due to thermal fluctuations .
notice that larger thermal fluctuations accelerate this trend above @xmath81 and retard it above @xmath82 , as seen from figure [ triangulo ] .
, @xmath83 , and @xmath78 .
the green and red lines are for @xmath84 and 39 sites , while the bold blue line is an extrapolation to the thermodynamic limit ( see text ) .
notice the 1/3 plateau of finite width @xmath85 @xcite . ]
interesting to note is that such a trend in figure [ triangulo ] resembles the mc solution of the ising model in the triangular lattice of our previous sections .
although it is not the same behavior , a tendency to form a 1/3 plateau is without any doubt .
no further sub - plateaus can possibly be observed in a triangular molecule because of lack of phase space .
domain walls in the ising dynamics are minimally formed by triangles , as seen before .
further supporting arguments come from numerical exact solutions for ( larger ) finite clusters . in figure [ exact_plateau ]
we see , for example , numerical results for the magnetization of the pure heisenberg model in the triangular lattice , in the absence of anisotropy , at @xmath23 , for @xmath84 and 39 sites @xcite .
the formation of a clear 1/3 plateau of finite width , @xmath85 , is seen from the graphic .
such results have been taken in the literature @xcite as the evidence of a definite plateau in the thermodynamic limit : the bold line in figure [ exact_plateau ] was drawn by connecting the midpoints of the finite - size steps for @xmath86 sites , except at the boundaries of the 1/3 plateau . a good estimate for the lower and upper fields of the boundaries of the exact 1/3 plateau in the triangular lattice
were obtained comparing spin - wave and exact diagonalization results .
they are given by @xmath87 and @xmath88 , respectively .
see more recent results in @xcite . to better understand our point , we should recall that both models , ising and heisenberg , show a similar behavior with a 1/3 plateau because the magnetic field already breaks the symmetry from @xmath89 to @xmath90 , thus making the strong quantum fluctuations inoperant in the isotropic heisenberg triangular case and , consequently , the @xmath91 state becomes a classical , collinear state , with long - range order .
in summary , we did an extensive investigation of the triangular lattice using monte carlo simulation techniques .
we studied this lattice using a 2d ising model for large magnetic moments coupled antiferromagnetically .
we were motivated by multiple steps observed in the magnetization of the spin - chain ca@xmath0co@xmath1o@xmath2 compound .
we found a very good agreement between our results and the experimental situation @xcite .
we demonstrated that the observed plateaus of the ferrimagnetic phase , below @xmath92 k , consist of metastable states and for this reason are time - dependent . in the experimental results such dependence was observed in the sweeping rate of the magnetic field , and in our case this dependence comes from the number of monte carlo flips ( mcf ) . in fact , a close parallelism is found between these two approaches , as seen from our figures [ tvar ] , [ mcfvar ] , [ t4k_5k ] , and figures 3 , 4 , 5 from ref .
besides such agreement , we also found that in the limit of very large number of mcf , namely , a very slow rate of the field , the three sub - steps converge into one plateau , at @xmath93 , like in the experiments , thus giving a thermal decay of such metastable states into the equilibrium , final 1/3 plateau , which has long - range hexagonal order ( described by the green ordered zone of the snapshots , with an hexagon as unit cell ) .
further analysis using snapshots of some specific configurations revealed the presence of linked ( mobile ) domain - walls that give origin to the splitting of the 1/3 plateau below @xmath19 k , as explained in the text .
such a result disproves the possibility of quantum tunneling as generating the multiple steps , at least for temperatures above 4 k , and for the reported sweep rates . based on our results , we believe that at lower temperatures an arrested configuration may provide a slower dynamics of the domain - walls , that would explain a @xmath40-independent time - decay process below 4 k , by critical slowing down , an aspect that certainly must still be checked down .
our solution has been checked by further considering the intrinsic dipolar interactions , finding no relevant contribution from such couplings .
we therefore discarded them as possible contribution of hysteresis phenomena which are strongly present at the ferrimagnetic region , below 10 k , in the experiments @xcite .
we believe that other ingredients and methods , like 3d simulations , must be considered to deal with such irreversible aspects , but this is beyond our model of the magnetization plateaus .
another point that we wanted to emphasize was linked to exact solutions of the heisenberg model in the triangular lattice for finite clusters .
we did a comparison with our previous numerical ising solution and found supporting arguments with the presence of the 1/3 plateau in the isotropic heisenberg case at @xmath23 .
both models share the 1/3 plateau , although for different reasons . in the quantum - mechanical case
the applied magnetic field strengthens the frustration effects in the triangular lattice up to a point of having sufficient overlap with the classical , collinear , three - fold degenerate , 1/3 plateau @xcite .
the point here is that the presence of the 1/3 plateau in the isotropic quantum limit does not require a strong uniaxial anisotropy .
although we did not a thermal study of the exact solution of the heisenberg model , a feature hard to be done , we think that at low temperatures the 1/3 plateau would survive to thermal fluctuations in order to be observed numerically .
such a comparison , therefore , strongly supports our thermal , classical study of the spin - chain compound , ca@xmath0co@xmath1o@xmath2 .
while this paper was considered for publication we became aware of another work , by kudasov _
@xcite , on the same problem , namely , a study of the dynamics of magnetization in frustrated spin - chain system ca@xmath0co@xmath1o@xmath2 using a 2d model .
although they present similar results for the magnetization steps , they are complementary to ours in some respects .
this work was partially supported by fondecyt grant number 1070224 , conicyt ph.d .
fellowship millennium science initiative under project p06 - 022-f and also supported by cnpq ( brasil ) .
one of us ( gm ) would like to acknowledge the universidad tcnica federico santa mara , in valparaso , chile , for a sabbatical stay .
we also thank dr .
andreas honecker for sharing his data on the triangular lattice .
20 m. oshikawa , m. yamanaka , and i. affleck , phys .
78 * , 1984 ( 1997 ) .
d. c. cabra , a. honecker , and p. pujol , phys .
* 79 * , 5126 ( 1997 ) .
t. vekua , d. c. cabra , a. dobry , c. gazza , and d. poilblanc , phys .
lett . * 96 * , 117205 ( 2006 ) .
i. j. hamad , l. o. manuel , g. martnez , and a. e. trumper , phys .
b * 74 * , 094417 ( 2006 ) .
m. f. collins and o. a. petrenko , can . j. phys .
* 75 * , 605 ( 1997 ) .
h. fjellvg , e. gulbrandsen , s. aasland , a. olsen , and b. c. hauback , j. solid state chem .
* 124 * , 190 ( 1996 ) .
s. aasland , h. fjellvg , and b. hauback , solid state commun .
* 101 * , 187 ( 1997 ) .
h. kageyama , k. yoshimura , k. kosuge , h. mitamura , and t. goto , j. phys .
. jpn . * 66 * , 1607 ( 1997 ) .
a. maignan , c. michel , a. c. masset , c. martin , and b. raveau , eur .
j. b * 15 * , 657 ( 2000 ) .
b. martnez , v. laukhin , m. hernando , j. fontcuberta , m. parras , and j. m. gonzlez - calbet , phys .
b * 64 * , 012417 ( 2001 ) .
b. raquet , m. n. baibich , j. m. broto , h. rakoto , s. lambert , and a. maignan , phys .
b * 65 * , 104442 ( 2002 ) . v. hardy , m. r. lees , a. maignan , s. hbert , d. flahaut , c. martin , and d. m. paul , j. phys . : condens . matter * 15 * , 5737 ( 2003 ) .
r. vidya , p. ravindran , h. fjellvg , a. kjekshus , and o. eriksson , phys .
* 91 * , 186404 ( 2003 ) . v. hardy , s. lambert , m. r. lees , and d. mck .
paul , phys .
b * 68 * , 014424 ( 2003 ) .
whangbo , d. dai , h .- j .
koo , and s. jobic , solid state commun .
* 125 * , 413 ( 2003 ) .
a. maignan , v. hardy , s. hbert , m. drillon , m. r. lees , o. petrenko , d. mc k. paul , and d. khomskii , j. mater . chem .
* 14 * , 1231 ( 2004 ) .
r. frsard , c. laschinger , t. kopp , and v. eyert , phys .
b * 69 * , 140405(r ) ( 2004 ) .
v. eyert , c. laschinger , t. kopp , and r. frsard , chem .
phys . lett . *
385 * , 249 ( 2004 ) . v. hardy , m. r. lees , o. a. petrenko , d. mck .
paul , d. flahaut , s. hbert , and a. maignan , phys . rev .
b * 70 * , 064424 ( 2004 ) .
d. flahaut , a. maignan , s. hebert , c. martin , r. retoux , and v. hardy , phys .
b * 70 * , 094418 ( 2004 ) . v. hardy , d. flahaut , m. r. lees , and o. a. petrenko , phys . rev .
b * 70 * , 214439 ( 2004 ) .
a. villesuzanne and m .- h .
whangbo , inorg . chem . * 44 * , 6339 ( 2005 ) .
k. takubo , t. mizokawa , s. hirata , j .- y .
son , a. fujimori , d. topwal , d. d. sarma , s. rayaprol , and e .- v .
sampathkumaran , phys .
b * 71 * , 073406 ( 2005 ) .
k. yamada , z. honda , j. luo , and h. katori , j. alloys compd .
* 423 * , 188 ( 2006 ) .
hua wu , m. w. haverkort , z. hu , d. i. khomskii , and l. h. tjeng , phys .
lett . * 95 * , 186401 ( 2005 ) .
hua wu , z. hu , d. i. khomskii , and l. h. tjeng , phys .
b * 75 * , 245118 ( 2007 ) .
w. wernsdorfer , m. murugesu , and g. christou , phys .
96 * , 057208 ( 2006 ) .
m. r. wegewijs , c. romeike , h. schoeller , and w. hofstetter , new j. phys .
* 9 * , 344 ( 2007 ) . for a review ,
see _ quantum tunneling of magnetization _ , edited by l. gunther and b. barbara ( kluwer , dordrecht , 1995 ) .
yuri b. kudasov , phys .
lett . * 76 * , 027212 ( 2006 ) .
x. y. yao , s. dong , and j .-
m liu , phys .
b * 73 * , 212415 ( 2006 ) .
x. y. yao , s. dong , h. yu , and j .- m liu , phys .
b * 74 * , 134421 ( 2006 ) .
m. vzquez , k. nielsch , p. vargas , j. velzquez , d. navas , k. pirota , m. hernndez - velez , e. vogel , j. cartes , r. b. wehrspohn and u. gosele , physica b * 343 * 395 ( 2004 ) ; m. vzquez , m. hernndez - velez , k. pirota , a. asenjo , d. navas , j. velzquez , p. vargas and c. ramos , eur .
j. b * 40 * , 489 ( 2004 ) .
d. p. landau and k. binder , _ a guide to monte carlo simulations in statistical physics _ ,
edition ( cambridge university press , 2005 ) .
k. y. choi , y. h. matsuda , h. nojiri , u. kortz , f. hussain , a. c. stowe , c. ramsey , and n. s. dalal , phys .
lett . * 96 * , 107202 ( 2006 ) .
s. carretta , p. santini , g. amoretti , f. troiani , and m. affronte , phys .
b * 76 * , 024408 ( 2007 ) . k. y. choi , n. s. dalal , a. p. reyes , p. l. kuhns , y. h. matsuda , h. nojiri , s. s. mal , and u. kortz , phys . rev .
b * 77 * , 024406 ( 2008 ) .
a. honecker , _
private communication_. b. bernu , p. lecheminant , c. lhuillier , and l. pierre , phys .
b * 50 * , 10048 ( 1994 ) .
a. honecker , j. phys .
: condens .
matter * 11 * , 4697 ( 1999 ) .
a. honecker , j. schulenburg , and j. richter , j. phys .
: condens .
matter * 16 * , s749 ( 2004 ) .
u. schollwck , j. richter , d. j. j. farnell , r. f. bishop ( eds . ) , _ quantum magnetism _ ( springer , berlin , 2004 ) . yu .
b. kudasov , a. s. korshunov , v. n. pavlov , and d. a. maslov , phys .
b * 78 * , 132407 ( 2008 ) . | it is known that the spin - chain compound ca@xmath0co@xmath1o@xmath2 exhibits very interesting plateaus in the magnetization as a function of the magnetic field at low temperatures .
the origin of them is still controversial . in this paper
we study the thermal behavior of this compound with a single - flip monte carlo simulation on a triangular lattice and demonstrate the decisive influence of metastable states in the splitting of the ferrimagnetic 1/3 plateau below 10 k. we consider the [ co@xmath1o@xmath2]@xmath4 chains as giant magnetic moments described by large ising spins on planar clusters with open boundary conditions . with this simple frozen - moment model
we obtain stepped magnetization curves which agree quite well with the experimental results for different sweeping rates .
we describe particularly the out - of - equilibrium states that split the low - temperature 1/3 plateau into three steps .
they relax thermally to the 1/3 plateau , which has long - range order at the equilibrium .
such states are further analyzed with snapshots unveiling a domain - wall structure that is responsible for the observed behavior of the 1/3 plateau .
a comparison is also given of the exact results in small triangular clusters with our monte carlo results , providing further support for our thermal description of this compound . |
since the existence of the adiabatic geometric phase that was first revealed in pioneer work of berry @xcite , the concepts of geometric phases have been extensively generalized along many directions @xcite , and now the applications of geometric phases can be found in various physical fields @xcite .
recently , the close relation between geometric phases and quantum phase transitions ( qpts ) has been revealed gradually @xcite and increasing interest has been drawn to the role of geometric phases in detecting qpts for various many - body systems @xcite . essentially , quantum phase transitions happened at zero temperature are characterized by the dramatic changes in the ground - state properties of a many - body system . unlike classical phase transitions driven by thermal fluctuations , qpts are driven by pure quantum fluctuations .
traditionally , qpts are analyzed by resorting to notions such as the order parameter and symmetry breaking within the landau - ginzburg paradigm , however , this scheme can not give a complete description of properties of the ground state in many - body systems .
in the past few years , a lot of efforts have been devoted to the study of qpts from other perspectives , such as quantum order or topological order @xcite , entanglement measures @xcite , and quantum fidelity based on the concept of quantum information sun , zanardiprl , chenshu , gu08 . generally ,
in the vicinity of qpts , the changes in the ground state driven by external parameters of the hamiltonian will lead to an energy - level crossing or avoided energy - level crossing between the ground state and the excited state @xcite , and the features of level structures can be captured by the geometric phase of the ground state because the features of energy - levels crossing or avoided crossing correspond to the divergence or extremum property of the berry curvature . from the geometrical perspective ,
the geometric phase is a reflection of the global curvature in the parameter space of the hamiltonian . in the present work
, we shall use the geometric phase of ground state to detect the qpts for a period - two inhomogeneous anisotropic @xmath0 spin-@xmath2 chain in a transverse field , in which the nearest - neighbor interactions and the degree of anisotropy will take alternating parameters between the neighbor sites @xcite . to the best of our knowledge ,
the previous studies of geometric phase as a witness of qpts mainly concerned homogeneous spin chains for simplicity carollo , zhu , hamma . on the other hand , inhomogeneous systems will exhibit rich phase diagrams and it would be interesting to investigate whether the geometric phase is able to characterize the quantum phase transition in these more complicate systems .
so far , many methods have been introduced to investigate the inhomogeneous spin chains in different limited conditions @xcite ; however an explicit expression of the ground state , which is necessary for the derivation of the geometric phase , is still lacking . in our scheme , by mapping the spin hamiltonian on a fermionic system through the jordan - wigner transformation and the fourier transformation , we derive a general canonical transformation to realize the diagonalization of the fermionic system hamiltonian coupled in the @xmath1- space and construct the exact expression of the ground state .
our results show that there exist more than one critical points at some parameter region and the critical points correspond to the divergence or extremum property of the berry curvature of the hamiltonian parameter space .
the system under consideration is an inhomogeneous periodic anisotropic @xmath0 spin-@xmath3 chain @xcite , which consists of @xmath4 cells with two sites in each cell , and in an external magnetic field .
its hamiltonian is given by @xmath5 , \label{spinham}\end{aligned}\ ] ] where @xmath6 are the local spin operators , @xmath7 is the exchange coupling , @xmath8 is the anisotropy in the in - plane interaction on the @xmath9 site in the @xmath10th cell and @xmath11 is the external field .
we assume periodic boundary conditions and choose @xmath4 to be odd for convenience .
this model encompasses a group of other well - known spin models as its special cases @xcite , such as quantum ising model in a transverse field for @xmath12 and @xmath13 , the transverse field @xmath14 model for @xmath15 and @xmath13 , and the uniform transverse field anisotropic @xmath0 model for @xmath16 and @xmath13 . in order to obtain the geometric phase in this system , we consider rotating this model by applying a rotation of angle @xmath17 about the @xmath18 axis to each spin @xcite , i.e. , @xmath19 $ ] , and then we have @xmath20 , in which @xmath21 is the relevant rotation operator and we have set @xmath22 for simplicity .
it can be verified that @xmath23 , and @xmath24 is @xmath25 periodic in @xmath17 because the quadratic form about the @xmath26 and @xmath27 axes appears symmetric in eq .
( [ spinham ] ) . considering the unitarity of the rotation operator @xmath28 ,
the critical behavior and energy spectrum of the family of hamiltonians parametrized by @xmath17 are obviously @xmath17 independent .
the spin hamiltonian can be mapped exactly on a spinless fermion model through the jordan - wigner transformation @xmath29 \,c_{l , a}^{\dag } , \notag \\ s_{l , b}^{+}\ ! & = & \!\exp \left [ i\pi \!\!\left ( \sum_{l^{\prime } = 1}^{l-1}\!\sum_{m^{\prime } = a , b}c_{l^{\prime } , m^{\prime } } ^{\dag } c_{l^{\prime } , m^{\prime } } + c_{l , a}^{\dag } c_{l , a}\right ) \right ] \,c_{l , b}^{\dag } , \notag \\ & & \end{aligned}\]]where @xmath30 are the spin ladder operators and @xmath31 and @xmath32 are the fermion creation and annihilation operators .
the original hamiltonian @xmath24 is transformed into @xmath33\ .\end{aligned}\]]in the fermion case , the periodic boundary conditions @xmath34,@xmath35 on the spin degrees of freedom imply that @xmath36c_{1,m}$ ] , in which ( @xmath37 ) is just the total fermion number @xmath38 .
thus the boundary conditions on the fermionic system are @xmath39 , and the fermionic system will obey periodic or antiperiodic conditions depending on whether @xmath38 is even or odd @xcite .
however , the differences between the two boundary conditions are negligible in the thermodynamic limit where the second - order qpts occur @xcite . without loss of generality
, we assume the periodic boundary condition on the fermionic system , which means that @xmath38 is always even and @xmath40 .
this periodic boundary condition enables us to introduce a fourier transformation , @xmath41to the hamiltonian @xmath42 , in which @xmath43 and @xmath44 . here
@xmath45 ( @xmath46 ) is defined as the coordinate of site @xmath47 ( @xmath48 ) on the @xmath10th cell in the one - dimensional lattice with the lattice parameter @xmath49 .
hence , the hamiltonian @xmath42 transformed into the momentum space is given by @xmath50-[(\frac{j_{a}\gamma _
{ a}}{2}e^{2i\varphi + ika}-\frac{j_{b}\gamma _ { b}}{2}e^{2i\varphi -ika})a_{-k}b_{k } \notag \\ & + & h.c.]\}\ .\end{aligned}\]]this hamiltonian has a quadratic form in fermion operators and can be exactly diagonalized .
we note that the hamiltonian @xmath51 can be expressed as @xmath52 with matrix @xmath53 and @xmath54 is a @xmath55 hermitian matrix .
therefore , we can always find a unitary transformation matrix @xmath56 which can be inserted in the hamiltonian as @xmath57 and then transform the matrix @xmath54 into a diagonal matrix @xmath58 .
that is to say , the term @xmath59 is equivalent to introducing the following canonical transformation and define a set of new operators , i.e. , @xmath60where @xmath61and @xmath62using the set of quasiparticle operators @xmath63 and @xmath64 , we can write the hamiltonian @xmath51 in the explicit diagonal form as @xmath65where @xmath66 are the eigenvalues of the hamiltonian matrix @xmath54 , and now , they are the relevant quasiparticle energy spectrums as follows : @xmath67furthermore , it can be verified that the general canonical transformation eq .
( [ cantran ] ) can be reduced to the familiar bogoliubov transformation in the case of the uniform anisotropic @xmath0 model .
now , let us focus on the geometric phase of the ground state .
we have introduced the family of hamiltonians parameterized by @xmath68 , and this family of hamiltonians @xmath69 can be described as a result of adiabatic rotation of the physical system .
the geometric phase of the ground state will be accumulated when the system finish a cyclic evolution , corresponding to varying the angle @xmath68 from 0 to @xmath70 [ @xmath71 is @xmath70 periodic in @xmath68 ] .
the hamiltonian @xmath51 in eq .
( [ diagham ] ) has been diagonalized in the set of quasiparticle number operators , which allows us to determine all the eigenvalues and eigenvectors .
we note that the energy spectrums @xmath73 , @xmath74 and @xmath75 , @xmath76 .
the ground state , denoted as @xmath77 and @xmath78 fermions occupied but with @xmath79 and @xmath80 fermions occupied .
explicitly , the ground state can be constructed as follows @xmath81where @xmath82 is the normalized factor and @xmath83 and @xmath84 are the vacuum states of the sublattices @xmath47 and @xmath48 , respectively .
it is easy to check that @xmath85 , @xmath86 and @xmath87 , @xmath88 for all @xmath1 .
the corresponding ground - state energy @xmath89 is @xmath90the geometric phase of the ground state , denoted @xmath91 , is given by @xmath92substituting eq .
( [ gs ] ) into eq .
( [ bg ] ) , we have @xmath93the factor of @xmath3 before the normalized factor @xmath94 is due to the repeated calculations about the @xmath1 and @xmath95 operators .
straightforward calculation is tedious .
nevertheless the result can be derived concisely from the following consideration .
we note that for each term of @xmath96 and @xmath97 in the integrand yield the same results of @xmath98 . in the same way
, the terms of @xmath99 and @xmath100 yield the results of @xmath101 , the terms of @xmath102 and @xmath103 yield the results of @xmath98 , and the terms of @xmath104 and @xmath105 yield the results of @xmath101 . finally , the overall result is @xmath106 \
\notag \\ & = & 2\pi \sum_{k>0}\left ( \cos \theta _ { k}+\cos \beta _ { k}\right ) .\end{aligned}\]]to study the quantum criticality , we are interested in the properties under the thermodynamic limit when the size of the spin lattice @xmath107 . in this case
, we introduce the notation of the geometric phase density as @xmath108 , thus , we have @xmath109 in this case , the summation @xmath110 has been replaced by the integral @xmath111 with @xmath112 . to better understand the qpts of this inhomogeneous periodic model and how the geometric phase of the ground state is used as a witness to detect them , we present numerical results for the derivative of its geometric phase @xmath113 as a function of different parameters ( @xmath114 ) in the hamiltonian . in fig.1
, we plot it as a function of @xmath115 and @xmath11 with fixed parameters @xmath116 and @xmath117 , which describes an inhomogeneous periodic ising model in a transverse field @xmath11 . for an inhomogeneous periodic ising model in a transverse field @xmath11 , as a function of the hamiltonian parameters @xmath118 and @xmath11 , in which ( @xmath119 ) .
the curves correspond to different lattice sizes @xmath120 ; ( b ) @xmath121 ; ( c ) @xmath122 ; ( d ) @xmath123.,width=316 ] as a function of @xmath124 and @xmath11 with the fixed parameters @xmath116 , @xmath125 and lattice sizes @xmath123 ; ( b ) @xmath126 as a function of @xmath127 and @xmath11 with the fixed parameters @xmath116 , @xmath128 , @xmath129 ; ( c ) @xmath126 as a function of @xmath130 and @xmath11 with the fixed parameters @xmath131 , @xmath132 , @xmath133 ; ( d ) @xmath126 as a function of @xmath127 and @xmath11 with the fixed parameters @xmath116 , @xmath134 , @xmath135.,width=316 ] for the inhomogeneous periodic @xmath0 model(@xmath136 ) as a function of the transverse field @xmath11 .
the curves correspond to different lattice sizes @xmath137 .
( b ) and ( c ) the positions of the first extremum point changes and tends as @xmath138 towards the first qpt point @xmath139 ; the positions of the second extremum changes and tends as @xmath140 towards the second qpt point @xmath141.,width=316 ] as shown in fig.1 , the peak of curves for @xmath142 becomes sharp with the increasing of the lattice size @xmath4 .
a notable feature is that the divergence of the curve in the thermodynamic limit only exists in the parameter region of @xmath143 and @xmath117 , which correspond to the case of the uniform quantum ising model , while in the other parameters regions , the curves only show extremum points . in fig.2
, we illustrate the derivative of the geometric phase of the ground state in various cases of inhomogeneous periodic systems .
an interesting thing is that there may exist two quantum phase transition points in some parameter regions for the inhomogeneous period - two systems .
the number of transition points and the corresponding divergence or extremum properties of curves are dependent on the parameters of the hamiltonian , which is quite different from those of the quantum ising model and anisotropic @xmath0 model in a transverse field @xcite . as shown in figs .
2(a ) and 2(c ) , the derivatives of the geometric phase only display the extremum instead of the divergence properties even under the thermodynamic limit condition . on the other hand , in figs .
2(b ) and 2(d ) , the extremum and divergence properties can coexist in some parameter regions . in order to further understand the divergence or extremum property of @xmath144 ,
we choose a section of fig .
2(c ) plotted in fig .
3 , in which the hamiltonian parameter takes @xmath116 , @xmath132 , @xmath145 and @xmath146 . in this case , the transition point of qpt is characterized by the extremum point . as shown in fig .
3 , there is no real divergence even in the thermodynamic limit but it tends to two extremum points with the increasing of the lattice size @xmath4 .
the transition points in the thermodynamic limit can also be obtained by the finite - size analysis of positions of extremum points for different size systems .
our results show that the position of the first extremum point approaches the first qpt point @xmath147 in a way of @xmath148 with the transition point @xmath149 and the second one approaches as @xmath150 with the transition point @xmath151 .
as shown above , the geometry phase can be used as a detector for the more complicated qpts in the inhomogeneous system .
this is because there exists an intrinsic connection between the geometric phase and the energy - level structure .
furthermore , similar connection is also reflected in the fidelity .
the relation between the fidelity and berry phase has been unveiled in terms of geometric tensors @xcite .
the intrinsic relationship between the fidelity and the characterization of qpts has also been studied in ref .
for a general hamiltonian of the quantum many - body system undergoing qpts given by @xmath152 where @xmath153 is supposed to be the driving term with @xmath154 as the control parameter , the second derivative of the ground - state energy can be expressed as @xcite @xmath155 here @xmath89 is the ground - state energy , @xmath156 are the eigenstates of @xmath157 , and @xmath158 is the ground state . on the other hand ,
the geometric phase of the system can be obtained by introducing another parameter @xmath159 to the hamiltonian eq .
( [ hlambda ] ) , i.e. , @xmath160 which is generated by a unitary transformation @xmath161 . here
, @xmath162 is unitary and satisfies @xmath163\neq 0 $ ] to ensure the nontrivial transformation .
obviously , such a transformation keeps the energy - level structures invariant and the critical behavior of the system is thus @xmath159 independent .
the eigenvalues are only characterized by the parameter @xmath164 . on the other hand
, we note that the geometric phase adiabatically undergoing a closed path @xmath165 in the @xmath159 space can be expressed as @xmath166 where @xmath167 is the berry curvature given by @xmath168 from the expressions of eqs .
( [ 2de ] ) and ( [ curvature ] ) , it is not hard to find that for both of them the singularities may come from the vanishing energy gap in the thermodynamic limit . in the inhomogeneous @xmath0 model , we find that a gapless excitation occurs only when @xmath169 or @xmath170 , which demands @xmath171 .
hence , this condition can be achieved only in the thermodynamic limit @xmath172 and for the appropriate parameters of the hamiltonian , i.e. , @xmath173 .
apart from these special cases , there exist no solution for @xmath174 or @xmath175 and a non - zero energy gap opened .
consequently , the berry curvature in the thermodynamic limit only develops extremum points instead of divergence . in summary , we present an exact diagonalization approach for an inhomogeneous periodic anisotropic @xmath0 model in a transverse field . by introducing a general canonical transformation
, we construct an explicit expression for the ground state , and based on this , we study the geometric phase of the ground state and qpts for this model .
different from the ising chain and anisotropic @xmath0 chain in a transverse field , the inhomogeneous periodic spin chains exhibit a richer behavior of qpts .
our results show that there may exist more than one phase transition point at some parameter regions . in the language of geometric phase , detecting the qpts of a many - body system driven by the external parameter @xmath164 is equivalent to finding a path @xmath165 in the parameter space of the hamiltonian , in which the berry curvature comes to the divergence or extremum points@xcite .
p. zanardi and m. rasetti , phys .
a * 264 * , 94 ( 1999 ) ; j. pachos , p. zanardi and m. rasetti , phys .
a * 61 * , 010305(r ) ( 1999 ) ; j. a. jones et al . ,
nature * 403 * , 869 ( 2000 ) ; l. m. duan , j. i. cirac , and p. zoller , science * 292 * , 1695 ( 2001 ) ; s. l. zhu and z. d. wang , phys .
. lett . * 89 * , 097902 ( 2002 ) ; _ ibid._,*91 * 187902 ( 2003 ) .
g. chen , j. li , and j .- q .
liang , phys .
a * 74 * , 054101 ( 2006 ) ; x. x. yi and w. wang , _ ibid_. * 75 * , 032103 ( 2007 ) ; h. t. cui and j. yi , _ ibid_. * 78 * , 022101 ( 2008 ) ; a. i. nesterov and s. g. ovchinnikov , phys .
e * 78 * , 015202 ( r ) ( 2008 ) .
t. j. osborne and m. a. nielsen , phys .
a * 66 * , 032110 ( 2002 ) ; a. osterloh l. amico , g. falci , and r. fazio , nature ( london ) , * 416 * , 608 ( 2002 ) ; g. vidal j. i. latorre , e. rico , and a. kitaev phys .
rev . lett.*90 * , 227902 ( 2003 ) .
s. j. gu , h. m. kwok , w. q. ning and h. q. lin , phys .
b * 77 * , 245109 ( 2008 ) ; w. l. you , y. w. li , and s. j. gu , phys . rev .
e * 76 * , 022101 ( 2007 ) ; s. chen , l. wang , s. j. gu and y. wang , phys .
e * 76 * , 061108 ( 2007 ) . | the notion of geometric phase has been recently introduced to analyze the quantum phase transitions of many - body systems from the geometrical perspective . in this work ,
we study the geometric phase of the ground state for an inhomogeneous period - two anisotropic @xmath0 model in a transverse field .
this model encompasses a group of familiar spin models as its special cases and shows a richer critical behavior .
the exact solution is obtained by mapping on a fermionic system through the jordan - wigner transformation and constructing the relevant canonical transformation to realize the diagonalization of the hamiltonian coupled in the @xmath1-space .
the results show that there may exist more than one quantum phase transition point at some parameter regions and these transition points correspond to the divergence or extremum properties of the berry curvature . |
when a system is quenched from a homogeneous high - temperature disordered state into a low - temperature regime , well defined domains of equilibrium ordered phases form randomly and then grow with time ( `` coarsen '' ) in a self - similar way , until the domain size becomes comparable to the size of the system @xcite .
a dynamic scaling hypothesis suggests that , at late times , the evolution of the system is characterized by a single length scale @xmath22 that represents a typical linear size of the domains .
it is well established , at least for systems with a scalar order parameter , that @xmath23 with @xmath24 for _ non - conserved _ dynamics and @xmath25 for _ conserved _ dynamics @xcite .
the ising spin model evolving with glauber spin flip dynamics @xcite demonstrates behavior characteristic of the former while evolution according to kawasaki spin dynamics @xcite exemplifies the latter .
a generalization of the glauber - ising model is the @xmath0-state potts model @xcite , which has @xmath0 distinct , but equivalent , ordered phases .
experimental realizations are known for @xmath26 i.e. the ising model and for @xmath27 @xcite .
the @xmath28 case describes several cellular structures @xcite , e.g. polycrystals @xcite , soap froths @xcite , magnetic bubbles @xcite and foams @xcite .
such coarsening systems are amongst those which have received considerable attention in recent years with regard to their _ persistence _ properties @xcite .
the persistence probability @xmath29 is simply the probability that a given stochastic variable @xmath30 , with zero mean , does not change sign in the time interval @xmath31 $ ] .
theoretical and computational studies of persistence include spin systems in one @xcite and higher @xcite dimensions , diffusion fields @xcite , fluctuating interfaces @xcite , phase - ordering dynamics @xcite , lotka - volterra models @xcite and reaction - diffusion systems @xcite .
experimental studies include the coarsening dynamics of breath figures @xcite , foams @xcite , soap froths @xcite , twisted nematic liquid crystals @xcite , and one - dimensional gas diffusion @xcite .
persistence in non - equilibrium critical phenomena has also been studied in the context of the global order parameter @xmath32 ( e.g. the total magnetization of a ferromagnet ) , regarded as a stochastic process @xcite . in many systems of physical interest
, the persistence decays algebraically according to @xmath33 , where @xmath20 is the persistence exponent and is , in general , nontrivial .
the non - triviality of @xmath20 emerges as a consequence of the coupling of the field @xmath30 to its neighbors , since such coupling implies that the stochastic process at a fixed point in space and time is non - markovian . for the @xmath0-state potts model at @xmath34 ,
the fraction of spins which have never flipped up to time @xmath3 ( i.e that fraction that have _ persisted _ in their original phase ) , has been observed @xcite to obey a power law decay @xmath35 , where @xmath6 has been obtained exactly , in one - dimension ( 1d ) , by derrida _
@xcite for arbitrary @xmath0 , @xmath36 ^ 2 .
\label{eq1}\ ] ] there is a very direct way @xcite of relating the glauber dynamics of the 1d @xmath0-state potts model at zero temperature to reaction - diffusion models @xcite .
consider uncorrelated initial conditions where each of the @xmath0 phases is present with equal density @xmath37 . during any time interval @xmath38
, each spin @xmath39 has a probability @xmath40 of becoming equal to its right neighbor , @xmath40 of becoming equal to its left neighbor and a probability @xmath41 of retaining its own value .
the domain walls therefore perform random walks . upon contact ,
two domain walls react instantaneously , either _ annihilating _ @xmath42 with probability @xmath43 or _ coalescing _ @xmath2 with probability @xmath44 , these numbers being the probabilities that the states on the far sides of the walkers are the same ( annihilation ) or different ( coalescence ) .
this reaction - diffusion model , together with the fact that in the initial condition each bond is occupied by a domain wall with probability @xmath45 , is completely equivalent to the spin problem .
the fraction of spins which have never flipped up to time @xmath3 is then simply the fraction of sites which have never been visited by a random walker .
the probabilistic algorithm for implementing the potts model through the annihilation or coalescence of random walkers allows @xmath0 to be a real number but is restricted to @xmath46 . however , an equivalent ising spin representation of the potts persistence problem @xcite permits a study @xcite of @xmath6 in the regime @xmath47 . in the present work we study the fraction of random walkers which have never encountered another walker , using the reaction - diffusion representation of the 1d glauber - potts model at zero temperature . to facilitate our discussion ,
we consider all the random walkers at time @xmath48 to be uninfected .
all the walkers are , however , considered to be infectious , so that for @xmath49 any contact between walkers ( assuming they survive the encounter ) leads to mutual contamination .
walkers that avoid all contact remain uninfected .
our goal , therefore , is to address the fraction of uninfected walkers @xmath50 up to time @xmath3 .
this problem has been addressed in some detail by a number of different authors @xcite , although under the guise of a different interpretation .
the general approach used involves monitoring the motion of a tagged test particle , released in the system at time @xmath51 .
if the tagged particle diffuses with diffusion constant @xmath52 and the other particles in the system diffuse with diffusion constant @xmath53 , the tagged particle is viewed as an external impurity diffusing through a homogeneous background .
one is then interested in addressing the survival probability of the tagged particle i.e. , the probability is has not been absorbed by another particle in the system .
of course in our terminology , this corresponds to the probability that the tagged particle remains uninfected .
it is well established @xcite that @xmath50 decays according to a power law @xmath54 , where @xmath55 is the ratio of the two diffusion constants in the problem .
when @xmath56 , corresponding to a static impurity , the model clearly maps onto the problem of persistent spins and in this case , @xmath57 .
the exponent @xmath58 is therefore considered a generalization of the standard persistence exponent @xmath6 , characterizing the survival probability of a mobile particle under the coarsening dynamics of the potts model .
a recent study of the survival probability of a mobile particle , moving according to either deterministic or stochastic rules , in given in @xcite , where the external fluctuating field takes the form of either the solution to the diffusion equation , the coarsening dynamics of the @xmath0-state potts model or spatially uncorrelated brownian signals . in the limit @xmath59
the potts model is equivalent to the @xmath60 single species coalescence process .
the dynamics of this system are particularly simple , since each particle sees only the `` cage '' formed by its two nearest neighbors .
it is immaterial that these neighbors may coalesce with more distant particles .
consequently , the kinetics of each particle involves only itself and its two nearest neighbors and one is therefore able to calculate @xmath61 exactly @xcite : @xmath62}\ . \label{inf}\ ] ] ( note that for @xmath63 , this of course reduces to the static persistence exponent , @xmath64 . ) in the limit of the ising model ( @xmath65 ) , there is , however , no corresponding exact result for general @xmath66 .
an exact result is only known for @xmath63 i.e. , @xmath67 .
a mean - field smoluchowski @xcite approach predicts for @xmath65 and general @xmath66 that , @xmath68 , in good agreement with numerical simulations @xcite .
however , when extended to the large @xmath0 limit , this smoluchowski approach differs substantially from the exact @xmath69 result eq .
( [ inf ] ) .
more recently , however , a perturbation theory developed by monthus @xcite has evaluated the exponent @xmath58 to first order in @xmath70 for arbitrary @xmath66 and at first order in @xmath66 for arbitrary @xmath0 . in this paper
, we will not study the exponent @xmath58 in complete generality since we are primarily interested in the special case where all of the particles are equally mobile @xmath71 and henceforth we therefore write @xmath72 . within the context of the current work , where we are interested in the case @xmath73 , the random walkers represent the motion of domain walls undergoing zero temperature coarsening .
the fraction of uncollided domain walls ( i.e. , uninfected walkers ) has previously been interpreted as `` domain wall persistence '' or equally , the probability that two adjacent domains survive . under this interpretation ,
majumdar _ et al .
_ @xcite have studied the quantity @xmath50 numerically for various @xmath0 .
they observed a power law decay , @xmath74 in agreement with the results of monthus @xcite and obtained a spectrum of values for @xmath5 : @xmath75 , where @xmath76 . in this paper , we extend and improve these numerical results by presenting ( sec .
ii ) more accurate data than previously achieved @xcite and exploring a larger range of @xmath0 .
we also relate the value of @xmath5 to the static persistent exponent @xmath6 via @xmath7 , where @xmath8 is a new exponent that we recently introduced in @xcite .
we additionally consider ( sec .
iii ) the coupled diffusion - limited reaction @xmath9 , @xmath10 in one - dimension with equal initial densities , @xmath77 .
we find that the density of particles decays in this model according to @xmath78 , independent of the initial walker density , characteristic of `` potts like '' behavior @xcite .
the fraction of sites unvisited by either an @xmath11 or a @xmath12 particle decays as @xmath33 with @xmath15 .
we discuss these exponents within the context of the @xmath0-state potts model and observe that the fraction of walkers which remain uninfected by their _
own _ species decays according to @xmath79 , where @xmath80 .
we also study the fraction of walkers which remain uninfected by _ both _ species and find that this too obey a power law , with @xmath81 . in analogy to sec .
ii we express the exponent @xmath19 is terms of the static persistence exponent @xmath20 and similarly obtain @xmath21 .
we conclude in sec .
iv with a summary of the results and a discussion of some open questions .
we investigate numerically the fraction of uninfected walkers @xmath50 in the 1d @xmath0-state potts model , at zero temperature , as a function of @xmath0 .
our simulations are performed on a 1d lattice of size @xmath82 with periodic boundary conditions . at @xmath83 a random walker
is placed at each lattice site , so that @xmath84 .
we choose this initial high density of walkers to accelerate the system s evolution into the asymptotic regime .
our model is updated using the direct method @xcite i.e. at each computational step , a particle is picked at random and moved with probability @xmath85 to a neighboring site , where @xmath53 is the diffusion constant .
if the destination site is occupied , the two particles either annihilate , with probability @xmath43 , or coalesce , to become a single particle , with probability @xmath44 , in accordance with the reaction - diffusion dynamics of the potts model detailed in sec . i. for each move of a particle , time @xmath3 is increased by @xmath86 , where @xmath87 is the current number of particles in the system .
such sequential dynamics can be chosen without loss of generality as parallel dynamics exhibit similar asymptotic behavior @xcite .
our simulations are performed for @xmath88 time steps and our results are averaged over @xmath89 runs .
[ fig1 ] clearly shows that the fraction of uninfected walkers decays according to a power - law , @xmath90 . in table
i we present our numerical values for @xmath5 in the range @xmath91 .
these values were obtained by performing a linear regression on log - log curves such as those presented in fig.[fig1 ] .
the regression was taken in the range @xmath92 .
this range was chosen to avoid initial transients and to avoid statistical fluctuations between different runs which become prominent as @xmath3 increases , especially for large @xmath0 . in fig .
[ fig2 ] we plot @xmath19 as a function of @xmath0 , where we choose to represent the data on a linear - log plot merely for clarity of presentation . [ cols="^,^",options="header " , ] table vi . numerical results for the exponent @xmath8 in various 1d models for which @xmath93^{\gamma}$ ] ( results for the first two models from ref
@xcite ) .
finally , we note that all values of @xmath8 obtained so far lie in the range @xmath94 spanned by the @xmath0-state potts model with @xmath95 . up to now our analysis of @xmath8 has been purely numerical , and naturally a more fundamental understanding of this exponent is required .
we thank satya majumdar for drawing our attention to some related work .
this work was supported by epsrc ( uk ) . | the dynamics of the one - dimensional @xmath0-state potts model , in the zero temperature limit , can be formulated through the motion of random walkers which either annihilate ( @xmath1 ) or coalesce ( @xmath2 ) with a @xmath0-dependent probability .
we consider all of the walkers in this model to be mutually infectious .
whenever two walkers meet , they experience mutual contamination .
walkers which avoid an encounter with another random walker up to time @xmath3 remain uninfected .
the fraction of uninfected walkers is known to obey a power law decay @xmath4 , with a nontrivial exponent @xmath5 [ c. monthus , phys .
rev .
e * 54 * , 4844 ( 1996 ) ; s. n. majumdar and s. j. cornell , _ ibid . _ * 57 * , 3757 ( 1998 ) ] .
we probe the numerical values of @xmath5 to a higher degree of accuracy than previous simulations and relate the exponent @xmath5 to the persistence exponent @xmath6 [ b. derrida , v.hakim and v. pasquier , phys .
rev .
lett . * 75 * , 751 ( 1995 ) ] , through the relation @xmath7 where @xmath8 is a new exponent introduced in [ s. j. odonoghue and a. j. bray , cond - mat/0111133 ] .
our study is extended to include the coupled diffusion - limited reaction @xmath9 , @xmath10 in one - dimension with equal initial densities of @xmath11 and @xmath12 particles .
we find that the density of walkers decays in this model as @xmath13 .
the fraction of sites unvisited by either an @xmath11 or a @xmath12 particle is found to obey a power law , @xmath14 with @xmath15 .
we discuss these exponents within the context of the @xmath0-state potts model and present numerical evidence that the fraction of walkers which remain uninfected decays as @xmath16 , where @xmath17 when infection occurs between like particles only , and @xmath18 when we also include cross - species contamination .
we find that the relation between @xmath19 and @xmath20 in this model can also be characterized by an exponent @xmath8 , where similarly , @xmath21 . 2 |
the blood circulation , the transport of cells and substances in the human body as well as some liquid - cooling systems and oil - recovery / oil - transport processes in engineering , are modeled by the equations of fluid motion posed in thin domains . the present paper studies the stokes equation with varying viscosity in a tube structure . in two - dimensional case a tube structure ] ( or pipe - wise structure ) is some connected union of thin rectangles with heights of order @xmath2 and with bases of order @xmath1 with a smoothened boundary @xmath3 ; here @xmath2 is a small positive parameter .
an asymptotic expansion for the case of a constant viscosity have been constructed in @xcite .
it is `` compiled '' of some poiseuille flows inside the rectangles glued by some boundary layer solutions , in the neighborhood of the junctions .
here we consider a more general case when the viscosity is not constant but depends on a longitudinal variable for each rectangle .
this situation models a blood flow in a vessel structure where the viscosity depends on the concentration of some substances diluted in blood or some blood cells . indeed , the asymptotic analysis of the convection - diffusion equation set in such domains @xmath4 shows that in the case of the neumann ( impermeability ) condition at the lateral boundary and small reynolds numbers , the concentration is asymptotically close to the one - dimensional description , that is the convection - diffusion equation set on the graph .
the solution of the problem on the graph is the leading term of the asymptotic expansion , and it evidently on depends on the longitudinal variable . on the other hand ,
the viscosity often depends on the concentration of the diluted substances or distributed cells , and so , it depends on the longitudinal variable . of course
, the fluid motion equation is coupled with the diffusion - convection equation in this case . however ,
if the velocity is small ( in our case , it is of order @xmath5 ) , then neglecting the convection , in comparison with the diffusion term or iterating with respect to the small term{l}-\mathrm{div}(\nu(c_{\varepsilon}){\mathcal{d}}u_{\varepsilon})+\nabla p_{\varepsilon}=f(x)\\ \mathrm{div}u_{\varepsilon}=0,\\ -\delta c_{\varepsilon}+ u_{\varepsilon}.\nabla c_{\varepsilon}=0 \end{array}\ ] ] where @xmath6 is the concentration , @xmath7 is the fluid velocity and @xmath8 is the pressure.if @xmath7 is small with respect to @xmath6 , we can write an expansion for @xmath6 with respect to the small parameter that is the ratio of magnitudes of @xmath9 and @xmath10 . then for the terms of this expansion , the third equation may be solved before the fluid motion equations .
another possible approach is the successive approximations ( fixed point iterations ) : @xmath11{l}-\mathrm{div}(\nu(c^{(n)}_{\varepsilon}){\mathcal{d}}u^{(n)}_{\varepsilon } ) + \nabla p^{(n)}_{\varepsilon}=f(x)\\ \mathrm{div}u^{(n)}_{\varepsilon}=0,\\ -\delta c^{(n)}_{\varepsilon}+ u^{(n-1)}_{\varepsilon}.\nabla c^{(n)}_{\varepsilon}=0 \end{array}\ ] ] where @xmath12 is the number of the iteration.in both approaches , we get a problem with the variable viscosity . ] , we get the steady state diffusion equation ; in absence of the source term in the right hand side , it has a piecewise - linear asymptotic solution on the graph for the concentration .
so , in this simplified situation , the diffusion equation can be solved before the fluid motion equation , and we obtain for the flow , the stokes or navier - stokes equation with a variable viscosity depending ( via concentration ) on the longitudinal variable .
these arguments provide the motivation for the stokes equation with variable viscosity ; to our knowledge , it has not been studied earlier from the asymptotic point of view.there are , of course , many other practical problems involving fluids with variable viscosity . for example , the presence of bacteria in suspension ( see @xcite ) may change locally the viscosity . in the first part , we consider the case of a flow in one rectangular channel with the periodicity condition at the end of the channel .
an asymptotic expansion of solution is constructed and justified ( a regular ansatz ) . in the second part
, we consider the case of a flow in one rectangular channel with the inflow / outflow boundary conditions at the ends . in this case as well we construct an asymptotic expansion of solution , that contains a regular ansatz and the boundary layer correctors . in the third part
, we construct an asymptotic expansion for solution of the stokes equation set in a tube structure . here
as well we associate to each rectangle a regular ansatz and then we glue all the regular ansatzes with help of the boundary layer correctors exponential decaying from the junctions of rectangles .
this procedure is similar to the procedure described in @xcite .
in the fourth part we consider some numerical experiments comparing the numerical and asymptotic solutions .
all constructed expansions are justified by calculation of residual terms and application of _ a priori _ estimates .
consider a small parameter @xmath2 , @xmath13 , @xmath14 , and define then a thin domain @xmath15 assume that incompressible , viscous fluid fills the domain @xmath16 . let @xmath17 be the exterior force applied to the fluid .
[ h ] rectangle.eps consider the following steady state stokes problem:@xmath18{ll}-\mathrm{div}(\nu(x_{1}){\mathcal{d}}u_{\varepsilon})+\nabla p_{\varepsilon } = f(x ) & \text{in } d_{\varepsilon},\\ \mathrm{div}u_{\varepsilon}=0 & \text{in } d_{\varepsilon},\\ u_{\varepsilon}(x_{1},\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ u_{\varepsilon}(x_{1},-\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ u_{\varepsilon}(0,x_{2})=\varepsilon^{2}\varphi_{0}(\frac{x_{2}}{\varepsilon } ) & \text{for } x_{2}\in(-\frac{\varepsilon}{2},\frac{\varepsilon}{2}),\\ u_{\varepsilon}(1,x_{2})=\varepsilon^{2}\varphi_{1}(\frac{x_{2}}{\varepsilon } ) & \text{for } x_{2}\in(-\frac{\varepsilon}{2},\frac{\varepsilon}{2 } ) , \end{array } \right .
\label{1.1}\ ] ] the unknowns of this system are the velocity @xmath7 and the pressure @xmath8 of the fluid .
the non homogeneous boundary conditions for the velocity are given by the functions @xmath19 whose second component is equal to zero , and the first satisfies @xmath20 where @xmath21)$ ] .
here @xmath22 , @xmath23 satisfies the following conditions @xmath24 where @xmath25)$ ] .
moreover there exist @xmath26 @xmath27 such that @xmath28 for all @xmath29 and @xmath30 for all @xmath31 ( and so , @xmath32 ) . in order to obtain the variational formulation of problem ,
let us introduce the following space@xmath33 and assume that @xmath34 . applying the extension theorem ( see @xcite ) , one can find a function @xmath35 such that @xmath36 , @xmath37 and @xmath38 and @xmath39 . changing the unknown function @xmath7 by @xmath40 we give the variational formulation of problem : @xmath41
we say that @xmath7 is a weak solution of problem if @xmath42 and satisfies .
if @xmath7 is a weak solution for problem , then there exists a distribution @xmath43 such that @xmath44 satisfies @xmath45 this problem in sense of distributions . _ proof . _
if we take @xmath46 , then from @xmath47 we have that @xmath48 from de rham lemma , it follows that there exists a distribution @xmath8 , unique up to an additive constant , such that @xmath49 [ th2.1]the variational problem admits a unique solution @xmath50 .
_ the riesz s theorem gives the existence and the uniqueness of solution because the norms @xmath51 and @xmath52 are equivalent . as a consequence
, we have : let @xmath50 be the solution of the variational problem .
then the following inequality holds@xmath53 where @xmath54 stands for the poincar - friedrichs inequality constant and @xmath55 is the lower bound of the viscosity .
mention that @xmath54 can be estimated by @xmath56 , where @xmath57 is independent of @xmath2 ( see @xcite ) .
consider the case when @xmath58 then we have : the following inequality holds@xmath59 where @xmath60 is a constant independent of @xmath61 _ proof . _ from de rham lemma
, it follows that there exists a distribution @xmath8 , unique up to an additive constant , such that @xmath62 it means that @xmath63 and so , @xmath64 @xmath65 where @xmath66 is the constant of the a priori estimate for @xmath67 ( see above ) and @xmath54 is the poincar - friedrichs constant ( it is of order @xmath2 ) .
this inequality gives the estimate of @xmath68 in the @xmath69norm .
an asymptotic analysis of problem shows that an asymptotic solution is given by a poiseuille type flow , with two boundary layer correctors localized in some neighborhoods of the ends of the channel .
didactically , it would be better to separate the construction of the poiseuille type flow for varying viscosity and the construction of the boundary layer correctors .
that s why we simplify problem @xmath70 , and replace @xmath71 , @xmath72 by the periodicity condition with respect to @xmath73 .
introduce the sobolev space @xmath74 here @xmath75 is the completion ( with respect to the @xmath76-norm ) of the space of the @xmath77)$]-functions vanishing at the boundary @xmath78 and 1-periodic in @xmath73 . as in the beginning of the section , @xmath34 .
we say that @xmath79 is a weak solution of the periodic problem @xmath18{ll}-\mathrm{\operatorname{div}}(\nu(x_{1}){\mathcal{d}}u_{\varepsilon})+\nabla p_{\varepsilon}=f(x ) , & \text{in } d_{\varepsilon},\\ \mathrm{\operatorname{div}}u_{\varepsilon}=0 , & \text{in } d_{\varepsilon},\\ u_{\varepsilon}(x_{1},\frac{\varepsilon}{2})=0 , & \text{for } x_{1}\in(0,1),\\ u_{\varepsilon}(x_{1},-\frac{\varepsilon}{2})=0 , & \text{for } x_{1}\in(0,1),\\ u_{\varepsilon}\text { is } 1-\text{periodic in } x_{1 } , & \end{array } \right .
\label{a.1}\ ] ] if and only if it satisfies the integral identity @xmath80 as in theorem [ th2.1 ] , we apply the riesz theorem and prove the existence and the uniqueness of @xmath79 , solution of problem .
the _ a priori _ estimate is given by @xmath81 where @xmath82 and @xmath55 is the lower bound of the viscosity . if @xmath7 is a weak solution for problem then there exists a distribution @xmath43 such that @xmath44 satisfies @xmath45 in sense of distributions , and @xmath59 where @xmath60 is a constant independent of @xmath61 _ proof .
_ by @xmath83 @xmath84 from de rham lemma , it follows that there exists a distribution @xmath8 , unique up to an additive constant , such that @xmath62 it means that @xmath85 and so , @xmath86 @xmath65 where @xmath66 is the constant of the a priori estimate for @xmath67 ( see above ) and @xmath54 is the poincar - friedrichs constant ( it is of order @xmath2 ) .
this inequality gives the estimate of @xmath68 in the @xmath69norm .
let us first construct the asymptotic expansion for the solution of the periodic problem ; then we will study the non periodic one .
define the infinite layer @xmath87 assume that @xmath88 , @xmath89 , @xmath90 is 1-periodic in @xmath73 .
[ ptb ] infini.eps denote @xmath91 , @xmath92 .
an asymptotic solution is written as:@xmath18{lcl}u_{1}^{k}(x_{1},x_{2 } ) & = & \displaystyle{\sum_{j=0}^{k}}\varepsilon ^{j+2}u_{1,j}\left ( x_{1},\frac{x_{2}}{\varepsilon}\right ) , \\ u_{2}^{k}(x_{1},x_{2 } ) & = & \displaystyle{\sum_{j=0}^{k}}\varepsilon ^{j+3}u_{2,j}\left ( x_{1},\frac{x_{2}}{\varepsilon}\right ) , \\
p^{k}(x_{1},x_{2 } ) & = & \displaystyle{\sum_{j=0}^{k}}\varepsilon^{j+1}p_{j}\left ( x_{1},\frac{x_{2}}{\varepsilon}\right ) + \displaystyle{\sum _ { j=0}^{k}}\varepsilon^{j}q_{j}(x_{1 } ) , \end{array } \right .
\label{a.2}\ ] ] with @xmath93 and @xmath94 1-periodic functions in @xmath73 , such that @xmath95 . substituting , in , equating the coefficients of the same powers of @xmath2 and denoting @xmath96 we obtain:@xmath18{l}-\frac{\partial}{\partial x_{1}}\left ( \nu(x_{1})\frac{\partial u_{1,j-2}}{\partial x_{1}}\right ) -\frac{\nu(x_{1})}{2}\frac{\partial^{2}u_{1,j}}{\partial\xi_{2}^{2}}-\frac{\nu(x_{1})}{2}\frac{\partial^{2}u_{2,j-2}}{\partial\xi_{2}\partial x_{1}}+\frac{\partial p_{j-1}}{\partial x_{1}}+\frac{\partial q_{j}}{\partial x_{1}}=f_{1}\delta_{j0},\\ \\
\frac{1}{2}\frac{\partial}{\partial x_{1}}\left [ \nu(x_{1})\left ( \frac{\partial u_{1,j-1}}{\partial\xi_{2}}+\frac{\partial u_{2,j-3}}{\partial x_{1}}\right ) \right ] + \nu(x_{1})\frac{\partial^{2}u_{2,j-1}}{\partial \xi_{2}^{2}}-\frac{\partial p_{j}}{\partial\xi_{2}}=0,\\ \\ \frac{\partial u_{1,j}}{\partial x_{1}}+\frac{\partial u_{2,j}}{\partial \xi_{2}}=0,\\
\\ u_{1,j}\left ( x_{1},\pm\frac{1}{2}\right ) = 0,\\ \\ u_{2,j}\left ( x_{1},\pm\frac{1}{2}\right ) = 0 . \end{array } \right .
\label{a.3}\ ] ] denote : @xmath97 ( it satisfies @xmath98 , @xmath99 ) and @xmath100 ( here @xmath101 ) .
denote : @xmath18{l}d^{-1}:f\longrightarrow\displaystyle{\int_{-\frac{1}{2}}^{\xi_{2}}f(x_{1},\tau)\mathrm{d}\tau,}\\ \tilde{d}^{-1}:f\longrightarrow\displaystyle{\int_{-\frac{1}{2}}^{\xi_{2}}f(x_{1},\tau)\mathrm{d}\tau-\int_{-\frac{1}{2}}^{\frac{1}{2}}\int_{-\frac { 1}{2}}^{\theta}f(x_{1},\tau)\mathrm{d}\tau\mathrm{d}\theta,}\\ d^{-2}:f\longrightarrow\displaystyle{\int_{-\frac{1}{2}}^{\xi_{2}}\int _ { -\frac{1}{2}}^{\theta}f(x_{1},\tau)\mathrm{d}\tau\mathrm{d}\theta-\left ( \xi_{2}+\frac{1}{2}\right ) \int_{-\frac{1}{2}}^{\frac{1}{2}}\int_{-\frac { 1}{2}}^{\theta}f(x_{1},\tau)\mathrm{d}\tau\mathrm{d}\theta.}\end{array } \right.\ ] ] the unknowns of @xmath102 are given by the following relations : @xmath18{lcl}u_{1,j } & = & -d^{-2}\left\ { \frac{\partial^{2}u_{2,j-2}}{\partial\xi _ { 2}\partial x_{1}}+\frac{2}{\nu(x_{1})}\left ( \frac{\partial}{\partial x_{1}}\left ( \nu(x_{1})\frac{\partial u_{1,j-2}}{\partial x_{1}}\right ) -\frac{\partial p_{j-1}}{\partial x_{1}}\right ) \right\ } \\ & & + \frac{2}{\nu(x_{1})}n_{1}(\xi_{2})\left ( \frac{\partial q_{j}}{\partial x_{1}}-f_{1}\delta_{j0}\right ) , \\ & & \\ u_{2,j } & = & d^{-1}d^{-2}\left\ { \frac{\partial^{3}u_{2,j-2}}{\partial \xi_{2}\partial x_{1}^{2}}+\frac{\partial}{\partial x_{1}}\left ( \frac{2}{\nu(x_{1})}\left ( \frac{\partial}{\partial x_{1}}\left ( \nu(x_{1})\frac{\partial u_{1,j-2}}{\partial x_{1}}\right ) -\frac{\partial p_{j-1}}{\partial x_{1}}\right ) \right ) \right\ } \\ & & -\frac{\partial}{\partial x_{1}}\left ( \frac{2}{\nu(x_{1})}\left ( \frac{\partial q_{j}}{\partial x_{1}}-f_{1}\delta_{j0}\right ) \right ) n_{2}(\xi_{2}),\\ & & \\ p_{j } & = & \tilde{d}^{-1}\left\ { \frac{1}{2}\frac{\partial}{\partial x_{1}}\left [ \nu(x_{1})\left ( \frac{\partial u_{1,j-1}}{\partial\xi_{2}}+\frac{\partial u_{2,j-3}}{\partial x_{1}}\right ) \right ] + \nu(x_{1})\frac{\partial^{2}u_{2,j-1}}{\partial\xi_{2}^{2}}\right\ } .
\end{array } \right . \label{a.4}\ ] ] _ proof .
_ integrating twice @xmath103 and using boundary conditions @xmath104 , we get @xmath105 .
this relation gives an expression of the unknown @xmath106 via @xmath94 .
all other functions contained by this relation are either known from previous computations or equal to zero .
we integrate next the incompressibility condition @xmath107 with respect to @xmath108 with the boundary condition@xmath109 and get @xmath110 . finally , integrating @xmath111 we get @xmath112 .
the unknown function @xmath94 is determined from the boundary condition @xmath113 .
actually , the boundary conditions @xmath104 and @xmath114 are satisfied by the definition of @xmath115 and @xmath116 , while @xmath113 gives : @xmath117 in particular , for @xmath118 , we have the darcy equation for the leading term of the pressure @xmath119:@xmath120 and so @xmath119 is the 1-periodic function given by the formula @xmath121 then @xmath122 , @xmath123 and @xmath124.for @xmath125 @xmath126 is the 1-periodic solution of equation @xmath127 it follows : @xmath128 , @xmath129 . for @xmath125 , @xmath130 yields : @xmath131 for @xmath132 , becomes @xmath133 where @xmath134 , @xmath135 is 1-periodic solution of given by : @xmath136 it follows : @xmath18{l}u_{1,2}(x_{1},\xi_{2 } ) = \frac{2}{\nu(x_{1 } ) } \frac{\partial^{2}}{\partial x_{1}^{2 } } ( \frac{\partial q_{0}}{\partial x_{1 } } - f_{1 } ) d^{-2}\left ( n_{1}(\frac{x_{2}}{\varepsilon } ) -\langle n_{1}\rangle_{2 } \right ) + \frac { 2}{\nu(x_{1 } ) } n_{1}(\frac{x_{2 } } { \varepsilon } ) \frac{\partial q_{2}}{\partial x_{1}},\\ \\
u_{2,2}(x_{1},\xi_{2})=- \frac{\partial}{\partial x_{1}}\left [ \frac{2}{\nu(x_{1})}\frac{\partial^{2}}{\partial x_{1}^{2 } } \left ( \frac{\partial q_{0}}{\partial x_{1 } } - f_{1}\right ) \right ] d^{-1}d^{-2}\left ( n_{1}(\frac{x_{2}}{\varepsilon})-\langle n_{1}\rangle_{2}\right ) - \frac{\partial } { \partial x_{1}}\left ( \frac{2}{\nu(x_{1 } ) } n_{2}(\frac{x_{2}}{\varepsilon } ) \frac{\partial q_{2}}{\partial x_{1}}\right ) , \end{array } \right.\ ] ] and we get the following theorem . the asymptotic solution of problem is given by:@xmath18{lcl}u_{1}^{k}(x_{1},x_{2 } ) & = & \varepsilon^{2}\frac{2}{\nu(x_{1})}\left ( \frac{\partial q_{0}}{\partial x_{1}}-f_{1}\right ) n_{1}(\frac{x_{2}}{\varepsilon})\\ & & + \varepsilon^{4}\left ( \frac{2}{\nu(x_{1})}\frac{\partial^{2}}{\partial x_{1}^{2}}(\frac{\partial q_{0}}{\partial x_{1}}-f_{1})d^{-2}\left ( n_{1}(\frac{x_{2}}{\varepsilon})-\langle n_{1}\rangle_{2}\right ) + \frac { 2}{\nu(x_{1})}n_{1}(\frac{x_{2}}{\varepsilon})\frac{\partial q_{2}}{\partial x_{1}}\right ) + o(\varepsilon^{5}),\\ & & \\ u_{2}^{k}(x_{1},x_{2 } ) & = & -\varepsilon^{5}\left ( \frac{\partial}{\partial x_{1}}\left [ \frac{2}{\nu(x_{1})}\frac{\partial^{2}}{\partial x_{1}^{2}}\left ( \frac{\partial q_{0}}{\partial x_{1}}-f_{1}\right ) \right ] d^{-1}d^{-2}\left ( n_{1}(\frac{x_{2}}{\varepsilon})-\langle n_{1}\rangle _ { 2}\right ) \right .
\\ & & \left .
+ \frac{\partial}{\partial x_{1}}\left ( \frac{2}{\nu(x_{1})}n_{2}(\frac{x_{2}}{\varepsilon})\frac{\partial q_{2}}{\partial x_{1}}\right ) \right ) + o(\varepsilon^{6}),\\ & & \\ p^{k}(x_{1},x_{2 } ) & = & q_{0}+\varepsilon^{2}\left ( q_{2}(x_{1})+\frac{\partial}{\partial x_{1}}\left ( \frac{\partial q_{0}}{\partial x_{1}}-f_{1}\right ) \left ( n_{1}(\frac{x_{2}}{\varepsilon})-\langle n_{1}\rangle_{2}\right ) \right ) + o(\varepsilon^{3 } ) , \end{array } \right .
\label{a.7}\ ] ] where @xmath119 and @xmath135 are periodic solutions of obtained as the unique solution of respectively . solution given by can be written as follows : @xmath18{lcl}u_{1}^{k}(x_{1},x_{2 } ) & = & -\varepsilon^{2}\frac{2\langle f_{1}\rangle_{1}}{\langle\nu(x_{1})\rangle_{1}}n_{1}(\xi_{2})-\varepsilon^{4}\left ( \frac { 2}{\nu(x_{1})}\frac{\langle f_{1}\rangle_{1}}{\langle\nu\rangle_{1}}\nu^{^{\prime\prime}}(x_{1})d^{-2}\left ( n_{1}(\frac{x_{2}}{\varepsilon } ) -\langle n_{1}\rangle_{2}\right ) \right .
\\ & & \left .
+ \frac{24{\mathcal{r}}\langle f_{1}\rangle_{1}}{\langle\nu ( x_{1})\rangle_{1}}\frac{\nu^{^{\prime}}(x_{1})}{\nu(x_{1})}n_{1}(\frac{x_{2}}{\varepsilon})\right ) + o(\varepsilon^{5}),\\ & & \\ u_{2}^{k}(x_{1},x_{2 } ) & = & \varepsilon^{5}\left ( \frac{\partial}{\partial x_{1}}\left [ \frac{2}{\nu(x_{1})}\frac{\langle f_{1}\rangle_{1}}{\langle \nu\rangle_{1}}\nu^{^{\prime\prime}}(x_{1})\right ] d^{-1}d^{-2}\left ( n_{1}(\frac{x_{2}}{\varepsilon})-\langle n_{1}\rangle_{2}\right ) \right .
\\ & & \left .
+ \frac{24{\mathcal{r}}\langle f_{1}\rangle_{1}}{\langle\nu ( x_{1})\rangle_{1}}(\frac{\nu^{^{\prime}}(x_{1})}{\nu(x_{1})})^{\prime}n_{2}(\frac{x_{2}}{\varepsilon})\right ) + o(\varepsilon^{6}),\\ & & \\ p^{k}(x_{1},x_{2 } ) & = & \displaystyle{\int_{0}^{x_{1}}\left ( f_{1}(\tau)-\frac{\langle f_{1}\rangle_{1}}{\langle\nu\rangle_{1}}\nu(\tau)\right ) \mathrm{d}\tau-\left\langle \int_{0}^{x_{1}}\left ( f_{1}(\tau)-\frac{\langle f_{1}\rangle_{1}}{\langle\nu\rangle_{1}}\nu(\tau)\right ) \mathrm{d}\tau\right\rangle _ { 1}}\\ & & + \varepsilon^{2}\left ( -12{\mathcal{r}}{\langle f_{1}\rangle_{1}}\left ( \frac{\nu(x_{1})}{\langle\nu\rangle_{1}}-1\right ) -\frac{\langle f_{1}\rangle_{1}}{\langle\nu\rangle_{1}}\nu^{^{\prime}}(x_{1})\left ( n_{1}(\frac{x_{2}}{\varepsilon})-\langle n_{1}\rangle_{2}\right ) \right ) + o(\varepsilon^{3 } ) .
\end{array } \right.\ ] ] introduce the following function:@xmath11{lcl}f^{k}(x_{1},x_{2 } ) & = & \left ( \frac{\partial}{\partial x_{1}}\left ( \nu(x_{1})\frac{\partial u_{1,k-1}}{\partial x_{1}}\right ) + \frac{\nu(x_{1})}{2}\frac{\partial^{2}u_{2,k-1}}{\partial x_{1}\partial\xi_{2}}-\frac{\partial p_{k}}{\partial x_{1}}\right ) e_{1}\\ & & \\ & + & \left ( \frac{1}{2}\frac{\partial}{\partial x_{1}}\left [ \nu ( x_{1})\left ( \frac{\partial u_{1,k}}{\partial\xi_{2}}+\frac{\partial u_{2,k-2}}{\partial x_{1}}\right ) \right ] + \nu(x_{1})\frac{\partial ^{2}u_{2,k}}{\partial\xi_{2}^{2}}\right ) e_{2}\\ & & \\ & + & \varepsilon\left ( \frac{\partial}{\partial x_{1}}\left ( \nu ( x_{1})\frac{\partial u_{1,k}}{\partial x_{1}}\right ) + \frac{\nu(x_{1})}{2}\frac{\partial^{2}u_{2,k}}{\partial x_{1}\partial\xi_{2}}\right ) e_{1}+\left ( \frac{1}{2}\frac{\partial}{\partial x_{1}}\left ( \nu ( x_{1})\frac{\partial u_{2,k-1}}{\partial x_{1}}\right ) \right ) e_{2}\\ & & \\ & + & \varepsilon^{2}\left ( \frac{1}{2}\frac{\partial}{\partial x_{1}}\left ( \nu(x_{1})\frac{\partial u_{2,k}}{\partial x_{1}}\right ) \right ) e_{2}. \end{array } \label{rol}\ ] ] let @xmath137 be the asymptotic solution given by and @xmath138 the solution of .
then the following estimate holds : @xmath139 _ proof .
_ denote @xmath140 .
we obtain the following problems for @xmath141 : @xmath142{ll}-\mathrm{div}(\nu(x_{1}){\mathcal{d } } u^{k } ) + \nabla p^{k}= -\varepsilon ^{k+1}f^{k } ( x_{1},x_{2 } ) & \quad in \quad d_{\varepsilon},\\ \mathrm{div } u^{k}= 0 & \quad in \quad d_{\varepsilon},\\ u^{k}(x_{1},\frac{\varepsilon}{2})=0 & \quad for \quad x_{1}\in(0,1),\\ u^{k}(x_{1},-\frac{\varepsilon}{2})= 0 & \quad for \quad x_{1}\in(0,1),\\ u^{k } \quad is \quad1-periodic \quad in \quad x_{1}. & \end{array } \right.\ ] ] the a priori estimates give us : @xmath143 and @xmath144 let us consider now the non periodic case , i.e. let us construct an asymptotic expansion of the solution for problem .
assume @xmath145).\ ] ] define an asymptotic solution by:@xmath18{lcl}\hat{u}^{k}(x_{1},x_{2 } ) & = & { u}^{k}(x_{1},x_{2})+{u}_{bl0}^{k}(\frac { x}{\varepsilon})+{u}_{bl1}^{k}(\frac{x_{1}-1}{\varepsilon},\frac{x_{2}}{\varepsilon})\\ \hat{p}^{k}(x_{1},x_{2 } ) & = & { p}^{k}(x_{1},x_{2})+{p}_{bl0}^{k}(\frac { x}{\varepsilon})+{p}_{bl1}^{k}(\frac{x_{1}-1}{\varepsilon},\frac{x_{2}}{\varepsilon } ) \end{array } \right .
\label{b.1}\ ] ] the expressions of @xmath146 , @xmath147 are the same as above : , , since the functions given by do not satisfy exactly the boundary conditions @xmath148 we add the boundary layer correctors .
they correspond to the left end for @xmath149 and to the right end for @xmath150 and their expressions are given by:@xmath18{lcl}{u}_{bli}^{k}(\frac{x_{1}-i}{\varepsilon},\frac{x_{2}}{\varepsilon } ) & = & \displaystyle{\sum_{j=0}^{k}}\varepsilon^{j+2}u_{j}^{i}\left ( \frac{x_{1}-i}{\varepsilon},\frac{x_{2}}{\varepsilon}\right ) \\ { p}_{bli}^{k}(\frac{x_{1}-i}{\varepsilon},\frac{x_{2}}{\varepsilon } ) & = & \displaystyle{\sum_{j=0}^{k}}\varepsilon^{j+1}p_{j}^{i}\left ( \frac{x_{1}-i}{\varepsilon},\frac{x_{2}}{\varepsilon}\right ) \end{array } \right .
\label{b.3}\ ] ] here @xmath151 to obtain problems for the boundary layers corresponding to the left side , we introduce the domain @xmath152 .
the problem@xmath18{ll}-\frac{\nu_{0}}{2}\delta_{\xi}u_{j}^{0}(\xi)+\nabla_{\xi}p_{j}^{0}(\xi)=0 & \text{if } \xi\in\pi^{+},\\ \mathrm{div}_{\xi}u_{j}^{0}=0 & \text{if } \xi\in\pi^{+},\\ u_{j}^{0}=0 & \text{if } \xi_{2}=\pm\frac{1}{2},\\ u_{j}^{0}(0,\xi_{2})=\varphi_{01}\delta_{j0}-\left ( \begin{array } [ c]{c}u_{1,j}(0,\xi_{2})\\ u_{2,j-1}(0,\xi_{2 } ) \end{array } \right ) & \text{if } \xi_{2}\in(-\frac{1}{2},\frac{1}{2 } ) , \end{array } \right .
\label{b.4}\ ] ] with the compatibility condition@xmath153 where @xmath154 , will give the boundary layer correctors for the velocity and for the pressure corresponding to the left end ( see @xcite ) .
this condition generates a boundary condition for @xmath94:@xmath155 in a similar way we introduce the boundary layer correctors corresponding to the right side .
the boundary layers for the velocity and pressure are defined on @xmath156 .
the analogous boundary condition for @xmath94 is satisfied automatically because of the conservation of the average of @xmath157 .
indeed,@xmath158 on the other hand , the condition @xmath159 is equivalent to the equation on @xmath94 , i.e. this equation on @xmath94 is equivalent to the conservation law for the average @xmath160 .
the problem satisfied by the asymptotic solution of order @xmath161 is as follows : @xmath18{ll}-\mathrm{div}(\nu(x_{1}){\mathcal{d}}\hat{u}^{k})+\nabla\hat{p}^{k}=f-\varepsilon^{k+1}f^{k}(x_{1},x_{2 } ) & \text{in } d_{\varepsilon},\\ \mathrm{div}\hat{u}^{k}=0 & \text{in } d_{\varepsilon},\\ \hat{u}_{\varepsilon}(x_{1},\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ \hat{u}_{\varepsilon}(x_{1},-\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ \hat{u}_{\varepsilon}(0,x_{2})=\varepsilon^{2}\varphi_{0}(\frac{x_{2}}{\varepsilon})+\varepsilon^{k+3}u_{2,k}(0,\xi_{2})e_{2}+\displaystyle{\sum _ { j=0}^{k}}\varepsilon^{j+2}u_{j}^{(1)}\left ( -\frac{1}{\varepsilon},\frac{x_{2}}{\varepsilon}\right ) & \text{for } x_{2}\in(-\frac{\varepsilon } { 2},\frac{\varepsilon}{2}),\\ \hat{u}_{\varepsilon}(1,x_{2})=\varepsilon^{2}\varphi_{1}(\frac{x_{2}}{\varepsilon})+\varepsilon^{k+3}u_{2,k}(1,\xi_{2})e_{2}+\displaystyle{\sum
_ { j=0}^{k}}\varepsilon^{j+2}u_{j}^{(0)}\left ( \frac{1}{\varepsilon},\frac{x_{2}}{\varepsilon}\right ) & \text{for } x_{2}\in(-\frac{\varepsilon } { 2},\frac{\varepsilon}{2 } ) , \end{array } \right .
\label{b.6}\ ] ] here @xmath162 is given by .
mention that the boundary conditions for @xmath163 on @xmath164 are not yet satisfied exactly because the traces of each boundary layer on the opposite side of the rectangle are exponentially small but do not vanish completely .
these traces should be eliminated by a small additional corrector .
let us construct a new function @xmath165 which satisfies the same boundary conditions as @xmath163 on @xmath164.let us describe its construction : let @xmath166 be a solution of the following problem:@xmath18{ll}u^{k}\in\left ( h^{1}(d_{\varepsilon})\right ) ^{2 } , & \\
\mathrm{div}u^{k}=0 & \text{in } d_{\varepsilon},\\ { u}^{k}(x_{1},\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ { u}^{k}(x_{1},-\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ { u}^{k}(0,x_{2})=\varepsilon^{k+3}u_{2,k}(0,\xi_{2})e_{2}+\displaystyle{\sum _ { j=0}^{k}}\varepsilon^{j+2}u_{j}^{(1)}\left ( -\frac{1}{\varepsilon},\frac{x_{2}}{\varepsilon}\right ) & \text{for } x_{2}\in(-\frac{\varepsilon } { 2},\frac{\varepsilon}{2}),\\ { u}^{k}(1,x_{2})=\varepsilon^{k+3}u_{2,k}(1,\xi_{2})e_{2}+\displaystyle{\sum _ { j=0}^{k}}\varepsilon^{j+2}u_{j}^{(0)}\left ( \frac{1}{\varepsilon},\frac{x_{2}}{\varepsilon}\right ) & \text{for } x_{2}\in(-\frac{\varepsilon } { 2},\frac{\varepsilon}{2 } ) .
\end{array } \right . \label{b.7}\ ] ] problem has at least one solution with the property @xmath167 _ proof . _
define @xmath168 , where @xmath169 , @xmath170{c}\varepsilon\left ( u^{k}\right ) _
{ 1}\\ \left ( u^{k}\right ) _ { 2}\end{array } \right ) , \quad\mbox{with } ( y_{1},y_{2 } ) = ( x_{1},\frac{x_{2}}{\varepsilon } ) .\ ] ] obvious computations lead to the following problem for @xmath171 : @xmath18{ll}\mathrm{div}_{y } w^{k}_{\varepsilon}= 0 & \quad\text{in } d,\\ w^{k}_{\varepsilon}(y_{1},\frac{1}{2})=0 & \quad\text{for } y_{1}\in(0,1),\\ w^{k}_{\varepsilon}(y_{1},-\frac{1}{2})= 0 & \quad\text{for } y_{1}\in(0,1),\\ w^{k}_{\varepsilon}(0 , y_{2})= \varepsilon^{k+3 } u_{2,k}(0,y_{2})e_{2 } + \left ( \begin{array } [ c]{c}\displaystyle{\sum_{j=0}^{k}}\varepsilon^{j+3 } u^{(1)}_{1,j}\left ( -\frac { 1}{\varepsilon},y_{2}\right ) \\ \displaystyle{\sum_{j=0}^{k}}\varepsilon^{j+2 } u^{(1)}_{2,j}\left ( -\frac { 1}{\varepsilon},y_{2}\right ) \end{array } \right ) & \quad\text{for } y_{2}\in(-\frac{1}{2},\frac{1}{2}),\\ w^{k}_{\varepsilon}(1 , y_{2})= \varepsilon^{k+3 } u_{2,k}(1,y_{2})e_{2 } + \left ( \begin{array } [ c]{c}\displaystyle{\sum_{j=0}^{k } } \varepsilon^{j+3 } u^{(0)}_{1,j}\left ( \frac { 1}{\varepsilon},y_{2}\right ) \\ \displaystyle{\sum_{j=0}^{k}}\varepsilon^{j+2 } u^{(0)}_{2,j}\left ( \frac { 1}{\varepsilon},y_{2}\right ) \end{array } \right ) & \quad\text{for } y_{2}\in(-\frac{1}{2 } , \frac{1}{2}),\\ & \end{array } \right.\ ] ] here @xmath172 is the divergence in @xmath173variables . as in @xcite
we can prove that there exists a function @xmath174 so that @xmath175 with the constant @xmath60 independent of @xmath2 . using the properties of the boundary layer correctors ( their exponential decay rate ) ,
we get : @xmath176 direct computations give @xmath177 .
combining these two estimates we achieve the proof .
the function@xmath178 satisfies the same boundary conditions as @xmath179 in @xmath180 .
the problem for the new functions @xmath165 , @xmath181 is an obvious consequence of and : @xmath18{ll}-\mathrm{div}(\nu(x_{1}){\mathcal{d}}\hat{u}^{k})+\nabla\hat{p}^{k}=f-\varepsilon^{k+1}f^{k}(x_{1},x_{2})-\mathrm{div}(\nu(x_{1}){\mathcal{d}}{u}^{k } ) & \text{in } d_{\varepsilon},\\ \mathrm{div}\hat{u}^{k}=0 & \text{in } d_{\varepsilon},\\ \hat{u}_{\varepsilon}(x_{1},\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ \hat{u}_{\varepsilon}(x_{1},-\frac{\varepsilon}{2})=0 & \text{for } x_{1}\in(0,1),\\ \hat{u}_{\varepsilon}(0,x_{2})=\varepsilon^{2}\varphi_{0}(\frac{x_{2}}{\varepsilon } ) & \text{for } x_{2}\in(-\frac{\varepsilon}{2},\frac { \varepsilon}{2}),\\ \hat{u}_{\varepsilon}(1,x_{2})=\varepsilon^{2}\varphi_{1}(\frac{x_{2}}{\varepsilon } ) & \text{for } x_{2}\in(-\frac{\varepsilon}{2},\frac { \varepsilon}{2 } ) . \end{array } \right .
\label{b.10}\ ] ] let @xmath182 be the asymptotic solution given by and @xmath138 the exact solution of . then the following estimates hold:@xmath18{l}\vert\hat{u}^{k}-u_{\varepsilon}\vert_{{h(d_{\varepsilon})}^{2}}=o\left ( \varepsilon^{k+\frac{3}{2}}\right ) , \\
\vert\nabla p^{k}-\nabla p_{\varepsilon}\vert_{h^{-1}(d_{\varepsilon})}=o\left ( \varepsilon^{k+\frac{3}{2}}\right ) .
\end{array } \right . \label{b.11}\ ] ] _ proof .
_ from it follows : @xmath183 from , and it follows : @xmath184 .
the estimate for pressure is a consequence of @xmath185 and the a priori estimate .
in this section we are going to construct an asymptotic expansion to the solution of problem , stated in a tube structure containing one bundle .
we justify the error estimate .
let us define a tube structure containing one bundle .
let @xmath186 be @xmath12 closed segments in @xmath187 , which have a single common point @xmath188 ( i.e. the origin of the co - ordinate system ) , and let it be the common end point of all these segments .
let @xmath189 be @xmath12 bounded segments in @xmath187 containing the point @xmath188 , the middle point of all segments , and such that @xmath190 is orthogonal to @xmath191 ( for simplicity assume that the length @xmath192 of each @xmath190 is equal to 1 ) .
let @xmath193 be the image of @xmath190 obtained by a homothetic contraction in @xmath194 times with the center @xmath188 .
denote @xmath195 the open rectangles with the bases @xmath193 and with the heights @xmath191 , denote also @xmath196 the second base - side of each rectangle @xmath195 and let @xmath197 be the end of the segment @xmath191 which belongs to the base @xmath196 ( see fig .
define the graph of the tube structure as the bundle of segments @xmath191 having a common point @xmath188 ( see fig . [
f4])@xmath198 denote below @xmath199 .
let @xmath200 , @xmath201 , be the images of the bounded domains @xmath202 ( such that @xmath203 contains the end of the segment @xmath197 and is independent of @xmath2 ) obtained by a homothetic contraction in @xmath194 times with the center @xmath197 .
define the tube structure associated with the bundle @xmath204 as a bounded domain ( see fig .
[ f3 ] ) : @xmath205 here the prime stands for the set of the interior points . assume that @xmath206 ( the result may be generalized for the case of the piecewise smooth boundary @xmath207 with no reentrant corners ) .
assume that the bases @xmath208 of @xmath209 @xmath210 , are some parts of @xmath207 .
we add the domains @xmath200 , @xmath201 , to smoothen the boundary of the tube structure .
[ ptb ] segments.eps [ ptb ] rectangles.eps [ ptb ] structure-tubulaire.eps consider the following system of equations:@xmath18{ll}-\mathrm{div}(\nu(x){\mathcal{d}}u_{\varepsilon})+\nabla p_{\varepsilon } = f(x ) & \text{in } { \mathcal{b}}^{\varepsilon},\\ \mathrm{div}u_{\varepsilon}=0 & \text{in } { \mathcal{b}}^{\varepsilon},\\ u_{\varepsilon}=g & \text{on } \partial{\mathcal{b}}^{\varepsilon}. \end{array } \right .
\label{1}\ ] ] here , @xmath211 on the lateral boundary of the rectangles composing @xmath212 ; moreover @xmath211 anywhere with the exception of the sides @xmath196 of the rectangles @xmath195 ( these sides are assumed to belong to the boundary of the tube structure ) ; @xmath213 , and for each @xmath214 , @xmath215 on @xmath216 , the vector valued functions @xmath217 do not depend on @xmath2 .
let @xmath17 be a vector - valued function of @xmath218 .
the solvability condition gives the relation @xmath219 introduce the local system of coordinates @xmath220 associated with the segment @xmath191 such that the direction of the axis @xmath221 coincides with the direction of the segment @xmath222 , i.e. @xmath223 is the longitudinal coordinate .
the axes @xmath220 form a cartesian coordinate system .
denote @xmath224 the infimum of radius of all circles with the center @xmath188 such that every point of it belongs only to not more than one of the rectangles @xmath195 , @xmath225 and @xmath226 is the maximal diameter of the domains @xmath227 .
we finally introduce the notation @xmath228 consider the right hand side vector valued function @xmath17 `` concentrated '' in some neighborhoods of the nodes @xmath197 and diffused in the rectangles , i.e. @xmath229 here @xmath230),\phi_{j}\in c_{0}^{1}(q)$ ] , @xmath231 where @xmath232 is a ball @xmath233 .
assume that @xmath234 such that @xmath235 for all @xmath236\cup\lbrack|e_{j}|-\beta;|e_{j}|]$ ] , where @xmath237 is a positive constant such that @xmath238 ; @xmath239 and there exist @xmath240 such that @xmath241 for all @xmath242 .
without loss of generality we may assume that @xmath243 for all @xmath244\cup\lbrack|e_{j}|-\beta;|e_{j}|]$ ] .
let @xmath245 be space of the divergence free vector valued functions from @xmath246 .
let @xmath247 be the subspace of vector valued functions of @xmath245 vanishing at the boundary .
assume that @xmath248 can be continued in @xmath212 as a vector valued @xmath249 of @xmath245 .
the variational formulation for is as follows : find @xmath250 such that @xmath251 , and such that it satisfies to the integral identity @xmath252 the riesz theorem will give the existence and the uniqueness of such a solution because the norms @xmath253 and @xmath254 are equivalent .
we have as a consequence that @xmath255 where @xmath256 is independent on @xmath2 ( see @xcite ) and @xmath257 is the lower bound of the viscosity ..
if @xmath7 is a weak solution for problem then there exists a distribution @xmath258 such that @xmath44 satisfies @xmath259 in the sense of distributions .
the following inequality holds in the case of @xmath260 : @xmath261 here @xmath60 is a constant independent of @xmath61 the proof is similar to that of the propositions of section [ sect2 ] .
we construct the main part of the asymptotic expansion in a form@xmath262@xmath11{lcl}p^{a } & = & \displaystyle{\sum_{l=0}^{k}\varepsilon^{l+1}\left\ { \sum_{e = e_{j};j=1,\dots , n}p_{l}^{e}(x^{e , l})\chi_{\varepsilon}(x)+\sum _ { i=0}^{n}p_{l}^{blo_{i}}\left ( \frac{x - o_{i}}{\varepsilon}\right ) \right\ } } + \\ & & + \displaystyle{\sum_{e = e_{j};j=1,\dots , n}\sum_{l=0}^{k}\varepsilon ^{l}q_{l}^{e}(x_{1}^{e})\chi_{\varepsilon}(x)+\sum_{i=0}^{n}q_{0}^{e_{i}}(x_{1}^{e_{i}}=0)\left ( 1-\chi_{\varepsilon}(x)\right ) \theta_{i}(x)}\end{array } \label{6}\ ] ] @xmath263 .
later , in the end of the section we will add an exponentially small corrector multiplying the boundary layers by a cut - off function @xmath264 in the subdomain where the boundary layers are just exponentially small ( see ( [ 15 ] ) , ( [ 16 ] ) ) . the last sum in ( [ 1 ] ) is taken for all nodes @xmath265 and the value @xmath266 is calculated at the point @xmath267 ; function @xmath268 is supposed to be continuous on the graph @xmath204 .
here @xmath269 is a function equal to zero at the distance less than @xmath270 from @xmath197 , @xmath201 , equal to zero on the rectangle @xmath195 if @xmath271 or if @xmath272 ; we suppose that function @xmath273 is equal to one on this rectangle if @xmath274 and @xmath275 , and we define @xmath273 by the relations @xmath276 if @xmath277 and @xmath278 if @xmath279 .
here @xmath280 is a differentiable on @xmath281 function of one variable , it is independent of @xmath2 , it is equal to zero on the segment @xmath282 $ ] and it is equal to one on the union of intervals @xmath283 .
moreover @xmath273 is equal to zero on every @xmath200 .
the functions @xmath284 , @xmath285 are defined as follows@xmath286 the relation between the vector - columns @xmath287 and @xmath288 ( here @xmath289 is the transposition symbol ) is given by @xmath290 where @xmath291 is an orthogonal matrix of passage from the canonic base to the local one . then applying the results of section [ sect2 ] , for every channel @xmath209 we get @xmath292 , @xmath293 and @xmath294 defined up to the scalar constants @xmath295 , @xmath296 .
indeed , denote @xmath297 the solution of equation ( [ a.5 ] ) with @xmath298 then the general solution of equation ( [ a.5 ] ) has a form : @xmath299 where @xmath300 and @xmath301 are the undetermined constants ; @xmath302 where the second component of @xmath303 does not depend on @xmath300 , @xmath301 ( see ( @xmath304)@xmath305 ) ; the same property holds for @xmath306 ( see ( @xmath304)@xmath307 ) ; the first component @xmath308 ( see ( @xmath304)@xmath45 ) depends on @xmath300:@xmath309 for @xmath310\bigcup[|e_{j}|-{\beta},|e_{j}|]$ ] , @xmath311 , @xmath312 and @xmath313 may be taken equal to zero because the right hand side @xmath314 is equal to zero for this values of @xmath223 , while the flow rate @xmath315 is constant on @xmath191 .
@xmath316 is the function introduced in section 2 . to get the problems for the boundary layers , we introduce the domain @xmath317 , where @xmath318 are the half - infinite strips obtained from @xmath195 by infinite extension behind the base @xmath319 and by homothetic dilatation in @xmath320 times ( with respect to the point @xmath188 ) ;
let @xmath321 be obtained from @xmath318 by a symmetric reflection relatively to the line containing @xmath193 and let @xmath322 , where @xmath323 is obtained from @xmath202 by a translation ( such that the point @xmath197 becomes @xmath188 ) . since @xmath324 for all @xmath325\cup \lbrack|e_{j}|-\beta;|e_{j}|]$ ]
, the boundary layer solution is a pair constituted of a vector valued function @xmath326 and a scalar function @xmath327 satisfying to the stokes system:@xmath18{l}-\frac{\nu_{0}}{2}\delta_{\xi}u_{l}^{blo_{0}}+\nabla_{\xi}p_{l}^{blo_{0}}=\phi_{0}(\xi)\delta_{l,0}+\\ \ \ \ + \displaystyle{\sum_{e = e_{j};j=1,\dots , n}}\left\ { c_{l}^{e}\left ( \nu_{0 } \delta_{\xi}\left ( \chi_{j}(\xi_{1}^{e})\gamma_{j}\left ( n_{1}(\xi_{2}^{e}),0\right ) ^{t}\right ) -\nabla_{\xi}\left ( \chi_{j}(\xi_{1}^{e})\xi_{1}^{e}\right ) \right ) -d_{l+1}^{e}\nabla_{\xi}\chi_{j}(\xi_{1}^{e})\right\ } , \\ \\ \mathrm{div}_{\xi}u_{l}^{blo_{0}}=-\displaystyle{\sum_{e = e_{j};\ j=1,\dots , n}}\mathrm{div}_{\xi}\left ( c_{l}^{e}\left ( \chi_{j}(\xi_{1}^{e})\gamma _ { j}\left ( n_{1}(\xi_{2}^{e}),0\right ) ^{t}\right ) \right ) , \text { \ \ if } \xi\in\omega_{o_{0}},\\ { u_{l}^{blo_{0}}}{|_{\partial\omega_{o_{0}}}}=0 , \end{array } \right .
\label{8}\ ] ] and for @xmath328@xmath18{l}-\frac{\nu_{0}}{2}\delta_{\hat{\xi}}u_{l}^{blo_{j}}+\nabla_{\hat{\xi}}p_{l}^{blo_{j}}=\\ \ \ \
= \phi_{j}(\hat{\xi})\delta_{l,0}+\hat{c}_{l}^{e}\left ( \nu_{0 } \delta_{\hat{\xi}}\left ( \chi_{j}(\hat{\xi}_{1}^{e})\right ) \hat{\gamma}_{j}\left ( n_{1}(\hat{\xi}_{2}^{e}),0\right ) ^{t}-\nabla_{\hat{\xi}}\left ( \chi_{j}(\hat{\xi}_{1}^{e})\hat{\xi}_{1}^{e}\right ) \right ) -\hat{d}_{l+1}^{e}\nabla_{\hat{\xi}}\chi_{j}(\hat{\xi}_{1}^{e}),\\ \\ \mathrm{div}_{\hat{\xi}}u_{l}^{blo_{j}}=-\hat{c}_{l}^{e_{j}}\mathrm{div}_{\hat{\xi}}\left ( \chi_{j}(\hat{\xi}_{1}^{e_{j}})\hat{\gamma}_{j}\left ( n_{1}(\hat{\xi}_{2}),0\right ) ^{t}\right ) , \text { \ \ if } \hat{\xi}\in \omega_{o_{j}},\\ { u_{l}^{blo_{j}}}_{|_{\partial\omega_{o_{j}},\hat{\xi}_{1}^{e_{j}}=0}}=g_{j}\delta_{l,0},\\ { u_{l}^{blo_{j}}}_{|_{\partial\omega_{o_{j}},\hat{\xi}_{1}^{e_{j}}\neq0}}=0 .
\end{array } \right .
\label{9}\ ] ] the variable @xmath329 is opposite to @xmath330 , i.e. to the first component of the vector @xmath331 .
so @xmath332 , where @xmath333 and @xmath334 is the diagonal matrix with the diagonal elements @xmath335 .
the constants @xmath336 , @xmath337 are defined in such a way that the functions @xmath338 and @xmath339 are equal , i.e.@xmath340 assume that every term in the sum @xmath341 in is defined only in the branch of @xmath342 , corresponding to @xmath343 , and it vanishes in @xmath344 .
the solutions of these boundary layer problems decay exponentially at infinity and the constants @xmath300 , @xmath336 , @xmath345 and @xmath337 are chosen from the conditions of existence of such solutions ( see @xcite ) .
let us define first @xmath346 from the condition of exponential decaying of @xmath326 at infinity:@xmath347 i.e. @xmath348 where the upper index @xmath1 corresponds to the first component of the vector .
then we find @xmath336 , and @xmath337 as defined in .
then we determine the constants @xmath349 from the condition of the exponential decaying of @xmath350 at infinity . to this end , consider first problem without the last term in equation @xmath351 , i.e.@xmath18{l}-\frac{\nu_{0}}{2}\delta_{\xi}\bar{u}_{l}^{blo_{0}}+\nabla_{\xi}\bar{p}_{l}^{blo_{0}}=\\ \ \ \ \
= \phi_{0}(\xi)\delta_{l,0}+\displaystyle{\sum_{e = e_{j};j=1,\dots , n}}\left\ { c_{l}^{e}\left ( \nu_{0}\delta_{\xi}\left ( \chi_{j}(\xi_{1}^{e})\right ) \gamma_{j}\left ( n_{1}(\xi_{2}^{e}),0\right ) ^{t}-\nabla_{\xi } \left ( \chi_{j}(\xi_{1}^{e})\xi_{1}^{e}\right ) \right ) \right\ } , \\ \\ \mathrm{div}_{\xi}\bar{u}_{l}^{blo_{0}}=-\displaystyle{\sum_{e = e_{j};j=1,\dots , n}}\mathrm{div}_{\xi}c_{l}^{e}\left ( \chi_{j}(\xi_{1}^{e})\gamma_{j}\left ( n_{1}(\xi_{2}^{e}),0\right ) ^{t}\right ) , \quad\xi \in\omega_{o_{0}},\\ \\ { \bar{u}_{l}^{blo_{0}}}{|_{\partial{\omega}_{o_{0}}}}=0 .
\end{array } \right . \label{12}\ ] ] here the constants @xmath300 are just defined by and and satisfy the condition @xmath352 i.e. @xmath353 indeed , the choice of constants @xmath354 and @xmath300 from and condition give relation .
it is known that there exists the unique solution @xmath355 of this problem such that @xmath356 stabilizes to zero at infinity on every branch of @xmath357 and @xmath358 stabilizes on every branch of @xmath357 associated to @xmath191 , to its own constant @xmath359 .
these constants are defined uniquely up to one common additive constant , which we fix here by a condition @xmath360 .
then we define@xmath18{l}{d}_{l+1}^{e_{j}}=-\bar{p}_{l}^{blo_{0}\infty j},\\ \\ u_{l}^{blo_{0}}=\bar{u}_{l}^{blo_{0}},\\ \\ { p}_{l}^{blo_{0}}=\bar{p}_{l}^{blo_{0}}+\displaystyle{\sum_{e = e_{j};j=1,\dots , n}}{d}_{l+1}^{e_{j}}\chi_{j}(\xi_{1}^{e_{j } } ) \end{array } \right .
\label{14}\ ] ] on every branch of @xmath357 , associated with @xmath191 , i.e. @xmath361 obviously , this pair @xmath362 satisfies .
the boundary layer functions @xmath363 and @xmath364 ; @xmath201 are not defined in the vicinity of @xmath188 . therefore we should change a little bit the formulas of @xmath365 and @xmath366 far from the nodes @xmath197 , @xmath201 . let @xmath367 be a smooth function defined on each segment @xmath191 , let it be one if @xmath368 and let it be zero if @xmath369 .
let @xmath370 for each rectangle @xmath195 and let @xmath371 on each @xmath200 .
set @xmath372 on @xmath200 and all half - rectangles having common points with @xmath200 and extend it by zero on the remaining part of @xmath195 .
then we define @xmath365 and @xmath366 as @xmath373@xmath11{lcl}\bar{p}^{a } & = & \displaystyle{\sum_{l=0}^{k}\varepsilon^{l+1}\left\ { \sum_{e = e_{j};j=0,1,\dots , n}p_{l}^{e}(x^{e , l})\chi_{\varepsilon}(x)+\sum _ { i=0}^{n}p_{l}^{blo_{i}}\left ( \frac{x - o_{i}}{\varepsilon}\right ) \eta^{(i)}(x)\right\ } } \\ & & + \displaystyle{\sum_{l=0}^{k}\varepsilon^{l}q_{l}^{e}(x_{1}^{e})\chi_{\varepsilon}(x)+\sum_{i=0}^{n}q_{0}^{e_{i}}(x_{1}^{e_{i}}=0)\left ( 1-\chi_{\varepsilon}(x)\right ) \theta_{i}(x).}\end{array } \label{16}\ ] ] let us mention that the last term ( sum ) in ( [ 16 ] ) corresponds to the boundary layer function @xmath327 for @xmath374 in this section we estimate the error between the exact solution and the asymptotic one . substituting the asymptotic expansions , , into , we get the relations @xmath18{ll}-\mathrm{div}(\nu(x){\mathcal{d}}\bar{u}^{a})+\nabla\bar{p}^{a}=f(x ) & \\ \ \ \ \ \ \ -\displaystyle{\sum_{e = e_{j};j=1,\dots , n}}\left\ { { \varepsilon } ^{k+1}\gamma_{j}\left ( f_{e_{j}}^{k}\right ) ^{t}\chi_{\varepsilon } ( x)-\varepsilon^{k}(\nabla_{\xi}\chi_{j}(\xi_{1}^{e_{j}})d_{k+1}^{e_{j}}\right . &
\\ \qquad\qquad\left . -\nabla_{\tilde{\xi}}\chi_{j}(\tilde{\xi}_{1}^{e_{j}}){d}_{k+1}^{e_{j}})\right\ } + \psi , & \text{in } { \mathcal{b}}^{\varepsilon},\\ \mathrm{div}\bar{u}^{a}=\psi & \text{in } { \mathcal{b}}^{\varepsilon},\\ \bar{u}^{a}=g & \text{on } \partial{\mathcal{b}}^{\varepsilon } , \end{array } \right .
\label{17}\ ] ] where @xmath375 is the residual described in , and @xmath376 is defined by @xmath377{ll}\displaystyle 0 & \mathrm{in}~{\mathcal{b}}^{\varepsilon}\cap\left\ { x^{e_{i}}_{1}<\beta\right\ } , \vspace{0.1cm}\\ \displaystyle-\nabla_{x}\eta_{i}(x_{1}^{e_{i } } ) .
u^{(blo_{0})}\left ( \frac{x - o_{0}}{\varepsilon}\right ) & \mathrm{in}~{\mathcal{b}}^{\varepsilon } \cap\left\ { \beta < x_{1}^{e_{i}}<\frac{3|e_{i}|}{8}\right\ } , \vspace{0.1cm}\\ 0 & \mathrm{in}~{\mathcal{b}}^{\varepsilon}\cap\left\ { \frac{3|e_{i}|}{8}<x_{1}^{e_{i}}<\frac{5|e_{i}|}{8}\right\ } , \vspace{0.1cm}\\ \displaystyle-\nabla_{x}\eta_{i}(x_{1}^{e_{i}}).u^{(blo_{i})}\left ( \frac{x - o_{i}}{\varepsilon}\right ) & \mathrm{in}~{\mathcal{b}}^{\varepsilon } \cap\left\ { \frac{5|e_{i}|}{8}<x_{1}^{e_{i}}<|e_{i}|\right\ } , \vspace { 0.1cm}\\ 0 & \mathrm{in}~\gamma_{i}^{\varepsilon},i=1,\dots , n,\vspace{0.1cm}\\ & \end{array } \right.\ ] ] @xmath378 , @xmath379 with a positive constant @xmath380 and @xmath381 because @xmath382 . the exponentially decaying residuals
@xmath383 and @xmath376 appear from the truncation of the boundary layer terms by the function @xmath264 : it is different from 1 in the part of the domain @xmath212 where @xmath384 , @xmath385 and their derivatives are exponentially small .
we are going to prove the estimate @xmath386 we can not apply directly the a priori estimates because @xmath387 is not divergence free . let us construct a function @xmath388 satisfying the following properties : @xmath18{ll}\displaystyle \hat{u}^{a}\in\left ( h_{0}^{1}({\mathcal{b}}^{\varepsilon } ) \right ) ^{2 } , & \vspace{0.1cm}\\ \displaystyle \mathrm{div}_{x } \hat{u}^{a}=\psi & \mathrm{in}\qquad { \mathcal{b}}^{\varepsilon},\vspace{0.1cm}\\ \hat{u}^{a}=0 & \mathrm{in}\quad{\mathcal{b}}^{\varepsilon}\cap\{x^{e}_{1}<\hat{d}_{0}\varepsilon\}\bigcup\gamma_{j}^{\varepsilon},j=1,\dots , n\vspace{0.1cm}\\ & \end{array } \right . \label{19}\ ] ] problem has at least one solution , satisfying @xmath389 _ proof . _ due to @xmath307 we can consider the problem as separate problem on each @xmath390 .
denote by @xmath391 the restriction of @xmath392 on @xmath390 , obviously @xmath393 for @xmath394 such that @xmath395 . for all @xmath396 such that @xmath397 introduce the new variable @xmath398 ; obviously @xmath399 .
define a new function @xmath400 , by @xmath401 .
obvious computations lead to the following problem for @xmath402 : @xmath18{ll}\displaystyle \mathrm{div}_{y } \mu_{\varepsilon}^{e_{j}}=\psi((|e_{j}|-\hat { d}_{0}\varepsilon)y_{1}^{e_{j}}+ \hat{d}_{0}\varepsilon,\varepsilon y_{2}^{e_{j } } ) & \mathrm{in}\qquad(0,1)\times(-\frac{1}{2},\frac{1}{2 } ) , \vspace{0.1cm}\\ \mu_{\varepsilon}^{e_{j}}=0 & \mathrm{on}\qquad\partial((0,1)\times(-\frac { 1}{2},\frac{1}{2 } ) ) , \vspace{0.1cm}\\ & \end{array } \right .
\label{20}\ ] ] applying the result of @xcite , chap .
iii , p. 127
we get : there exist a solution of [ 20 ] ) such that @xmath403 expressing the norm of @xmath404 with respect to the norm @xmath405 we obtain @xmath406 i.e. @xmath407 .
so , @xmath408 .
define @xmath409 .
then @xmath410 satisfies the following problem : @xmath18{ll}-\mathrm{div}(\nu(x){\mathcal{d}}u^{a})+\nabla\bar{p}^{a}=f(x)+ \mathrm{div}(\nu(x){\mathcal{d}}\hat{u}^{a})+\psi & \\
-\displaystyle{\sum_{e = e_{j};j=1,\dots , n}}\left\ { { \varepsilon}^{k+1}\gamma_{j}\left ( f_{e_{j}}^{k}\right ) ^{t}\chi_{\varepsilon}(x)-\varepsilon ^{k}(\nabla_{\xi}\chi_{j}(\xi_{1}^{e_{j}})d_{k+1}^{e_{j}}-\nabla_{\tilde{\xi}}\chi_{j}(\tilde{\xi}_{1}^{e_{j}})\tilde{d}_{k+1}^{e_{j}})\right\ } , & \text{in } { \mathcal{b}}^{\varepsilon},\\ \mathrm{div}\bar{u}^{a}=0 & \text{in } { \mathcal{b}}^{\varepsilon},\\ \bar{u}^{a}=g & \text{on } \partial{\mathcal{b}}^{\varepsilon } , \end{array } \right . \label{22}\ ] ] let @xmath411 be the asymptotic solution given by , and @xmath44 the solution of , the following estimates hold:@xmath18{l}\vert\bar{u}^{a}-u_{\varepsilon}\vert_{{h^{1}({\mathcal{b}}^{\varepsilon})}^{2}}=o\left ( \varepsilon^{k+\frac{1}{2}}\right ) , \\ \vert\nabla\bar{p}^{a}-\nabla p_{\varepsilon}\vert_{{h^{1}({\mathcal{b}}^{\varepsilon})}}=o\left ( \varepsilon^{k+\frac{1}{2}}\right ) . \end{array } \right.\ ] ] applying a priori estimates and we get @xmath412 and so , @xmath413 the estimate for the pressure is obtained from the a priori estimates .
these estimates justify the construction of the asymptotic expansion .
the main result can be easily generalized in case when the length of @xmath190 is different from 1 .
formula shows that only @xmath414 could be different from zero .
the same analysis can be provided in the case of a multi - bundle structure , that is the union of a finite number of thin domains of @xmath212 type ( see @xcite , section 4.5.2 ) , and in this case the constants @xmath415 should be determined from a system of linear algebraic equations ( see ( 4.5.43),(4.5.44 ) in @xcite ) .
we consider the stokes flow in a rectangular domain @xmath416 with @xmath417 : @xmath418{ll}\displaystyle-\mathrm{div}(\nu(x_{1})du)+\nabla p=0 & \displaystyle\text { in } ( 0,1)\times(0,\varepsilon)\\ \displaystyle\mathrm{div } u=0 & \displaystyle\text { in } ( 0,1)\times ( 0,\varepsilon)\\ \displaystyle u(0,x_{2})=\varepsilon^{2 } \frac{x_{2}}{\varepsilon}\left ( 1-\frac{x_{2}}{\varepsilon}\right ) & \displaystyle\text { on } ( 0,\varepsilon ) \\ & \\ \displaystyle u(1,x_{2})=\varepsilon^{2 } \frac{x_{2}}{\varepsilon}\left ( 1-\frac{x_{2}}{\varepsilon}\right ) & \displaystyle\text { on } ( 0,\varepsilon ) \\
\displaystyle u(x_{1},0)=u(x_{1},\varepsilon ) & \displaystyle\text { on } ( 0,1 ) \end{array } \right.\ ] ] here @xmath419 . from
we get for @xmath118 : @xmath18{l}\displaystyle\frac{\partial}{\partial x_{1}}\left ( \frac{1}{\nu(x_{1})}\frac{\partial q_{0}}{\partial x_{1}}\right ) = 0\\ \\ \displaystyle u_{0,1}=\varepsilon^{2}\frac{1}{\nu(x_{1})}\frac{\partial q_{0}}{\partial x_{1}}\frac{x_{2}}{\varepsilon}\left ( 1-\frac{x_{2}}{\varepsilon } \right ) \\ \displaystyle u_{0,2}=0\\ \displaystyle p_{0}=0 \end{array } \right.\ ] ] with the compatibility condition @xmath420 we have : @xmath421 since we assume that @xmath422 we get : @xmath18{l}\displaystyle q_{0}(x_{1})=-x_{1}(x_{1}+2)+3\\ \\
\displaystyle u_{0,1}=\varepsilon^{2}\frac{x_{2}}{\varepsilon}\left ( 1-\frac{x_{2}}{\varepsilon}\right ) \\
u_{0,2}=0\\ p_{0}=0 \end{array } \right.\ ] ] solving numerically the problem ( by comsol ) we get the following results : for the first component of the velocity we have : we see that the error is of order of @xmath2 et @xmath423 for the pressure and the velocity respectively.*2 .
* consider now the stokes problem in a t - shape domain @xmath424\times(0,\varepsilon)\cup(0,\varepsilon)\times ( -0.45,0.55)$ ] @xmath18{ll}\displaystyle u(1,x_{2})=g_{1}\left ( \frac{x_{2}}{\varepsilon}\right ) = \varepsilon^{2 } \frac{x_{2}}{\varepsilon}\left ( 1-\frac{x_{2}}{\varepsilon } \right ) & \text { on } ( 0,\varepsilon)\\ & \\
\displaystyle u(x_{1},-0.45)=g_{2}\left ( \frac{x_{1}}{\varepsilon}\right ) = \varepsilon^{2 } \frac{x_{1}}{\varepsilon}\left ( 1-\frac{x_{1}}{\varepsilon } \right ) & \text { on } ( 0,\varepsilon)\\ & \\
\displaystyle u(x_{1},-0.45)=g_{3}\left ( \frac{x_{1}}{\varepsilon}\right ) = 2\varepsilon^{2 } \frac{x_{1}}{\varepsilon}\left ( 1-\frac{x_{1}}{\varepsilon } \right ) & \text { on } ( 0,\varepsilon)\\ & \\ & \end{array } \right.\ ] ] as inflow / outflow conditions , @xmath417 , @xmath425 anywhere else on the boundary and @xmath426 .
@xmath427 are defined in such a way that @xmath23 is equal to @xmath428 in each rectangle and constant near the nodes .
the functions @xmath429 and @xmath430 have the following shape : * * acknowledgments .
* * the authors were partially supported by the following grants : `` strutture sottili '' of the program `` collaborazioni interuniversitarie internazionali '' ( 2004 - 2006 ) of the italian ministry of education , university and research ; sfr momad of the university of saint etienne and enise ( the ministry of the research and education of france ) , the joint french - russian pics cnrs grant `` mathematical modeling of blood diseases '' and by the grant no .
14.740.11.0875 `` multiscale problems : analysis and methods '' of the ministry of edication and research of russian federation . 99 blanc f. , gipouloux o. , panasenko g. , zine a.m. , _ asymptotic analysis and partial asymptotic decomposition of the domain for stokes equation in tube structure
_ , mathematical models and methods in applied sciences , 1999 , vol.9 , 9 , 1351 - 1378 .
cardone g. , panasenko g.p .
, sirakov y. , _ asymptotic analysis and numerical modeling of mass transport in tubular structures _ , mathematical models and methods in applied sciences ( m3as ) 20 , n. 4 ( 2010 ) 1 - 25 . | the stokes equation with the varying viscosity is considered in a thin tube structure , i.e. in a connected union of thin rectangles with heights of order @xmath0 and with bases of order @xmath1 with smoothened boundary .
an asymptotic expansion of the solution is constructed : it contains some poiseuille type flows in the channels ( rectangles ) with some boundary layers correctors in the neighborhoods of the bifurcations of the channels .
the estimates for the difference of the exact solution and its asymptotic approximation are proved .
keywords : stokes equation , thin tube structures , variable viscosity , asymptotic expansion , boundary layer correctors .
ams subject classification ( 2000 ) : 35b27 , 35q30 , 76m45 , 65n55 . |
in this paper , we study the quantum dynamics of two interacting oscillator systems in two different , but very well - known scenarios . in case one ,
both oscillators are assumed to be linear , and in case two , one oscillator is assumed to be linear , while the other oscillator is assumed to be non - linear ( angular momentum oscillator ) .
we study heisenberg equations and the time evolution of the wave - functions @xmath0 for both cases to discuss the quantum interactive behaviors between the two oscillators therein , e.g. , entanglement , spontaneous emission ( in case two ) , and quantum versus classical differences ( in case two ) .
entanglement , an implicit non - classical ingredient of interacting quantum systems , is the key to enabling quantum computation , which allows for the solution of certain classes of problems more efficiently than the classical counterpart ( see , e.g. , refs . ) . for case two , by setting the relevant commutator relation in each oscillator to one or zero , we utilize an interesting quantum - classical scheme @xcite to clearly characterize the quantum features in the system .
in order to investigate the time evolution of the wave - function and the notion of entanglement in case one , we adopt the coupled - boson representation,@xcite which was discussed by schwinger in terms of the correspondence between two one - dimensional linear oscillators and an angular momentum oscillator . based on this representation , we will easily obtain exact energy eigenvalues and eigenstates of case one .
the coupled - boson representation has been used in context with similar types of problems to construct an ideal model for quantum - mechanical interference@xcite and to study the current operator in two coupled linear oscillator ( the rate of exchange of occupation number between oscillators).@xcite we adopt here @xmath1 as an entanglement measure for the total system in a pure state,@xcite where @xmath2 is a reduced density matrix of individual oscillator @xmath3 .
we will show that this measure increases with the occupation number @xmath4 for a given initial state for case one , and with the maximum quantum number @xmath5 of the angular momentum oscillator for case two , respectively ; this increase with occupation number results from considering a dissipationless and coherent process , whereas we know that entanglement will diminish when loss is taken into account , e.g. , resulting from the ( thermal ) interaction between the system and its environment.@xcite for case two , we will discuss entanglement for the cases where the angular momentum oscillator is in the states @xmath6 explicitly . in section [ sec : linear_linear_osc ] , we study the interaction between two linear oscillators .
the heisenberg equations of motion are analyzed in the coupled boson representation , leading to the straightforward diagonalization of the coupled oscillator hamiltonian ; from this diagonalized hamiltonian , the exact eigenstates and eigenvalues of the coupled oscillator hamiltonian are obtained , thus allowing for the calculation of the reduced density matrix , and the subsequent study of the entanglement under the two initial conditions of simple product state ( a disentangled state ) and exact eigenstate for the coupled oscillator system ( an entangled state ) . in section [ sec : linear_non_linear ] , we study the interaction between a linear and a non - linear , angular momentum oscillator . we make use of the seminal work of ref . to analyze the quantum - mechanical and the classical dynamics of this interaction , and to connect the quantum manifestation of spontaneous emission in this system with the notion of entanglement .
further , the total system hamiltonian is diagonalized for @xmath7 using rotational generating operators of the group @xmath8.@xcite from this diagonalization , the reduced density matrix , and thus the measure of entanglement , is calculated for simple product ( disentangled ) initial states .
unlike the linear - linear oscillator system , this coupled system displays periodicity in the measure of entanglement for @xmath9 and @xmath10 , whereas for @xmath11 and beyond , the measure of entanglement is aperiodic ; this aperiodicity is shown to be a result of the complex structure of the diagonalized multi - level hamiltonian .
let us begin with case one .
the coupled oscillator system under investigation is described in the _ rotating wave approximation _ by the hamiltonian @xmath12 where @xmath13 denotes a coupling strength , and we assume that @xmath14 .
schwinger found that the quantum - mechanical angular momentum can be obtained by using creation and annihilation operators of two one - dimensional harmonic oscillators in terms of @xmath15 .
this is explicitly given as follows:@xcite @xmath16 where @xmath17 with @xmath18 is obviously an occupation number operator of each linear oscillator @xmath3 . from @xmath19
the operators @xmath20 , where @xmath21 , satisfy the usual angular - momentum commutation relations . the operator @xmath22 yields the quantum number of angular momentum , @xmath23 .
in terms of the angular momentum operators in ( [ eq : bosonic_rep ] ) the hamiltonian in ( [ eq : hamiltonian1 ] ) is rewritten as @xmath24 where @xmath25 .
since @xmath26 = 0 $ ] , we easily obtain a constant of motion @xmath22 with @xmath27 : @xmath28\ ; = \ ; 0\,,}\ ] ] which means that the total oscillator occupation number is conserved . also , we find that @xmath29 and so @xmath30 is a constant of motion for @xmath31 . with the aid of the identity , @xmath32\ , + \ , \frac{\vartheta^2}{2!}\ , [ \hat{u},\,[\hat{u},\,\hat{v}]]\ , + \ , \frac{\vartheta^3}{3!}\ , [ \hat{u},\,[\hat{u},\,[\hat{u},\,\hat{v}]]]\ , + \ , \cdots\,,}\ ] ] we evaluate @xmath33 , where @xmath34 , to arrive at the diagonalized hamiltonian of eq .
( [ eq : coupled_boson_hamiltonian ] ) as @xmath35 where @xmath36 ( note that in case of @xmath37 , we have @xmath38 for any non - zero @xmath13 ) . from eqs .
( [ eq : bosonic_rep ] ) , ( [ eq : diagonal_hamiltonian ] ) we now obtain two renormalized uncoupled harmonic oscillators with the renormalized energy splitting for each oscillator , @xmath39 where @xmath40 , @xmath41 .
here , @xmath42 , and @xmath43 explicitly reads @xmath44 where the rotation matrix is @xmath45 thus , from the schwinger approach , we realize that the coupling term in eq .
( [ eq : hamiltonian1 ] ) can be equivalently represented as the @xmath30 component of angular momentum , and therefore , @xmath46 in eq .
( [ eq : hamiltonian1 ] ) can be diagonalized by a unitary rotational operator with respect to the angular momentum operator @xmath47 through @xmath48.@xcite then , using eqs .
( [ eq : uncoupled_hamiltonian ] ) , ( [ eq : rotation_mat ] ) we easily obtain the original annihilation operators as @xmath49 where @xmath50 but @xmath51 . here
, neglecting the `` non - local '' observable @xmath52 , we obviously have @xmath53 \ne 1 $ ] at @xmath54 .
now , we easily acquire the exact eigenstates and eigenvalues of the hamiltonian in ( [ eq : hamiltonian1 ] ) as @xmath55 we see here that for @xmath56 ( i.e. , @xmath31 ) , the eigenstates @xmath57 will also be eigenstates of @xmath58 .
let us now study the entanglement in the time evolution of the two coupled linear oscillators , @xmath59 . to this end , we adopt,@xcite as a measure of entanglement , @xmath60\,,}\ ] ] where @xmath61 \equiv \mbox{tr}_{\nu}\ , ( \hat{\rho}^{(\nu)})^2 $ ] is the purity measure of the state of each oscillator @xmath3 . here , it clearly holds that @xmath62 = \mathcal{p}[\hat{\rho}^{(2)}]$].@xcite as can also easily be verified , @xmath63 $ ] takes on a maximum value of @xmath64 if the oscillator @xmath3 is in a pure state ; it then follows that @xmath65 for a simple product ( disentangled ) state , latexmath:[${\textstyle \left|\psi\right\ > = @xmath63 $ ] takes on its minimum value @xmath67 , where @xmath68 is a number of accessible orthogonal states @xmath69 of each oscillator @xmath3 , if the oscillator @xmath3 is in the maximally mixed state given by @xmath70 ; then , the corresponding total wave - function @xmath71 is a maximally entangled state in form of @xmath72 , and @xmath73 has its maximum value @xmath74 .
we will explicitly determine @xmath75 in its time evolution with mathematically manageable initial states @xmath76 given below .
we obtain the reduced density matrix , @xmath77 , from the system state evolution , @xmath78 , by constructing the system density matrix @xmath79 from the known state of the system , and then taking the trace of @xmath80 with respect to all subsystem matrix elements within the system but those of subsystem @xmath3 . for our oscillator system under study ,
the state evolution is given by @xmath81 where @xmath82 .
therefore , in constructing @xmath83 from ( [ eq : schroedinger_pic_psi1 ] ) , and taking the trace of @xmath84 with respect to oscillator @xmath85 , we formally acquire the reduced density matrix for oscillator @xmath64 as @xmath86 where @xmath87 . using @xmath78 in eq .
( [ eq : schroedinger_pic_psi1 ] ) , we explicitly obtain @xmath88 where @xmath89 , and @xmath90 from eq .
( [ eq : eigenstates ] ) has been used in eq .
( [ eq : schroedinger_pic_psi1 ] ) to obtain the explicit result .
now , in transforming eq .
( [ eq : reduced_density_mat1 ] ) to the coupled - boson representation discussed in eq .
( [ eq : bosonic_rep ] ) , we relabel @xmath91 and @xmath92 in terms of the eigenstates of @xmath93 , namely , @xmath94 and @xmath95 , respectively , where @xmath96 and @xmath97 . since the rotation with respect to @xmath98 conserves the quantum number @xmath5 as noted in eq .
( [ eq : equation_of_motion1 ] ) , we also find that @xmath99 taking into account that @xmath100 and @xmath101 for a given @xmath102 , respectively , it follows that eq .
( [ eq : reduced_density_mat1 ] ) can be reexpressed as @xmath103 where @xmath104 and @xmath105 , respectively . then , in defining @xmath106 and @xmath107 we relabel @xmath108 in terms of @xmath109 , where @xmath110 and @xmath111 , respectively .
this clearly shows that two coupled harmonic oscillators can naturally be treated as an angular momentum oscillator .
we now make use of the coupled - boson algebra to evaluate the reduced density matrix @xmath112 below for two physically interesting , transparent , and mathematically manageable initial states , @xmath76 , to illustrate the entanglement measure @xmath75 , namely , the case of a simple product state , @xmath113 ( case i ) , and the case of an eigenstate of the exact hamiltonian in eq .
( [ eq : coupled_boson_hamiltonian ] ) , @xmath114 of eq .
( [ eq : eigenstates ] ) ( case ii ) . in terms of physical interest , the simple product state in case
i is an initial condition which inherently assumes no entanglement at time zero , whereas case ii , using the exact eigenstate of the hamiltonian , implicitly includes entanglement .
let us begin with the case i : @xmath115 . in the coupled - boson representation , this state becomes @xmath116 where @xmath117 and @xmath118 .
we now have @xmath119 and @xmath120 in eq .
( [ eq : reduced_density_mat2 ] ) which then reduces to the diagonal form , @xmath121 where @xmath122 results from eq .
( [ eq : constraints ] ) with @xmath123 . in eq .
( [ eq : reduced_density_mat3 ] ) , all the matrix elements can be written in terms of @xmath124 with ( @xmath125 ) and ( @xmath126 ) , where @xmath127 is formally explicitly expressed as @xcite @xmath128 here we see that @xmath129 , and @xmath130 . particularly for the case that @xmath131 , this reduces further to @xmath132 which will be used later .
equivalently , @xmath127 can be expressed in terms of the tabulated jacobi polynomial @xcite @xmath133 in the form @xmath134 reexpressing @xmath109 in eq .
( [ eq : reduced_density_mat3 ] ) in terms of @xmath127 , we obtain @xmath135 where @xmath136 with the property that @xmath137 [ here , we used the relation that @xmath138 ; @xmath139 can be explicitly evaluated for any set of the harmonic oscillator initial condition numbers , @xmath140 .
we refer the reader to the appendix [ sec : appendix1 ] for the explicit expression of eq .
( [ eq : f_function ] ) obtained from the substitution of eq .
( [ eq : rotation_mat3 ] ) , which will be useful below . from ( [ eq : reduced_density_mat3_1 ] ) we obtain the measure of entanglement given by @xmath141 we note generally that @xmath142 in eq .
( [ eq : f_function ] ) is a time - dependent oscillatory function with a modal distribution coefficient @xmath143 governing the oscillatory strength .
this clearly influences the measure of entanglement as noted in eq .
( [ eq : entanglement_measure_case_1 ] ) and as seen in fig .
[ fig : reduced_mat_time_evolution ] for specific values of @xmath144 .
let us now apply eqs .
( [ eq : reduced_density_mat3_1 ] ) and ( [ eq : entanglement_measure_case_1 ] ) for specific values of @xmath144 . for @xmath145 , the total occupation number @xmath146
, there are two possible simple product states @xmath147 describing this initial condition .
we can easily evaluate @xmath148 : for @xmath149 , from eqs .
( [ eq : rotation_mat2 ] ) and ( [ eq : f_function ] ) we obtain @xmath150 , and then the @xmath151 diagonal matrix from eq . ( [ eq : reduced_density_mat3_1 ] ) , @xmath152 similarly , for @xmath153 , we acquire the diagonal matrix @xmath154 whose elements are exchanged from eq .
( [ eq : reduced_density_mat4 ] ) . due to the fact that @xmath155 = \mathcal{p}[\hat{\rho}_2^{(1)}(t ) ] \leq 1 $ ] ,
the reduced density matrices , @xmath156 and @xmath154 , represent mixed states , respectively , with both directly reflecting the appearance of entanglement between the two coupled oscillators ; this is noted in fig .
[ fig : reduced_mat_time_evolution ] with the properties of @xmath157 and @xmath158 as given in eq .
( [ eq : entanglement_measure_case_1 ] ) .
this appearance of entanglement in the time evolution from the disentangled initial product state for each case is clearly attributed to the interaction hamiltonian in ( [ eq : hamiltonian1 ] ) , i.e. , @xmath159 . likewise , for @xmath160 ( or @xmath161 ) , we also obtain the entanglement in the time evolution resulting from the interaction hamiltonian . due to their complex forms of eqs .
( [ eq : reduced_density_mat3_1 ] ) and ( [ eq : entanglement_measure_case_1 ] ) [ see also ( [ eq : appendix_1 ] ) and ( [ eq : appendix_2 ] ) ] , respectively , we simply plot the exact numerical results of @xmath75 for @xmath162 with latexmath:[$|\psi(0)\ > = clearly see that @xmath75 increases with @xmath164 ( or @xmath144 ) .
two points deserve comment here .
first , we see that for a given total occupation number @xmath165 [ constant of motion ; cf .
( [ eq : equation_of_motion1 ] ) ] , @xmath75 is periodic ( see fig . [
fig : reduced_mat_time_evolution ] ) , as is the reduced density matrix @xmath148 due to the clear oscillator dependence on time as seen in @xmath142 in ( [ eq : f_function ] ) ; on the other hand , this is not the case for a single oscillator coupled with a thermal reservoir modeled by a sea of infinite oscillators @xcite , for then , dissipative decay of its occupation number into the the reservoir is effective .
second , in calculating @xmath166 and comparing results to the corresponding expressions in ( [ eq : schroedinger_pic_psi2 ] ) - ( [ eq : reduced_density_mat3 ] ) , one easily arrives at the fact that @xmath167 for any @xmath144 , which confirms the relationship valid for any system state @xmath78 that @xmath168 = \mathcal{p}[\hat{\rho}^{(1)}]$ ] , thus indicating the equality of mixture .
next , we consider case ii , in which the initial state is given by an eigenstate of the exact hamiltonian in eq .
( [ eq : coupled_boson_hamiltonian ] ) , @xmath169 with @xmath117 and @xmath118 . from eq .
( [ eq : reduced_density_mat2 ] ) , we easily obtain the reduced density matrix and the measure of entanglement , @xmath170 where @xmath171 here , @xmath172 is time - independent , whereas @xmath139 from eq .
( [ eq : f_function ] ) is not as obtained in the previous case .
also , we can easily show that @xmath167 . then , as in eq .
( [ eq : entanglement_measure_case_1 ] ) , we have @xmath173 particularly for @xmath174 , namely @xmath175 , from eq . ( [ eq : rotation_mat2 ] ) with @xmath176 , we easily find that @xmath177 where @xmath178 .
therefore , from eq .
( [ eq : eigen_initial_state_reduced_density_mat1 ] ) we arrive at the fact that the occupation number @xmath179 in oscillator @xmath64 is described by the binomial distribution @xmath180 with a trial probability @xmath181 , which remains unchanged in the time evolution . along the same line , for oscillator @xmath85
, we also have @xmath182 and , from eq .
( [ eq : eigen_initial_state_reduced_density_mat1 ] ) , the binomial distribution @xmath183 with @xmath184 for the occupation number @xmath179 in oscillator 2 .
these reduced density matrices , @xmath185 and @xmath186 , represent mixed states for @xmath187 , respectively , due to the fact that @xmath188 = \mathcal{p}[\hat{\rho}_2^{(1 ) } ] \leq 1 $ ] , thus indicating the entanglement between the two linear oscillators , @xmath189 obtained from eq .
( [ eq : entanglement_measure_case_2 ] ) .
furthermore , as in the previous case , @xmath75 increases with @xmath144 ( see fig .
[ fig : reduced_mat_time_evolution2 ] ) ; it is noted in the figure that @xmath190 as @xmath191 .
the second coupled system under investigation consists of a linear oscillator and a ( non - linear ) angular momentum oscillator .
this system approximates the interaction between a field mode and an atomic @xmath164-level system under idealized conditions resulting from neglecting dissipation and the coupling between other surrounding atomic systems . following the seminal work of ref . , which considered the analysis quantum - mechanically and also classically to compare one case with the other systematically , it is convenient to employ , for the linear oscillator , dimensionless coordinate @xmath192 and momentum @xmath193 , where @xmath194 and @xmath195 , satisfying @xmath196_{\mathsf{p } } = i$ ] , and , for the angular momentum oscillator , dimensionless angular momentum variables @xmath197 with @xmath198_{\mathsf{p } } = i\,\epsilon_{rst}\,{\mathtt{j}}_t$ ] , where the brackets @xmath199_{\mathsf{p}}$ ] represent the commutators for the quantum - mechanical description , and @xmath200 times the poisson brackets for the classical description , respectively .
the hamiltonians of the two individual oscillators are then expressed as @xmath201 respectively , and the interaction hamiltonian is given in the _ rotating wave approximation _ by @xmath202 leading to the total hamiltonian @xmath203 here , we have @xmath204 ( classically , hermitian conjugation corresponds to complex conjugation ) .
these non - hermitian variables and @xmath205 clearly obey @xmath206_{\mathsf{p}}\ ; = \ ; 1\,,\ ; \left[\left(\mathtt{j}_{+}\right)_2,\,\left(\mathtt{j}_{-}\right)_2\right]_{\mathsf{p}}\ ; = \ ; \left(\mathtt{j}_z\right)_2\,,}\ ; \left[\left(\mathtt{j}_{+}\right)_2,\,\left(\mathtt{j}_z\right)_2\right]_{\mathsf{p}}\ ; = \ ; -\left(\mathtt{j}_{+}\right)_2\,,\ ; \left[\left(\mathtt{j}_{-}\right)_2,\,\left(\mathtt{j}_z\right)_2\right]_{\mathsf{p}}\ ; = \ ; \left(\mathtt{j}_{-}\right)_2\ ] ] with all other brackets vanishing . from now on , let us restrict ourselves to the resonant case , @xmath207 . in order to make later calculations simpler , we then introduce the reduced variables@xcite specified , with the aid of the identity ( [ eq : identity_1 ] ) [ for the classical description , the commutators therein have to be obviously replaced by @xmath200 times poisson brackets ] and the rules in ( [ eq : commutator_rel1 ] ) , by @xmath208 and , similarly , @xmath209 respectively .
these reduced variables are easily seen to satisfy the same bracket relations as the corresponding unreduced variables , namely , @xmath210_{\mathsf{p}}\ ; = \ ; 1\;,\ ; \left[\left(j_{+}\right)_2,\,\left(j_{-}\right)_2\right]_{\mathsf{p}}\ ; = \ ; \left(j_z\right)_2\;,}\ ; \left[\left(j_{+}\right)_2,\,\left(j_z\right)_2\right]_{\mathsf{p}}\ ; = \ ; -\left(j_{+}\right)_2\;,\ ; \left[\left(j_{-}\right)_2,\,\left(j_z\right)_2\right]_{\mathsf{p}}\ ; = \ ; \left(j_{-}\right)_2\,.\ ] ] the interaction hamiltonian in ( [ eq : int_hamiltonian1 ] ) is now rewritten as @xmath211 and the equations of motion in terms of the reduced variables will be given by @xmath212_{\mathsf{p}}\,,\ ] ] where @xmath213 stands for any of the above reduced variables . from eqs .
( [ eq : commutator_rel2 ] ) , ( [ eq : eq_of_motion1 ] ) it follows that @xmath214 and then @xmath215 which , as shown previously,@xcite hold both quantum - mechanically in the heisenberg picture and classically on the basis of the poisson bracket in phase space . from the equation of motion in eq .
( [ eq : eq_of_motion1 ] ) it turns out that the sum of the ( dimensionless ) energy of both oscillators , @xmath216 with @xmath217 ( without the zero - point energy of the linear oscillator in the quantum - mechanical description ) , and the interaction hamiltonian @xmath218 are constants of motion , respectively , which are determined by a given initial state .
as well , the square of a given total angular momentum of the angular momentum oscillator , @xmath219 , is also a constant of motion .
then , substituting the expressions @xmath220\;,\;\ ; 2\ , \left(j_{-}\right)_2\,\left(j_{+}\right)_2\ ; = \ ; \mathbf{j}^2\ , -\ , \left(j_z\right)_2 ^ 2\ , -\ , \left[\left(j_{+}\right)_2,\ , \left(j_{-}\right)_2\right]\ ] ] ( note that there are no subscripts @xmath221 of the brackets ) into eq .
( [ eq : eq_of_motion3 ] ) , we obtain @xmath222\right)\ , \left(j_z\right)_2\ , + \ , \left[\left(j_{+}\right)_2,\ , \left(j_{-}\right)_2\right]\ , -\
, { \mathbf{j}}^2\right\}\,.\ ] ] let us introduce the notation @xmath223\ , = \
, \lambda_1\;,\;\ ; \left[\left(j_{+}\right)_2,\ , \left(j_{-}\right)_2\right]\ , = \ , \lambda_2\ , \left(j_z\right)_2\,,\ ] ] where @xmath224 with @xmath225 correspond to the quantum - mechanical and the classical description of each oscillator , respectively .
( [ eq : eq_of_motion4 ] ) is now reduced to @xmath226 where the constant of motion @xmath227 with @xmath5 being classically ( @xmath228 ) the total ( continuous ) angular momentum and quantum - mechanically ( @xmath229 ) the corresponding quantum number . here
, @xmath230 is , clearly , an operator for @xmath231 , with the property @xmath232 in general . eq .
( [ eq : eq_of_motion5 ] ) is a non - linear differential equation for @xmath230 in both the classical and quantum - mechanical description .
since @xmath233_{\mathsf{p } }
= [ \dot{\left(j_z\right)}_2 , { \mathtt{e}}]_{\mathsf{p } } = 0 $ ] ( note that @xmath216 ) , it follows that in the quantum - mechanical consideration of eq .
( [ eq : eq_of_motion5 ] ) the constant of motion @xmath234 can be treated as a @xmath235-number .
we now consider eq .
( [ eq : eq_of_motion5 ] ) with two interesting initial states .
first , for the ground state with @xmath236 and @xmath237 , then @xmath238 ; it immediately follows that eq .
( [ eq : eq_of_motion5 ] ) , at @xmath239 , becomes @xmath240 quantum - mechanically , this state corresponds obviously to @xmath241 .
from the fact that each oscillator with the minimum energy can not give up further energy , we necessarily have @xmath242 here , which indicates that only @xmath243 simultaneously are physically allowed . for the quantum description , the zero - point fluctuation @xmath244 would yield a positive force on the angular momentum oscillator , and
the fluctuation @xmath229 of the angular momentum oscillator would lead to energy transfer into the linear oscillator ; however , the two formal processes are always cancelled such that they are not part of the real ( or measurable ) physical processes of energy transfer.@xcite therefore , the `` semiclassical '' option with @xmath245 and @xmath229 , and vice versa , is not physically admissible for the ground state @xmath246 , and is thus inappropriate for the system analysis .
another initial state under investigation is the state @xmath247 with @xmath236 and @xmath248 , quantum - mechanically @xmath249 .
then , eq . ( [ eq : eq_of_motion5 ] ) , at @xmath239 , reduces to @xmath250 independent of @xmath251 !
this demonstrates that only for the quantum description of the linear oscillator ( @xmath252 ) is spontaneous emission available , which arises from the zero - point fluctuation .
interestingly , it points out that for @xmath253 , there exists an unstable equilibrium leading to the absence of spontaneous emission ; the unstable equilibrium in the classical description ( @xmath254 ) and the presence of spontaneous emission for @xmath252 were thoroughly discussed in ref . . here
, we connect the spontaneous emission process with the notion of entanglement . in considering the time evolution of the wave - function for the spontaneous emission in the schrdinger picture , @xmath255 with the hamiltonian @xmath256 in eq .
( [ eq : total_hamiltonian ] ) , we now necessarily have the entangled state for @xmath257 , i.e. , early time beyond the initial time , which is given by @xmath258 this entanglement is explicitly shown in a compact form for the two - level case @xmath259 later in eq .
( [ eq:2_level_schroedinger1 ] ) and in the more complex form for the three - level case @xmath260 in ( [ eq:3_level_density_mat ] ) .
therefore , for the quantum description @xmath261 , the entanglement is always temporally present in the spontaneous emission process . here ,
interestingly enough , we still have the spontaneous emission for large enough @xmath5 , which immediately leads to the fact that the entanglement persists even as the _ coupled _ angular momentum oscillator is taken to the limit of a large number of levels , a limit which would go over to the classical limit for an _ uncoupled _ angular momentum oscillator .
in essence , the state @xmath262 of the uppermost excited state of a @xmath5-level angular momentum oscillator is always coupled to the vacuum state of the linear oscillator . from eqs .
( [ eq : int_hamiltonian2 ] ) , ( [ eq : eq_of_motion2 ] ) we also obtain @xmath263 where @xmath264 , and subsequently , with the aid of eqs . ( [ eq : qm_commutator_rel ] ) , ( [ eq : lambda1_2 ] ) , the expression @xmath265\ , + \ , \left(\mathtt{e}\ , + \ , \frac{\lambda_1}{2}\right ) \lambda_1\ , \lambda_2\ , \kappa^2\ , -\ , k^2\,.}\ ] ] in ref .
, it was indicated that in the classical description ( @xmath254 ) eq . ( [ eq : differential_eq_for_j3_dot_square2 ] ) has the form of an equation for the vertical position @xmath266 of a classical spherical pendulum , where the constant of motion @xmath267 represents an angular momentum about the vertical axis through the center .
it turns out that for @xmath236 and @xmath248 we get @xmath268 from ( [ eq : int_hamiltonian2 ] ) with @xmath269 and @xmath270 ( note that in the quantum description the constant of motion @xmath271 ) and then @xmath272 from ( [ eq : differential_eq_for_j3_dot_square2 ] ) . with @xmath273 in eq .
( [ eq : eq_of_motion8 ] ) , this reveals that the initial state @xmath274 corresponds to an unstable equilibrium of the pendulum . here
, it is also worthwhile pointing out here that the semiclassical treatment of jaynes - cummings @xcite reproduces this spontaneous emission for @xmath9 only , with the same decay rate as that given by the quantum - mechanical description ; this is accomplished by adding a phenomenological damping term to the classical equation of motion of the linear oscillator which would result from the interaction with the instantaneous expectation value of the dipole moment of the angular - momentum oscillator .
therefore , the jaynes - cummings model just provides an `` artificial '' picture of the spontaneous emission without describing the actual physical processes involved therein ( e.g. , entanglement ) , and also without offering a direct way for its extension to more than the two - level case @xmath275 . in order to study the appearance of the entanglement between the linear and the angular momentum oscillators in the quantum description , we discuss the state evolution of the total system in the schrdinger picture given by @xmath276 from eqs .
( [ eq : hamiltonian_of_two_osc ] ) - ( [ eq : total_hamiltonian ] ) with the reduced variables in eqs .
( [ eq : reduced_variables1 ] ) - ( [ eq : reduced_variables2 ] ) for the resonant case , @xmath277
. we will explicitly consider below the measure of entanglement for the cases of @xmath278 of the angular momentum oscillator . to this end , as will be seen , the @xmath279 irreducible matrix representation for the total system is employed , which is also systematically applicable for @xmath280 . here
, we assume that the given total energy ( a constant of motion ) is given by @xmath281 .
we first consider the case of @xmath9 , namely , a two - level angular momentum oscillator interacting with a linear oscillator . for the total energy @xmath282 ,
the basis @xmath283 allows us to rewrite eq .
( [ eq : hamiltonian_resonance_case2 ] ) in explicit operator form@xcite as @xmath284 this can be written in a @xmath151 matrix form as @xmath285 where @xmath286 denotes the pauli matrix . using the identity , @xmath287 with @xmath288 , we arrive at the diagonalized hamiltonian for eq .
( [ eq : hamiltonian_2_levels ] ) as @xmath289 accordingly , we acquire the energy eigenstates as @xmath290 with the corresponding eigenvalues , @xmath291 and @xmath292 , respectively . then , just as in discussion related to eq .
( [ eq : schroedinger_pic_psi1 ] ) for the two coupled linear oscillators , the state evolution for the initial state @xmath293 is given by @xmath294 this state explicitly includes the possibility of spontaneous emission ( @xmath236 ) . from eq .
( [ eq:2_level_schroedinger1 ] ) , we easily acquire the reduced density matrix of the angular momentum oscillator as @xmath295 which represents a mixed state .
therefore , it immediately follows that @xmath296\ ; = \ ; \sin^4\left(\frac{\kappa t}{\sqrt{2 } } \sqrt{n_1 + 1}\right)\ , + \ , \cos^4\left(\frac{\kappa t}{\sqrt{2 } } \sqrt{n_1 + 1}\right)}\,,\ ] ] with @xmath297 $ ] which oscillates between @xmath298 and @xmath299 in the time evolution , thus indicating the appearance of entanglement between the linear and the angular momentum oscillators ( see fig .
[ fig : reduced_mat ] ) .
furthermore , from eq .
( [ eq : density_matrix_2_level_non_linear_osc ] ) we find , after a minor calculation , that @xmath300 therefore , from eqs .
( [ eq : eq_of_motion5 ] ) , ( [ eq : j_z_time_evolution ] ) , it follows that @xmath301 which obviously yields the consistent result with that in eq .
( [ eq : eq_of_motion8 ] ) for the case that @xmath302 and @xmath303 .
next , we consider the case of @xmath10 .
for the total energy @xmath304 , from eq .
( [ eq : hamiltonian_resonance_case2 ] ) with the basis @xmath305 , the hamiltonian takes up the @xmath306 irreducible matrix representation @xmath307 based on the generating operators of the group @xmath8,@xcite where @xmath308 ( see appendix [ sec : appendix2 ] ) , the hamiltonian in eq .
( [ eq : hamiltonian_3_level_matrix ] ) can be reexpressed as @xmath309 where @xmath310 and @xmath311 are generating operators of @xmath8 , and their matrix forms are explicitly given in eq .
( [ eq : su_n_algebra ] ) . here , it is interesting to note that for a large @xmath312 such that @xmath313 , the hamiltonian in eq .
( [ eq : hamiltonian_3_level_matrix ] ) can be transformed approximately to @xmath314 which formally corresponds to the exact hamiltonian for @xmath9 in eq .
( [ eq : hamiltonian_2_levels ] ) with @xmath315 .
thus , for a large @xmath316 , we have an approximate diagonalized representation @xmath317 for @xmath10 , based on the previous result . for the case of general @xmath316 , proceeding from eq .
( [ eq : hamiltonian_3_level_matrix ] ) , we easily obtain the energy eigenstates with respect to the basis @xmath318 as @xmath319 and the corresponding energy eigenvalues , @xmath320 respectively .
therefore , the diagonalized hamiltonian is given by @xmath321 in order to diagonalize the hamiltonian @xmath46 in eq .
( [ eq : hamiltonian_resonance_case2 ] ) for higher manifolds of @xmath5 , it is instructive to systematically analyze the above diagonalization of @xmath46 and confirm for @xmath322 . in seeking the rotational operation that diagonalizes @xmath46 in eq .
( [ eq : hamiltonian_3_level ] ) to achieve @xmath317 , it is necessary to consider the unitary , euler - like transformation @xmath323 comprising three sequential , non - commuting rotations with respect to @xmath324 defined in ( [ eq : su_n_algebra2 ] ) , with euler angles @xmath325 .
then , as previously accomplished for the case of @xmath9 in eqs .
( [ eq : hamiltonian_2_levels ] ) - ( [ eq:2_level_1 ] ) , here we similarly choose @xmath326 to obtain @xmath317 in eq .
( [ eq : hamiltonian_3_level_diagonal ] ) .
this clearly shows that the diagonalization of the hamiltonian in eq .
( [ eq : hamiltonian_3_level ] ) can not be described by a simple rotation such as @xmath327 of an angular momentum oscillator,@xcite whereas for @xmath9 , the rotation @xmath328 was used in eq .
( [ eq : identities1 ] ) to diagonalize @xmath46 in eq .
( [ eq : hamiltonian_2_levels ] ) . however , for the case of @xmath10 and large @xmath312 given in eq .
( [ eq : approximate_hamiltonian ] ) , the diagonalization can be approximately accomplished by @xmath329 .
finally , it can be shown by induction , for general @xmath5 , that the diagonalization of the hamiltonian in eq .
( [ eq : hamiltonian_resonance_case2 ] ) is generally characterized by @xmath330 rotation angles @xmath331 with respect to the set @xmath332 as given in eq .
( [ eq : su_n_generators ] ) , where @xmath308 .
let us now consider the state evolution with the initial condition @xmath333 ; just as in eq .
( [ eq : schroedinger_pic_psi1 ] ) , from eqs .
( [ eq:3_eigenstates ] ) , ( [ eq:3_eigenvalues ] ) we find that @xmath334 from this state , we obtain , after some calculations , the reduced density matrix of the angular momentum oscillator , @xmath335\,.}\end{aligned}\ ] ] this reduced density matrix represents a mixed state with @xmath168 \leq 1 $ ] and @xmath336 , directly reflecting the entanglement of the total state @xmath78 [ here , eq .
( [ eq:3_level_density_mat ] ) for @xmath10 is comparable to eq .
( [ eq : density_matrix_2_level_non_linear_osc ] ) for @xmath9 ] .
furthermore , by noting that @xmath338 for a given @xmath339 with the diagonalized form @xmath317 in eq .
( [ eq : hamiltonian_3_level_diagonal ] ) , we can easily show that the state evolution in eq .
( [ eq:3_level_schroedinger1 ] ) displays periodicity with period @xmath340 as noted in fig .
[ fig : reduced_mat ] .
this means that the time evolution operator @xmath341 , expressed here with respect to the basis @xmath342 given in appendix [ sec : appendix2 ] , as @xmath343 where @xmath344 and @xmath345 for all @xmath346 , has the periodic property that @xmath347 .
let us consider the case of @xmath11 , where the hamiltonian of eq .
( [ eq : hamiltonian_resonance_case2 ] ) for the total energy @xmath348 is given by @xmath349 the diagonalized form of eq .
( [ eq : hamiltonian_4_level ] ) can be obtained , after tedious calculations similar to those described in the previous case , as @xmath350 where @xmath351 , and the @xmath352 s with @xmath353 denote energy eigenvectors .
due to the complex form of the reduced density matrix of the angular momentum oscillator , @xmath354 $ ] , obtained from @xmath355 with @xmath356 , we simply plot the exact numerical results of @xmath75 in figs . [
fig : reduced_mat4_1 ] , [ fig : reduced_mat4_2 ] .
based on the numerical analyses@xcite and the statistical approximation in terms of classical random variables@xcite , it has been shown that @xmath357 with @xmath358 exhibits aperiodic behavior in its time evolution for all @xmath359 , even for @xmath5 large enough ( i.e. , in the classical limit for an uncoupled angular momentum oscillator ) , while the classical counterpart displays periodic motion on the spherical pendulum [ cf .
( [ eq : differential_eq_for_j3_dot_square2 ] ) with @xmath360 and @xmath361 ; note that for this initial state , the angular momentum @xmath267 about the vertical axis through the center vanishes ] .
this shows that the aperiodicity in the time evolution would be of non - classical origin .
we indicate this aperiodicity for any @xmath362 here by noting the impossibility of having the periodic property @xmath363 for arbitrary @xmath339 ; from eq .
( [ eq : hamiltonian_4_level_diagonal ] ) , we find , after some calculations , that @xmath364 since here the phase factors , @xmath365 and @xmath366 , are incommensurable with respect to each other , we immediately note that not all @xmath367 in the expression @xmath368 can simultaneously vanish periodically , which leads to the aperiodicity in the state evolution @xmath369 ( see figs .
[ fig : reduced_mat4_1 ] , [ fig : reduced_mat4_2 ] ) . along the same line
, it can be shown that we have the diagonalized hamiltonian for any @xmath5 in form @xmath370 where @xmath371 , and @xmath372 with @xmath373 ; @xmath374 s denote energy eigenvectors . here ,
each @xmath375 is , in general , incommensurable with any other @xmath376 , where @xmath377 , and for the case that @xmath4 is odd , one of @xmath378 s always vanishes .
therefore , we have @xmath379 thus showing that @xmath380 for any @xmath339 unless all @xmath381 are commensurable with each other . from this , we easily find that the aperiodic behavior in the state evolution , @xmath382 , will survive and even increase with @xmath383 , i.e. , @xmath359 [ note that for @xmath10 , we obviously have @xmath384 only , which leads to the periodicity in the state evolution : @xmath385 for the case that @xmath386 .
in summary , we have investigated the fundamental dynamics of two interacting oscillators : in one scenario , two linear oscillators , and in the other scenario , a linear and a non - linear oscillator have been considered .
for the first scenario , based on the coupled boson representation we compactly described the equation of motion in the heisenberg picture and also systematically studied the quantum entanglement in the time evolution of the total wave - function @xmath78 for various initial states in the schrdinger picture . from this
, we found that the appearance of entanglement in the time evolution from the disentangled initial simple product state is attributed to the interaction between the two individual oscillators . also , the measure of entanglement increases with the total occupation number .
for the second scenario , quantum versus classical behaviors have been studied , based on the heisenberg equation developed in terms of relevant reduced kinematics operator variables and parameterized commutator relations . by setting the corresponding commutator relations to one or zero , respectively ,
the heisenberg equations are shown to describe the full quantum or classical motion of the interaction system , thus allowing us to discern the differences between the fully quantum and fully classical dynamical picture .
in addition , for this second scenario , in the fully quantum - mechanical description , we considered special examples of @xmath387 for the coupled angular momentum state , demonstrating the explicit appearances of entanglement .
the entanglement increases with @xmath5 and so persists even as @xmath388 , a limit which would go over to the classical picture for an _ uncoupled _ angular momentum oscillator .
this entanglement occurs because the uppermost excited state of the @xmath5-level angular momentum oscillator is always _ coupled _ to the vacuum state of the linear oscillator , a purely quantum coupling which manifestly gives rise to spontaneous emission .
we have also shown that the dynamics of this scenario can be systematically described by the irreducible matrix representation based on the generating operators of the group @xmath8 , while that of the first scenario can be given , based on the coupled boson representation , simply by the rotations of an angular momentum oscillator .
for the coupled linear - angular momentum oscillator , this system was shown to display periodicity in the measure of entanglement for @xmath9 and @xmath10 , whereas for @xmath11 and beyond , the measure of entanglement was shown to be aperiodic ; this aperiodicity is apparent from the form of the diagonalized multi - level hamiltonian and the resulting structure of the time evolution operator .
the authors acknowledge the support of the office of naval research and the national science foundation for this work .
from eqs . ( [ eq : rotation_mat3 ] ) and ( [ eq : f_function ] ) with the relation that @xmath389 , we find that @xmath390 particularly for the case that @xmath56 , namely @xmath391 , eq .
( [ eq : appendix_1 ] ) reduces to @xmath392 where @xmath393 .
the set of the generating operators is given by @xcite @xmath394 where @xmath395 , and @xmath396 here , @xmath397 , and @xmath398 and @xmath399 denote @xmath400 and @xmath401 , respectively ; @xmath402 , and @xmath403 , where @xmath404 denotes @xmath405 .
@xmath406 represents a projection operator for @xmath407 , and a transition operator for @xmath408 . for @xmath409 ( i.e. , @xmath9 )
, the generators @xmath410 exactly correspond to @xmath411 . for @xmath412 ( i.e. , @xmath10 ) , three generators @xmath413 , for example , have the matrix representations , @xmath414 respectively . similarly , for @xmath415 generators
@xmath416 , we have @xmath417 the set @xmath418 can be used as a well - defined basis of the space of @xmath419 matrices . 1 p.w .
shor , `` polynomial - time algorithms for prime factorization and discrete logarithms on a quantum computer '' in _ proceedings of the @xmath420th annual symposium on the foundations of computer science _
( ieee , los alamitos , ca , 1994 ) , pp .
124 - 134 .
grover , `` a fast quantum mechanical algorithm for database search '' in _ proceedings of the @xmath421th annual acm symposium on the theory of computing _ ( acm press , new york , 1996 ) , pp .
212 - 219 .
senitzky , `` nonperturbative analysis of the resonant interaction between a linear and a nonlinear oscillator '' , _ _ * 3 * , 421 ( 1971 )
. j. schwinger , _ quantum mechanics : symbolism of atomic measurements _ , edited by b .- g .
englert ( springer , berlin , 2001 ) .
a. shalom and j. zak , `` a quantum mechanical model of interference '' , _ phys .
a _ * 43 * , 13 ( 1973 ) .
iafrate and m. croft , `` interaction between two coupled oscillators '' , _ _ * 12 * , 1525 ( 1975 ) . for a detailed discussion on general entanglement measures ,
see , e.g. , o. rudolph , `` a new class of entanglement measures '' , _ j. math
. phys . _ * 42 * , 5306 ( 2001 ) .
see , e.g. , d. giulini , e. joos , c. kiefer , j. kupsch , i .- o . stamatescu , and h.d .
zeh , _ decoherence and the appearance of a classical world in quantum theory _
( springer , berlin , 1996 ) . g. mahler and v.a .
weberruss , _ quantum networks : dynamics of open nanostructures _ ( 2nd ed .
springer , new york , 1998 ) .
hughston , r. jozsa , and w.k .
wootters , `` a complete classification of quantum ensembles having a given density matrix '' , _ phys . lett .
a _ * 183 * , 14 ( 1993 ) . in an appendix
therein , it was shown that for pure states @xmath422 of the coupled system the two reduced density matrices , @xmath185 and @xmath186 , have the same non - zero eigenvalues with the same multiplicities . from this
we easily find that @xmath62 = \mathcal{p}[\hat{\rho}^{(2)}]$ ] .
see , e.g. , _ handbook of mathematical functions _ , edited by m. abramowitz and i. a. stegun ( national bureau of standards , 1964 ) .
see , e.g. , w.h .
louisell , _ radiation and noise in quantum electronics _
( mcgraw - hill , new york , 1964 ) .
senitzky , `` quantum optics .
two - frequency interactions '' , _ phys .
rev . _ * 183 * , 1069 ( 1969 ) .
senitzky , `` comment on energy balance for a dissipative system '' , _ _ * 51 * , 5166 ( 1995 ) .
jaynes and f.w .
cummings , `` comparison of quantum and semiclassical radiation theories with application to the beam maser '' , _ ieee _ * 51 * , 89 ( 1963 ) . in ref .
212 , the diagonalization of eq .
( [ eq : hamiltonian_resonance_case2 ] ) for @xmath9 only was described in the same basis .
e. abate and h. haken , `` exakte behandlung eines laser - modells '' , _
z. naturforsh . _ * 19a * , 857 ( 1964 ) ; m. tavis and f.w .
cummings , `` exact solution for an n - molecule - radiation - field hamiltonian '' , _ phys .
rev . _ * 170 * , 379 ( 1968 ) ; d.f .
walls and r. barakat , `` quantum - mechanical amplification and frequency conversion with a trilinear hamiltonian '' , _ _ * 1 * , 446 ( 1970 ) . | the fundamental quantum dynamics of two interacting oscillator systems are studied in two different scenarios . in one case ,
both oscillators are assumed to be linear , whereas in the second case , one oscillator is linear and the other is a non - linear , angular - momentum oscillator ; the second case is , of course , more complex in terms of energy transfer and dynamics .
these two scenarios have been the subject of much interest over the years , especially in developing an understanding of modern concepts in quantum optics and quantum electronics . in this work , however , these two scenarios are utilized to consider and discuss the salient features of quantum behaviors resulting from the interactive nature of the two oscillators , i.e. , coherence , entanglement , spontaneous emission , etc . , and to apply a _ measure of entanglement _ in analyzing the nature of the interacting systems .
the heisenberg equation for both coupled oscillator scenarios are developed in terms of the relevant reduced kinematics operator variables and parameterized commutator relations .
for the second scenario , by setting the relevant commutator relations to one or zero , respectively , the heisenberg equations are able to describe the full quantum or classical motion of the interaction system , thus allowing us to discern the differences between the fully quantum and fully classical dynamical picture . for the coupled linear and angular - momentum oscillator system in the fully quantum - mechanical description , we consider special examples of two , three , four - level angular momentum systems , demonstrating the explicit appearances of entanglement .
we also show that this entanglement persists even as the _ coupled _ angular momentum oscillator is taken to the limit of a large number of levels , a limit which would go over to the classical picture for an _ uncoupled _ angular momentum oscillator .
_ key words _ : entanglement : coupled - boson representation : spontaneous emission |
the problem of counting the number of ways in which a given permutation can be factored into a given number of transpositions is of perennial interest in algebraic combinatorics .
usually , one considers this problem in the presence of constraints on the factors , e.g. that they should be transpositions of a certain type , or should collectively generate a certain group , etc . by varying these constraints ,
one obtains enumeration problems which enjoy surprising connections with other branches of mathematics .
the earliest enumerative study of transposition factorizations was carried out by hurwitz @xcite in the nineteenth century .
motivated by a problem from enumerative algebraic geometry , namely the counting of almost simple ramified covers of the sphere by other riemann surfaces , hurwitz published an explicit formula for the number of _ minimal transitive factorizations _ of an arbitrary permutation into transpositions .
there are two constraints in the hurwitz factorization problem : `` minimality , '' which requires that the number of transpositions used should be as small as possible , and `` transitivity , '' which requires that the factors should act transitively on the points @xmath0 hurwitz s formula for the number of minimal transitive factorizations of a permutation @xmath1 of cycle type @xmath2 is @xmath3 as a particularly beautiful special case , hurwitz s formula yields that the number of factorizations of a full cycle in @xmath4 into @xmath5 transpositions is @xmath6 the number of trees on @xmath7 labelled vertices .
this case of hurwitz s formula was independently rediscovered and popularized by dnes @xcite .
the general hurwitz formula was independently rediscovered by goulden and jackson @xcite , to whom the first rigorous proof is due .
a key feature of the hurwitz factorization problem is centrality : the number of minimal transitive factorizations of @xmath8 depends only on the cycle type of @xmath9 this remains true for transitive factorizations of arbitrary length .
a different choice of constraints leading to a non - central factorization problem was considered by stanley @xcite , who initiated the study of what he termed _ reduced decompositions_. these are minimal factorizations in which the transpositions allowed to be used as factors are the coxeter generators @xmath10 reduced decompositions are also referred to as _ sorting networks _ because of their relation to the bubblesort algorithm familiar to computer scientists .
they could also be called _ minimal coxeter factorizations_. the enumeration of reduced decompositions is complicated by its non - centrality , and has spawned its own extensive literature , see @xcite for a beautiful introduction .
the asymptotic behaviour of random reduced decompositions is the subject of an intriguing set of conjectures due to angel et al .
@xcite .
the coxeter factorization problem naturally fits into a wider class of constrained factorization problems , in which the factors are chosen from a specified set of transpositions which generate @xmath11 a second example from from this class was considered by pak @xcite , who initiated the study of _
star factorizations_. these are factorizations in which the transpositions allowed to be used as factors have the form @xmath12 for example , the unique minimal star factorization of @xmath13 is @xmath14 thanks to recent work of irving and rattan @xcite , goulden and jackson @xcite , and fray @xcite , the combinatorics of star factorizations is now completely understood .
recently , gewurz and merola @xcite posed the problem of enumerating transposition factorizations under a constraint on the order of the factors .
a factorization @xmath15 of @xmath1 into a product of @xmath16 transpositions is called _ primitive _ if @xmath17 i.e. if its factors appear in weakly increasing order with respect to the larger element in each .
for example , @xmath13 can be factored into a product of two transpositions in three ways , @xmath18 but only the first two of these factorizations are primitive .
gewurz and merola obtain the interesting result that the number of primitive factorizations of the cycle @xmath19 into @xmath5 transpositions ( i.e. the number of minimal primitive factorizations ) is the catalan number @xmath20 this result should be considered in tandem with hurwitz s @xmath21-count of unrestricted factorizations of @xmath22 in this extended abstract prepared for fpsac 2010 , we will give an overview of the authors ongoing work on the enumeration of primitive factorizations .
first we study minimal primitive factorizations , and obtain an analogue of hurwitz s formula , i.e. an explicit formula which counts minimal primitive factorizations of an arbitrary permutation ( theorem [ thm : minimaltype ] and corollary [ cor : minimaltotal ] below )
. then we present a link between the primitive factorization problem and jucys - murphy elements .
this connection explains the centrality of the primitive factorization problem , and allows us to use character theory to enumerate primitive factorizations of a full cycle into any number of transpositions ( theorem [ thm : centralfactorial ] below ) .
finally , we discuss a surprising connection between the primitive factorization problem and the theory of matrix models : generating functions enumerating primitive factorizations may be expressed as integrals over groups of unitary matrices against the haar measure ( theorem [ thm : matrixintegral ] below ) .
it turns out that these integrals are of independent interest and have a long history in mathematical physics .
any primitive factorization of @xmath8 into @xmath16 transpositions has the form @xmath23 where @xmath24 is a weak @xmath25-part composition of @xmath26 we will say that the above factorization is of type @xmath27 if the frequencies @xmath28 coincide with the parts of @xmath29 after reordering . for example , there are three primitive factorizations of @xmath30 of type @xmath31 namely @xmath32 let us now enumerate minimal primitive factorizations by type .
let @xmath33 denote the set of all weakly increasing sequences @xmath34 of @xmath16 positive integers such that @xmath35 for @xmath36 and @xmath37 it is not difficult to show that @xmath38 given a partition @xmath39 one may introduce a refinement @xmath40 of the catalan number by declaring @xmath40 to be the number of sequences in @xmath33 of type @xmath41 then , by definition , @xmath42 these refined catalan numbers have previously been studied by haiman @xcite and stanley @xcite in connection with parking functions , and are known explicitly : @xmath43 where @xmath44 is the multiplicity of @xmath45 in @xmath41 finally , given a pair of partitions @xmath46 , introduce the set of sequences of partitions @xmath47 thus sequences in @xmath48 are obtained by breaking parts of @xmath49 in such a way that , after sorting , one obtains @xmath41 if @xmath48 is non - empty , then @xmath29 is said to be a refinement of @xmath50 [ thm : minimaltype ] let @xmath1 be a permutation of reduced is the partition obtained by subtracting one from the length of each of its cycles .
thus the size of the reduced cycle type of @xmath8 is the length of a minimal factorization of @xmath8 into transpositions . ]
cycle type @xmath51 , and let @xmath29 be another partition of @xmath26 the number of primitive factorizations of @xmath8 of type @xmath29 is @xmath52 as an example , consider the permutation @xmath53 this permutation has reduced cycle type @xmath54 so the length of a minimal primitive factorization of @xmath8 is eight . by theorem [ thm :
minimaltype ] , the number of minimal primitive factorizations of @xmath8 of type @xmath55 is @xmath56 the proof of theorem [ thm : minimaltype ] is bijective , and we refer the reader to our full - length article @xcite for details . as a corollary of this result
, we obtain an elegant formula which counts the total number of minimal primitive factorizations of an arbitrary permutation .
gewurz and merola s result is recovered as the case @xmath57 of this corollary .
[ cor : minimaltotal ] let @xmath1 be a permutation of non - reduced cycle type @xmath58 the total number of primitive factorizations of @xmath8 into @xmath59 transpositions is @xmath60
we will now give an algebraic explanation of the fact that the primitive factorization problem is central .
let @xmath61 $ ] denote the group algebra of the symmetric group , and @xmath62 its center .
the center is a commutative algebra with canonical basis @xmath63 consisting of the conjugacy classes of @xmath64 for this reason we call @xmath62 the _ class algebra_. let @xmath65 denote the algebra of symmetric functions over @xmath66 we define a specialization @xmath67 as follows
. for @xmath68 put @xmath69 thus @xmath70,$ ] with @xmath71 the elements so defined are called _ jucys - murphy elements_. they were introduced independently by jucys @xcite and murphy @xcite .
these simple elements have many remarkable properties , some of which we will make use of here . diverse applications of the jucys - murphy elements are found in the work of okounkov @xcite and okounkov and vershik @xcite .
although @xmath72 for @xmath73 the jm elements do belong to the gelfand - zetlin subalgebra of @xmath61.$ ] this is the maximal commutative subalgebra of @xmath61 $ ] generated by the class algebras @xmath74 where @xmath75 are embedded in @xmath61 $ ] in the canonical way .
this is clear , since @xmath76 is by definition the difference of an element of @xmath77 and an element of @xmath78 consequently , the jm elements commute with one another , and we may define the alphabet @xmath79 and evaluate symmetric functions on this alphabet .
it is a remarkable result of jucys that , for any @xmath80 @xmath81 thus symmetric functions of jm elements are central , and we have the _ jm specialization _ @xmath82 the jm specialization gives a very clean proof of the centrality of the primitive factorization problem .
let @xmath83 denote the complete homogoneous symmetric function of degree @xmath84 i.e. @xmath85 the sum of all degree @xmath16 monomials in the variables @xmath86 then , by jucys result , @xmath87 and we may write @xmath88 for some coefficients @xmath89 on the other hand , we have that @xmath90 thus , for any permutation @xmath1 of cycle type @xmath91 we have that @xmath92 in other words , the combinatorial problem of counting primitive factorizations is equivalent to the algebraic problem of resolving @xmath93 with respect to the canonical basis of the class algebra @xmath94 more generally , for any partition @xmath29 one may consider the resolution @xmath95 where @xmath96 is the monomial symmetric function of type @xmath41 we then have , for any @xmath97 @xmath98
the above algebraic encoding of the primitive factorization problem allows us to enumerate primitive factorizations of a full cycle @xmath99 this is thanks to the remarkable properties of jm elements in irreducible representations of @xmath61,$ ] which amount to the fact that , while it is difficult to compute the the coordinates of @xmath100 with respect to the class basis of @xmath101 it is easy to compute its coordinates with respect to the character basis of @xmath94 the following remarkable result is due to jucys @xcite , see @xcite for a complete proof . given a young diagram @xmath29 and a cell @xmath102 in @xmath103 recall that the _ content _ of @xmath29 is defined to be the column index of @xmath29 less its row index .
let us associate to each young diagram @xmath29 its alphabet @xmath104 of contents .
let @xmath105 denote the product of all hook - lengths of @xmath41 let @xmath106 be the characters of the irreducible representations of @xmath107 which form a basis of @xmath94 then , for any symmetric function @xmath80 the character expansion of @xmath100 is @xmath108 thus , at the combinatorial level , the character expansion of @xmath100 is implemented by the substitution rule @xmath109 let us use the above character expansion result to enumerate primitive factorizations of a full cycle @xmath110 of any given length .
we already know that the number of minimal primitive factorizations of @xmath111 i.e. those consisting of @xmath5 transpositions , is the catalan number @xmath112 we will now solve the problem when the number of transpositions used is allowed to be arbitrary .
as we will see momentarily , this problem has an unexpectedly simple and beautiful solution .
let @xmath113 be an indeterminate , and form the generating function @xmath114 this generating function is an element of the algebra @xmath115 $ ] of single - variable formal power series with coefficients in the class algebra @xmath94 by jucys character expansion result , we have @xmath116 where the last line follows from the generating function @xmath117 for the elementary symmetric functions .
note that this computation shows that the generating function @xmath118 is actually a _ rational _ function over @xmath94 now , given @xmath119 in order to obtain the generating function @xmath120 we simply take the corresponding traces of the conjugacy class @xmath121 in each irreducible representation , to obtain a rational function over @xmath122 @xmath123 up until this point , the partition @xmath124 has been generic , but now we restrict to the special case @xmath125 the partition of @xmath7 with a single part .
a classical result from representation theory informs us that the trace of @xmath126 in an irreducible representation can only be non - zero in `` hook '' representations : @xmath127 now , the content alphabet of a hook diagram may be obtained immediately , @xmath128 so that @xmath129 for example , if @xmath130 this is a rational function of the form @xmath131 thus , as an irreducible rational function , @xmath132 has the form @xmath133 where @xmath134 are some constants to be determined momentarily . before finding the above coefficients ,
let us consider the generating function @xmath135 the coefficients in this generating function are complete symmetric functions evaluated on the alphabet @xmath136 of square integers .
reason dictates that they ought to be close relatives of the stirling numbers @xmath137 the stirling number @xmath138 has the following combinatorial interpretation : it counts the number of partitions @xmath139 of an @xmath140-element set into @xmath141 disjoint non - empty subsets .
stirling numbers are given by the explicit formula @xmath142 the numbers @xmath143 are known as _ central factorial numbers_. the central factorial numbers were studied classically by carlitz and riordan , see ( * ? ? ?
* exercise 5.8 ) for references .
they have the following combinatorial interpretation : @xmath144 counts the number of partitions @xmath145 of a set of @xmath140 marked and @xmath140 unmarked points into @xmath141 disjoint non - empty subsets such that , for each block @xmath146 if @xmath45 is the least integer such that either @xmath45 or @xmath147 appears in @xmath146 then @xmath148 central factorial numbers are given by the explicit formula @xmath149 now let us determine the unknown constants @xmath150 by the above discussion , the generating function @xmath132 has the form @xmath151 on the other hand , by the results of the previous sections , @xmath152 consequently , we must have @xmath153 and we have proved the following result .
[ thm : centralfactorial ] for any @xmath154 the number of primitive factorizations of @xmath19 into @xmath155 transpositions is @xmath156 where @xmath144 denotes the carlitz - riordan central factorial number .
equivalently , we have the generating function @xmath157
finally , we come to what is perhaps the most intriguing aspect of the primitive factorization problem : its connection with matrix models .
the theory of matrix models has its origins in an area of mathematical physics known as quantum field theory , see @xcite for a solid introduction . for our purposes ,
the following grossly oversimplified description of matrix model theory suffices : 1 .
pick an interesting subset @xmath158 of the space @xmath159 of all @xmath160 complex matrices .
2 . put an interesting probability measure @xmath161 on @xmath162 3 .
select an interesting random variable ( measurable function ) @xmath163 4 .
compute the expected value @xmath164 of the random variable @xmath165 ( possibly after rescaling ) as a power series in @xmath166 @xmath167 5 .
realize that the coefficients @xmath168 occurring in the above perturbative expansion have an interesting combinatorial interpretation .
this informal discussion is meant to convey the impression that , from a combinatorial perspective , matrix integrals may sometimes play the role of generating functions .
it sometimes happens that a traditional generating function is difficult to obtain , but that by running the above steps in reverse one can concoct a matrix integral which encodes a sequence of interest
. furthermore , some matrix models have special features that are very useful , and these features can be used to extract combinatorial information in the generating function spirit .
for example , it might be that an integral of interest can be exactly evaluated , thereby yielding an explicit generating function for the sequence it encodes ( see @xcite for a famous example ) .
even if this is not the case , it often happens that matrix integrals interact well with more advanced analytical tools , such as orthogonal polynomials or integrable systems of differential equations ( see e.g. @xcite ) , and this may again yield insight into combinatorial structure .
let us present a matrix model for primitive factorizations .
for our space of matrices we select the group @xmath169 of @xmath160 complex unitary matrices .
since @xmath169 is compact , it carries a unique left and right translation invariant probability measure , the haar measure @xmath170 which we take for our probability measure of interest .
now we will select an interesting random variable , or rather class of random variables , @xmath171 it would certainly be nice if we could compute the expected value @xmath172 of any polynomial function of the entries of @xmath173 and @xmath174 since we can approximate a large class of functions on @xmath169 by polynomials in matrix coefficients ( think stone - weierstrass / peter - weyl ) . by linearity of the integral , it suffices to consider the case where @xmath175 is a monomial .
furthermore , an easy argument using the invariance of the haar measure shows that the expected value of such a monomial will be zero unless @xmath175 is of equal degree in the entries of @xmath173 and @xmath176 ( think integrals of the form @xmath177 over the unit circle @xmath178 ) .
thus we need only consider integrals of the form @xmath179 where the lowercase @xmath180 s are matrix elements and @xmath181 are functions .
integrals of this form are called _
@xmath7-point correlation functions _ of matrix elements .
they are actually of considerable interest in mathematical physics @xcite and free probability theory @xcite .
it is known that , provided @xmath182 the computation of the @xmath7-point functions can be reduced to the computation of `` permutation correlators '' @xmath183 finally , recall that we have defined @xmath184 and that this quantity can be non - zero only for @xmath16 of the form @xmath185 therefore let us introduce the notation @xmath186 it is not unreasonable to think of @xmath187 as a combinatorially motivated analogue of the usual hurwitz number @xcite @xmath188 obtained by replacing the transitivity constraint with the primitivity constraint .
unfortunately , we will not be able to say much about the proof of this result here .
suffice to say that that theorem [ thm : matrixintegral ] arises from two points of view regarding the orthogonal projection of @xmath191 onto the commutant @xmath192 the first point of view involves the permutation correlators and the second involves understanding the element @xmath193 in the left - regular representation of @xmath61;$ ] the equivalence of the two is , in a sense , a manifestation of the schur - weyl duality between the representation theories of @xmath4 and @xmath194 we refer the interested reader to our articles @xcite for further details regarding the proof of theorem [ thm : matrixintegral ] , and its applications .
let us conclude with the following comparison of unrestricted and primitive factorizations of a full cycle : @xmath195 } \bigg { ( } \frac{\sinh z/2}{z/2 } \bigg{)}^{n-1 } \\
\tilde{a}_{g,(n ) } & = { \operatorname{cat}}_{n-1}{2n-2 + 2 g \choose 2n-2 } \bigg { [ } \frac{z^{2g}}{(2 g ) ! }
\bigg { ] } \bigg { ( } \frac{\sinh z/2}{z/2 } \bigg{)}^{2n-2}. \end{aligned}\ ] ] the first of these formulas is due to jackson @xcite ( see also @xcite ) , while the second is a consequence of theorem [ thm : centralfactorial ] together with riordan s exponential generating function for the central factorial numbers .
j. dnes , _ the representation of a permutation as the product of a minimal number of transpositions and its connection with the theory of graphs _
* 4 * ( 1959 ) , 63 - 70 .
b. shapiro , m. shapiro , a. vainshtein , _ ramified covers of @xmath199 with one degenerate branching point and the enumeration of edge - ordered graphs _ , adv . in math .
( ams transl . )
* 34 * ( 1997 ) , 219 - 228 . | [ [ abstract . ] ] abstract .
+ + + + + + + + + a factorization of a permutation into transpositions is called `` primitive '' if its factors are weakly ordered .
we discuss the problem of enumerating primitive factorizations of permutations , and its place in the hierarchy of previously studied factorization problems .
several formulas enumerating minimal primitive and possibly non - minimal primitive factorizations are presented , and interesting connections with jucys - murphy elements , symmetric group characters , and matrix models are described .
[ [ rsum . ] ] rsum .
+ + + + + + + une factorisation en transpositions dune permutation est dite `` primitive '' si ses facteurs sont ordonns .
nous discutons du problme de lnumration des factorisations primitives de permutations , et de sa place dans la hirarchie des problmes de factorisation prcdemment tudis .
nous prsentons plusieurs formules numrant certaines classes de factorisations primitives , et nous soulignons des connexions intressantes avec les lments jucys - murphy , les caractres des groupes symtriques , et les modles de matrices . |
as a promising mechanism of quark confinement , dual superconductivity picture of qcd vacuum has been intensively investigated up to today since the early proposal in 1970s @xcite .
the key ingredient of this picture is the existence of _ monopole condensation _ which causes the dual meissner effect . as a result
, the color electric flux between quark and antiquark is squeezed to form the string
. then the non - zero string tension plays the role of the proportional constant of the linear potential realizing quark confinement . for this picture to be true ,
the monopole condensed vacuum is expected to give more stable vacuum than the perturbative one . on the other hand ,
savvidy @xcite has argued based on the general theory of the renormalization group that the _ spontaneous generation of color magnetic field _ should occur in yang - mills theory .
however , nielsen and olesen @xcite have shown immediately that the explicitly calculated effective potential for the magnetic field in the background gauge has a pure imaginary part in addition to the real part which agrees with the prediction of the renormalization group equation .
in other words , the savvidy vacuum with magnetic condensation is unstable due to the existence of the _ tachyon mode _ corresponding to the lowest landau level realized by the applied external color magnetic field .
it should be remarked that the off - diagonal gluons were treated as if they were massless in these calculations .
now we recall that the infrared ( ir ) abelian dominance @xcite and magnetic monopole dominance in the maximal abelian ( ma ) gauge were confirmed in 1990s by numerical simulations @xcite where the ma gauge is the same as the background gauge adopted in @xcite .
the infrared abelian dominance @xcite follows from _ non - zero mass of off - diagonal gluons _ , see @xcite for numerical simulations and @xcite for analytical studies .
the off - diagonal gluon mass should be understood as the dynamical mass or the spontaneously generated mass , since the original yang - mills lagrangian does not include the mass term .
if sufficiently large dynamical mass is generated for off - diagonal gluons to compensate for the tachyon mode due to spontaneously generated magnetic field , therefore , the instability might disappear to recover the stability of the savvidy vacuum .
the main purpose of this paper is to clarify the physical origin of the dynamical mass for the off - diagonal gluons and thereby to obtain thorough understanding for the infrared abelian dominance and monopole dominance . in this connection , we discuss how the above two pictures ( dual superconductivity and magnetic condensation ) are compatible with each other in the gauge - independent manner .
this issue is also related to how to treat the ir divergence originating from the massless gluons and the tachyon mode corresponding to the lowest landau level . in other words ,
we examine whether a self - consistent picture can be drawn among magnetic condensation , massive off - diagonal gluons , abelian dominance , monopole dominance and stability of savvidy vacuum ( or disappearance of tachyon mode ) . in this paper
we adopt as a technical ingredient the cho - faddeev - niemi ( cfn ) decomposition @xcite to extract the topological configurations for the su(2 ) yang - mills theory in the ma gauge .
we adopt the cho - faddeev - niemi ( cfn ) decomposition for the non - abelian gauge field @xcite for extracting the topological configurations explicitly , such as magnetic monopoles ( of wu - yang type ) and multi - instantons ( of witten type ) . by introducing a three - component vector field @xmath3 with unit length , i.e. , @xmath4 , in the su(2 ) yang - mills theory ,
the non - abelian gauge field @xmath5 is decomposed as @xmath6 where we have used the notation : @xmath7 , @xmath8 and @xmath9 . by definition , @xmath10 is parallel to @xmath11 , while @xmath12 is orthogonal to @xmath11 .
we require @xmath13 to be orthogonal to @xmath11 , i.e. , @xmath14 .
we call @xmath15 the restricted potential , while @xmath16 is called the gauge - covariant potential and @xmath17 can be identified with the non - abelian magnetic potential as shown shortly . in the naive abelian projection , uniform in spacetime , e.g. , @xmath18 . then @xmath19 has the off - diagonal part @xmath20 and the diagonal part @xmath21
. see @xcite for details . ]
@xmath15 corresponds to the diagonal component , while @xmath16 corresponds to the off - diagonal component , apart from the magnetic part @xmath17 .
accordingly , the non - abelian field strength @xmath22 is decomposed as @xmath23 where we have introduced the covariant derivative , @xmath24 \equiv \hat{d}_\mu : = \partial_\mu + g \mathbb{v}_\mu \times , $ ] and defined the two kinds of _ abelian _ field strength : @xmath25 here @xmath26 is the magnetic field strength which is proportional to @xmath11 and is rewritten in the form @xmath27 since @xmath28 is shown to be locally closed and it can be exact locally with the abelian magnetic potential @xmath29 .
this is because @xmath28 represents the color magnetic field generated by magnetic monopoles as follows .
the parameterization of the unit vector @xmath30 leads to the expression @xcite @xmath31 where @xmath32 is the jacobian of the transformation from the coordinates @xmath33 on @xmath34 to @xmath35 parameterized by @xmath36 .
then the surface integral of @xmath28 over a closed surface is quantized in agreement with the dirac quantization condition : @xmath37 where @xmath35 is a surface of a unit sphere with area @xmath38 .
hence @xmath39 gives a number of times @xmath40 is wrapped by a mapping from @xmath41 to @xmath35 .
this fact is mathematically characterized as the non - trivial second homotopy class , @xmath42 .
the abelian magnetic potential @xmath29 is given by @xmath43 $ ] with arbitrary @xmath44 .
see @xcite for more information .
the original gauge transformation , @xmath45 \bm{\omega } , $ ] is replaced by two sets of gauge transformations . and
@xmath14 , see @xcite for more details . ] : ( passive or quantum gauge transformation ) @xmath46 \bm{\omega } , \quad \delta_\omega \mathbb{x}_\mu = d_\mu[\mathscr{a } ] \bm{\omega } - \bm{n } ( \bm{n } \cdot d_\mu[\mathscr{a } ] \bm{\omega } ) , \\ ( \longrightarrow & \delta_\omega \mathbb{b}_\mu = 0 , \quad \delta_\omega \mathbb{v}_\mu = \bm{n } ( \bm{n } \cdot d_\mu[\mathscr{a } ] \bm{\omega } ) ) . \end{aligned}\ ] ] : ( active or background gauge transformation ) @xmath47 \bm{\omega ' } - ( \bm{n } \cdot \partial_\mu \bm{\omega ' } ) \bm{n } , \quad \delta_{\omega } ' \mathbb{v}_\mu = d_\mu[\mathbb{v } ] \bm{\omega ' } ) .
\end{aligned}\ ] ] the gauge transformation ii does not mix @xmath13 with the abelian variables @xmath11 and @xmath48 . the transformation law for the field strength can be obtained in the similar way .
the massiveness of the off - diagonal gluons implies the abelian dominance @xcite .
first , we show that the magnetic condensation defined below leads to the massive off - diagonal gluons at least in the ma gauge .
therefore , we argue that the abelian dominance follows from the magnetic condensation . in the last stage
we argue that this result is also consistent with the monopole dominance and monopole condensation . by making use of the cfn decomposition ,
the ( euclidean ) yang - mills lagrangian is rewritten as @xmath49 where we have used a fact that @xmath50 , @xmath26 , @xmath51 , and hence @xmath52 are all parallel to @xmath11 , while @xmath53 is orthogonal to @xmath11 ( which follows from the fact @xmath54 ) . by collecting the terms in @xmath13 ,
the lagrangian reads @xmath55 where we have defined @xmath56 with @xmath57 -(d_\rho[\mathbb{v } ] d_\rho[\mathbb{v}])^{ab } \delta_{\mu\nu } + ( d_\mu[\mathbb{v } ] d_\nu[\mathbb{v}])^{ab } .
\label{defq}\end{aligned}\ ] ] the second term of @xmath58 is cast into the form , @xmath59 d_\rho[\mathbb{v}])^{ab } = -\partial_\rho^2 \delta^{ab } + g^2 [ ( v_\rho^c)^2 \delta^{ab } - v_\rho^a v_\rho^b ] - g \epsilon^{abc } \partial_\rho v_\rho^c - 2 g \epsilon^{abc } v_\rho^c \partial_\rho , \end{aligned}\ ] ] where @xmath60 and @xmath61 .
we focus on the quadratic term in @xmath13 without derivatives , @xmath62 x_\mu^a x_\mu^b , \label{crossterm}\end{aligned}\ ] ] which can be diagonalized to yield the mass term for the off - diagonal gluons .
therefore , _ if the vacuum condensation _ @xmath63 _ occurs , then the off - diagonal gluons acquire the non - zero mass _ @xmath64 given by @xmath65 note that @xmath66 has the gauge invariance i , while it does not have the gauge invariance ii , but the minimum value of the spacetime average of @xmath67 , i.e. , @xmath68 can have a definite non - zero value @xcite , see also @xcite .
hereafter , we call the condensation @xmath69 _ magnetic condensation _ for convenience .
the magnetic potential @xmath12 is related to the magnetic monopole in the sense explained above .
if such magnetic condensation takes place at all , it could be related to the ( magnetic ) monopole condensation .
we postpone the relationship between magnetic condensation and ( magnetic ) monopole condensation to a subsequent paper @xcite . or @xmath70 , paying special attention to the magnetic potential @xmath71 or @xmath72 alone .
this is partly because the electric ( photon ) condensation @xmath73 will be incompatible with the residual u(1 ) invariance , and partly because we can not obtain the final closed expression for the effective potential including the electric u(1 ) potential @xmath74 . ]
next , we discuss how such magnetic condensation can occur actually in gluodynamics . at first sight
, such a situation looks very difficult to be realized , since the original yang - mills lagrangian does not include the quartic polynomials in the field @xmath12 of the type @xmath75 , although the other type @xmath76 is contained as a part of @xmath77 in the tree lagrangian .
we restrict our arguments to the euclidean space .
then a mathematical identity for the @xmath78 gauge potential , yields the lower bound on @xmath79 , @xmath80 i.e. , @xmath81 where @xmath82 and @xmath83 therefore , the magnetic condensation in question @xmath84 follows from the other magnetic condensation @xmath85 , since @xmath86 thus , @xmath87 and @xmath88 are not independent . to obtain the definite value of @xmath87 , we need to know @xmath89 in addition to @xmath88 which will be calculated below .
the naive generating functional is given by ( @xmath90 : source term ) @xmath91 = \int \mathcal{d}a_\mu \mathcal{d}\mathbb{x}_\mu \mathcal{d}\bm{n}\delta(\bm{n } \cdot \bm{n}-1 ) \delta(\bm{n } \cdot \mathbb{x}_\mu ) \delta(d_\mu[\mathbb{v } ] \mathbb{x}_\mu ) e^ { - \int_{x } \mathscr{l}_{ym } - \int_{x } \mathscr{l}_{j } } .
\end{aligned}\ ] ] the measure @xmath92 is shown to be invariant under the i , ii gauge transformations .. however , the determinant does not influence the one - loop result , see @xcite , and it is neglected in the following . ] two conditions @xmath93 and @xmath54 are also invariant under the gauge transformations i and ii .
the additional condition @xmath94 \mathbb{x}_\mu = 0 $ ] is covariant under the gauge transformation ii , i.e. , @xmath95 \mathbb{x}_\mu ) = g ( d_\mu[\mathbb{v } ] \mathbb{x}_\mu ) \times \bm{\omega'}$ ] , while it is neither invariant nor covariant under the gauge transformation i. then it can be identified with the gauge fixing ( gf ) condition for the off - diagonal part perpendicular to @xmath11 ( generalized differential ma gauge ) according to @xcite , since @xmath96 \mathbb{x}_\mu=0 $ ] .
in fact , it is obtained by minimizing @xmath97 w.r.t .
the gauge transformation i : @xmath98 \mathbb{x}_\mu .\end{aligned}\ ] ] we can define two sets of brst transformations @xmath99 corresponding to gauge transformations i , ii by introducing the nakanishi - lautrup ( nl ) auxiliary fields @xmath100 and the faddeev - popov ( fp ) ghost @xmath101 and antighost fields @xmath102 , respectively .
the brst transformations defined in this way are shown to be nilpotent @xmath103 .
we can obtain the two fp ghost terms @xmath104 and @xmath105 for the ma gf condition , @xmath106 = & \int \mathcal{d}\mathbb{c } \mathcal{d}\bar{\mathbb{c } } \mathcal{d}\mathbb{n } \mathcal{d}\mathbb{c } ' \mathcal{d}\bar{\mathbb{c } } ' \mathcal{d}\mathbb{n } ' \int \mathcal{d}\mathbb{x}_\mu \int \mathcal{d}\bm{n}\delta(\bm{n } \cdot \bm{n}-1 ) \delta(\bm{n } \cdot \mathbb{x}_\mu ) \int \mathcal{d}a_\mu \nonumber\\ & \times e^ { - \int_{x } [ \mathscr{l}_{ym}+\mathscr{l}_{gf+fp}+\mathscr{l}_{gf+fp } ' ] - \int_{x } \mathscr{l}_{j } } , \end{aligned}\ ] ] where ( apart from the gf + fp term for the parallel part ) @xmath107\mathbb{x}_\mu ) \nonumber\\
= & \mathbb{n } \cdot d_\mu[\mathbb{v}]\mathbb{x}_\mu + i \bar{\mathbb{c } } \cdot d_\mu[\mathbb{v}]d_\mu[\mathscr{a } ] \mathbb{c } + g^2 ( i\bar{\mathbb{c } } \cdot ( \bm{n } \times \mathbb{x}_\mu ) ) ( ( \bm{n } \times \mathbb{x}_\mu ) \cdot \mathbb{c } ) , \nonumber\\ \mathscr{l}_{gf+fp } ' = & -i \bm{\delta}'(\bar{\mathbb{c } } ' \cdot d_\mu[\mathbb{v}]\mathbb{x}_\mu ) = \mathbb{n } ' \cdot d_\mu[\mathbb{v}]\mathbb{x}_\mu + ig \bar{\mathbb{c } } ' \cdot ( d_\mu[\mathbb{v}]\mathbb{x}_\mu \times \mathbb{c } ' ) .
\label{gf+fp2}\end{aligned}\ ] ] in this framework , it is not difficult to show @xcite that the effective action written in terms of @xmath11 and @xmath48 is obtained in the one - loop level by integrating out @xmath13 and the other fields as @xmath91 = & \int \mathcal{d}\bm{n}\delta(\bm{n } \cdot \bm{n}-1 ) \int \mathcal{d}a_\mu e^ { - s_{eff } } , \\ s_{eff } = & \int_{x } \frac{1}{4 } [ \mathbb{f}_{\mu\nu } + \mathbb{h}_{\mu\nu } ] ^2 + \frac{1}{2}\ln \det \ { ( \mathbb{q}){}_{\mu\nu } \ } - \ln \det \ { -d_\mu[\mathbb{v}]d_\mu[\mathbb{v } ] \ } \nonumber\\ & + \frac{1}{2 } \ln \det \ { d_\mu[\mathbb{v } ] ( \mathbb{q}^{-1}){}_{\mu\nu } d_\nu[\mathbb{v } ] \ } + \cdots , \\ [ \mathbb{q}]_{\mu\nu}^{ab } : = & - ( d_\rho[\mathbb{v } ] d_\rho[\mathbb{v}])^{ab } \delta_{\mu\nu } + g \epsilon^{abc } n^c k_{\mu\nu } , \quad k_{\mu\nu } : = 2(f_{\mu\nu } + h_{\mu\nu } ) . \label{defq3}\end{aligned}\ ] ] it is possible @xcite to show in the similar way to @xcite that the coupling constant in this framework of the yang - mills theory under the cfn decomposition obeys exactly the same beta function as the original yang - mills theory , although this issue has already been discussed based on the momentum shell integration of the wilsonian renormalization group @xcite .
now we proceed to estimate the logarithmic determinants to obtain the effective potential .
however , only in the pure magnetic case , i.e. , @xmath108 ( @xmath109 ) , we can obtain a closed form for the effective potential of the ( spacetime - independent constant ) field @xmath110 . in case of massless gluons ,
the eigenvalue @xmath111 of the operator @xmath112 is given by @xmath113 with @xmath114 , corresponding to two transverse physical modes and two unphysical ( longitudinal and scalar ) modes respectively .
the contribution from two unphysical gluon modes are cancelled by that from the ghost and antighost , since the operator @xmath115d_\mu[\mathbb{v}]$ ] has exactly the same eigenvalues .
only the two physical modes contribute to the mode sum @xmath116 in the heat kernel method as @xmath117 and hence the logarithmic determinants are calculated as @xmath118d_\mu[\mathbb{v } ] \ } + \frac{1}{2 } \ln \det \ { d_\mu[\mathbb{v } ] ( \mathbb{q}^{-1}){}_{\mu\nu } d_\nu[\mathbb{v } ] \ } \nonumber\\ = & - \int d^4x { 1 \over ( 4\pi)^2 } \lim_{s \rightarrow 0 } { d \over ds } { \mu^{2s } \over \gamma(s ) } \int_{0}^{\infty } d\tau \tau^{s-2 } \ 2g\|\mathbb{h}\| \left [ { e^{-3g\|\mathbb{h}\| \tau } \over 1-e^{-2g\|\mathbb{h}\| \tau } } + { e^{g\|\mathbb{h}\| \tau } \over 1-e^{-2g\|\mathbb{h}\| \tau } } \right ] . \label{det1}\end{aligned}\ ] ] then the effective potential @xmath119 for @xmath120 is obtained as ) is apparently the same as the old result obtained by nielsen and olesen @xcite in which it is assumed that the external magnetic field @xmath121 has the specific direction in color space and the magnitude is uniform in space - time from the beginning . in the present derivation using the cfn decomposition , the direction @xmath3 of the color magnetic field @xmath122 can be chosen arbitrary at every spacetime point @xmath123 , and the lorentz and color rotation symmetry is not broken . moreover
, this formalism enables us to specify the physical origin of magnetic condensation as arising from the magnetic monopole through the relation , @xmath124 . ]
@xmath125 + \text{pure imaginary part } , \label{effpot}\end{aligned}\ ] ] where @xmath126 . the origin of the pure imaginary part in ( [ effpot ] ) is the last term in ( [ level ] ) and ( [ det1 ] ) , corresponding to the lowest landau level @xmath127 , which does not decrease for large @xmath128 leading to the infrared divergence at @xmath129 , whereas the ultraviolet divergence at @xmath130 can be handled by isolating the @xmath131 pole _ a la _ dimensional regularization . the real part of the effective potential has a minimum at @xmath132 away from the origin . neglecting the pure imaginary part , therefore , the magnetic condensation occurs at the minimum of the effective potential , @xmath133 .\end{aligned}\ ] ] the appearance of the pure imaginary part due to a tachyon mode is easily understood as follows .
the formal gaussian integration over the off - diagonal gluons leads to @xmath134 where the determinant is given as the product of all the eigenvalues , @xmath135 a tachyon mode @xmath136 makes the @xmath137 negative , since the other eigenvalues are positive .
hence , the logarithm produces the pure imaginary part .
once the dynamical mass generation occurs through the magnetic condensation , the eigenvalues are shifted upward by @xmath138 , i.e , @xmath139 . especially , the lowest tachyon mode @xmath140 is modified to @xmath141 .
if @xmath138 is large enough so that @xmath142 , the tachyon mode is removed , i.e. , @xmath143 , and all the eigenvalues of @xmath144 become positive .
the effective potential can be calculated by summing up the modes with positive eigenvalues alone as in the massless case .
thus the infrared divergence disappears and the pure imaginary part is eliminated in the effective potential , if the ratio @xmath145 is greater than one , @xmath146 .
even when the off - diagonal gluons are massive , we can see that the effective potential @xmath119 still has a nontrivial minimum in @xmath147 , although the location @xmath148 of the minimum could change , @xmath149 .
this is because the existence of the nontrivial minimum originates from the asymptotic freedom , i.e. , positivity of the coefficient @xmath150 of the quantum correction part @xmath151 in the one - loop effective potential @xmath119 where @xmath152 agrees with the first coefficient of the @xmath153-function , @xmath154 . in the massive case
, it is shown that @xmath155 is modified into @xmath156 \ge b_0 $ ] , depending only on the ratio @xmath157 , where @xmath158 is the generalized riemann zeta function .
since @xmath159 is positive for arbitrary value of @xmath160 , the effective potential @xmath119 still has a nontrivial minimum at @xmath161 without the pure imaginary part for @xmath162 .
( of course , the constant @xmath163 is also changed and hence the location of the minimum @xmath148 could change , but it does not affect the existence of the minimum . ) in fact , the bound @xmath164 , i.e. , @xmath165 follows from the identification ( [ mass ] ) and the inequality ( [ ineq ] ) , @xmath166 , provided that the expectation value @xmath88 is given by the minimum @xmath148 of the one - loop effective potential for massive gluons .
( a possible zero mode will be discussed in @xcite . )
the mass generation for off - diagonal gluons due to magnetic condensation modifies the energy mode , @xmath167 thus , the tachyon mode is removed and the instability of the savvidy vacuum no longer exists . another view is obtained for the effective potential of the magnetic component as shown below .
we apply the faddeev - niemi decomposition @xcite to the magnetic potential @xmath17 where we introduce the two complex scalar fields @xmath168 and a tweibein @xmath169 with @xmath170 to parameterize the plane of a four - dimensional space in which the two vector fields @xmath171 and @xmath172 lie : @xmath173 where @xmath174 is the complex combination satisfying @xmath175 , @xmath176 @xmath177 . by making use of this decomposition ,
two magnetic variables are rewritten as @xmath178 ^ 2 = { 1 \over 2 } x^2 y^2 , \\
b_\mu^a b_\mu^a = & |\psi_1|^2+|\psi_2|^2 = { 1 \over 2 } ( x^2+y^2 ) , \end{aligned}\ ] ] where we have defined @xmath179 the effective potential @xmath119 is drawn as function of new variables @xmath180 or @xmath181 in fig .
[ fig : potential ] .
the circle ( cylinder ) intersects with the four hyperbolae if the radius exceeds the minimal value , @xmath182 , i.e. , @xmath183 , see fig . [
fig : potential ] .
thus the result of the one - loop calculation is consistent with the non - vanishing vacuum condensation @xmath84 .
this mechanism of providing the mass for the gauge boson ( off - diagonal gluons ) is very similar to the higgs mechanics of providing the mass for the gauge particles through the non - vanishing vacuum expectation value ( vev ) of the scalar field @xmath184 which is obtained as the location @xmath185 of the absolute minimum of the higgs potential @xmath186 .
however , the tree potential has no non - trivial minima and the potential has non - trivial minima only through the radiative corrections based on the coleman
weinberg mechanism @xcite . as functions of @xmath187 and @xmath188 . ( a ) classical case @xmath189 : all the minima are on the two lines @xmath190 or @xmath191 , ( b ) quantum case ( one - loop level ) : all the minima lie in the four hyperbolae @xmath192 forming the flat directions.,title="fig:",height=226 ] as functions of @xmath187 and @xmath188 . ( a ) classical case @xmath189 : all the minima are on the two lines @xmath190 or @xmath191 , ( b ) quantum case ( one - loop level ) : all the minima lie in the four hyperbolae @xmath192 forming the flat directions.,title="fig:",height=226 ] in order to show quark confinement , we need to calculate the wilson loop average @xmath193 \rangle$ ] . the non - abelian wilson loop operator @xmath194 $ ] for a closed loop @xmath195 is defined as the trace of the path - ordered product @xmath196 of the exponential of the line integral of the gauge potential along the closed loop @xmath195 : @xmath197 : = { \rm tr } \left [ \mathscr{p } \exp \left\ { ig \oint_{c } dx^\mu \mathscr{a}_\mu(x ) \right\ } \right ] .\end{aligned}\ ] ] a version of the non - abelian stokes theorem @xcite for su(2 ) and @xcite for su(n ) allows us to rewrite the wilson loop into the surface integral over the surface @xmath198 with the boundary @xmath195 : @xmath197 = & \int d\mu_s(\bm{n } ) \exp \left\ { i { j \over 2 } \int_{s : \partial s = c } d\sigma^{\mu\nu } g_{\mu\nu } \right\ } , \\
g_{\mu\nu}(x ) = & \partial_\mu [ \bm{n}(x ) \cdot \mathscr{a}_\nu(x ) ] - \partial_\nu [ \bm{n}(x ) \cdot \mathscr{a}_\mu(x ) ] - { 1 \over g } \bm{n}(x ) \cdot [ \partial_\mu \bm{n}(x ) \times \partial_\nu \bm{n}(x ) ] , \end{aligned}\ ] ] where we have used the adjoint orbit representation @xcite @xmath199 and @xmath200 is the product of the invariant measure on @xmath201 , @xmath202 over the surface @xmath198 with @xmath203 characterizing the representation of su(2 ) group , @xmath204 . under the gauge transformation , @xmath11 transforms in the adjoint representation , @xmath205 then @xmath206 has the manifestly gauge invariant form , @xmath207 , $ ] with the covariant derivative @xmath208 . in the cfn decomposition , @xmath206 is identified with @xmath209 and the wilson loop is rewritten in terms of the abelian components composed of @xmath48 and @xmath11 . if there are no string singularities in @xmath210 ( say the photon part ) , then the magnetic monopole current defined by @xmath211 reduces to @xmath212 where @xmath213 obeys the topological conservation law @xmath214 and is not the noether current .
it is obvious that @xmath206 has the same form as the t hooft polyakov tensor under the identification @xmath215 thus the wilson loop is the probe of the magnetic monopole .
these results suggest that the massive fields @xmath13 are redundant for large loop @xmath195 ( decouple in the long distance ) and the infrared abelian dominance in the wilson loop average is justified to hold .
thus we obtain a consistent scenario of abelian dominance and monopole dominance for quark confinement @xcite .
it is important to remark that the _ off - diagonal gluon condensation _
@xcite in the present framework @xmath216 yields the _ mass term _ for the field @xmath217 : @xmath218 in view of ( [ crossterm ] ) .
therefore , the off - diagonal gluon condensation yields the faddeev - skyrme model with knot solitons , based on intricate arguments .
we do not need the procedure for taking the static limit . ]
@xmath219 ^ 2 . \ ] ] however , the systematic derivation of the off - diagonal gluon condensation is still lacking .
the claim @xmath220 is also very suggestive for the existence of the faddeev - skyrme model .
we have applied the cho - faddeev - niemi decomposition to the su(2 ) yang - mills theory and performed the brst quantization .
first , we have pointed out that the magnetic condensation , @xmath84 , can provide the mass for the off - diagonal gluons .
therefore the existence of the magnetic condensation enables us to explain the infrared abelian dominance which has been confirmed by numerical simulations in the last decade as a key concept for understanding the dual superconductivity .
second , the magnetic condensation in question can occur if the other magnetic condensation takes place , i.e. , @xmath221 , based on a simple mathematical identity .
third , we have explicitly calculated the effective potential of the color magnetic field @xmath222 in one - loop level .
the resulting effective potential has a non - zero absolute minimum supporting the magnetic condensation @xmath84 .
we have applied the change of variables to visualize the potential just as the higgs potential of the scalar field , apart from the savvidy instability .
note that the magnetic condensation is not exactly equivalent to the monopole condensation .
yet , two condensations are intimately connected to each other .
in fact , a scenario for the off - diagonal gluons to acquire their mass for realizing the monopole condensation has already been demonstrated in @xcite where the physical origin of the off - diagonal gluon mass was not specified .
therefore , _ monopole condensation as the origin of dual superconductivity follows from the magnetic condensation _
@xmath84 , if we regard the resulting massiveness of the off - diagonal gluons .
as the magnetic condensation is sufficient to provide the mass for off - diagonal gluons , the magnetic condensation and monopole condensation can give a self - consistent picture for supporting the dual superconductivity .
however , the existence of the pure imaginary part in the effective potential @xmath119 was a signal of the savvidy instability which was discovered by nielsen and olesen @xcite .
in this paper we have shown that the mass generation for off - diagonal gluons due to magnetic condensation modifies the calculation of the effective potential and eliminates the pure imaginary part corresponding to the tachyon mode .
thus , the mass generation for off - diagonal gluons due to magnetic condensation recover the stability of the savvidy vacuum . to go beyond the one - loop calculation presented in this paper
, we can introduce the antisymmetric auxiliary tensor field @xmath223 as in @xcite and include its condensation as an independent degrees of freedom to search for the true vacuum .
this is also necessary to discuss the confining string @xcite .
extension of the magnetic condensation to su(3 ) is also indispensable to discuss the realistic world .
these issues will be tackled in a subsequent paper @xcite .
the author would like to thank atsushi nakamura , takeharu murakami and toru shinohara for helpful discussions .
this work is supported by grant - in - aid for scientific research ( c)14540243 from japan society for the promotion of science ( jsps ) , and in part by grant - in - aid for scientific research on priority areas ( b)13135203 from the ministry of education , culture , sports , science and technology ( mext ) .
999 y. nambu , phys .
d 10 , 4262 - 4268 ( 1974 ) .
+ g. t hooft , in : high energy physics , edited by a. zichichi ( editorice compositori , bologna , 1975 ) .
+ s. mandelstam , phys .
report 23 , 245 - 249 ( 1976 ) .
+ a.m. polyakov , phys .
b 59 , 82 - 84 ( 1975 ) . nucl .
b 120 , 429 - 458 ( 1977 ) .
kondo , [ hep - th/0105299 ] , phys .
b 514 , 335345 ( 2001 ) .
kondo , [ hep - th/0306195 ] , phys .
b 572 , 210 - 215 ( 2003 ) .
kondo , t. murakami , t. shinohara and t. imai , [ hep - th/0111256 ] , phys .
d 65 , 085034 ( 2002 ) . | we show that a certain type of color magnetic condensation originating from magnetic monopole configurations is sufficient to provide the mass for off - diagonal gluons in the su(2 ) yang - mills theory under the cho
faddeev niemi decomposition .
we point out that the generated gluon mass can cure the instability of the savvidy vacuum .
in fact , such a novel type of magnetic condensation is shown to occur by calculating the effective potential .
this enables us to explain the infrared abelian dominance and monopole dominance by way of a non - abelian stokes theorem , which suggests the dual superconductivity picture of quark confinement .
finally , we discuss the implication to the faddeev - skyrme model with knot soliton as a low - energy effective theory of yang - mills theory . ' '' '' ' '' '' 1.0 cm * magnetic condensation , * 0.3 cm * abelian dominance , * 0.3 cm * and instability of savvidy vacuum * 0.3 cm * * 0.5 cm * kei - ichi kondo@xmath0 * 0.5 cm _ @xmath1department of physics , faculty of science , chiba university , chiba 263 - 8522 , japan _
key words : magnetic condensation , abelian dominance , monopole condensation , quark confinement , savvidy vacuum , pacs : 12.38.aw , 12.38.lg ' '' '' 0.1 cm @xmath2 e - mail : [email protected] 0.5 cm |
the total hilbert space for @xmath1 particles with @xmath26 accessible spin components and interaction strength @xmath10 is constructed as the tensor product @xmath27 of the spatial hilbert space @xmath28 and the spin hilbert space @xmath29 .
the hamiltonian @xmath23 ( [ hammy ] ) acts as the identity on @xmath30 , but it can be used to reduce @xmath31 into a direct sum of energy eigenspaces @xmath32 @xmath33 where @xmath34 is the spectrum of @xmath23 for a particular @xmath10 . because @xmath17 is a symmetry of @xmath23 , the space @xmath31
can also be reduced into a direct sum of irreps of @xmath17 labeled by the triplet @xmath35 , i.e.@xmath36 where @xmath37 labels the degeneracy of @xmath38-type irreps .
the three labels @xmath39 of the irreps correspond to the three factors of @xmath17
. first , the center - of - mass excitation quantum number @xmath40 labels the irreps for @xmath21 .
second , the parity quantum number @xmath41 labels the irrep for parity inversion .
the relative parity quantum number @xmath42 will be used instead of @xmath43 to remove the contribution due to the parity from the center - of - mass excitations since inversion of the just the center - of - mass coordinate is contained with the @xmath21 symmetry factor .
the third factor of @xmath17 is the symmetric group @xmath19 . here
i briefly summarize some notation for @xmath19 and its irreps as necessary for this article ; consult a standard reference like @xcite for details .
the group @xmath19 has @xmath15 elements corresponding to the permutations of @xmath1 objects .
the set of elements of @xmath19 decomposed into conjugacy classes for each partition @xmath44 of @xmath1 , e.g. for four particles there are five partitions @xmath45 , @xmath46 , @xmath47 , @xmath48 and @xmath49 . for each conjugacy class
there is also an irrep of @xmath19 , and these are often depicted as standard young diagrams with @xmath1 boxes , e.g. @xmath50 .
the dimension of each irrep @xmath51 is the number of standard young tableaux for that diagram , i.e. the diagram s boxes are labeled such that rows of boxes always have an equal or increasing number as one moves to the right , and vertical boxes have always have increasing numbers . for a given @xmath1 ,
the irrep @xmath52 is one - dimensional and totally symmetric and the irrep @xmath53 is one - dimensional and totally antisymmetric .
all other irreps are multidimensional .
conjugate representations , in which rows and columns are exchanged , have the same dimension . within an irrep ,
each element of @xmath19 is represented by a @xmath51-dimensional matrix .
the traces of these matrices are called the characters of the representation , and all elements of @xmath19 from the same conjugacy class have the same character .
assuming there are no other symmetries of the hamiltonian than @xmath17 for the parameter range @xmath54 , then the degeneracy of any eigenvalue @xmath55 is constrained to be the dimension of an irrep of @xmath17 and there is a unique correspondence between each energy degeneracy subspace @xmath32 and a specific irrep @xmath56 .
this correspondence will not change as @xmath10 is adiabatically tuned within the range @xmath54 , even though energy of the subspace may vary .
therefore the energy levels in the subspace @xmath57 do not cross or reorder as @xmath10 is varied from non - interacting to hard - core repulsion .
if they did cross at some @xmath10 , this would imply an additional symmetry of @xmath23 for that particular value of interaction strength . to my knowledge
there is no analytic or numerical evidence for such an `` accidental '' symmetry ( see for example @xcite for the three - body case ) , but a proof of this assumption that @xmath17 exhausts the symmetries of @xmath23 for the parameter range @xmath54 would be most welcome .
if there are no spin degrees of freedom , or only one spin component is accessible , the spin hilbert space @xmath30 is one - dimensional and the decomposition of @xmath31 into @xmath17 irreps is the complete reduction of the problem . in that case , only the irreps @xmath52 for bosons and @xmath53 for fermions are physically realizable .
if the particles have multiple accessible spin components , then there are several different ways to proceed .
one way is to reduce the spin hilbert space @xmath30 with respect to @xmath19 and then takes its direct product with the spatial hilbert space @xmath31 .
note that the total spin operator @xmath58 is invariant with respect to @xmath19 so subspaces with the same total spin and spin component will belong to an irrep of @xmath19 .
techniques for the reduction of a spin system are well - known @xcite ; an example with four spin-@xmath59 fermions is provided in table [ spintable ] .
then the total hilbert space @xmath60 can be symmetrized to account for fermions or bosons by taking the direct product of the @xmath19 irreps for the spin with the @xmath19 factor of the @xmath17 irrep for the spatial part .
the reduction of the direct product of two @xmath19 irreps will contain a single copy of @xmath52 if the irreps are the same and a single copy of @xmath53 if they are conjugate irreps .
this method is essentially a variation of the technique described in @xcite and extended in @xcite .
.this table summarizes the decomposition of the spin hilbert space @xmath61 for @xmath62 spin-@xmath59 particles into irreps of @xmath63 .
for example , this shows that there are three ways to construct a spin state with @xmath64 and @xmath65 , and they transform among themselves under particle permutations as the irrep @xmath46 . in total
, @xmath30 can either be decomposed as @xmath66 , where @xmath67 is the irrep of @xmath68 , or as @xmath69 . [
cols="^,^,^,^,^",options="header " , ] | spectroscopic labels for a few particles with spin that are harmonically trapped in one - dimension with effectively zero - range interactions are provided by quantum numbers that characterize the symmetries of the hamiltonian : permutations of identical particles , parity inversion , and the separability of the center - of - mass .
the exact solutions for the non - interacting and infinitely repulsive cases are reduced with respect to these symmetries .
this reduction explains how states of single - component and multi - component fermions and bosons transform under adiabatic evolution from non - interacting to strong hard - core repulsion .
these spectroscopic methods also clarify previous analytic and numerical results for intermediate values of interaction strength .
several examples , including adiabatic mapping for two - component fermionic states in the cases @xmath0 , are provided .
the quantum system of a few particles in one dimension , confined by an external field and interacting via a short - range potential , has been a touchstone model for nuclear , atomic , statistical , and mathematical physics since the 1930 s .
this system possesses enough symmetry and regularity to be analytically tractable in many cases , but also manifests a rich enough phenomenological structure to serve as physically significant model for realistic systems .
this class of models has been a playground for studying the interplay of integrability and solvability , the emergence of universality in few - body phenomena , and the transition from few - body to many - body physics .
this article focuses specifically on the quantum system of @xmath1 particles of mass @xmath2 confined in a one - dimensional harmonic trap with frequency @xmath3 and interacting via an effectively zero - range potential .
the hamiltonian for the model is @xmath4 with the particle coordinates @xmath5 expressed in units of @xmath6 .
this hamiltonian has received intense attention recently because it serves as a model for neutral cold atoms with interactions tunable through the manipulation of feshbach and confinement - induced resonances trapped in extremely prolate , approximately harmonic optical wells @xcite .
the model has been used to investigate few - body and many - body properties of one - dimensional trapped ultracold gases of bosons @xcite , fermions @xcite , and mixtures of bosons and fermions @xcite .
the so - called unitary limit @xmath7 of hard - core repulsive interactions is known as the tonks - girardeau gas and has been realized in the laboratory with bosons @xcite , as has the metastable super - tonks - girardeau limit @xmath8 @xcite .
recent experiments in which the state of a few trapped , interacting fermions can be controlled and measured @xcite have generated excitement that tunable fermi gases will allow simulation of solid - state phenomena such as the transition to ferromagnetism @xcite .
the solutions of hamiltonian ( [ hammy ] ) form a complete basis for the spatial states of @xmath1 distinguishable particles , and these solutions can be symmetrized to account for identical bosons or fermions , whether single species or multi - component .
an exact , complete set of solutions of ( [ hammy ] ) are known when @xmath9 for all @xmath10 @xcite , and these solutions have become the benchmark for further theoretical investigations . for @xmath11 , a complete set of exact solutions
are known only for the interaction - free case @xmath12 and for the hard - core limit @xmath13 solutions are also energy eigenstates at @xmath14 , but they no longer form a complete basis due to the existence of bound states . ] . in the @xmath12 case , the system is equivalent to the well - known isotropic oscillator in @xmath1 dimensions . in the @xmath13 limit , @xmath15 independent solutions can be constructed from each @xmath12 fermionic solution @xcite by generalizing the famous bose - fermi mapping @xcite .
this mapping has also been generalized to construct exact solutions at @xmath16 for multi - component fermions @xcite , multi - component bosons @xcite and boson - fermion mixtures @xcite .
both sets of exact spatial solutions can be described using a unified spectroscopic framework .
the spectroscopic labels correspond to the irreducible representations ( irreps ) of the total symmetry group @xmath17 of the hamiltonian .
the group @xmath17 is isomorphic to the direct product @xmath18 , where @xmath19 is permutation symmetry , @xmath20 corresponds to parity and @xmath21 to contact symmetry of the one - dimensional harmonic oscillator in the separated center - of - mass variable .
the symmetry group @xmath17 holds for any interaction strength @xmath22 , so @xmath17 irrep labels serve as quantum numbers that are respected by the interparticle interaction for all values of @xmath10 , not just for the exact limits .
in this article , i argue that these spectroscopic labels can be used to help implement symmetrization , understand the transformation of states under rapid or adiabatic changes in the interaction parameter , and improve approximation schemes for states in the intermediate interaction regime .
many previous works have exploited the symmetry and representation theory of @xmath19 to construct symmetrized states , e.g. @xcite , but this treatment differs in three substantial ways .
first , the symmetrization and reduction into @xmath17 irreps in the @xmath12 and @xmath7 limits is facilitated by using the jacobi hypercylindrical coordinate system , defined below .
this coordinate system naturally incorporates the contact and dilation symmetries ( c.f .
@xcite ) of @xmath23 .
second , a method is used to construct spatial states with the right properties for symmetrizing multi - component spin systems that does not require explicitly constructing the spin state .
third , and perhaps most importantly , the group @xmath17 exhausts the symmetry of @xmath23 ( except for @xmath24 or @xmath25 ) , whereas @xmath19 does not .
therefore , for intermediate values of @xmath10 , each possible set of spectroscopic labels uniquely identifies a distinct degenerate energy subspace , and each subspace has the dimension of the corresponding @xmath17 irrep . when @xmath12 and @xmath13 the symmetry group of ( [ hammy ] ) is larger than @xmath17 but contains it as a subgroup .
that means the degenerate energy eigenspaces are reducible into irreps of @xmath17 , and they are usually not simply reducible .
the reduction problem for these two exact limits is solved in this article .
comparing the pattern of irreps of @xmath17 at @xmath12 and @xmath13 provides the generalization of the boson - fermion mapping to multicomponent cases or mixed symmetry . under adiabatic evolution of the interaction strength
, states follow this mapping .
the results of this method agree with recent calculations by other methods @xcite .
additionally , once this map is known , one can construct approximate solutions for intermediate interaction strengths from the extreme exact cases using variational methods ( c.f .
@xcite ) . in a subsequent work
i will show how this same spectroscopic framework also allows the dynamics arising from rapid variations of the interaction parameter , or quenches , to be described accurately and efficiently .
finally , the spectroscopic notation developed here clarifies results achieved by a variety of analytic and numerical approximation methods for intermediate values of @xmath10 @xcite .
these methods typically produce spectral graphs that are dense with crossing lines , but these lines can be assigned unique labels with this system . when the graphs are partitioned into @xmath17 irreps ,
there are no level crossings .
exact diagonalization methods that exploit this spectroscopic system will require fewer basis states to get the same accuracy because there are no interaction matrix elements between states in different @xmath17 irreps . |
for quite some time , the additivity conjecture has been one of the most notorious open problems in quantum information theory ; it has been settled only recently in a breakthrough paper by hastings @xcite .
the original conjecture can be stated in several equivalent ways @xcite ; one possible formulation is via the _ minimum output entropy _ of a quantum channel @xmath1 , defined as @xmath2 where the minimization is over all input states @xmath3 , and @xmath4 is von neumann entropy .
the intuition is that @xmath5 is a measure of noisiness of the channel @xmath1 .
the original additivity conjecture stated that @xmath6 for all channels @xmath1 and @xmath7 ; that is , the minimum output entropy of a pair of channels should be the sum of the individual minimum output entropies . during the years that the problem has been studied , it turned out to be convenient to generalize the additivity problem to _
@xmath8-rnyi entropies _ : for @xmath9 , @xmath10 , and density matrices @xmath3 , define @xcite @xmath11 and then the question is whether @xmath12 holds true in general , for all @xmath9 . due to the limit
@xmath13 , ( [ addconjp ] ) is a natural generalization of ( [ addconj ] ) .
quite surprisingly , it turned out that the conjectured equalities ( [ addconjp ] ) and ( [ addconj ] ) are both false in general .
they were subsequently disproved by constructing counterexample channels , first for @xmath14 @xcite , then for @xmath15 @xcite , then for @xmath16 and @xmath17 @xcite , and finally for @xmath18 @xcite , killing the original conjecture ( [ addconj ] ) . for detailed expositions of the problem and its history , see for example @xcite or @xcite . despite those no - go results
, it has been shown that additivity holds for many interesting classes of channels and several values of @xmath8 , for example for the cases that one of the channels is the identity channel @xcite or a unital qubit channel @xcite .
even if additivity fails in general , its validity in special cases is still interesting in its own and potentially useful for channel coding problems , cf .
the main goal of this paper is to show that some of those results for special channels have a natural interpretation within a more general framework . to motivate our more general definition , notice first that the additivity conjecture can equivalently be stated as the assertion that _ entanglement does not help _ to produce pure outputs .
in fact , equation ( [ addconj ] ) holds if and only if the map @xmath19 attains its global minimum at an _ unentangled _ input state @xmath3 : since @xmath20 for density operators @xmath21 and @xmath3 , we get @xmath22 if @xmath23 and @xmath24 are the minimizers for the two channels , i.e. @xmath25 and similarly for @xmath24 .
this means that the inequality `` @xmath26 '' in ( [ addconj ] ) is always true . on the other hand , as von neumann entropy is concave , the global minimum will be attained at some _ pure _ input state ; also , @xmath23 and @xmath24 can be chosen pure .
thus , the fact that @xmath27 is indeed the global minimizer , i.e. `` @xmath28 '' holds in ( [ addconj ] ) , is equivalent to the fact that no other entangled input state can produce even smaller output entropy .
equation ( [ addconjp ] ) can be reformulated in a similar way : additivity for @xmath8-rnyi entropy with @xmath16 holds true if and only if the function @xmath29 attains its global maximum at an unentangled input state @xmath3 .
thus , we have two variations of the same general problem : _ compute the trace of a convex function of the output , and decide whether this expression attains its global maximum at an unentangled input state .
_ for von neumann entropy @xmath4 , this function is @xmath30 , while for the @xmath8-rnyi entropy @xmath31 with @xmath16 , this function is @xmath32 .
it is natural to ask what happens if the problem is generalized .
what if one takes another convex function , different from @xmath30 or @xmath32 ?
we use definition [ defentfn ] to study the generalized problem . in this definition and
all of the following , applying a ( convex ) function @xmath33\to{{\mathbb r}}$ ] to a density matrix @xmath34 is meant in the sense of spectral calculus : diagonalizing @xmath35 , we define @xmath36 such that in particular @xmath37 , where the sum is over all eigenvalues @xmath38 of @xmath21 . [ defentfn ]
let @xmath33\to{{\mathbb r}}$ ] be a convex function , and let @xmath1 and @xmath7 be quantum channels .
we say that @xmath39 is _ additive _ for @xmath40 if there exists some unentangled input state @xmath41 such that @xmath42 for all input states @xmath21 . there are some simple consequences of this definition
. first note that if @xmath39 is convex as a real function , then it is automatically a `` convex trace function '' on the density operators in the sense that @xmath43 see @xcite .
thus , @xmath44 attains its maximum on pure input states , i.e. ( [ eqdef ] ) is equivalent to the existence of pure states @xmath45 and @xmath46 such that @xmath47 yet , in contrast to von neumann or @xmath8-rnyi entropy , there is in general no way to further simplify the expression on the left - hand side by splitting the function of the tensor product into two addends or factors .
clearly , this definition captures the additivity problems as special cases : [ lemma2 ] let @xmath40 be a pair of quantum channels .
then additivity of @xmath8-rnyi entropy @xmath48 holds if and only if the function @xmath49 is additive for @xmath40 , where @xmath50 hence proving the additivity conjecture for a pair of channels is equivalent to showing that @xmath30 is additive .
is there any reason why this more general approach could help ?
in fact , there is a popular example in matrix analysis where a similar strategy turned out to be successful , which is lwner s theory of operator convex functions ( @xcite ) .
a real function @xmath39 is called _ operator convex _ if @xmath51 for all self - adjoint operators @xmath3 and @xmath21 and @xmath52 .
this is an operator inequality , meaning that the difference of the right- and left - hand side is positive semidefinite .
clearly , operator convex functions are convex , but the converse turns out to be false . for example , @xmath53 is convex , but not operator convex .
given some convex function @xmath39 , it can be difficult to decide directly from the definition ( [ eqoperatorconvex ] ) whether @xmath39 is operator convex .
by contrast , it turns out that there is a simple characterization of the set of _ all _ operator convex functions , which can be stated elegantly in terms of integral representations or complex analysis .
this is an unexpected result , since the definition ( [ eqoperatorconvex ] ) itself involves only a linear - algebraic inequality .
thus , it seems reasonable to hope that something similar might happen in the case of the additivity problem , at least for special classes of ( highly symmetric ) channels : possibly the class of additive functions for a channel pair , defined by the linear - algebraic inequality ( [ eqdef ] ) , is also simple to characterize . as we will show in this paper , this speculation turns out to be true for the werner - holevo channel in @xmath0 dimensions at least .
we start by giving some simple examples .
[ lemaffine ] let @xmath54 and @xmath55 .
a convex function @xmath33\to{{\mathbb r}}$ ] is additive for a pair of channels if and only if the function @xmath56 is additive for that pair of channels . in particular , linear functions
@xmath57 are additive ( for every pair of quantum channels )
. * proof . *
it is clear that scaling a function with @xmath54 does not change the location of its global maximum
. if @xmath58 and @xmath59 are arbitrary quantum channels , and if @xmath3 is an arbitrary state on @xmath60 , then it holds for @xmath61 @xmath62 thus , @xmath63 is constant , and can be added to any function without changing its additivity properties . in particular ,
@xmath39 itself is additive , as every unentangled input @xmath41 satisfies ( [ eqdef ] ) . ' '' '' @xmath64 [ exidentity ] for channels of the form @xmath65 , every convex function is additive .
that is , if @xmath1 is an arbitrary quantum channel , and @xmath66 is the identity channel on some hilbert space , then every convex function @xmath33\to{{\mathbb r}}$ ] is additive for @xmath67 . in particular , as is well - known , von neumann entropy and the @xmath8-rnyi entropies are additive for such channels for all @xmath9 . *
the proof closely follows the lines of @xcite .
suppose @xmath68 .
we may choose @xmath69 to be pure .
let @xmath70 be a unitary dilation of @xmath1 , such that @xmath71 as seen in example [ lemaffine ] , we may assume that @xmath72 . since the expression right of @xmath73 is a pure state , the spectrum will not change if we replace @xmath73 by @xmath74 up to possible multiplicity of the eigenvalue zero .
thus , @xmath75 where @xmath76 . by convexity
, this expression is maximized if @xmath77 is pure , i.e. @xmath78 . ' '' '' @xmath64 it is well - known @xcite that the minimum output @xmath8-rnyi entropy of a pair of quantum channels is at least as large as that of one of its constituents , i.e. @xmath79 . the following lemma generalizes this statement , and yields an analogous property for all convex functions .
[ lemsinglechannel ] ' '' '' if @xmath33\to{{\mathbb r}}$ ] is a convex function with @xmath72 , then @xmath80 and similarly for @xmath7 . * remark . * if @xmath81 , then the bound is @xmath82 , where @xmath83 and @xmath84 denote the dimensions of the output hilbert spaces of @xmath1 and @xmath7 respectively
. * proof . *
if @xmath85 denotes an arbitrary pure state on the output hilbert space of @xmath7 , and @xmath86 is the spectrum of @xmath87 , then the spectrum of @xmath88 is @xmath89 , with @xmath90 zeroes .
thus , @xmath91 the equality in @xmath92 follows from example [ exidentity ] . ' '' '' @xmath64 the first counterexample channel to the additivity conjecture for the @xmath8-rnyi entropy ( for @xmath14 ) has been given by werner and holevo @xcite . in dimension @xmath93 , the werner - holevo channel @xmath94 is defined as @xmath95 it has the useful covariance property @xmath96 for every unitary @xmath97 . as a simple example , we derive a necessary condition for additivity for this channel in dimension @xmath98 : [ exhw ] if @xmath33\to{{\mathbb r}}$ ] is a convex function with @xmath99 then @xmath39 is not additive for the werner - holevo channel pair @xmath100 .
* proof . *
if @xmath101 are arbitrary pure states , the output @xmath102 has a @xmath103-fold degenerate eigenvalue @xmath104 , and a @xmath105-fold degenerate eigenvalue @xmath106 . due to the covariance property ( [ hwcovariance ] ) , this is true for all pure states and does not depend on @xmath107 or @xmath108 .
thus , @xmath109 on the other hand , if we input a maximally entangled state @xmath110 , it is shown in @xcite that the output @xmath111 has a single eigenvalue @xmath112 and a @xmath113-fold degenerate eigenvalue @xmath114 , such that @xmath115 comparing both expressions for @xmath98 , we see that @xmath39 is not additive for @xmath100 if the stated inequality holds . ' '' '' @xmath64 it is clear that we get more similar inequalities for @xmath116 , but in most cases , these inequalities seem to be weaker .
the following lemma shows that the multiplicativity problem of the minimum output rank also fits into definition [ defentfn ] .
we need this result later in the proof of example [ lemcontinuous ] .
[ lemminoutputrank ] ' '' '' the convex function @xmath117 \end{array } \right.\ ] ] is not for all channels additive .
* the function @xmath118 is related to the minimum output rank of quantum channels @xmath1 : @xmath119 where @xmath93 is the dimension of the output hilbert space of @xmath1 .
it has been shown in @xcite that the minimum output rank is not multiplicative ; there exist channels @xmath1 and @xmath7 such that @xmath120 which means that @xmath121 does not achieve its global minimum at tensor product input states @xmath3 .
consequently , @xmath122 achieves its global maximum at entangled input states @xmath3 . ' '' '' @xmath64
since every convex function on @xmath123 $ ] is bounded , the sup norm distance @xmath124 } |f(x)-g(x)|$ ] can be used as a distance measure on the set of convex functions @xmath125 on the unit interval .
this way , we get a notion of `` open '' and `` closed '' sets in @xmath125 .
formally , we get the relative topology of @xmath125 within the larger banach space of bounded functions on @xmath123 $ ] .
the next lemma shows that the set @xmath126 of additive functions for a fixed pair of channels is a closed cone , where `` cone '' refers to the simple property that @xmath127 holds for every @xmath128 .
[ lemopencone ] with respect to the @xmath129-norm topology , the set of additive functions on a pair of channels @xmath40 is a closed cone . *
the cone property is trivial : a function @xmath39 is additive for @xmath40 if and only if @xmath130 is additive for @xmath40 .
on the other hand , a function @xmath39 is _ not _ additive for @xmath40 if and only if there exists an entangled state @xmath3 , such that @xmath131 it is clear that there is some @xmath132 such that for any convex function @xmath133\to{{\mathbb r}}$ ] with @xmath134 , equation ( [ eqnonadditive ] ) still holds if @xmath39 is replaced by @xmath135 .
this shows that the set of non - additive functions for @xmath40 is open . ' '' '' @xmath64 in the following , we will often show that a sequence of additive convex functions @xmath136 on @xmath123 $ ] converges pointwise to a limit function @xmath39 , and then refer to lemma [ lemopencone ] to conclude that @xmath39 must be additive , too .
in fact , it is shown in ( * ? ? ?
* corollary 1.3.8 ) that in this case , pointwise convergence implies uniform convergence , and the limit function must be convex .
hence this kind of reasoning is justified .
it is a natural question whether the set of additive or non - additive functions has interesting properties .
one useful property is convexity .
it is not clear in general if the set of additive functions for a given arbitrary pair of channels is convex .
however , convexity holds for the special class of _ unitarily covariant _ channels . in accordance with @xcite
, we call a channel @xmath1 unitarily covariant if for every unitary @xmath97 , there exists a unitary @xmath137 such that @xmath138 sometimes a different class of channels is studied : a channel @xmath1 is called _ irreducibly covariant _ ( cf .
@xcite ) if there are irreducible unitary representations @xmath139 of a group @xmath140 such that @xmath141 for all @xmath142 and all @xmath3 .
unitarily covariant channels need not be irreducibly covariant , and vice versa ; for example , if @xmath1 and @xmath7 are @xmath93-dimensional unitarily covariant channels with @xmath143 , then the tensor product channel @xmath144 is irreducibly covariant with respect to @xmath145 , but it is in general not unitarily covariant . for a counterexample in the opposite direction ,
define a channel @xmath7 on @xmath146 via @xmath147 , where @xmath148 is some normalized vector .
then @xmath7 is unitarily covariant in the sense of equation ( [ equnicov ] ) ( with @xmath149 for every @xmath97 ) , but it is not irreducibly covariant , since any group representation @xmath150 satisfying equation ( [ equniirred ] ) must leave the subspace spanned by @xmath85 invariant .
the werner - holevo channel ( [ hwchannel ] ) is an example of a unitarily covariant channel due to ( [ hwcovariance ] ) .
[ lemconvexity ] ' '' '' if @xmath1 and @xmath7 are unitarily covariant channels , then the set of additive functions on @xmath40 is convex ( and due to lemma [ lemopencone ] , a closed convex cone ) . * proof .
* let @xmath39 and @xmath135 be additive convex functions for @xmath40 .
we have to prove that @xmath151 is also additive for @xmath40 . due to the unitary covariance of @xmath1 and @xmath7 , the eigenvalues of @xmath152 and @xmath153 are the same for every unitary @xmath97 and @xmath137 .
thus , @xmath63 depends only on the schmidt coefficients of the pure state @xmath3 ( and similarly for @xmath135 ) .
since @xmath39 is additive , the expression @xmath63 attains its global maximum at every pure unentangled input state @xmath3 at once .
the same is true for @xmath135 ; thus , @xmath39 and @xmath135 have a global maximizer in common .
it follows that @xmath151 must have the same global maximizer , namely , an unentangled state . ' '' ''
@xmath64 the minimum output entropy additivity conjecture is known to hold true for the werner - holevo channel , defined in ( [ hwchannel ] ) , in arbitrary dimensions . according to datta @xcite and alicki and fannes @xcite ,
the same is true for the additivity of the @xmath8-rnyi entropy for @xmath154 , but additivity does not hold if @xmath14 ( cf .
moreover , additivity also holds in the domain @xmath155 , as remarked in @xcite .
due to lemma [ lemma2 ] , those additivity results are related to the functions @xmath30 and @xmath32 for @xmath154 as well as @xmath156 for @xmath155 .
an interesting observation is that all these functions are operator convex as defined in ( [ eqoperatorconvex ] ) .
thus , the following theorem contains many known results on the werner - holevo channel as special cases : [ hwtheorem ] ' '' ''
every operator convex function @xmath157 is additive for the werner - holevo channel ( tensored with itself ) in dimension @xmath158 .
we conjecture that this is also true for the werner - holevo channel in larger dimensions @xmath116 and for more than two factors ; yet , it seems that the original calculations in @xcite can not be so easily adapted to that general case .
also , numerically it seems that it is sufficient that @xmath39 is operator convex on @xmath123 $ ] ( instead of @xmath159 ) , but the proof is more difficult . * proof .
* it is well - known @xcite that every operator convex function @xmath135 on @xmath160 has an integral representation of the form @xmath161 where @xmath162 is some probability measure on @xmath163 $ ] .
therefore , if @xmath39 is operator convex on @xmath159 , then it is in particular operator convex on @xmath164 , and we can shift the above expression by substituting @xmath165 to obtain @xmath166 where @xmath167 , and @xmath168 . moreover , the measure @xmath162 must vanish on @xmath169 $ ] , because @xmath170 has a pole in the positive reals for every @xmath171 $ ] , but @xmath39 is by assumption defined on all of @xmath159 . for the same reason , @xmath162 must vanish at @xmath172 . according to lemma [ lemconvexity ] , it is thus sufficient to show that the functions @xmath173 and @xmath174 are additive for every @xmath175 $ ] ; then , it follows that @xmath39 must be additive , too .
but the function @xmath173 is trivially additive ( as shown in lemma [ lemaffine ] ) .
let @xmath176 be the werner - holevo channel in dimension @xmath98 as defined in ( [ hwchannel ] ) . from @xcite , we know the eigenvalues of the output @xmath177
if the input has schmidt coefficients @xmath178 : there are @xmath179 eigenvalues of the form @xmath180 and @xmath158 eigenvalues of the form @xmath181 where @xmath182 , and @xmath183 .
hence , @xmath184 since the set @xmath185 is convex , the convex function @xmath186 attains its maximum on the extremal points , i.e. those points where , up to permutation , @xmath187 and @xmath188 . due to the simple form of the functions @xmath174 ,
it is easy to show with some analysis that the function @xmath189 attains its global maximum for @xmath190 $ ] at @xmath191 , corresponding to @xmath192 .
in fact , this expression is constant in @xmath193 for @xmath194 , and it is increasing in @xmath193 if @xmath195 . in summary , @xmath196 attains its global maximum on the states with schmidt coefficients @xmath197 , i.e. on the unentangled states .
thus , @xmath174 is additive for every @xmath175 $ ] for two copies of the werner - holevo channel in dimension @xmath158 .
the claim follows . ' '' '' @xmath64 consider the set @xmath198 of functions that are additive for _ all _ pairs of unitarily covariant channels @xmath40 . according to lemma [ lemconvexity ] , the set @xmath198 is a closed convex cone .
it is an interesting problem to determine the set @xmath198 explicitly . in the light of theorem [ hwtheorem ] , and due to the fact that the most natural closed convex subset of the convex functions is the set of operator convex functions , the following conjecture seems natural : [ conjaddop ] the set of functions @xmath198 that are additive for all unitarily covariant channels agrees with the set of operator convex functions on some interval @xmath199 .
it seems that for a _
fixed _ pair of channels , the set of additive functions does not have a simple description in general , and several natural conjectures on the structure of the set of additive functions fail .
for example , it is easy to construct convex functions @xmath39 and @xmath135 such that @xmath39 and @xmath151 are additive for the werner - holevo channel pair @xmath100 , but such that @xmath135 is not additive for @xmath100 . also , there are additive functions @xmath39 and @xmath135 such that @xmath200 is not additive ( cf .
theorem [ thepiecewiseaffine ] ) . in the following
, we will prove some more results on the set of functions that are additive for certain sets of channels .
we will assume that the channel sets have the following property : [ defchannelclass ] in the remainder of the paper , a _ channel class _ @xmath201 is a set of channels which is closed with respect to tensor products , and which contains all maximally depolarizing channels .
that is , * @xmath202 , * @xmath203 for all @xmath204 , where @xmath205 .
examples of channel classes are * the set of all channels , and * the set of irreducibly covariant channels : tensor products of irreducibly covariant channels are again irreducibly covariant @xcite , and @xmath206 is irreducibly covariant if @xmath21 is proportional to the identity .
the set of unitarily covariant channels is not a channel class .
however , the set of all channels which can be written as tensor products of unitarily covariant channels is a channel class .
we are interested in the set of functions that are additive for all channels in a given channel class @xmath201 ( we call them the `` functions that are additive for @xmath201 '' ) .
according to lemma [ lemopencone ] , the set of those functions is a closed cone for every channel class @xmath201 .
but we can say more .
[ thetensorstructure ] if a convex function @xmath33\to{{\mathbb r}}$ ] is additive for a channel class @xmath201 ( as defined in definition [ defchannelclass ] ) , then @xmath207 is additive for @xmath201 for every @xmath208 , too . moreover , if @xmath201 is additionally closed with respect to tensor products with @xmath206 for all @xmath21 , then the function @xmath209\mapsto \sum_{i=1}^n f(\mu_i\cdot x)\ ] ] is additive for @xmath201 as well for every probability vector @xmath210 ) .
* let @xmath211 be an arbitrary probability vector , and let @xmath39 be additive for @xmath201 .
let @xmath212 be arbitrary channels .
we have to show that the function @xmath213 is additive for @xmath40 .
let @xmath21 be a @xmath214 density operator with eigenvalues @xmath215 .
consider the channel @xmath216 , and let @xmath107 be an input vector for this tripartite channel .
it has a schmidt decomposition @xmath217 where @xmath218 and @xmath219 are orthonormal bases on the input hilbert spaces for the channels @xmath144 and @xmath206 respectively .
the corresponding output is @xmath220 the trace of a convex function on that output attains its maximum , due to convexity , in the extremal case where , up to permutation , @xmath187 and @xmath221 .
this means that we may choose the input to be unentangled between @xmath144 and @xmath206 . in this case
, if the output @xmath222 has spectrum @xmath223 , then the output @xmath224 has spectrum @xmath225 . since @xmath39 is additive for @xmath201 , it is in particular additive for the channel pair @xmath226 as long as @xmath201 is closed with respect to tensor products with @xmath206 .
if this is the case , the expression @xmath227 attains its global maximum at an unentangled input state @xmath107 .
but @xmath228 and so the expression @xmath229 attains its global maximum at unentangled input states @xmath107 .
it follows that @xmath230 is additive for @xmath40 .
in particular , we get that @xmath207 is additive for @xmath201 for every @xmath208 if we insert @xmath231 . ' '' '' @xmath64 as a simple example application , we find that functions which are additive for _ all _ channels must be continuous at zero : [ lemcontinuous ] if a convex function @xmath33\to{{\mathbb r}}$ ] is additive for all channels , then it is continuous at zero
. * proof .
* let @xmath201 be the class of all channels .
suppose that @xmath39 is additive for @xmath201 , but _ not _ continuous at zero . since @xmath39 is convex , the limit @xmath232 exists and is less than @xmath233 .
if @xmath39 is additive for @xmath201 , then the function @xmath234 is additive for @xmath201 as well due to example [ lemaffine ] . since @xmath39 is continuous on @xmath164 and @xmath235 , \end{array } \right.\
] ] the sequence of functions @xmath236 converges to the function @xmath118 introduced in lemma [ lemminoutputrank ] .
but we know from theorem [ thetensorstructure ] that the functions @xmath237 are additive for @xmath201 for every @xmath208 . moreover , according to lemma [ lemopencone ] , the set of additive functions for @xmath201 is closed . thus , @xmath118 must be additive , which contradicts lemma [ lemminoutputrank ] . ' '' '' @xmath64 we will later see that this result is not valid for @xmath238 : in lemma [ lemdelta1 ] , we show that there exist additive functions that are discontinuous at @xmath238 .
here is another interesting example which in some sense `` interpolates '' between the von neumann and @xmath8-rnyi entropies : [ lemdistortedentropy ] ' '' '' let @xmath239 and @xmath201 a channel class . if the function @xmath240 is additive for @xmath201 , then the function @xmath156 is additive for @xmath201 as well ; consequently , the minimum output @xmath8-rnyi entropy is additive for all channel pairs in @xmath201 .
* notice that @xmath241 is convex on @xmath123 $ ] if and only if @xmath239 , which explains the choice of the interval for @xmath8 .
suppose that @xmath241 is additive for @xmath201 .
then , @xmath242 is additive for @xmath201 as well for every @xmath208 according to theorem [ thetensorstructure ] . as multiplication with a constant
does not affect additivity , it follows that @xmath243 is additive for @xmath201 as well , and so is @xmath244 taking the limit @xmath245 , the claim follows from lemma [ lemopencone ] . ' '' '' @xmath64 here are some more consequences of theorem [ thetensorstructure ] .
the proofs are very similar to the proof of example [ lemdistortedentropy ] and thus omitted .
[ lemvonneumann ] let @xmath33\to{{\mathbb r}}$ ] be a convex function .
* if @xmath246 with @xmath247 and @xmath39 is additive for a class of channels @xmath201 , then @xmath30 is additive for @xmath201 as well , i.e. the minimum output von neumann entropy is additive for @xmath201 .
* if @xmath39 has a non - linear analytic extension to a complex neighbourhood of zero , then there exist channels @xmath40 such that @xmath39 is not additive for @xmath40 .
this shows that von neumann entropy plays some kind of special role : if any function that behaves like @xmath30 for small @xmath248 is additive , then von neumann entropy is automatically additive as well .
the second part of the lemma concerns possible functions @xmath39 that are additive for _ all _ channels : this possibility is ruled out for many functions , for example , say , for @xmath249 for @xmath250 . the main idea to prove the second part is the fact that , after subtracting a linear function , analytic functions can be approximated by a monomial @xmath251 , but the functions @xmath252 violate additivity for some channels as shown , for example , in @xcite .
a convex function on @xmath123 $ ] is automatically continuous on @xmath164 , but it may be discontinuous at the endpoints .
we have shown in example [ lemcontinuous ] that functions that are additive for all channels are continuous at @xmath253 .
in contrast , the following simple arguments show that additive functions may be discontinuous at @xmath238 .
[ lemdelta1 ] if @xmath1 and @xmath7 are quantum channels such that @xmath144 outputs a pure state , then every convex function is additive for @xmath40 .
consequently , if @xmath254\to{{\mathbb r}}$ ] are convex functions that differ only at @xmath238 , then @xmath39 is additive for any pair of channels if and only if @xmath135 is additive for that pair of channels .
* let @xmath33\to{{\mathbb r}}$ ] be a convex function , and suppose there exists some input state @xmath255 and a pure state @xmath108 such that @xmath256 . denoting the minimum output entropy of a channel @xmath1 by @xmath257 as in the introduction , it is well - known @xcite and in fact proven in the present paper in lemma [ lemsinglechannel ] that @xmath258 and so @xmath1 ( and by the same argument , @xmath7 ) outputs a pure state , too . taking the tensor product of the corresponding inputs
, we get an unentangled ( pure tensor product ) input state @xmath259 for @xmath144 such that @xmath260 is pure as well . due to the schur convexity @xcite of the map @xmath261
, the state @xmath259 is a maximizer of the map @xmath262 , and so @xmath39 is additive for @xmath40 .
let now @xmath1 and @xmath7 be arbitrary quantum channels . if @xmath144 outputs a pure state , then both @xmath39 and @xmath135 are additive for @xmath40 . on the other hand ,
if @xmath144 does not output a pure state , then the eigenvalues of every output are strictly less than @xmath263 , and @xmath264 for every @xmath21 . ' '' ''
@xmath64 thus , modifying a convex function at @xmath238 does not affect its additivity property .
for example , the following function @xmath265 is additive for all channels : @xmath266
the simplest functions that have not yet been studied before in the context of additivity are the piecewise linear functions .
more in detail , while linear functions @xmath267 are additive for all channels according to lemma [ lemaffine ] , the simplest examples of functions with unknown additivity properties are those functions @xmath39 which are the maximum of two linear functions , @xmath268 with @xmath269 , and @xmath270 to ensure continuity and convexity .
we call @xmath271 the _ kink _ of @xmath39 . fig .
[ abbkink ] shows what such functions look like . are those functions additive ? in this section ,
we give a partial answer to this question .
we first note a simple consequence of theorem [ thetensorstructure ] : there we have shown that if @xmath39 is additive , then @xmath207 must be additive as well .
it is natural to conjecture that more generally , @xmath272 must then always be additive for every @xmath273 $ ] . while it is not clear if this holds true in general , we can prove it for the case that @xmath39 is differentiable at zero : [ lemlambda ] let @xmath33\to{{\mathbb r}}$ ] be a convex function which is differentiable at zero . if @xmath201 is a class of channels which is closed with respect to tensor products with @xmath206 for all @xmath21 ( cf .
definition [ defchannelclass ] ) , and if @xmath39 is additive for @xmath201 , then the function @xmath209\mapsto \sum_{i=1}^n f(\mu_i\cdot x)\ ] ] is additive for @xmath201 as well for every sub - probability vector @xmath274 , i.e. if @xmath275 for every @xmath276 and @xmath277 .
in particular , @xmath272 is additive for @xmath201 for all @xmath273 $ ] . * proof . * without loss of generality , we may assume that @xmath72 , otherwise we can add some constant to @xmath39 without changing its additivity properties .
let @xmath278 , then @xmath279 is a probability vector for every @xmath280 . according to theorem [ thetensorstructure ] , the function @xmath281 must then be additive for @xmath201 for every @xmath280 . since @xmath39 is by assumption differentiable at zero , the limit @xmath282 exists and equals @xmath283 .
hence @xmath284 due to the closedness property of the additive functions as shown in lemma [ lemopencone ] , it follows that the function @xmath285 is additive for @xmath201 . but
@xmath286 is a linear function that we may subtract without affecting additivity due to example [ lemaffine ] . ' '' ''
@xmath64 we now use this lemma to prove our result on piecewise linear functions : if those functions are additive or not depends only on the location of the kink .
[ thepiecewiseaffine ] ' '' '' there is a global constant @xmath287 such that the following holds true : if @xmath39 is the maximum of two linear functions as plotted in fig .
[ abbkink ] , with kink at @xmath271 , then @xmath288 similarly , for every channel class @xmath201 which is closed with respect to tensor products with @xmath206 for all @xmath21 , there is a constant @xmath289 with the same property .
it is natural to conjecture that @xmath290 holds ; in this case , no function of this type would be additive for all channels . * proof .
* for simplicity , we assume that @xmath201 is the class of all channels ; the more general case is completely analogous .
it is sufficient to consider the piecewise linear functions @xmath291 since every function which is the maximum of two linear functions can be transformed into this form without affecting its additivity properties , if it has kink at @xmath271 .
explicitly , if @xmath39 is defined as in ( [ eqpiecewise ] ) , then the function @xmath135 defined by @xmath292 has this form , and shares the additivity property with @xmath39 due to example [ lemaffine ] .
thus , additivity of @xmath39 ( resp .
@xmath135 ) depends only on the location of the kink .
now suppose @xmath293 is additive for some @xmath294 $ ] . as @xmath293
is differentiable at zero , it follows from lemma [ lemlambda ] that @xmath295 is additive as well for every @xmath296 .
it is elementary to see that @xmath297 and so @xmath298 is additive , or equivalently @xmath299 for every @xmath300 .
this shows that there is some constant @xmath301 $ ] such that @xmath302 is additive if and only if the kink @xmath271 is larger than or equal to @xmath303 . finally ,
if @xmath304 , then @xmath302 is not additive according to example [ exhw ] .
this shows that @xmath305 . ' '' ''
@xmath64 here is a recipe how to improve the lower bound on @xmath303 ( or in the best case to prove that @xmath290 ) : find an example of a pair of channels such that the maximum output eigenvalue @xmath306 is attained at an entangled input state . then @xmath307 .
in fact , the proof above ( or rather its reference to example [ exhw ] ) exploits this fact for a pair of werner - holevo channels in dimension @xmath0 .
in this paper , we have studied the problem whether a given convex trace function , if it is applied to the output of a bipartite quantum channel , attains its maximum at an unentangled input state .
this problem generalizes the minimum output entropy additivity problem in a natural way : for example , there is a single channel bound on the output capacity ( lemma [ lemsinglechannel ] ) , additivity always holds if one of the channels is the identity channel ( example [ exidentity ] ) , and the study of the minimum output rank ( lemma [ lemminoutputrank ] ) and the largest output eigenvalue ( theorem [ thepiecewiseaffine ] ) have natural interpretations in our more general framework . in theorem [ hwtheorem ]
, we have shown that all operator convex functions on @xmath159 are additive for the werner - holevo channel in @xmath0 dimensions , which contains the well - known additivity results for this channel as special cases .
since the set of functions that are additive for all unitarily covariant channels is convex ( lemma [ lemconvexity ] ) , it is natural to conjecture that this set of functions can be classified further , possibly in a way as stated in conjecture [ conjaddop ] .
we have also shown some additional structural properties of the set of additive functions ( e.g. lemma [ lemopencone ] or theorem [ thetensorstructure ] ) , drawing new connections between functions like @xmath241 and the @xmath8-rnyi entropies , and also yielding partial reasons why von neumann entropy seems to play a special role for additivity ( cf .
lemma [ lemvonneumann ] ) . even though the original additivity conjecture has recently been disproved @xcite
, it is still interesting to study additivity for special classes of channels .
moreover , the transition from additivity to non - additivity ( say , the dimensionality of the channels ) is still not well understood , and the history of the additivity problem shows that introducing new entropy notions ( like @xmath8-rnyi entropy ) can be useful .
this is why we are confident that our framework of additive convex functions might be helpful in some instances of this problem . 0.25 cm * acknowledgments . *
the author would like to thank n. ay , j. eisert , d. gross , t. krger , r. seiler , a. szkoa , r. werner , and c. witte for helpful discussions and especially ra .
siegmund - schultze for his never - ending enthusiasm for the additivity conjecture .
t. cubitt , a. w. harrow , d. leung , a. montanaro , and a. winter , _ counterexamples to additivity of minimum output @xmath8-rnyi entropy for @xmath8 close to @xmath104 _ , commun .
phys . * 284 * , 281 ( 2008 ) . | we study a natural generalization of the additivity problem in quantum information theory : given a pair of quantum channels , then what is the set of convex trace functions that attain their maximum on unentangled inputs , if they are applied to the corresponding output state ?
we prove several results on the structure of the set of those convex functions that are `` additive '' in this more general sense .
in particular , we show that all operator convex functions are additive for the werner - holevo channel in @xmath0 dimensions , which contains the well - known additivity results for this channel as special cases
. 0.5 cm |
cosmological models with a scalar field and an exponential potential are of fundamental importance in the study of the universe . these models are motivated by the fact that they arise naturally in alternative theories of gravity and occur as the low - energy limit in supergravity theories @xcite . by using qualitative techniques ,
the well - known power - law inflationary solution has been shown to be an attractor for all initially expanding bianchi models ( except a subclass of the bianchi type ix models which will recollapse ) in the class of spatially homogeneous bianchi models @xcite . more recently
, cosmological models which contain both barotropic matter and a scalar field with an exponential potential have been studied @xcite , partially motivated by the fact that there exist spatially flat isotropic scaling solutions in which the energy density due to the scalar field is proportional to the energy density of the perfect fluid @xcite . in @xcite the stability of these cosmological scaling solutions within the class of
spatially homogeneous cosmological models with a perfect fluid subject to the equation of state @xmath2 ( where @xmath3 is a constant satisfying @xmath4 ) was studied and it was found that when @xmath5 , and particularly for realistic matter with @xmath6 , the scaling solutions are unstable ; essentially they are unstable to curvature perturbations , although they are stable to shear perturbations .
in addition , in @xcite homogeneous and isotropic spacetimes with non - zero spatial curvature were studied .
it is clearly of interest to study more general cosmological models .
one class of models of particular interest are those with _ positive spatial curvature_. these models have attracted less attention since they are more complicated mathematically .
positive - curvature friedmann - robertson - walker ( frw ) models @xcite , kantowski - sachs models @xcite and bianchi type ix models @xcite have been studied using qualitative methods , although rigorous analyses using a new set of compact variables have not been carried out .
the bianchi type ix models are known to have very complicated dynamics , exhibiting the characteristics of chaos @xcite , and are hence beyond the scope of the present study .
recently @xcite positive - curvature frw models and kantowski - sachs models with a perfect fluid and a cosmological constant have been investigated using qualitative methods and utilizing compactified variables .
the outline of this paper is as follows .
we shall first comprehensively study the qualitative properties of the class of positive - curvature frw models with a barotropic fluid and a non - interacting scalar field with an exponential potential , extending and generalising work by turner @xcite , who used different basic variables .
we shall then analyse the qualitative properties of the kantowski - sachs models .
positive - curvature frw models and kantowski - sachs models belong to the class of spherically symmetric models , and hence the present work is a natural extension of recent work @xcite .
indeed , it turns out that understanding the dynamics of the kantowski - sachs models is crucial for understanding the global dynamics of general spherically symmetric similarity models @xcite .
the matter content of the models is taken to be a perfect fluid and a scalar field with exponential potential .
the corresponding energy - momentum tensor is @xmath7 where @xmath0 is a non - negative constant , and the pressure is given by @xmath2 , with the equation - of - state parameter in the range @xmath8 . the fluid energy density @xmath9 and the scalar field @xmath10 are functions of a timelike coordinate @xmath11 .
a dot denotes differentiation with respect to @xmath11 and throughout units are used in which @xmath12 .
the matter components are assumed to be non - coupled , and thus they are separately conserved : @xmath13 for convenience , we define @xmath14
we start our investigation of closed cosmological models with a perfect fluid and a scalar field by looking at the closed frw models .
the line element for these models can be written @xmath15 the expansion of the fluid congruence is given by @xmath16 , and the evolution equation for the curvature @xmath17 is @xmath18 the conservation equations yield @xmath19 from the field equations we obtain @xmath20 \label{eq : friedmannfrw},\\ \dot{\theta}&=&-\frac{1}{3}\left\{\theta^2+\frac{1}{2}d^2 + \frac{3}{2}\left[3x^2 - 3v+3(\gamma-1)\mu\right]\right\ } .\end{aligned}\ ] ] assuming @xmath21 , the friedmann equation , eq .
( [ eq : friedmannfrw ] ) , shows that @xmath22 is a dominant quantity .
thus , compact variables can be defined according to @xmath23 note also that the curvature is given by @xmath24 the friedmann equation becomes @xmath25 defining a new independent variable , @xmath26 , the evolution equation for @xmath27 @xmath28 decouples .
thus , a reduced set of evolution equations is obtained : @xmath29 ,
\nonumber\\ u'&=&3q_0u\left[-1+\left(u^2+\frac{\gamma}{2}\omega_d\right)\right]+ \sqrt{\frac{3}{2}}\kappa w^2 , \label{eq : dynsysfrw}\\ w'&=&3q_0\left(u^2+\frac{\gamma}{2}\omega_d\right)w - \sqrt{\frac{3}{2}}\kappa uw \nonumber.\end{aligned}\ ] ] there is also an auxiliary evolution equation @xmath30\omega_d , \ ] ] and it is straight - forward to consider the set of variables @xmath31 , rather than @xmath32 @xcite . note that by setting @xmath33 , and identifying @xmath34 , @xmath35 , the evolution equations corresponding to closed frw models with a cosmological constant are obtained @xcite .
it is also useful to consider the deceleration parameter , given by @xmath36 , \ ] ] for @xmath37 . from this expression
, we can see that there is an inflationary region ( @xmath38 ) in the state space whenever @xmath39 .
however , as will be seen below , it is only when @xmath1 that there exist attractors that are inflationary .
note also that for @xmath40 , @xmath41 so that @xmath42 whenever @xmath43 in which case @xmath44 is itself monotonic . when @xmath38 , that is in the inflationary region , @xmath44 need not be monotonic for example , see the orbits close to @xmath45 in figs .
[ fig : frw0 ] and [ fig : frw1 ] .
the dynamical system eqs .
( [ eq : dynsysfrw ] ) is symmetric under the transformation @xmath46 thus , it is sufficient to discuss the behaviour in one part of the state space , the dynamics in the other part being obtained by eq .
( [ eq : symmfrw ] ) . furthermore , note that @xmath47 is a monotonic function in the regions @xmath48 and @xmath49 for @xmath50 . as there are no equilibrium points with @xmath51 when @xmath5 , @xmath52 acts as a monotonic function in the interior of the state space .
consequently there can be no periodic or recurrent orbits in the interior state space and global results can be deduced .
in addition , from the expression for the monotonic function @xmath52 we can see immediately that either @xmath53 or @xmath54 asymptotically .
a number of equilibrium points can be found for the dynamical system , eqs .
( [ eq : dynsysfrw ] ) . in
what follows , @xmath55 denotes the sign of @xmath44 , whereas @xmath56 is a density parameter associated with the scalar field . in table [ tab : equifrw ] , the various equilibrium points are summarized .
the subscripts on the labels have the following significance : the left subscript gives the sign of @xmath44 and indicates whether the corresponding model is expanding ( + ) or contracting ( @xmath57 ) .
the right subscript gives the sign of @xmath58 ; i.e. , the sign of @xmath59 .
.equilibrium points of the closed frw models . [ cols="<,<,^,^,^,<",options="header " , ] lll + expanding from a singularity ( @xmath49 ) & @xmath60 & @xmath61 + contracting from a dispersed state ( @xmath48 ) & @xmath62 & @xmath1 + + contracting to a singularity ( @xmath48 ) & @xmath63 & @xmath61 + expanding to a dispersed state ( @xmath49 ) & @xmath45 & @xmath1 + the eigenvalues for each of the equilibrium points are given in table [ tab : eigenks ] .
the sources and sinks of the dynamical system when @xmath5 are listed in table [ tab : stabks ] ( all of the other equilibrium points are saddles ) .
thus , there is always two segments of the equilibrium set @xmath60 that act as sources for orbits .
similarly there are two segments on @xmath63 that are sinks . when @xmath64 , these are the only attractors , and all solutions start from and recollape to a singularity ( @xmath65 ) . when @xmath1 , which then implies that @xmath66 for @xmath5 , the equilibrium points @xmath67 are attractors .
thus , for @xmath1 , there are also ever - expanding ( @xmath68 ) and ever - collapsing ( @xmath69 ) solutions . from the expression for the monotonic function @xmath52 we deduce that _
all _ orbits asymptotically have @xmath70 , @xmath53 or @xmath54 .
indeed , the existence of the monotonic function ensures that there are no periodic orbits and that generically orbits asymptote towards the local attractors ( sinks and sources ) .
therefore , we can determine the global dynamics of the models .
to * summarize * , when @xmath64 all solutions start from and recollape to a singularity ( @xmath65 ) .
thus , in this case solutions can neither isotropize nor inflate .
when @xmath1 , there are also ever - expanding ( @xmath71 ) ( and ever - collapsing @xmath72 ) solutions in addition to the recollapsing solutions .
again , the @xmath73 points correspond to power - law inflation when @xmath1 . consequently , in this case there is a subclass of solutions that isotropize and inflate .
the global asymptotic dynamics is similar to that in the case of positive - curvature frw models .
however , due to the presence of shear , the intermediate or transient dynamics can be quite different . in the kantowski - sachs case
the state space is four - dimensional and so we can not display the phase portraits graphically ( as in the frw case ) .
however , as an illustration we shall present the phase portraits in the three - dimensional fluid vacuum and the massless scalar field invariant sets in order to compare intermediate behaviours .
the fluid vacuum ( @xmath74 ) is an invariant submanifold , as seen from eq .
( [ eq : om ] ) . using the friedmann equation to eliminate @xmath75
, we obtain a three - dimensional dynamical system in @xmath76 : @xmath77 , \\
q_+'&=&-(1-q_0 ^ 2)(1-q_+^2)-3q_0q_+(1-q_+^2-u^2 ) , \\
u'&=&(1-q_0 ^ 2)q_+u+ \left(\sqrt{\frac{3}{2}}\kappa-3q_0u\right)(1-q_+^2-u^2 ) .\end{aligned}\ ] ] from table [ tab : qks ] , it is immediately seen that the equilibrium points that are contained in this submanifold are @xmath78 , @xmath67 , and @xmath79 .
the state space is depicted in figs . [
fig : ksvac1 ] , [ fig : ksvac2 ] and [ fig : ksvac3 ] .
note that @xmath80 , where @xmath81 is defined by @xmath82 is an invariant submanifold , and that both @xmath67 and @xmath79 are contained in this submanifold .
the massless case corresponds to the invariant submanifold @xmath83 , which leads to a three - dimensional system in ( @xmath84 ) : @xmath85 , \nonumber \\
q_+'&=&-(1-q_0 ^ 2)(1-q_+^2 ) -\frac{3}{2}(2-\gamma)q_0q_+\omega_d , \\
u'&= & u\left[(1-q_0 ^ 2)q_+-\frac{3}{2}(2-\gamma)q_0\omega_d\right ] .
\nonumber \end{aligned}\ ] ] from table [ tab : equiks ] , it is immediately seen that the equilibrium points that are contained in this submanifold are @xmath78 and @xmath86 .
the state space is depicted in fig .
[ fig : ksmassless ] .
we have studied closed cosmological models with a perfect fluid satisfying a linear equation of state with @xmath87 and a scalar field with an exponential potential . we have utilized a new set of normalised variables which lead to the compactification of state space , enabling us to apply the theory of dynamical systems to determine the qualitative properties of the models . in all cases
we have been able to find monotonic functions which , together with a local analysis of the equilibrium points , enable us to determine the global properties of the models .
we first studied the closed frw cosmological models .
we found that when @xmath64 , all solutions start from and recollape to a singularity ( @xmath65 ) . in this case
solutions generically do not inflate .
when @xmath1 , solutions can either recollapse ( @xmath65 ) or expand forever ( @xmath71 ) towards power - law inflation solutions ( or collapse forever @xmath72 ) ; consequently , in this case there is a subclass of solutions that inflate .
a number of phase portraits were displayed .
these results generalise previous qualitative work on positive - curvature frw models with a scalar field ( only ) @xcite and with a scalar field plus a barotropic perfect fluid @xcite in which compactified variables were not utilized , and rigorous analyses of perfect fluid ( only ) models using compactified variables @xcite , and completes and generalises more recent work using different compactified variables @xcite .
we also note that positive - curvature frw models with a perfect fluid and a positive cosmological constant have been investigated recently using qualitative methods and utilizing compactified variables @xcite . in the case of the kantowski - sachs models we again found that when @xmath64 all solutions start from and recollape to a singularity ( @xmath65 ) and can consequently neither isotropize nor inflate .
when @xmath1 , there are also ever - expanding ( @xmath71 ) ( and ever - collapsing @xmath72 ) solutions in addition to the recollapsing solutions , where again the @xmath73 points correspond to the flat frw power - law inflationary solution . consequently , in this case there is a subclass of solutions that isotropize and inflate .
the investigation of kantowski - sachs models complements the study of bianchi models @xcite and completes the analysis of spatially homogeneous models .
collins @xcite studied perfect fluid kantowski - sachs models qualitatively using expansion - normalised variables ( for which the state space was non - compact ) and showed that all models start at a big bang and recollapse to a final `` big crunch '' singularity .
this work was generalised recently by goliath and ellis @xcite in which kantowski - sachs models with a perfect fluid and a cosmological constant were investigated using qualitative methods and utilizing the compactified variables of uggla and zur - muhlen @xcite ; particular attention was focussed upon whether the models isotropize , thereby explaining the presently observed near - isotropy of the universe .
more importantly , kantowski - sachs models with a scalar field and an exponential potential , but without barotropic matter , have been studied qualitatively @xcite , although compactified variables were not utilized . to conclude an analysis of positive - curvature spatially homogeneous cosmological models with a perfect fluid and a scalar field with an exponential potential
, bianchi type ix models would need to be studied .
however , such a study is beyond the scope of the current paper .
for example , bianchi type ix models are known to have very complicated dynamics , exhibiting the characteristics of chaos @xcite . however ,
partial results are known .
bianchi type ix models with a scalar field ( only ) have been studied qualitatively , with an emphasis on whether these models can isotropize @xcite . scalar - field models with matter have also been studied @xcite . for example , it has been shown that the power - law inflationary solution is an attractor for all initially expanding bianchi type ix models except for a subclass of the models which recollapse @xcite .
however , compact variables have not been utilized and the analyses were not rigorous . a more rigorous treatment of the class of bianchi type ix models with a non - tilted perfect fluid ( only ) using compactified variables has been possible @xcite . although an appropriately defined normalised hubble variable is found to be monotonic , enabling some results to be obtained , several problems remain open .
more rigorous global results are possible . for example , bianchi type ix models with matter have been shown to obey the `` closed universe recollapse '' conjecture @xcite , whereby initially expanding models enter a contracting phase and recollapse to a future `` big crunch '' .
in addition , ringstrm has proven that a curvature invariant is unbounded in the incomplete directions of inextendible null geodesics for generic vacuum bianchi models @xcite , and rigorously shown that the mixmaster attractor is the past attractor of bianchi type ix models with an orthogonal perfect fluid @xcite . a complete qualitative analysis of the special class of locally rotationally symmetric bianchi type ix
perfect fluid models , which do not exhibit oscillatory or chaotic behaviour near to the initial or final singularities , has been given in @xcite , based upon an appropriately defined set of bounded variables .
the kantowski - sachs models exhibit similar global properties to the positive - curvature frw models ; in particular , for @xmath64 all initially expanding models reach a maximum expansion and thereafter recollapse , whereas for @xmath1 models generically recollapse or expand forever towards a flat isotropic power - law inflationary solution .
the bianchi type ix models share these qualitative properties
. however , the intermediate behaviour of the kantowski - sachs models can be quite different to that of the frw models . in order to illustrate the possible intermediate dynamics of the kantowski - sachs models
, we studied the special cases of no barotropic fluid , and a massless scalar field in secs .
[ sec : vac ] and [ sec : massless ] , respectively ( see figs . [
fig : ksvac1 ] [ fig : ksmassless ] ) .
finally , we remark that the dynamics of the kantowski - sachs models obtained here will be crucial for understanding the global dynamics of general self - similar spherically symmetric models @xcite .
we would like to thank peter turner for his help with the analysis of the positive - curvature frw models @xcite .
ac would like to acknowledge financial support from nserc of canada .
mg would like to thank the department of mathematics and statistics at dalhousie university for hospitality while this work was carried out . | closed , spatially homogeneous cosmological models with a perfect fluid and a scalar field with exponential potential are investigated , using dynamical systems methods .
first , we consider the closed friedmann - robertson - walker models , discussing the global dynamics in detail .
next , we investigate kantowski - sachs models , for which the future and past attractors are determined .
the global asymptotic behaviour of both the friedmann - robertson - walker and the kantowski - sachs models is that they either expand from an initial singularity , reach a maximum expansion and thereafter recollapse to a final singularity ( for all values of the potential parameter @xmath0 ) , or else they expand forever towards a flat power - law inflationary solution ( when @xmath1 ) . as an illustration of the intermediate dynamical behaviour of the kantowski - sachs models ,
we examine the cases of no barotropic fluid , and of a massless scalar field in detail .
we also briefly discuss bianchi type ix models . |
the low energy limit of string theory provides a definite form of dilaton gravity action @xcite with a specific dilaton coupling and additional 2-form fields @xcite .
thus , it may be worth examining the exact cosmological solutions of the 4-dimensional low energy string gravity in the framework of dilaton gravity , instead of general relativity .
there have been many studies on the cosmological evolution of the low energy string gravity with 2-form fields @xcite . in the early cosmological evolution , it is necessary to consider the effect of matter as well as the 2-form fields in the low energy string gravity . without matter , it would be convenient to choose the einstein frame redefining the metric as usual
. however , in the presence of matter , the frame changes of metric make the dilaton field couple with matter fields differently depending on their spins .
thus , the matter energy - momentum tensors would be non - trivial functions of the dilaton field in a different frame .
therefore , we have considered a fluid - like matter distribution in the fundamental string frame @xcite , in which the gravitational constant determined by the dilaton field evolves in time . because the relativistic matter should be dominant over the non - relativistic matter in the early evolution of the universe , we have examined the exact cosmological solutions with the 2-form fields and/or the radiation - like fluid in the string frame .
the vanishing beta functions originated from the world sheet conformal symmetry in the closed string sector of string theory give the well - known low energy string gravity action in the critical @xmath0-dimension @xcite . assuming a ricci - flat compactification of the internal @xmath1-dimensional space decoupled from our @xmath2-dimensional spacetime @xcite , and adding the matter lagrangian @xmath3 which is decoupled from the dilaton field in the string frame
, we have the following form of action in 4-dimensional spacetime from the low energy limit of string theory @xcite , @xmath4 + \int d^{4}x\sqrt{-g } { \cal l}_{m } , \label{action}\end{aligned}\ ] ] where we have redefined the dilaton field @xmath5 with @xmath6 as @xmath7 in the string frame , and @xmath8 are the field strengths of the 2-form antisymmetric fields @xmath9 in ns - ns sector as @xmath10 . the equations of motion for the 2-form fields @xmath9 , the dilaton field @xmath6 , and the metric @xmath11 are obtained as @xmath12 @xmath13 @xmath14 where @xmath15 is the matter energy - momentum tensor . taking the trace part of equation ( [ g - eq ] ) , we have an expression for @xmath16 . plugging it into the dilaton equation ( [ d - eq ] ) , we can rewrite the dilaton equation of motion equivalently as @xmath17 where @xmath18 is the trace of the matter energy - momentum tensor . as we can see from eq.([s - eq ] ) , the driving force of the dilaton field is given by the antisymmetric 2-form field strengths and the trace of matter energy momentum tensor . @xmath8 solution of eq.([h - eq ] ) is found as @xmath19 where @xmath20 is a new smooth scalar field , @xmath21 is the covariant constant antisymmetric tensor such that @xmath22 , with @xmath23 .
the bianchi identity @xmath24 for the antisymmetric field strengths @xmath8 gives the equation of motion for the scalar field @xmath20 as @xcite @xmath25 defining the dual strength vector @xmath26 as @xmath27 , we find that @xmath28 we examine the cosmological evolution of the eqs.([h - eq],[d - eq],[g - eq ] ) with the spatially flat robertson - walker metric @xmath29 and the cosmological ansatz : @xmath30 , @xmath31 .
then , eq.([psi - eq ] ) gives @xmath32 where @xmath33 is a constant which measures the strength of the 2-form fields .
the solution eq.([h - sol2 ] ) gives a uniform non - vanishing dual strength vector such as @xmath34 using the metric ( [ rw ] ) and the explicit solution of @xmath35 from eqs.([h2-sol],[b - sol ] ) , eqs.([g - eq],[s - eq ] ) are reduced to a set of cosmological equations . denoting the time derivative as an over - dot ,
the cosmological equations with uniform matters in the string frame are given as @xmath36 @xmath37 @xmath38 where the hubble parameter is @xmath39 , and the cosmic acceleration parameter is @xmath40 . with the uniform matter density @xmath41 and the uniform matter pressure @xmath42 , the energy momentum tensors are @xmath43 , and @xmath44 . in eqs.([cosg00-eq],[cosg11-eq ] ) , it seems that the 2-form fields play a role of matter with the energy density @xmath45 and the pressure @xmath45 having the equation of state @xmath46 .
however , the @xmath47 term in the action ( [ action ] ) gives an additional contribution , the conformal density @xmath48 defined as @xmath49 , to the right hand side of eq.([cosd - eq ] ) . by taking a time derivative eq.([cosg00-eq ] ) and using eqs.([cosg11-eq],[cosd - eq ] )
, it is found that the matter energy conservation holds as @xmath50 which is independent of the 2-from fields @xmath9 and the dilaton field @xmath6 in the string frame . with the equation of state for matter @xmath51 , eq.([matter - eq ] )
gives the matter density evolving as @xmath52 where @xmath53 is a positive constant which measures the strength of the matter density .
the 2-form fields having @xmath54 behavior in eqs.([cosg00-eq],[cosg11-eq],[cosd - eq ] ) dominate the universe with a small scale factor @xmath55 .
as the universe expands , the rapidly decreasing 2-form fields lay down their role in the late time evolution of the universe , and matter with the density ( [ matter - sol ] ) or a cosmological constant will take over the dominance .
first , let us begin with the vacuum solution of eqs.([cosg00-eq],[cosg11-eq],[cosd - eq ] ) , without the 2-form fields and matter .
the equation ( [ cosd - eq ] ) can be written as @xmath56 with a new time function @xmath57 such that @xmath58 , this equation becomes a simple equation having a general linear solution in @xmath59 , @xmath60 where @xmath61 and @xmath62 are two positive integration constants which should be determined by an initial condition .
if @xmath63 , then we have the trivial vacuum solution such that @xmath64 is a constant , and the scale factor @xmath65 is also a constant with @xmath66 . if @xmath67 , then we have @xmath68 the new time function @xmath59 is a monotonically increasing function of time @xmath69 . as the time
@xmath69 elapses , @xmath59 increases rapidly for a small scale factor , but very slowly for a large scale factor @xmath65 .
thus , the gravitational constant , which is the inverse of @xmath70 , decreases as @xmath71 in vacuum . from eq.([cosg00-eq ] ) , we obtain @xmath72 and its integration as @xmath73 where @xmath74 is a positive integration constant . thus , using @xmath75 , we obtain the cosmological time @xmath69 as a function of @xmath59 as @xmath76 where @xmath77 is a constant time .
thus , we find the exact vacuum solutions without matter and the 2-form fields as @xmath78 where @xmath79 and @xmath80
. therefore , the both vacuum solutions with @xmath67 give an ever shrinking universe to @xmath81 in the future . especially , the solution with the lower signs meets a curvature singularity at a finite time @xmath82 .
let us consider the solution of eqs.([cosg00-eq],[cosg11-eq],[cosd - eq ] ) with the 2-form fields only .
the equation ( [ cosd - eq ] ) can be written as @xmath83 with a new time function @xmath57 such that @xmath58 , this equation becomes the simple harmonic equation in @xmath59 , @xmath84 where @xmath85 is an integration constant , and @xmath86 .
we find that @xmath59 should be in a compact domain , @xmath87 . from eq.([cosg00-eq ] ) , we have the solution of @xmath88 as follows , @xmath89 using the solution ( [ ab - phi - sol ] ) of @xmath6 , from eq.([htau - sol ] ) , we find the velocity of the cosmic scale factor as follows @xmath90 we can integrate eq.([ab - vel - eq ] ) to have the scale factor @xmath65 as a function of @xmath59 @xmath91 for the upper and the lower signs of eq.([htau - sol ] ) respectively
. the cosmological time @xmath92 for each solution is given by the integration over @xmath93 , @xmath94 the infinite future of the cosmological time @xmath69 corresponds to @xmath95 .
the cosmological time @xmath96 and the scale factor @xmath97 in the einstein frame are obtained as latexmath:[\[\begin{aligned } t_{e}(\tau)= \int_{0}^{\tau } \phi(\tau ) s_{\pm}^{3}(\tau ) d\tau ,
\quad s_{e}(\tau)= \phi(\tau ) s_{\pm}(\tau ) = because the time derivative of @xmath97 about @xmath96 is given as @xmath99 it is found that the upper / lower sign solution corresponds to an ever expanding / shrinking solution , respectively , in the einstein frame .
the second derivative of @xmath97 about @xmath96 is the same for the both solutions , and always negative as @xmath100 thus , both solutions decelerate always in the einstein frame .
the exact solutions in 4-spacetime dimension seem to be consistent with the phase - plane analysis @xcite , and with the analytic fourier mode solutions @xcite .
next , let us examine the cosmological solution of eqs.([cosg00-eq],[cosg11-eq],[cosd - eq ] ) with the radiation - like fluid in the string frame . because the uniform relativistic matter has the equation of state @xmath101 with @xmath102 , the trace of the matter energy momentum tensors vanishes
thus , without the 2-form fields @xmath103 , the dilaton equation ( [ cosd - eq ] ) gives the same solution ( [ phi - vac - sol ] ) as in the vacuum .
if @xmath63 , then we have the well - known decelerating expansion solution with a constant dilaton field @xmath64 as in general relativity .
thus , let us consider the @xmath67 case , @xmath104 . using a relativistic matter density @xmath41 and a pressure density @xmath42 such that @xmath105 from eq.([matter - sol ] ) , and with the dilaton solution ( [ phi - vac - sol],[phidot - vac - sol ] ) ,
we obtain the following equation for the scale factor @xmath65 from eq.([cosg00-eq ] ) , @xmath106 introducing another time function @xmath107 which is also an increasing function of the time @xmath69 , we can rewrite this equation as follows , @xmath108 integrating this , we have the solution for the time parameters @xmath107 as @xmath109 where @xmath110 is an integration constant .
thus , the range for the time @xmath107 is determine as @xmath111 from eq.([t - sol ] ) , we have a relation between the two time parameters @xmath107 and @xmath59 as @xmath112 integrating this equation , the solution for the upper signs is obtained as @xmath113 where @xmath114 is a positive integration constant , and the integration constant @xmath110 may be chosen as @xmath115 such that @xmath116 .
the scale factor @xmath117 for the upper signs is a downward concave function in time , having a minimum as shown in fig.1 . versus @xmath118 with @xmath119 in eq.([a0-s - sol]),width=259 ]
the cosmological time @xmath69 is related to the time parameter @xmath107 as follows @xmath120 integrating this equation , it is found that the cosmological time @xmath69 is a monotonically increasing function of the time parameter @xmath107 with @xmath121 at @xmath122 as shown in fig.2 . versus @xmath118 from eq.([time - eq]),width=259 ] for the upper sign bounce solution ( [ a0-s - sol ] ) , as the time @xmath107 approaches to @xmath123 , the dilaton field vanishes and the scale factor grows infinitely .
therefore , the gravitational constant grows infinitely as the cosmological time approaches to the initial cosmological time @xmath121 .
we can examine the cosmological evolution with the help of eq.([cosg11-eq ] ) also .
using the solution ( [ phi - vac - sol ] ) of @xmath6 and eq.([a0-vel - eq ] ) , we find the acceleration of the scale factor from eq.([cosg11-eq ] ) as @xmath124 for the lower signs in eqs.([a0-vel - eq],[a0-acc - eq ] ) , the velocity and the acceleration of the scale factor are both negative always , giving an ever shrinking universe to meet the singularity with the scale factor @xmath81 in the future .
however taking the upper signs gives the interesting bounce solution as we have obtained in eq.([a0-s - sol ] ) . from eqs.([a0-vel - eq],[a0-acc - eq ] ) , the velocity and the acceleration of the scale factor in the time @xmath69 show the following behavior , @xmath125 thus , the scale factor should have a downward concave form in time as seen in fig.1 , having a minimum size at @xmath126 .
if a universe started its evolution at a finite size with @xmath127 , it would shrink with a positive acceleration to the minimum size at @xmath128 , which is the turning point .
then , the universe begins an accelerating expansion again until @xmath129 , and continues a decelerating expansion afterward . if we choose @xmath130 , @xmath131 as the value satisfying the equation @xmath132 .
then , the dilaton field @xmath70 increases from the value @xmath133 at the minimum size @xmath130 , and approaches to a constant value @xmath134 finally in the expanding universe . at the minimum size @xmath130
, the matter density has the maximum value @xmath135 as @xmath136 unlike other examples of bouncing scenarios in string cosmology @xcite , the bouncing solution found with the radiation - like fluid decoupled from the dilaton field in the string frame does not satisfy the property of duality symmetry which is typical of string cosmology @xcite .
finally , we consider the solutions of eqs.([cosg00-eq],[cosg11-eq],[cosd - eq ] ) with both the radiation - like fluid and the 2-form fields . the equation ( [ cosd - eq ] ) with the 2-form fields and the radiation - like fluid gives the same solution ( [ ab - phi - sol ] ) of the case with the 2-form fields only . from eq.([cosg00-eq ] ) , we have the solution of @xmath88 as follows , @xmath137 using the solution ( [ ab - phi - sol ] ) of @xmath6 , from eq.([htau - sol ] ) we find the velocity of the cosmic scale factor as follows @xmath138 similarly as in eq.([a0-vel - eq1 ] ) , introducing a new time function @xmath139 , we can rewrite this equation as @xmath140 integrating this , we have the solution for the upper signs as @xmath141 where @xmath114 is a positive integration constant , and the integration constant @xmath110 may be also chosen as @xmath115 such that @xmath142 .
this scale factor @xmath143 for the upper signs is a downward concave function in time , having a minimum size at @xmath144 , like the previous upper sign solution for the relativistic matter only case .
thus , it is found that adding the 2-form fields does not change the qualitative behaviour of the exact solutions ( [ a0-s - sol ] ) for the radiation - like fluid only .
we have discussed a particular class of exact cosmological solutions of the 4-dimensional low energy string gravity with the 2-form fields and/or the uniform relativistic matter in the string frame ( see @xcite for the general exact solutions with arbitrary equation of state of the fluid sources ) . in the vacuum without matter and the 2-form fields , the exact cosmological solutions always give monotonically shrinking universes if the dilaton field is not a constant .
however , in the presence of a radiation - like fluid and/or the 2-form fields , the exact solutions exhibit a minimum size of the universe in the string frame . throughout the evolution
, the exact solutions do not show any singularity , except the initial cosmological curvature singularity in the string frame . as further studies , in the early stage of cosmological evolution
, it may be necessary to consider the effects of the cosmological constant or a dilaton potential with the radiation - like fluid .
furthermore , in the late stage of the cosmological evolution , the effects of the dominant non - relativistic fluid should be considered also . especially , after the transition from the radiation - dominated to the matter - dominated regime , the qualitative behaviour of the considered solutions is expected to change in the latest stage of the cosmological evolution , and the dilaton field might be relevant for phenomenological applications , such as a possible interpretation of the dilaton as the cosmic dark - energy field @xcite . | we have discussed a particular class of exact cosmological solutions of the 4-dimensional low energy string gravity in the string frame . in the vacuum without matter and the 2-form fields
, the exact cosmological solutions always give monotonically shrinking universes if the dilaton field is not a constant .
however , in the presence of the 2-form fields and/or the radiation - like fluid in the string frame , the exact cosmological solutions show a minimum size of the universe in the evolution , but with an initial cosmological curvature singularity in the string frame . |
the van der waals gradient theory @xcite was originally developed at the end of the 19-th century as an `` effective '' physical picture to describe the critical region .
this theory considers a local helmholtz free - energy composed of two parts .
the first one describes the homogeneous behaviour of the system , while the second part characterizes the non - homogeneous one . in addition , the first part is proportional to the fluid density @xmath0 while the second one is proportional to the square of the density gradient .
much effort has been devoted to apply the van der waals gradient theory to various fluids . at present , it can be considered as the most simple and comprehensive model in the physics of interfaces and capillarity .
the associated hydrodynamic equations can be written as @xmath1 } , \label{korteweg}\end{aligned}\ ] ] where @xmath2 is a function of the density @xmath0 and of @xmath3 @xcite and @xmath4 is the current velocity .
this formulation of the van der waals theory was originally due to korteweg @xcite , who proposed a continuum mechanical model in which the cauchy stress tensor apart from the standard cauchy - poisson term contains an additional term defined as @xmath5 where @xmath6 is the unit tensor . as already mentioned by dunn and serrin @xcite , the modern terminology concerning the korteweg model refers to elastic materials of grade @xmath7 , where the particular case of @xmath8 has been well studied in recent years @xcite .
equations ( [ continuity ] ) and ( [ korteweg ] ) have been linked recently @xcite to a nonlinear schrdinger equation viewed as a particular set of hydrodynamic equations describing the so - called nonlinear madelung fluid @xcite .
this link between capillarity and the schrdinger equation can shed more light onto the so - called quantum - like approach to many - particle systems such as beams in particle accelerators and beam - plasma systems . the standard procedure in this direction is to approximate the physical systems characterized by an overall interaction with a suitable van der waals mean field theory . to avoid misunderstanding ,
it is worthwhile to note that schrdinger equation alone does not provide an entire quantum mechanical picture . it should be necessarily supplemented by a theory of quantum measurement and consequently by a proper physical interpretation of wave packets . in the quantum - like approach ,
the many - particle systems are described in an effective way as a whole .
based on the above considerations , it appears interesting and attractive to explore the possibility of a rigorous derivation from kinetic theory of the general hydrodynamic picture thus discussed .
the starting point of our analysis is the equation for the microscopic phase space density @xmath9 @xmath10 } = 0 , \label{micphspdens}\ ] ] for a system consisting of @xmath11 particles , which occupies volume @xmath12 in the configuration space . here
@xmath13 and @xmath14 are the coordinates and the canonically conjugate momenta , @xmath15 is the particle mass and @xmath16 is the microscopic force , which apart from the external force includes a part specifying the type of interaction between particles .
suppose that at some initial time @xmath17 the microscopic phase space density is known to be @xmath18 .
then , the formal solution of equation ( [ micphspdens ] ) for arbitrary time @xmath19 can be written as @xmath20 where @xmath21 is the evolution operator , specifying the hamiltonian flow .
the choice of the initial @xmath22 is based on the knowledge of the microscopic characteristics of the system . due to the extremely complex particles dynamics ,
full consistent description is not feasible .
therefore , the detailed information on the microscopic level is incomplete .
if our system is a complex one in the sense that both the external forces and the collective forces are highly nonlinear , a dynamic instability of motion is likely to occur on a characteristic time scale @xmath23 .
the only information available to an outside observer by means of a macroscopic measuring device is a coarse - grained density distribution with a smoothing function , which takes into account the dynamic instability of motion .
thus , we assume @xmath24 where for simplicity the explicit dependence on the momentum variables @xmath14 has been suppressed . to take into account the initial preparation of the system , one has to displace the initial time @xmath17 at @xmath25 and perform an average over the past history of the system
. then equation ( [ micphspdens ] ) becomes @xcite @xmath26 } \nonumber \\ % \end{equation } % % % % % \begin{equation } = { \frac { 1 } { \tau } } { \left ( { \widetilde{n}}_m - n_m \right)}. \label{micphspdensms}\end{aligned}\ ] ] since the collision time is supposed to be much smaller than the time @xmath23 , the collision integral can be dropped and the kinetic equation for the one - particle distribution function @xmath27 can be written as @xmath28 the right - hand - side of equation ( [ onepartdfsms ] ) is regarded as a `` collision integral '' , and it can be represented as @xmath29 } , \label{sdcolintgen } % \end{equation}\end{aligned}\ ] ] where @xmath30 with @xmath31 . as a first very interesting step , we consider the diffusion approximation @xmath32 } + { \frac { 1 } { 2 } } \nabla_k \nabla_l { \left [ b_{kl } { \left ( { \bf x } ; t \right ) } f \right ] } , \label{sdcolintil}\ ] ] where @xmath33 and a summation over repeated indices is implied . for hamiltonian systems the well - known relation @xmath34 holds , which gives @xmath35 for @xmath36 . in passing , it is worthwhile to mention that using the principle of maximum information entropy formulated by jaynes , it can be shown @xcite that the smoothing function @xmath37 is of the form @xmath38}. \label{sdsmoothfun}\end{aligned}\ ] ] the quantities @xmath39 are the first and the second moment of @xmath40 at the instant of time @xmath41 , provided that @xmath40 measured at the instant @xmath19 equals @xmath13 [ i.e. @xmath42 .
in addition , @xmath43 is the covariance matrix defined as @xmath44}. \label{sdcovarmat}\ ] ] the generalized kinetic equation ( [ onepartdfsms ] ) has a form analogous to the bhatnagar - gross - krook ( bgk ) equation , widely used in the kinetic theory of gases @xcite .
there is however , an important conceptual difference between the two equations . in the bgk equation
the function @xmath45 should be replaced by the equilibrium distribution function @xmath46 describing the global equilibrium and the characteristic time @xmath23 should be replaced by the corresponding relaxation time . the smoothed distribution function in equation ( [ onepartdfsms ] )
characterizes a local quasi - equilibrium state within the smallest unit cell of continuous medium , while @xmath23 is the corresponding time scale .
rather than following the standard approach in deriving the hydrodynamic picture , we introduce the characteristic function @xmath47 instead .
it is straightforward to verify that @xmath48 satisfies the following equation @xmath49 note that the local moments @xmath50 can be obtained from the characteristic function according to the relation @xmath51 where @xmath52 .
the well - known hydrodynamic quantities , such as the mass density @xmath0 , the mean velocity @xmath53 of a fluid element and the hydrodynamic stress tensor @xmath54 can be defined as @xmath55 @xmath56 @xmath57 here , @xmath58 , implies the thermodynamic limit . defining also the deviation from the mean velocity as @xmath59 and using the evident relation @xmath60
we can represent the stress tensor @xmath61 according to the relation @xmath62 here @xmath63 is the internal stress tensor .
equation ( [ charfunceq ] ) and the one obtained after differentiating with respect to @xmath64 evaluated at @xmath65 , yield the smoluchowski equation and the equation for the momentum balance , respectively .
these can be written in the form @xmath66 } = { \frac { 1 } { 2 } } \nabla_k \nabla_l { \left ( b_{kl } \varrho \right ) } , \label{contineq}\ ] ] @xmath67 let us consider the time inversion transformation specified by @xcite @xmath68 , @xmath69 and @xmath70 .
we argue that there exists a backward velocity @xmath71 such that @xmath72 the transformed smoluchowski equation ( [ contineq ] ) can be represented according to @xmath73 } = - { \frac { 1 } { 2 } } \nabla_k \nabla_l { \left ( b_{kl } \varrho \right)}. \label{contineqb}\ ] ] summing up and subtracting equations ( [ contineq ] ) and ( [ contineqb ] ) , we obtain the continuity equation @xmath74 and the fick s law @xmath75 here @xmath76 are the current and the osmotic velocity , respectively .
it is worthwhile to mention that since the mean velocity of a fluid element is a generic function of time @xmath19 , it can be split into odd and even part .
note that from equation ( [ currentosm ] ) it follows that @xmath77 , where @xmath4 is the odd part , while @xmath78 is the even part .
equation ( [ momentbal ] ) for the balance of momentum can be written alternatively as @xmath79 after performing a time inversion in equation ( [ momentbalb ] ) , we obtain @xmath80 where @xmath81 denotes the transformed internal stress tensor after performing the time inversion . summing up the last two equations , we arrive at the sought - for equation for the current velocity @xmath82 where @xmath83 in order to find the explicit form of the internal stress tensor ( [ internstdef ] ) , we observe that the maximum entropy of the system is realized , provided the small source in the generalized kinetic equation ( [ onepartdfsms ] ) vanishes .
this condition is equivalent to the condition of detailed balance in the case , where the collision integral ( small source ) is approximated by a fokker - planck operator .
the condition of detailed balance implies that the distribution function factorizes in the form @xmath84 where @xmath85 is a normalizable function . from the above considerations
, it follows directly that @xmath86 where @xmath87 is the boltzmann constant and @xmath88 is the temperature .
+ in the simplest case , where the external force vanishes and the diffusion tensor is diagonal and isotropic , @xmath89 , we obtain @xmath90 where @xmath91 .
thus , the hydrodynamic equations describing a free capillary fluid have been recovered .
+ in the case , where an external force is applied , the korteweg stress tensor contains an additional term proportional to the drift coefficient @xmath92 . on the other hand from the principle of detailed balance
, it follows that the drift coefficient is proportional to the external force .
the physical implication of the latter is that the additional term in the korteweg stress tensor can be regarded as a coupling between the external field and the mean field of purely hydrodynamical origin .
since a detailed information about the system on the microscopic level is incomplete , one possible way to take into account its initial preparation , i. e. an eventual dynamic instability of motion that might have set in and/or other large - scale characteristics , is to introduce a suitable smoothing procedure . as a result ,
the kinetic equation providing an unified kinetic , hydrodynamic and diffusion description contains a small source and is therefore irreversible .
although the effective collision integral ( small source ) can be represented as a kramers - moyal expansion , for the purposes of the present paper it suffices to consider the right - hand - side of the generalized kinetic equation as approximated with a properly defined fokker - planck operator .
the latter form of the collision term is adopted as a starting point in the derivation of the hydrodynamic equations for capillary fluids .
+ the hydrodynamic approximation is further obtained in a standard manner from the characteristic function for the local moments of the distribution function .
an important feature of the approach is that the fick s law emerges naturally from the transformation properties of the hydrodynamic quantities under time inversion .
the osmotic velocity is uniquely specified by the first two infinitesimal moments of the smoothing function and in a sense is a measure of the irreversibility .
+ the main result of the analysis performed in this paper , the hydrodynamic equations for free capillary fluids have been derived from kinetic theory .
if an external force is present , the korteweg stress tensor has to be modified accordingly .
an additional term proportional to the drift coefficient emerges implying a coupling between the external field and the mean field of purely hydrodynamical origin . | based on the generalized kinetic equation for the one - particle distribution function with a small source , the transition from the kinetic to the hydrodynamic description of many - particle systems is performed .
the basic feature of this new technique to obtain the hydrodynamic limit is that the latter has been partially incorporated into the kinetic equation itself .
the hydrodynamic equations for capillary fluids are derived from the characteristic function for the local moments of the distribution function .
the fick s law appears as a consequence of the transformation law for the hydrodynamic quantities under time inversion . |
ratchet effects , i.e. , directed transport without a biased field due to a broken spatial - temporal symmetry , are of great interest to the understanding of molecular motors in biological systems and have been under intense investigation in a variety of simple model systems @xcite .
early studies focused on the role of external noise , which was then replaced by deterministic chaotic dynamics with dissipation @xcite .
recently , it was shown that even purely hamiltonian dynamics with both regular and chaotic phase space structures @xcite or with complete chaos @xcite is capable of generating ratchet effects .
as such , there has been a great interest in studying simple model hamiltonian systems in an effort to observe and understand ratchet effects .
the kicked rotor ( kr ) model @xcite and its variants are ideal systems for such studies .
for example , delta - kicked systems with an alternating kicking period and an asymmetric potential @xcite , or with multiple external fields @xcite have been considered in connection with ratchet effects .
these studies are stimulating since the classical dynamics of the kr ( i.e. , the standard map ) and the associated quantum dynamics have long served as paradigms for classical and quantum chaos @xcite . indeed , the kr has been realized in several atom optics experimental groups @xcite , and is also of considerable interest in other fields such as condensed matter physics @xcite , molecular physics @xcite , and quantum information science @xcite .
although not often thought of as such , the basic idea that underlies the above - mentioned studies of ratchet effects in the kr , i.e. , to manipulate symmetry properties to induce a directed current , is not new from a coherent control perspective .
indeed , fifteen years ago kurizki , shapiro , and brumer @xcite proposed a simple coherent control scenario that can generate photocurrents in semiconductors without a biased voltage .
this proposal and its various extensions @xcite have been experimentally realized @xcite . in this paper
we introduce a totally different type of hamiltonian ratchet effect in a simple delta - kicked model system , which we call a ratchet accelerator " . without a net external force ( i.e.
, the external force averaged over space is zero ) , the ratchet accelerator can generate directed classical anomalous transport , with the net current ( defined below ) that accelerates linearly with time .
this is made possible by taking advantage of the large transporting islands in the classical phase space of a modified kicked rotor system recently proposed by gong , wrner , and brumer @xcite , in conjunction with a second kicking field that breaks the spatial symmetry .
the corresponding quantum dynamics is also studied , shedding considerable light on quantum - classical differences and correspondence in the presence of anomalous diffusion .
for example , we show that due to the tunneling between a chaotic sea and a relatively large regular region of the classical phase space , the quantum directed current can be in the opposite direction of the classical directed current .
this paper is organized as follows . in sec .
ii we describe the background and the motivation of our model as a ratchet accelerator .
the classical results of directed anomalous diffusion are presented in sec .
iii , followed by the analogous quantum results in sec .
iv . concluding remarks are given in sec .
consider the following model hamiltonian @xmath0 where @xmath1 is the momentum , @xmath2 is the particle mass , @xmath3 is the conjugate position , @xmath4 and @xmath5 are the strength of the first and second kicking fields , @xmath6 is the time interval between kicks , and the function @xmath7 will be determined below .
the associated classical map is given by @xmath8 where @xmath9 is the scaled dimensionless position variable , @xmath10 is the scaled dimensionless momentum variable , @xmath11 , @xmath12 , and @xmath13 represents the phase space location of a classical trajectory at @xmath14 .
the corresponding quantum unitary evolution operator for propagating from time @xmath15 to time @xmath14 is given by @xmath16 \exp[-if(n)k_{1}\cos(\tilde{x})]\exp[ik_{2}\sin(2\tilde{x } ) ] , \label{qmapkr}\end{aligned}\ ] ] with dimensionless parameters @xmath17 , @xmath18 , @xmath19 .
let @xmath20 , where @xmath21 , then one has @xmath22=i\tau$ ] , and @xmath23 , indicating that @xmath24 plays the role of an effective planck constant .
note that @xmath25 and @xmath26 . clearly , if @xmath27 and @xmath28 then eq . ( [ ratcheth ] ) reduces to the standard kicked rotor model @xcite and the classical map of eq .
( [ clamapp ] ) becomes the standard map . if @xmath27 , @xmath29 , where @xmath30 for @xmath31 , @xmath32 , and @xmath33 for @xmath34 , @xmath35 , where @xmath36 an integer , then the system becomes that in our previous work @xcite .
our accelerator , introduced here , assumes @xmath37 and @xmath29 .
that is , the system is kicked by two fields , one of which reverses the sign of its kicking potential after every two kicks .
note first the existence of `` accelerating trajectories '' in the classical delta - kicked dynamics .
in particular , for particular values of @xmath38 , the standard map [ @xmath39 , @xmath28 ] can generate trajectories whose momentum increases ( or decreases ) linearly with time ( at least on the average ) .
these trajectories are a class of _ transporting _ trajectories @xcite . to see this consider the initial conditions : @xmath41 for @xmath42 , where @xmath43 and @xmath44 are integers . clearly , these phase space points are shifted by a constant value ( @xmath45 ) in @xmath46 after each iteration , resulting in a quadratic increase of rotational energy .
these accelerating trajectories are rather stable insofar as they may persist for values of @xmath47 close to @xmath48 ( with their average momentum shift after each iteration oscillating around the constant value @xmath45 ) , thus giving rise to transporting regular islands @xcite , also called the `` accelerator modes '' in the standard map case @xcite . dramatically ,
if @xmath29 and @xmath39 , then there can exist much larger transporting islands @xcite .
these islands are associated with the marginally stable points @xmath49 for @xmath50 .
trajectories on these islands will be shifted by a constant value [ @xmath51 in @xmath46 after each kick .
if classical trajectories are launched from the transporting regular islands , they simply jump to other similar islands located in adjacent phase space cells . for trajectories initially outside the transporting regular islands , the `` stickiness '' of the boundary between the islands and the chaotic sea induces anomalous diffusion over the phase space .
for example , the square of the variance of momentum increases nonlinearly , but not quadratically .
this is intrinsically different from the case of normal chaotic diffusion in which the square of the variance of momentum increases linearly with the number of kicks .
as in previous studies on delta - kicked systems , one needs to break the spatial symmetry of the kicking potential to induce ratchet effects .
more specifically , our choice of the above model hamiltonian is motivated by the desire to have large transporting islands in the classical phase space @xcite and to break the spatial symmetry of the transporting regular islands .
as shown below , the symmetry can be indeed broken by a second kicking field , and yet the remaining transporting islands are still of a significant size and have unidirectional transporting properties .
this being the case , the mechanism of the hamiltonian ratchet effects described below becomes transparent without using previous analyses of the relationship between broken spatial - temporal symmetries and ratchet effects @xcite .
to demonstrate the feasibility of our basic idea , in this section we present our results using some specific computational examples .
figure [ clamap]a displays the classical phase space structure of the standard map with @xmath52 , @xmath39 , and @xmath28 . the regular islands in fig .
[ clamap]a are not transporting since the momentum of the trajectories launched from these regular islands is bounded and oscillates periodically .
figure [ clamap]b shows that if the first kicking field reverses its potential after every two kicks , then the regular non - transporting islands in the standard map are destroyed and new regular islands emerge .
a simple computation reveals that those islands seen in fig .
[ clamap]b are transporting .
in particular , classical trajectories launched from regular islands on the left side ( @xmath53 ) of the phase space cell will accelerate , with their @xmath46 increased by @xmath54 after each kick .
due to the spatial symmetry , these islands always have partners with the opposite transporting property : trajectories launched from those islands on the right side ( @xmath55 ) of the same phase space cell will accelerate in the opposite direction , with their @xmath46 decreased by about @xmath56 after each kick .
these properties of the transporting regular islands seen in fig .
[ clamap]b indicate that they are associated with the marginal stable points @xmath57 , @xmath58 in the case of @xmath59 .
dramatically , upon introducing the second kicking field in eq .
( [ clamapp ] ) , e.g. , @xmath60 , the transporting islands no longer appear in pairs as in fig .
[ clamap]b .
rather , as seen in fig .
[ clamap]c , the spatial symmetry of the phase space structures is clearly broken , but the size of the main transporting islands that persist remains significant .
specifically , all the trajectories launched from the regular structures clearly seen in fig .
[ clamap]c will still accelerate linearly with time ; the remaining part of the phase space is exclusively occupied by the chaotic sea .
that is , there does not exist similar transporting islands that decreases @xmath46 in a linear fashion . for a classical ensemble that uniformly covers the entire phase space cell shown in fig .
[ clamap]c , the spatial and temporal periodicity of the system requires that the average acceleration rate is zero @xcite , i.e. , @xmath61 where @xmath62 and @xmath63 denote the areas of the chaotic sea and the regular transporting region , respectively , with their associated average acceleration rate denoted by @xmath64 and @xmath65 .
consider now an initial classical ensemble with @xmath66 and @xmath67 uniformly distributed between @xmath68 and @xmath69 .
such an ensemble is easy to simulate both classically and quantum mechanically . since this initial ensemble is entirely located in the chaotic sea ( see fig .
[ clamap]c ) , eq .
( [ vec ] ) predicts that the average acceleration rate of this classical ensemble is given by @xmath70 .
this also indicates that the larger the transporting island is , the faster the directed acceleration would be . in the case of fig .
[ clamap]c , @xmath71 , so more trajectories travel to the regime of negative @xmath67 .
that is , even though the average force of the kicking fields is zero and the initial classical ensemble is both spatially symmetric and time - reversal invariant , all the trajectories tend to be accelerated in one direction . note that , by changing the strength of the second kicking field or varying the relative phase between the two kicking fields ( e.g. , the second kicking field is replaced by @xmath72 , where @xmath73 is a relative phase factor ) , one can also obtain @xmath74 .
hence , the acceleration can be also obtained in the reverse direction . for an initial classical ensemble with @xmath66 and @xmath67 uniformly distributed from @xmath68 to @xmath75 , fig .
[ cla - p]a displays the time dependence of the classical net current @xmath76 , where @xmath77 denotes the ensemble average .
it is indeed seen that the classical net current is in the negative direction , and accelerates linearly with time .
the acceleration rate is also consistent with eq .
( [ vec ] ) .
figure [ cla - p]b displays a @xmath78 plot of the time dependence of the associated momentum variance @xmath79
. a linear fit of the result in fig .
[ cla - p]b gives @xmath80 , where @xmath81 is the number of kicks .
this confirms that the diffusion dynamics is anomalous and immediately gives @xmath82 .
hence , as far as the first and second order statistical moments are concerned , the momentum fluctuations of the classical ensemble become less and less important as time increases . in this sense , for sufficiently large times , almost all trajectories in the classical ensemble are accelerated in the negative direction .
how then does this ratchet accelerator manifests itself in the corresponding quantum system and how small the effective planck constant should be in order to observe the analogous ratchet effects in the quantum dynamics ?
to carry out the quantum calculations we impose a periodic boundary condition on the system , i.e. , @xmath83 , where @xmath84 is the wavefunction of the system .
this periodic boundary condition suggests that @xmath67 is now to be understood as an angle variable , and is consistent with the fact that the classical dynamics is invariant if @xmath85 .
further , our previous work @xcite suggests that this choice of boundary conditions should not affect the essence of the quantum dynamics .
figure [ q - current ] shows the net quantum current , i.e. , the expectation value of @xmath86 for two values of the effective planck constant @xmath24 , with the initial state given by @xmath87 . comparing the results here with that in fig .
[ cla - p]a , one sees that the magnitude of the quantum directed current can be much smaller than that of the classical directed current .
interestingly , the directed current in the quantum case displays much more complicated behavior than the linear acceleration seen classically ( some slow oscillatory behavior of the quantum current over much longer time scales is also observed , but should be of much less experimental interest ) .
this can be qualitatively explained in terms of a combination of a dynamical localization effect @xcite and the quantum tunneling between the chaotic sea and the transporting islands @xcite . in particular , since the dynamical localization in quantum delta - kicked systems always suppresses the diffusion , the quantum directed current is not expected to grow all the time .
more importantly , since the chaotic sea and the transporting islands have opposite acceleration signs [ see eq .
( [ vec ] ) ] , the quantum tunneling between them @xcite necessarily decreases the directed current .
indeed , in the case of @xmath88 , one sees that for most of the time the quantum current is in the opposite direction of the classical current .
note that , unlike a previous observation of tunneling - induced current reversal in quantum brownian motion @xcite , here the tunneling is between a chaotic sea and a regular region and the system is totally isolated from the environment . the results in fig .
[ q - current ] also suggest that @xmath89 is small enough to observe the signature of a ratchet accelerator in a quantum delta - kicked system .
for example , the quantum current for @xmath89 does increase almost linearly for as many as 2000 kicks , with the average increase rate about three times smaller than in the classical case .
this means that for experiments using cold sodium atoms @xcite the directed and relatively large quantum current obtained with 2000 kicks would correspond to an average drift velocity on the order of 1 m / s .
note however , this value of @xmath24 is about one order of magnitude smaller than what is being examined in atom optics experiments @xcite , but may be achievable in the future .
it is also interesting to examine the lineshape of dynamical localization with the spatial symmetry of the transporting regular islands broken by the second kicking field .
figure [ pdis ] displays the associated distribution function @xmath90 after 4000 kicks for the same two values of @xmath24 .
it is clearly seen that for both cases @xmath90 is asymmetric . due to the stickiness of the boundary between the transporting islands and the chaotic sea and the associated quantum tunneling , the probability of finding the system in state @xmath91 with a large and positive @xmath2 can be much higher than that of finding the system in state @xmath92 .
this does not contradict the previous observation that after an initial stage the quantum current for @xmath89 is negative , because the tail of the statistical distribution function @xmath90 has little effect on the net current , the statistical moment of the lowest order .
the huge difference in @xmath90 between the two cases shown in fig .
[ pdis ] with slightly different values of @xmath24 is also noteworthy .
this extreme sensitivity of @xmath90 to the value of @xmath24 is more or less related to the fact that the values of @xmath24 we consider are quite comparable to the area of the main transporting island seen in fig .
[ clamap]c .
using a simple model hamiltonian system we have demonstrated a new ratchet effect by considering directed classical anomalous diffusion .
we have shown that it is possible to achieve directed acceleration with a fixed acceleration rate with two kicking fields that have zero average gradient .
this is the case even though the initial classical ensemble is both spatially symmetric and time - reversal invariant .
we have also presented some computational examples of the quantum dynamics with different values of the effective planck constant .
it is found that quantum coherence effects may induce a reversal of directed current .
it is also found that directed quantum current that accelerates linearly with time without using a biased external field should also be achievable in a quantum system that is still far from the classical limit .
we have previously shown that quantum anomalous diffusion in a simple delta - kicked system can be faster than the underlying classical anomalous diffusion @xcite .
however , the spatial symmetry therein always gives a zero momentum expectation value and therefore one necessarily considers the momentum variance in order to make a meaningful quantum - classical comparison .
by contrast , our ratchet accelerator model here allows for a study of the quantum - classical differences in the momentum expectation value itself . as such
, the role of the quantum tunneling between a chaotic sea and transporting regular islands becomes even more clear : while it can enhance the increase rate of the momentum variance @xcite , it suppresses the net acceleration rate of the average momentum in our ratchet accelerator model .
it would be of great interest to consider a molecular realization of the ratchet accelerator , where the variable @xmath3 is understood as an angle variable and @xmath1 is understood as an angular momentum variable . since a microwave field @xcite will create a kicking potential of the form @xmath93 and an off - resonance laser field @xcite can create a kicking potential of the form @xmath94 , a diatomic subject to both microwave and off - resonance laser fields , with their polarization direction perpendicular to one another , is a possible candidate for realizing the ratchet accelerator we propose .
however , since in this case the diatomic should be at least described by a three - dimensional rigid rotor , which differs from the planetary kicked rotor in a number of aspects @xcite , detailed calculations should be carried out to guide the experimental studies .
nevertheless , we believe that this ratchet accelerator model may be of importance for manipulating the rotational motion of diatomics and may provide a means of orienting molecules while accelerating rotation without using a biased external field .
100 current address : department of chemistry and the james franck institute , university of chicago , chicago , il 60637 .
p. reimann , phys . rep . * 361 * , 57 ( 2002 ) . r.d . astumian and p. hnggi ,
phys . today * 55 * , 33 ( 2002 ) .
p. jung , j.g .
kissner , and p. hnggi , , 3436 ( 1996 ) .
mateos , , 258 ( 2000 ) .
m. porto , m. urbakh , and j. klafter , , 491 ( 2000 ) .
s. flach , o. yevtushenko , and y. zolotaryuk , , 2358 ( 2000 ) .
t. dittrich , r. ketzmerick , m.f .
otto , and h. schanz , ann .
( berlin ) , * 9 * , 755 ( 2000 ) .
h. schanz , m.f .
otto , r. ketzmerick , and t. dittrich , , 070601 ( 2001 ) . transporting trajectories can be transporting in either momentum or position , whereas accelerating trajectories are transporting in momentum and hence accelerate .
t. cheon , p. exner , and p. seba , j. phys .
japan * 72 * , 1087 ( 2003 ) .
s. denisov , j. klafter , and m. urbakh , , 046203 ( 2002 ) .
monteiro , p.a .
dando , n.a.c .
hutchings , and m.r .
isherwood , , 194102 ( 2002 )
. t. jonckheere , m.r .
isherwood , and t.s .
monteiro , phys .
. lett . * 91 * , 253003 ( 2003 ) .
jones , m. goonasekera , h.e .
saunders - singer , and d.r .
meacher , preprint quant - ph 0309149 ( 2003 ) .
g. casati and b. chirikov , _ quantum chaos : between order and disorder _
( cambridge university press , new york , 1995 ) .
moore , j.c .
robinson , c.f .
bharucha , b. sundaram , and m.g .
raizen , , 4598 ( 1995 ) ; h. ammann , r. gray , i. shvarchuck , and n. christensen , , 4111 ( 1998 ) ; j. ringot , p. szriftgiser , and j.c .
garreau , d. delande , , 2741 ( 2000 ) ; m.b .
darcy , r.m .
godun , m.k .
oberthaler , d. cassettari , and g. s. summy , , 074102 ( 2001 ) .
s. fishman , d.r .
grempel , and r.e .
prange , , 509 ( 1982 ) .
g. benenti , g. casati , i. guarneri , and m. terraneo , , 014101 ( 2001 ) .
r. blmel , s. fishman , and u. smilansky , j. chem .
* 84 * , 2604 ( 1986 ) . i. sh .
averbukh and r. arvieu , , 163601 ( 2001 ) .
m. machholm and n.e .
henriksen , , 193001 ( 2001 ) .
m. leibscher , i.sh .
averbukh , and h. rabitz , , 213001 ( 2003 ) .
facchi , s. pascazio , and a. scardicchio , , 61 ( 1999 ) .
b. georgeot and d.l .
shepelyansky , , 2890 ( 2001 ) .
g. kurizki , m. shapiro , and p. brumer , , 3435 ( 1989 ) ; m. shapiro and p. brumer , _ principles of the quantum control of molecular processes _ ( john wiley , new york , 2003 ) .
r. atanasov , a. hache , j.l.p .
hughes , h.m .
van driel , and j.e .
sipe , , 1703 ( 1996 ) .
bhat and j.e .
sipe , , 5432 ( 2000 ) .
e. dupont , p.b .
corkum , h.c .
liu , m. buchanan , and z.r .
wasilewski , , 3596 ( 1995 ) .
a. hache , y. kostoulas , r. atanasov , j.l.p .
hughes , j.e .
sipe , and h.m .
van driel , , 306 ( 1997 ) .
stevens , a.l .
smirl , r.d.r .
bhat , a. najmaie , j.e .
sipe , and h.m .
van driel , , 136603 ( 2003 ) .
j. gong , h.j .
wrner , and p. brumer , , 026209 ( 2003 ) .
j. gong , h.j .
wrner , and p. brumer , , 056202 ( 2003 ) .
a. iomin , s. fishman , and g.m .
zaslavsky , , 036215 ( 2002 ) .
r. roncaglia , l. bonci , b.j .
west , and p. grigolini , , 5524 ( 1995 ) .
s. denisov and s. flach , , 056236 ( 2001 ) .
s. denisov , j. klafter , m. urbakh , and s. flach , physica d * 170 * , 131 ( 2002 ) .
s. denisov , s. flach , a.a .
ovchinnikov , o. yevtushenko , and y. zolotaryuk , 041104 ( 2002 ) .
the quantum tunneling between the chaotic sea and the transporting island implies that the floquet state of the kicked quantum system can simultaneously occupy these classical phase space structures , a counter - intuitive result first demonstrated in l. hufnagel , r. ketzmerick , m.f .
otto , and h. schanz , , 154101 ( 2002 ) .
p. reimann , m. grifoni , and p. hnggi , , 10 ( 1997 ) .
j. gong and p. brumer , , 3590 ( 2001 ) . | directed classical current that increases linearly with time without using a biased external field is obtained in a simple model hamiltonian system derived from a modified kicked rotor model , by breaking the spatial symmetry of the transporting regular islands in the classical phase space .
a parallel study of the corresponding quantum dynamics suggests that although quantum coherence effects suppress the directed current and can even induce a current reversal , directed quantum current that increases linearly with time can nonetheless be realized in a quantum system that is far from the classical limit . |
in strange nuclear physics and hadron physics , kaonic nuclei ( nuclear system with anti - kaons , @xmath6 ) have been a hot topic because they are interestingly expected to have several exotic properties , such as the formation of dense state , due to strong @xmath1 attraction @xcite .
to reveal such _ expected _ exotic properties involved in kaonic nuclei , we have focused on the simplest kaonic nucleus `` @xmath0 '' which is a prototype system of kaonic nuclei . in the experimental side , new results are being reported from j - parc experiments @xcite . especially , j - parc e27 collaboration has reported some signal in their @xmath0 search experiment using deuteron target @xcite . in the theoretical side ,
the three - body system of @xmath0 has been investigated with various ways as summarized in ref .
the binding energy and decay width rather depend on approaches and employed potentials .
however , all theoretical studies result that the @xmath0 can be bound with less than 100 mev binding energy .
since there is @xmath7 threshold at 103 mev below @xmath5 threshold , those calculations indicate that the @xmath0 should be a resonant state located between the two thresholds .
in addition , it is known by studies of the @xmath8 that the @xmath1 couples strongly to the @xmath9 @xcite .
thus , we consider that _ 1 .
resonance _ and _ 2 .
coupled - channel problem _ are key ingredients in the theoretical study of the @xmath0 .
we , here , employ a coupled - channel complex scaling method ( cccsm ) since this approach can simultaneously treat these two ingredients .
it should be noted that the complex scaling method has greatly succeeded in studies of resonant states of stable / unstable nuclei @xcite .
we have applied the cccsm to the two - body system of @xmath1-@xmath10 @xcite .
( @xmath11 means @xmath12 and @xmath13 hyperons . ) since the cccsm is found to be a useful tool also for the study of hadronic system through those studies , we now tackle the three - body system of @xmath0 with the same method .
we briefly explain our method to investigate the @xmath0 resonance . here , note that the `` @xmath0 '' called in theoretical studies means a coupled - channel system of @xmath5-@xmath7-@xmath14 with quantum numbers of @xmath2 and @xmath3 .
basically , we follow the usual prescription of the complex scaling method @xcite to calculate complex eigenvalues of the three - body system of @xmath0 .
the hamiltonian for the @xmath0 is complex - scaled , with a complex - scaling operator @xmath15 in which the coordinate and the conjugate momentum are transformed as @xmath16 and @xmath17 , respectively . diagonalizing the complex - scaled hamiltonian with a @xmath18-integrable basis functions , which are the correlated gaussian functions @xcite in our study
, complex eigenvalues are obtained . among those eigenvalues , the eigenvalues which are independent of the scaling angle @xmath19
correspond to the resonant states .
the @xmath0 is a coupled - channel system of @xmath5 , @xmath7 and @xmath14 .
we treat such a multi - channel problem as a single - channel problem with help of feshbach method @xcite in the complex scaling method .
we call our method as a coupled - channel complex scaling method with feshbach projection ( _ cccsm+feshbach method _ ) @xcite . in feshbach method , a model space ( @xmath20 space ) and outer space of the model space ( @xmath21 space ) are assigned .
then , the schrdiner equation is given as a coupled - channel equation of wave functions for @xmath20 and @xmath21 spaces . by the elimination the @xmath21-space wave function , a schrdiner equation only for the @xmath20-space wave function @xmath22
is obtained as @xmath23 \phi_p = e \phi_p$ ] .
( the operator @xmath24 is a kinetic - energy operator for the @xmath20 space . ) here , the effective potential for @xmath20 space , @xmath25 , is formally written as @xmath26 with @xmath27 , where @xmath28 is the green function for the @xmath21 space and \{@xmath29 } indicates the coupled - channel potential for the @xmath20 and @xmath21 spaces with @xmath30=@xmath20 or @xmath21 .
certainly , a single - channel schrdiner equation for the @xmath20 space can be obtained with the feshbach method .
however , the problem is how to represent the @xmath28 in actual calculations .
we overcome this matter with a unique nature of the complex scaling method .
the closure relation is proven to hold also in the csm , including resonant states explicitly as well as bound and non - resonant continuum states .
( _ extended closure relation _ , ecr @xcite ) .
the ecr is known to be well described approximately with a set of finite number of the eigenstates \{@xmath31 } which are obtained by the diagonalization of a complex - scaled hamiltonian @xmath32 with a gaussian basis function ; @xmath33 @xcite . with help of the ecr
, the complex - scaled green function @xmath34 is expressed as @xmath35 where the energy eigenvalue @xmath36 and eigenstate @xmath37 are obtained by the diagonalization of the complex - scaled hamiltonian @xmath38 with gaussian basis functions . since the non - scaled green function for the @xmath21 space is obtained from the complex - scaled green function @xmath39 by such an inverse transformation as @xmath40 , the effective potential for the @xmath20 space is @xmath41 since the @xmath39 given as eq .
( [ gq_th ] ) is represented with gaussian functions , the effective potential @xmath25 is also composed of gaussian functions .
therefore , it can be used as usual in the complex scaling method with gaussian basis function .
we apply this technique to the calculation of the @xmath0 system . setting the @xmath1 channel as @xmath20 space and the @xmath9 and @xmath42 channels as @xmath21 space
, we construct an effective @xmath1 potential @xmath43 . in other words ,
the @xmath10 channels are eliminated at the step of two - body calculation .
the @xmath43 plugged in the three - body hamiltonian of the @xmath5 , we solve the single - channel problem of the @xmath5 system with the complex scaling method . the effective @xmath1 potential derived with the cccsm+feshbach method
has an energy dependence .
the self - consistency for @xmath1 energy has to be taken into account when bound and resonant states are considered .
such a self - consistency is dealt with a manner proposed in an earlier study @xcite : the definition of the @xmath1 energy in the @xmath0 is non - trivial because the two - body @xmath1 system is a subsystem of the three - body @xmath5 system .
we define the @xmath1 energy ( @xmath44 ) in two ways by considering extreme two pictures ; the anti - kaon is regarded as a field ( ansatz 1 ) or it is considered as a particle ( ansatz 2 ) .
details of calculation of the @xmath44 are given in eqs .
( 20 ) and ( 21 ) in ref . @xcite .
also in the present study , these two ansatz are examined .
we remark that this self - consistency is realized for the _ complex _
@xmath1 energy since the complex energy of the resonance pole is directly treated in the current work , while the real @xmath1 energy is considered in the earlier work with a variational approach @xcite .
( left ) complex - energy eigenvalue distribution when @xmath1 energy for the effective potential is fixed to the @xmath45 energy .
( right ) pole position of @xmath0 obtained by the self - consistent calculation , where @xmath46 is varied from 90 mev to 120 mev and two ansatz for @xmath1 energy are examined.,title="fig : " ] ( left ) complex - energy eigenvalue distribution when @xmath1 energy for the effective potential is fixed to the @xmath45 energy .
( right ) pole position of @xmath0 obtained by the self - consistent calculation , where @xmath46 is varied from 90 mev to 120 mev and two ansatz for @xmath1 energy are examined.,title="fig : " ] we show our result of the @xmath0 calculated with the cccsm+feshbach method . as a @xmath1(-@xmath10 ) potential , we use chiral su(3)-based potentials having a single - range gaussian form in the coordinate space , which was proposed in our previous work @xcite . here , the results obtained with a version of our potentials , called nrv2c with @xmath47 mev , are shown as a typical case .
first , we consider the case where the @xmath1 energy in the energy - dependent @xmath1 potential is fixed to that for the @xmath8 .
in other words , the @xmath1 energy is not self - consistent in the @xmath0 , but it is self - consistent for the isolated two - body system of @xmath1-@xmath9 which forms the @xmath8 resonance . the distribution of obtained complex - energy eigenvalues is shown in the left panel of fig . [ fig1 ] . in the figure
, the origin of the real energy axis corresponds to the @xmath5 three - body threshold .
it is known that in the complex scaling method the continuum states appear on so - called @xmath48 line ( @xmath49 ) when the scaling angle is @xmath19 @xcite .
therefore , the eigenvalues along the @xmath48 line running from the origin indicate the scattering continuum states of @xmath5 three - body system
. there are eigenvalues on the other line .
the starting point of this line is ( @xmath50 , @xmath51 ) mev which is nearly equal to the complex energy of the @xmath45 @xcite .
( @xmath45 denotes the higher pole of the @xmath8 . )
therefore , the eigenvalues on the second line indicate the @xmath52 two - body scattering continuum states .
there is a complex eigenvalue isolated from the two lines mentioned above , as marked with a red circle in the figure .
this eigenstate means the @xmath0 resonance . in the case where the @xmath1 energy is fixed at that of the @xmath45 , the pole energy of the @xmath0 resonance
is found to be @xmath53 mev .
next , we take into account the self - consistency for the @xmath1 energy in the three - body system @xmath0 , following the two ansatz as explained in the section [ e - dep ] . as a result of the search for self - consistent solutions ,
such solutions of the @xmath0 resonance have been successfully found .
the resonance energy is obtained to be @xmath54 mev for ansatz 1 and ( @xmath55 , @xmath56 ) mev for ansatz 2 .
the binding energy @xmath57 is not so dependent on the ansatz , whereas the decay width @xmath58 strongly depends on it .
compared with the earlier work of the variational calculation with a chiral su(3)-based potential @xcite that is a similar type to the potential used here , the present calculation gives slightly deeper binding and narrower width .
we examine the @xmath46 dependence of the solutions , since the @xmath46 value is a parameter of our @xmath1 potential .
the resonance pole of the @xmath0 system is found for each @xmath46 value when it is varied from 90 mev to 120 mev .
( see the right panel of fig .
[ fig1 ] . )
however , since we find that the pole for @xmath59 mev in ansatz 1 is unstable for the scaling angle @xmath19 , we discard this solution .
consequently , the binding energy is obtained to be small , 20 - 30 mev .
the decay width is obtained to spread widely , 20 - 65 mev .
we comment on the present result .
it is found that both ansatz give similar binding energy .
however , they give rather different decay width ; ansatz 1 tends to give extremely small decay width , compared to ansatz 2 .
the ansatz 2 seems to provide the decay width similar to that of the earlier study @xcite .
complex eigenvalue distribution of @xmath5 with @xmath4 and @xmath3 .
av4 @xmath60 and a phenomenological @xmath1 potentials are employed .
`` k@xmath61-d '' means the scattering continuum states of @xmath62 and deuteron . ] as mentioned in introduction , recently j - parc e27 collaboration has reported their experimental result of @xmath0 search via @xmath63 reaction @xcite .
we consider that this reaction excites @xmath4 state , not @xmath2 state of the @xmath0 that we have calculated so far , since the target of deuteron is spin 1 state and the @xmath64 reaction is expected not to flip the spin so much @xcite .
hence , we have investigated a @xmath5 system with the quantum numbers of @xmath4 and @xmath3 in the same way as mentioned above . here , we set the spin and isospin of two nucleons to be @xmath65 and @xmath66 in the trial wave function of @xmath5 , respectively .
we employ the av4 @xmath60 potential in which the tensor force is incorporated into the central force , and an energy - independent phenomenological @xmath1 potential @xcite .
the @xmath1 energy in the effective @xmath1 potential is fixed to be that of @xmath45 resonance for the simplicity .
[ fig_s=1 ] shows the complex eigenvalue distribution obtained with the cccsm+feshbach method . as marked in the figure
, we can see the @xmath45 resonance , the deuteron bound state , and @xmath45+@xmath67 and deuteron+@xmath62 scattering continuum states .
there are no @xmath5 three - body resonance states . as a result of our calculation
, the @xmath5 could not form any resonances with the quantum numbers @xmath68 which are considered to be excited via the @xmath63 reaction .
we have investigated a prototype of kaonic nuclei , @xmath0 , with a coupled - channel complex scaling method combined with feshbach method . in the method ,
a coupled - channel problem is reduced to a single - channel problem to be handled more easily , by utilizing the extended closure relation in the complex scaling method . as a result of the search for the @xmath0 resonance pole on the complex energy plane ,
the binding energy and the mesonic decay width of @xmath0 are obtained to be 20 - 30 mev and 20 - 65 mev , respectively , where we use an energy - dependent chiral su(3)-based @xmath1 potential .
we consider that the @xmath0 is shallowly bound as suggested by earlier studies with variational methods using the same kinds of @xmath1 potentials @xcite .
according to the recent report of j - parc e27 collaboration , a bump structure is found at @xmath69 mev in the @xmath70 invariant - mass spectrum with two protons tagged @xcite .
this means that the binding energy is about @xmath71 mev similarly to the results of the past two experiments @xcite , _ if the observed state is a @xmath0 state_. if this is the case , there are large discrepancy between experimental observations and theoretical results . in addition
, there has been still an essential question `` what is the object observed in the experiments ? '' . to resolve these questions , further studies from both experimental and theoretical sides are necessary . in our future plan
, we will carry out the @xmath5-@xmath7-@xmath14 coupled - channel calculation without any channel elimination . by the explicit treatment of @xmath72 channels we will obtain more accurate result on the @xmath0 , and expect to clarify the role of the @xmath72 three - body dynamics as well as that of the @xmath5 dynamics .
this work was developed through several activities ( meetings and workshops ) at the j - parc branch of the kek theory center .
funding : this work is funded by jsps kakenhi grant number 25400286 and partially by grant number 24105008 .
conflict of interest : the authors declare that they have no conflict of interest .
a. dot and t. myo , nucl .
phys . a * 930 * ( 2014 ) 86 .
y. suzuki and k. varga , _ stochastic variational approach to quantum - mechanical few - body problems ( lecture notes in physics monographs ) _ ( springer , 1998 ) . | we have investigated a prototype of kaonic nuclei , `` @xmath0 '' , with a coupled - channel complex scaling method ( cccsm ) . combining the cccsm with feshbach projection method
, we can handle a coupled - channel problem effectively as a single - channel problem . by using an energy - dependent chiral su(3)-based @xmath1 potential , the @xmath0 ( @xmath2 and @xmath3 )
is obtained to be shallowly bound with the binding energy of 20 - 30 mev .
the mesonic decay width depends on the interaction parameters and ansatz ; the decay width is ranging from 20 to 65 mev . in case of @xmath4 state ,
no three - body @xmath5 resonant states are found . |
here s why : these are , first of all , the general principles of quantum mechanics , special relativity , and locality , that lead one to relativistic quantum field theory @xcite . in addition , we require invariance under the nonabelian gauge symmetry @xmath0 , the specific matter content of quarks six spin 1/2 dirac fermions which are color triplets and renormalizability .
these requirements determine the theory completely , up to a very small number of continuous parameters as discussed below .
deeper consideration reduces the axioms further .
theoretical physicists have learned the hard way that consistent , non - trivial relativistic quantum field theories are difficult to construct , due to the infinite number of degrees of freedom ( per unit volume ) needed to construct local fields , which tends to bring in ultraviolet divergences . to construct a relativistic quantum theory ,
one typically introduces at intermediate stages a cutoff , which spoils the locality or relativistic invariance of the theory .
then one attempts to remove the cutoff , while adjusting the defining parameters , to achieve a finite , cutoff - independent limiting theory .
renormalizable theories are those for which this can be done , order by order in a perturbation expansion around free field theory .
this formulation , while convenient for mathematical analysis , obviously begs the question whether the perturbation theory converges ( and in practice it never does ) . a more straightforward procedure , conceptually , is to regulate the theory as a whole by discretizing it , approximating space - time by a lattice @xcite .
this spoils the continuous space - time symmetries of the theory .
then one attempts to remove dependence on the discretization by refining it , while if necessary adjusting the defining parameters , to achieve a finite limiting theory that does not depend on the discretization , and does respect the space - time symmetries .
the redefinition of parameters is necessary , because in refining the discretization one is introducing new degrees of freedom . the earlier , coarser theory results from integrating out these degrees of freedom , and
if it is to represent the same physics it must incorporate their effects , for example in vacuum polarization . in this procedure
, the big question is whether the limit exists .
it will do so only if the effects of integrating out the additional short - wavelength modes , that are introduced with each refinement of the lattice , can be captured accurately by a re - definition of parameters already appearing in the theory .
this , in turn , will occur only in a straightforward way if these modes are weakly coupled .
( another simple possibility is that short - distance modes of different types cancel in vacuum polarization .
this is what occurs in supersymmetric theories .
other types of ultraviolet fixed points are in principle possible , but difficult to imagine or investigate . ) but this is true , if and only if the theory is asymptotically free . one can investigate this question , i.e. , whether the couplings decrease to zero with distance , or in other words whether the theory is asymptotically free , within weak coupling perturbation theory @xcite .
one finds that only nonabelian gauge theories with simple matter content , and no non - renormalizable couplings , satisfy this criterion @xcite .
supersymmetric versions of these theories allow more elaborate , but still highly constrained , matter content . summarizing the argument , only those relativistic field theories which are asymptotically free
can be argued in a straightforward way to exist . and
the only asymptotically free theories in four space - time dimensions involve nonabelian gauge symmetry , with highly restricted matter content .
so the axioms of gauge symmetry and renormalizability are , in a sense , gratuitous .
they are implicit in the mere _ existence _ of non - trivial interacting quantum field theories .
thus qcd is a member of a small aristocracy : the closed , consistent embodiments of relativity , quantum mechanics , and locality . within this class
, it is among the least affected members .
as i just discussed , qcd can be constructed by an explicit , precisely defined discretization and limiting procedure .
this provides , in principle , a method to compute any observable , in terms that could be communicated to a turing machine .
in fact marvelous things can be accomplished , in favorable cases , using this direct method . for a stirring example , see figure 4 below .
the computational burden of the direct approach is , however , heavy at best .
when one can not use importance sampling , as in addressing such basic questions as calculating scattering amplitudes or finding the ground state energy at finite baryon number density , it becomes totally impractical . for these reasons
various improved perturbation theories continue to play an enormous role in our understanding and use of qcd .
the most important and well - developed of these , directly based on asymptotic freedom , applies to hard processes and processes involving heavy quarks @xcite .
it is what is usually called `` perturbative qcd '' , and leads to extremely impressive results as exemplified by figures 1 - 3 , below .
the scope of these methods is continually expanding , to include additional `` semi - hard '' processes , as will be discussed in many talks at the conference .
when combined with some additional ideas , they allow us to address major questions regarding the behavior of the theory at high temperatures and large densities , as i ll touch on below .
chiral perturbation theory @xcite , which is based on quite different aspects of qcd , is extremely useful in discussing low - energy processes , though it is difficult to improve systematically .
a proper discussion of this , or of many other approaches each of which offers some significant insight ( traditional nuclear physics , bag model , reggeism , sum rules , large n , skyrme model , ... ) would not be appropriate here .
i would however like to mention a perturbation theory which i think is considerably underrated , that is strong coupling perturbation theory @xcite .
it leads to a simple , appealing , and correct understanding of confinement @xcite , and even its existing , crude implementations provide a remarkably good caricature of the low - lying hadron spectrum .
it may be time to revisit this approach , using modern algorithms and computer resources .
there are significant applications of qcd to nuclear physics , accelerator physics , cosmology , extreme astrophysics , unification , and natural philosophy .
i ll say just a few words about each , in turn . _ nuclear physics _
: understanding atomic nuclei was of course the original goal of strong interaction physics . in principle qcd
provides answers to all its questions .
but in practice qcd has not superseded traditional nuclear physics within its customary domain .
the relationship between qcd and traditional nuclear physics is in some respects similar to the relationship between qed and chemistry .
the older disciplines retain their integrity and independence , because they tackle questions that are exceedingly refined from the point of view of the microscopic theories , involving delicate cancellations and competitions that manufacture small net energy scales out of much larger gross ones .
qcd offers many insights and suggestions , however , of which we will hear much at this conference .
there is also an emerging field of extreme nuclear physics , including the study of nuclei with hard probes and heavy - ion collisions , where the influence of qcd is decisive .
_ accelerator physics _ : most of what goes on at high - energy accelerators is described by qcd .
this application has been so successful , that experimenters no longer speak of `` tests of qcd '' , but of `` qcd backgrounds '' !
two- and even three - loop calculations of such `` backgrounds '' are in urgent demand .
what can one add to that sincere testimonial ?
_ cosmology _ : because of asymptotic freedom , hadronic matter becomes not impenetrably complex , but rather profoundly simpler , under the extreme conditions predicted for the early moments of the big bang .
this stunning simplification has opened up a large and fruitful area of investigation . _
extreme astrophysics _ : the physics of neutron star interiors , neutron star collisions , and collapse of very massive stars involves extreme nuclear physics .
it should be , and i believe that in the foreseeable future it will be , firmly based on microscopic qcd . _ unification and natural philosophy _ : see below .
let me enumerate some major ones : radiative corrections , running couplings , confinement , spontaneous ( chiral ) symmetry breaking , anomalies , instantons .
much could be said about each of these , but i will just add a few words about the first . the lamb shift in qed
is rightly celebrated as a triumph of quantum field theory , because it shows quantitatively , and beyond reasonable doubt , that loop effects of virtual particles are described by the precise , intricate rules of that discipline .
but in qcd , we probably have by now 50 or so cases where two- and even three - loop effects are needed to do justice to experimental results and the rules are considerably more intricate
! a straightforward accounting of the parameters in qcd would suggest 8 : the masses of six quarks , the value of the strong coupling , and the value of the p and t violating @xmath1 parameter .
the fact that there are only a small finite number of parameters is quite profound .
it is a consequence of the constraints of gauge invariance and renormalizability ( or alternatively , as we saw , existence , by way of asymptotic freedom ) . in reality
there are not 8 parameters , but only 6 .
@xmath2 is eliminated by dimensional transmutation @xcite .
this means , roughly stated , that because the coupling runs as a function of distance , one can not specify a unique numerical value for it .
it will take any value , at some distance or other .
one can put ( say ) @xmath3 , thereby determining a length scale @xmath4 .
what appeared to be a choice of dimensionless coupling , is revealed instead to be a choice of unit of length , or equivalently ( with @xmath5 ) of mass . only the dimensionless _
ratio _ of this mass to quark masses can enter into predictions for dimensionless quantities .
so what appeared to be a one - parameter family of theories , with different couplings , turns out to be a single theory measured using differently calibrated meter - sticks .
the @xmath1 term is eliminated , presumably , by the peccei - quinn mechanism @xcite . to assure us of this ,
it would be very nice to observe the quanta associated with this mechanism , namely axions @xcite . in any case , we know for sure that the @xmath1 term is very small . for purposes of strong interaction physics , within qcd itself , we can safely set it to zero , invoking p or t symmetry , and be done with it
. this economy of parameters would already be quite impressive , given the wealth of phenomena described .
however if we left it at that we would be doing a gross injustice to qcd , and missing one of its most striking features . to make my point , let me call your attention to a simplified version of qcd , that i call `` qcd lite '' .
qcd lite is simply qcd truncated to contain just two flavors of quarks , both of which are strictly massless , and with the @xmath1 parameter set to zero .
these choices are natural in the technical sense , since they can be replaced by symmetry postulates .
indeed , assuming masslessness of the quarks is tantamount to assuming exact @xmath6 chiral symmetry , and @xmath7 is tantamount to assuming the discrete symmetries p or t. ( actually , once we have set the quark masses to zero , we can dial away @xmath1 by a field redefinition . ) now there are two especially remarkable things about qcd lite @xcite .
the first is that it is a theory which _ contains no continuous free parameters at all_. its only inputs are the numbers 3 ( colors ) and 2 ( flavors ) .
the second is that it _ provides an excellent semi - quantitative theory of hadronic matter_. indeed , in reality the strange and heavier quarks have very little influence on the structure or masses of protons , neutrons , atomic nuclei , pions , rho mesons , ... .
leaving them out would require us to abridge , but not to radically revise , the rosenfeld tables .
this is proved by the remarkable quantitative success , at the 5 - 10% level , of lattice gauge theory in the ` quenched ' approximation @xcite . for in this approximation , the influence of the heavier quarks on the lighter ones is systematically ignored .
the only major effect of putting the u and d masses to zero is to make the pions , which are already quite light by hadronic standards , strictly massless .
the perturbation to the proton mass , for instance , can be related using chiral perturbation theory to the value of the so - called @xmath8 term , a directly measurable quantity @xcite .
when this is done , one finds that the u and d quark masses are responsible for only about 5% of the proton mass .
thus qcd lite provides a truly remarkable realization of john wheeler s program , `` getting its from bits '' .
for here we encounter an extremely rich and complex class of physical phenomena including , in principle , nuclear and particle spectra that can be calculated , accurately and without ambiguity , using as sole inputs the numbers 3 and 2 .
i will not waste a lot of words on this , showing instead a few pictures , each worth many thousands of words .
figure 1 @xcite displays graphically that many independent types of experiments at different energy scales have yielded determinations of the strong coupling constant , all consistent with the predicted running . the overall accuracy and consistency of this phenomenology
is reflected in the precision with which this coupling is determined , to wit 5% ( at the z mass ) .
a remarkable feature of the theory is that a wide range of possible values for the coupling at relatively small energies focuses down to quite a narrow range at the highest accessible energies .
thus any `` reasonable '' choice of the scale at which the coupling becomes numerically large leads , within a few per cent , to a unique value of the coupling at the z mass .
our successful qcd predictions for high energy experiments have no wiggle - room ! while figure 1 is impressive , it does not do complete justice to the situation . for several of the experimental ` points '
each represents a summary of hundreds of independent measurements , any one of which might have invalidated the theory , and which display many interesting features .
figures 2 and 3 partially ameliorate the omission .
figure 2 @xcite shows some of the experimental data on deep inelastic scattering all subsumed within the ` dis ' point in figure 1 unfolded to show the complete @xmath9 and @xmath10 dependence .
our predictions @xcite for the pattern of evolution of structure functions with @xmath9 decrease at large @xmath10 , increase at small @xmath10 are now confirmed in great detail , and with considerable precision . particularly spectacular is the rapid growth at small @xmath10 .
this was predicted @xcite , in the form now observed , very early on .
however , even at the time we realized this rise could not continue forever .
the proliferating partons begin to form a dense system , and eventually one must cease to regard them as independent . there is a very interesting many - body problem developing here , which seems ripe for experimental and theoretical investigation , and may finally allow us to make contact between microscopic qcd and the remarkably successful regge - pole phenomenology .
figure 3 @xcite displays a comparison of the experimental distribution of jet energies , in 3-jet events , with the qcd prediction .
shown is the energy fraction of the second hardest jet , compared to its kinematic maximum .
for a detailed explanation , see @xcite .
the rise at @xmath11 reflects the singularity of soft gluon bremsstrahlung , and matches the prediction of qcd ( solid line ) . for comparison ,
the predictions for hypothetical scalar gluons are shown by the dotted line .
this is as close to a direct measurement of the core interaction of qcd , the basic quark - gluon vertex , as you could hope to see .
the other piece of this figure displays a related but more sophisticated comparison , using the ellis - karliner angle .
finally figure 4 @xcite shows the comparison of the qcd predictions to the spectrum of low - lying hadrons .
unlike what was shown in the previous two figures , and most of the points in figure 1 , this tests the whole structure of the theory , not only its perturbative aspect .
the quality of the fit is remarkable .
note that only one adjustable parameter ( the strange quark mass ) and one overall choice of normalization go into the calculation .
otherwise it s pure `` its from bits '' .
improvements due to enhanced computing power and to the use of domain wall quarks @xcite , that more nearly respect chiral symmetry , are on the horizon .
since there seems to be much confusion ( and obfuscation ) on the point , let me emphasize an aspect of figure 4 that ought to be blindingly clear : what you do nt see in it .
you do nt see massless degrees of freedom with long - range gauge interactions , nor parity doublets .
that is , confinement and chiral symmetry breaking are simply true facts about the solution of qcd , that emerge by direct calculation . the numerical work has taken us way beyond abstract discussion of these features .
finally , to justify the adverb in `` _ most _ perfect '' i must briefly recall for you some prominent flaws in our other best theories of physics , which qcd does not share .
quantum electrodynamics is of course extremely useful incomparably more useful than qcd and successful in practice .
but there is a worm in its bud .
it is not asymptotically free . treated outside of perturbation theory , or extrapolated to extremely high energy , qed becomes internally inconsistent .
modern electroweak theory shares many of the virtues of qed , but it harbors the same worm , and in addition contains many loose ends and continuous free parameters .
general relativity is the deepest and most beautiful theory of all , but it breaks down in several known circumstances , producing singularities that have no meaningful interpretation within the theory . nor
does it mesh seamlessly with the considerably better tested and established framework we use for understanding the remainder of physics . specifically , general relativity is notoriously difficult to quantize . finally , it begs the question of why the cosmological term is zero , or at least fantastically small when measured in its natural units .
superstring theory promises to solve some and conceivably all of these difficulties , and to provide a fully integrated theory of nature , but i think it is fair to say that in its present form superstring theory is not defined by clear principles , nor does it provide definite algorithms to answer questions within its claimed scope , so there remains a big gap between promise and delivery .
the behavior of qcd at high temperature , and low baryon number density , is relevant to cosmology indeed , it describes the bulk of matter filling the universe , during the first few seconds of the big bang and to the description of both numerical and physical experiments .
there are ambitious experimental programs planned for rhic , and eventually lhc , to probe this physics .
it is also , i think , intrinsically fascinating to ask what happens to empty space , if you keep adding heat ?
the equilibrium thermodynamics of qcd at finite temperature ( and zero chemical potential ) is amenable to direct simulation , using the techniques of lattice gauge theory .
figure 5 @xcite does not quite represent the current , rapidly evolving , state of the art , but it does already demonstrate some major qualitative points .
the chiral symmetry breaking condensate , clearly present ( as previously advertised ) at zero temperature , weakens and seems to be gone by @xmath12 .
likewise at these temperatures there is a sizable increase in the value of the polyakov loop , indicating that the force between distant color sources has considerably weakened .
furthermore the energy density increases rapidly , approaching the value one would calculate for an ideal gas of quarks and gluons .
the pressure likewise increases rapidly , but lags somewhat behind the ideal gas value .
all these phenomena indicate that at these temperatures and above a description using quarks and gluons as the degrees of freedom is much simpler and more appropriate than a description involving ordinary hadrons . indeed
, the quarks and gluons appear to be quasi - free .
that is what one expects , from asymptotic freedom , for the high - energy modes that dominate the thermodynamics .
it is an interesting challenge to reproduce the pressure analytically @xcite .
since the only scale in the problem is the large temperature , if one can organize the calculation so as to avoid infrared divergences , asymptotic freedom will legitimize a weak coupling treatment .
even more interesting would be to do this by a method that also works at finite density , since the equation of state is of great interest for astrophysics and is _ not _ accessible numerically .
there is no doubt , in any case , that qcd predicts the existence of a quark - gluon plasma phase , wherein its basic degrees of freedom , normally hidden , come to occupy center stage .
while transition to a quark - gluon plasma at asymptotically high temperatures is not unexpected , the abruptness and especially the _ precocity _ of the change is startling .
below 150 mev the only important hadronic degrees of freedom are the pions .
why does this rather dilute pion gas suddenly go berserk ?
the change is enormous , quantitatively .
the pions represent precisely 3 degrees of freedom .
the free quarks and gluons , with all their colors , spins , and antiparticles , represent 52 degrees of freedom .
there are many ideas for detecting signals of quark - gluon plasma formation in heavy ion collisions , which you will be hearing much of .
i would like discuss briefly a related but more focused question , on which there has been dramatic progress recently @xcite .
this is the question , whether there is a true phase transition in qcd accessible to experiment .
one might think that the answer is obviously `` yes '' , since there are striking differences between ordinary hadronic matter and the quark - gluon plasma .
this is not decisive , however .
let me remind you that the dissociation of ordinary atomic gases into plasmas is not accompanied by a phase transition , even though these states of matter are very different ( so different , that at princeton they are studied on separate campuses ) .
similarly , confinement of quarks is believed to go over continuously , at high temperature , into screening certainly , no one has demonstrated the existence of an order parameter to distinguish between them . what about chiral symmetry restoration ?
for massless quarks , there is a definite difference between the low - temperature phase of broken chiral symmetry and the quasi - free phase with chiral symmetry restored , so there must be a phase transition . a rather subtle analysis using the renormalization group indicated that for two massless quarks one might have a second - order phase transition , while for three massless
quarks it must be first - order @xcite .
this is the pattern observed in numerical simulations .
the real world has two very light quarks and one ( the strange quark ) whose mass is neither clearly small nor clearly large compared to basic qcd scales . here
, then , a sharp question emerges .
if the strange quark is effectively heavy , and the other quarks are taken strictly massless , we should have a second - order transition .
if the strange quark is effectively light , we should have a first - order transition . which is the case , for the physical value of the strange quark mass ?
although this has been a controversial question , there seems to be an emerging consensus among lattice gauge theorists that it is second - order .
unfortunately , this means that with small but finite @xmath13 and @xmath14 quark masses we will not have a sharp phase transition at all , but only a crossover . and not a particularly sharp one , at that .
for while we are ordinarily encouraged to treat these masses as small perturbations , they are responsible for the pion masses , which are far from negligible at the temperatures under discussion .
so the relevant correlation length never gets very large .
nevertheless it was interesting to point out @xcite that in the @xmath15 plane one could naturally connect the first- and second- order behaviors , as in figure 6 .
the line of first - order transitions ends at a tricritical point .
this is a true critical point , with diverging correlation lengths and large fluctuations . the first - order line ,
since it is a locus of discontinuities , and therefore the existence of a tricritical point where it terminates , are features which survive the small perturbation due to non - zero light quark masses .
all these statements can be tested against numerical simulation .
stephanov , rajagopal and shuryak @xcite have brought the subject to a new level of interest , taking off from the simple but brilliant observation that one expects similar behavior in the @xmath16 plane .
the big advantage of this is , that while @xmath17 is not a control parameter one can vary experimentally , the chemical potential @xmath18 is .
they have proposed quite specific , characteristic signatures for passage near this transition in the thermal history of a fireball , such as might be obtained in heavy ion collisions @xcite .
the signatures involve enhanced fluctuations and excess , non - thermal production of low - energy pions .
i believe it ought to be possible to refine the prediction , by locating the tricritical point theoretically . for
while it is notoriously difficult to deal with large chemical potentials at small temperature numerically , there are good reasons to be optimistic about high temperatures and relatively small chemical potentials , which is our concern here @xcite .
if all these strands can be brought together , it will be a wonderful interweaving of theory , experiment , and numerics .
the behavior of qcd at high baryon number density , and low temperature , is of direct interest for describing neutron star interiors , neutron star collisions , and events near the core of collapsing stars .
unfortunately , it has proved quite difficult to calculate this behavior directly numerically using lattice gauge theory techniques .
this is because in the presence of a chemical potential the functional integral for the partition function is no longer positive definite ( or even real ) configuration by configuration , so importance sampling fails , and the calculation converges only very slowly . on the other hand , there has been remarkable progress on this problem over the last year or two using analytical techniques .
this has shed considerable new light on many aspects of qcd .
we have new , fully calculable mechanisms for confinement and chiral symmetry breaking @xcite .
amazingly , we find that two famous , historically influential `` mistakes '' from the prehistory of qcd the han - nambu @xcite assignment of integer charge to quarks , and the sakurai @xcite model of vector mesons as yang - mills fields @xcite emerge from the microscopic theory at high density . and we find that in the slightly idealized version of qcd with three degenerate light quarks , there need be no phase transition separating the calculable high density phase from ( the appropriate version of ) nuclear matter @xcite ! at the request of the organizers i will be giving a separate seminar on these developments @xcite , so here i will be telegraphic . why might we expect qcd to become analytically tractable at high density ? at the crudest heuristic level , it is a case of asymptotic freedom meets the fermi surface .
let us suppose , optimistically , that a weak coupling treatment is going to be appropriate , and see where it leads .
if the coupling is weak and the density large , our first approximation to the ground state is large fermi balls for all the quarks . due to the pauli exclusion principle
, the modes deep within the ball will be energetically costly to excite , and the important low - energy degrees of freedom will be the modes close to the fermi surface .
but these modes will have large momentum . thus their interactions , generically , will either hardly deflect them , or will involve large momentum transfer .
in the first case we do nt care , while the second involves a weak coupling , due to asymptotic freedom . on reflection ,
one perceives two big holes in this argument .
first , it does nt touch the gluons .
they remain massless , with singular interactions and strong couplings in the infrared that do not appear to be under control .
second , as we learn in the theory of superconductivity , the fermi surface is generically unstable , even at weak coupling .
this is because pairs of particles ( or holes ) of equal and opposite momenta are low - energy excitations which can all scatter into one another .
thus one is doing highly degenerate perturbation theory , and in that circumstance even a small coupling can have large qualitative effects .
fortunately , our brethren in condensed matter physics have taught us how to deal with the second problem , and its proper treatment also cures the first .
there is an attractive interaction between quarks on the opposite sides of the fermi surface , and they pair up and condense . in favorable cases and in particular , for three degenerate or nearly degenerate flavors this color superconductivity produces a gap for all the fermion excitations , and also gives mass to all the gluons .
thus a proper weak - coupling treatment automatically avoids all potential infrared divergences , and our optimistic invocation of asymptotic freedom provides , at asymptotically high density , its own justification .
the different components of the standard model have a similar mathematical structure , all being gauge theories .
their common structure encourages the speculation that they are different facets of a more encompassing gauge symmetry , in which the different strong and weak color charges , as well as electromagnetic charge , would all appear on the same footing .
the multiplet structure of the quarks and leptons in the standard model fits beautifully into small representations of unification groups such as @xmath19 or @xmath20 .
there is the apparent difficulty , however , that the coupling strengths of the different standard model interactions are widely different , whereas the symmetry required for unification requires that they share a common value .
the running of couplings suggests an escape from this impasse @xcite . since the strong , weak , and electromagnetic couplings run at different rates ,
their inequality at currently accessible scales need not reflect the ultimate state of affairs .
we can imagine that spontaneous symmetry breaking a soft effect has hidden the full symmetry of the unified interaction .
what is really required is that the fundamental , bare couplings be equal , or in more prosaic terms , that the running couplings of the different interactions should become equal beyond some large scale . using simple generalizations of the formulas derived and tested in qcd , which are none other than the ones experimentally validated in figure 1
, we can calculate the running of all the couplings , to see whether this requirement is met . in doing
so one must make some hypothesis about the spectrum of virtual particles .
if there are additional massive particles ( or , better , fields ) that have not yet been observed , they will contribute significantly to the running of couplings once the scale exceeds their mass .
let us first consider the default assumption , that there are no new fields beyond those that occur in the standard model .
the results of this calculation are displayed in figure 7 @xcite .
considering the enormity of the extrapolation this works remarkably well , but the accurate experimental data indicates unequivocally that something is wrong .
there is one particularly attractive way to extend the standard model , by including supersymmetry .
supersymmetry can not be exact , but if it is only mildly broken ( so that the superpartners have masses @xmath21 1 tev ) it can help explain why radiative corrections to the higgs mass parameter , and thus to the scale of weak symmetry breaking , are not enormously large . in the absence of supersymmetry power counting would indicate a hard , quadratic dependence of this parameter on the cutoff .
supersymmetry removes the most divergent contribution , by canceling boson against fermion loops . if the masses of the superpartners are not too heavy , the residual finite contributions due to supersymmetry breaking will not be too large .
the minimal supersymmetric extension of the standard model , then , makes definite predictions for the spectrum of virtual particles starting at 1 tev or so .
since the running of couplings is logarithmic , it is not extremely sensitive to the unknown details of the supersymmetric mass spectrum , and we can assess the impact of supersymmetry on the unification hypothesis quantitatively .
the results , as shown in figure 8 @xcite , are quite encouraging .
a notable result of the unification of couplings calculation , especially in its supersymmetric form , is that the unification occurs at an energy scale which is enormously large by the standards of traditional particle physics , perhaps approaching @xmath22 gev . from a phenomenological viewpoint , this is fortunate .
the most compelling unification schemes merge quarks , antiquarks , leptons , and antileptons into common multiplets , and have gauge bosons mediating transitions among all these particle types .
baryon number violating processes almost inevitably result , whose rate is inversely proportional to the fourth power of the gauge boson masses , and thus to the fourth power of the unification scale . only for such large values of the scale
is one safe from experimental limits on nucleon instability . from a theoretical point of view the large scale
is fascinating because it brings us , from the internal logic of particle physics , to the threshold of quantum gravity .
i find it quite remarkable that the logarithmic running of couplings , discovered theoretically and now amply verified within qcd , permits a meaningful quantitative discussion of these extremely ambitious and otherwise thinly rooted ideas , and even allows us to discriminate between different possibilities ( especially , susy vs. non - susy ) .
since qcd is our most perfect example of a fundamental theory of nature , it is appropriate to use it as a basis for drawing broad conclusions about how nature works , or , in other words , for `` natural philosophy '' .
_ alien _ : as has been the case for all fundamental physical theories since galileo , qcd is formulated in abstract mathematical terms . in particular , there are no hints of moral concepts or purposes . nor , in thousands of rigorous experiments , have we encountered any signs of active intervention in the unfolding of the equations according to permanent laws .
_ weird _ : quantum mechanics is notoriously weird , and qcd incorporates it in its marrow .
less remarkable , but to me no less weird , is the need to define qcd through a limiting procedure .
this would seem to be a rather difficult and inefficient way to run a universe . _ comprehensible _ : qcd wonderfully illustrates einstein s remark , `` the most incomprehensible thing about nature is that it is comprehensible . ''
one hundred years ago people did not know there were such things as an atomic nuclei and a strong interaction ; just over fifty years ago the pion and kaon were discovered ; just over twenty - five years ago the strong interaction problem still seemed hopelessly intractable .
that we , collectively , have got from there to here so quickly , against overwhelming odds , is an extraordinary achievement .
it is a tribute to our culture , and to the glory of the human mind . and
it is there that where we must locate , for now , the most incomprehensible thing about nature .
* figure 1 * experimental verification of the running of the coupling , as predicted in @xcite .
the determinations , running from left to right , are from : corrections to bjorken sum rule , corrections to gross - llewyllen - smith sum rule , hadronic width of @xmath23 lepton , b @xmath24 threshold production , prompt photon production in pp and @xmath25 collisions , scaling violation in deeply ineslastic scattering , lattice gauge theory calculations for heavy quark spectra , heavy quarkonium decays , shape variables characterizing jets at different energies ( white dots ) , total @xmath26 annihilation cross section , jet production in semileptonic and hadronic processes , energy dependence of photons in z decay , w production , and electroweak radiative corrections . from @xcite .
* figure 5 * top part : evolution of the energy and of the pressure of 2-flavor qcd as a function of temperature , showing precocious and rapid approach to a quasi - free quark - gluon plasma .
bottom part : evolution of the chiral condensation order parameter and of the polyakov loop , which is a measure of the inverse induced mass of an inserted color source , and vanishes in a confined phase .
one sees clear signals of chiral symmetry restoration and deconfinement @xcite .
* figure 6 * connecting the second- and first- order chiral symmetry restoration transitions predicted for two , respectively three , light quark flavors .
the end of the first - order line is a tricritical point .
a similar diagram may be valid for fixed strange quark mass and varying chemical potential .
* figure 7 * running of the couplings extrapolated toward very high scales , using just the fields of the standard model .
the couplings do not quite meet .
experimental uncertainties in the extrapolation are indicated by the width of the lines @xcite .
* figure 8 * running of the couplings extrapolated to high scales , including the effects of supersymmetric particles starting at 1 tev . within experimental and theoretical uncertainties
, the couplings do meet @xcite . | i discuss why qcd is our most perfect physical theory .
then i visit a few of its current frontiers .
finally i draw some appropriate conclusions . |
a * proper colouring * of a graph is a labeling of its vertices with colours such that no two vertices sharing the same edge have the same colour .
a colouring using at most @xmath3 colours is called a * proper @xmath3-colouring*. the smallest number of colours needed to colour a graph @xmath4 is called its * chromatic number * , and it is denoted by @xmath5 . in this paper we are concerned with another notion of colouring , first introduced by chartrand et al . @xcite . for
a given ( not necessarily proper ) @xmath3-colouring @xmath6 $ ] of the vertex set of @xmath7 , let @xmath8 be the * neighbourhood colour set * of a vertex @xmath9 .
( in this paper , @xmath10:= \{1 , 2 , \ldots , k\}$ ] . )
the colouring @xmath11 is a * set colouring * if @xmath12 for every pair of adjacent vertices in @xmath4 .
the minimum number of colours , @xmath3 , required for such a colouring is the * set chromatic number * @xmath13 of @xmath4 .
one can show that @xmath14 indeed , the upper bound is trivial , since any proper colouring @xmath11 is also a set colouring : for any edge @xmath15 , @xmath16 , the neighbourhood of @xmath17 , contains @xmath18 whereas @xmath19 does not . on the other hand ,
suppose that there is a set colouring using at most @xmath3 colours .
since there are at most @xmath20 possible neighbourhood colour sets , one can assign a unique colour to each set obtaining a proper colouring using at most @xmath20 colours .
we get that @xmath21 , or equivalently , @xmath22 . with slightly more work
, one can improve this lower bound by 1 ( see @xcite ) , which is tight ( see @xcite ) .
let us recall a classic model of random graphs that we study in this paper .
the * binomial random graph * @xmath0 is the random graph @xmath4 with vertex set @xmath23 $ ] in which every pair @xmath24}{2}$ ] appears independently as an edge in @xmath4 with probability @xmath2 .
note that @xmath1 may ( and usually does ) tend to zero as @xmath25 tends to infinity . [ cols="^,^ " , ] now we state the main result of the paper .
[ thm : main ] suppose that @xmath1 is such that @xmath26 for some @xmath27 .
let @xmath28 .
then , the following holds a.a.s . 1 .
if @xmath2 is a constant , then @xmath29 2 .
if @xmath30 and @xmath31 for some @xmath32 $ ] , then @xmath33 3 . if @xmath34 , then @xmath35 note that the result is asymptotically tight for dense graphs ( that is , for @xmath36 ; see part ( i ) and part ( ii ) for @xmath37 ) . for sparser graphs ( part ( ii ) for @xmath38 ) the ratio between the upper and the lower bound is a constant that gets large for @xmath39 small . on the other hand ,
the trivial lower bound of @xmath40 ( see ) gives us the following : a.a.s .
@xmath41 provided that @xmath42 as @xmath43 , and @xmath30 ; @xmath44 otherwise ( see @xcite ) .
so the lower bound we prove is by a multiplicative factor of @xmath45 larger than the trivial one , provided that @xmath46 . if @xmath47 for some @xmath48 , then our bound is by a factor of @xmath49 better than the trivial one .
this seemingly small improvement is important to obtain the asymptotic behaviour in the case @xmath50 , and in particular , to obtain the zig - zag for constant @xmath2 .
the upper and the lower bounds are proved in section [ sec : upper_bound ] and section [ sec : lower_bound ] , respectively .
let us also mention that , in fact , the two bounds proved below are slightly stronger .
in particular , the upper bound holds for @xmath51 , the point where the trivial bound of @xmath52 becomes stronger .
we will use the following version of * chernoff s bound*. suppose that @xmath53 is a binomial random variable with expectation @xmath54 . if @xmath55 , then @xmath56 \le \exp \left ( -\frac{\delta^2 \mu}{2 } \right),\ ] ] and if @xmath57 , @xmath58 \le \exp\left(-\frac{\delta^2 \mu}{2+\delta}\right).\ ] ] these inequalities are well known and can be found , for example , in @xcite .
we will also use * suen s inequality * that was introduced in @xcite and revised in @xcite . for a finite set @xmath59 ,
let @xmath60 , and for any @xmath61 , let @xmath62 be some event with the corresponding indicator random variable @xmath63 .
( in our application , @xmath62 will be the event that vertices @xmath64 and @xmath65 have the same neighbourhood colour sets . )
let @xmath66 be the random variable counting how many such events occur .
the associated * dependency graph * has @xmath67 as its vertex set , and @xmath68 if and only if @xmath69 .
suen s inequality asserts that @xmath70 where @xmath71,\\ \delta & = & \max_{(x_1,y_1 ) \in { \mathcal i } } \sum_{(x_2 , y_2 ) \sim ( x_1 , y_1 ) } { \mathbb{p}}(a_{x_2 , y_2}).\end{aligned}\ ] ]
we start by proving and .
since @xmath72 ^ 2 + [ 1-(1-p)^\ell]^2 = 2\left [ ( 1-p)^{\ell } -1/2\right]^2 + 1/2,\ ] ] it follows that @xmath73 , and consequently also holds . now , let @xmath74 where @xmath75 .
observe that @xmath76 ^ 2 + [ 1-(1-p)^\ell]^2 = \frac{[1-(1-p)^{\delta}]^2 + 1}{2 } \le \frac{[1-(1-p)]^2 + 1}{2 } = \frac{p^2 + 1}{2}.\ ] ] implying the upper bound in .
we keep the definition of function @xmath77 for constant @xmath2 introduced above ( see ( [ eq : r ] ) ) .
we extend it here for sparser graphs as follows : suppose that @xmath2 tends to zero as @xmath43 , and that @xmath31 for some @xmath78 $ ] . then , we define @xmath77 such that @xmath79 , that is , @xmath80 since it follows from that @xmath81 .
the upper bound in theorem [ thm : main ] follows immediately from the next lemma .
[ lem : upper_bound ] suppose that @xmath1 is such that @xmath82 for some fixed @xmath27 .
let @xmath28 .
then , a.a.s .
@xmath83 before we move to the proof , let us note that the lower bound for @xmath2 is not necessary , and the result can be extended to sparser graphs .
the reason it is introduced here is that for sparser graphs , the trivial upper bound of @xmath5 is stronger ; note that a.a.s .
@xmath84 provided that @xmath42 as @xmath43 , and @xmath30 ; @xmath85 otherwise .
the proof is straightforward .
let @xmath86 be any function tending to infinity with @xmath25 ( slowly enough ) . before exposing the edges of the ( random ) graph @xmath4 , we partition
( arbitrarily ) the vertex set into @xmath87 sets , each consisting of @xmath88 * important * vertices , and one remaining set of vertices , these being not important .
( for expressions such as @xmath87 that clearly have to be an integer , we round up or down but do not specify which : the choice of which does not affect the argument . )
note that @xmath89 and so there are enough vertices to perform this operation . all vertices in a given set receive the same colour , and hence the total number of colours is equal to @xmath90 . for a given pair of vertices , @xmath91
, we need to estimate from above the probability @xmath92 that they have the same neighbourhood colour sets .
we do it by considering sets of important vertices that neither @xmath64 nor @xmath65 belong to .
let @xmath93 be the set of ( important ) vertices of the same colour , and let @xmath94 .
then , either both @xmath64 and @xmath65 are not connected to any vertex from @xmath93 , yielding the contribution @xmath95 ^ 2 $ ] to the probability @xmath92 , or both @xmath64 and @xmath65 are connected to at least one vertex from @xmath93 , giving the contribution @xmath96 ^ 2 $ ] .
thus , @xmath97 ^ 2 + [ 1-(1-p)^{\ell_0}]^2 \big)^{r \lg n + \omega - 2 } = s^{r \lg n + \omega - 2}.\ ] ] hence , the expected number of pairs of adjacent vertices that are _ not _ distinguished by their neighbourhood colour sets is at most @xmath98 where the last equality follows from the definition of @xmath99 .
finally , by , we get that @xmath100 and so @xmath101 tends to zero as @xmath102 .
hence , the lemma follows by markov s inequality .
before we move to the proof of the lower bound , we need the following technical lemma .
let @xmath103 and @xmath104 be fixed positive real numbers . first we show that there exist unique numbers @xmath106 and @xmath105 satisfying .
since @xmath114 is increasing on @xmath115 $ ] , we get @xmath116 clearly , @xmath117 is also increasing and continuous on @xmath118 $ ] , and thus , there is a unique real number @xmath106 such that @xmath119 and @xmath120 to finish the proof of , set @xmath121 and observe that @xmath122 , since @xmath123 .
now we move to the proof of .
let @xmath124 since for every real number @xmath125 , @xmath126 we get @xmath127 furthermore , @xmath128 and so @xmath129 in order to show the inequality in it suffices to prove @xmath130 which is equivalent to @xmath131 and subsequently to @xmath132 set @xmath133 now , showing ( and hence also ) is equivalent to showing @xmath134 the latter inequality immediately from hlder s inequality ( see , for example , @xcite ) : indeed , let @xmath135 ( observe that @xmath136 and @xmath137 are well - defined since @xmath138 and @xmath139 . ) then , since @xmath140 , @xmath141 , and @xmath142 , hlder s inequality yields @xmath143 as required .
finally note that the equality in follows from applied with @xmath144 .
the proof of the lemma is finished .
as we did in the previous section , we keep the definition of the function @xmath77 for constant @xmath2 ( see ( [ eq : r ] ) ) .
we extend it here for sparser graphs as follows ( in a different way than in the previous section ) : suppose that @xmath2 tends to zero as @xmath43 and that @xmath31 for some @xmath78 $ ] .
this time , @xmath77 is defined such that @xmath145 , that is , @xmath146 since @xmath73 . first , let us note that , since the expected degree tends to infinity faster than @xmath151 , it follows immediately from chernoff s bound and the union bound that a.a.s .
all vertices have degree at most , say , @xmath152 .
hence , since we aim for a statement that holds a.a.s . , we may assume that the maximum degree of @xmath4 is at most @xmath152 .
in fact , the argument is slightly more delicate and will be explained soon .
suppose that we are given a colouring of the vertices .
we partition all colours into * important * and * unimportant * ones : a colour is important if the number of vertices of that colour is at most @xmath153 . first , let us show that unimportant colours can distinguish only a few edges .
formally , we claim the following : + + * claim : * a.a.s .
each set of @xmath153 vertices dominates all but at most @xmath153 vertices .
+ + _ proof of the claim . _
note that the expected number of pairs of disjoint sets of size @xmath153 with no edge between them is at most @xmath154 the claim follows from the first moment method .
@xmath155 hence ,
if @xmath2 is constant , then @xmath156 unimportant colours can distinguish only @xmath157 vertices .
all remaining vertices will have _ all _ unimportant colours present in their neighbourhood colour sets ; as a result , no edge in the graph induced by these vertices can be distinguished by unimportant colours . on the other hand , if @xmath30 , then at most @xmath158 unimportant colours can distinguish at most @xmath159 vertices , since @xmath160 . as a consequence of the claim , we may concentrate on important colours from now .
suppose that a colouring @xmath161 $ ] using @xmath3 important colours is fixed ; @xmath162 is the set of vertices coloured with important colours . moreover , let us fix a set @xmath163 of @xmath164 vertices that are ( possibly ) distinguished by unimportant colours .
our goal is to estimate the probability @xmath165 that important colours distinguish endpoints of edges in @xmath166 $ ] , which is the graph induced by those vertices that are coloured with unimportant colours and that are adjacent to at least one vertex from each unimportant colour class .
since the number of configurations to investigate is at most @xmath167 it is enough to estimate @xmath165 from above by , say , @xmath168 the result will then follow immediately by the union bound .
the expected number of edges in @xmath166 $ ] is @xmath169 , and so it follows from chernoff s bound that with probability at most @xmath170 the number of edges is smaller than , say , @xmath171 . on the other hand , if the maximum degree in @xmath166 $ ] is larger than @xmath152 ( for some configuration @xmath172 ) , then we stop the whole argument and claim no lower bound for @xmath13 . recall that at the beginning of the proof , we showed that a.a.s
clearly , if this is the case , then ( deterministically ) the degree of each vertex in @xmath166 $ ] is at most @xmath152 .
hence , we may condition on the event that the graph @xmath166 $ ] has the following two properties : ( i ) the number of edges is at least @xmath174 , and ( ii ) no vertex has degree more than @xmath152 .
it is important that no edge between @xmath175 and @xmath176 has been exposed yet .
let us focus on constant @xmath2 first .
suppose that the number of important colours is equal to @xmath177 where the error term follows from and from our assumption that @xmath178 from which we get , as before , @xmath179 .
suppose that for a given configuration @xmath172 , the probability @xmath92 that two adjacent vertices @xmath91 from @xmath175 are not distinguished by important colours is @xmath180 for some @xmath181 .
( recall that @xmath182 is the lower bound for @xmath92 which can be attained when all colour classes have size @xmath88 . )
we will use suen s inequality to obtain an upper bound for @xmath165 .
let @xmath183 for any @xmath61 , let @xmath62 be the event ( with the corresponding indicator random variable @xmath63 ) that the neighbourhood colour sets ( restricted to important colours only ) of @xmath64 and @xmath65 are equal .
let @xmath66 .
we wish to estimate the probability that @xmath184 .
denote by @xmath185 ( for @xmath186 ) the number of vertices coloured by colour @xmath187 .
suppose first that @xmath188 .
this means @xmath189 ^ 2 + [ 1-(1-p)^{\kappa_i}]^2 \right ) = s^k.\ ] ] in this case , @xmath190 where the last equality follows from the definition of @xmath99 .
observe that for @xmath191 and @xmath192 in @xmath67 we get @xmath193 ^ 3 + [ 1-(1-p)^{\kappa_i}]^3 \right).\ ] ] thus , since the number of pairs @xmath191 and @xmath192 is bounded by @xmath194 , we have @xmath195 ^ 3 + [ 1-(1-p)^{\kappa_i}]^3 \right)\\ & = | { \mathcal i}| \cdot s^k \cdot 4pn \cdot \frac{\prod_{i=1}^k \left ( [ ( 1-p)^{\kappa_i}]^3 + [ 1-(1-p)^{\kappa_i}]^3 \right)}{\prod_{i=1}^k \left ( [ ( 1-p)^{\kappa_i}]^2 + [ 1-(1-p)^{\kappa_i}]^2 \right)}\\ & = \mu \cdot 4pn \cdot \frac{\prod_{i=1}^k \left ( [ ( 1-p)^{\kappa_i}]^3 + [ 1-(1-p)^{\kappa_i}]^3 \right)}{\prod_{i=1}^k \left ( [ ( 1-p)^{\kappa_i}]^2 + [ 1-(1-p)^{\kappa_i}]^2\right ) } .\end{aligned}\ ] ] now , using , we apply corollary [ cor : ratio ] with @xmath196 if @xmath197 , and @xmath198 otherwise , to get @xmath199 now we will prove that @xmath200 . using taylor expansion at @xmath201 , one can show that for any @xmath202 $ ] we have @xmath203 furthermore , since together with @xmath204 implies @xmath205 we get @xmath206 , and hence @xmath207 finally , @xmath208 it follows from suen s inequality ( see ( [ eq : suen ] ) ) that @xmath209 and @xmath210 , as needed ( see ) .
suppose now that @xmath211 .
in this case , @xmath212 is larger than before but , unfortunately , @xmath213 grows faster than @xmath212 ( as @xmath125 grows ) and eventually becomes larger than @xmath212 . in order to avoid this undesired situation , we make the dependency graph sparser so that @xmath212 is still of order @xmath214 .
let us note that if @xmath215 , where @xmath17 is defined so that @xmath216 , then suen s inequality can be avoided and we can simply use chernoff s bound to obtain the desired upper bound for @xmath217 : indeed , it follows from the claim proved above that there exists a matching in @xmath166 $ ] consisting of at least @xmath218 edges ; otherwise , the remaining @xmath219 vertices in @xmath175 would form an independent set that could be split into two sets of equal size and , clearly , no edge will be present between them , contradicting the claim .
since the events are now independent , and the expected number of edge endpoints not distinguished is @xmath220 the desired bound holds , since @xmath221 by chernoff s bound .
it remains to consider the case @xmath222 .
our goal is to scale the degree in the dependency graph down by a multiplicative factor of @xmath223 let @xmath224 be a random subgraph of @xmath67 : each @xmath61 is independently put into @xmath224 with probability @xmath225 . since @xmath226 , @xmath227 , and thus a.a.s .
moreover , since the maximum degree in the dependency graph is at most @xmath152 , by chernoff s bound together with a union bound over all vertices , it follows that a.a.s .
the maximum degree in a random subgraph of it is at most @xmath229 therefore , the deterministic ( non - constructive ) conclusion is that there exists a subgraph of the dependency graph with at least @xmath230 pairs and the maximum degree at most @xmath231 . we restrict ourselves to this subgraph , stressing one more time that no edge between @xmath175 and @xmath176 is exposed yet .
now , @xmath232 moreover , @xmath233 let @xmath234 .
note that @xmath235 , @xmath236 is increasing on the interval @xmath237 $ ] attaining @xmath238 , and then is decreasing on the interval @xmath239 $ ] going back to @xmath240 .
hence , if @xmath241 , then @xmath242 , as a function of @xmath125 , is maximized for @xmath243 .
then , as a function of @xmath105 , since @xmath241 , @xmath244 , and @xmath245 , @xmath246 is maximized for @xmath247 .
we are back to the case @xmath188 that we already checked . on the other hand , if @xmath248 , then , since @xmath181 , @xmath236 and therefore @xmath213 is maximized again for @xmath188 .
hence , in both cases we have @xmath249 finally , @xmath250 hence , suen s inequality can be applied as before , and the proof for constant @xmath2 is finished .
the case @xmath30 can be verified exactly the same way .
in fact , it is slightly easier since @xmath251 , @xmath252 , and we do not have to worry about an increasing value of @xmath105 ( and therefore , neither about an increasing value of @xmath99 ) .
we point out only the adjustments of the proof .
recall that the definition of @xmath99 is extended to the case @xmath30 .
the number of colours is now @xmath253 , and we have @xmath254 . for the case @xmath188 , we have @xmath255 as desired for the union bound ( see ) .
since now @xmath251 , the argument for @xmath213 is much easier : @xmath256 since @xmath257 .
( note that for @xmath258 but still @xmath30 , we have @xmath259
. however , this causes no problem , as @xmath260 and so @xmath261 .
otherwise , that is , if @xmath262 , we have @xmath263 and @xmath261 follows easily . ) finally , as before , and again using the same lower bound on @xmath264 , we have @xmath265 and suen s inequality can be applied .
now let us consider the case @xmath266 .
the definition of @xmath17 is not affected , and for @xmath215 we use chernoff s bound since the events are independent .
the only difference is that the new value of @xmath182 has to be used to get @xmath267 as desired . for the case @xmath222 ,
the definition of @xmath225 remains the same and again , after adjusting the value of @xmath182 we get @xmath268 .
the argument for @xmath261 is not affected .
finally , @xmath269 provided that @xmath270 . for
slightly sparser graphs , that is , when @xmath271 , observe that we only have that @xmath272 , but in fact we can show a stronger bound for @xmath213 : it follows that @xmath273 and thus @xmath274 hence , @xmath275 , and so @xmath276 , as needed for suen s inequality to be useful .
the proof is finished . | in this paper we study the set chromatic number of a random graph @xmath0 for a wide range of @xmath1 .
we show that the set chromatic number , as a function of @xmath2 , forms an intriguing zigzag shape .
[ section ] [ theorem]lemma [ theorem]definition [ theorem]conjecture [ theorem]proposition [ theorem]algorithm [ theorem]corollary [ theorem]observation [ theorem]open problem [ theorem]remark |
10 -0.2ex plus0.05ex minus0.05ex here we refer to standard textbooks like ballhausen c j , _ introduction to ligand field theory _ 1962 mcgraw - hill ; sugano s , tanabe y , kamimura h , _ multiplets of transition - metal ions in crystals _ 1972 academic press haverkort m w , hu z , tanaka a , ghiringehelli g , roth h , cwik m , lorenz t , schler - langeheine c , streltsov s v , mylnikova a s , anisimov v i , de nadai c , brookes n b , hsieh h h , lin h j , chen c t , mizokawa t , taguchi y , tokura y , khomskii d i , tjeng l h 2005 phys .
lett . * 94 * 056401 rckamp r , benckiser e , haverkort m w , roth h , lorenz t , freimuth a ,
jongen l , mller a , meyer g , reutler p , bchner b , revcolevschi a , cheong s w , sekar c , krabbes g , grninger m 2005 new journal of physics * 7 * 144 | we present a study of the magnetic susceptibility of la@xmath0sr@xmath1coo@xmath2 single crystals in a doping range @xmath3 .
our data shows a pronounced magnetic anisotropy for all compounds .
this anisotropy is in agreement with a low - spin ground state ( @xmath4 ) of co@xmath5 for @xmath6 and a high - spin ground state ( @xmath7 ) of co@xmath8 .
we compare our data with a crystal - field model calculation assuming local moments and find a good description of the magnetic behavior for @xmath9 .
this includes the pronounced kinks observed in the inverse magnetic susceptibility , which result from the anisotropy and low - energy excited states of co@xmath8 and are not related to magnetic ordering or temperature - dependent spin - state transitions .
transition - metal oxides are known for their complex interplay between different degrees of freedom like spin , charge and orbitals . in some of these systems
another interesting property is found : ions like co@xmath5 in a crystal - field environment can occur in different spin states .
additionally , transitions between different spin states are possible for some compounds . a prominent example showing this phenomenon is lacoo@xmath10 , which has been examined and discussed since the middle of the last century ( see e.g. refs .
@xcite ) .
the existence of different spin states arises from the competition between crystal - field effects and on - site coulomb interaction .
crystal fields lift the degeneracy of the @xmath11 states .
if the crystal field is strong , this can lead to a violation of hund s rules . in the case of cubic symmetry and in a one - electron picture , the five - fold degenerate @xmath11 states are split into a three - fold degenerate @xmath12and a two - fold degenerate @xmath13 level .
the splitting between @xmath12and @xmath13 states is called @xmath14 . with a strong crystal field ,
the electrons will be forced into the low - spin state ( ls ) . regarding a @xmath15 system , this state consists of an antiparallel alignment of spins with @xmath16 and @xmath4 . on the other hand ,
the coulomb interaction manifests itself in the effect of hund s coupling .
a parallel arrangement of spins minimizes the electron - electron repulsion because the pauli principle forces the electrons to occupy different orbitals . in a weak cubic crystal field
, this effect will dominate and lead to the high - spin state ( hs ) , which is the configuration with the highest total spin possible in accordance with the pauli principle . for a @xmath15 system , this configuration is @xmath17 with @xmath18 . for co@xmath5
the crossover between these two different spin states occurs roughly at an energy difference of @xmath19ev . in the case of lacoo@xmath10 ,
a low - spin ground state for co@xmath5 was found@xcite .
interestingly , the difference between the crystal field energies and the promotional energies is so small that an excited state with different spin state can be reached by thermal excitation .
this excited state has been a subject of debate for a long time .
apart from the hs state described above , the intermediate - spin state ( is , @xmath20 @xmath21 ) was also discussed@xcite and reported in many experiments , but it was shown that it might have been confused with a spin - orbit coupled hs state@xcite . for the layered cobaltates
la@xmath0sr@xmath1coo@xmath2 , much less is known about the spin state of co@xmath5 . based on magnetic
measurements@xcite a hs ground state for @xmath22 and a spin - state transition to an is ground state for @xmath23 was proposed .
this conclusion was based on a curie - weiss analysis of the susceptibility in a temperature range of 100k to 300k and nmr measurements@xcite .
unrestricted hartree - fock calculations@xcite showed a slightly different picture . here , three different magnetic phases were found .
an antiferromagnetic hs phase ( @xmath24 ) , a ferromagnetic hs phase ( @xmath25 ) and an antiferromagnetic ls - hs - ordered phase ( @xmath26 ) were proposed .
these calculations were also based on the results of the curie - weiss analysis in ref .
@xcite . taking into account
that la@xmath0sr@xmath1coo@xmath2 is an anisotropic material with rather strong spin - orbit coupling compared to the tetragonal crystal field splitting , the validity of the curie - weiss law is questionable .
the aim of this paper is to analyze the magnetic susceptibility from a different perspective , concentrating on the spin state of co@xmath5 .
the single crystals used for the magnetic measurements have been grown using the floating - zone technique in an image furnace .
a strontium doping range of @xmath27 was covered .
resistivity measurements revealed that la@xmath0sr@xmath1coo@xmath2 is a strong insulator for all sr doping concentrations .
the @xmath28 sample turned out to possess the highest resistivity , which is in accordance with the charge ordering of co@xmath8 and co@xmath5 at @xmath29k that has already been reported@xcite .
the magnetization was measured with a quantum design vibrating sample magnetometer ( vsm ) .
the field was aligned parallel to the coo@xmath30 planes as well as perpendicular to these planes ( the crystallographic @xmath31 direction ) .
the corresponding components @xmath32 and @xmath33 are plotted in fig .
[ fig : chi ] .
a short - range antiferromagnetic order has been found in the compounds @xmath34 by neutron measurements@xcite .
this frustrated short - range order is also reflected by the difference between field cooled ( fc ) and zero - field cooled ( zfc ) measurements at low temperatures .
the susceptibility is smaller for zfc than fc below a certain freezing temperature .
sr@xmath1coo@xmath2 for two different directions of the magnetic field .
the insert is an expanded view of the low - temperature region to show the difference between fc and zfc measurements for @xmath35 . the curves with the lower susceptibility refer to the zfc measurements .
] figure [ fig : sus ] shows the inverse magnetic susceptibility for both orientations of field .
the main feature of the susceptibility in the paramagnetic regime is the pronounced magnetic anisotropy , both in magnitude and form of the curves .
the direction of the anisotropy is the same for all crystals , finding @xmath32 to be bigger than @xmath33 .
sr@xmath1coo@xmath2 for two different directions of the magnetic field .
the form of the curves and the magnetic anisotropy strongly deviates from curie - weiss behavior . ] here , the magnetic anisotropy arises from band structure and spin - orbit coupling . for mott and charge - transfer insulators with well - localized moments ,
band - structure and covalency effects can be approximated by an effective crystal field .
the crystal field reflects the anisotropy of the lattice and lifts the degeneracy of the @xmath11 states , resulting in a new set of linear combinations of the unperturbed wave functions as a basis .
the expectation values of the components of the orbital moment result from these new linear combinations .
thus , the crystal s anisotropy may result in an anisotropy of the orbital moment .
spin - orbit coupling ties the spin moment to this anisotropy .
the spin - orbit coupling hamiltonian will be written as @xmath36 where the sum over @xmath37 runs over all electrons .
the dot product between spin and orbital momentum tends to align these moments antiparallel for each electron . for an anisotropic orbital momentum ,
the spin is aligned in the direction of maximum orbital momentum@xcite .
the full many - body ground - state for a @xmath38 or @xmath39 configuration in a crystal - field calculation including spin - orbit coupling and a tetragonal distortions is not simple @xcite but well known . in order to get
an intuitive picture one would like to fall back to a single electron description . in the limit of full spin polarization
this can be done and gives important results . in the following
we will first discuss the magnetic anisotropy of la@xmath0sr@xmath40coo@xmath41 in terms of a one - electron picture .
we will show that each spin state has a different magnetic anisotropy from which , by comparison to the experiment , the spin states of the co ion can be concluded . in order to obtain also a quantitative description and to verify our simple argumentation we will present a full many - body crystal - field calculation afterwards .
within a cubic crystal structure , the @xmath11 states split into @xmath13 orbitals and @xmath12 orbitals .
the basis can be chosen as @xmath42 and @xmath43 , respectively .
a partially filled @xmath12 shell can produce a pseudo orbital moment of @xmath44 .
though the individual real wave functions @xmath45 , @xmath46 and @xmath47 of the basis themselves have a completely quenched orbital moment , their linear combinations are in general complex . writing @xmath48 , @xmath49 and @xmath50 as the orbital wave function with the orbital moment quantized along the axes @xmath51 , @xmath52 and @xmath53 , respectively , one finds @xmath54 in the limit of very large crystal field splittings @xmath14 and with a partially filled @xmath12 shell , the moment is isotropic , despite the large orbital moment .
in fact , the @xmath12 electrons are sometimes compared to @xmath55 electrons@xcite .
introducing a tetragonal distortion , the orbital moment becomes anisotropic .
the tetragonal crystal field splits the cubic @xmath13 states into non - degenerate @xmath56 and @xmath57 levels , while the @xmath12 states split into a non - degenerate @xmath58 state and a two - fold degenerate @xmath59 level . as a basis for these states ,
the real orbital functions can also be used .
the @xmath53 axis of the system is taken to be identical with the @xmath31 axis of the crystal . in the case of la@xmath0sr@xmath1coo@xmath2 ,
the oxygen octahedron is elongated in the @xmath31 direction .
the order of levels and occupations for the ground states of the two cobalt ions is illustrated in fig .
[ fig : occ ] . levels in a tetragonal crystal field arising from an elongated oxygen octahedron .
the two sketches on the left show the occupation for the ls and hs state of co@xmath5 in a one - electron picture .
the right sketch refers to hs co@xmath8 .
the real orbitals on the right can be used as a basis.,width=377 ] the co@xmath5 low - spin state is , except for a small van vleck susceptibility , nonmagnetic since it does not carry any moment . in the high - spin state , co@xmath5 has spin and orbital moment .
the orbital degeneracy is not completely quenched : the degenerate @xmath60 and @xmath61 are occupied by three electrons .
these real orbitals can thus be recombined to form complex orbitals carrying orbital moment in the @xmath53 direction as described in eq .
( [ eq : co3 ] ) .
note that the linear combinations in equations ( [ eq : co21 ] ) and ( [ eq : co22 ] ) can not be formed because the @xmath62 orbital is not degenerate with the @xmath60 and @xmath61 orbitals .
therefore the orbital moment is larger in the @xmath53 direction making this the easy axis .
this anisotropy should be reflected in the susceptibility with @xmath33@xmath63@xmath32 , which obviously contradicts the anisotropy found in the measurements .
a co@xmath5 hs system shows the wrong magnetic anisotropy .
next , we discuss the is state of co@xmath5 . due to the elongation of the oxygen octahedra , the @xmath13 electron occupies the @xmath64 orbital .
this also effects the splitting of the @xmath12 states .
the @xmath60 and @xmath61 orbitals remain degenerate but are not degenerate with the @xmath62 orbital .
this arises from the different charge distributions in relation to the @xmath64 orbital .
we have five electrons to fill in the @xmath12 states , which can also be treated as one hole in the @xmath12 states .
this hole is attracted to the @xmath13 electron in the @xmath64 orbital .
fig .
[ fig : is ] shows the charge distribution of this hole and the @xmath64 orbital . in the left sketch , both
the electron in the @xmath64 orbital and a hole in the @xmath62 orbital is drawn .
the sketch in the center shows the @xmath64 orbital and the hole in a linear combination of the degenerate @xmath60 and @xmath61 orbitals .
regarding the distances between the electron and the hole , on can conclude that the state with the hole in the @xmath60 and @xmath61 orbitals is lower in energy . as the order of levels is reversed when we are speaking about holes instead of electrons ,
this means that the @xmath62 orbital is lowered in energy .
thus , the order and occupation of levels is the one shown in fig .
[ fig : is ] .
the magnetic anisotropy is the same as in the case of co@xmath5 hs which we treated in the last paragraph . with three electrons in the degenerate level of the @xmath60 and @xmath61 orbitals
, we can also make use of eq .
( [ eq : co3 ] ) to show that the direction of anisotropy does not fit the measurements for la@xmath0sr@xmath1coo@xmath2 . neither the hs nor the is state of co@xmath5 show the correct magnetic anisotropy .
but before we draw further conclusions , co@xmath8 should be discussed .
although co@xmath8 has also been found in the ls state in some intramolecular compounds@xcite , for bulk crystals it can be safely assumed that co@xmath8 is in the hs state@xcite .
orbital ( black ) and a hole in the @xmath62 orbital ( white ) .
the charge distribution in the center refers to an electron in the @xmath65 orbital and a hole in a linear combination of the @xmath60 and @xmath61 orbitals .
the splitting and the occupation of the @xmath11 levels for co@xmath5 in the is state is shown on the right.,width=404 ] regarding the ground state produced by the crystal field in fig .
[ fig : occ ] , where the lowest level is filled , no linear combinations like in eq .
( [ eq : co3 ] ) can be formed .
the orbital moment would be completely quenched and the magnetic moment would be determined by an isotropic spin . if , however , spin - orbit coupling is of the same magnitude as the splitting of the @xmath12 orbitals , it will mix these states . mixing in a state with orbital moment in the @xmath53 direction would require placing a hole in the @xmath66 orbital ( see eq .
[ eq : co3 ] ) .
this would cost an energy equal to the splitting within the @xmath12 orbitals .
it is more favorable to mix in a state with orbital moment in the @xmath51 ( or @xmath52 ) direction by placing a hole in the @xmath67 ( or the @xmath68 ) orbital ( see eq .
[ eq : co21 ] ) .
this costs only half of the crystal - field splitting within the @xmath12 orbitals .
an orbital moment in the @xmath51 ( or @xmath52 ) direction costs less energy than in the @xmath53 direction .
so , the @xmath62 plane will be the easy plane for co@xmath8 in this elongated tetragonal structure@xcite .
the spin moment will also be found in the @xmath62 plane . comparing these findings to the measurement ,
we see that co@xmath8 shows the correct direction of magnetic anisotropy . now
a conclusion about the spin state of co@xmath5 can be drawn . for @xmath9 , at least half of the cobalt ions in the crystal are co@xmath5 . taking into account that co@xmath8 has a smaller spin moment than hs co@xmath5 , the latter would dominate the anisotropy .
the opposite is , however , the case in the data .
thus , for this doping range , the spin state of co@xmath5 is identified with the ls state and the magnetization must be assigned to co@xmath8 , which shows the correct anisotropy . for @xmath69
, we still have 40% co@xmath5 and , because of the pronounced anisotropy , we can follow the same argumentation here . regarding @xmath70 in fig .
[ fig : sus ] , the anisotropy is found to be rather small compared to the other compounds .
it should be stressed that the direction of anisotropy in layered perovskites depends on the direction in which the oxygen octahedron is distorted . in some systems like k@xmath30cof@xmath2 the octahedra
are compressed in the @xmath31 direction , resulting in an easy - axis anisotropy@xcite ; co@xmath8 in almost cubic symmetry also tends to have an easy - axis anisotropy like in coo .
however , in our case of la@xmath0sr@xmath1coo@xmath2 , the oxygen octahedron surrounding the cobalt ion is elongated by the inherent tetragonal structure .
this situation is comparable to the change of anisotropy in coo films on different substrates@xcite . for a quantitative analysis of the data and a confirmation of the results found in the qualitative discussion of the anisotropy ,
a full - multiplet calculation was carried out within the crystal field approximation@xcite .
a hamiltonian for the @xmath11 shell of the co@xmath8 ion was set up , including crystal field , spin - orbit coupling and magnetic field .
the slater integrals @xmath71 , @xmath72 and @xmath73 characterizing the coulomb interaction and the spin - orbit coupling constant were taken from a hartree - fock approximation @xcite .
any excitations from the oxygen @xmath74 levels and mixing with higher states were neglected , as these effects are not significant when considering thermal energies .
the parameters for the crystal field were treated semi - empirically .
regarding the tetragonal distortion as a perturbation of the cubic symmetry , the effect of the hybridization of the co @xmath11 shell with the surrounding oxygen @xmath74 shells only increases the splitting of the cubic @xmath12 and @xmath13 states .
hybridization can thus be included into the crystal field parameters and need not be treated separately .
the hamiltonian was diagonalized and eigenstates and eigenvalues were used to obtain the temperature dependence of the magnetization .
magnetic exchange was included in the calculation on a mean - field level .
the results can be seen in fig .
[ fig : fit ] , where the inverse magnetization is plotted .
( black lines ) , plotted inverse .
a mean - field magnetic exchange was added to the calculation .
note that the magnetization of the ion was weighted with the content of co@xmath8 ( @xmath75 ) . ]
the crystal - field parameters were chosen as follows : the @xmath14 splitting and the splitting within the @xmath13 levels were fixed at 1.5ev and 0.2ev , respectively .
these two values are of minor importance for the calculation , because the anisotropy of the magnetization arises from the @xmath12 electrons , as described in the discussion above .
the splitting of the @xmath12 states showed up to be crucial for the form and the anisotropy of the calculated curves .
the values of the splitting within the @xmath12 states were chosen as 80mev , 60mev , 68mev , and 50mev for the compounds
@xmath760.5 , 0.6 , 0.7 , and 0.8 , respectively .
these @xmath12 crystal field splittings for co are in good agreement with previous measurements on strained coo thin films @xcite .
they might appear to be surprisingly small compared to the values found for the titanates @xcite in the order of 200 mev .
the difference between co and ti can be understood by considering the radial extent of the @xmath77 wave function and the transition metal - oxygen ( tm - o ) distance .
the co @xmath77 wave function is much smaller than the ti @xmath77 wave function , reducing the covalency and thereby the size of the crystal - field splitting . at the same time , the average tm - o distances of la@xmath0sr@xmath40coo@xmath41 and latio@xmath78 are almost equal ( @xmath792.02 for co - o @xcite and @xmath792.04 for ti - o @xcite ) .
this is related to the occupation of the anti - bonding @xmath13 orbitals in a hs co@xmath8 compound which pushes the o atoms further away .
it can be seen in fig .
[ fig : fit ] that the calculation gives a good description for @xmath80 and @xmath81 .
the magnetization of the half - doped sample @xmath28 is reproduced well for higher temperatures . at temperatures below the freezing temperature
, the calculation can not describe the system because magnetic order is not included in the model .
the @xmath82 sample is fitted well for lower temperatures . above @xmath83k , the magnetization in the measurement
is enhanced by another effect and results in a deviation from the calculation .
this effect can be seen in all curves in fig .
[ fig : sus ] , but at significantly higher temperatures .
this increase of magnetization could arise from the thermal population of the co@xmath5 hs or is state .
but surely further measurements at higher temperatures are needed to confirm this assumption .
nevertheless , the full - multiplet calculation of a pure co@xmath8 system already succeeds in describing the main features of the magnetization for @xmath84 .
this confirms that co@xmath5 is a ls system in this doping range . in summary ,
the magnetic susceptibility of la@xmath0sr@xmath1coo@xmath2 has been analyzed for two different directions of the external magnetic field .
a definite anisotropy of the magnetic moment is found experimentally with @xmath32@xmath63@xmath33 .
this direction of anisotropy does not match with the behavior of hs or is co@xmath5 , whereas co@xmath8 in the hs state shows the correct single - ion anisotropy .
thus , the spin state of co@xmath5 for @xmath85 must be the ls state .
this conclusion is also confirmed by a full - multiplet calculation .
we acknowledge financial support by the deutsche forschungsgemeinschaft through sfb608 . |
studies of @xmath0 transition amplitudes ( electrocouplings ) represent an important part of the efforts in exploration of strong interaction dynamics in the non - perturbative regime of strong quark - gluon coupling .
the evolution of @xmath0 electrocouplings with photon virtualities q@xmath2 opens up access to the relevant degrees of freedom in the n@xmath1 structure at different distance scales .
it allows us to explore how the strong interaction of quarks and gluons generates different @xmath3 states and how it evolves with distance from the pqcd regime to quark - gluon confinement @xcite . in this contribution
we discuss the current status of the @xmath0 electrocoupling studies from exclusive meson electroproduction off protons and their impact on our understanding of the n@xmath1 structure .
nucleon resonance electroexcitations are described by two transverse ( @xmath4 , @xmath5 ) and one longitudinal ( @xmath6 ) electrocoupling amplitudes @xcite .
these electrocouplings offer an access to the resonance structure .
the clas detector afforded excellent opportunities to study n@xmath1 electroexcitation with great precision and has contributed the lion s share of the world s data on all essential exclusive meson electroproduction channels in the resonance excitation region .
clas provides nearly complete coverage of the final hadron phase space @xcite .
electrocouplings have been obtained from the clas data of @xmath7 , @xmath8 exclusive channels at q@xmath2 up to 5 gev@xmath2 , @xmath9 channel at q@xmath2 @xmath10 4.0 gev@xmath2 , and from @xmath11 electroproduction at q@xmath2 @xmath10 1.5 gev@xmath2 @xcite .
several phenomenological reaction models were developed for evaluation the @xmath0 electrocouplings in independent analyses of n@xmath12 @xcite , n@xmath13 @xcite and @xmath14 electroproduction @xcite .
coupled - channel analyses are also making progress toward the extraction of resonance electrocouplings in global multi - channel data fits @xcite .
the @xmath15 and @xmath11 exclusive channels provided a major part of the information on electrocouplings of nucleon resonances .
the clas data considerably extends information available on @xmath16 , @xmath8 electroproduction off protons .
a total of nearly 120,000 data points on unpolarized differential cross sections , longitudinally polarized beam asymmetries , and longitudinal target and beam - target asymmetries were obtained with the clas detector at w @xmath10 1.7 gev and 0.2 gev@xmath2 @xmath10 q@xmath2 @xmath10 5.0 gev@xmath2 @xcite .
recently , preliminary @xmath7 electroproduction data ( 36,000 data points ) have become available at 1.6 gev @xmath10 w @xmath10 2.0 gev and 1.8 gev@xmath2 @xmath10 q@xmath2 @xmath10 4.0 gev@xmath2 @xcite .
the data were analyzed within the framework of two conceptually different approaches : a ) the unitary isobar model , and b ) a model , employing dispersion relations @xcite .
all well established @xmath3 states in the mass range @xmath17 @xmath10 1.8 gev were incorporated into the @xmath15 channel analyses .
the two approaches provide good description of the @xmath15 data in the range : @xmath18 @xmath10 1.7 gev and @xmath19 @xmath10 5.0 gev@xmath2 , resulting in @xmath20/d.p .
@xmath10 2.0 @xcite .
this good description of a large body of different observables allowed us to obtain reliable information on the @xmath21 , @xmath22 , and @xmath23 resonance electrocouplings .
the @xmath24 electroproduction data from clas @xcite provide for the first time information on nine independent one - fold - differential and fully - integrated cross sections in each bin of @xmath18 and @xmath19 in a mass range @xmath18 @xmath10 2.0 gev , and at photon virtualities of 0.25 gev@xmath2 @xmath10 @xmath19 @xmath10 1.5 gev@xmath2 .
the analysis of these data allowed us to establish all essential mechanisms contributing to @xmath11 electroproduction : @xmath25 , @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath30 meson - baryon channels and direct production of the @xmath11 final state without formation of intermediate unstable hadrons .
the data - driven reaction model jm has been developed @xcite with the goal of extracting resonance electrocouplings and the @xmath31 and @xmath32 hadronic decay widths . the contributions from well established n@xmath1 states in the mass range up to 2.0 gev were included into the amplitudes of @xmath31 and @xmath32 meson - baryon channels .
resonant contributions were treated employing unitarized version of the breit - wigner ansatz @xcite .
the jm model provides a reasonable description of @xmath24 differential cross sections at @xmath18 @xmath10 1.8 gev and @xmath19 @xmath10 1.5 gev@xmath2 .
the successful description of @xmath11 electroproduction cross sections allowed us to isolate the resonant contributions and to determine both resonance electrocouplings and @xmath33 and @xmath32 decay widths fitting them to the measured observables .
@xmath34 magnetic form factor in ash convention @xcite normalized to the dipole form 3@xmath35 .
experimental data are taken from the review @xcite .
the dashed and solid curves correspond to quark core and combined quark core and meson baryon cloud contributions @xcite .
right : dseqcd results on quark core contribution to the p @xmath36 @xmath34 transition magnetic form factor obtained with contact qq - interaction ( solid line ) and accounting for a quark anomalous magnetic moment ( long - dashed blue line ) in comparison with a coupled channel approach @xcite inferred from experimental data ( short - dashed green line ) @xcite .
the dseqcd expectations for realistic qq - interaction is shown as dotted red line .
experimental values of the dressed magnetic transition form factor are shown as points .
[ p33],title="fig:",width=170 ] @xmath34 magnetic form factor in ash convention @xcite normalized to the dipole form 3@xmath35 .
experimental data are taken from the review @xcite .
the dashed and solid curves correspond to quark core and combined quark core and meson baryon cloud contributions @xcite .
right : dseqcd results on quark core contribution to the p @xmath36 @xmath34 transition magnetic form factor obtained with contact qq - interaction ( solid line ) and accounting for a quark anomalous magnetic moment ( long - dashed blue line ) in comparison with a coupled channel approach @xcite inferred from experimental data ( short - dashed green line ) @xcite . the dseqcd expectations for realistic qq - interaction is shown as dotted red line .
experimental values of the dressed magnetic transition form factor are shown as points .
[ p33],title="fig:",width=245 ] the p @xmath36 @xmath34 magnetic transition form factor in ash convention @xcite determined from analyses of exclusive @xmath15 electroproduction data @xcite and normalized to the dipole form is shown in fig .
[ p33 ] ( left ) .
physics analyses of these results revealed the structure of @xmath37 state as combined contribution from external meson - baryon cloud and internal core of three dressed quarks in the ground state bound in the flavor decuplet .
the models accounting for only quark degrees of freedom allowed one to describe the data at q@xmath2 @xmath38 3.0 gev@xmath2 but underestimate considerably the experimental values at q@xmath2 @xmath10 1.0 gev@xmath2 .
a good description of the data in the full area of photon virtualities shown in fig . [ p33 ] ( left ) has been achieved accounting for the contribution from both external meson - baryon cloud derived from experimental @xmath15 data within the coupled - channel approach @xcite and the inner quark core .
first evaluations of bare quark core contributions to the p @xmath36 @xmath34 magnetic transition form factor have recently become available from dyson - schwinger equations of qcd ( dseqcd ) @xcite .
the results @xcite are shown in fig .
[ p33 ] ( right ) in comparison with the quark core contribution employing the coupled - channel approach @xcite inferred from the experimental data .
consistent results of both approaches on the quark - core contribution demonstrate promising potential of dseqcd in describing the @xmath37 resonance quark content starting from the qcd lagrangian .
dseqcd models allow us to explore how n@xmath1 masses are generated non - perturbatively , relating resonance electrocouplings to dressed quark mass function @xcite .
( left ) and @xmath39 ( right ) @xmath21 electrocouplings obtained from analyses of n@xmath12 electroproduction @xcite ( red circles ) and @xmath11 exclusive channel ( black triangles ) @xcite including recent preliminary results in the area of q@xmath2 from 0.5 to 1.5 gev@xmath2 ( blue squares ) .
the results at the photon point are taken from @xcite .
quark core contributions to @xmath40 electrocoupling estimated within the framework of the light - front quark models @xcite are shown in the left part as solid and dashed lines , respectively . the quark core contribution to @xmath39 electrocoupling obtained within the framework of a quark model @xcite
is shown in the right part by solid line .
absolute values of the meson - baryon cloud contributions inferred from the experimental data in a coupled - channel approach @xcite are shown by thick solid lines in magenta .
[ p11d13],title="fig : " ] ( left ) and @xmath39 ( right ) @xmath21 electrocouplings obtained from analyses of n@xmath12 electroproduction @xcite ( red circles ) and @xmath11 exclusive channel ( black triangles ) @xcite including recent preliminary results in the area of q@xmath2 from 0.5 to 1.5 gev@xmath2 ( blue squares ) . the results at the photon point are taken from @xcite .
quark core contributions to @xmath40 electrocoupling estimated within the framework of the light - front quark models @xcite are shown in the left part as solid and dashed lines , respectively . the quark core contribution to @xmath39 electrocoupling obtained within the framework of a quark model @xcite
is shown in the right part by solid line .
absolute values of the meson - baryon cloud contributions inferred from the experimental data in a coupled - channel approach @xcite are shown by thick solid lines in magenta .
[ p11d13],title="fig : " ] for the first time electrocouplings of @xmath40 and @xmath39 resonances were obtained in independent analyses of n@xmath12 ( 0.2 gev@xmath2 @xmath10 q@xmath2 @xmath10 5.0 gev@xmath2 ) @xcite and @xmath11 ( 0.25 @xmath10 q@xmath2 @xmath10 0.6 gev@xmath2 ) @xcite exclusive pion electroproduction off protons .
recently these electrocouplings have been determined from the clas @xmath11 electroproduction data @xcite in a range of q@xmath2 from 0.5 to 1.5 gev@xmath2 .
the published @xcite and preliminary results are shown in fig .
[ p11d13 ] .
the @xmath40 and @xmath39 electrocouplings determined from @xmath15 and @xmath11 channels are consistent .
consistent results on electrocouplings of prominent n@xmath1-states determined in independent analyses of major meson electroproduction channels with different backgrounds strongly suggest the reliable extraction of these fundamental quantities .
masses , total decay widths of @xmath40 and @xmath39 resonances and their branching fractions to the @xmath31 final states determined in our analysis @xcite are consistent with the rpp results @xcite and coupled - channel analysis by bonn - gatchina group @xcite .
analyses of the @xmath40 and @xmath39 electrocouplings within the framework of quark models @xcite and a coupled - channel approach @xcite revealed a combined contribution to the structure of these states from both an external meson - baryon cloud and an inner core of three dressed quarks in the first radial excitation and an orbital excitation of l=1 for the @xmath40 and @xmath39 states , respectively .
the studies of these electrocouplings offer further insight to resonance structure allowing us to explore how the bound systems of three dressed quarks with different quantum numbers are generated and how the meson - baryon cloud depends on the isospin and spin - parity of the states .
a successful description of quark core contribution to @xmath40 electrocouplings starting from qcd lagrangian was achieved in the exploratory dseqcd approach @xcite , which well reproduces the estimates for quark core contributions obtained from the results on @xmath40 electrocouplings employing the coupled - channel approach @xcite .
electrocouplings of @xmath41 resonance .
the maid results from analysis of n@xmath12 electroproduction data are shown as triangles @xcite , while squares stand for preliminary clas results from the @xmath11 electroproduction data @xcite .
preliminary clas results on @xmath21 and @xmath23 electrocouplings of @xmath42 from the @xmath11 electroproduction data @xcite are shown in the middle and right panels , respectively .
the evaluation of these electrocouplings within the framework of a quark model @xcite and a bethe - salpeter approach @xcite are shown as solid and dashed lines .
[ s31f15],title="fig:",width=3 ] electrocouplings of @xmath41 resonance .
the maid results from analysis of n@xmath12 electroproduction data are shown as triangles @xcite , while squares stand for preliminary clas results from the @xmath11 electroproduction data @xcite .
preliminary clas results on @xmath21 and @xmath23 electrocouplings of @xmath42 from the @xmath11 electroproduction data @xcite are shown in the middle and right panels , respectively . the evaluation of these electrocouplings within the framework of a quark model @xcite and a bethe - salpeter approach @xcite
are shown as solid and dashed lines .
[ s31f15],title="fig:",width=3 ] electrocouplings of @xmath41 resonance .
the maid results from analysis of n@xmath12 electroproduction data are shown as triangles @xcite , while squares stand for preliminary clas results from the @xmath11 electroproduction data @xcite .
preliminary clas results on @xmath21 and @xmath23 electrocouplings of @xmath42 from the @xmath11 electroproduction data @xcite are shown in the middle and right panels , respectively .
the evaluation of these electrocouplings within the framework of a quark model @xcite and a bethe - salpeter approach @xcite are shown as solid and dashed lines .
[ s31f15],title="fig:",width=3 ] preliminary results on electrocouplings of @xmath42 , @xmath43 , @xmath44 , @xmath45 , and @xmath46 states were obtained from the clas @xmath11 electroproduction data @xcite .
most of the excited states with masses above 1.6 gev decay preferentially to the final n@xmath47 states , making the @xmath11 electroproduction data the primary source of information on electrocouplings of high - mass resonances . examples of available high - mass state electrocouplings are shown in fig .
[ s31f15 ] .
analysis of @xmath11 channel confirmed the @xmath44 electrocouplings determined previously from n@xmath12 electroproduction data @xcite and provided accurate results on the transverse electrocouplings and the first information on the longitudinal electrocouplings of all the aforementioned high - lying excited proton states .
analyses of these electrocouplings within the framework of quark models @xcite demonstrated that the transition amplitudes to n@xmath1 and @xmath48 states of different spin - parity and isospin further extend our knowledge on the n@xmath1 structure .
the models @xcite have limited success in describing resonance electrocouplings .
none of them is able to reproduce the electrocouplings of these resonances .
electrocouplings of a large number of excited nucleon states in the mass range up to 1.8 gev have become available from analyses of exclusive meson electroproduction off protons .
consistent results on resonance electrocouplings from independent analyses of major exclusive channels with different non - resonant contributions strongly suggest the reliable extraction of these fundamental quantities from the data .
the considerable extension of the information on resonance electrocouplings provides new opportunities to explore , how n@xmath1 states of different quantum numbers are generated by strong interaction in the non - perturbative regime . | the transition @xmath0 amplitudes ( electrocouplings ) for prominent excited nucleon states obtained in a wide area of photon virtualities offer valuable information for the exploration of the n@xmath1 structure at different distances and allow us to access the complex dynamics of non - perturbative strong interaction .
the current status in the studies of @xmath0 electrocouplings from the data on exclusive meson electroproduction off protons measured with the clas detector at jefferson lab is presented .
the impact of these results on exploration of the n@xmath1 structure is discussed . |
continued evolution in technology has made computing devices an integral part of one s life .
the most common manifestation of this is the ` mobile phone ' .
modern technology has transformed the mobile phone from a mere communication device to a versatile computing device .
these hand - held devices have enabled us not only to access information , but also to provide information to others on the move .
modern mobile phones , equipped with powerful sensors , have endowed capabilities to provide and create near real - time information .
this real - time information is useful for oneself and for others .
an established approach for sharing and provision of information and creating useful applications in a distributed environment is service oriented architecture ( soa ) @xcite . realizing soa over mobile devices has the potential to convert mobile phones owned by common people from mere _ information subscribers _ to _ information providers _ and beyond .
the major advantage of this is that it can be used in scenarios where there is little or no preexisting infrastructure .
examples of such scenarios include war - front , post - disaster relief management . in such scenarios ,
mobile based soa has the potential to enable ground teams to provide runtime information to commanding units , help teams at disaster sites to exchange data , analyse damage and examine various statistics using mobile devices . in such systems of soa over mobile devices
all three elements of the soa triangle : service providers , service consumers , and service registries are realised over mobile devices .
web services are the proven way towards implementation of a `` service oriented architecture '' .
advancement in mobile device technology has motivated researchers to explore the possibilities of effectively hosting web services over mobile devices , and thereby trying to realize service oriented systems in mobile environments
. there has been substantial work towards enabling mobile devices to host web services @xcite@xcite@xcite .
an important aspect of service oriented systems , `` service discovery '' , however , remains a challenge in mobile environments .
there is literature available on service discovery for distributed environments @xcite@xcite@xcite , but one catering specifically to mobile environments is still lacking .
several challenges specific to hosting web services over mobile devices need to be taken into account in such service discovery mechanisms .
these include , but are not limited to battery and network constraints , limited computational power of mobile devices .
moreover , such dynamic mobile services are prone to uncertainty ( owing to network outage , battery issues , physical damage ) and frequent changes in functionality ( primarily owing to the change of context ) , and hence make frequent service updates a necessity to effectively function as web - services .
the role of the _ service registry _ therefore , becomes one of prominence to properly manage such dynamism .
traditional service registry solutions for web - services such as uddi @xcite , ebxml @xcite , can not be directly utilised in such environments that require frequent updates .
what contributes to this is the exhaustive data model of such registry offerings that is hard to analyze and parse for mobile devices at run - time . to the best of our knowledge ,
the current work is the first attempt to comprehensively investigate these issues and design a dynamic service registry that facilitates service discoveries in mobile environments . as mentioned earlier
, the ultimate aim is to realize a service oriented architecture over mobile devices without involving high end servers .
hence , the proposed architecture provides all registry related information and operations using mobile devices itself , without requiring high - end computers or high management costs .
further , in order to support scalability , fault tolerance , and fault localization , we propose a distributed and category based service registry . to demonstrate the feasibility of the approach
, we have engineered a prototype deployment .
this includes heterogeneous and loosely coupled mobile devices deployed in a collaborative manner to manage the service registry along with native hosted services .
we also compare the proposed approach with the traditional uddi system for managing service registry from the perspective of mobile devices .
the evaluation shows propitious results in favour of our approach wherein the latter is shown to have acceptable battery requirements , low data communication costs , promising scalability , and little or no hindrance to the working of native applications of mobile devices . this work is a significant extension of our previous work @xcite . in our previous work
@xcite , we had introduced an xmpp based model to maintain a service registry for mobile environments .
the main focus of the work was to introduce the registry architecture and the communication mechanism followed during service discovery from the registry . in the presented work
, we provide a holistic service registry framework that makes use of xmpp based service registry framework at the core .
we defined the roles for mobile devices involved in the service registry framework to provide a scalable mobile registry solution .
we have further extended the service registry operations to cater the specific needs of the mobile environment .
we further provide detailed descriptions of the various registry operations that facilitate the realisation of a dynamic mobile service registry .
we further evaluated the proposed approach by realizing it through a working prototype and deployed it over mobile devices of volunteers .
we further present a detailed literature survey that covers various categories of service registries .
the rest of the paper is organized as follows : section [ sec : motive ] presents a motivation and requirements for the novel approach .
section [ sec : arch ] provides details of the proposed approach and design concepts .
various registry operations are discussed in section [ subsec : operation ] .
prototype implementation details and inline comparison with uddi are presented in section [ sec : implementation ] .
section [ sec : evaluation ] includes notes on the experimental evaluation of the approach .
this is followed by section [ sec : related ] that presents a survey of related work .
finally , section [ sec : conclude ] concludes the paper with a brief discussion on future possibilities .
kotler et al .
@xcite suggested services as `` activities or benefits offered for sale that are essentially intangible and do not result in the ownership of anything '' .
a mobile service defined in this work is a service that is offered from mobile phones of providers ; this may also include information provided by the mobile sensors , third party software , or human users .
this allows different machines to exchange information with each other over a network , without necessarily requiring a user interface . in general ,
the service may be a component or sub - part of the web application that is usually used by human users .
for example , a chatting web application provides gui to human users to communicate with another human .
while a presence service embedded in the web application detects the presence of other machines , this presence service does not require any human intervention .
_ alice is a high risk cardiovascular patient .
recently , she got an ecg sensor implanted in her body @xcite that monitors her cardiovascular health and provides statistics and information as a mobile service via her mobile phone .
this service can be consumed by her cardiologist and she can be provided with proper prescriptions as per her current health .
one day she had a sudden cardiac arrest on her way to another city .
alarming variations in her ecg signals were observed by the service on her mobile device and the service discovered the nearest ambulance through the latter s exposed mobile service .
further , her mobile service automatically provided access to her latest ecg signals to the ambulance support medical staff and enabled them to prepare well in advance for the patient .
the ambulance was able to discover her current location through another service on her mobile device that provided gps coordinates .
further , when the ambulance was on its way , the ambulance s mobile service provided the doctors at the nearest hospital with the latest information on the situation . simultaneously , the hospital was able to make use of alice s ecg mobile service to gather her ecg history and prior to her arrival the doctors at the hospital had a chance to study her medical profile and case in detail . on its way to the hospital
, the ambulance was able to make use of the services exposed by other travelers on their respective mobile devices to avoid the busy route and opt for the path with less traffic .
meanwhile , the insurance company was contacted by alice s mobile service and her hospital information was provided , so that the financial aspect could be taken care of even before her arrival .
alice s cardiologist was also able to provide details of his / her prescriptions via his / her mobile service to the doctors in the hospital so that the latter could learn about her medications and allergies if any . _ with rapid advancements in mobile technologies and wireless networking ,
mobile devices have become perhaps the most suitable and economical solutions for the provision of dynamic , transient , contextual , personalized services .
these mobile provisioned services can make the service access handy and convenient for service consumers .
further , the provisioning of mobile services is an economical solution that requires little or no pre - existing infrastructure . in the discussed scenario , alice , her cardiologist , the ambulance
, hospital staff can make use of each other s mobile services in critical situations and can provide assistance to alice .
this explains the importance of mobile services ; subsequently , however , the above scenario also raises a question : `` how is the mobile service consumer able to discover the appropriate service among such large number of services and that too in an uncertain environment as a mobile environment ? '' .
web services hosted on mobile devices are mainly useful for sharing contextual , personal , proximal information . mobile devices in such environments are mostly distributed arbitrarily and make service discovery and management of service registry a cumbersome process . in this section ,
we discuss the need for a novel service registry architecture for mobile environments . in order to provide an effective service registry for mobile environments , two approaches are possible .
the first is a classical centralized service registry approach where all information on the available mobile services is maintained at one place and this is usually over a powerful computing device ; the second approach is a decentralized service registry approach . here
, the registries are maintained by a system of distributed nodes in such a way that each node caters to a fraction of the services and there is a large degree of redundancy . between these two approaches ,
the decentralised service registry approach appears to be more appropriate for mobile environments .
there are several reasons for this such as the issue of a single - point - of - failure in the case of a centralised system , the lack of a definite guarantee of continuous reliable connections between mobile devices and the central server , the difficulties of rapid and regular updates in large centralised registries thus giving rise to obsolete information and so on .
there are , of course , drawbacks in the decentralised system as well .
it is these drawbacks that we will discuss and attempt to overcome in the rest of this paper .
cloud offloading is another approach that is often used for facilitating services over mobile devices . in cloud offloading , the service logic ,
usually , resides on the cloud and the mobile devices may work as the proxy for these services .
the associated concerns of cloud offloading @xcite ( significant network delay and latency , rigid sla requirements etc .
) however do not make it a potential candidate for dynamic mobile service registry .
sanaei et.al .
@xcite discuss these challenges in detail .
a decentralised service registry may be realised using either traditional service registry approaches that are commonly used in legacy wired systems ( such as uddi @xcite , ebxml @xcite ) or a new approach especially catering to the vagaries of mobile environments may be adopted .
though possible , adopting traditional registry approaches such as uddi @xcite from w3c is ill suited to dynamic mobile environments .
the traditional registry architecture comprises uddi data entities ( businessentity , businessservice , bindingtemplate , tmodel , publisherassertion , subscription ) , various uddi services and api sets , uddi nodes for supporting node api set , uddi registries .
such a base architecture is quite ` heavyweight ' and makes it difficult to host uddi over mobile and resource constrained devices .
further , services offered over mobile devices tend to behave in an anarchic manner ; as the changes in the functional , non - functional , other aspect of the services may be quite frequent owing to regular change in context and networking environment of the device .
this requires frequent updates to the service registry .
uddi , on the other hand , is designed around concepts of soap / wsdl , heavyweight technologies that make frequent updates a cumbersome process.as a consequence , the information on the uddi registry quickly becomes obsolete . a new approach ,
therefore , is imperative for maintaining an effective registry system for mobile environments . before we get into detailed discussions on the proposed approach , here is a quick point - wise summary of the requirements for effective service registries for mobile environments .
this is along the lines of dustdar et al .
@xcite who did something similar for articulating general requirements for web service registries .
_ r1 : management of transient web services : _ the very nature of mobile devices makes hosted web services repeatedly and randomly enter and leave the network .
a service registry should be such that it supports such dynamic and frequent arrival and departure of service providers .
_ r2 : lightweight : _ a service registry designed for mobile environments should be lightweight . a lightweight service registry would complement the power ( i.e. battery ) and computational constraints of mobile devices . furthermore , a lightweight service registry is agile and is easier to integrate with diversified mobile environments .
_ r3 : minimum communication overhead : _ given the battery and network constraints in mobile devices , emphasis should be towards a registry system with minimum communication overhead .
_ r4 : distributed service registry : _ as the number of mobile devices ( and therefore potential web services over these ) are increasing exponentially , a centralized service registry system has limited utility and gets outdated very quickly . hence , a distributed service registry system is required to support scalability .
_ r5 : enabling run time search : _ an important enabler of mobile based service oriented systems is support for run time search .
this is necessary owing to the frequent arrival of new and often more competent services and/or failure of existing services .
conforming to the above points could potentially ensure a service registry suitable for mobile environments .
in mobile based soa environments , each mobile device can perform the functionalities of both a service provider and a service consumer . as a service consumer
, a mobile device discovers web - services and invokes them after negotiating with the providers . as a service provider , a mobile device hosts services and publishes hosted services with service registries .
however , as stated earlier , an effective mobile registry to publish and discover such mobile services in dynamic environments is still lacking .
the mobile services are provided by mobile devices and may be consumed by another mobile device in a peer to peer manner .
our approach suggests a service registry system that comprises a light - weight registry server at each participating mobile device .
the registry at each mobile device contains minimal information that is _ just _ sufficient to uniquely identify the registered entity .
a registry server ( at each mobile device ) manages either of two types of registries : 1 .
service registry : registered mobile services are managed in the service registry . the service registry contains an entry for each service as : _ service name , service access point , service i d , service description , service groups , availability , service location , service provider , other service information_. 2 . group registry : registered services are categorized into service groups . these service groups are managed in a group registry . the group registry contains the following information for a service group : _ group name , group domain , group description , registrant , groupid , group access point , other group information_. the organization of registries , as proposed , is shown in figure [ fig : regent ] .
the service information that is just enough to identify a registered service and that is less likely to change is kept in the registries and service information that is more likely to change but does not affect the discovery process of the service is kept in the vicinity of the provider .
the service binding description , and contextual descriptions are provider specific and are likely to change in the mobile environment .
therefore , these descriptions of the service are kept in close vicinity of the mobile service provider .
close vicinity here implies that the description is hosted on the same mobile device as the service or a third party repository , where these descriptions can be updated rapidly .
we define several registry related operations in section [ subsec : operation ] that are performed using xml streams .
these xml streams are inspired by xmpp @xcite ( extensible messaging and presence protocol ) , a well known and established communication protocol .
xmpp is already in wide use in mobile environments in several instant messaging applications .
the proposed approach provides service registry operations that facilitate effective discovery of a service .
details like the non - functional descriptions and quality of service values of the services have deliberately been kept out of the proposed system to make it as lightweight as possible and hence suitable for mobile environments .
there are four primitive steps in the proposed mobile service registry approach that are presented in figure [ fig : approach ] : \1 ) _ mobile service registry access point is retrieved : _ as shown in figure [ fig : approach].a , a registry requester ( represented by a mobile icon with * m * ) accesses the public registry to retrieve the access point details of mobile service registry .
2 ) _ mobile service registry is accessed via navigator nodes : _ as shown in figure [ fig : approach].b , the navigator nodes ( represented by mobile icons with * n * ) are contacted for the `` group registry '' .
this group registry contains the list of service groups .
service group of the required service provider is discovered in the group registry .
3 ) _ service group is contacted via registry nodes : _ as shown in figure [ fig : approach].c , the registry nodes ( represented by mobile icon with * r * ) are accessed via groupid for retrieving the service provider s information .
`` service registry '' is traversed for the required service provider .
4 ) _ the service provider is contacted : _ as shown in the figure 2.d , finally , the required service provider is discovered and it is contacted for service negotiation and service binding . these steps are discussed in detail in the next subsection .
we first start with the roles performed by the mobile devices .
a mobile device potentially performs the following roles ( as shown in figure [ fig : approach].a ) : navigator node and registry node .
navigator nodes are the entry points for the mobile registry architecture as shown in figure [ fig : approach ] .
these navigator nodes are accessible by service consumers via public access points .
we have devised the mobile registry architecture as a service itself .
the mobile registry and its public access points can be registered with any global public registries just to make them globally discoverable ( as shown in figure [ fig : approach].a ) .
the motive to use global public registry is to provide the access point details of the mobile registry architecture ; mobile devices would need to use the global public registry just once to retrieve the access point details of the architecture .
( these global public registries could be any existing service registries as discussed in section [ sec : related ] .
these registries are assumed to be well in place , hence is not discussed in detail .
the global public registry is not suitable for mobile services , the reasons are already discussed in section [ subsec : need ] . ) there can be multiple navigator nodes connected to the public access points via a common access channel , as shown in figure [ fig : approach].b .
the common access channel can be viewed as a communication bus that enables various mobile devices to communicate .
the common access channel gives mobile devices the liberty to join and leave the network at any time without disturbing other navigator nodes .
whenever a new mobile device joins as a navigator node , the group registry is updated / downloaded via the access channel and the shared domain ontology becomes accessible .
the idea of a common access channel can be realized using existing networking technologies as suggested in rfc1112 and rfc5771 .
navigator nodes are the mobile devices that manage the group registry ( refer figure [ fig : regent ] ) .
as discussed in the earlier section , the group registry manages service groups .
these service groups are uniquely identified by group identifiers or groupid .
the group registry comprises the list of service groups present in the network along with the respective group identifiers and other group details .
the navigator nodes are further responsible for categorizing the registered services into the various service groups and to navigate the service providers to the assigned service groups .
navigator nodes rely on existing ontological approaches to categorize services on the basis of their domains of offering ( or offered service type ) .
all navigator nodes have access to the shared domain ontologies @xcite for categorization of the offered services . whenever a service provider needs to register its
offered service , the group registry is referred first for matching the service with its service group . in case the service offered by the provider does not belong to any of the existing service groups , a new service group is created by referring the domain ontology and updating the group registry with a new group entry .
we have used an existing classification method for classifying services into groups @xcite .
the motivation behind adopting this method is that the method does not require a training set for classification and dynamic changes to the classification parameters is possible without having to retrain the classifier .
this is of particular importance in mobile environments as it provides run time updates to the domain ontology without disturbing the classifiers .
the service group classification method follows generic steps .
first , the mapping criteria are parsed from user supplied service description which is in the textual format . then
, the domain ontology is mapped to the mapping criteria .
the process calculates a matching categorization score of the service description with the ontological context as defined by allahyari et al .
@xcite .
for example , the mobile service providers hosting services for : a. ) doctor s rating and b. ) hospital building floor map , share the same service group _ hospital _ , hence they are identified by the same _
groupid_. however a new mobile service provider offering contact information of pizza outlets would fall in a separate service group .
we do not dwell upon the classification approach in this paper .
the interested reader is referred to @xcite for more details on this .
registry nodes are the mobile devices that manage the service registries ( refer figure [ fig : regent ] ) .
as discussed in the previous section , a service registry manages the registered services that are uniquely identified by service identifiers .
this enables a service provider to provide multiple services over a single mobile device .
the service registry comprises the list of registered services in a service group , their availability information along with the service details that are just sufficient to manage and identify the registered services .
of these details , the real time availability information of the registered services is what mainly contributes to overcoming the uncertainty of mobile environments .
this availability information managed at the registry node gives the much needed reliability to the services hosted on mobile devices .
registry nodes are responsible for managing the up - to - date service registry , responding to service registry related queries , and performing registry related operations .
the service group can be seen as an overlay group of registered mobile service providers and registry nodes that are identified by the groupid .
we have devised this group identifier as a multi - cast address for the service group members .
the requests sent to the service group are received by all the member mobile service providers , however , only the group member acting as the registry node responds to the requests ( as shown in figure [ fig : approach].c ) .
to improve query response time , a replica of the service registry that is retrieved from the registry node is managed at all the mobile devices hosting services in the service group .
selective updates are performed to keep the local replica updated . during service discovery ,
the local replica is first referred to , in case discovery fails at the local replica then the registry node is contacted . maintaining replicas of this kind
do add a little overhead to the architecture but in the larger context the local replicas reduce traffic meant for discovery over the network substantially .
further , the local replicas ease the transition of service providers into full fledged registry nodes in the eventuality of a registry node failure ( details on this in the subsequent subsection ) .
mobile devices acting as navigator nodes or registry nodes can also depict uncertain behavior and are prone to failure .
hence , in the case of existing registry node failure , a new registry node can be elected from the member mobile service providers , without compromising on the consistency of other service groups or navigators .
( the same approach is also applied to navigator nodes . )
heartbeat operations are used to detect registry node failure .
any mobile service provider can become a registry node by participating in an election and declaring its candidature .
our method of registry node election is inspired by the leader election problem of distributed computing .
the algorithm is discussed in more detail in section [ subsec : mobops ] .
_ potential service consumers _ use the proposed mobile registry architecture for discovering their desired services .
this service discovery is a three step process . in the first step as shown in figure [ fig : approach].b
, the service consumer sends the discovery request to the mobile registry architecture via a well known access point or uri ( uniform resource identifier ) for the desired service .
this request for service discovery is first handled by the navigator node .
the navigator node searches its group registry and provides the matching service group details along with the groupid for the requested service .
groupid _ acts as a multi - casting address for all the service providers that belong to the group . in the second step as shown in figure [ fig : approach].c , the service consumer contacts the service group using the groupid of the required service .
the registry node of the service group responds with the available matching services and their corresponding service details . in the third step
as shown in figure [ fig : approach].d , service consumer contacts the mobile service provider offering the required service to retrieve the technical description and performs the service negotiation for service access . _ mobile service providers _ use the proposed mobile registry architecture for registering their offered services .
the service registration process primarily comprises two steps .
first , the mobile service provider sends a service registration request to the mobile registry architecture via the well known access points .
the navigator nodes first handle the registration request and respond with the matching service group along with the groupid and other group details registered in the group registry .
second , the mobile service provider contacts the service group and registers its service with the service registry of the registry node . alternatively ,
if there is no matching service group in the group registry , the navigator node creates a new service group . in this case
, the registrant mobile service provider becomes the registry node of the newly formed service group .
in this section , we describe the operations and functionalities that provide registry operations such as registration , discovery , service updates , and service binding in the proposed architecture .
an inline comparison with uddi is also presented .
the following registry operations make use of several components discussed in an earlier work of ours @xcite . in the proposed approach
, we have two types of registrations : 1 ) group registration ( at navigator node ) 2 ) service registration ( at the registry node ) .
these registrations are shown in the figure [ fig : registerproc ] . _
group registration : _ the service group is registered in the group registry at the navigator node . a mobile service provider ( registry client )
contacts the mobile registry framework via the navigator node to register the service .
however , if a matching service group is not yet registered in the group registry , a new service group registration request is initiated .
the mobile service provider sends a group registration request to the navigator node along with the service details .
hereafter , the navigator refers the domain ontology and based on the service details , a matching group is mapped .
this matched service group is updated in the group registry along with its groupid .
the registry is then shared among all the navigator nodes .
subsequent to the successful registration process , a `` groupid '' is sent to the newly registered mobile device ( in a ` result ' type iq stanza ) . at this point
the registrant is the registry node of the newly formed service group .
figure [ fig : registerproc ] shows the group registration process .
_ service registration : _ services are registered in the service registry at the registry node . the mobile service provider fetches the matching service group at the navigator node and contacts the registry node of the matching service group via the groupid .
hereafter , the registry node receives service related information from the mobile service provider and generates a serviceid for the new service .
the registration process is completed when the information of the new service is updated at the service registry of the registry node .
this updated registry is made available to the service providers in the service group for provider initiated updates ( pull based updates ) . upon successful registration
a `` serviceid '' is sent to the newly registered mobile service provider ( in a ` result ' type iq stanza ) .
figure [ fig : registerproc ] shows the service registration process .
+ _ web service registration in uddi : _ here
we quickly discuss the registration process in the traditional uddi registry system so that one can appreciate the significance of the proposed approach . in uddi
, registration is done using the _ publisher apis set _ exposed by the uddi , such as save_service , save_business , save_binding , save_t_model .
these apis are used to save detailed information on the web service , which may not be necessary in case of the mobile based web services .
moreover this information would tend to become heavy for mobile devices to process or transport .
a typical uddi registry @xcite primarily consists of the following information : businessentity , businessservice , bindingtemplate and tmodels for a registered service .
the information is passed on by the web - service provider to the uddi registry through publisher apis ( the information is transported as xml tags , we are not showing the xml for the uddi structure owing to space constraints here . for the benefit of interested readers
, we have uploaded details on this at : http://goo.gl/cn8vap ) .
deploying such a uddi registry over a mobile device would tend to become heavy owing to the limited computational power and network constraints in mobile devices .
the uddi registry would also significantly lag behind in managing the dynamic nature of mobile devices .
two possible cases are included in the prototype : 1 .
discovery initiated by the registered service provider , 2 .
discovery initiated by an external mobile service consumer . _ discovery by registered service provider : _ the registered service provider first matches the service group of the required service with the local replica of the service registry . if the group matches then it fetches the required service locally . in case
the service is found locally , a selective update is performed from the registry node to obtain information on the latest availability status of the service . in case
the service is not found locally , the query is propagated to the registry node and subsequently the local replica is updated with the latest service registry status .
_ discovery by external service consumer : _ the service consumer first contacts the navigator node and retrieves the matching service group information .
subsequent to this the service discovery is forwarded to the matching service group .
afterwards , the registry node of the service group responds with the matching service information .
the registered service provider from the other service group also follows this process . figure [ fig : registerdisc ] shows the service discovery process . _
service discovery in uddi : _
web service discovery in a uddi registry is done via public inquiry apis of the uddi , such as find_service , find_binding , find_business , find_tmodel .
the service discovery is performed centrally by the uddi registry server , which requires high computational capability .
this is because the consumer requests the uddi registry server which in turn does the query search centrally and responds to the consumer with the results .
the complexity and structured nature of the uddi data structure would makes searching tedious were it adopted in a mobile environment .
though traditional uddis enable consumers to query the registry and are effective in a centralized system , they are ill suited to the mobile environments that are mostly distributed . web service binding information is necessary to call a particular web service .
it includes the technical information on a web service , such as the access endpoint , required parameter values , return type etc . in the proposed architecture , binding information is exchanged directly between the service consumer and the service provider ( as shown in the figure [ fig : approach].d ) .
the service provider can provide the wsdl / wadl document or it s global url in the binding information as well .
the functional description of the service is kept in the close vicinity of the service provider .
the reason being that the mobile services tend to change frequently and this might result in a change in the technical description of the same .
therefore , a proximal location of the functional description facilitates mobile service providers to readily change the service operations dynamically without violating the service registry information ( figure [ fig : servicebind ] shows the service binding process ) .
furthermore , other types of descriptions viz .
non - functional , contextual , business descriptions are usually present on the same device as the service provider to keep descriptions up - to - date without increasing traffic over the mobile registry architecture .
.summary of operations performed in proposed approach and uddi . [ cols="<,<,<",options="header " , ] in our third experiment , we sent ( 50 * 4= ) 200 new service registration requests from other mobile devices and virtual instances to the registry architecture .
figure [ fig : total ] depicts the total response time behavior for a service provider .
also noteworthy here is the fact that the mobile devices were continuously moving with the volunteers , hence the devices were randomly joining and leaving the network .
therefore , we observed a few outliers in the response time behavior .
we concluded that the average service registration time ( including outliers ) is near 5 seconds which seems acceptable for practical purposes .
the fourth experiment evaluated the effect of directory size on the discovery time .
we registered multiple services in a service group on the registry node . in order to test the scalability of our prototype , we ran service discovery operations for various directory sizes : 500 , 1000 , 5000 , 10000 , 50000 , 100000 .
we discovered that the discovery time increases as the size of the directory increases .
however , in spite of this even with 100000 registered services the query response time was under 1 second which is acceptable for all practical purposes .
we sent service discovery requests from four mobile devices and virtual instances to varied numbers of registered services and the response time was calculated for these requests .
figure [ fig : scale1 ] shows the average response time for these discovery requests for various registry sizes . through the whole experiment , the mobile device acting as the registry node was moving continuously within the network .
further , the size of the registry increased linearly with the increasing number of registered services .
even with thousands of services registered , the registry size was under 10 mb ( shown in figure [ fig : scale2 ] ) .
it should be noted here that we made use of sqlite for implementing the service and group registries over android mobile devices .
our fifth experiment aimed at evaluating the feasibility of the proposed approach when running simultaneously with other phone activities . in this experiment
, we conducted a comparative study on the difference in response time for the service discovery requests a ) when a volunteer was answering a phone call and , b ) when the device was idle .
for this experiment , we sent 50 requests for group discovery to the volunteer s mobile device .
first , we observed the response behavior when the device was idle in the volunteer s pocket . during this phase the volunteer was randomly moving within the network
the response time behavior during this period is shown in figure [ fig : wocall ] .
next , we made a phone call on the volunteer s mobile phone and observed the response time during the call .
the results demonstrate that there is a change in response time , but this change is well within acceptable limits .
response time behavior during the call is shown in figure [ fig : call ] .
there is an initial peak in the response time when the call is made . from the initial results ,
we conclude that the mobile device takes a little time to respond to the first discovery request .
this is due to the fact that some processing time is required to _ awaken _ the sleeping mobile application .
we feel , therefore , that _ piggybacking of incoming requests _ could be a good approach to reduce the energy overhead .
the sixth experiment involved an evaluation of the reliability of the approach .
we toggled the availability status of the service provider from _ available _ to _ unavailable _ and back to _ available _ with a time difference of 10 seconds .
these toggles were repeated 120 times at the registry node . during this time
, we continuously probed the registry node from the service consumers to get the status of the service provider .
the initial results shown by the experiment have less than 1% false negative , where false negative implies - registry node returns availability information as _ available _ when the service provider has updated it to _ unavailable _ and vice versa .
although the scale of these experiments was limited , the results were promising
. it will be interesting to explore the performance of this approach for a much larger number of service providers , and registry clients and for much longer durations ( a few days ) .
( interested readers may refer to https://goo.gl/4vg895 for further details about the architecture and experimental setup . )
nonetheless , we present the speculated trend of our experiments with a large number of devices : 1 .
effects on battery : with increasing number of devices there would be more numbers of registry requests and responses ; hence more battery power would be needed . however , in such scenarios the service groups and navigator nodes would play a crucial role .
more specialized service groups would be formed as suggested in section [ subsec : concept ] , this would keep the number of devices in a service group within limits in turn keeping the battery usage in check ( irrespective of any number of registered services ) .
effects on data exchange : with increasing number of devices there would be more exchange of data .
however , the split service groups would keep a check on the data exchange .
further , keeping in view the current data usage by modern smartphones , the data usage by large number of devices should be in acceptable limits .
effects on service registration time : we believe that with increasing number of devices and service registrations , the response time would increase but would be well within the accepted range . also , this is a onetime process for a service
, it would be acceptable for all practical purposes .
effects on service discovery time : figure [ fig : scale1 ] shows the discovery time for a large number of registered services varying from one thousand to a few hundreds thousand .
this is still within acceptable limits .
5 . effects on registry size : the registry size will increase linearly for increasing number of register services .
trend is shown in the figure [ fig : scale2 ] . while other experiments are dependent on the individual mobile device without depending on the other device s behavior .
further , the effects of other contextual parameters on increasing number of devices ( e.g. network , isp , terrain , carrier type etc . ) would be interesting to look into but unfortunately these are not within the scope of current focus .
the proposed architecture caters to the issues of providing a dynamic service registry in mobile environments .
it manage to incorporate the requirements outlined in section [ sec : req ] .
_ r1 : management of transient web services : _ the approach effectively manages the availability information on each registered web service and is capable of dynamically updating it .
this helps in satisfying requirement _ r1 _ and keeping the service registry up - to - date irrespective of random entry and exit of transient web services .
_ r2 : lightweight : _ the architecture seeks information that is just enough to uniquely identify and manage a web service from the registrant mobile device for registry related operations .
the registries manage information that is less likely to change and whatever change does happen is updatable via a watchdog process .
this keeps the registry architecture as lightweight as possible and thus satisfies requirement _
r2_. _ r3 : minimum communication overhead : _ we have made use of just three xml stanzas in order to minimize communication overhead , satisfying requirement _
r3_. during our experiments , we observed the exchange of just a few kilobytes of data for hundreds of request transfers .
this small overhead also helped us minimize battery utilization , as reflected in table [ table : power ] and [ table : network ] .
furthermore , the approach uses just one stanza `` iq stanza '' for easier service registration and de - registration .
_ r4 : distributed service registry : _ the proposed approach manages the service registry over dispersed registry nodes and navigator nodes . thus satisfying requirement _ r4 _ and improving fault localization . these features ,
contribute to a light weight and autonomous service registry effective for mobile environments .
_ r5 : run time search : _ the dynamic availability information , minimal data transfer , and faster response time helped in performing effective run time searches and in the process satisfied requirement _
r5_. further , the proposed dynamic mobile service registry is compatible with existing uddi and other registries .
the dynamic registry can register external uddi and other service registries as one of the registered services along with their access points .
contrary the proposed registry can be registered with other registries and uddi as any of the registered service .
there are certain important assumptions that have been made in the proposed service registry framework .
these are listed as follows : 1 ) the load balancing of the incoming registry requests is assumed to be handled at the device level
. there could be a threshold capacity limit , at each device depending on its hardware capacity . on exceeding this capacity limit ,
the incoming requests are not entertained and these surplus requests are handled by other registry nodes that have access to the common access channel .
2 ) there is no reward system suggested in the current state of the work .
hence , it is assumed that the mobile device owners are self - motivated to provide their respective devices for serving as the registry .
we may look into a system of rewards / incentive as part of future work .
3 ) the updates are performed to keep the local registry replica updated ( at the navigator / registry nodes ) in an automated manner without manual intervention
. a few important limitations of the current work are listed as follows : a ) the current design of the framework does not deal with the privacy issues of the mobile phone users associated with service registry and service provisioning .
b ) the current design of the mobile registry does not handle the qos ( quality of service ) aspect of the mobile web service .
the qos may be handled by providing a link to the data server ( external to the framework ) that could have qos and other details on the service description .
c ) though we have a system in place to detect the unavailability of registry nodes through heartbeat signals , the behavior of the mobile device owner , poor network connectivity , and physical damage to the mobile device could result in the abrupt unavailability of the registry / navigator node .
these have not been dealt with in this work .
d ) finally , the current evaluation of the approach was conducted within a supervised lab environment .
the registry power and data requirements , service discovery performance , and other results were therefore well within acceptable range
. there could be slight variations in these if the experiments were carried out at a much larger scale including thousands of mobile devices .
service discovery is an important aspect of service - oriented architecture .
two types of approaches are primarily adopted for service discovery : registry based approach and registry - less approach .
the registry - less approach usually makes use of overlay networks , hash tables , and other broadcasting / multi - casting techniques . in the proposed work ,
we perform service discovery using the former i.e. `` registry based approach '' .
we surveyed existing literature from the late 90 s .
we classified the registry based approach into two broad categories : the centralized registry approach and the distributed registry approach .
these two can further be classified into those for mobile environments and those for non - mobile environment .
our survey includes works from the areas of soa , peer - to - peer networking , mobile ad - hoc networking .
the centralized service registry approach is used in several popular technologies : service location protocols @xcite , sun s jini architecture @xcite , service discovery services @xcite , microsoft s universal plug and play ( upnp ) @xcite .
these service discovery infrastructures rely on a central registry for discovering capable services .
the service information is stored at a centralized registry .
all registry related operations are performed by a single entity .
one well known example of centralized service registry architecture for web services is uddi ( universal description discovery and integration ) @xcite .
uddi is not the service registry itself .
however , uddi is the specification of a framework for describing web - services , registering web - service , and discovering web - service .
several data structures and apis have been published for describing , registering and querying web - services through the uddi .
the ebxml ( electronic business xml ) @xcite standard is another example of a centralized web service registry architecture .
hoschek @xcite presented a grid based hyper registry for web services in peer - to - peer networks .
the registry is an xquery based centralized database that manages dynamic distributed contents .
juric et al.@xcite proposed an extension to the uddi for incorporating version support for services .
they presented modifications to the category tag of business service and tmodel of the uddi infoset with the intent to introduce service interface versions in uddi and wsdl .
bernstein and vij @xcite proposed the use of xmpp for intercloud topology , security , authentication and service invocations . in some ways
this work is similar to ours .
however , our work focuses on registry management in mobile based service oriented architecture .
we focus on managing the service registry in a distributed manner over resource constrained mobile devices .
seto et al .
@xcite proposed a service registry for ubiquitous networks to dynamically discover service resources .
their registry divides the service operation into the source , transformation , and sink , specifies physical meta - data to manage devices , and associate a keyword with it .
feng et al .
@xcite proposed a registry framework to include interoperability among various semantic web service models .
for this , they made use of registry meta model for interoperability-7 and several mapping rules for handling semantic mismatch between services .
feng et al .
@xcite proposed a service evolution registry where providers can register their service evolution information and consumers can be sent an alert regarding the service evolution .
their work manages service versions along with their dependencies and discrepancies .
more recently @xcite presents the idea of using object relational databases in the service registry .
the registry extends the search to include search on the basis of service commitments and service expectations .
some existing work @xcite@xcite discusses the possibility of centralised service registries for mobile environments .
diehl et al .
@xcite talk about centralized service registries that store the service domain , service types , location and access rights to manage service mobility and adoption of services in wireless networks .
beck et al .
@xcite propose an adaptable service framework for mobile devices that relies on a central service registry for dynamic service registration and discovery .
doulkeridis et al .
@xcite discuss the idea of managing contextual information in service registries .
the main focus of the work is to ease service discovery in mobile environments by maintaining context aware service registries .
deepa and swamynathan @xcite talk about a directory based architecture that make use of two integrated architectures : backbone - based and cluster based .
their work was intended to facilitate service discovery in mobile ad - hoc networks and to achieve improved network traffic , response time , and hit ratio .
chen et al .
@xcite and sivashanmugam et al .
+ @xcite discuss a few initial approaches to maintain uddi in a federated environment .
one approach @xcite supports qos based discovery from requesters and provides an aggregated result from the federated registries
. the other approach @xcite suggests the use of various metadata and ontologies to manage uddi in a federated environment .
verma et al.@xcite propose meteor - s wsdi , that focusses on providing registries in distributed and federated environments .
extended registries ontology was used to provide access to these distributed registries and organizing them in domain based categorization .
the approach discussed in @xcite presents a distributed service registry for grid application .
the approach utilises xpath queries and ontological trees for domain based service discovery .
baresi and miraz @xcite talk about an approach to enable heterogeneous federated registries to exchange service information .
the approach is based on the publish and subscribe model .
ad - uddi @xcite is a distributed registry architecture that adopts an active monitoring mechanism .
the approach extends uddi to incorporate automated service updates in a federated registry environment .
treiber and dustdar @xcite propose an active web service registry that make use of atom news formats .
rss software is used in the approach to form an active distributed registry .
shah et al .
@xcite also propose an rss - based distributed service registry in the move to achieve global soa .
the proposed registry is intended to provide dynamic discovery using rss and tries to resolve synchronization issues in rss .
jaiswal et al .
@xcite introduce a decentralized registry using the chord protocol for peer - to - peer environments .
the registry comprises distributed hash tables of web - service names and web - service ips .
their method claims to cater to demand driven web - service provisioning .
another direction in distributed service registry systems is one meant for cloud environments @xcite and @xcite .
lin et al .
@xcite present a hadoop - based service registry for the cloud environment .
the work proposes geographical knowledge service registries that are designed to simplify service registration , improve discovery and other registry operations for cloud services .
elgazzar et al .
@xcite propose to manage a local service registry at the provider s site for the offered services .
this local service registry is managed in a distributed manner and has two types of services : local and remote .
the paper proposes discovery - as - a - service in the cloud environment .
das gupta et al .
@xcite in a more recent paper , discuss about the possibility of a federated registry system for p2p networks .
the work makes use of multi - agent based distributed service discovery for non - deterministic and dynamic environments .
they propose that super peer nodes manage the distributed service registry and other peers register their services with these registries .
zhang et al .
@xcite discuss the integration of peer to peer technology with soa .
they talk about self - organizing , semi - structured p2p frameworks for support and propose to use a private service registry at each peer for discovering the manufacturing services .
the local private registry of the peer is traversed first to discover a service .
handorean and roman @xcite propose probably the first work that discusses the possibility of distributed service registries in mobile environments . in their approach ,
the availability of services is shown in the registry along with an atomic update facility to maintain consistency .
konark @xcite is a distributed xml based registry that has a tree structure .
a top - down approach is used for the tree , with generic classification of services at the top and specific classification at the bottom .
every node maintains a service registry , where it stores information about its own services and also about services that other nodes provide .
the approach provides a semantic service registry and enables servers as well as clients to register and discover services in a distributed manner .
schmidt and parashar @xcite present a distributed hash table based approach for distributed registries in peer to peer networks .
the approach supports service discovery based on keywords , wild cards on an internet scale .
indexing of keywords associated with the web service description document is managed at the peers .
tyan and mahmoud @xcite discuss an approach for service discovery in mobile ad - hoc environments .
the work considers a registry as a tree and makes this registry available to every node in the network .
the approach makes use of a location aware routing protocol and divides the network into hexagonal grids with a gateway for each that have the service registry .
golzadeh and niamanesh @xcite discuss an approach for a service registry system for mobile ad - hoc networks .
the approach divides the manet into clusters with one head for each cluster that acts as a directory for the cluster .
the head node has two types of registries : the service provider s registry and the other head node registry .
a decentralized service registry for location based web services is discussed in @xcite .
the approach is relied on cellular network system and base transceiver station for retrieving the local registry address .
these addresses are broadcasted by base station and interested mobile devices download the registry address for location based services .
one of the latest works by jo et al .
@xcite makes use of bloom filter to manage distributed service registry for mobile iot environments .
proposed work uses hierarchical bloom filters for reducing message exchanges among registries in move to find available services .
although there is a good amount of work that has been done towards developing effective service registries , an architecture that enables mobile devices to host mobile service registries and that takes the distinct features of mobile environment into account such as intermittent connectivity , dynamic nature , frequent service description changes is still lacking .
our work focuses on registry management in mobile based service oriented architecture .
we focus on managing the service registry in a distributed manner on resource constrained mobile devices .
this service registry contains minimal information about the registered services in a manner that is just enough to uniquely discover the services .
the proposed work is an extension of some of our previous work @xcite .
in this paper we looked into , perhaps , the most challenging aspect of implementing an soa in mobile environments : effective service registries .
our studies show that traditional approaches for implementing service registries ( such as uddi ) can not be directly adopted in mobile environments , given the dynamic , volatile and uncertain nature of such environments . a novel approach to manage service registries ` solely ' over mobile devices was proposed that effectively addressess issues specific to mobile environments and enables run time service discoveries .
we evaluated the approach by developing a prototype and deploying it over real mobile devices . to emulate real world usage as closely as possible we requested volunteers to deploy our prototype on their personal mobile devices and continue doing their routine tasks .
the experimental results indicate that the proposed solution is an effective enabler for soa in mobile environments .
we performed several experiments to confirm the efficacy of the prototype across several parameters such as : timely performance , battery consumption , effect of the random / nomadic behaviour of people carrying mobile devices , conflict with native mobile apps , reliability .
future work in this direction would be towards mobile service registries that focus on qos factors unique to mobile environments .
future work will also tackle security related issues , dynamic service group splitting in mobile service registries .
further , the availability of the service registries will be a prime focus .
we would like to thank tanveer ahmed and dheeraj rane for their valuable insights.the work was supported by the ministry of human resource development - government of india .
54 natexlab#1#1[1]`#1 ` [ 2]#2 [ 1]#1 [ 1]http://dx.doi.org/#1 [ ] [ 1]pmid:#1 [ ] [ 2]#2 , , & ( ) . . in _ _ ( pp . ) . .
, , , & ( ) . .
( ed . ) . .
, & ( ) . . in _ _ ( pp . ) . . , & ( ) .
. in _ _ ( pp . ) . volume of _ _ . . , , & ( ) .
in _ _ mobide 99 ( pp . ) . : . , &
( ) . . in _ _ ( pp . ) . .
, , , , & ( ) . . , _ _ , .
, , , & ( ) . . in _ _ . , , , , & ( ) .
. in _ _ ( pp . ) . . . , , , & ( ) . . ,
_ , . . , & ( ) .
. in _ _ ( pp . ) . .
, , , , , & ( ) . . in _ _ ( pp . ) . . , , & ( ) .
. in _ _ ( pp . ) . volume of _ _ . , & ( ) . . ,
, , & ( ) . . in , & ( eds . ) , _ _ ( pp . ) . volume of _ _ . . , & ( ) . . , _ _ , . .
, , & ( ) . . , _
, , & ( ) . . in _ _ ( pp . ) . .
, , , , , , & ( ) . . in _ _ ( pp . ) . . , & ( ) .
. in _ _ ( pp . ) . .
( ) . . , _
, , , & ( ) . . in _ _ ( pp . ) . . , & ( ) .
. in _ _ coordination 02 ( pp . ) . : .
, , , & ( ) . . in _ _ ( pp . ) . volume . .
( ) . . in _ _ ( pp . ) . .
, , , & ( ) . . in _ _ ( pp . ) . . , , & ( ) .
. in _ _ ( pp . ) . volume of _ _ . . , , , & ( ) . . ,
, , , & ( ) . . . , & ( ) . . in _ _ ( pp . ) . volume of _ _ . .
, , , & ( ) . . in _ _ ( pp . ) . .
( ) . . , _
[ arxiv : http://dx.doi.org/10.1137/0215074 ] .
, , , & ( ) . . , _
, , , , & ( ) . . , _
_ , . . , , , , , , , , , & ( )
( ) . . in _ _ ( pp . ) . . , , , & ( ) . . , _ _ ,
. internet engineering task force ( ietf ) proposed standard . ( ) . .
internet engineering task force ( ietf ) proposed standard . , , , &
( ) . . , _
& ( ) . . , _ _ , . .
, , & ( ) . . in _ _ iiwas 11 ( pp . ) .
, , , & ( ) . . in _
_ ( pp . ) .
, , & ( ) . . in _ _ ( pp . ) . .
, , & ( ) . . in _ _ ( pp . ) . .
, , , & ( ) . . in _ _ ( pp . ) . .
. , & ( ) . . ,
_ , . . , & ( )
_ , . . , , , , , & ( ) . . , _
_ , . . , & ( ) .
. in _ _ . .
( ) . . , _
, , , , , , & ( ) .
. in _ _ codes / isss 10 ( pp . ) .
http://doi.acm.org/10.1145/1878961.1878982 . .
, , , & ( ) . . , _ | advancements in technology have transformed mobile devices from being mere communication widgets to versatile computing devices .
proliferation of these hand held devices has made them a common means to access and process digital information .
most web based applications are today available in a form that can conveniently be accessed over mobile devices
. however , web - services ( applications meant for consumption by other applications rather than humans ) are not as commonly provided / consumed over mobile devices . facilitating this and in effect realizing a service - oriented system over mobile devices has the potential to further enhance the potential of mobile devices .
one of the major challenges in this integration is the lack of an efficient service registry system that caters to issues associated with the dynamic and volatile mobile environments .
existing service registry technologies designed for traditional systems fall short of accommodating such issues . in this paper , we propose a novel approach to manage service registry systems provided ` solely ' over mobile devices , and thus realising an soa without the need for high - end computing systems .
the approach manages a dynamic service registry system in the form of light weight and distributed registries .
we assess the feasibility of our approach by engineering and deploying a working prototype of the proposed registry system over actual mobile devices .
a comparative study of the proposed approach and the traditional uddi ( universal description , discovery , and integration ) registry is also included .
the evaluation of our framework has shown propitious results in terms of battery cost , scalability , hindrance with native applications .
service - oriented systems ; mobile computing ; mobile web services ; web service registry ; web service discovery . |
the hot gas in elliptical galaxies , groups , and clusters of galaxies is thought to be enriched with metals by material ejected by supernovae .
hence , the abundances in the hot gas record the star formation history of these systems and therefore probe the rates of type ia and type ii supernovae as well as the shape of the imf ( for reviews see renzini 1997 , 1999 ) .
since the hot gas in these systems emits at temperatures @xmath7-@xmath8 @xmath9 ( @xmath10 - 10 kev ) , x - ray observations are the preferred means to determine the abundances in the hot gas .
spatial analyses of the x - ray surface brightness profiles of groups of galaxies using _ rosat _ @xcite indicate that the hot gas typically extends out to distances in excess of 100 kpc ( e.g. david et al .
1994 ; pildis , bregman , & evrard 1995 ; davis et al . 1996
; ponman et al . 1996 ; trinchieri , fabbiano , & kim 1997 ) and also suggest that the gas is divided into distinct components corresponding to the halo of the central galaxy and an extended component for the surrounding group ( mulchaey & zabludoff 1998 ; also see ikebe et al .
these studies typically obtain average temperatures of @xmath11 - 2 kev from analysis of the _ rosat _ pspc spectra and also find evidence for positive temperature gradients in some groups ( trinchieri et al .
1997 ; mulchaey & zabludoff 1998 ) . despite the evidence for temperature gradients , in most cases the abundances reported by the _
studies are obtained from isothermal spectral models from which very sub - solar fe abundances are inferred ( a notable exception being ngc 533 trinchieri et al .
1997 ) .
similarly , analyses of the _ asca _ data @xcite of groups which assume isothermal gas typically find very sub - solar fe abundances and , in some cases , si / fe ratios in excess of solar ( fukazawa et al .
1996 , 1998 ; davis , mulchaey , & mushotzky 1999 ) .
although the point spread function ( psf ) of _ asca _ is quite large and asymmetric , there have been several attempts to deconvolve the _ asca _ data to allow analysis of the spectra as a function of radius with the results for galaxy clusters being decidedly mixed ( e.g. markevitch et al .
1998 ; kikuchi et al .
recently , finoguenov et al . and finoguenov & ponman have applied their own method of deconvolution to the _ asca _ data of a small number of groups . by assuming the hot gas to be in a single phase
, these authors have fitted isothermal models at several radii for each group .
similar to the previous studies , finoguenov et al . and
finoguenov & ponman obtain very sub - solar fe abundances and si / fe ratios in excess of solar .
the very sub - solar fe abundances for the hot gas in groups of galaxies deduced by these studies are in many cases less than those of galaxy clusters and thus pose a problem for the popular models of chemical enrichment which assume ratios of type ia to type ii supernova and an imf similar to those of the milky way .
( e.g. renzini 1997 ) . to reconcile these observations with these models
it has to be postulated that galaxy groups accrete a significant amount of primordial gas after they have spent most of their evolution expelling gas .
the unattractive features of this scenario are discussed by renzini .
these results for groups are in every way analogous to those initially obtained for ellipticals with _ rosat _
( e.g. davis & white 1996 ) and _ asca _ ( e.g. matsumoto et al . 1997 ) .
recently , however , buote & fabian ( 1998 , hereafter bf ) and buote ( 1999 , hereafter b99 ) have demonstrated that models having at least two temperature components in the hot gas are required to explain the _ asca _ spectra of the brightest ellipticals .
these multiphase models agree with the positive temperature gradients indicated by the _ rosat _
data unlike the isothermal models ( b99 ) .
perhaps most importantly bf and b99 also showed that the fe abundances determined from the multiphase spectral models of the brightest ellipticals are consistent with solar in stark contrast to the very sub - solar values obtained from isothermal models .
this strong dependence of the fe abundance on the spectral model indicates that one can not obtain a reasonable estimate of the fe abundance by using an isothermal model if the hot gas consists of multiple temperature components .
similar problems also arise when determining the @xmath5/fe ratios ( b99 ) .
since the spectral properties of the brightest ellipticals are so similar to groups , it should be expected that similar results for the fe abundance also apply to groups of galaxies . to investigate this issue we selected for analysis the groups from the samples of mulchaey & zabludoff and davis et al .
having the highest quality _ asca _ data suitable for detailed spectral analysis .
in addition , we included the group ngc 1132 since mulchaey & zabludoff argue that it has properties similar to bright groups despite the fact that it has no large galaxies aside from ngc 1132 .
our final sample consists of 12 galaxies whose properties are listed in table [ tab.prop ] .
the paper is organized as follows . in section [ obs ]
we present the observations and discuss the data reduction .
we define the models used for broad - band spectral fitting in section [ models ] and discuss calibration and fitting issues in section [ issues ] .
the results for the broad - band spectral fitting analysis are presented in section [ results ] .
we discuss the implications of our results in section [ disc ] and present our conclusions in section [ conc ] .
finally , in appendix [ bias ] we use simulated _ asca _ observations to examine in detail the `` fe bias '' and `` si bias '' associated with the spectral fitting of ellipticals , groups , and clusters of galaxies .
[ cols="<,^,^,^,^,^,^,^,^,^ " , ] best - fitting 1 t model parameters resulting from fits to simulated _ asca _ data of the models in table [ tab.models ] .
all elements not listed have abundances tied to fe in their solar ratios
. see text for discussion of the _ asca _ simulations .
the x - ray studies of the hot gas in elliptical galaxies ngc 720 by buote & canizares and ngc 4636 by trinchieri et al .
were the first to show that the _ rosat _ pspc spectra of ellipticals could not distinguish between isothermal models with very sub - solar fe abundances and two - temperature models with solar fe abundances .
using simulated pspc data buote & canizares further demonstrated that if the galaxy spectrum is intrinsically characterized by two temperatures in the hot gas with solar fe abundances but is instead fitted with an isothermal model then sub - solar fe abundances necessarily result ; i.e. the isothermal model gives a strongly biased measurement of the fe abundance kev for the hot gas and @xmath12 kev for a bremsstrahlung component .
although this shows the degeneracies in metallicity depending on the spectral model of a faint ( low @xmath13 ) elliptical , this does not demonstrate a bias resulting from fitting a spectrum with two temperatures in the hot gas ( i.e. with @xmath14 - 2 kev ) apart from any bremsstrahlung component . ] .
the strong dependence of the fe abundance on spectral model has now been demonstrated using _ asca _
data for elliptical galaxies by bf and b99 and for groups of galaxies in this paper .
these papers also show that multiphase models are clearly favored over isothermal models for these systems . as a result
, investigators hoping to obtain average fe abundances for these objects by fitting isothermal models in actuality always infer strongly biased ( i.e. very sub - solar ) fe abundances for ellipticals ( e.g. matsumoto et al .
1997 ) and for groups ( e.g davis et al .
since this `` fe bias '' is crucial for interpreting the x - ray spectra of these systems , in this appendix we attempt to elucidate this bias by using simulations of simple multiphase models that are consistent with the current _ asca _ observations of ellipticals , groups , and clusters .
we investigate the dependence of the fe bias on temperature by considering objects spanning the temperatures of the brightest ellipticals , groups , and clusters of galaxies .
in addition , we use _ asca _ simulations of these systems to show the dependence of the @xmath5/fe ratios on the spectral model .
thus , this appendix updates and extends the brief studies of the fe bias that appear in section 2.2 of buote & canizares and in section 3.2.2 of bf .
the key properties of the four reference models we use for this study are listed in table [ tab.models ] .
each of these models corresponds to regions of positive temperature gradient ( e.g. see section 5 of b99 ) .
models 1 and 2 are based on the two - temperature and cooling flow models of the elliptical galaxy ngc 1399 within @xmath15 kpc ( b99 ) which are also quite similar to the corresponding models of several of the lower - temperature groups in this paper ( see sections [ 2 t ] and [ cf ] ) . to explore the high - temperature groups we base model 3 on the cooling flow models of mkw 9 and ngc 4104 ( section [ cf ] ) .
finally , we obtained the _ asca
_ data of a478 from the heasarc archive and have analyzed the spectra within a @xmath16 radius km s@xmath17 mpc@xmath17 , and @xmath18 . ] in a similar manner to the ellipticals and groups ( see section [ obs ] ) : the result is listed as model 4 and is consistent with previous studies ( e.g. johnstone et al .
a478 was selected because ( 1 ) it is one of the brightest clusters having evidence for substantial excess absorption and ( 2 ) with respect to other cooling flow clusters a comparatively large fraction of the emission measure arises from the cooling flow component ; i.e. a478 should be one of the most favorable clusters for examining systematics when fitting an isothermal model .
we simulated _ asca _ observations of the models in table [ tab.models ] using xspec . because of its better energy resolution and sensitivity near 1 kev we focus on the sis data .
for consistency of presentation we use the same pi grouping for each simulated sis spectrum ( i.e. that of the sis data of ngc 1399 see b99 ) .
our choices of exposure times for the simulations are motivated by a desire to insure ( 1 ) sufficiently high s / n for each case so that noise does not noticeably affect the fitted parameter values , ( 2 ) that the sizes of the residuals ( in terms of @xmath19 ) are similar to the best available _ asca _ data , and ( 3 ) that the simulated data for each model have similar s / n . although background is included in the simulations , the s / n criteria stated above insure that its effect is only noticeable at the highest energies which are not of interest to our study .
in figure [ fig.sim.n1399 ] we plot the simulated _ asca _ sis data of models 1 and 2 from table [ tab.models ] and also show the best - fitting isothermal model and its residuals .
these 1 t models with the @xmath5/fe ratios fixed at solar give best - fitting values of @xmath20 for the simulated 2 t model and @xmath21 for the simulated cf+1 t model .
( these metallicities do not differ significantly from the fe abundances obtained when the @xmath5/fe ratios are allowed to vary in the fits see table [ tab.1tvar ] . )
notice that both the pattern of residuals and the metallicities derived from the 1 t fits are very similar for both models 1 and 2 and thus the fe bias is essentially the same for isothermal fits to two - temperature and cooling flow spectra .
this fe bias is insensitive to the presence of excess absorption on the colder component .
if instead we consider a model identical to model 1 but with @xmath22 @xmath23 , then the metallicity obtained from fitting the 1 t model to the simulated _ asca _ data has a best - fitting value @xmath24 which is consistent with the value obtained from the simulated model 1 within the @xmath25 errors .
below 1 kev the pattern of residuals for this model differs from model 1 such that the residual peak below 1 kev is slightly less pronounced than in figure [ fig.sim.n1399 ] .
the pattern of residuals of model 3 ( figure [ fig.sim.hight ] ) is less complex than for the lower temperature models 1 and 2 .
the best - fitting metallicity using the 1 t model ( @xmath5/fe fixed at solar ) is @xmath26 which represents only a 40% under - estimate of the true value in contrast to the @xmath27 under - estimate observed for models 1 and 2 .
the fe bias decreases more dramatically as we move to still higher temperatures with model 4 ( figure [ fig.sim.hight ] ) . for this model
we obtain a best - fitting value of @xmath28 using the 1 t model ( @xmath5/fe fixed at solar ) which is only a @xmath29 under - estimate of the true value .
if we allow the @xmath5/fe ratios to vary ( table [ tab.1tvar ] ) the best - fitting fe abundance rises to @xmath30 ; i.e. the fe bias for hot clusters is insignificant .
a physical explanation of the fe bias is as follows . for models 1 and 2 which have average temperatures @xmath10 kev
, the colder temperature components preferentially excite emission lines in the fe l complex below 1 kev ( from ions fe 17 - 21 ) while the hotter temperature components preferentially excite the fe l lines from @xmath10 - 1.4 kev ( fe 21 - 24 ) . since these components contribute approximately equally to the emission measure
, the net result is that the spectral shape in the highly temperature - sensitive fe l energy region is flattened with respect to a 1 t model .
that is , the spectrum of a 1 t model which has the average temperature of the two components of models 1 or 2 ( and has the same fe abundance ) will preferentially excite the fe l lines right at 1 kev at the expense of the other fe l lines near @xmath31 kev and @xmath32 kev and thus will be more peaked at 1 kev than either model 1 or 2 . to partially compensate for this , i.e. to flatten the spectral shape near 1 kev , the 1 t model decreases its fe abundance so that the contribution of the flatter bremsstrahlung continuum is increased in the model .
however , the distinctive residual pattern in figure [ fig.sim.n1399 ] demonstrates that the fe l lines are sufficiently temperature sensitive ( and the sis has sufficient energy resolution ) to thwart this attempt to successfully fit the 1 t model to the simulated two - temperature and cooling flow spectra .
the strong temperature - sensitive fe l emission lines are the key to this fe bias . as one moves to higher temperatures ( models 3 and 4 ) the fe l lines become weaker and
the bias decreases rapidly for average temperatures above 2 kev .
moreover , at higher temperatures the strong 6.5 kev fe k emission line complex becomes increasingly more important and dominates the determination of the fe abundance . a final note : since it is the variation of the spectral shape of the fe l complex with temperature which gives rise to the fe bias , one does not obtain a biased measurement of the fe abundance using a 1 t model if the spectrum consists of multiple spectral components which have the same temperature but different fe abundances . in this case
the 1 t model yields a value for the fe abundance that is a weighted average over the various components .
prominent residuals in the 1 t fits ( @xmath5/fe fixed at solar ) are located near the strong emission lines of o , ne , mg , si , and s ( figures [ fig.sim.n1399 ] and [ fig.sim.hight ] ) . if the abundances of these elements are allowed to vary separately from fe we obtain the results listed in table [ tab.1tvar ] .
although in most instances the o , ne , and ( to a lesser extent ) the mg abundances differ significantly from fe , their new values do not correspond to a substantial reduction of the residuals over @xmath33 - 1.5 kev .
the 1 t fits for models 1 and 2 produce zero o and ne abundances and mg / fe ratios less than solar .
these results are very sensitive to the presence of excess absorption on the colder temperature component .
if instead we consider a model identical to model 1 but with @xmath34 @xmath23 , then we obtain @xmath35 , @xmath36 , and @xmath37 with @xmath38 . for models 3 and 4
the 1 t fits yield o , ne , and mg abundances very different from models 1 and 2 and from each other .
although the 1 t fit for model 3 gives zero o abundance its large ne abundance is qualitatively different from models 1 and 2
. the ratios of o / fe , ne / fe , and mg / fe obtained for model 4 are all very much in excess of solar .
hence , the o , ne , and mg abundances are very sensitive to the temperature model and to the absorption model , and thus the systematic errors introduced when using a 1 t model to measure these abundances are more complex in origin than the fe bias discussed above . in contrast , the k@xmath5 emission lines of si and s are at sufficiently high energies so that they are not affected by the absorption model .
the residuals near the si ( @xmath39 - 2.0 kev ) and s ( @xmath40 - 2.7 kev ) lines are readily apparent for all the models in figures [ fig.sim.n1399 ] and [ fig.sim.hight ] , especially for model 4 .
the 1 t fits with variable @xmath5/fe ratios listed in table [ tab.1tvar ] for models 1 - 3 indicate modest ratios of @xmath41 for @xmath42 and more substantial ratios of @xmath43 - 1.5 solar for @xmath44 .
model 4 has the largest ratios with @xmath45 and @xmath46 .
thus , the inferred values of si / fe tend to increase with temperature while the s / fe ratio is relatively constant .
these excess si / fe and s / fe ratios inferred by 1 t models arise from two effects : ( 1 ) the colder temperature components present in the two - component models produce stronger si and s emission , and ( 2 ) the 1 t model under - estimates the continuum at energies above @xmath47 kev . issue # 1 is more important for the higher temperature systems like clusters ( model 4 ) while issue # 2 dominates for the lower temperature systems like elliptical galaxies ( models 1 - 2 ) .
therefore , attempts to infer the si / fe and s / fe ratios using isothermal models on spectra that are intrinsically composed of multiple temperature components will necessarily yield values for these ratios that are biased in excess of the true values , the effect of which is most important for higher temperature clusters of galaxies .
this bias may have played a role in the excess si / fe ratios obtained by mushotzky et al . and
may in part explain the trend discovered by fukazawa et al .
that the si / fe ratio rises with temperature for clusters galaxies . although fukazawa et al . attempt to eliminate any problems associated with multiphase gas by excluding the central regions of their clusters , it is quite possible that their arbitrary method of defining the affected regions does not fully remove the effects of multiphase gas and the resulting si bias we have described .
anders e. , grevesse n. , 1989 , geochimica et cosmochimica acta , 53 , 197 arimoto n. , matsushita k. , ishimaru y. , ohashi t. , renzini a. , 1997 , apj , 477 , 128 arnaud k. , 1996 , in jacoby g. and barnes j. , eds . , astronomical data analysis software and systems v , asp conf .
series volume 101 , p17 arnaud , m. , rothenflug , r. , 1985 , a&as , 60 , 425 arnaud , m. , raymond , j. , 1992 , apj , 398 , 394 baluciska - church m. , mccammon d. , 1992 , 400 , 699 bauer f , bregman j. n. , 1996 , apj , 457 , 382 beers t. c. , kriessler j. r. , bird c. m. , huchra j. p. , 1995
, aj , 109 , 874 bregman j. n. , hogg d. e. , roberts m. s. , 1992 , apj , 387 , 484 brighenti f. , mathews w. g. , 1997 , apj , 486 , l83 brighenti f. , mathews w. g. , 1998 , apj , 495 , 239 brighenti f. , mathews w. g. , 1999a , apj , in press ( astro - ph/9808175 ) brighenti f. , mathews w. g. , 1999b , apj , in press ( astro - ph/9811258 ) buote d. a. , 1999 , mnras , in press ( astro - ph/9811080 ) ( b99 ) buote d. a. , canizares c. r. , 1994 , apj , 427 , 86 buote d. a. , fabian a. c. , 1998 , mnras , 296 , 977 ( bf ) buote d. a. , canizares c. r. , fabian a. c. , 1999 , mnras , in press ( astro - ph/9804290 ) canizares c. r. , fabbiano g. , trinchieri g. , 1987 , apj , 312 , 503 ciotti l. , dercole a. , pellegrini s. , renzini a. , 1991 , apj , 376 , 380 david l. p. , jones c. , forman w. , daines s. , 1994 , apj , 428 , 544 davis d. s. , white iii r. e. , 1996 , apj , 470 , l35 davis d. s. , mushotzky r. f. , mulchaey j. s. , worrall d. , birkinshaw m. , burstein d. , 1996 , apj , 460 , 601 davis d. s. , mulchaey j. s. , mushotzky r. f. , 1999 , apj , 511 , 34 de vaucouleurs g. , de vaucouleurs a. , corwin h.g . , buta r. j. , paturel g. , fouqu p. , 1991
, third reference catalogue of bright galaxies ( austin : univ . texas press ) ( rc3 ) dickey j. m. , lockman f. j. , 1990 , ara&a , 28 , 215 faber s. m. , wegner g. , burstein d. , davies r. l. , dressler a. , lynden - bell d. , terlevich r. j. , 1989 , apjs , 69 , 763 fabian a. c. , 1994 , ara&a , 32 , 277 fabian a. c. , nulsen p. e. j. , canizares c. r. , 1984 , nature , 310 , 733 fabian a. c. , arnaud k. a. , bautz m. w. , tawara y. , 1994 , apj , 436 , l63 feldman u. , 1992 , physica scripta , 46 , 202 finoguenov a. , ponman t. j. , 1999 , mnras , in press ( astro - ph/9901100 ) finoguenov a. , jones c. , forman w. , david l. , 1999 , apj , in press ( astro - ph/9810107 ) fujita y. , fukumoto j. , okoshi k , 1997 , apj , 488 , 585 fukazawa y. , et al . , 1996 , pasj , 48 , 395 fukazawa y. , ishida m. , ebisawa k. , 1997 , crab calibrations on the xrt+gis energy response functions , in asca news vol . 5 , ( nasa / gsfc : greenbelt ) , 3 fukazawa y. , et al . , 1998 , pasj , 50 , 187 gendreau k. , yaqoob t. , 1997 , asca xrt calibration issues , in asca news vol . 5 , ( nasa / gsfc : greenbelt ) , 8 hickson p. , mendes de oliveira c. , huchra j. p. , palumbo g. g. c. , 1992 , apj , 399 , 353 ikebe y. , et al . , 1996 , nature , 379 , 427 ikebe y. , et al . , 1997 ,
apj , 481 , 660 irwin j. a. , bregman j. n. , evrard a. e. , 1999 , apj , in press ( astro - ph/9901406 ) ishimaru y. , arimoto n. , 1997 , pasj , 49 , 1 johnstone r. m. , fabian a. c. , edge a. c. , thomas p. a. , 1992 , mnras , 255 , 431 kikuchi k. , furusho t. , ezawa h. , yamasaki n. y. , ohashi t. , fukazawa y. , ikebe y. , 1999 , pasj , in press ( astro - ph/9903431 ) kim d .- w . , fabbiano g. , matsumoto h. , koyama k. , trinchieri g. , 1996 , apj , 468 , 175 kunieda h. , et al . , xrt status report , in asca news vol . 5 , ( nasa / gsfc : greenbelt ) , 3 liedahl d. a. , osterheld a. l. , goldstein w. h. , 1995 , apj , 438 , l115 markevitch m. , forman w. r. , sarazin c. l. , vikhlinin a. , 1998 , apj , 503 , 77 matsumoto h. , koyama k. , awaki h. , tsuru t. , loewenstein m. , matsushita k. , 1997 , apj , 482 , 133 mewe r. , gronenschild e. h. b. m. , van den oord g. h. j. , 1985 , a&as , 62 , 197 mulchaey j. s. , zabludoff a. i. , 1998 , apj , 496 , 73 mulchaey j. s. , zabludoff a. i. , 1999 , apj , in press ( astro - ph/9810458 ) mushotzky r. f. , loewenstein m. , arnaud k. a. , tamura t. , fukazawa y. , matsushita k. , kikuchi k. , hatsukade i. , 1996 , apj , 466 , 686 nulsen p. e. j. , 1986 , mnras , 221 , 377 nulsen p. e. j. , 1998 , in soker n. ed . , galactic and cluster cooling flows , asp conf .
series , vol .
115 , ( asp : san francisco ) , 135 odea c. p. , baum s. a. , maloney p. r. , tacconi l. j. , sparks w. b. , 1994 , apj , 422 , 467 pellegrini s. , ciotti l. , 1998 , a&a , 333 , 433 pildis r. a. , bregman j. n. , evrard a. e. , 1995 , apj , 443 , 514 ponman t. j. , bourner p. d. , j. , ebeling h. , bhringer h. , 1996 , mnras , 283 , 690 ramella m. , geller m. j. , huchra j. p. , thorstensen j. r. , 1995 , aj , 109 , 1469 raymond j. c. , smith b. w. , 1977 , apjs , 35 , 419 renzini a. , 1994 , in durret f. , mazure a. , tran thanh van j. eds . , clusters of galaxies , ( gis - sur - yvette : editions frontires ) , 221 renzini a. , 1997 , apj , 488 , 35 renzini a. , 1999 , in walsh j. and rosa m. , chemical evolution from zero to high redshift , ( berlin : springer ) , in press ( astro - ph/9902361 ) renzini a. , ciotti l. , dercole a. , pellegrini s. , 1993 , apj , 419 , 52 takahasi t. , et al . , extended source analysis with asca , in asca news vol . 3 , ( nasa / gsfc : greenbelt ) , 34 tanaka y. , inoue h. , & holt s. s. , 1994 , pasj , 46 , l37 thomas p. a. , fabian a. c. , nulsen p. e. j. , 1987 , mnras , 228 , 973 tonry j. l. , davis m. , 1981 , apj , 246 , 666 trinchieri g. , fabbiano g. , kim d .- w . , 1997 ,
a&a , 318 , 361 trinchieri g. , kim d .- w . , fabbiano g. , canizares c. , 1994 , apj , 428 , 555 trmper j. , 1983 , adv . space res . , 2 , 241 white d. a. , fabian a. c. , 1995 , mnras , 273 , 72 white r. e. iii , sarazin c. l. , 1987 , apj , 318 , 621 xu h. , makishima k. , fukazawa y. , ikebe y. , kikuchi k. , ohashi t. , tamura t. , 1998 , apj , 500 , 738 zabludoff a. i. , mulchaey j. s. , 1998 , apj , 496 , 39 | we analyze the _ asca _ spectra accumulated within @xmath0 kpc radii of 12 of the brightest groups of galaxies . upon fitting isothermal models ( 1 t ) jointly to the _ asca _ sis and gis spectra we obtain fits for most groups that are of poor or at best marginal quality and give very sub - solar metallicities similar to previous studies , @xmath1 @xmath2 .
two - temperature models ( 2 t ) provide significantly better fits for 11 out of the 12 groups and in every case have metallicities that are substantially larger than obtained for the 1 t models , @xmath3 @xmath2 . although not very well constrained , for most of the groups absorption in excess of the galactic value is indicated for the cooler temperature component of the 2 t models .
a simple multiphase cooling flow model gives results analogous to the 2 t models including large metallicities , @xmath4 @xmath2 .
the nearly solar fe abundances and also solar @xmath5/fe ratios indicated by the 2 t and cooling flow models are consistent with models of the chemical enrichment of ellipticals , groups , and clusters which assume ratios of type ia to type ii supernova and an imf similar to those of the milky way .
thus , we have shown that the very sub - solar fe abundances and si / fe enhancements obtained from most previous studies within @xmath6 kpc of galaxy groups are an artifact of their fitting isothermal models to the x - ray spectra which also has been recently demonstrated for the brightest elliptical galaxies . owing to the importance of these results for interpreting x - ray spectra , in an appendix we use simulated _ asca _ observations to examine in detail the `` fe bias '' and `` si bias '' associated with the spectral fitting of ellipticals , groups , and clusters of galaxies .
galaxies : general galaxies : evolution x - rays : galaxies . |
in many multimedia applications , a stream of data packets is required to be sequentially encoded and decoded under strict latency constraints .
for such a streaming setup , both the fundamental limits and optimal schemes can differ from classical communication systems . in recent years
, there has been a growing interest in the characterization of fundamental limits for streaming data transmission @xcite . in @xcite ,
coding techniques based on tree codes were proposed for streaming setup with applications to control systems . in @xcite ,
khisti and draper established the optimal diversity - multiplexing tradeoff ( dmt ) for streaming over a block - fading multiple - input multiple - output channel . in @xcite ,
the same authors proposed a coding technique using finite memory for streaming over discrete memoryless channels ( dmcs ) that attains the same reliability as previously known semi - infinite coding techniques with growing memory . in @xcite ,
the error exponent was studied in a streaming setup of distributed source coding .
we note that these prior works assumed that the code operates in the large deviations regime in which the rate is bounded away from capacity ( or the rate pair is strictly inside the optimal rate region for compression problems ) and the error probability decays exponentially as the blocklength increases .
other interesting asymptotic regimes include the central limit and moderate deviations regimes .
let @xmath2 denote the blocklength of a single message henceforth .
in the central limit regime , the rate approaches to the capacity at a speed proportional to @xmath3 and the error probability does not vanish as the blocklength increases . in the moderate deviations regime ,
the rate approaches to the capacity strictly slower than @xmath3 and the error probability decays sub - exponentially fast as the blocklength increases . for block coding problems ,
both regimes have received a fair amount of attention recently . these works
aim to characterize the fundamental interplay between the coding rate and error probability .
the most notable early work on channel coding in the central limit regime ( also known as second - order asymptotics or the normal approximation regime ) is that of strassen @xcite , who considered dmcs and showed that the backoff from capacity scales as @xmath4 when the error probability is fixed .
strassen also deduced the constant of proportionality , which is related to the so - called _ dispersion _
hayashi @xcite considered dmcs with cost constraints as well as discrete channels with markovian memory .
@xcite refined the asymptotic expansions and also compared the normal approximation to the finite blocklength ( non - asymptotic ) fundamental limits . for a review and extensions to multi - terminal models ,
the reader is referred to @xcite . for the moderate deviations regime , he _
et al . _
@xcite considered fixed - to - variable length source coding with decoder side information .
altu and wagner @xcite initiated the study of moderate deviations for channel coding , specifically dmcs .
polyanskiy and verd @xcite relaxed some assumptions in the conference version of altu and wagner s work @xcite and they also considered moderate deviations for additive white gaussian noise ( awgn ) channels .
however , this line of research has not been extensively studied for the streaming setup . to the best of our knowledge , there has been no prior work on the streaming setup in the moderate deviations and central limit regimes with the exception @xcite where the focus is on source coding . in this paper
, we study streaming data transmission over a dmc in the moderate deviations and central limit regimes .
our streaming setup is illustrated in fig .
[ fig : streaming ] . in each block of length @xmath2 ,
a new message is given to the encoder at the beginning , and the encoder generates a codeword as a function of all the past and current messages and transmits it over the channel .
the decoder , given all the past received channel output sequences , decodes each message after a delay of @xmath0 blocks .
this streaming setup introduces a new dimension not present in the block coding problems studied previously .
in the special case of @xmath5 , the setup reduces to the block channel coding problem .
if @xmath6 , however , there exists an inherent tension in whether we utilize a block only for the fresh message or use it also for the previous messages with earlier deadlines .
it is not difficult to see that due to the memoryless nature of the model , a time sharing scheme will not provide any gain compared to the case of @xmath5 .
a natural question is whether a joint encoding of fresh and previous messages would improve the performance when @xmath6 .
our results indicate that the fundamental interplay between the rate and error probability can be greatly improved when delay is allowed in the streaming setup . in the moderate deviations regime
, the moderate deviations constant is shown to improve over the block coding or non - streaming setup by a factor of @xmath0 . in the central limit regime , the second - order coding rate is shown to improve by a factor of approximately @xmath1 for a wide range of channel parameters .
for both asymptotic regimes , we propose coding techniques that incorporate a joint encoding of fresh and previous messages . for the moderate deviations regime ,
we propose a coding technique in which , for every block , the encoder jointly encodes all the previous and fresh messages and the decoder re - decodes all the previous messages in addition to the current target message . for the error analysis of this coding technique , we develop a refined and non - asymptotic version of the moderate deviations upper bound in ( * ? ? ?
* theorem 3.7.1 ) that allows us to uniformly bound the error probabilities associated with the previous messages . on the other hand , for the central limit regime
, we can not apply such a coding technique whose memory is linear in the block index . in the error analysis in the central limit regime
, we encounter a summation of constants as a result of applications of the central limit theorem .
if the memory is linear in the block index , this summation causes the upper bound on the error probability to diverge as the block index tends to infinity .
hence , for the central limit regime , we propose a coding technique with _ truncated _ memory where the memory at the encoder varies in a periodic fashion .
our proposed construction judiciously balances the rate penalty imposed due to the truncation and the growth in the error probability due to the contribution from previous messages . by analyzing the second - order coding rate of our proposed setup ,
we conclude that the channel dispersion parameter also decreases approximately by a factor of @xmath0 for a wide range of channel parameters .
furthermore , we explore interesting variants of the basic streaming setup in the moderate deviations regime .
first , we consider a scenario where there is an erasure option at the decoder and analyze the undetected error and the total error probabilities , extending a result by hayashi and tan @xcite .
next , by utilizing the erasure option , we analyze the rate of decay of the error probability when a variable decoding delay is allowed .
we show that such a flexibility in the decoding delay can dramatically improve the error probability in the streaming setup .
this result is the analog of the classical results on variable - length decoding ( see e.g. , @xcite ) to the streaming setup . finally , as a simple example for the case where the message rates are not constant , we consider a scenario where the rate of the messages in odd block indices and the rate of the messages in even block indices are different and analyze the moderate deviations constants separately for the two types of messages .
this setting finds applications in video and audio coding where streams of data packets do not necessarily have a constant rate .
the rest of this paper is organized as follows . in section [ sec :
model ] , we formally state our streaming setup .
the main theorems are presented in section [ sec : main ] and proved in section [ sec : proof ] . in section [ sec : extension ] , the moderate deviations result for the basic streaming setup is extended in various directions .
we conclude this paper in section [ sec : conclusion ] .
the following notation is used throughout the paper .
we reserve bold - font for vectors whose lengths are the same as blocklength @xmath2 . for two integers @xmath7 and @xmath8 , @xmath9 $ ]
denotes the set @xmath10 . for constants
@xmath11 and @xmath12 $ ] , @xmath13 denotes the vector @xmath14 and @xmath15 denotes @xmath16}$ ] where the subscript is omitted when @xmath17 , i.e. , @xmath18}$ ] .
this notation is naturally extended for vectors @xmath19 , random variables @xmath20 , and random vectors @xmath21 .
@xmath22 for an event @xmath23 denotes the indicator function , i.e. , it is 1 if @xmath23 is true and 0 otherwise .
@xmath24 and @xmath25 denote the ceiling and floor functions , respectively .
for a dmc @xmath26 and an input distribution @xmath27 , we use the following standard notation and terminology in information theory : * information density : @xmath28 where @xmath29 denotes the output distribution .
we note that @xmath30 depends on @xmath27 and @xmath31 but this dependence is suppressed .
the definition can be generalized for two vectors @xmath32 and @xmath33 of length @xmath34 as follows : @xmath35 * mutual information : @xmath36\\ & = \sum_{x\in \mathcal{x}}\sum_{y\in \mathcal{y } } p(x)w(y|x)\log\frac{w(y|x)}{pw(y)}.\end{aligned}\ ] ] * unconditional information variance : @xmath37.\end{aligned}\ ] ] * conditional information variance : @xmath38.\end{aligned}\ ] ] * capacity : @xmath39 where @xmath40 denotes the probability simplex on @xmath41 . *
set of capacity - achieving input distributions : @xmath42 * channel dispersion @xmath43 where @xmath44 is from ( * ? ? ?
* lemma 62 ) , where it is shown that @xmath45 for all @xmath46 .
consider a dmc @xmath26 .
a streaming code is defined as follows : [ def : basic ] an @xmath47-streaming code consists of * a sequence of messages @xmath48 each distributed uniformly over @xmath49 $ ] , * a sequence of encoding functions @xmath50 that maps the message sequence @xmath51 to the channel input codeword @xmath52 , and * a sequence of decoding functions @xmath53 that maps the channel output sequences @xmath54 to a message estimate @xmath55 , that satisfies @xmath56 i.e. , the probability of error averaged over all block messages does not exceed @xmath57 .
we note that a streaming code with a _ fixed _
blocklength @xmath2 consists of a _ sequence _ of encoding and decoding functions since a stream of messages is sequentially encoded and decoded .
[ fig : streaming ] illustrates our streaming setup for the case with @xmath58 . in the beginning of block
@xmath59 , new message @xmath60 is given to the encoder .
the encoder generates a codeword @xmath61 as a function of all the past and current messages @xmath62 and transmits it over the channel in block @xmath63 . since @xmath58 ,
the decoder decodes message @xmath60 at the end of block @xmath64 , as a function of all the past received channel output sequences @xmath65 .
in this section , we state our main results .
the following two theorems present achievability bounds for the moderate deviations and the central limit regimes , respectively , which are proved in section [ sec : proof ] .
[ thm : md ] consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .. ] then , there exists a sequence of @xmath71-streaming codes such that for some @xmath72 , @xmath73 corresponds to an upper bound on the moderate deviations constant . in the special case of @xmath5 ,
the moderate deviations constant is shown to be the channel dispersion @xmath74 in @xcite . ]
@xmath75 [ thm : cl ] consider a dmc @xmath26 with @xmath66 .
for any @xmath76 and @xmath77 , there exists a sequence of @xmath71-streaming codes such that is termed second - order coding rate in this paper .
this is slightly different from what is common in the literature where instead @xmath78 is known as the second - order coding rate @xcite . ]
@xmath79 and @xmath80 the following corollary , whose proof is in appendix [ appendix : asymptotic ] , elucidates a closed - form and interpretable expression for the upper bound on the error probability in .
[ coro : cl_asymp ] consider a dmc @xmath26 with @xmath66 .
for any @xmath76 , there exists a sequence of @xmath71-streaming codes such that @xmath81 and @xmath82 where @xmath83 defined in the following has the property that for every @xmath84 , @xmath83 tends to 1 as @xmath85 tends to infinity : @xmath86 fig .
[ fig : const ] illustrates how fast the constant @xmath83 in corollary [ coro : cl_asymp ] converges to 1 as @xmath85 increases . for @xmath58
, we can see that @xmath83 is less than 1.1 when @xmath87 and is less than 1.05 when @xmath88 .
hence , the effect of the constant @xmath83 is not significant for a wide range of @xmath89 and @xmath0 .
theorems [ thm : md ] and [ thm : cl ] illustrate that the fundamental interplay between the rate and probability of error can be greatly improved when delay is allowed in the streaming setup . in the moderate deviations regime ,
the moderate deviations constant improves by a factor of @xmath0 . assuming that @xmath83 can be approximated sufficiently well by @xmath90 , for the central limit regime
, the second - order coding rate @xmath91 is improved ( reduced ) by a factor of @xmath1 . another way to view this via the lens of the channel dispersion @xmath74 ; this parameter is approximately reduced by a factor of @xmath0 .
consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .
we denote by @xmath92 an input distribution that achieves the dispersion .
for each @xmath59 and @xmath93 , generate @xmath94 in an i.i.d .
manner according to @xmath92 .
the generated codewords constitute the codebook @xmath95 . in block
@xmath63 , after observing the true message sequence @xmath62 , the encoder sends @xmath96 .
consider the decoding of @xmath60 at the end of block @xmath97 . in our scheme ,
the decoder not only decodes @xmath60 , but also re - decodes @xmath98 at the end of block @xmath99 . for @xmath100
$ ] has been already decoded at the end of block @xmath101 .
nevertheless , the decoder re - decodes @xmath102 at the end of @xmath99 , because the decoder needs to decode @xmath102 to decode @xmath60 and the probability of error associated with @xmath102 becomes lower ( in general ) by utilizing recent channel output sequences .
] let @xmath103 denote the estimate of @xmath104 at the end of block @xmath99 .
the decoder decodes @xmath104 sequentially from @xmath105 to @xmath106 as follows : * given @xmath107}$ ] , the decoder chooses @xmath103 according to the following rule .
, @xmath108 is null . ]
if there is a unique index @xmath109 that satisfies for @xmath110 denote the set of message indices mapped to the @xmath8-th codeword according to the encoding procedure . for @xmath111 and @xmath112 , we denote by @xmath113 the set of codewords @xmath114 . ] @xmath115}(\hat{g}_{t_k , [ 1:j-1 ] } , g_{[j : t_k ] } ) , \mathbf{y}_{[j : t_k ] } ) & > ( t_k - j+1 ) \cdot
\log m_n \label{eqn : dec_rule}\end{aligned}\ ] ] for some @xmath116}$ ] , let @xmath117 .
in is defined in terms of @xmath92 and @xmath31 .
this dependence is suppressed henceforth . ] if there is none or more than one such @xmath118 , let @xmath119 . *
if @xmath120 , repeat the above procedure by increasing @xmath8 to @xmath121 . if @xmath106 , the decoding procedure terminates and the decoder declares that the @xmath63-th message is @xmath122 .
now , fix an arbitrary @xmath131 . by applying the chain of inequalities ( * ? ? ?
* eq . ( 53)-(56 ) ) , we have @xmath132^+\right\}\cr & \leq \mathbbm{1}\left\{\sum_{l=1}^{n(t_k - j+1 ) } i(x_l;y_l)\leq ( t_k - j+1)n(c-\lambda\rho_n)\right\ } + \exp\left\{-(t_k - j+1)n(1-\lambda)\rho_n\right\}. \label{eqn : spr}\end{aligned}\ ] ] combining the bounds in and , we obtain @xmath133 for sufficiently large @xmath2 , where @xmath134 is some non - negative constant dependent only on the input distribution @xmath135 and channel statistics @xmath136 and @xmath44 is from the moderate deviations upper bound in lemma [ lemma : nonasymptotic_md ] , which is relegated to the end of this subsection . also see remark [ rem : md ] .
now , we have @xmath137\cr & \leq \sum_{j=1}^k \left ( \exp\left\{-(t_k - j+1)n\rho_n^2\lambda^2\left(\frac{1}{2v}-\lambda \rho_n \tau\right)\right\}+\exp\left\{-(t_k - j+1)n(1-\lambda)\rho_n\right\ } \right)\\ & \leq \sum_{j = t}^{t_k } \left ( \exp\left\{-jn\rho_n^2\lambda^2\left(\frac{1 } { 2v}-\lambda \rho_n \tau\right)\right\}+\exp\left\{-jn(1-\lambda)\rho_n\right\}\right ) \\ & \leq \frac{\exp\left\{-tn\rho_n^2\lambda^2\left(\frac{1 } { 2v}-\lambda \rho_n \tau\right)\right\}}{1-\exp\{-n\rho_n^2\lambda^2\left(\frac{1 } { 2v}-\lambda \rho_n \tau\right)\}}+\frac{\exp\left\{-tn(1-\lambda)\rho_n\right\}}{1-\exp\left\{-n(1-\lambda)\rho_n\right\ } } \label{eqn : md_sumup}\end{aligned}\ ] ] for sufficiently large @xmath2 , which leads to @xmath138\leq -\frac{t\lambda^2}{2v}.\end{aligned}\ ] ] finally , by taking @xmath139 , we have @xmath138\leq -\frac{t}{2v}.\end{aligned}\ ] ] hence , there must exist a sequence of codes @xmath95 that satisfies , which completes the proof .
the following lemma used in the proof of theorem [ thm : md ] corresponds to a non - asymptotic upper bound of the moderate deviations theorem ( * ? ? ?
* theorem 3.7.1 ) , whose proof is in appendix [ appendix : finite_md ] . [
lemma : nonasymptotic_md ] let @xmath140 be a sequence of i.i.d .
random variables such that @xmath141 = 0 $ ] , @xmath142 = \sigma^2>0 $ ] , and its cumulant generating function @xmath143 $ ] for @xmath144 is analytic around the origin and satisfies that @xmath145 } | h'''(s ) |$ ] is finite . for a sequence @xmath146 satisfying the moderate deviations constraints , i.e. , @xmath147 and @xmath148 , the following bound holds : @xmath149 for sufficiently large @xmath2 .
[ rem : md ] let us comment on the assumption in lemma [ lemma : nonasymptotic_md ] that @xmath150 is finite . in our application
, @xmath151 then , we have @xmath152\\ & = -s i(x_1;y_1 ) + \log \e \left [ \big(\frac{w(y_1 |x_1)}{p_xw(y_1 ) } \big)^s\right ] . \end{aligned}\ ] ] by differentiating thrice , we can show that @xmath153 is continuous in @xmath154 . restricting @xmath154 to @xmath155 $ ]
means that @xmath153 is a continuous function over a compact set .
hence its maximum is attained and is necessarily finite . consider a dmc @xmath26 with @xmath66 .
we remark that in the moderate deviations regime , for every block , the encoder maps _ all _ the previous messages to a codeword . for the central limit regime ,
we propose a coding strategy where the encoder maps only _ some _ recent messages to the codeword in each block .
similar idea of incorporating truncated memory was used in @xcite with the focus on reducing the complexity . here
, we use a different memory structure from @xcite .
let @xmath156 and @xmath157 denote the maximum and the minimum numbers of messages that can possibly be mapped to a codeword in each block , respectively .
we choose the size @xmath67 of message alphabet as follows : @xmath158 for some @xmath76 . to make the above choice of @xmath67 valid
, we assume @xmath159 .
furthermore , we assume that the minimum encoding memory is at least @xmath0 , i.e. , @xmath160 .
we denote by @xmath92 an input distribution that achieves the dispersion .
our encoder has a periodically time - varying memory @xmath161 $ ] with a period of @xmath162 blocks , after an initialization step of the first @xmath163 blocks .
let us first describe our message - codeword mapping rule for the case of @xmath164 and @xmath165 , which is illustrated in fig .
[ fig : mapping ] . for the first nine blocks , the encoder maps all the previous messages to a codeword .
since the maximum encoding memory is nine in this example , we _ truncate _ the messages that are mapped to a codeword on and after the tenth block , so that the encoding memory is periodically time - varying from four to nine with a period of six blocks .
for instance , let us consider the first period from the tenth block to the fifteenth block . in the tenth block , the encoder maps the messages @xmath166 to a codeword , thus ensuring that the encoding memory is four . in block
@xmath167 $ ] , the encoder maps the messages @xmath168 to a codeword and hence the encoding memory becomes the maximum memory of nine when @xmath169 .
now , let us formally describe the encoding procedure for the general case . for each @xmath170 $ ] and @xmath93 , generate @xmath94 in an i.i.d .
manner according to @xmath92 . in block
@xmath170 $ ] , the encoder sends @xmath96 .
let @xmath171 for @xmath172 denote the set of @xmath173 block indices in the @xmath174-th period on and after the @xmath175-st block , i.e. , @xmath176 .
for each @xmath177 and @xmath178 , , a total of @xmath179 messages , i.e. , @xmath180 , are mapped to a codeword . ]
generate @xmath181 in an i.i.d .
manner according to @xmath92 . in block
@xmath182 , the encoder sends @xmath183})$ ] . on the other hand
, we note that our message - codeword mapping rule is also periodic in the ( vertical ) axis of message index .
we can group the messages according to the maximum block index to which a message is mapped .
let @xmath184 for @xmath185 denote the @xmath174-th group @xmath186 of messages that are mapped to a codeword up to block @xmath187 , which is illustrated in fig .
[ fig : mapping ] for the example of @xmath164 and @xmath165 .
this grouping rule is useful for describing the decoding rule .
the decoding rule of @xmath188 at the end of block @xmath99 is exactly the same as that for the moderate deviations regime .
hence , from now on , let us focus on the decoding of @xmath189 for @xmath190 at the end of block @xmath99 . at the end of block @xmath99 ,
the decoder decodes not only @xmath60 , but also all the messages in the previous group and the previous messages in the current group , for @xmath100 $ ] has been already decoded at the end of block @xmath101 . nevertheless , the decoder re - decodes some of the previous messages at the end of @xmath99 . ]
i.e. , @xmath191 .
let @xmath103 denote the estimate of @xmath104 at the end of block @xmath99 .
let us first describe our decoding procedure for the example of @xmath58 , @xmath164 , and @xmath165 illustrated in fig .
[ fig : mapping ] . consider the decoding of @xmath192 at the end of block @xmath99 .
for the cases @xmath193 and @xmath194 can be stated in a similar manner . ]
the decoder decodes not only @xmath60 , but also all the messages @xmath195 in @xmath196 and the previous messages @xmath197 in @xmath198 .
the underlying rules of our decoding procedure can be summarized as follows : * since messages @xmath199 in @xmath200 , which we do not want to decode , are involved in blocks @xmath201 , we do not utilize the channel output sequences in those blocks for decoding . *
for the decoding of the @xmath8-th message for @xmath202 $ ] , among the channel output sequences from block @xmath203 to block @xmath99 , we utilize the channel output sequences in which the @xmath8-th message is involved . according to the above rules , the blocks to be considered for the decoding of messages @xmath204 are as follows : a. for @xmath205 , blocks are involved is @xmath99 if @xmath206 , and it is @xmath207 otherwise .
in other words , the last block index to which the messages in @xmath196 are involved is @xmath208 . ]
indexed from 10 to @xmath209 , b. for @xmath104 for @xmath210 $ ] , blocks indexed from @xmath8 to @xmath211 , and c. for @xmath212 for @xmath213 $ ] , blocks indexed from @xmath8 to @xmath99 . in particular , since the pairs of the first block index and the last block index to be considered for the decoding of messages @xmath205 are the same , we decode @xmath205 simultaneously . by keeping this in mind , our decoding procedure for @xmath214 for the example of @xmath58 , @xmath164 and @xmath165 is formally stated as follows : a. if there is a unique index vector @xmath215}$ ] that satisfies , the following notation is used for the set of codewords .
let @xmath216 for @xmath110 denote the set of message indices mapped to the @xmath8-th codeword according to the encoding procedure . for @xmath111 and @xmath112
, we denote by @xmath113 the set of codewords @xmath114 . ]
@xmath217}(g_{[7:\nu_k]}),\mathbf{y}_{[10:\nu_k]})&>(\nu_k-6)\cdot \log m_n\label{eqn : dec_rule_cl1_toy}\end{aligned}\ ] ] for some @xmath218}$ ] , let @xmath219}=g_{[7:10]}$ ] .
if there is none or more than one such @xmath215}$ ] , let @xmath219}=(1,\cdots , 1)$ ] .
b. the decoder sequentially decodes @xmath118 from @xmath220 to @xmath221 as follows : * given @xmath222}$ ] , the decoder chooses @xmath103 according to the following rule . if there is a unique index @xmath109 that satisfies @xmath223}(\hat{g}_{t_k,[7:j-1 ] } , g_{[j:\nu_k ] } ) , \mathbf{y}_{[j:\nu_k ) ] } ) & > ( \nu_k - j+1 ) \cdot \log m_n \label{eqn : dec_rule_cl2_toy}\end{aligned}\ ] ] for some @xmath224}$ ] , let @xmath117 . if there is none or more than one such @xmath118 , let @xmath119 .
* if @xmath220 , repeat the above procedure by increasing @xmath8 to @xmath225 . if @xmath221 , proceed to the next decoding procedure .
c. the decoder sequentially decodes @xmath118 from @xmath226 to @xmath106 as follows : * given @xmath227}$ ] , the decoder chooses @xmath228 according to the following rule . if there is a unique index @xmath109 that satisfies @xmath115}(\hat{g}_{t_k,[7:j-1 ] } , g_{[j : t_k ] } ) , \mathbf{y}_{[j : t_k ] } ) & > ( t_k - j+1 ) \cdot
\log m_n \label{eqn : dec_rule_cl3_toy}\end{aligned}\ ] ] for some @xmath116}$ ] , let @xmath117 .
if there is none or more than one such @xmath118 , let @xmath119 .
* if @xmath120 , repeat the above procedure by increasing @xmath8 to @xmath121 . if @xmath106 , the whole decoding procedure terminates and the decoder declares that the @xmath63-th message is @xmath122 . the above description of the decoding procedure for the example in fig .
[ fig : mapping ] is naturally extended for the general case . in general , the procedure for decoding of @xmath189 for @xmath190 at the end of block @xmath99 consists of the following three steps : ( i ) simultaneous non - unique decoding of the first @xmath229 messages in the previous group , ( ii ) sequential decoding of the remaining @xmath230 messages in the previous group , and ( iii ) sequential decoding of the messages in the current group up to the current block .
let us describe the decoding rule when @xmath193 in the following : a. if there is a unique index vector @xmath231 that satisfies @xmath232}(g^{\min(a , t_k)}),\mathbf{y}_{[b:\min ( a , t_k)]})&>\min(a , t_k)\cdot \log m_n\label{eqn : dec_rule_cl1}\end{aligned}\ ] ] for some @xmath233}$ ] , let @xmath234}=g^b$ ] . if there is none or more than one such @xmath231 , let @xmath235}=(1,\cdots,1)$ ] .
b. the decoder sequentially decodes @xmath118 from @xmath236 to @xmath237 as follows : * given @xmath107}$ ] , the decoder chooses @xmath103 according to the following rule . if there is a unique index @xmath109 that satisfies @xmath238}(\hat{g}_{t_k,[1:j-1 ] } , g_{[j:\min(a , t_k ) ] } ) , \mathbf{y}_{[j:\min(a , t_k ) ] } ) & > ( \min(a , t_k)-j+1 ) \cdot \log m_n \label{eqn : dec_rule_cl2}\end{aligned}\ ] ] for some @xmath239}$ ] , let @xmath117 . if there is none or more than one such @xmath118 , let @xmath119 .
* if @xmath240 , repeat the above procedure by increasing @xmath8 to @xmath121 . if @xmath237 , proceed to the next decoding procedure .
c. the decoder sequentially decodes @xmath118 from @xmath241 to @xmath106 as follows : * given @xmath107}$ ] , the decoder chooses @xmath103 according to the following rule . if there is a unique index @xmath109 that satisfies @xmath115}(\hat{g}_{t_k,[1:j-1 ] } , g_{[j : t_k ] } ) , \mathbf{y}_{[j : t_k ] } ) & > ( t_k - j+1 ) \cdot
\log m_n \label{eqn : dec_rule_cl3}\end{aligned}\ ] ] for some @xmath116}$ ] , let @xmath117 .
if there is none or more than one such @xmath118 , let @xmath119 .
* if @xmath120 , repeat the above procedure by increasing @xmath8 to @xmath121 .
if @xmath106 , the whole decoding procedure terminates and the decoder declares that the @xmath63-th message is @xmath122 . by exploiting the symmetry of the message - codeword mapping rule , the decoding rule for @xmath242 proceeds similarly .
we note that the superscript in each error event represents the decoding step in which the error event is involved .
now , we have @xmath126 & \leq \pr(\mathcal{e}_{k}^{\mathrm{(i)}})+\pr ( \tilde{\mathcal{e}}_{k}^{\mathrm{(i)}})+ \sum_{j = b+1}^{a - b+1 } \pr(\mathcal{e}_{k , j}^{\mathrm{(ii ) } } ) + \sum_{j = b+1}^{a - b+1 } \pr ( \tilde{\mathcal{e}}_{k , j}^{\mathrm{(ii ) } } ) \cr & \quad \quad \quad\quad \quad+\sum_{j = a - b+2}^{k } \pr ( \mathcal{e}_{k , j}^{\mathrm{(iii ) } } ) + \sum_{j = a - b+2}^{k } \pr ( \tilde{\mathcal{e}}_{k , j}^{\mathrm{(iii ) } } ) .
\label{eqn : first_case_sum}\end{aligned}\ ] ] let us bound each term in the rhs of ( [ eqn : first_case_sum ] ) .
first , @xmath247 is upper - bounded as follows : @xmath248}(g^{\alpha } ) , \mathbf{y}_{[b:\alpha]})\leq \alpha\cdot \log m_n\right)\\ & \leq \pr\left(\sum_{l=1}^{n(\alpha - b+1 ) } i(x_l;y_l)\leq \alpha \cdot \log m_n\right)\\ & \overset{(a)}{\leq } \pr\left(\sum_{l=1}^{n(\alpha - b+1 ) } i(x_l;y_l)\leq ( \alpha - b+1)(nc - l\sqrt{n})\right)\\ & \overset{(b)}{\leq } q\left(\frac{\sqrt{\alpha - b+1}}{\sqrt{v}}l\right)+\frac{\tau_1}{\sqrt{(\alpha - b+1)n } } \label{eqn : berry_1}\end{aligned}\ ] ] for some non - negative constant @xmath249 that is dependent only on the input distribution @xmath92 and the channel statistics @xmath136 , where @xmath250 s are i.i.d .
random variables each generated according to @xmath251 , @xmath44 is from the choice of @xmath67 in , and @xmath252 is from the berry - esseen theorem ( e.g. , @xcite ) . similarly , we can show @xmath253 and @xmath254 next , @xmath255 is upper - bounded as follows : @xmath256}(g^{\alpha } ) , \mathbf{y}_{[b:\alpha ] } ) > \alpha \cdot \log m_n \mbox { for some } g^{\alpha } \mbox { such that } g^{b } \neq g^b)\\ & \leq m_n^{\alpha } \cdot \pr\left(\sum_{l=1}^{n(\alpha - b+1 ) } i(x_l;\bar{y}_l ) >
\alpha\cdot \log m_n\right)\\ & \overset{(a)}{= } m_n^{\alpha}\cdot \e\left[\exp\left\{-\sum_{l=1}^{n(\alpha - b+1 ) } i(x_l;y_l)\right\}\right .
\cdot \left.\mathbbm{1}\big\{\sum_{l=1}^{n(\alpha - b+1 ) } i(x_l;y_l)>\alpha \log m_n\big\}\right]\\ & \overset{(b)}{\leq } \frac{\tau_2}{\sqrt{(\alpha - b+1)n } } \ ] ] for some non - negative constant @xmath257 that is dependent only on the input distribution @xmath92 and channel statistics @xmath136 , where @xmath129 s are i.i.d .
random variables each generated according to @xmath251 @xmath258 , @xmath44 is due to an elementary chain of equalities given in appendix [ appendix : chain ] , and @xmath252 is from ( * ? ? ?
* lemma 47 ) .
similarly , we can show @xmath259 and @xmath260 by substituting the above bounds into the rhs of , we obtain @xmath137\cr & \leq \sum_{j = b}^{a - b+1 } \left(q\left(\frac{\sqrt{\alpha - j+1}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{(\alpha - j+1)n } } \right)\cr & \qquad\qquad+\sum_{j = a - b+2}^{k } \left(q\left(\frac{\sqrt{t_k - j+1}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{(t_k - j+1)n}}\right ) \\ & \leq \sum_{j=\alpha - a+b}^{\alpha - b+1 } \left(q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{jn } } \right)+\sum_{j = t}^{t_k - a+b-1 } \left(q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{jn}}\right ) \label{eqn : sumall}\\ & \overset{(a)}{\leq } \sum_{j = b}^{a - b+1 } \left(q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{jn } } \right)+\sum_{j = t}^{a - b+t } \left(q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{jn}}\right),\label{eqn : sumall_ub}\end{aligned}\ ] ] where @xmath44 is because if @xmath261 , which implies @xmath262 , the rhs of is upper - bounded as follows : @xmath263 and if @xmath264 , which implies @xmath265 , the rhs of is upper - bounded as follows : @xmath266 now , the rhs of is bounded as follows : @xmath267 where @xmath44 is from lemma [ lemma : sum_integral ] ( with the identification of @xmath268 ) , which is relegated to the end of this subsection , and @xmath252 is obtained by applying similar steps as in the proof of corollary [ coro : cl_asymp ] .
can be obtained by replacing @xmath0 by @xmath229 in the rhs of . ]
now let us choose @xmath269 and @xmath270 for @xmath271 . by substituting this choice of @xmath163 and @xmath229 into the rhs of and the rhs of
, we obtain @xmath272 and @xmath126\leq \sum_{j = t}^{\infty } q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+o(n^{-\delta/2 } ) , \label{eqn : ensemble_cl}\end{aligned}\ ] ] respectively . due to the symmetry of the decoding procedure , the bound holds for @xmath189 for @xmath242 . for @xmath188 , by defining the error events in the same way as for the moderate deviations regime and then applying similar bounding techniques used in the above , it can be verified that @xmath126&\leq \sum_{j = t}^{t_k } q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{jn}}\\ & \leq \sum_{j = t}^{a - b+t } q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+\frac{\tau_1+\tau_2}{\sqrt{jn } } \\ & \leq \sum_{j = t}^{\infty } q\left(\frac{\sqrt{j}}{\sqrt{v}}l\right)+o(n^{-\delta/2}).\end{aligned}\ ] ] hence , there must exist a sequence of codes @xmath95 that satisfies and , which completes the proof .
the following basic lemma is used in the proof of theorem [ thm : cl ] , whose proof is omitted .
[ lemma : sum_integral ] assume two integers @xmath273 and @xmath274 such that @xmath275 .
if @xmath276 is monotonically decreasing and integrable on @xmath277 $ ] , we have @xmath278 where @xmath279 denotes the antiderivative of @xmath276 .
in this section , we explore interesting variations of the basic streaming setup in section [ sec : model ] . for the brevity of the results , we focus on the moderate deviations regime .
consider the scenario where there is an erasure option at the decoder , i.e. , the decoder can output an erasure symbol instead of a message estimate . in the presence of an erasure option
, there are two types of error events : ( i ) the decoder declares an erasure and ( ii ) the decoder outputs an incorrect message , not an erasure . in many applications , the undetected error ( the latter event ) is more undesirable than an erasure ( the former event ) . in the following ,
we define a streaming code with an erasure option by taking into account the undetected error and the total error probabilities separately .
an @xmath280-streaming code with an erasure option consists of * a sequence of messages @xmath48 each distributed uniformly over @xmath49 $ ] , * a sequence of encoding functions @xmath50 that maps the message sequence @xmath51 to the channel input codeword @xmath52 , and * a sequence of decoding functions @xmath281 that maps the channel output sequences @xmath54 to a message estimate @xmath282 or an erasure symbol @xmath283 , that satisfies @xmath56 i.e. , the total error probability does not exceed @xmath57 , and @xmath284 i.e. , the undetected error probability does not exceed @xmath285 .
the following theorem presents upper bounds on the undetected error and the total error probabilities .
the proof of this theorem is provided in appendix [ appendix : erasure ] .
[ thm : erasure ] consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .
for any @xmath286 , there exists a sequence of @xmath287-streaming codes with an erasure option such that @xmath288 theorem [ thm : erasure ] indicates that for our proposed scheme , the undetected error probability decays much faster than the total error probability , i.e. , the exponent of the undetected error probability is the order of @xmath289 , whereas that of the total error probability is the order of @xmath290 .
we note that when @xmath5 and @xmath291 for @xmath292 and @xmath293 , theorem [ thm : erasure ] reduces to ( * ? ? ?
* theorem 1 ) . in the streaming setup
, both the exponents of the total error and the undetected error probabilities improve over the block coding or non - streaming setup in ( * ? ? ?
* theorem 1 ) by factors of @xmath0 .
we note that the decoding delay is assumed to be fixed to @xmath0 up to this point . in this subsection
, we relax this constraint by requiring the _ average _ decoding delay not to exceed @xmath0 .
a streaming code with average delay constraint is defined as follows : an @xmath294-streaming code with average delay constraint consists of * a sequence of messages @xmath48 each distributed uniformly over @xmath49 $ ] , * a sequence of encoding functions @xmath50 that maps the message sequence @xmath51 to the channel input codeword @xmath52 , and * a sequence of decoding functions @xmath295 that maps the channel output sequences @xmath296 to a message estimate @xmath297 or an erasure symbol @xmath298 for every @xmath127 $ ] that satisfies @xmath299 and @xmath300}{n}\leq t,\end{aligned}\ ] ] where @xmath301 for @xmath59 denotes the ( random ) decoding delay of the @xmath63-th message . is required to be decoded at the end of every block on and after the @xmath63-th block in this definition .
one may wonder why the decoder does not stop decoding @xmath60 after it outputs an estimate of @xmath60 , not an erasure .
we note that our definition includes such a operation as a special case by letting the decoder simply fix the estimate of @xmath60 once it outputs a message estimate . ] for block channel coding with feedback , it is known that the error exponent can be significantly improved by allowing variable decoding delay , e.g. , @xcite . for streaming setup , the following theorem , which is proved in appendix [ appendix : variable ] , shows that such an improvement can be obtained in the absence of feedback .
[ thm : variable ] consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .
for any @xmath302 , there exists a sequence of @xmath303-streaming codes with average delay constraint such that @xmath304 we note that the exponent of the error probability @xmath305 is of the order @xmath289 ( instead of @xmath290 as in ) , and hence it is improved tremendously by allowing variable decoding delay .
we note that the rates of the messages are assumed to be fixed across time thus far . in many practical streaming applications ,
however , a stream of data packets does not have a constant rate .
for example , in the mpec standard for video coding , i frames have higher rates than p frames in general .
similarly , in audio coding , voice packets have higher rates than silent packets . in this subsection , to obtain useful insights when the message rates vary across time , we assume a simple example where the rate of the messages in odd block indices and the rate of the messages in even block indices are different . a streaming code with alternating message rates is defined as follows : an @xmath306-streaming code with alternating message rates . ]
consists of * a sequence of messages @xmath48 where message @xmath307 for @xmath110 is distributed uniformly over @xmath308 $ ] and message @xmath309 for @xmath110 is distributed uniformly over @xmath310 $ ] , * a sequence of encoding functions @xmath311 that maps the message sequence @xmath312 \ } \cup \{g_{2j } : j\in [ 1:\lfloor k/2 \rfloor ] \ }
\in \mathcal{g}_1^{\lceil k/2 \rceil}\times \mathcal{g}_2^{\lfloor k/2 \rfloor } $ ] to the channel input codeword @xmath52 , and * a sequence of decoding functions @xmath313 that maps the channel output sequences @xmath54 to the message estimate @xmath314 , that satisfies @xmath315 and @xmath316 we note that the error probabilities are considered separately for the messages in the odd block indices and for the messages in the even block indices .
the following theorem , which is proved in appendix [ appendix : alternate ] , gives achievability bounds when the message rates are alternating .
[ thm : alternate ] consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .
for any @xmath317 , there exists a sequence of @xmath318-streaming codes with alternating message rates such that @xmath319 if @xmath0 is odd , and @xmath320 if @xmath0 is even .
theorem [ thm : alternate ] indicates that for our coding strategy , the average of the moderate deviations constants does not change even if the rates are alternating .
we note that the moderate deviations constants of the odd and even messages are asymmetric if @xmath0 is odd and they are symmetric otherwise . to illustrate the intuition behind this , let us consider the most recent @xmath0 blocks , which dominates the error probability .
if @xmath0 is odd , the average message rate in those @xmath0 blocks depends on whether the current target message index is odd or even and this leads to the asymmetry . on the other hand ,
if @xmath0 is even , it is fixed to @xmath321 regardless of the current target message index .
more details are provided in appendix [ appendix : alternate ] .
in this paper , we studied the fundamental interplay between the rate and error probability for a streaming setup with a decoding delay of @xmath0 blocks . in the moderate deviations regime
, the moderate deviations constant was shown to improve by at least a factor of @xmath0 .
we proposed a coding technique with infinite memory such that all the previous and fresh messages are jointly encoded in each block . on the other hand , in the central limit regime
, the second - order coding rate was shown to improve by approximately a factor of @xmath1 for a wide range of channel parameters . to ensure that the summation of berry - esseen constants ( e.g. , the last terms in the rhs of - ) does not diverge in the error analysis , we proposed a coding technique with truncated memory such that the encoding and decoding memories do not grow with the block index . furthermore , we generalized the moderate deviations result in various directions .
we first considered a scenario with an erasure option at the decoder and showed that both the exponents of the total error and the undetected error probabilities improve by factors of @xmath0 . by utilizing the erasure option
, we showed that the exponent of the total error probability can be improved to that of the undetected error probability ( in the order sense ) at the expense of a variable decoding delay .
finally , we considered a scenario where the message rates are alternating and showed that the same average moderate deviations constant as the case of constant rate can be obtained .
we note that all of our encoding strategies do not depend on @xmath0 .
hence , our coding techniques are directly applicable for multicast scenario where a sender transmits a common stream of data packets to multiple receivers with possibly different decoding constraints .
let us conclude with a final remark on proving a converse in our streaming setup .
our problem appears to be closely related to the bit - wise unequal protection ( uep ) problem in the sense that we need to capture the tension that arises when a common channel is used for more than one messages with individual error criteria .
for the seemingly simpler bit - wise uep problem @xcite for the block channel coding with the _ same _ decoding deadline , however , tight characterizations of various asymptotic fundamental limits ( e.g. , error exponents ) remain challenging open problems in general .
this indicates that a highly - nontrivial converse technique , perhaps along the lines of sahai s work @xcite , would be needed for our streaming setup where the messages have _ different _ decoding deadlines .
let @xmath322 . note that corollary [ coro : cl_asymp ]
is proved if we show @xmath323 to that end , we use the following bounds on the @xmath324-function : @xmath325 where @xmath326
. then , we have @xmath327 which completes the proof .
fix @xmath328 and @xmath329 . then , we have @xmath330 \\ & \overset{(b)}{=}\exp\left\{-n \left(s{\varepsilon}_n-\log\e [ \exp\{sz_1\}]\right)\right\ } \\ & = \exp\left\{-n \left ( s{\varepsilon}_n - h(s)\right)\right\ } \label{eqn : chernoff } .\end{aligned}\ ] ] where @xmath44 follows from markov s inequality and @xmath252 follows from the independence of @xmath331 s .
the third - order taylor series expansion of the cumulant generating function @xmath332 can be written as @xmath333 for some @xmath334 $ ] .
it is easy to check that @xmath335 , @xmath336 = 0 $ ] and @xmath337=\sigma^2 $ ] .
now , we take @xmath338 plugging this into and yields @xmath339 where the final inequality holds for all @xmath2 sufficiently large since @xmath147 and @xmath340 as @xmath341 and thus @xmath342 .
the following chain of equalities is used in the proof theorem [ thm : cl ] .
\label{eqn : chain4}\end{aligned}\ ] ]
consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .
we denote by @xmath92 an input distribution that achieves the dispersion .
fix @xmath286 .
the encoding procedure is the same as that for the basic streaming setup in section [ subsec : md_pf ] .
let us consider the decoding of @xmath60 at the end of block @xmath99 .
the decoding procedure is modified from that for the basic streaming setup in section [ subsec : md_pf ] as follows : * the decoding test is modified as follows : @xmath115}(\hat{g}_{t_k , [ 1:j-1 ] } , g_{[j : t_k ] } ) , \mathbf{y}_{[j : t_k]})>(t_k - j+1 ) \cdot ( \log m_n+ \gamma n\rho_n ) , \label{eqn : dec_rule_erasure}\end{aligned}\ ] ] i.e. , the threshold value is increased proportional to @xmath344 . *
if there is none or more than one @xmath118 that satisfies the decoding test for some @xmath116}$ ] , the decoder declares an erasure , i.e. , @xmath345 , and terminates the decoding procedure .
similarly as in section [ subsec : md_pf ] , we first consider the probability of error averaged over random codebook @xmath95 .
the error event @xmath123 for @xmath59 happens only if at least one of the following @xmath124 events occurs : @xmath346}(g^{t_k } ) , \mathbf{y}_{[j : t_k]})\leq ( t_k - j+1 ) \cdot ( \log m_n+ \gamma n\rho_n ) \},~ j\in [ 1:k ] \label{eqn : err1_ers}\\ \tilde{\mathcal{e}}_{k , j}'&:= \{i(\mathbf{x}_{[j : t_k]}(g^{j-1},g_{[j : t_k ] } ) , \mathbf{y}_{[j : t_k ] } ) > ( t_k - j+1 ) \cdot ( \log m_n+ \gamma n\rho_n ) \cr & ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ \mbox { for some } g_{[j : t_k ] } \mbox { such that } g_{j } \neq g_{j } \},~ j\in [ 1:k ] .
\label{eqn : err2_ers}\end{aligned}\ ] ] we note that and are obtained by replacing @xmath321 by @xmath347 in and , respectively .
then , we have @xmath126&\leq \sum_{j=1}^k \left(\pr(\mathcal{e}_{k , j}')+\pr(\tilde{\mathcal{e}}_{k , j } ' ) \right).\end{aligned}\ ] ] on the other hand , the undetected error event @xmath348 has the following relationship : @xmath349}\neq g_{[1:k ] } , \hat{g}_k\neq 0\ } \\ & = \cup_{j\in [ 1:k]}\{\hat{g}_{t_k , [ 1:j-1]}= g_{[1:j-1 ] } , \hat{g}_{t_k , j}\neq g_{j } , \hat{g}_k\neq 0\}. \end{aligned}\ ] ] hence , the undetected error probability is bounded as follows : @xmath350&\leq \sum_{j=1}^k \pr(\hat{g}_{t_k , [ 1:j-1]}= g_{[1:j-1 ] } , \hat{g}_{t_k , j}\neq g_{j } , \hat{g}_k\neq 0)\\ & \leq \sum_{j=1}^k \pr(\tilde{\mathcal{e}}_{k , j}').\end{aligned}\ ] ] now , for @xmath127 $ ] , let us bound @xmath351 and @xmath352 .
similarly as in section [ subsec : md_pf ] , @xmath129 s denote i.i.d .
random variables each generated according to @xmath251@xmath258 in the following .
first , we have @xmath353 for sufficiently large @xmath2 , where @xmath134 is some non - negative constant dependent only on the input distribution @xmath135 and channel statistics @xmath136 and @xmath44 is from lemma [ lemma : nonasymptotic_md ] in section [ subsec : md_pf ] .
next , we have @xmath354\\ & \leq m_n^{t_k - j+1}\cdot \exp\left\{-(t_k - j+1 ) \cdot ( \log m_n+\gamma n\rho_n)\right\}\\ & = \exp\left\{-(t_k - j+1)\gamma n\rho_n\right\},\end{aligned}\ ] ] where @xmath44 is obtained by applying a chain of equalities similar to that in appendix [ appendix : chain ] . hence
, we obtain @xmath126&\leq \sum_{j=1}^k \big ( \exp\left\{-(t_k - j+1)n\rho_n^2(1-\gamma)^2\left(\frac{1}{2v}-(1-\gamma ) \rho_n \tau\right)\right\}\cr & \qquad \qquad \qquad + \exp\left\{-(t_k - j+1)n\gamma\rho_n\right\ } \big)\\ & \leq \sum_{j = t}^{t_k } \left ( \exp\left\{-jn\rho_n^2(1-\gamma)^2\left(\frac{1 } { 2v}-(1-\gamma ) \rho_n \tau\right)\right\}+\exp\left\{-jn\gamma \rho_n\right\}\right ) \\ & \leq \frac{\exp\left\{-tn\rho_n^2(1-\gamma)^2\left(\frac{1 } { 2v}-(1-\gamma)\rho_n \tau\right)\right\}}{1-\exp\{-n\rho_n^2(1-\gamma)^2\left(\frac{1 } { 2v}-(1-\gamma ) \rho_n \tau\right)\}}+\frac{\exp\left\{-tn\gamma\rho_n\right\}}{1-\exp\left\{-n\gamma \rho_n\right\ } } \label{eqn : averg_total}\end{aligned}\ ] ] and @xmath350&\leq\frac{\exp\left\{-tn\gamma\rho_n\right\}}{1-\exp\left\{-n\gamma \rho_n\right\ } } \label{eqn : averg_undetec}\end{aligned}\ ] ] for sufficiently large @xmath2 . to show the existence of a deterministic code , we apply markov s inequality as follows : @xmath355 } { n}\right)<\frac{1}{2 } \\ & \pr\left(\limsup_{n\rightarrow \infty } \sum_{k=1}^n\frac{\pr(\hat{g}_k\neq g_k , \hat{g}_k\neq 0|\mathcal{c}_n)}{n}>2\limsup_{n\rightarrow \infty } \sum_{k=1}^n\frac{\e_{\mathcal{c}_n}[\pr(\hat{g}_k\neq g_k , \hat{g}_k\neq 0|\mathcal{c}_n)]}{n } \right)<\frac{1}{2}. \end{aligned}\ ] ] then ,
from the union bound , we have @xmath356 } { n } \mbox { or } \cr & \limsup_{n\rightarrow \infty } \sum_{k=1}^n\frac{\pr(\hat{g}_k\neq g_k , \hat{g}_k\neq 0|\mathcal{c}_n)}{n}>2\limsup_{n\rightarrow \infty } \sum_{k=1}^n\frac{\e_{\mathcal{c}_n}[\pr(\hat{g}_k\neq g_k , \hat{g}_k\neq 0|\mathcal{c}_n)]}{n } \big)<1.\end{aligned}\ ] ] therefore , there must exist a sequence of codes @xmath95 that satisfies @xmath357 and @xmath358 which completes the proof . consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .
we denote by @xmath92 an input distribution that achieves the dispersion .
fix @xmath302 and @xmath286 .
the encoding procedure is the same as that for the basic streaming setup in section [ subsec : md_pf ] .
let us consider the decoding of message @xmath60 at the end of block @xmath359 for @xmath360 . at the end of _
every _ block @xmath359 for @xmath361 . ]
if @xmath362 $ ] , the decoder outputs @xmath363 . for @xmath364 ,
the decoder outputs a message estimate @xmath365 or an erasure symbol @xmath366 according to the same decoding rule illustrated in appendix [ appendix : erasure ] with delay @xmath367 .
then , the error probability of @xmath60 after the random decoding delay @xmath368 ( averaged over the random codebook generation ) is bounded as follows : @xmath369&=\sum_{d = t}^{\infty } \e_{\mathcal{c}_n}\big[\pr(d_k = d , \hat{g}_{k+d-1,k}\neq g_{k } , \hat{g}_{k+d-1,k}\neq 0|\mathcal{c}_n)\big]\\ & \leq \sum_{d = t}^{\infty } \e_{\mathcal{c}_n}\big[\pr(\hat{g}_{k+d-1,k}\neq g_{k } , \hat{g}_{k+d-1,k}\neq 0|\mathcal{c}_n)\big]\\ & \overset{(a)}{\leq } \sum_{d = t}^{\infty}\frac{\exp\left\{-dn\gamma \rho_n\right\}}{1-\exp\left\{-n\gamma \rho_n\right\}}\\ & \leq \frac{\exp\left\{-tn\gamma\rho_n\right\}}{(1-\exp\left\{-n\gamma\rho_n\right\})^2},\end{aligned}\ ] ] where @xmath44 is from the upper bound on the undetected error probability with delay @xmath367 . on the other hand ,
the excess of delay averaged over the random codebook generation is bounded as @xmath370 & = \pr(d_k = t+1|\mathcal{c}_n)+2 \pr(d_k = t+2|\mathcal{c}_n)+\cdots\\ & = \pr(\hat{g}_{k+t-1,k}=0 , \hat{g}_{k+t , k}\neq 0|\mathcal{c}_n)\cr & \qquad \qquad \qquad+2 \pr(\hat{g}_{k+t-1,k}=0 , \hat{g}_{k+t , k}=0 , \hat{g}_{k+t+1,k}\neq 0|\mathcal{c}_n)+\cdots\\ & \leq \sum_{d = t+1}^{\infty } ( d - t)\cdot \pr\left(\hat{g}_{k+d-2,k } = 0|\mathcal{c}_n\right ) \\ & \leq \sum_{d = t+1}^{\infty } ( d - t)\cdot \pr \left(\hat{g}_{k+d-2,k } \neq g_k|\mathcal{c}_n\right)\\ & \overset{(a)}{\leq } \sum_{d = t+1}^{\infty } ( d - t)\cdot \big ( \frac{\exp\left\{-(d-1)n\rho_n^2(1-\gamma)^2\left(\frac{1 } { 2v}-(1-\gamma ) \rho_n \tau\right)\right\}}{1-\exp\{-n\rho_n^2(1-\gamma)^2\left(\frac{1 } { 2v}-(1-\gamma ) \rho_n \tau\right)\}}\cr & \qquad\qquad\qquad\qquad\qquad
+ \frac{\exp\left\{-(d-1)n\gamma\rho_n\right\}}{1-\exp\left\{-n\gamma \rho_n\right\}}\big),\label{eqn : delay_ub}\end{aligned}\ ] ] where @xmath44 is from the upper bound on the total error probability with delay @xmath371 . by following similar statements using markov s inequality in the proof of theorem [ thm : erasure
] , we can obtain @xmath372 } { n}\cr & \qquad\qquad\qquad\qquad \mbox { or } \limsup_{n\rightarrow \infty } \sum_{k=1}^n\frac{\e[d_k|\mathcal{c}_n]}{n}-t>2\limsup_{n\rightarrow \infty } \sum_{k=1}^n\frac{\e_{\mathcal{c}_n}[d_k - t|\mathcal{c}_n]}{n } \big)<1.\end{aligned}\ ] ] therefore , there must exist a sequence of codes @xmath95 that satisfies @xmath373 and tends to infinity . ]
@xmath374}{n}\leq t+o(1).\end{aligned}\ ] ] we note that implies @xmath375 by taking @xmath376 , this completes the proof . consider a dmc @xmath26 with @xmath66 and any sequence of integers @xmath67 such that @xmath68 , where @xmath69 and @xmath70 .
we denote by @xmath92 an input distribution that achieves the dispersion .
fix @xmath377 .
for each @xmath59 and @xmath378 , generate @xmath94 in an i.i.d .
manner according to @xmath92 .
the generated codewords constitute the codebook @xmath95 . in block
@xmath63 , after observing the true message sequence @xmath62 , the encoder sends @xmath96 .
consider the decoding of @xmath60 at the end of block @xmath97 .
similarly as in the decoding procedure in section [ subsec : md_pf ] for the basic streaming setup , the decoder not only decodes @xmath60 , but also re - decodes @xmath98 at the end of block @xmath99 .
let @xmath103 denote the estimate of @xmath104 at the end of block @xmath99 . for alternating message rates ,
we modify the decoding procedure for the basic streaming setup in section [ subsec : md_pf ] according to the following rules and the messages in even block indices have a higher rate of @xmath379 because @xmath377 . ] : * in the basic streaming setup , we consider the window of blocks @xmath380 $ ] for the decoding of @xmath104 for @xmath127 $ ] as shown in .
note that the last block index in each decoding window is @xmath99 . for alternating message rates
, we choose the last block index to be an odd number , to avoid the average message rate in each decoding window of blocks exceeding the capacity .
hence , the last block index in each decoding window is chosen to be @xmath99 , if both @xmath63 and @xmath0 are odd numbers or if both @xmath63 and @xmath0 are even numbers .
otherwise , it is chosen to be @xmath381 . * in the basic streaming setup , we note that the error event related to the smallest decoding window dominates the error probability . for alternating message rates ,
if the smallest decoding window ( after the last block index is chosen to be an odd number ) consists of an odd number of blocks , the average message rate in that window is away from the capacity .
hence , in that case , we adjust the threshold in the decoding test to ensure that the contribution of the smallest decoding window to the error probability is negligible . if @xmath63 is also an odd number , the decoding procedure is the same as that for the basic streaming setup in section [ subsec : md_pf ] , except that when @xmath106 , we perform the following decoding test instead of : @xmath382}(\hat{g}_{t_k , [ 1:k-1 ] } , g_{[k : t_k ] } ) , \mathbf{y}_{[k : t_k ] } ) & > ( t+r-1 ) \cdot \log m_n . \end{aligned}\
] ] if @xmath63 is an even number , the decoding procedure is the same as that for the basic streaming setup in section [ subsec : md_pf ] , except that the last block index @xmath99 in each decoding window is replaced by @xmath383 .
if @xmath63 is an odd number , the decoding procedure is the same as that for the basic streaming setup in section [ subsec : md_pf ] , except that the last block index @xmath99 in each decoding window is replaced by @xmath383 and when @xmath106 , we perform the following decoding test : @xmath384}(\hat{g}_{t_k , [ 1:k-1 ] } , g_{[k : t_k-1 ] } ) , \mathbf{y}_{[k : t_k-1 ] } ) & > ( t+r-2 ) \cdot \log m_n . \end{aligned}\ ] ] if @xmath63 is an even number , the decoding procedure is the same as that for the basic streaming setup in section [ subsec : md_pf ] .
let us first consider the case where both @xmath0 and @xmath63 are odd numbers .
the error event @xmath123 happens only if at least one of the following @xmath124 events occurs : @xmath385}(g^{t_k } ) , \mathbf{y}_{[j : t_k]})\leq ( t_k - j+1 ) \cdot
\log m_n \},~ j\in [ 1:k-1 ] \label{eqn : err1_al}\\ \tilde{\mathcal{e}}_{k ,
j}''&:= \{i(\mathbf{x}_{[j : t_k]}(g^{j-1},g_{[j : t_k ] } ) , \mathbf{y}_{[j : t_k ] } ) > ( t_k - j+1 ) \cdot
\log m_n \cr & ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ \mbox { for some } g_{[j : t_k ] } \mbox { such that } g_{j } \neq g_{j } \},~ j\in [ 1:k-1 ] \label{eqn :
err2_al}\\ \mathcal{e}_{k , k}''&:= \ { i(\mathbf{x}_{[k : t_k]}(g^{t_k } ) , \mathbf{y}_{[k : t_k]})\leq ( t+r-1 ) \cdot \log m_n \}\\ \tilde{\mathcal{e}}_{k , k}''&:= \{i(\mathbf{x}_{[k : t_k]}(g^{k-1},g_{[k : t_k ] } ) , \mathbf{y}_{[k : t_k ] } ) > ( t+r-1 ) \cdot \log m_n \cr & ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ \mbox { for some } g_{[k : t_k ] } \mbox { such that } g_{k } \neq g_{k } \}.\end{aligned}\ ] ] now , we have @xmath126&\leq \sum_{j=1}^{k } \left(\pr(\mathcal{e}_{k , j}'')+\pr(\tilde{\mathcal{e}}_{k , j } '' ) \right).\end{aligned}\ ] ] fix arbitrary @xmath131 . by applying similar steps as in the error analysis for the basic streaming setup in section [ subsec : md_pf ]
, we can show by @xmath386 in the rhs of . ]
@xmath387 for sufficiently large @xmath2 , where @xmath134 is some non - negative constant dependent only on the input distribution @xmath135 and channel statistics @xmath136 .
now , @xmath388 is bounded as follows : @xmath389^+\right\}\right]\\ & \overset{(b)}{\leq } \pr\left(\sum_{l=1}^{nt } i(x_l;y_l)\leq ( t+r-1)n(c-\lambda\rho_n)\right)+\exp\left\{-(t+r-1)n(1-\lambda)\rho_n\right\}\\ & \leq \pr\left(\sum_{l=1}^{nt } i(x_l;y_l)\leq ( t+r-1)nc\right)+\exp\left\{-(t+r-1)n(1-\lambda)\rho_n\right\}\\ & \overset{(c)}{\leq}\exp\left\{-tn\tau'\right\}+\exp\left\{-(t+r-1)n(1-\lambda)\rho_n\right\}\end{aligned}\ ] ] for some constant @xmath390 that depends only on the input distribution @xmath135 and channel statistics @xmath136 , where @xmath129 s are i.i.d .
random variables each generated according to @xmath130 , @xmath44 is from the identity ( * ? ? ?
* eq . ( 69 ) ) , @xmath252 follows from similar steps as - , and @xmath391 is from lemma [ lemma : central ] that is relegated to the end of this appendix .
thus , we obtain @xmath126&\leq\frac{\exp\left\{-(t+1)n\rho_n^2\lambda^2\left(\frac{1 } { 2v}-\lambda \rho_n \tau\right)\right\}}{1-\exp\{-n\rho_n^2\lambda^2\left(\frac{1 } { 2v}-\lambda \rho_n \tau\right)\}}+\frac{\exp\left\{-(t+1)n(1-\lambda)\rho_n\right\}}{1-\exp\left\{-n(1-\lambda)\rho_n\right\}}\cr & \qquad\qquad+\exp\left\{-tn\tau'\right\}+\exp\left\{-(t+r-1)n(1-\lambda)\rho_n\right\}\end{aligned}\ ] ] when both @xmath0 and @xmath63 are odd numbers .
now , let us consider the case where @xmath0 is an odd number and @xmath63 is an even number .
we remind that the decoding procedure is the same as that for the basic streaming setup in section [ subsec : md_pf ] , except that @xmath99 is replaced by @xmath383 . by applying similar steps as in the error analysis for the basic streaming setup in section [ subsec : md_pf ]
, we can show by @xmath392 in the rhs of . ]
\frac{\exp\left\{-(t-1)n\rho_n^2\lambda^2\left(\frac{1 } { 2v}-\lambda \rho_n \tau\right)\right\}}{1-\exp\{-n\rho_n^2\lambda^2\left(\frac{1 } { 2v}-\lambda \rho_n \tau\right)\}}+\frac{\exp\left\{-(t-1)n(1-\lambda)\rho_n\right\}}{1-\exp\left\{-n(1-\lambda)\rho_n\right\ } } \label{eqn : md_sumup_al23}\end{aligned}\ ] ] for sufficiently large @xmath2 . by following similar statements using markov s inequality in the proof of theorem [ thm : erasure
] , we can show that there must exist a sequence of codes @xmath95 that satisfies @xmath393 and @xmath394 which completes the proof for the case where @xmath0 is an odd number , by taking @xmath139 .
the proof for the case where @xmath0 is an even number can be done in a similar manner , and hence it is omitted .
[ lemma : central ] let @xmath140 be a sequence of i.i.d .
random variables with zero mean such that its cumulant generating function @xmath143 $ ] for @xmath144 is continuously differentiable . for @xmath395 ,
the following bound holds : @xmath396 where @xmath397 is the _ rate function _ defined as follows : @xmath398 for any @xmath399 , we obtain @xmath400 by applying the same steps used to obtain in the proof of lemma [ lemma : nonasymptotic_md ] . since @xmath399 is arbitrary ,
we obtain the following bound @xmath401 furthermore , because @xmath402 and @xmath403 , we conclude that @xmath404 .
s. c. draper and a. khisti , `` truncated tree codes for streaming data : infinite - memory reliability using finite memory , '' in _ proc . international symposium on wireless communication systems ( iswcs ) _ , nov .
2011 , pp . 136140 .
v. strassen , `` asymptotische abschtzungen in shannons informationstheorie , '' in _ trans .
third prague conf .
theory _ , prague , 1962 ,
689723 , http://www.math.cornell.edu/@xmath405pmlut/strassen.pdf . v. y. f. tan , _
asymptotic estimates in information theory with non - vanishing error probabilities_.1em plus 0.5em minus 0.4em foundations and trends in communications and information theory , 2015 , vol . 11 , no . 1 - 2
. y. polyanskiy and s. verd , `` channel dispersion and moderate deviations limits for memoryless channels , '' in _ proc . 48th annual allerton conference on communication , control , and computing _ , monticello , il , 2010 .
z. lin , v. y. f. tan , and m. motani , `` on error exponents and moderate deviations for lossless streaming compression of correlated sources , '' _ ieee trans .
inf . theory _ , submitted for publication .
[ online ] .
available : http://arxiv.org/abs/1507.03190 . | we consider streaming data transmission over a discrete memoryless channel . a new message is given to the encoder at the beginning of each block and the decoder
decodes each message sequentially , after a delay of @xmath0 blocks . in this streaming setup ,
we study the fundamental interplay between the rate and error probability in the central limit and moderate deviations regimes and show that i ) in the moderate deviations regime , the moderate deviations constant improves over the block coding or non - streaming setup by a factor of @xmath0 and ii ) in the central limit regime , the second - order coding rate improves by a factor of approximately @xmath1 for a wide range of channel parameters . for both regimes ,
we propose coding techniques that incorporate a joint encoding of fresh and previous messages . in particular ,
for the central limit regime , we propose a coding technique with truncated memory to ensure that a summation of constants , which arises as a result of applications of the central limit theorem , does not diverge in the error analysis .
furthermore , we explore interesting variants of the basic streaming setup in the moderate deviations regime .
we first consider a scenario with an erasure option at the decoder and show that both the exponents of the total error and the undetected error probabilities improve by factors of @xmath0 .
next , by utilizing the erasure option , we show that the exponent of the total error probability can be improved to that of the undetected error probability ( in the order sense ) at the expense of a variable decoding delay .
finally , we also extend our results to the case where the message rate is not fixed but alternates between two values . |
distant ( ) , isolated early - type galaxies have been studied extensively at optical and near - ir wavelengths .
the optical luminosities of early - type galaxies with rest - frame red - sequence colors have faded by @xmath51 magnitude since @xmath8 ( e.g. , bell 2004a ; mcintosh 2005 ; treu 2005 , 2006 ; van der wel 2005 ) . furthermore
, little star - formation activity is observed within these galaxies out to at least @xmath0 ( e.g. , bell 2005 ) indicating that this observed evolution is consistent with the passive aging of old stars .
however , the luminosity density of red - sequence galaxies has been observed to remain approximately constant over ( e.g. , bell 2004a ; faber 2006 ) , and since these galaxies appear to be fading passively with cosmic time , this suggests that an important fraction ( @xmath51/2 ) of the current @xmath9 stellar - mass content of the early - type galaxy population has emerged since @xmath10 .
one explanation for these observations is that relatively young , lower - mass galaxies continuously populate the red - sequence ( e.g. , van der wel 2004 ; thomas 2005 ) through either a downsizing mass - assembly scenario ( e.g. , cowie 1996 ; bundy 2005 ; cimatti 2006 ; de lucia 2006 ; lee 2006 ) and/or a scenario where lower - mass early - type galaxies have different merging histories than their higher - mass counterparts ( e.g. , khochfar & burkert 2003 ) . at @xmath11 ,
distant red galaxies ( drgs ; selected using @xmath12 colors ) and extremely red objects ( eros ; @xmath13 or @xmath14 selected ) , some of which are likely progenitors to early - type galaxies , appear to have large star - formation rates and inferred dynamical masses similar to those observed for massive local ellipticals ( e.g. , frster schreiber 2004 ; mccarthy 2004 ; van dokkum 2004 ; papovich 2006 ) .
these analyses suggest that the majority of the stellar mass within many massive early - type galaxies formed at @xmath15 , possibly during a dust - obscured far - infrared - luminous phase ( e.g. , chapman 2005 ; swinbank 2006 ) .
x - ray investigations of local early - type galaxies have revealed that the total luminosity is correlated with optical luminosity ( e.g. , trinchieri & fabbiano 1985 ; canizares 1987 ; brown & bregman 1998 ; osullivan 2001 ; ellis & osullivan 2006 ) .
emission from local early - type galaxies originates primarily from a combination of hot interstellar gas and low mass binaries ( lmxbs ) , and due to their relatively low star - formation rates ( sfrs ) per unit stellar mass ( typically sfr/@xmath16@xmath17 gyr@xmath18 for @xmath19 0.00.7 galaxies with @xmath20 ; e.g. , feulner 2006 ) , these galaxies are expected to have negligible contributions from high mass binaries ( hmxbs ; e.g. , grimm 2002 ) .
two fairly distinct classes of early - type galaxies , which can be divided by optical luminosity at , have been identified .
early - type galaxies with @xmath21 are referred to as `` optically luminous '' and those with @xmath22 are referred to as `` optically faint . ''
optically luminous early - type galaxies in the local universe are generally observed to have emission powered largely by hot ( ) interstellar gas at soft ( ) energies and lmxbs at hard ( ) energies ( e.g. , matsumoto 1997 ) .
figure 1a illustrates the spectrum of ngc 1600 ( sivakoff 2004 ) , a representative optically luminous early - type galaxy ; the hot - gas and lmxb components have been separated for illustrative purposes .
the luminosities for optically luminous early - type galaxies are correlated with their optical luminosities following @xmath23 .
stellar winds generate the gas and high - velocity interactions and type ia supernovae heat it to temperatures ( e.g. , sarazin 1997 ) .
galaxies with larger stellar masses and deeper gravitational potential wells generally have higher stellar velocity dispersions and thus provide more significant heating of the gas .
therefore , the relation is manifested as a result of proportionalities between the stellar mass in a particular galaxy ( traced by @xmath24 ) and the stellar velocity dispersion of that galaxy ( traced by ; e.g. , mahdavi & geller 2001 ) .
optically faint early - type galaxies in the local universe have soft emission characterized by a power - law ( ) component from lmxbs and a mild contribution from hot interstellar gas ( i.e. , a thermal plasma with @xmath25 kev ) . at hard energies ,
the emission is completely dominated by lmxbs .
the total luminosities of optically faint early - type galaxies are observed to be linearly correlated with their optical luminosities ( i.e. , @xmath26 ) .
this linear correlation is thought to arise largely as a result of the number of lmxbs in a given galaxy being proportional to the stellar mass within that galaxy ( e.g. , gilfanov 2004 ) . due to their low luminosities
, few optically faint early - type galaxies have been studied in detail in the band , and the presently available data for these galaxies are plagued by large scatter due to small numbers of lmxbs per galaxy and varying contributions from hot interstellar gas ( e.g. , david 2006 ) .
therefore , much of what is known about lmxbs in these galaxies comes from studying optically luminous early - type galaxies that have little contamination from hot interstellar gas ( i.e. , faint early - type galaxies ) .
figure 1b shows the spectrum of ngc 4697 ( e.g. , sarazin 2001 ) , an faint ( yet optically luminous ) early - type galaxy , which is expected to be representative of the typical optically faint early - type galaxy spectrum .
the two early - type galaxy classes discussed above ( optically luminous and faint ) generally differ in their relative spectral contributions from hot interstellar gas and lmxbs .
an important component of the emission from early - type galaxies , most notably in optically luminous sources , is the hot interstellar gas .
detailed observations of the gas have shown that it radiates strongly but does not appear to be cooling as expected ( e.g. , xu 2002 ; @xmath27 7.1 of mathews & brighenti 2003 and references therein ; tamura 2003 ; bregman 2005 ) .
the observed gas temperatures and densities of the central regions of optically luminous early - type galaxies ( e.g. , fukazawa 2006 ) imply short radiative cooling times of @xmath5@xmath28 yr on average .
simple cooling - flow models , which include heating from stellar winds and type ia supernovae , vastly overpredict the amount of cooled gas observed in the central regions of these galaxies ( see , e.g. , @xmath27 8 of mathews & brighenti 2003 ) ; this suggests some additional form of heating must be present
. attempts to mitigate this problem by including additional heating components in these models , such as transient heating from active galactic nuclei ( agns ) , have been unsuccessful in reproducing the observed properties of the gas ( e.g. , the radial profiles of electron temperature , density , and metallicity ; see , e.g. , brighenti & mathews 2003 ) . despite these modelling difficulties , both observational and theoretical studies
have suggested that energy feedback from transient agns must be important ( e.g. , binney & tabor 1995 ; ciotti & ostriker 1997 , 2001 ; brzan 2004 ; best 2005 , 2006 ; churazov 2005 ; di matteo 2005 ; scannapieco 2005 ; bower 2006 ; brighenti & mathews 2006 ; croton 2006 ; hopkins 2006 ; mcnamara 2006 ) ; these studies maintain that agn feedback plays a crucial role in regulating the growth of early - type galaxies and their central supermassive black holes .
observations of giant elliptical galaxies in the local universe have revealed significant cavities filled with radio lobes that are inferred to have been driven by nuclear agn outbursts ( e.g. , churazov 2001 , 2002 ; jones 2002 , 2005 ; forman 2005 ; nulsen 2005a , 2005b ; allen 2006 ) ; these observations suggest significant energy transfer and recurrence timescales much shorter than the average cooling time .
one of the goals of this paper is to characterize the evolution of the hot -emitting gas in normal early - type galaxies and to place constraints on the role of feedback by transient agns .
lmxbs provide another important contribution to the emission from early - type galaxies ( both optically luminous and faint ) , and many investigations have been dedicated to understanding their nature . observations in the optical band , most notably from the _ hubble space telescope _
( ) , have shown that many of the lmxbs in early - type galaxies are coincident with globular clusters ( e.g. , sarazin 2003 ) , and the number of lmxbs observed in a given galaxy appears to be linearly correlated with the globular - cluster specific frequency , @xmath29 ( i.e. , the number of globular clusters per unit optical luminosity ; e.g. , white 2002 ; kim & fabbiano 2004 ; irwin 2005 ; juett 2005 ) .
white ( 2002 ) examined several optically luminous , lmxb - dominated early - type galaxies with stellar populations having ages ranging from ( as measured via spectroscopy ) , and found that their luminosity ratios ( @xmath30[lmxb]@xmath31 ) have no dependence on galaxy age ; @xmath30(lmxb)@xmath31 was observed to be linearly proportional to @xmath29 . again using optically luminous early - type galaxies , irwin ( 2005 ) found that details of the relation imply that a significant number of lmxbs were formed in the fields of the early - type galaxies and that the field lmxb contribution to @xmath30(lmxb)@xmath31 is most significant , and even dominates over the globular cluster lmxb emission , for galaxies with smaller values of @xmath29 . in the optically luminous galaxies studied by white ( 2002 ) and
irwin ( 2005 ) @xmath29 is large , and lmxbs are mainly formed in the high - density central regions of globular clusters through tidal interactions between normal stars and either neutron stars or black holes . in optically faint low - mass systems
, however , @xmath29 is typically small ( e.g. , ashman & zepf 1998 ) , and it is therefore likely that lmxbs that formed within the galactic fields of these systems dominate @xmath30(lmxb)@xmath31 ( see , e.g. , fig . 3 of irwin 2005 ) ; these lmxbs are thought to have formed via the evolution of primordial binaries on timescales of following a star - formation event ( e.g. , verbunt & van den heuvel 1995 ) .
furthermore , if the downsizing mass - assembly scenario discussed above is correct , field lmxbs within optically faint galaxies could evolve significantly over cosmic time , as lmxb populations emerge in the wake of relatively recent bursts of star formation and eventually fade after @xmath51 gyr ( e.g. , ghosh & white 2001 ) .
therefore , another goal of this paper is to measure the properties of normal optically faint early - type galaxies over a significant range of redshift to place constraints on the evolution of the lmxb activity in these systems . at present , there have been few investigations of the redshift evolution of the properties of isolated early - type galaxies , which contrasts with the case for late - type galaxies ( e.g. , hornschemeier 2002 ; norman 2004 ; laird 2005 ; kim 2006 ; lehmer 2006 ) . brand ( 2005 )
used stacking analyses ( see also , e.g. , brandt 2001 ; nandra 2002 ; lehmer 2005a ) to investigate the evolution of optically luminous red galaxies within the botes field of the noao deep wide - field survey ( ndwfs ) over the redshift range .
the observations used in these analyses were @xmath55 ks in duration ( e.g. , murray 2005 ) and covered an area of @xmath51.4 deg@xmath32 . brand ( 2005 )
found that the average luminosities of these galaxies mildly increase with redshift , which is primarily due to a rise in agn activity with redshift . due to its relatively large sample size ( 3316 red galaxies ) ,
the brand ( 2005 ) investigation provided useful statistical constraints on the activity from distant powerful agns ; however , the sensitivity of this study was too low ( i.e. , the luminosity detection limit was @xmath33 at @xmath0 ) to distinguish effectively galactic emission ( i.e. , hot interstellar gas and lmxbs ) from that of luminous agns .
deep surveys , which utilize long exposures with or ( see , e.g. , brandt & hasinger 2005 for a review ) , are required to investigate the emission from distant `` normal '' ( i.e. , not predominantly powered by luminous agns ) early - type galaxies .
an ideally suited field for such an investigation is the extended deep field - south ( ) .
the is a @xmath50.3 deg@xmath32 , multiwavelength survey field composed of the central @xmath51 ms deep field - south ( cdf - s ; giacconi 2002 ; alexander 2003 ) and four flanking , contiguous @xmath5250 ks observations ( lehmer 2005b ) .
the sensitivity limits ( for detecting an individual unresolved point source ) over the are @xmath5@xmath34 in the most sensitive regions and @xmath5@xmath35 over the majority of the field ; these levels correspond to luminosity detection limits of @xmath5@xmath36 and @xmath5@xmath37 , respectively , at @xmath38 .
such sensitivity limits are sufficient to detect moderately - powerful agns and normal , luminous early - type galaxies ( e.g. , ngc 1600 ; sivakoff 2004 ) out to @xmath39 .
recently , mcintosh ( 2005 ; hereafter , m05 ) isolated a sample of 728 optically selected early - type galaxies within the from the gems ( galaxy evolution from morphologies and seds ; rix 2004 ) and ( classifying objects by medium - band observations in 17 filters ; wolf 2004 ) surveys , which overlap with @xmath50.23 deg@xmath32 ( @xmath577% ) of the .
the m05 sample spans a redshift range of , and due to its depth and relatively large comoving volume , it is ideal for investigating the evolution of early - type galaxies over the last half of cosmic history . in this paper
, we utilize the m05 sample and a supplementary sample of 64 additional early - type galaxies located within the ( selected using the same techniques discussed in m05 ) to characterize the redshift evolution of the properties of early - type galaxies .
our aim is to measure the average emission , via stacking techniques ( applied to detected and undetected sources ) , from distant normal early - type galaxy populations .
we investigate separately the cosmic - time evolution of ( 1 ) the hot interstellar gas content through optically luminous galaxy samples and ( 2 ) the lmxb populations through optically faint galaxy samples .
the galactic column density is @xmath40 for the ( stark 1992 ) .
all of the fluxes and luminosities quoted throughout this paper have been corrected for galactic absorption . unless stated otherwise , we make reference to optical magnitudes using the vega magnitude system .
@xmath41 = 70 , @xmath42 = 0.3 , and @xmath43 = 0.7 are adopted throughout this paper ( e.g. , spergel 2003 ) , which imply a look - back time of 6.3 gyr at @xmath44 .
we started with the m05 sample of 728 , early - type galaxies and a supplementary sample of 64 galaxies within the field ( i.e. , 792 total galaxies ) , which were selected following the same methods outlined in m05 using the same data .
the total sample was selected using ( 1 ) rest - frame @xmath45 red - sequence colors derived from and ( 2 ) quantitative optical morphologies ( via the srsic indices ; @xmath46 ) as observed from imaging through gems ( see @xmath27 2.3 of m05 for details ) .
the sample generated using these criteria is representative of the early - type galaxy population as a whole and is highly complete down to the selection limit of @xmath47 .
photometric redshifts for these early - type galaxies were provided by and are used throughout this paper . using secure redshifts from various spectroscopic surveys in the region ( e.g. , le fevre 2004 ; szokoly 2004 ; mignoli 2005 ; vanzella 2005 , 2006 ; silverman 2006 ) , which were available for @xmath520% of the m05 sample
, we confirmed that the photometric redshifts for our sample are highly accurate ( i.e. , @xmath550% of the galaxies have @xmath48 < 0.02 $ ] and @xmath575% have @xmath48 < 0.03 $ ] ) .
the selection of early - type galaxies via these quantitative methods isolates galaxies with visual morphologies ranging from galaxies , and a large majority of these are e / s0 galaxies ( bell 2004b ) .
we note that the properties of e and s0 galaxies are observed to be similar ( e.g. , forman 1985 ; blanton 2001 ; xu 2005 ) .
our primary goal was to investigate the redshift evolution of the properties of normal early - type galaxies . to achieve this better
, we filtered our original sample to include 539 galaxies within the redshift range .
this range was chosen to optimize our use of the sensitivity to allow for the detection and identification of low - luminosity agns over a broad range of cosmic look - back times ( @xmath49 gyr ) . as discussed in @xmath27 1.3 , the observations of
the are sufficiently sensitive to detect agns with luminosities over this entire redshift range ( @xmath50 ) . furthermore , given that the luminosities of agn - hosting optically luminous ( @xmath51 ) early - type galaxies in the local universe span @xmath52 ( mean @xmath53 ; osullivan 2001 )
, this limit should adequately prevent powerful agns from dominating stacked photon counts ( assuming the agn activity is not widespread ) .
we utilized multiwavelength observations of the region to obtain a census of the -detected agns and normal galaxies within our sample .
we matched galaxies in our sample to sources in the catalogs presented in alexander ( 2003)1 ms .
] for the @xmath51 ms observations and lehmer ( 2005b)250 ks . ] for the @xmath5250 ks observations . for a successful match
, we required that the optical and centroids be displaced from each other by less than 1.5 @xmath54 the radius of the positional error circles ( confidence ) , which are given in each respective catalog .
we note that investigations of luminous off - nuclear sources , which may lie outside our matching radius , have shown that there is a dearth of such sources within early - type galaxies ( e.g. , irwin 2004 ; lehmer 2006 ) .
the source catalogs were generated using wavdetect ( freeman 2002 ) at false - positive probability thresholds of and @xmath55 for the @xmath51 ms and the @xmath5250 ks , respectively .
however , as demonstrated in @xmath27 3.4.2 of alexander ( 2003 ) and @xmath27 3.3.2 of lehmer ( 2005b ) , legitimate lower significance sources , detected by running wavdetect at a false - positive probability threshold of , can be isolated by matching with relatively bright optical sources . since the sky surface density of @xmath56 ,
@xmath57 early - type galaxies is relatively low ( @xmath50.65 galaxies arcmin@xmath58 ) , we can effectively use this technique to isolate sources within our sample .
we estimate that when using wavdetect at a false - positive probability threshold of , we expect @xmath50.5 spurious sources . in total
, we detected 49 early - type galaxies in at least one of the , , or bandpasses .
we used the following criteria to identify candidate agns within our sample : 1 . _ hard x - ray emission _ :
our best discriminator of obscured ( @xmath59 @xmath40 ) agn activity is the presence of a hard spectrum , which is characterized by a relatively shallow effective spectral slope ( i.e. , @xmath60 ) .
we classified sources that were detected only in the bandpass as potential agn candidates . for sources detected in both the and bandpasses
, we required that the hardness ratio measured using these bandpasses , @xmath61 ( where @xmath62 is the _ observed _ count - rate in each respective bandpass ) , be greater than 0.5 ( corresponding to an effective photon index of @xmath60 ) ; these sources would have spectral slopes shallower than those expected for a pure lmxb population ( e.g. , church & baluciska - church 2001 ; irwin 2003 ) .
we found that all sources with detections were classified as agn candidates .
furthermore , there were a few sources that were detected in only the bandpass .
we argue that since these sources were not detected in the bandpass , our most sensitive bandpass , then there must be a significant contribution from the bandpass such that @xmath63 .
these sources were flagged as agn candidates and removed from our stacking analyses . using criterion 1 , we identified a total of 29 agn candidates . 2 .
_ x - ray to optical flux ratio _ : spectral hardness is a good indicator of obscured agn activity ; however , powerful unobscured agns with steep spectral slopes ( @xmath64 ) could still be missed if this were our only means for identifying candidate agns .
another useful discriminator of agn activity , which aids in the identification of luminous unobscured agn candidates , is the flux ratio , @xmath65 ( e.g. , maccacaro 1988 ; hornschemeier 2000 ; bauer 2004 ) .
we use the criterion @xmath66 ( see @xmath27 4.1.1 of bauer 2004 for justification ) to identify unobscured agn candidates in our sample ; two candidates , which were not identified by criterion 1 above , were found using this criterion .
we note that 16 out of the 29 agns that satisfied criterion 1 ( @xmath555% ) also satisfied criterion 2 .
_ radio to optical flux ratio _ : we used radio maps ( 1.4 ghz ; limiting 1@xmath67 depth of @xmath514@xmath68jy ) from the australia telescope compact array ( atca ; pi : a. koekemoer ; afonso 2006 ) , which cover the entire region , to identify additional potential agns .
we matched the positions of radio - detected sources to those of our early - type galaxies using a 20 matching radius and found 18 matches .
of the 18 radio - detected sources in our sample , we find that three are detected in the bandpass . of the three -detected sources ,
only cxoecdfs was not previously identified as an agn candidate using criteria 1 and 2 ; however , visual inspection of the radio source coincident with cxoecdfs reveals that it has a clear fr ii radio morphology .
we find that the three agn - like -detected radio sources have relatively large flux ratios , @xmath69/\nu f_\nu [ 5000 $ ] @xmath70)$ ] , as expected for agns ( e.g. , kinney 1996 ) .
out of the remaining 15 radio - detected matches , 13 sources that were not detected in the bandpass were found to have uncharacteristically high flux ratios for early - type galaxies , @xmath71/\nu f_\nu [ 5000 $ ] @xmath70 ) > -4 $ ] .
these 13 sources were classified as potential agns and were excluded from our stacking analyses described below .
we note that in addition to agns , star - forming galaxies ( e.g. , edge - on dusty galaxies ) that were misclassified as early - type galaxies may also satisfy this criterion ( see , e.g. , fig . 8 of barger 2006 ) ; therefore , this crierion also helps to guard against the inclusion of such sources . to summarize ,
we have detected 49 early - type galaxies in the band and have identified 32 -detected agn candidates : 29 from criterion 1 , two additional sources from criterion 2 , and one additional source from criterion 3 .
finally , using criterion 3 we identified an additional 13 agn candidates that were not detected in the band .
therefore , we identified a total of 45 agn candidates from our sample of 539 early - type galaxies .
the remaining 17 -detected sources that we do not classify as agn candidates are considered to be normal early - type galaxies .
these normal galaxies are included in our stacking analyses , and as we discuss in @xmath27 2.2 below , including these galaxies does not significantly affect our results . a more detailed discussion of the properties of our -detected sources is presented in @xmath27 3.1 below .
we note that the three criteria discussed above may not be completely sufficient to classify all -detected sources that are truly agns as agn candidates ( see , e.g. , peterson 2006 ) .
such a misclassification is possible for low - luminosity agns that are only detected in the more sensitive bandpass , which have emission too weak for an accurate classification .
furthermore , we also expect there to be agns that lie below the detection threshold that are not identified here .
however , in @xmath27 3.2.2 , we use the agn fraction as a function of luminosity to argue quantitatively that we do not expect misclassified agns ( detected only in the bandpass ) and low - luminosity agns below the detection limit to have a serious impact on our results .
a large majority of the normal galaxies in our sample were not detected individually in the deep observations over the .
we therefore implemented stacking analyses ( see @xmath27 2.2 for a description ) of galaxy samples , which were constructed to cover two optical luminosity ranges in multiple redshift bins .
we first divided the original sample of 539 early - type galaxies by luminosity into optically luminous ( @xmath72 ) and optically faint ( @xmath73 ) samples .
the two luminosity ranges were motivated by the observed physical differences between local early - type galaxies within these two luminosity ranges ( see discussion in @xmath27 1.1 ) and optical completeness limitations of the m05 sample ( i.e. , the @xmath57 completeness limit ) . as mentioned in @xmath27 1.1 and displayed in figure 1 , the optically luminous galaxies have soft spectra dominated by hot interstellar gas , while the optically faint early - type galaxies have spectra powered mainly by lmxbs with a small contribution from hot interstellar gas .
galaxies from both luminosity classes have hard emission dominated by lmxbs .
we divided the optically luminous sample into four redshift bins of equal comoving volume ( , , , and ) , and due to optical completeness limitations , we divided our optically faint sample into two redshift bins ( and ) .
although the subtends a relatively small solid angle , we note that each redshift bin has a radial depth that exceeds 500 mpc , which is a factor of @xmath745 times larger than the extent of the largest structures in the universe ( e.g. , springel 2006 ) .
we created histograms of the galaxy luminosity - distance distributions on scales of @xmath5100 mpc and did not observe any dominating density `` spikes . ''
this suggests that our samples are not being significantly affected by cosmic variance .
furthermore , provided the physical properties of field early - type galaxies are not strongly affected by their environments in a systematic way , we suggest that cosmic variance should not affect our results even if present . in figure 2a , we show rest - frame @xmath75-band luminosity , @xmath24 ( in solar units ; @xmath76 ) , versus redshift for the 539 galaxies ( _ dots _ and _ crosses _ ) that make up our `` general sample '' ; we indicate the divisions of the sample by optical luminosity ( _ shaded regions _ ) and redshift ( _ vertical dotted lines _ ) .
sources highlighted by large circles ( both _ open _ and _ filled _ ) are those that are individually detected in ; the open circles indicate -detected agn candidates , and the filled circles indicate the remaining detected sources , which are classified as normal early - type galaxies .
in addition to the general sample , we have created `` faded samples '' of early - type galaxies .
the faded samples were constructed using the general sample discussed above ; however , we have corrected the rest - frame @xmath75-band luminosities for passive evolution .
these corrections take into account the fact that the optical power output of early - type galaxies has been fading with cosmic time due to stellar evolution ( see @xmath27 1 ) .
for each galaxy , we have computed an evolved , @xmath9 @xmath75-band luminosity ( @xmath77 ) following the best - fit optical - size - dependent solutions for optical fading presented in tables 1 and 2 of m05 .
m05 present eight best - fit relations for @xmath78 ( i.e. , the difference between the absolute @xmath79-band magnitude at redshifts @xmath80 and @xmath9 ) versus @xmath80 , which differ in their evolutionary scenarios ( based upon formation scenarios and the pegase models of fioc & rocca - volmerange 1997 ) and whether or not the fit is constrained to @xmath81 . throughout this paper
we quote results based upon the best - fit solution that assumes @xmath82 and a single - burst evolutionary model with a formation redshift of @xmath83 and a metallicity of ( see @xmath27 3.2 of m05 for details ) .
we converted the @xmath79-band relation to the @xmath75-band following @xmath84 .
additional analyses , using the seven alternative evolutionary scenarios presented in m05 , were also performed , but no material differences were observed in our results .
figure 2b shows @xmath77 versus @xmath80 for our faded sample ; symbols , lines , and shaded regions have the same meaning as in figure 2a . in generating samples for stacking
, we made the same sample divisions ( of both luminosity and redshift ) as we did for the general sample , except we used @xmath77 to discriminate between optically luminous and faint galaxies .
this approach allows us to place constraints on the evolution of the activity from distant galaxies selected from the optical band as they would appear today .
furthermore , based on local estimates of stellar ratios , we can use @xmath77 to estimate the stellar masses of these systems . using a chabrier initial mass function ( chabrier 2003 )
, we estimate that the luminosity ranges @xmath85 and @xmath86 correspond roughly to stellar - mass ranges of @xmath87 and @xmath88 , respectively . in the presentation below , we focus our attention on results drawn from the faded samples since these are expected to be the most physically meaningful . in order to address the fact that the emission from early - type galaxies is dominated by different physical processes in different energy bands
, we performed stacking analyses in three bandpasses : kev ( sb1 ) , 3.1 of alexander ( 2003 ) . ]
( soft band ; sb ) , and ( hard band ; hb ) .
the shaded bars in the panels of figure 1 show the spectral coverage of these bandpasses at the median redshifts of our optically luminous ( @xmath89 ) and faint ( @xmath90 ) samples for the spectral energy distributions ( seds ) of ngc 1600 ( _ top panel _ ) and ngc 4697 ( _ bottom panel _ ) , respectively . for optically luminous galaxies , the sb1 and sb will effectively sample emission dominated by hot interstellar gas , which produces @xmath580% and @xmath570% of the total emission in each respective bandpass .
in contrast , 90% of the hb flux originates from lmxb emission . for optically faint galaxies , sb and hb will effectively sample the emission from lmxbs ; however , the emission observed in sb1 has roughly equal contributions from hot gas and lmxb emission . in our analyses
, we used data products from alexander ( 2003 ) for the @xmath51 ms and lehmer ( 2005b ) for the @xmath5250 ks ( see footnotes 1 and 2 ) .
the data products are publicly available for all energy bands except for sb1 in the @xmath5250 ks regions , which were generated specifically for the analyses here using the same methods described in lehmer ( 2005b ) .
the methodology of our stacking procedure was similar to that outlined in lehmer ( 2005a ) ; for completeness , we describe this procedure below .
we maximized our stacked signal by optimizing our choices of circular stacking aperture radius ( from which we extract photon counts for both sources and their backgrounds ) and inclusion radius ( i.e. , the maximum off - axis angle within which we include sources for stacking ) .
this process is needed because the point spread function ( psf ) increases in size with off - axis angle , which degrades the sensitivity for those sources that are far off - axis . for the optimization process ,
we stacked all early - type galaxies in our sample that were not detected individually in the band . in order to obtain a clean signal , we excluded galaxies that were located @xmath9110 from unrelated sources in the source catalogs and within the extent of extended sources , which are likely associated with galaxy groups or poor clusters ( see @xmath27 3.4 of giacconi 2002 and @xmath27 6 of lehmer 2005b ) .
sources that lie within both the @xmath51 ms and @xmath5250 ks were stacked using both observations , as long as the off - axis angle for each observation was within our chosen inclusion radius .
our optimization procedure was a two - step iterative process that was performed using the sb , the bandpass in which our signal was maximized .
in the first step of the optimization process ( _ step one _ ) we held the inclusion radius ( @xmath92 ) fixed and stacked sources using a variety of circular stacking apertures of constant radii ( i.e. , we did not vary the aperture radius as a function of off - axis angle ) .
we used 15 different circular stacking apertures with radii in the range of to obtain a relation for how the signal - to - noise ratio ( s / n ) varied as a function of aperture size . for a given aperture , we stacked the photon counts and effective exposure times from each galaxy position and summed them to obtain total source - plus - background counts , @xmath93 , and a total exposure time , @xmath94 , respectively .
we estimated total background counts , @xmath75 , using monte carlo simulations and the background maps described in @xmath27 4.2 of alexander ( 2003 ) for the @xmath51 ms and @xmath27 4 of lehmer ( 2005b ) for the @xmath5250 ks .
we shifted our aperture from each galaxy position randomly within a circular region of radiud @xmath525 , extracted background counts from each relevant background map , and summed the counts to obtain an estimate of the total background counts .
this procedure was repeated 10,000 times to obtain an estimate of the local background and its dispersion .
our best estimate of the total background counts , @xmath75 , was approximated by taking the mean background calculated from the 10,000 monte carlo trials .
for each of the 15 circular apertures , we computed the relevant signal - to - noise ratio ( s / n ) using the equation , s / n @xmath95 . after performing the stacking procedure for all 15 different apertures , we identified the aperture radius that produced the maximal s / n . in the second step of the optimization process ( _ step two _ ) , we held the optimized aperture determined from step one fixed , but this time we stacked sources using 15 different inclusion radii ranging from to obtain s / n as a function of inclusion radius . after performing the stacking procedure for all 15 different inclusion radii
, we identified the optimal inclusion radius . for a clean stacking signal
, we expect that s / n will be proportional to the inclusion radius ( i.e. , s / n @xmath96^{1/2 } \propto [ \pi r_{\rm incl}^2]^{1/2 } \propto r_{\rm incl}$ ] ) , and therefore we chose the optimal inclusion radius based on where the relation appears to deviate significantly from linear .
this method helps to guard against the inclusion of contaminating agns in the low - sensitivity regions at large off - axis angles . using the optimal inclusion radius determined with this method , we repeated step one .
this two - step process was run iteratively until a converged solution was obtained .
we found that a stacking aperture of @xmath515 and an inclusion radius of @xmath570 produced the optimal signal ; we note that this choice of optimized @xmath92 is somewhat smaller than that determined in lehmer ( 2005a ) due to differing exposures and the more conservative approach that we have adopted for choosing the optimal @xmath92 .
hereafter , we utilize these values for stacking aperture and inclusion radius in our analyses . in figure 3 ,
we show s / n as a function of inclusion radius ( fig . 3a ) and aperture size ( fig .
3b ) , which were obtained by holding the aperture radius constant at 15 and the inclusion radius constant at 70 , respectively . after excluding sources that were ( 1 ) classified as agn candidates ( via the three criteria outlined in @xmath27 2.1.1 ) , ( 2 ) located at off - axis angles greater than 70 , ( 3 ) within regions of extended emission , and ( 4 ) within 10 of an unrelated -detected sources , we were left with general and faded samples of 276 and 229 galaxies , respectively ; in both samples , 13 galaxies were -detected sources . using these samples , we stacked the properties following the procedure described above .
the sizes of the general and faded samples differ most notably because of the fading of high - redshift optically luminous galaxies out of the selected luminosity range .
we stacked samples both with and without -detected normal galaxies included , and no material differences were observed in our results ; therefore , -detected sources that were classified as normal galaxies were included in our stacking analyses .
figure 4 shows the spatial distribution of the 229 stacked sources in our faded sample over the region .
sources stacked using the @xmath51 ms and @xmath5250 ks observations are shown as circles and diamonds , respectively ; the sources indicated with filled circles were stacked using both observations .
for each stacked sample , we required the s / n be greater than or equal to 3 ( i.e. , @xmath7499.9% confidence ) for a detection . for stacked samples without significant detections ,
3@xmath67 upper - limits were placed on the source counts . using net counts ( i.e. , @xmath97 ) from our stacked samples and adopted seds ( see below ) , we calculated absorption - corrected fluxes and rest - frame luminosities . due to the fact that our 15 radius stacking aperture encircles only a fraction of the psf100% at off - axis angle @xmath98 to @xmath545% at @xmath99 . for the hb , this fraction varies from @xmath580% at @xmath98 to @xmath540% at @xmath99 .
] for sources at relatively large off - axis angle , we calculated aperture corrections @xmath100 for each stacked source @xmath101 .
since we are calculating average counts from the summed emission of many sources of differing backgrounds and exposure times , we used a single , representative exposure - time - weighted aperture correction , @xmath102 .
this factor , which was determined for each stacked sample , was calculated following : @xmath103 where @xmath104 is the effective exposure time for each stacked source .
the average aperture corrections ( @xmath102 ) for our samples were @xmath51.5 and @xmath51.8 for the and bands , respectively .
we note that at the mean redshifts of our samples , the @xmath515 radius aperture corresponds to projected physical radii in the range of . for early - type galaxies in the local universe ,
extended emission originating outside of these radii varies considerably among galaxies and generally represents only a small fraction ( ) of the total flux ( e.g. , fukazawa 2006 ) .
therefore , we do not make additional aperture corrections to account for extended emission .
we estimated mean observed count rates ( @xmath62 ) using the following equation : @xmath105 where @xmath93 , @xmath75 , and @xmath94 are defined above .
to convert count rate to flux , we used the galactic column density given in @xmath27 1.3 and the best - fit spectral energy distribution ( sed ) for ngc 1600 for the optically luminous galaxies ( see fig .
1a ) and a power - law sed with @xmath106 for the optically faint galaxies ( see @xmath27 1.1 for justification ) .
we note that for a range of reasonable sed choices , we find systematic fractional uncertainties in the count - rate to flux conversion of @xmath9110% for sb and hb and @xmath550% for sb1 .
mean rest - frame luminosities @xmath107 were calculated following : @xmath108 where @xmath109 is the luminosity distance , @xmath110 is the mean observed - frame flux , and @xmath111 is a unitless conversion factor , which relates the observed - frame flux to the rest - frame luminosity using our adopted seds . the fractional errors on @xmath112 due to uncertainties in @xmath113 range from @xmath510% to @xmath51% for our optically luminous and samples , respectively . due to
the relatively broad energy ranges our bandpasses encompass and the relatively small spectral shifts observed over the redshifts of sources in our samples , we used observed - frame fluxes to compute rest - frame luminosities . for our adopted seds , we found that values of @xmath111 varied between 0.8 and 1.1 , depending on the bandpass and redshift of the source .
for all bandpasses , sed choice contributes small systematic uncertainties in @xmath111 , which range from depending on redshift .
we note that the mean luminosity is expected to be closely representative of a typical galaxy within the confined optical luminosity ranges used here .
moreover , using the osullivan ( 2001 ) sample of local early - type galaxies , we find that @xmath114 and 1.2 for optically luminous and faint samples respectively . since we are only able to calculate mean quantities via the stacking used here , we utilize mean quantities throughout .
results from our stacking analyses are presented below in @xmath27 3.2 .
in table 1 , we present the properties of the -detected early - type galaxies .
source properties were determined following the methods outlined in alexander ( 2003 ) and lehmer ( 2005b ) for the and , respectively . using the matching criterion discussed in @xmath27 2.1.1 , we matched 39 of our early - type galaxies to sources that were included in the main catalogs of either alexander ( 2003 ) or lehmer ( 2005b ) , and an additional ten matches were identified using sources detected using wavdetect at a false - positive probability threshold of @xmath115 ( i.e. , 49 total detected sources )
. figure 5 shows the @xmath116-band magnitude ( from combo-17 ) versus flux for all 49 detected early - type galaxies in our sample ; normal galaxies and agn candidates are indicated as filled and open circles , respectively . as discussed in @xmath27 2.1.1 , we classified 32 ( @xmath565% ) -detected sources as agn candidates .
we note that the majority of the agn candidates have @xmath117 ; however , a few of these candidates have @xmath118 , including the fr ii source cxoecdfs .
agns with @xmath119 are either significantly obscured or relatively -weak agns . as discussed in @xmath27 1.2
, transient agn feedback may play an important role in heating the hot gas in normal early - type galaxies . in order to understand and constrain the influence of transient agn activity within early - type galaxies , we computed the -luminosity - dependent cumulative agn fraction , @xmath120 ( i.e. , the fraction of early - type galaxies harboring an agn with a luminosity of @xmath121 or greater ) .
we made use of the bandpass because of its ability to probe relatively unattenuated emission in a regime of the spectrum where we expect there to be minimal emission from normal galaxies ( see also @xmath27 2.1.1 for more details ) .
figure 2 illustrates that the majority of the agns in our samples ( _ open circles _ ) originate within optically luminous ( @xmath122 ) early - type galaxies , and therefore when computing @xmath120 , we used only optically luminous galaxies ; we note that the number of agns within our optically faint samples is too low to obtain a respectable constraint on @xmath120 . in order to quantify the redshift evolution of @xmath120 , we split our optically luminous samples into two redshift intervals of roughly equal size ( @xmath19 [ @xmath123 and @xmath124 [ @xmath125 ) .
only two redshift intervals were chosen due to statistical limitations on the number of detected agns .
we computed @xmath120 by taking the number of candidate agns with a luminosity of @xmath121 or greater and dividing it by the number of early - type galaxies in which we could have detected an agn with luminosity @xmath121 .
the latter number was computed by considering the redshift of each galaxy and its corresponding sensitivity limit , as obtained from spatially dependent sensitivity maps ( see @xmath27 4.2 of alexander 2003 and @xmath27 4 of lehmer 2005b ) ; these sensitivity maps were calibrated empirically using sources detected by wavdetect at a false - positive probability threshold of @xmath115 .
figure 6 shows @xmath120 as a function of @xmath121 for the two redshift bins considered here .
we find suggestive evidence for evolution in @xmath120 between @xmath124 0.42 and @xmath19 0.65 , which is consistent with the global trend observed for luminous agns in general ( e.g. , brandt & hasinger 2005 ) . at @xmath126 , where @xmath120 is most tightly constrained , we find that @xmath127 \times f_c(z=0.42)$ ] .
we note that although this value is poorly constrained , it is consistent with the @xmath6 evolution observed for luminosity functions of -selected agns ( e.g. , ueda 2003 ; barger 2005 ; hasinger 2005 ) , and is in agreement with the stacked constraints on optically - selected early - type galaxies set by brand ( 2005 ) .
we return to the discussion of -detected agns in @xmath27 3.2.2 when discussing the undetected agn contribution to our stacked samples and in @xmath27 4.1 when discussing the transient agn contribution to heating the hot interstellar gas in optically luminous early - type galaxies . in tables 2 and 3 , we summarize the average properties of our stacked samples of normal early - type galaxies . for illustrative purposes , we created figure 7 , which shows ( sb ) adaptively - smoothed stacked images of our faded samples .
we detect the average emission from all of our samples in sb , several in sb1 , and only two in hb .
the two samples from which we detect hb emission ( the @xmath128 optically luminous general sample and the @xmath129 optically luminous faded sample ) have and count - rate ratios ( i.e. , @xmath130 and @xmath61 ) that are broadly consistent with our adopted sed .
furthermore , all of our stacked samples have mean flux ratios ( @xmath65 ) that are consistent with those expected for normal galaxies ( see col.[5 ] of table 3 and fig .
5 ) . in figure 5
, we have plotted mean quantities from our samples as large filled squares and triangles , which represent our optically luminous and faint faded samples , respectively ; these symbols have been shaded with varying grayscale levels to indicate redshift , such that darker shading represents higher redshift samples . for the purpose of comparing our results with local early - type galaxies ,
we utilize the osullivan ( 2001 ; hereafter os01 ) sample .
the os01 sample was selected from the lyon - meudon extragalactic data archive ( leda ) using morphological type ( @xmath131 ; hubble types ) , distance ( @xmath132 km s@xmath18 ) , and apparent magnitude ( @xmath133 ) .
the leda catalog is known to be @xmath590% complete down to @xmath134 .
observations of these galaxies were available mainly from the pspc with a significant minority of the data originating from the ipc .
we also utilized data from david ( 2006 ) for six of the os01 galaxies having only upper limits .
figure 8 shows the luminosity versus @xmath75-band luminosity for galaxies included in the os01 sample with @xmath135 kpc ; luminosities have been normalized to the bandpass using the sed adopted in os01 ( i.e. , a mekal plasma sed with solar metallicity and a plasma temperature of 1 kev ) .
we have denoted field , central - cluster , brightest - group , and agn - hosting early - type galaxies as identified from the os01 sample ; several well - studied examples have been highlighted ( m32 , m87 , ngc 1399 , ngc 1600 , ngc 4697 , and ngc 5102 ) for reference .
as noted in @xmath27 1.1 , the and @xmath75-band luminosities of local early - type galaxies are observed to be correlated , and these correlations follow a power law : @xmath136 where @xmath137 and @xmath138 are fitting constants . using the os01 sample , we performed linear - regression analyses for galaxies with @xmath122 and separately to determine @xmath137 and @xmath138 for each luminosity regime . when doing these calculations , we excluded ( 1 ) sources at @xmath139 kpc , ( 2 ) sources with emission that may be significantly influenced by -emitting gas associated with galaxy clusters or groups such as central - cluster and brightest - group galaxies , ( 3 ) agn - hosting galaxies , ( 4 ) ngc 5102 , due to its anomalously low luminosity and evidently recent star - formation activity ( e.g. , os01
; kraft 2005 ; see fig . 8) , and ( 5 ) ngc 4782 , which has an anomalously large @xmath24 that drives the correlation ( i.e. , @xmath140 ; see , e.g. , os01 ) .
we utilized kendall s tau ( kendall 1938 ) to measure the correlation strengths and buckley - james regression ( buckley & james 1979 ; isobe 1986 ) to calculate the best - fit correlation parameters ( i.e. , @xmath137 and @xmath138 ) for each luminosity regime .
these tools were available through the astronomy survival analysis software package ( asurv rev .
1.2 ; isobe & feigelson 1990 ; lavalley 1992 ) .
we found correlation significances of @xmath56.8@xmath67 and @xmath52.8@xmath67 for the @xmath122 and galaxy samples , respectively .
we found that for the optically luminous sample and ( _ solid line _ in fig . 8) , and for the optically faint sample and ( _ dashed line _ in fig .
we note that the relation for optically faint galaxies is poorly constrained with @xmath575% of the galaxies having only upper limits ; however , since these galaxies are thought to be dominated by lmxbs , investigations of the lmxb luminosity per unit @xmath75-band luminosity can offer a consistency check for this relation . using observations of 14 e / s0 galaxies ,
kim & fabbiano ( 2004 ) found that for lmxbs with @xmath141 , @xmath142 using a fixed slope of @xmath143 .
this value is the most tightly constrained relation available at present and is consistent with other investigations of the discrete - source contribution ( e.g. , sarazin 1997 ; os01 and references therein ; gilfanov 2004 ) and our calculated relation for local optically faint early - type galaxies . for reference ,
the expected discrete - source contribution from lmxbs is presented in figure 8 as a shaded region , which represents the dispersion of the relation .
we note that the relation for local optically faint early - type galaxies ( _ dashed line _ ) is consistent with that expected for emission originating strictly from lmxbs . in figure 8 , we have plotted mean quantities from our samples using the same symbols and symbol - shading schemes that were adopted in figure 5 .
the plotted error bars for our stacked samples were computed by propagating ( 1 ) poisson errors on the source counts ( gehrels 1986 ) , ( 2 ) 1 @xmath67 errors on @xmath144 , and ( 3 ) systematic errors on the sed - dependent parameters ( i.e. , count - rate to flux conversion and @xmath111 ) ; these errors were propagated following the `` numerical method '' described in @xmath271.7.3 of lyons ( 1991 ) . in an initial evaluation of these results , we find that the average properties of our optically luminous samples appear to follow the local relation at all redshifts .
in contrast , our optically faint samples deviate from the local relation significantly ( 1.5@xmath67 at @xmath145 ) over the redshift range , suggesting there may be some evolution in the lmxb populations within these galaxies .
figures 9a and 9b show the mean luminosities ( ) of our stacked general and faded samples , respectively ; optically luminous and faint samples are indicated as dark filled squares and triangles , respectively .
-detected sources are shown as circles , which represent both normal galaxies ( _ filled circles _ ) and agn candidates ( _ open circles _ ) .
we have included the corresponding mean luminosities for local early - type galaxies from the os01 optically luminous and faint samples in matched optical - luminosity ranges ; as before , these samples were filtered to exclude galaxies classified as agns , central - cluster galaxies , brightest - group galaxies , ngc 5102 , and ngc 4782 . mean optical and luminosities for the os01 samples were computed using the kaplan - meier estimator ( e.g. , feigelson & nelson 1986 ) within asurv , which appropriately handles censored data .
when calculating these mean luminosities , we filtered the os01 sample to include only sources with distances @xmath146 mpc and @xmath147 mpc for the optically luminous and faint samples , respectively ; these distances represent approximate completeness limits for the optical - luminosity ranges used here .
we calculated mean luminosities of and @xmath5@xmath148 for the optically luminous and faint samples , respectively ; error bars here represent 1@xmath67 errors on the means ( computed with asurv ) .
these calculations were made using 102 optically luminous galaxies ( 60 upper limits ) and 48 optically faint galaxies ( 36 upper limits ) .
we attempted to improve the -detection fractions of these samples by using subsamples of galaxies that were created with distance limits smaller than the completeness limits quoted above ; however , we found no improvement in these fractions .
figures 10a and 10b show the mean luminosity ratio ( @xmath149 ) for our general and faded samples , respectively ( symbols have the same meaning as they did in fig .
9 ) ; the expected local discrete - source contribution and its dispersion are shown as a horizontal dotted line and the surrounding shaded region , respectively .
as observed in figure 8 , we find little evidence for evolution in our optically luminous samples ( both general and faded ) .
using the available data , we find that @xmath150 $ ] and @xmath151 \times ( l_{\rm x}/l_{b})_{z=0}$ ] for our optically luminous general and faded samples , respectively . in order to constrain further the allowed redshift evolution of @xmath152
, we utilized the @xmath153 statistic and a simple evolutionary model , @xmath154 .
for this single - parameter model , we constrained @xmath155 using 90% confidence errors ( @xmath156 ) .
we found best - fit parameters of @xmath157 ( @xmath158 for 4 degrees of freedom ) and @xmath159 ( @xmath160 for 4 degrees of freedom ) for our optically luminous general and faded samples , respectively .
for the optically faint early - type samples , we observe suggestive redshift evolution in @xmath152 , and by @xmath161 it has increased above the local relation by a factor of @xmath5@xmath162 and @xmath5@xmath163 for our general and faded samples , respectively . using the @xmath153 statistic and the same simple model for redshift evolution presented above , we find best - fit values of @xmath164 ( @xmath165 for 2 degrees of freedom ) and @xmath166 ( @xmath167 for 2 degrees of freedom ) for our optically faint general and faded samples , respectively .
we note that the evolution observed for optically faint early - type galaxies is largely driven by the value of @xmath152 at @xmath9 . due to the fact that the @xmath9 value for @xmath152 is based on 48 sources , with 36 ( 75% ) having only upper limits
, we can not rule out the possibility that @xmath152 at @xmath9 is significantly affected by systematic errors .
moreover , the total emission from optically faint galaxies is expected to vary significantly between galaxies due to low numbers of lmxbs and variable amounts of hot gas , and therefore large fractional errors are expected for @xmath152 .
we therefore consider this result to be only marginal at present .
as noted above , the local relation between and @xmath24 is nonlinear for optically luminous early - type galaxies . in order to investigate whether such nonlinearities have an effect on our overall results , we created figure 11 , which illustrates residuals to the local best - fit relations , @xmath168 , for optically luminous ( @xmath169 ; @xmath170 ) and faint ( @xmath171 ; @xmath172 ) samples .
figure 11 shows that the nonlinearities observed in the local relation do not affect our conclusions drawn from using @xmath152 as a proxy for evolution .
furthermore , due to the approximate equality of the mean values of @xmath24 for all samples of a given luminosity class , quantitative analyses that account for nonlinearities in the relations yield roughly identical results to those quoted using simply . in this section
we assess whether the stacked properties presented above suffer from contamination by agns with luminosities below our detection threshold . in @xmath27 3.1
( see also fig .
6 ) , we presented the cumulative agn fraction , @xmath120 , the fraction of early - type galaxies harboring an agn with a luminosity of @xmath121 or greater .
to first order , we can use the functional form of @xmath120 to generate a census of the agn population that we expect to be missing due to sensitivity limitations . as noted in @xmath27 3.1
, there is evidence that @xmath120 evolves with redshift .
we modelled this redshift evolution of @xmath120 using the functional form , @xmath173 ( see @xmath27 3.1 for justification ) .
we also assumed that the dependence of @xmath120 upon @xmath121 was similar at all redshifts over the range @xmath19 . using our faded sample of optically luminous early - type galaxies we computed @xmath174 , the average cumulative agn fraction over the redshift range ( @xmath175 ) , following the procedure outlined in @xmath27 3.1 ; this was done to obtain a better understanding of the overall shape of the @xmath176 curve .
figure 12a ( _ filled circles with error bars _ ) shows our computed values of @xmath177 as a function of @xmath178 .
we fit the @xmath177 data points using a quadratic relation ( i.e. , @xmath179 , where @xmath180 , @xmath181 , and @xmath182 ; _ thick solid curve _ in fig .
12a ) over the luminosity range @xmath183 ; this covers the same luminosity range for agns as the os01 local sample . using our best - fit relation for @xmath177 and our model for the redshift evolution of @xmath120 , we calculated @xmath184 for each of our optically luminous faded samples following : @xmath185 where @xmath80 is the median redshift of each sample , @xmath186 is the median redshift of our best - fit redshift - averaged relation , @xmath177 . in figure 12a
, we show our estimates of @xmath120 for each of our optically luminous faded samples ; these curves are annotated on the figure . for comparison ,
we have shown the agn fraction measured for drgs ( see @xmath27 1 ) at @xmath187 by rubin ( 2004 ; _ open diamond _ ) and have extrapolated our model for @xmath120 out to @xmath188 ( _ dot - dashed curve _ ) .
we note that even at @xmath187 our model agrees reasonably well with observations . in order to estimate the amount of agn contamination that may be contributing to our stacked signals , it is desirable to convert the cumulative agn fractions to differential forms ( i.e. , the fraction of galaxies harboring agns within discrete luminosity bins ) . using the model for @xmath120
presented above , we estimated the fractions of galaxies harboring an agn within luminosity bins of width @xmath189 for each of our optically luminous faded samples ; we refer to these as differential agn fractions , @xmath190 .
the histograms in figure 12b show our estimates of @xmath190 .
due to the deep coverage in the region , a large fraction of the luminous agns ( @xmath191 ) would have been removed from our samples before stacking ( see @xmath27 2.1.1 ) .
we are therefore only interested in the fraction of galaxies falling below our sensitivity limit . using the sensitivity maps described in @xmath27 3.1 ,
we determined the fraction of optically luminous galaxies within each of our stacked faded samples for which we could _ not _ have detected an agn of luminosity @xmath121 if present ; we refer to these fractions as @xmath192 ( i.e. , the fractions of galaxies below our sensitivity limit ) and show them in figure 12c . for each of our optically luminous faded samples we calculated the fraction of sources that harbor an undetected agn with luminosity @xmath121 ( in bins of width @xmath189 ) , @xmath193 , by multiplying @xmath190 by @xmath192 . in figure 12d
, we show @xmath193 as a function of @xmath194 . using @xmath193 , we computed the approximate agn contamination with the following summation : @xmath195 where the summation is over all bins of @xmath196 .
we find @xmath197 for our samples .
for our samples with @xmath198 , agns with @xmath199 contribute @xmath7470% of the total @xmath200 estimate .
furthermore , when extrapolating our model for @xmath120 down to much lower values of @xmath121 , we find no significant difference in our estimates of @xmath201 .
we note that this result is mildly dependent on our extrapolation of @xmath120 to values of @xmath202 . if there exists a large population of radiatively - inefficient low - luminosity agns in early - type galaxies that radiate within this luminosity range ( e.g. , advection - dominated accretion flows [ adafs ] ) , then we may be underestimating @xmath201 .
however , observations of @xmath550 early - type galaxies in the local universe have revealed that the majority of the central supermassive black holes in these galaxies are radiating at extremely low efficiencies and typically have observed @xmath203 ( e.g. , loewenstein 2001 ; david 2005 ; pellegrini 2005 ) . therefore a change in the shape of @xmath120 at @xmath204 is not expected .
we converted @xmath201 to a luminosity assuming a power - law model with an effective photon index @xmath205 , which was determined by stacking all 24 agn candidates in our optically luminous faded sample ; in this calculation we purposely made no attempt to correct for intrinsic absorption .
for these 24 agn candidates , we find that the mean band ratio @xmath206 , which corresponds to an effective photon index of @xmath207 ; this effective photon index is in agreement with the effective photon indices measured for sources in the deep field - north (; see fig . 14 of alexander 2003 ) with similar fluxes .
if we assume @xmath208 is a characteristic effective photon index for the agns we expect to be missing , then we find that agn contamination can account for @xmath54% and @xmath511% of the observed emission from our optically luminous faded samples at @xmath209 and @xmath210 , respectively ; this amount does not significantly affect the results presented in @xmath27 3.2.1 above .
we note that the sed used in this calculation has an important effect on the overall estimate of the agn contamination .
since our estimate for contamination decreases as @xmath211 decreases , the amount of contamination in our samples may be affected if our choice of @xmath205 is too flat . however , if we choose a steeper effective photon index such as @xmath212 ( the observed spectrum of the background ) , we still find that the estimated agn contribution to our signal is too low ( @xmath9125% at @xmath213 ) to make a substantive difference to our results .
furthermore , as discussed in @xmath27 2.1.1 , we have taken additional precautions to eliminate several agn candidates that were not detected in the bandpass ; these sources would not be taken into account in this estimate for agn contamination .
we also note that similar results are found when performing the above analyses using our optically luminous general samples . due to poor statistical constraints on agn activity in optically faint early - type galaxies , an estimate of @xmath120
could not be determined reliably using the present data .
studies of the agn fraction as a function of galactic stellar mass have found that agns are much more commonly observed in massive galaxies than lower - mass galaxies ( e.g. , kauffmann 2003 ) .
we therefore approximated a strict upper limit to the agn contribution to the emission from our optically faint early - type samples by using the same model for @xmath120 presented above for the more massive optically luminous galaxies . using this model
, we estimate that agns contribute @xmath214 20% and @xmath214 40% of the emission from our optically faint samples ( both general and faded ) assuming @xmath215 and @xmath212 , respectively .
these limits slightly reduce the significance of the quoted evolution for our optically faint faded samples such that @xmath216 is estimated to be [ @xmath217 and @xmath218 \times ( l_{\rm x}/l_{b})_{z=0}$ ] for @xmath215 and @xmath212 , respectively ; however , we note that these limits should be regarded as very conservative .
the above results suggest differing evolutionary histories for optically luminous and faint early - type galaxies . as shown in @xmath27 3.2.2 , our results are not expected to be significantly affected by an undetected population of agns .
therefore , changes in the emission with redshift are likely the result of global changes in the emission from hot interstellar gas and/or lmxbs . in the sections below ,
we discuss possible interpretations of the results for our optically luminous and faint early - type galaxies in turn . as discussed in @xmath27 2.2 , we chose the sb for our stacking analyses to sample directly the emission from hot interstellar gas .
therefore , the near constancy of @xmath152 with redshift can be largely explained as a general balance between the energy losses from the hot gas ( @xmath219 ) and the energy deposition from heating mechanisms ( @xmath220 ) over each cooling time , @xmath221 .
here we investigate the relative contributions from feedback mechanisms to constrain physical models of the heating of the hot gas in early - type galaxies .
as noted in @xmath27 1.2 , the typical inferred radiative cooling time for the central regions of an optically luminous early - type galaxy is @xmath222 yr .
since each of our optically luminous early - type galaxy redshift bins are larger than the cooling timescale ( i.e. , our redshift bins have temporal widths in the range of ) , we can estimate the redshift - dependent energy components following : @xmath223 @xmath224 here , @xmath225 is the redshift - dependent globally - averaged power output from the gaseous component of our optically luminous early - type galaxies .
the first two terms of equation 7b represent agn heating from both radiative and mechanical feedback power , respectively .
the radiative feedback power is represented as the product of the average agn luminosity per galaxy ( @xmath226 ) , its bolometric correction factor ( @xmath227 ; e.g. , marconi 2004 ; barger 2005 ) , and the efficiency factor describing the coupling between radiation and the hot interstellar gas ( @xmath228 ) .
we note that in cases where the compton temperature of the agn sed falls below the temperature of the hot interstellar gas , radiation from the central agn may effectively cool the gas and thereby drive @xmath228 to negative values ( see , e.g. , @xmath27 6 of nulsen & fabian 2000 ; ciotti & ostriker 2001 ) .
mechanical power ( e.g. , through agn jets ) is likely to be a very important feedback mechanism , and we have indicated its contribution as @xmath229 . finally , @xmath230 represents additional energy input from alternative forms of heating over each cooling time ( see below ) . using the hot - gas component of the sed for ngc 1600
( see _ dotted curve _ in fig .
1a ) , we estimate @xmath231 ; for our optically luminous samples , this amounts to a mean value of @xmath232 .
we measured @xmath226 using our redshift - dependent model for the differential fraction , @xmath190 , which was presented in @xmath27 3.2.2 ( see also fig .
duty cycles for agn activity in any given galaxy are expected to be shorter than the timescales represented by each of our redshift bins ; however , since we are considering large populations of early - type galaxies , we do not expect significant variations in the agn fraction ( as measured from the `` snapshot '' ) at any given time within each redshift bin . therefore using this model
, we can calculate @xmath226 following : @xmath233 where the summation is over bins of @xmath234 and covers the luminosity range @xmath235 .
we found @xmath236 and @xmath521 @xmath237 per galaxy at @xmath209 and @xmath213 , respectively . in figure 13 , we show @xmath226 as a function of redshift for our optically luminous faded samples ( _ filled squares _ ) .
for comparison , we also show stacking results from the brand ( 2005 ) samples of early - type galaxies ( _ open diamonds _ ) , which have mean @xmath116-band absolute magnitudes that are well matched to those of our optically luminous faded samples ; these mean luminosities are dominated by emission from agns and therefore provide a good estimate of @xmath238 ( @xmath58090% of the emission is from agns ) . using these data and our model for the evolution of the emission from transient agns , @xmath239
, we found @xmath240 ( see the _ dotted curve _ in fig .
13 ) . using the above relations and the assumption that @xmath241 , we arrive at the following relation : @xmath242 as discussed in @xmath27 1.2 , transient agn feedback is expected to play a significant role in the heating of the hot interstellar gas .
if we assume that agn feedback is largely responsible for keeping the gas hot , then we can neglect the last term of equation 9 ( i.e. , @xmath243 ) .
with this assumption and the observed constancy of @xmath244 with redshift , we infer that the strongly - evolving radiative power must be poorly coupled to the hot interstellar gas , such that @xmath245 .
this suggests that mechanical agn power ( i.e. , @xmath246 ) dominates the feedback over the redshift range @xmath19 and does not evolve in the same way as the radiative power .
we note that it is also possible that @xmath247 may play some non - negligible role in the heating of the hot gas .
additional heating sources may include inward thermal conduction from the large reservoirs of hot gas found in the outer regions of early - type galaxies ( e.g. , narayan & medvedev 2001 ; brighenti & mathews 2003 ) , type ia supernovae and stellar winds ( e.g. , loewenstein & mathews 1987 , 1991 ) , and infalling circumgalactic gas ( e.g. , brighenti & mathews 1998 ) ; however , the influence of these heating sources is presently not well constrained . for our optically faint early - type galaxy samples
, we found suggestive evidence for redshift evolution in @xmath7 over the redshift range @xmath124 0.00.5 ( see @xmath27 3.2.1 and figs .
10 and 11 ) .
although some of the observed emission may be due to agn activity ( @xmath27 3.2.2 ) , there remains suggestive evidence that normal activity is evolving with redshift , and we discuss possible scenarios explaining this evolution below .
one possible driver of @xmath7 evolution may come from global changes in the lmxb populations within optically faint early - type galaxies . as discussed in @xmath27 1.2 , lmxbs from primordial binaries within the galactic field
are expected to dominate the overall lmxb emission from optically faint early - type galaxies ; this differs from the lmxb emission from optically luminous early - type galaxies , which originates primarily from globular clusters .
lmxbs from primordial binaries emerge in the wake of star - formation epochs @xmath5110 gyr following a major star - formation event .
therefore , changes in the mean early - type galaxy stellar age with redshift should result in observed changes in the mean emission from these systems .
furthermore , these changes are expected to be most conspicuous @xmath51 gyr after major star - formation events ( e.g. , white & ghosh 1998 ; ghosh & white 2001 ) .
galaxy formation scenarios that favor a more recent emergence of the optically faint early - type galaxy population onto the red - sequence such as downsizing or mass - dependent merging histories ( see discussion and references in @xmath27 1 ) would predict significant evolution of the lmxb emission from these systems . a second source of evolution of the emission from optically faint early - type galaxies
could in principle come from cooling of the hot -emitting gas .
however , studies of optically faint early - type galaxies in the local universe have shown that the hot gas emission generally makes up a minority fraction ( typically @xmath540% ) of the total emission ( e.g. , david 2006 ) and is therefore less likely to be completely responsible for the observed evolution than lmxbs .
understanding the evolution of the properties of early - type galaxies could be greatly improved by ( 1 ) constraining better the properties of local optically faint early - type galaxies and making a census of their agn populations , ( 2 ) performing additional investigations using other deep fields that have complementary coverage , and ( 3 ) conducting deeper observations using or future missions .
these possibilities are discussed in more detail below . in figure 8 , we showed the values of and @xmath24 for local early - type galaxies from the os01 sample , which is the largest , uniformly - selected sample available for studying the correlations of local early - type galaxies . the majority ( @xmath575% ) of the isolated optically faint early - type galaxies have only upper limits , which has restricted our interpretation of the redshift evolution of these galaxies .
new observations of the galaxies having only upper limits could not only improve the characterization of the @xmath24 correlation at lower @xmath24 , but could also provide useful insight into the role of hot interstellar gas and lmxbs within the galactic field , which are presently not well constrained .
furthermore , stacking analyses of well - chosen samples of these galaxies could provide useful statistical insight into the mean properties of these galaxies and mitigate the effects of poor source statistics for individual galaxies . stacking analyses using additional samples of early - type galaxies
could effectively reduce the sizes of the errors on mean quantities and/or allow for the analyses of samples in more finely partitioned bins of redshift and optical luminosity .
an important requirement for such studies is to obtain an adequate census of the underlying agn population , which may significantly influence the stacking results . in order to remove effectively agns from stacked signals at @xmath50 , relatively deep observations are necessary .
using the luminosity - dependent agn fractions determined in @xmath27 3.2.2 , we suggest that agns will provide significant contamination for exposures of @xmath91100 ks .
therefore , studies of distant early - type galaxies using multiwavelength data ( including coverage ) from already existing fields such as the @xmath52 ms or the @xmath5200 ks all - wavelength extended groth strip international survey ( aegis ; nandra 2005 ; davis 2006 ) would improve the present situation .
finally , deeper observations of already existing fields ( most notably the ) would provide an improved census of the low - to - moderate luminosity agn population at higher redshifts and connect the properties of these relatively passive early - type populations with those of their higher redshift progenitors ( e.g. , drgs , eros , and distant submillimeter - emitting galaxies ; e.g. , alexander 2005 ) .
in addition to the improvement that additional observations could provide , it is also worth noting that future missions such as and should allow the first investigations of the evolution of the normal early - type galaxy luminosity function with redshift .
using stacking analyses , we have investigated the evolution of 539 , @xmath19 0.10.7 early - type galaxies located in the .
these galaxies were selected using a combination of red - sequence colors and srsic indices as a part of the combo-17 and gems surveys ( m05 ) .
we classified our original early - type galaxy sample as the `` general sample '' and generated an additional `` faded sample , '' which was corrected for the passive fading of old stellar populations . using these samples , we analyzed separately optically luminous ( @xmath2 ) and faint ( @xmath3 ) populations , which are expected to have soft spectra dominated by hot interstellar gas and lmxbs , respectively .
our primary goal was to use stacking analyses to measure and constrain the redshift evolution of the average emission from normal early - type galaxies . to achieve this
, we used a variety of techniques to identify powerful agns , which we removed from our stacking analyses .
our key results are as follows : 1 .
we detected 49 early - type galaxies in the band and classified 32 of these as agn candidates based on their , optical , and radio properties ( see @xmath27 2.1.1 for details ) ; the remaining 17 -detected sources had multiwavelength properties consistent with normal galaxies .
in addition to the 32 -detected agn candidates , we identified 13 galaxies with agn - like radio to optical flux ratios , which we characterized as potential agns .
we found that the majority of the agn candidates were coincident with optically luminous early - type hosts .
the inferred agn fraction for our optically luminous galaxies shows evidence for evolution with redshift in a manner consistent with the @xmath6 evolution expected from other studies of agn evolution .
2 . when stacking the counts from our normal optically luminous early - type galaxy samples , we found that the to optical mean luminosity ratio , @xmath7 , stays roughly constant over the redshift range , which indicates that the -emitting gas has not significantly evolved over the last @xmath56.3 gyr ( i.e. , since @xmath248 ) . using the data available
, we found that @xmath249 = @xmath250 $ ] and @xmath151 \times ( l_{\rm x}/l_{b})_{z=0}$ ] for our general and faded samples , respectively .
we interpret the lack of evolution of optically luminous early - type galaxies to be due to an energy balance between the heating and cooling of the hot gas over each cooling time .
when assuming that the heating is largely due to transient agn activity , we found that mechanical feedback dominates the heating out to @xmath0 versus radiative power , which we inferred to be very poorly coupled to the gas .
furthermore , this result suggests that the radiative and mechanical agn power evolve differently with cosmic time .
3 . for our optically faint early - type galaxy samples
, we found suggestive evidence that @xmath7 increases with redshift .
by @xmath19 0.5 , @xmath7 is measured to be @xmath162 and @xmath163 times larger than that measured at @xmath251 for our general and faded samples , respectively ; however , due to poor statistical constraints on the local relation and the undetected agn population , we could not confidently rule out the null hypothesis .
we hypothesized that evolution of the optically faint early - type galaxy emission may be due to the evolution of lmxbs in galaxies that have recently joined the red - sequence and/or the cooling of hot gas within these galaxies .
we thank ann hornschemeier , paul nulsen , ewan osullivan , craig sarazin , ohad shemmer , john silverman , and the anonymous referee for useful suggestions , which have improved the quality of this paper .
we gratefully acknowledge the financial support of nsf career award ast-9983783 ( b.d.l .
, w.n.b . ) , x - ray center grant g04 - 5157a ( b.d.l .
, w.n.b . , a.t.s . ) , the royal society ( d.m.a . ) , the emmy noether program of the deutsche forschungsgemeinscaft ( e.f.b . ) , the fellowship program ( f.e.b . ) , and nsf grant ast 03 - 07582 ( d.p.s . ) .
[email protected] & 0.68 & @xmath25215.5 & @xmath25214.7 & 1.26 & @xmath2520.94 & 0.7 & 11.1 & 42.3 & e - cdf - s 03 & a + [email protected] & 0.15 & @xmath25215.4 & @xmath214 @xmath25214.9 & @xmath214 0.78 & @xmath2522.63 & 1.4 & 10.4 & @xmath214 40.8 & e - cdf - s 03 & n + [email protected] & 0.22 & @xmath214 @xmath25215.5 & @xmath25214.6 & @xmath253 1.46 & @xmath2521.93 & 1.4 & 10.4 & 41.4 & e - cdf - s 02 & a + [email protected] & 0.49 & @xmath214 @xmath25215.6 & @xmath214 @xmath25214.9 & @xmath41 & @xmath2521.04 & 1.4 & 10.2 & @xmath214 42.0 & e - cdf - s 03 & a + [email protected] & 0.47 & @xmath25215.8 & @xmath25214.6 & 2.31 & @xmath2520.77 & 0.1 & 10.3 & 42.0 & e - cdf - s 02 & a + + [email protected] & 0.68 & @xmath25215.3 & @xmath25214.9 & 0.57 & @xmath2520.31 & 1.2 & 10.3 & 42.3 & cdf - s & a , r + [email protected] & 0.58 & @xmath25215.2 & @xmath25214.9 & 0.51 & @xmath2521.11 & 1.5 & 10.8 & 42.1 & e - cdf - s 02 & a + [email protected] & 0.58 & @xmath25215.3 & @xmath25214.8 & 0.66 & @xmath2520.83 & 1.4 & 10.6 & 42.2 & e - cdf - s 02 & a + [email protected] & 0.63 & @xmath25215.9 & @xmath25215.6 & 0.54 & @xmath2521.13 & 1.4 & 10.4 & 41.5 & cdf - s & a + [email protected] & 0.43 & @xmath214 @xmath25215.6 & @xmath25214.9 & @xmath253 1.07 & @xmath2520.67 & 1.4 & 9.6 & 41.8 & e - cdf - s 03 & a + + [email protected] & 0.61 & @xmath214 @xmath25215.6 & @xmath25214.8 & @xmath253 1.49 & @xmath2520.54 & 1.4 & 10.0 & 42.2 & e - cdf - s 03 & a + [email protected] & 0.59 & @xmath25215.1 & @xmath25213.7 & 3.74 & 0.01 & @xmath2520.3 & 10.8 & 43.0 & e - cdf - s 02 & a + [email protected] & 0.66 & @xmath25216.2 & @xmath214 @xmath25215.0 & @xmath214 3.18 & @xmath214 @xmath2521.32 & 1.4 & 10.9 & @xmath214 42.1 & e - cdf - s 03 & n , s + [email protected] & 0.66 & @xmath25215.9 & @xmath214 @xmath25215.3 & @xmath214 0.67 & @xmath2521.35 & 1.4 & 10.7 & @xmath214 41.8 & cdf - s & n + [email protected] & 0.23 & @xmath25215.8 & @xmath214 @xmath25214.9 & @xmath214 1.85 & @xmath214 @xmath2521.99 & 1.4 & 10.2 & @xmath214 41.2 & e - cdf - s 02 & n + + [email protected] & 0.57 & @xmath25215.5 & @xmath214 @xmath25214.9 & @xmath214 1.07 & @xmath2520.46 & 1.4 & 9.8 & @xmath214 42.1 & e - cdf - s 02 & a + [email protected] & 0.34 & @xmath25216.3 & @xmath25215.3 & 1.46 & @xmath2521.95 & 0.4 & 10.4 & 41.1 & cdf - s & a + [email protected] & 0.47 & @xmath25216.4 & @xmath214 @xmath25215.6 & @xmath214 1.50 & @xmath214 @xmath2522.06 & 1.4 & 10.4 & @xmath214 41.2 & cdf - s & n , s + [email protected] & 0.69 & @xmath214 @xmath25216.2 & @xmath214 @xmath25215.4 & @xmath41 & @xmath2521.66 & 1.4 & 10.6 & @xmath214 41.8 & cdf - s & a , s + [email protected] & 0.60 & @xmath25216.2 & @xmath25214.1 & 11.80 & 0.34 & @xmath2521.4 & 10.1 & 42.4 & cdf - s & a + + [email protected] & 0.64 & @xmath214 @xmath25216.3 & @xmath25215.5 & @xmath253 1.33 & @xmath2521.56 & 1.4 & 10.6 & 41.6 & cdf - s & a + [email protected] & 0.48 & @xmath25214.7 & @xmath25214.3 & 0.60 & @xmath2520.43 & 1.3 & 10.5 & 42.5 & e - cdf - s 02 & a + [email protected] & 0.58 & @xmath25216.1 & @xmath214 @xmath25215.2 & @xmath214 1.37 & @xmath2521.90 & 1.4 & 10.9 & @xmath214 41.8 & cdf - s & n + [email protected] & 0.22 & @xmath25215.2 & @xmath25215.1 & 0.29 & @xmath2522.05 & 1.9 & 10.5 & 41.1 & cdf - s & a , r + [email protected] & 0.66 & @xmath25216.4 & @xmath214 @xmath25215.5 & @xmath214 1.68 & @xmath2521.59 & 1.4 & 10.6 & @xmath214 41.7 & cdf - s & n , s + + [email protected] & 0.33 & @xmath25215.1 & @xmath214 @xmath25214.7 & @xmath214 0.56 & @xmath2521.19 & 1.4 & 10.2 & @xmath214 41.8 & e - cdf - s 01 & n + [email protected] & 0.12 & @xmath25215.7 & @xmath25215.0 & 0.90 & @xmath2522.74 & 0.9 & 10.3 & 40.4 & cdf - s & a + [email protected] & 0.66 & @xmath214 @xmath25215.5 & @xmath214 @xmath25214.8 & @xmath41 & @xmath2520.91 & 1.4 & 10.5 & @xmath214 42.4 & e - cdf - s 01 & a , s + [email protected] & 0.63 & @xmath25215.8 & @xmath25215.2 & 0.72 & @xmath2520.90 & 1.0 & 10.5 & 41.9 & cdf - s & a + [email protected] & 0.64 & @xmath214 @xmath25216.4 & @xmath25215.9 & @xmath253 0.73 & @xmath214 @xmath2522.01 & 1.4 & 10.9 & 41.2 & cdf - s & a , s + + [email protected] & 0.17 & @xmath25215.3 & @xmath214 @xmath25214.8 & @xmath214 0.69 & @xmath2520.22 & 1.4 & 8.3 & @xmath214 41.0 & e - cdf - s 01 & a + [email protected] & 0.47 & @xmath25216.5 & @xmath214 @xmath25215.5 & @xmath214 1.90 & @xmath214 @xmath2522.18 & 1.4 & 10.7 & @xmath214 41.3 & cdf - s & n , s + [email protected] & 0.17 & @xmath25215.5 & @xmath214 @xmath25214.9 & @xmath214 1.05 & @xmath2522.87 & 1.4 & 10.7 & @xmath214 41.0 & e - cdf - s 01 & n + [email protected] & 0.24 & @xmath25215.1 & @xmath214 @xmath25214.9 & @xmath214 0.44 & @xmath2522.49 & 1.4 & 11.1 & @xmath214 41.3 & e - cdf - s 04 & n + [email protected] & 0.60 & @xmath25215.3 & @xmath25215.0 & 0.51 & @xmath2520.42 & 1.5 & 10.1 & 42.0 & e - cdf - s 01 & a + + [email protected] & 0.11 & @xmath25215.7 & @xmath214 @xmath25215.2 & @xmath214 0.56 & @xmath2522.89 & 1.4 & 10.2 & @xmath214 40.2 & cdf - s & n + [email protected] & 0.12 & @xmath25215.5 & @xmath214 @xmath25215.0 & @xmath214 0.86 & @xmath2522.54 & 1.4 & 10.0 & @xmath214 40.5 & e - cdf - s 01 & n + [email protected] & 0.56 & @xmath214 @xmath25215.6 & @xmath25215.3 & @xmath253 0.45 & @xmath214 @xmath2521.50 & 1.4 & 10.8 & 41.7 & e - cdf - s 01 & a , s + [email protected] & 0.26 & @xmath25215.0 & @xmath214 @xmath25214.8 & @xmath214 0.39 & @xmath2521.31 & 1.4 & 10.1 & @xmath214 41.5 & e - cdf - s 04 & n + [email protected] & 0.57 & @xmath25216.0 & @xmath214 @xmath25215.3 & @xmath214 1.24 & @xmath214 @xmath2521.50 & 1.4 & 10.8 & @xmath214 41.7 & e - cdf - s 04 & n + + [email protected] & 0.16 & @xmath214 @xmath25215.8 & @xmath25215.4 & @xmath253 0.65 & @xmath214 @xmath2521.31 & 1.4 & 9.0 & 40.4 & e - cdf - s 01 & a , r , s + [email protected] & 0.64 & @xmath214 @xmath25215.7 & @xmath25214.2 & @xmath253 4.15 & @xmath2520.09 & 1.4 & 10.4 & 42.5 & e - cdf - s 04 & a + [email protected] & 0.69 & @xmath214 @xmath25215.8 & @xmath25214.6 & @xmath253 2.60 & @xmath2520.50 & 1.4 & 10.5 & 42.2 & e - cdf - s 01 & a + [email protected] & 0.65 & @xmath25214.9 & @xmath25213.9 & 1.57 & 0.46 & 0.5 & 10.3 & 43.0 & e - cdf - s 04 & a + [email protected] & 0.14 & @xmath25214.6 & @xmath214 @xmath25214.8 & @xmath214 0.19 & @xmath2522.52 & 1.4 & 10.8 & @xmath214 40.9 & e - cdf - s 01 & n + + [email protected] & 0.14 & @xmath25214.4 & @xmath214 @xmath25214.9 & @xmath214 0.11 & @xmath2522.46 & 1.4 & 10.9 & @xmath214 40.8 & e - cdf - s 01 & n + [email protected] & 0.51 & @xmath25214.9 & @xmath25214.5 & 0.66 & @xmath2520.12 & 1.3 & 10.1 & 42.4 & e - cdf - s 01 & a + [email protected] & 0.16 & @xmath214 @xmath25215.6 & @xmath25214.4 & @xmath253 2.40 & @xmath2522.02 & 1.4 & 10.4 & 41.3 & e - cdf - s 01 & a + [email protected] & 0.36 & @xmath214 @xmath25215.5 & @xmath214 @xmath25214.9 & @xmath41 & @xmath2521.97 & 1.4 & 10.5 & @xmath214 41.7 & e - cdf - s 01 & a , s + + + 0.25 @xmath254 0.08 & 45 & 5 & 15.0 & 30.5 @xmath254 7.4 & 94.5 @xmath254 11.9 & @xmath214 27.0 & 9.8 & 19.7 & 0.9 + 0.47 @xmath254 0.03 & 52 & 2 & 20.4 & 15.6 @xmath254 6.6 & 31.4 @xmath254 9.1 & @xmath214 32.4 & 4.1 & 5.5 & 2.1 + 0.58 @xmath254 0.02 & 51 & 2 & 18.2 & 12.9 @xmath254 6.2 & 33.5 @xmath254 9.0 & @xmath214 31.2 & 3.6 & 6.2 & 2.2 + 0.66 @xmath254 0.02 & 74 & 3 & 30.0 & 16.6 @xmath254 7.3 & 40.7 @xmath254 10.6 & 34.4 @xmath254 13.4 & 3.5 & 5.7 & 3.2 + + + + + 0.24 @xmath254 0.11 & 27 & 1 & 11.9 & 10.1 @xmath254 5.4 & 25.5 @xmath254 7.8 & @xmath214 27.1 & 3.4 & 5.7 & 2.9 + 0.46 @xmath254 0.03 & 27 & 0 & 9.9 & @xmath214 14.5 & 12.7 @xmath254 6.5 & @xmath214 22.0 & 2.7 & 3.2 & 0.5 + + + + + 0.25 @xmath254 0.08 & 41 & 4 & 14.1 & 26.3 @xmath254 7.0 & 75.3 @xmath254 10.9 & @xmath214 26.2 & 8.7 & 16.2 & 1.0 + 0.47 @xmath254 0.03 & 44 & 2 & 16.5 & @xmath214 17.0 & 25.4 @xmath254 8.3 & @xmath214 30.0 & 2.9 & 4.9 & 2.3 + 0.58 @xmath254 0.02 & 30 & 2 & 9.5 & 12.5 @xmath254 5.5 & 28.0 @xmath254 7.6 & 21.9 @xmath254 8.8 & 4.9 & 7.2 & 3.6 + 0.66 @xmath254 0.02 & 55 & 3 & 20.5 & 17.6 @xmath254 6.7 & 42.2 @xmath254 9.8 & @xmath214 33.5 & 4.6 & 7.3 & 2.9 + + + + + 0.22 @xmath254 0.09 & 28 & 2 & 11.2 & 14.3 @xmath254 5.8 & 43.5 @xmath254 8.9 & @xmath214 26.3 & 4.9 & 10.1 & 2.6 + 0.46 @xmath254 0.03 & 31 & 0 & 12.5 & 11.9 @xmath254 5.7 & 15.3 @xmath254 7.1 & @xmath214 24.2 & 3.9 & 3.4 & 0.3 + + + 0.25 @xmath254 0.08 & @xmath25216.5 @xmath254 0.6 & @xmath25216.2 @xmath254 0.2 & @xmath214 @xmath25216.0 & @xmath2523.00 & 39.7 @xmath254 0.6 & 40.2 @xmath254 0.2 & @xmath214 40.4 & 10.5 & 29.7 @xmath254 0.1 + 0.47 @xmath254 0.03 & @xmath25216.9 @xmath254 0.7 & @xmath25216.8 @xmath254 0.4 & @xmath214 @xmath25216.1 & @xmath2522.63 & 39.9 @xmath254 0.8 & 40.1 @xmath254 0.4 & @xmath214 40.8 & 10.4 & 29.7 @xmath254 0.1 + 0.58 @xmath254 0.02 & @xmath25217.0 @xmath254 0.8 & @xmath25216.8 @xmath254 0.3 & @xmath214 @xmath25216.1 & @xmath2522.29 & 40.1 @xmath254 0.8 & 40.4 @xmath254 0.4 & @xmath214 41.1 & 10.5 & 29.9 @xmath254 0.1 + 0.66 @xmath254 0.02 & @xmath25217.1 @xmath254 0.8 & @xmath25216.9 @xmath254 0.3 & @xmath25216.4 @xmath254 0.5 & @xmath2522.21 & 40.1 @xmath254 0.8 & 40.4 @xmath254 0.4 & 40.8 @xmath254 0.6 & 10.6 & 29.9 @xmath254 0.1 + + + + + 0.24 @xmath254 0.11 & @xmath25217.1 @xmath254 0.9 & @xmath25216.7 @xmath254 0.4 & @xmath214 @xmath25215.9 & @xmath2522.70 & 39.4 @xmath254 0.9 & 39.6 @xmath254 0.4 & @xmath214 40.5 & 9.8 & 29.9 @xmath254 0.2 + 0.46 @xmath254 0.03 & @xmath214 @xmath25216.8 & @xmath25217.0 @xmath254 0.7 & @xmath214 @xmath25216.0 & @xmath2551.79 & @xmath214 40.5 & 40.0 @xmath254 0.7 & @xmath214 41.0 & 9.6 & 30.4 @xmath254 0.2 + + + + + 0.25 @xmath254 0.08 & @xmath25216.5 @xmath254 0.6 & @xmath25216.3 @xmath254 0.2 & @xmath214 @xmath25216.0 & @xmath2523.08 & 39.7 @xmath254 0.6 & 40.1 @xmath254 0.2 & @xmath214 40.4 & 10.4 & 29.7 @xmath254 0.1 + 0.47 @xmath254 0.03 & @xmath214 @xmath25216.6 & @xmath25216.8 @xmath254 0.4 & @xmath214 @xmath25216.1 & @xmath2522.67 & @xmath214 40.2 & 40.1 @xmath254 0.4 & @xmath214 40.9 & 10.4 & 29.8 @xmath254 0.2 + 0.58 @xmath254 0.02 & @xmath25216.7 @xmath254 0.7 & @xmath25216.5 @xmath254 0.3 & @xmath25216.1 @xmath254 0.5 & @xmath2522.21 & 40.3 @xmath254 0.8 & 40.6 @xmath254 0.4 & 41.0 @xmath254 0.6 & 10.5 & 30.2 @xmath254 0.1 + 0.66 @xmath254 0.02 & @xmath25216.9 @xmath254 0.7 & @xmath25216.7 @xmath254 0.3 & @xmath214 @xmath25216.1 & @xmath2522.22 & 40.3 @xmath254 0.7 & 40.6 @xmath254 0.4 & @xmath214 41.2 & 10.5 & 30.1 @xmath254 0.1 + + + + + 0.22 @xmath254 0.09 & @xmath25217.0 @xmath254 0.7 & @xmath25216.4 @xmath254 0.3 & @xmath214 @xmath25215.9 & @xmath2522.71 & 39.5 @xmath254 0.7 & 39.8 @xmath254 0.3 & @xmath214 40.4 & 9.8 & 30.0 @xmath254 0.1 + 0.46 @xmath254 0.03 & @xmath25217.1 @xmath254 0.8 & @xmath25217.0 @xmath254 0.6 & @xmath214 @xmath25216.0 & @xmath2551.96 & 40.1 @xmath254 0.8 & 40.0 @xmath254 0.6 & @xmath214 40.9 & 9.8 & 30.2 @xmath254 0.2 + 2.2 for further details ) . the emission from ngc 1600 was best fit using a power - law ( @xmath256 ) for the discrete component and a mekal plasma ( @xmath257 kev ; @xmath258 ) for the unresolved component .
ngc 4697 was best fit by a thermal bremsstrahlung ( @xmath259 kev ) for the lmxb component and a mekal plasma ( @xmath260 kev ; @xmath261 ) for the hot interstellar gas . for reference
, we have shown the energy ranges of our adopted bandpasses redshifted to @xmath262 and @xmath263 , the median redshifts of our optically luminous and faint samples , respectively .
we note that for luminous ( optically luminous ) early - type galaxies within the redshift ranges considered in this study , the ( sb1 ) and ( sb ) bandpasses primarily trace hot interstellar gas and the ( hb ) bandpass is primarily dominated by lmxbs .
however , for faint ( optically faint ) early - type galaxies , all bandpasses are generally dominated by lmxb emission.,width=529 ] -band luminosity versus redshift for our general sample of 539 early - type galaxies .
larger circles indicate -detected agn candidates ( _ open _ ) and normal galaxies ( _ filled _ ) .
sources denoted with crosses are ( 1 ) in close proximity ( @xmath21410 ) to an -detected source , ( 2 ) within the boundaries of extended sources , ( 3 ) at large off - axis angles ( i.e. , @xmath2537 from all aimpoints ) , and/or ( 4 ) found to have relatively large radio to optical flux ratios ( see @xmath27 2.1.1 ) ; these sources have been removed from our stacking analyses ( see @xmath27 2.2 for details ) .
the shaded bands show the luminosity ranges of our optically luminous and faint samples ; vertical dotted lines indicate the evenly - spaced comoving volume intervals chosen in constructing our stacking samples .
( * b * ) evolved , @xmath9 , @xmath75-band luminosity of our sample ( see discussion in @xmath27 2.1.2 ) , which constitutes our faded sample .
symbols and boundaries are the same as in figure 3a . using a chabrier initial mass function ( chabrier 2003 )
, we estimate that the luminosity ranges @xmath264 and @xmath86 correspond roughly to stellar mass ranges of @xmath265 and @xmath88 , respectively . , width=529 ] 1 ms and @xmath5250 ks .
positions of stacked sources from our faded sample are shown as open circles ( @xmath51 ms ) and open diamonds ( @xmath5250 ks ) ; filled circles represent galaxy positions that have been stacked using both the @xmath51 ms and @xmath5250 ks observations ( see @xmath27 2.2 for additional details ) .
aim points of each observation are indicated as plus signs , and the surrounding 70 inclusion radii are indicated with dashed circles .
the apparent lack of sources in the north - eastern corner ( i.e. , the upper left - hand corner ) of the image is partially due to missing coverage from the gems imaging . for reference , we have outlined the @xmath5160 arcmin@xmath32 goods - s region ( _ solid rotated rectangle _ ; giavalisco 2004).,width=529 ] -band magnitude versus flux for -detected early - type galaxies in our sample .
agn candidates and normal galaxies are plotted as open symbols ( _ circles and stars _ ) and filled circles , respectively ; sources with upper limits were detected in either the or bandpasses .
diagonal dotted lines represent lines of constant to optical flux ratio ( i.e. , @xmath266 + 1 , @xmath2521 , and @xmath2522 ) ; luminous agns generally have @xmath267 ( _ shaded region _ ) .
agn - candidates were classified following the three criteria discussed in @xmath27 2.1.1 .
the sources marked with five - pointed stars are radio - detected agn candidates , including the fr ii source cxoecdfs .
furthermore , mean @xmath116-band magnitudes and fluxes for our stacked optically luminous ( _ large filled squares _ ) and faint ( _ large filled triangles _ ) faded samples are plotted with varying grayscale levels to indicate redshift ; darker levels indicate larger redshift values . , width=529 ] , as a function of luminosity for our optically luminous faded samples with @xmath124 0.100.55 ( _ filled circles _ ) and ( _ open diamonds _ ) . each data point with printed fractions
represents the number of agn candidates detected with luminosity of @xmath268 or greater divided by the number of early - type galaxies with coverage sufficient to detect an agn of @xmath121 ( see @xmath27 3.1 for further details).,width=529 ] ) samples , and the bottom panels show our optically faint ( @xmath269 ) samples .
these images were generated using the ciao tool csmooth with a minimum significance of 2.5@xmath67 .
the images are @xmath515 ( @xmath530.5 pixels ) per side , and each pixel is 0492 .
our circular stacking aperture of radius 15 is shown in each image centered on the optical centroid of our stacked sources .
additional sample information , including the number of stacked galaxies , total exposure , signal - to - noise ratio ( s / n ) , and logarithm of the mean @xmath9 @xmath75-band luminosity ( @xmath77 ) are annotated on each smoothed image .
, width=680 ] -band luminosity , @xmath24 , for @xmath135 kpc local ellipticals with the average properties of our faded samples plotted .
small symbols and upper limits are sources from the os01 local sample , and the different symbols correspond to field galaxies ( _ circles _ ) , central - cluster galaxies ( _ upward - pointing open triangles _ ) , brightest - group galaxies ( _ downward - pointing open triangles _ ) , and agns ( _ crosses _ ) .
the best - fit relations for the luminosity intervals @xmath270 and @xmath271 are shown as solid and dashed lines , respectively ; the shaded region shows the expected discrete - source contribution ( kim & fabbiano 2004 ; see also the discussion in @xmath27 3.2.1 ) .
the local ellipticals m32 , m87 , ngc 1399 , ngc 1600 , ngc 4697 , and ngc 5102 have been marked , for reference .
large shaded symbols represent our optically luminous ( _ squares _ ) and faint ( _ triangles _ ) faded samples and have grayscale levels corresponding to the mean redshift of each sample , such that darker shading represents higher redshifts .
error bars in represent 1@xmath67 errors on the mean .
we quote @xmath75-band luminosities as @xmath77 to illustrate any potential evolution of the mean luminosities.,width=529 ] ) and faint ( @xmath73 ) samples , respectively .
error bars in redshift represent the standard deviation of the redshift for sources in each stacked sample . for comparison ,
we have plotted the corresponding mean luminosities ( and errors on the means computed with asurv ) of normal early - type galaxies from the os01 local sample ( _ gray filled square and triangle _ ) .
all mean values ( ours and those of os01 ) were calculated after excluding agns , central - cluster galaxies , and brightest - group galaxies , and should reflect the average properties of isolated field early - type galaxies .
the solid curve illustrates the median detection limit ( using the median sensitivity limit of for our total sample ) . for reference , the luminosities of the local ellipticals m32 , m87 , ngc 1399 , ngc 1600 , ngc 4697 , and ngc 5102 have been indicated.,width=529 ]
-band mean luminosity ratio ( @xmath272 ) versus redshift for our general ( * a * ; see figure 2a ) and faded ( * b * ; see figure 2b ) samples .
symbols have the same meaning as in figure 9 ; the dotted line and shaded region represent the expected local discrete - source contribution and its dispersion ( from kim & fabbiano 2004 ) . for the faded sample , we used @xmath77 when computing the to@xmath75-band mean luminosity ratios . for reference , the local ellipticals m32 , m87 , ngc 1399 , ngc 1600 , ngc 4697 , and ngc 5102 have been plotted.,width=529 ] for our general ( * a * ; see figure 2a ) and faded ( * b * ; see figure 2b ) samples . for optically luminous galaxies ( @xmath72 ) we used @xmath169 and @xmath170 , and for optically faint galaxies ( @xmath73 ) , we adopted @xmath171 and @xmath172 .
symbols have the same meaning as in figure 9 , and the redshifts of the local samples have been offset from @xmath9 for viewing ease .
the dotted horizontal line indicates the zero residual . for reference , the local ellipticals m32 , m87 , ngc 1399 , ngc 1600 ,
ngc 4697 , and ngc 5102 have been plotted.,width=529 ] or greater ) , @xmath120 , versus @xmath194 for our optically luminous faded samples .
the cumulative agn fraction computed over the entire redshift range , @xmath177 , is indicated as filled circles with 1@xmath67 error bars , and the thick solid curve represents our best - fit quadratic relation to the data ( see @xmath27 3.2.2 for details ) .
this relation was used to estimate @xmath120 for our optically luminous faded samples ( _ annotated curves _ ) . for reference ,
we have plotted the observed agn fraction for @xmath187 drgs ( rubin 2004 ; _ open diamond _ ) and our model extrapolated out to @xmath273 ( _ dot - dashed curve _ ) .
( * b * ) differential agn fractions ( i.e. , the fraction of galaxies harboring an agn in discrete bins of width @xmath274 ) , @xmath190 , versus @xmath194 .
( * c * ) fraction of early - type galaxies for which we could _ not _ have detected an agn with a luminosity of @xmath121 , @xmath192 , versus @xmath178 .
( * d * ) fraction of galaxies harboring agns in our optically luminous faded samples that would remain undetected due to sensitivity limitations , @xmath275 , versus @xmath194 ; these galaxies would not have been removed from our stacking analyses .
, width=377 ] , versus redshift for our optically luminous faded samples ( _ filled squares _ ) ; these luminosities were derived following equation 8 .
error bars on @xmath226 are 1@xmath67 errors , which were derived by propogating errors on the fit to @xmath177 ( _ solid curve _ in fig .
12 ) through to equation 8 . for comparison ,
we have plotted the brand ( 2005 ) stacking constraints for their early - type galaxy samples ( _ open diamonds _ ) , which have mean @xmath116-band absolute magnitudes similar to those of our optically luminous faded samples .
the dotted curve represents the @xmath6 model fit to our data.,width=529 ] | we investigate the evolution over the last 6.3 gyr of cosmic time ( i.e. , since @xmath0 ) of the average properties of early - type galaxies within the extended deep field - south ( ) .
our early - type galaxy sample includes 539 objects with red - sequence colors and srsic indices larger than @xmath1 , which were selected jointly from the combo-17 ( classifying objects by medium - band observations in 17 filters ) and gems ( galaxy evolution from morphologies and seds ) surveys .
we utilize the deep observations over the and stacking analyses to constrain primarily the average emission from `` normal '' early - type galaxies ( i.e. , those that are not dominated by luminous active galactic nuclei [ agns ] ) . in our analyses ,
we study separately optically luminous ( @xmath2 ) and faint ( @xmath3 ) galaxy samples , which we expect to have soft ( ) emission dominated by hot ( @xmath41 kev ) interstellar gas and low mass binary ( lmxb ) populations , respectively .
we detect individually 49 ( @xmath59% ) of our galaxies in the band , and classify these sources as either normal early - type galaxies ( 17 galaxies ) or agn candidates ( 32 galaxies ) .
the agn fraction of our optically luminous samples evolves with redshift in a manner consistent with the @xmath6 evolution observed in other investigations of -selected agns . after removing potential agns from our samples , we find that the mean luminosity ratio ( @xmath7 ) for optically luminous early - type galaxies does not evolve significantly over the redshift range .
this lack of evolution implies a general balance between the heating and cooling of the hot interstellar gas .
if transient agn activity is largely responsible for maintaining this balance , then we infer that mechanical power must be dominating the feedback out to @xmath0 .
furthermore , in this scenario the average mechanical agn power must remain roughly constant over the last half of cosmic time despite significant evolution in the average agn radiative luminosity .
for our optically faint early - type galaxies , we find suggestive evidence that @xmath7 increases with redshift over the range ; however , due to limited statistical constraints on both the local @xmath7 ratio and the agn contamination of our samples , we consider this result to be marginal . |
gg tau is a young stellar system within the taurus molecular cloud complex which is located at a distance of @xmath1140pc ( elias 1978 ) .
it consists of a pair of close binary stars ( leinert et al .
1991 ) separated by @xmath110
( @xmath11500 au ) .
the main component of this binary pair consists of a k7-m0 star ( strom et al .
1989 ; hartigan et al . 1994 ) and an m4 star ( roddier et al . 1996 ) separated by 0.255 ( @xmath138 au ) , and has collectively been classified as an emission - line ( `` classical '' ) t tauri star ( herbig & bell 1988 ) . associated with this system , which we will refer to simply as gg tau , is a large , massive circumstellar disk , which has recently been imaged using millimeterwave interferometers by a number of groups ( simon & guilloteau 1992 ; kawabe et al 1993 ; koerner , sargent & beckwith 1993 ; dutrey , guilloteau & simon 1994 ) .
they found a well - resolved , extended feature , whose morphology and kinematics are consistent with a rotating _
circumbinary _ disk ( dutrey , guilloteau & simon 1994 ; koerner , sargent & beckwith 1993 ) .
dutrey et al .
have also found that the disk , of outer radius @xmath1800 au , appears to have a central hole or cavity of radius @xmath1180 au .
roddier et al .
( 1996 ) have recently presented near - infrared images acquired using adaptive - optics techniques , which clearly show a partially evacuated cavity of radius @xmath1200 au in the disk .
they also find some infrared excess from both of the stars inside the cavity , perhaps from warm inner disks surrounding each star .
the circumstellar / circumbinary disk radiates very strongly at millimeter and sub - millimeter wavelengths ( beckwith et al 1990 ; beckwith & sargent 1991 ; moriarty - schieven et al 1994 ) .
indeed , the published 800 @xmath0 flux densities are nearly 25% of the _ iras _ 100 @xmath0 flux density of 5.2 jy .
we present here evidence that the sub - mm flux density of gg tau has varied .
we interpret our observations as a sudden brightening or `` flare '' of the disk .
we will present a set of simple models that can match the observations , and discuss their implications .
ccccc [ flux - tab ] 12 & 1983 & 25 & 1.27 @xmath2 0.13 & iras psc 25 & 1983 & 25 & 1.65 @xmath2 0.15 & iras psc 60 & 1983 & 60 & 3.03 @xmath2 0.27 & iras psc 100 & 1983 & 100 & 5.16 @xmath2 0.57 & iras psc 450 & jan 1994 & 17.5 `` & 4.54 @xmath2 0.35 & msb & dec 1994 & 17.5 '' & 4.16 @xmath2 0.23 & msb 800 & 29 nov , 3 dec 1989 , 7 dec 1990 & 17 `` & 1.25 @xmath2 0.08 & bs & sept 1991 & 16.5 '' & 1.11 @xmath2 0.12 & mswkt & jan 1994 & 16.5 `` & 1.71 @xmath2 0.04 & msb & sept 1994 & 16.5 '' & 1.59 @xmath2 0.06 & msb & dec 1994 & 16.5 " & 1.65 @xmath2 0.08 & msb 1100 & 2 dec 1989 & 22 `` & 0.800 @xmath2 0.051 & bs & sept 1991 & 18.5 '' & 0.74 @xmath2 0.12 & mswkt & jan 1994 & 18.5 `` & 1.07 @xmath2 0.03 & msb & sept 1994 & 18.5 '' & 0.83 @xmath2 0.03 & msb & dec 1994 & 18.5 " & 0.85 @xmath2 0.03 & msb 1300 & 29 , 30 april 1988 & 11 `` & 0.593 @xmath2 0.053 & bscg & jan 1994 & 19 '' & 0.69 @xmath2 0.03 & msb & sept 1994 & 19 `` & 0.63 @xmath2 0.02 & msb & dec 1994 & 19 '' & 0.63 @xmath2 0.03 & msb 2000 & jan 1994 & 28 " & 0.32 @xmath2 0.06 & msb the data to be discussed in this report are presented in table 1 , and include new observations as well as observations of beckwith et al ( 1990 ; hereafter bscg ) , beckwith & sargent ( 1991 ; hereafter bs ) , and moriarty - schieven et al ( 1994 ; hereafter mswkt ) .
the new observations presented here were obtained at the james clerk maxwell telescope ( jcmt ) located near the summit of mauna kea , in january , september , and december 1994 .
the photometry was obtained using the ukt14 bolometer equipped with filters passing effective wavelengths of 444 , 790 , 1090 , 1260 and 1920 @xmath0 , using a 65 mm aperture ( corresponding to 17.5 `` , 16.5 '' , 18.5 `` , 19 '' and 28 `` ( fwhm ) respectively on the sky ) , chopping in azimuth by 60 '' at a frequency of 7.8 hz .
cccc [ crl618-tab ] crl 618 & 450 & 13 @xmath2 2 & & 800 & 4.3 @xmath2 0.2 & 4.3 @xmath2 0.2 & 1100 & 2.6 @xmath2 0.1 & 2.8 @xmath2 0.2 & 1300 & 2.3 @xmath2 0.1 & 2.7 @xmath2 0.2 & 2000 & 1.7 @xmath2 0.1 & 2.2 @xmath2 0.2 l1551-irs5 & 450 & 41 @xmath2 4 & & 800 & 5.7 @xmath2 0.3 & 1100 & 2.4 @xmath2 0.1 & 1300 & 1.7 @xmath2 0.1 & 2000 & 0.7 @xmath2 0.2 no planets were available for the run of january 1994 , so both atmospheric opacity and flux calibration were estimated from observations of crl 618 .
pointing , which varied by @xmath32 - 3 " , was checked at every calibration observation performed at least once per hour .
flux densities were also checked using l1551-irs5 . assumed flux densities for these two objects were from sandell ( 1994 ) , and are shown in table 217% at 1300@xmath0 and @xmath129% at 2 mm .
this would tend to enhance the flux increase we see between the 1992 and 1994 data sets . ] . the quoted errors for these observations ( table 1 ) include statistical and calibration uncertainties , the latter of which includes the uncertainty in the flux density of crl 618 . during the run of september 1994 , the zenith optical depth at 225 ghz as measured by the radiometer operated by the nearby caltech submillimeter observatory ( cso ) , was found to be extremely stable , with maximum excursions ranging from @xmath4 = 0.11 to 0.12 .
thus we were able to point , focus , and obtain photometry of l1551-irs5 ( only a few degrees away ) while it was at virtually the same airmass as gg tau .
we were also able to observe mars @xmath12 hours later when it too was at nearly the same airmass as gg tau ( 1.03 compared to 1.00 ) .
these observations , assuming flux densities from sandell ( 1994 ) for l1551-irs5 and from the planetary flux density model of wright ( 1976 ) , were then applied _ directly _ to the photometric observations of gg tau without having to estimate atmospheric opacity or instrumental sensitivity .
( the differences in airmass between mars and gg tau would only introduce an error of @xmath13% even assuming @xmath5 1 . )
consistent values of flux density were found from both calibrators , and quoted errors reflect both the statistical and calibration uncertainties . during the run of december 1994 , we obtained the zenith optical depth at 225 ghz from the cso radiometer and found the sky to be stable over the course of the observations .
two observations of mars at airmasses of 1.5 and 1.1 allowed us to estimate the atmospheric optical depth at the observed wavelengths over the relevant airmass .
we also had observations of l1551 irs 5 at virtually the same airmass as gg tau .
we used these two calibration methods ( i.e. deriving @xmath4 and sensitivity from mars , and direct comparison with l1551-irs5 ) to derive consistent values of flux density for gg tau .
the data from bscg were obtained at the institut de radio astronomie millimtrique ( iram ) 30 m telescope at pico veleta in april 1988 , and were calibrated for flux density using mars and uranus .
the effective wavelength of the observations was 1300@xmath0 .
bs observed using a bolometer on the cso in late 1989 and 1990 , calibrating for atmospheric opacity using hl tau and for flux density on mars in 1990 and on w3(oh ) , crl618 , irc+10216 , 3c273 , 3c84 , and hl tau in 1989 ( no planets were available ) .
the effective wavelengths of these observations were 1056 and 769@xmath0 with fwhm of 22 and 17 respectively .
the observations from mswkt were obtained at the jcmt in september 1991 at effective wavelengths of 790 and 1090 @xmath0 and fwhm of 16.5 and 18.5 respectively , calibrating flux density from mars and atmospheric opacity from crl618 .
figure [ ggtau_fig1 ] displays the flux density of gg tau at 800 , 1100 and 1300 @xmath0 as a function of the date of observation .
( the 800@xmath0 observation of bs is an average of observations made in late 1989 and late 1990 , and so is placed at 1990.4 . ) the errors shown on figure [ ggtau_fig1 ] include both statistical and calibration uncertainties , and are often dominated by the latter .
( note that these errors do not include the uncertainty of the absolute flux of mars ( about 5% ) based on the model of wright ( 1976 ) . )
it is apparent that from september 1991 to january 1994 the flux density of gg tau has increased by more than 50% at 800@xmath0 and by @xmath640% at 1100@xmath0 , but by only @xmath117% at 1300@xmath0 . from january to september 1994 , there was a marginal decrease at 800@xmath0 , and a decrease to 1991 values at 1100 and 1300 @xmath0 .
it thus appears that the circumbinary disk of gg tau experienced a significant increase in its measured luminosity , a `` flare '' , between late 1991 and late 1993 , affecting predominantly sub - millimeter wavelengths .
one must use caution , however , before accepting such a claim .
errors in calibration , etc . , could account for apparent changes in measured flux density . as noted in 2 , all of these data except for the run of january 1994 and of december 1989 ( from bs ) were calibrated for flux density from mars , and by using the assumed martian flux densities calculated from the model of wright ( 1976 ) .
the december 1989 observation at 1100@xmath0 was calibrated from several sources .
the january 1994 run was calibrated on crl618 , whose flux density was checked on mars during the september 1994 run .
thus , with minor exceptions , all of the calibration has been done in a consistent manner , and we are confident that the derived flux densities are correct within errors .
cccc [ sanity - check - tab ] 04113 + 2758 & 461 @xmath2 53 & 510 @xmath2 40 & 1.11 @xmath2 0.15 04169 + 2702 & 281 @xmath2 53 & 230 @xmath2 30 & 0.82 @xmath2 0.19 gg tau & 737 @xmath2 117 & 1070 @xmath2 30 & 1.45 @xmath2 0.23 04361 + 2547 & 188 @xmath2 27 & 200 @xmath2 20 & 1.06 @xmath2 0.19 04365 + 2535 & 438 @xmath2 38 & 390 @xmath2 40 & 0.89 @xmath2 0.12 five of the sources including gg tau which were observed in september 1991 by mswkt were re - observed at 1100@xmath0 during our january 1994 run , providing a consistency check .
these data are shown in table 3 . in four of the five cases , the ratio of 1994 to 1991 flux densities
is well within one standard deviation of being unity .
only gg tau has an elevated flux density .
finally we note that the 800@xmath0 and 1100@xmath0 flux densities measured by bs and bscg in 1989/1990 on different telescopes using different instruments are consistent with our 1991 observations and inconsistent with those of post-1991 .
the ratio of our september 1994 800@xmath0 data to the average bs 1989 data is [email protected] , and the corresponding 1100@xmath0 data is [email protected] .
other effects could potentially mimic a change in flux density .
if the source is extended , a smaller aperture or beamwidth would detect less of the emission .
however , interferometric observations , including those of dutrey , guilloteau & simon ( 1994 ) , have resolved the structure of gg tau , and have found it to be @xmath73 " , i.e. much smaller than any of the beamwidths used .
the component to the south associated with gg tau(c ) ( the smaller of the two binary systems ) is not visible at 2.7 mm in the dutrey et al maps , indicating that there is little contribution to the millimeter emission from that source .
in addition , the jcmt beam used by mswkt is identical to ours , and would cover the same area .
finally , it must be noted that the observations of bs are made at slightly _
shorter _ wavelengths ( 769@xmath0 compared to 790@xmath0 , and 1056@xmath0 compared to 1090@xmath0 ) .
since flux density goes as @xmath8 , however , the effect of this should be to _ increase _ the flux density seen by bs , which is opposite to the effect seen .
figure [ ggtau_fig2 ] displays the spectral energy distribution ( sed ) of gg tau using the pre-1992 data from table 1 , the 1994 january and december data , and the iras flux densities .
mswkt have shown that the slope of the sed at millimeter / sub - millimeter wavelengths is steeper for embedded sources compared to more evolved , optically visible , t tauri stars ( like gg tau ) .
indeed gg tau had the shallowest sed of all the sources in their study . from figure [ ggtau_fig2 ]
it can be seen that the slope of the sed appears to have changed as gg tau has undergone the `` flare '' . in table 4
we show the slope @xmath9 of the sed ( from @xmath8 ) derived from each data set ( bs includes a 624@xmath0 data point plus the data from bscg ) .
the sed appears to have steepened abruptly after the `` flare '' , and continues to be steeper than pre-``flare '' .
any model ought to account for a number of effects , including an increase of the 800 @xmath0 flux density of at least 50% over a timescale of @xmath72 years , with relatively little effect at longer wavelengths , resulting in a steepening of the sed during the flare and a more rapid return at longer wavelengths to `` normal '' .
let us consider the existing disk models for gg tau s continuum emission . in 1990 , beckwith et al presented a source model for gg tau that consisted of a geometrically thin disk , with a large optically thick regime ( out to 45 au ) and an overall disk size of 100 au .
their model accounts for the fluxes from 12 to 1300 @xmath0 quite well ( fig .
[ ggtau_fig3 ] , solid line ) . however , dutrey et al ( 1994 ) reported that the disk had a large central 180 au radius `` hole '' , based on 2.6 mm interferometer observations .
they presented an alternate model , where a cold dust ring at approximately 180 au was the dominant emission source of the 2.6 mm emission .
beyond the ring , which contained some 90% of the disk mass , the disk continued with a very flat surface density law .
in addition , they replaced the optically thin disk with a flared disk , whose scale height was roughly 30 au at 180 au radius .
they argued that the temperature profile was likely to be t@xmath10r@xmath11 , because the flared disk would intercept a larger fraction of the stellar luminosity than spatially thin disk models .
it effectively would act more like an envelope than a disk when intercepting the stellar radiation field .
artymowicz and lubow ( 1995 ) showed how such a circumbinary disk might arise and actually be dynamically important for the evolution of the central binary .
we point out that if the dutrey et al .
idea of a dense dust ring is correct , the temperature profile is likely to be steeper in the dust ring ( and the outer disk colder ) than their model assumes .
this occurs because the dust ring , which has a large extinction , attenuates the radiation field of the central stars quite strongly ( wolfire and churchwell 1994 ) . while the inner temperature would likely remain at 25 k
, the temperature of the outer disk would fall below the 20 k that is assumed in the dutrey et al .
model . the outer disk , thus , could be effectively shielded from the central heating source .
in addition , their model does not produce any substantial flux below 100@xmath0 ( fig .
[ ggtau_fig3 ] , dashed line ) , implying that at least one additional component ( either a smaller circumbinary disk or circumstellar disk ) must be present to explain the iras data . as an illustration of the type of disk that might be present , we present in fig .
[ ggtau_fig3 ] ( dotted line ) a disk model which is simply a truncated beckwith et al . model .
this optically thick ( at 50 @xmath0 ) disk has a mass of less than 10@xmath12 m@xmath13 . with an outer radius of only 15 au
, it lies well inside the estimated deprojected separation of 50 au reported by dutrey et al .
roddier et al .
( 1996 ) detected a near - infrared excess toward both stars in the cavity region , likely due to circumstellar disks .
also , the h@xmath9 emission from gg tau which led to its classification as a t tauri star is believed to be indirect evidence of a circumstellar disk , as opposed to the circumbinary disk seen at millimeter wavelengths .
given the current large uncertainties of the disk(s ) physical properties , we feel that it is unrealistic to construct a detailed model at this time .
therefore , we will investigate flat disk models . for these
, we adopt a single temperature power law for the disk , and the radii indicated by dutrey et al .
, as appropriate for the submillimeter region .
we adopt the procedures as described in butner , natta , & evans ( 1994 ) to parameterize the disk properties .
we varied the density , density profiles , temperature radial profiles , and dust optical properties to get a sense of how sensitive the predicted spectral energy distributions are to the input parameters .
we found the current spectral energy distribution could be fit by models within a broad range of parameter space . in fig .
[ ggtau_fig4 ] , we present three such models , where the `` before '' and `` after '' models are indicated by solid and dashed lines respectively . in each case , we limit ourselves to changing a single input parameter . in fig
[ ggtau_fig4](a ) , our model is based on a power law with a temperature profile given by t(r)@xmath10 r@xmath14 ; @xmath15 . the only change between the models is the disk s luminosity , or effectively the disk temperature at the inner radius . the second model ( fig .
[ ggtau_fig4](b ) ) has @xmath16 , and again , the only change between models is an increase in the disk luminosity .
both fit the pre - brightening phase reasonably well . by adjusting a single parameter , the total disk luminosity ( or equivalently the temperature at the disk inner edge )
, we can then produce the post - brightening observations .
the change in the disk luminosity is modest , going from @xmath170.01 to @xmath170.02 l@xmath18 in both cases .
the corresponding disk temperatures changes are modest , only a few degrees with the disks remaining very cold ( 10 - 20 k ) .
an alternative to varying the temperature and large - scale disk properties is to hold these constant , and instead vary the dust properties . in fig .
[ ggtau_fig4](c ) , we present `` pre- '' and `` post- '' brightening models where the dust opacity law ( parameterized as @xmath19 ) steepens from @xmath20 0.2 to @xmath20 1.0 .
we note that our @xmath20 0.2 model , while extremely high , is in agreement with the estimate of @xmath21 in beckwith and sargent s optically thin disks ( their @xmath22 models ) .
it is possible to produce the brightening effect in other ways of course , such as varying the inner radius of the dust ring .
the size scales involved ( the nominal model disk mass is a substantial fraction of 1 m@xmath18 and the radius is order 200 au ) , however , would argue for longer timescales than we see for this event .
we emphasize that the fits in fig .
[ ggtau_fig4 ] are simple `` toy '' models intended as a starting point for detailed calculations , not as a claim justifying a particular mechanism .
our `` toy '' models illustrate that the observations do not define the parameters sufficiently to identify physical causes for the observed brightening .
the effect is dramatic enough , however , that it is interesting to speculate .
what would cause a sudden increase in luminosity ( or temperature across the entire disk ) and/or a change in the dust properties of a circumstellar disk ?
various possibilities suggest themselves .
for example , an inner disk might still be accreting material ( artymowicz and lubow , 1995 ) .
if this disk were to undergo an outburst , it might have increased the energy available for the outer circumbinary disk to intercept . to date , no convincing reports of such dramatic changes have been reported at visual wavelengths for gg tau ( herbst , herbst & grossman 1994 ) .
a standard fu - ori - type outburst with its 100 times increase in luminosity certainly has not been seen .
another possibility is some sort of event in the circumbinary disk itself .
one case that appeals to dynamical events is a change in the disk scale height .
the change in scale height would permit the disk to intercept more radiation , thus allowing a natural way to increase the disk s luminosity .
the distances involved a 10 or 15 au change in scale height is required to account for a two - fold increase in total disk luminosity are large enough to be a concern .
the required velocities are more than 20 km / sec , and it is difficult to imagine what type of event would cause parts of the disk to achieve such large - scale motions . altering the grain properties would require a sudden density enhancement for that mechanism to be viable .
a sudden accretion event in the disk , or interaction with gg tau / c , the binary system which lies a few arcseconds to the south ( cf dutrey et al 1994 ) , might provide the trigger .
co(2 - 1 ) data presented by koerner et al .
( 1993 ) illustrates that the gas motions in the area may be more complex than the co(1 - 0 ) indicates , and the near - ir images of roddier et al .
( 1996 ) show bridge - like structures between the outer ring and inner stars .
if there had been an accretion shock as material fell on the disk , there might be additional energy
either for heating the grains or altering their structure enough to affect the optical properties . finally , we note that one other t tauri star , v773 tauri ( hd 283447 ) , is known to be variable at mm - wavelengths ( dutrey et al .
1996 ; jensen et al .
( 1994 ) ; bscg ; osterloh & beckwith 1995 ) .
it is also bright at cm - wavelenghts ( oneal et al .
1990 ; feigelson et al . 1994 ) and is rapidly and strongly variable at these wavelengths .
its variability and large brightness temperature ( phillips , lonsdale & feigelson 1991 ) suggest that the emission is synchrotron radiation related to a strong magnetic field .
it is conceivable that the sub - millimeter variability of gg tau may also be related to non - thermal processes .
however , unlike v773 tau , gg tau s variability seems greatest at short wavelengths , and indeed gg tau has not been detected at centimeter - wavelengths despite at least two sensitive surveys ( bieging , cohen & schwartz 1984 ; white , jackson & kundu 1993 ) .
thus it seems most likely that the outburst from gg tau is thermal in nature and likely related to the circumstellar disk and/or circumbinary ring . while at drao
, g.m - s . was supported by a research associateship from the national research council of canada .
h. m. b. gratefully acknowledges the support of the carnegie institute of washington through a carnegie fellowship , as well as support from a nasa origins of the solar systems program grant ( nagw-4097 ) .
he further thanks the max - planck - gesellschaft arbeitsgruppe `` staub in sternentstehungsgebieten '' for their support while he was on leave at jena , germany during the writing of the final draft of this paper . | we have monitored the millimeter and submillimeter emission from the young stellar object gg tau , a t tauri binary system surrounded by a massive circumbinary disk .
we find that between 1992 and 1994 , the flux has increased significantly at 800 , 1100 , and 1300 @xmath0 , resulting in a steepening of the observed spectral energy distribution at those wavelengths .
such an increase appears consistent with a modest increase in disk luminosity ( a factor of two ) .
the increase in the effective disk temperature might arise from a slight change in the disk heating processes .
alternatively , the flux increase may reflect a sudden change in the underlying dust optical properties . |
the ads - cft correspondence due to maldacena @xcite is a concrete realisation of t hooft s proposal relating yang - mills ( at large-@xmath4 ) to string theory .
the correspondence relating @xmath0 supersymmetric yang - mills ( sym ) theory to type iib strings propagating on @xmath5 has been tested in many different ways .
the first tests involved matching computations in the supergravity limit of the string theory and the corresponding ones in the conformal field theory ( cft ) following the proposal in @xcite .
recently , it was realised that the anomalous dimensions of operators in this theory are given by the energy spectrum of a spin - chain@xcite .
the @xmath0 theory has a high degree of symmetry and thus it is of interest to understand versions with lesser symmetry .
orbifolds of this theory provide one obvious class of cfts@xcite .
these lead to theories which are dual to type iib string theory propagating on @xmath6 , where @xmath7 are five manifolds that are orbifolds of @xmath8 .
more generally , @xmath7 must be a sasaki - einstein manifold . a new class of such manifolds that lead to cft s with @xmath9 supersymmetry
have also been constructed recently .
these manifolds have been called @xmath10 spaces cones over these spaces are ricci - flat non - compact six dimensional manifolds with @xmath11 holonomy and are natural generalisations of the conifold @xcite .
thus , these are examples where both sides of the ads - cft correspondence are understood even though we do not fully understand string theory on ads spaces . from a field
theoretic perspective , leigh and strassler@xcite considered ( multi - parameter ) marginal deformations of @xmath0 supersymmetric yang - mills theory that preserve @xmath9 supersymmetry and are conformal .
we shall refer to these theories are known as the leigh - strassler ( ls ) theories . while there have been attempts @xcite to understand
the anticipated dual string theories for these field theories , the precise correspondence is not known in all generality . an important development due to lunin and maldacena @xcite was the construction of the gravity duals for the @xmath12-deformed @xmath0 sym which is a sub - class in the generalised ls family of @xmath9 theories .
rational values of @xmath13 in the deformation turn out to be related to @xmath14 orbifolds of @xmath8 with discrete torsion@xcite .
the most general leigh - strassler @xmath9 preserving deformations of @xmath0 sym is given ( in superfields ) by the following superpotential : @xmath15 the @xmath0 limit occurs when @xmath16 and @xmath17 .
the @xmath12-deformed theory is obtained when @xmath18 and @xmath19 . for generic values of the deformations , the global symmetry of the deformed theory
is the trihedral group @xmath1 with its centre being a @xmath20 sub - group of the @xmath2 symmetry .
this is in contrast to the situation when @xmath21 , where one has a @xmath22 symmetry which is intimately related to the toric nature of these spaces .
the ls deformations are parametrised by the four couplings : @xmath23 ( the yang - mills coupling constant ) , @xmath24 , @xmath25 and @xmath26 .
while these are all marginal at the classical level , they are not all marginal in the quantum theory . however , it has been argued by leigh and strassler that in a subspace of the four - dimensional space of couplings , the theory is conformal .
in particular , the vanishing the four beta functions is related to the vanishing of the anomalous dimension of the scalars fields . the exact expression for the subspace
is not known .
however , it is known to two loops and is given by @xmath27 in the large @xmath4 limit that we pursue in this paper , the above condition simplifies to @xmath28 we also choose to work with real @xmath12 which is possible at the one - loop level . in principle , this can be done at higher loops as well since the imaginary part of @xmath12 can be gotten rid of by redefining @xmath24 .
perturbative studies of the @xmath12-deformed theory have been carried out by several authors@xcite .
chiral primaries of these theories are known well at least in the planar ( large-@xmath4 ) case@xcite .
the ultraviolet finiteness of the @xmath12-deformed theory has been studied in ref .
@xcite and a proof of its ultraviolet finiteness at the planar level has also appeared recently@xcite . for the general ls deformations , the anomalous dimensions for chiral operators with low values of scaling dimension has been carried out ( using super - graphs ) in ref .
@xcite as well as ref .
this paper focuses on a planar one - loop computation of anomalous dimensions of single trace operators in order to systematically search for operators that are protected in the ls theory . and
@xmath29 , one has to consider @xmath30 operators .
it must be pointed out that for the @xmath12-deformed theory in the planar limit , ref .
@xcite has obtained an integrable dilatation operator and diagonalized it using a bethe ansatz . ] . in this paper , we systematically search for operators whose anomalous dimensions vanish at planar one - loop in the ls theory .
the operators are organised into representations of @xmath1 .
we find that protected operators appear in one of the three representations , @xmath31 or @xmath32 depending on the value of the scaling dimension , @xmath33 modulo three .
the paper is organised as follows . in section two ,
we present some of the background needed for the paper .
in particular , we present the f - term superpotential in component form and show that it contains double trace operators . in section three , we compute the anomalous dimensions for operators of the form @xmath34 .
we present the details of our computation as well as verify that all contributions that appear from non f - term interactions cancel and that our results are gauge independent . in section four , we compute the anomalous dimensions for operators up to dimension six in the @xmath12-deformed theory as a preliminary to the computing in the general leigh - strassler theory .
we verify that the results are consistent with expected results up to dimension six operators . in section five which contains the main results of this paper , after organising the operators using the trihedral group
, we compute the planar one - loop anomalous dimension in the leigh - strassler theory .
we present our conclusions and outlook in section six .
some technical details are relegated to the appendix for completeness .
we present here the details of the lagrangian for the general leigh - strassler deformation in component form .
it turns out that unlike the @xmath0 lagrangian , this lagrangian * can not * be written in terms of a single trace though the superfield lagrangian is written as a single trace .
there are some terms that can only be written as a double trace
these terms are however suppressed by a power of @xmath35 but can not be neglected in the large @xmath4 limit as we will see . the simplest way to see
the appearance of double trace operators is to consider trace identity ( for @xmath36 generators in the fundamental representation ) : @xmath37 notice that both @xmath38 and @xmath39 are both of the form @xmath40 for some @xmath41 .
we thus see that @xmath42 can not be written in single trace form ( using the above identity ) unless the operators @xmath41 and @xmath43 are traceless ( as in the @xmath0 limit ) . the f - term interactions ( involving bosonic fields ) for the ls theory
is given by the potential by @xmath44 and the @xmath45-deformed commutator is @xmath46_q \equiv q z_1 z_2 - \bar{q } z_2 z_1 $ ] . ]
@xmath47_q \bar{z_1}^2 - \bar{h } h ' [ \bar{z_2},\bar{z_3}]_q z_1 ^ 2 - |h|^2 [ z_2,z_3]_q [ \bar{z_2},\bar{z_3}]_q\big ) \nonumber \\ - \frac{1}{n } \big [ |h'|^2\ , \textrm{tr}(\bar{z_1}^2)\ , \textrm{tr } ( z_1 ^ 2 ) + \ h \bar{h ' } \,\textrm{tr}([z_2,z_3]_q)\ , \textrm{tr } ( \bar{z_1}^2 ) - \bar{h } h ' \ , \textrm{tr}(z_1 ^ 2)\ , \textrm{tr}([\bar{z_2},\bar{z_3}]_q ) \nonumber \\ -|h|^2\ , \textrm{tr}([z_2,z_3]_q)\ , \big ( \textrm{tr } [ \bar{z_2},\bar{z_3}]_q \big ) \big ] + \textrm{cyclic permutations}\hspace{1.65in}\end{aligned}\ ] ] the first line are the single trace operators while the last two lines are the double trace operators . when @xmath19 , one can see that the double trace terms are proportional to @xmath48 which vanishes when @xmath49 . thus the @xmath0 sym theory does not have double trace terms in its component lagrangian . note that the double trace operators also do not exist when the gauge group is @xmath50 .
since the d - terms are unaffected by the leigh - strassler deformations to the superpotential , they are the identical to the one obtained in the @xmath0 theory ( written in terms of @xmath51 superfields ) .
the detailed lagrangian is given in appendix a. the @xmath0 sym theory has a @xmath52-symmetry which is @xmath53 . in the @xmath12-deformed theory
, this is broken down to @xmath22 each of the three scalars has charge one under only one of three @xmath54 s with @xmath2 being identified with the diagonal . in the leigh - strassler theory , the @xmath22
is further broken down to @xmath55 .
the ls theory has another symmetry given by the cyclic permutation @xmath56 of the three scalar fields .
this however , does not commute with the @xmath57 .
the trihedral group , @xmath58 , is a discrete subgroup of @xmath59 that captures the essential non - abelian nature .
the centre of this group is a @xmath60 .
the @xmath2 charge which is proportional to the scaling dimension for chiral primaries becomes a @xmath20 valued charge .
specifically , in our conventions , the value of the scaling dimension , @xmath33 , modulo three is the @xmath20 charge .
historically , the appearance of a 27 parameter non - abelian discrete subgroup was first noticed in ref .
our attention to the appearance of the @xmath1 was drawn from ref .
@xcite which attributed it to s. benvenuti .
we have found the trihedral group extremely useful in organising the chiral operators .
for instance , it reduced the number of free parameters for an operator at dimension six from @xmath30 to three different operators with @xmath61 , @xmath62 and @xmath62 parameters .
one may ask whether @xmath1 remains a symmetry of the quantum theory . as argued by leigh and strassler ,
the full quantum effective potential will continue to be of the same form as the classical one .
in other words , quantum corrections renormalised the coefficients @xmath24 , @xmath25 and @xmath12 .
thus , @xmath63 will remain a symmetry of the superpotential . for @xmath1 to be a symmetry in the quantum theory , however it is also necessary that the khler potential also respects this symmetry .
the tree - level khler potential does respect the symmetry and one needs to check if this remains true at higher orders in perturbation theory .
we have not resolved this issue completely and hope to report on this in the future . since our one - loop computation makes use of only tree - level interactions ,
the use of @xmath1 is a valid one .
another useful invariance of the action is the following : @xmath64 this is _ not _ a symmetry since it acts on the couplings as well .
this however leads to restrictions on the possible renormalisation of coupling constants .
the propagators for the various fields are as follows , where feynman gauge has been chosen to write down the gauge propagator .
@xmath65 we have explicitly verified the gauge independence of our results by working in the landau gauge as well .
in @xmath9 gauge theories , it is known that holomorphy is the basis for certain non - renormalisation theorems@xcite . in order to prove properties that make use of holomorphy
, one usually works in superfields and regularisation schemes that are compatible with holomorphy . in our context
where we are computing anomalous dimensions of operators involving scalars that arise from chiral superfields , holomorphy implies that the only interaction terms that contribute to the anomalous dimension are those that arise from @xmath67-terms as we will explicitly verify . in computing the anomalous dimension of the operator @xmath68 ,
we compute the two - point function of this operator with its conjugate operator , which we denote by @xmath69 and study its singularity when the two operators are coincident .
one expects @xmath70 where @xmath33 is the naive scaling dimension of operator and @xmath71 its anomalous dimension .
thus the anomalous dimension is computed extracting the logarithmic singularities and summing over all such contributions . for the family of operators
@xmath72 , we find that , at large @xmath4 ( i.e. , in the planar limit ) , the one - loop contribution to the anomalous dimension from all interactions take the following form ( on using dimensional regularisation ) : @xmath73 when the sum of all contributions is such that the coefficient of @xmath74 vanishes , we obtain a candidate for the chiral primary .
recall that for chiral primaries , the scaling dimension is determined entirely by its @xmath2-charge and hence should receive * no * corrections . for a true chiral primary
, @xmath71 vanishes to all orders .
so the vanishing of the planar one - loop contribution to any operator does not imply that it is a chiral primary since it could obtain contributions at higher orders .
however , such operators provide us with candidates for chiral primaries .
since we are working in component form , we need to explicitly verify that all non - holomorphic contributions to the anomalous dimensions of chiral fields cancel out .
these contributions should vanish irrespective of whether the operator is a chiral primary or not .
while this is expected @xcite , we use this computation as a non - trivial check of our results .
such contributions come from three kinds of terms : * * d - term : * figures [ fig:1 ] and [ fig:2 ] arise from the d - term interaction vertex + @xmath75[z_j,\bar{z}_j]\big)$ ] . + [ z_1,\bar{z}_1]\big)$ ] .
the figure to the right schematically shows how the logarithmic divergence was extracted .
the interaction vertex is labelled by a filled - in circle.,height=124 ] + [ z_2,\bar{z}_2]\big)$],height=96 ] * * gluon exchange : * figure [ fig:3 ] indicates the contribution from the gluon - scalar interaction vertex @xmath76 + \partial_\mu \bar{z}_i [ a^\mu , z_i]\big)$ ] .
+ + this diagram is gauge dependent and is logarithmically divergent in the feynman gauge , but non - divergent in the landau gauge . *
* self - energy : * figure [ fig:4 ] indicates the contribution arising from the self - energy correction to all scalar propagators .
this one is also a gauge dependent contribution .
+ as we will see the three contribution cancel for all operators that we are considering here .
we now provide some details of this cancellation .
we have also verified that the cancellation holds in both the feynman and landau gauge though we will provide details for the feynman gauge .
[ fig:1 ] has contributions coming from the interaction term @xmath77[z_1,\bar{z_1 } ] \big)$ ] .
we evaluate the loop correction by doing one momentum integral and then taking the inverse fourier transform to get the answer in position space .
we consider the fields pairwise and find the loop correction due to the interaction .
we calculate loops involving interaction vertex separately in momentum space and then fourier transform to position space .
we multiply this with the contribution @xmath78 from the part which does not involve interaction vertex .
this is schematically explained in the diagram on the right in fig .
[ fig:1 ] for an operator of general form @xmath79 , there are @xmath80 contributions from this vertex .
but when we have fields of only one flavor , say @xmath81 , there is an additional term giving a total of @xmath82 from this interaction vertex .
however when @xmath83 , there are no contributions from this vertex . similar contributions for fields of other two flavors ensures that the combinatoric factor is symmetric in @xmath84 . hence the contribution from this diagram is @xmath85 \end{aligned}\ ] ] where @xmath86 and @xmath87 are the combinatorial factors with all the properties described above , given by ( @xmath88 is the kronecker delta function and is non - vanishing only when @xmath89 ) @xmath90 and @xmath91 the momentum integral in eqn.([anom1 ] ) is evaluated in appendix .
we find the contribution from this figure as @xmath92 \gamma^4[1 - \epsilon]}{(4 \pi)^d \gamma^2[2 - 2 \epsilon ] } \big [ \int \frac{d^dp}{(2\pi)^d } \frac{e^{i p\cdot x}}{(p^2)^{2\epsilon } } \big ] \nonumber \\ & = & \frac{g^2 n^{k+l+m+1}\big(j_{klm}+g_{klm } \big)}{256 \pi^6 + 3 \log|x|^2 \big ) \ , \end{aligned}\ ] ] where @xmath93 is euler constant .
[ fig:2 ] involves the d - term interaction@xmath94 [ z_2,\bar{z_2 } ] \big)$ ] and similar terms obtained by cyclic permutation of flavor indices .
here we notice that when the operator has fields of all three flavors the combinatorial factor must be 3 .
when the there are fields of two flavors this factor must be 2 .
there is no such interaction when there are only fields of single flavor and hence there is no contribution from this diagram .
the integral to be evaluated is the same as in fig .
[ fig:1 ] .
the contribution from this interaction vertex is @xmath95 the interaction vertex is @xmath96+\textrm{c.c}$ ] .
the calculation of corrections is again done by taking propagators pairwise , as in the previous cases .
we realise that out of the different types of contractions possible only two are giving rise to any divergence .
to find the combinatorial factor for fig .
[ fig:3 ] can be easily identified as the some of the combinatorial factors of the above two diagrams .
the net contribution from this diagram is @xmath97 this contribution arises out of the one - loop correction to the scalar propagator @xmath98 .
the calculation of one - loop corrected scalar propagator is given in the appendix .
this correction is represented by the blob in fig .
[ fig:4 ] . here
again we calculate the contribution from fig .
[ fig:4 ] by taking lines pairwise , where one of the two lines has the blob .
( this is needed to match the numerical factors in the calculation of other diagrams . )
this blob can appear on any of the @xmath99 lines .
taking into consideration that the gauge group is @xmath36 the combinatorial factor from fig .
[ fig:4 ] is seen to be @xmath87 . multiplying this by a factor of @xmath100 from the rest of the @xmath101 lines , the one - loop correction to scalar propagator is obtained in momentum space from eqn .
( [ betloop ] ) as @xmath102}{(4 \pi)^{2-\epsilon } ( p^2)^{1+\epsilon}}\end{aligned}\ ] ] inserting this into the fig .
[ fig:4 ] and calculating the loop @xmath103}{(4 \pi)^{2-\epsilon } } \int \frac{d^dp}{(2 \pi)^d } \frac{1 } { ( p - q)^2 ( p^2)^{1+\epsilon } } \nonumber \\ & = & -2 \frac { n^3 \big(|h|^2 + |h'|^2/2 \big ) \gamma(\epsilon ) \gamma(2 \epsilon ) b[1 - 2 \epsilon , 1 - \epsilon ] b[1 - \epsilon , 1 - \epsilon]}{(4 \pi)^{4 - 2 \epsilon}\ b[1 , 1 + \epsilon]\ \gamma(2 + \epsilon)\ ( q^2)^{2\epsilon}}\end{aligned}\ ] ] the @xmath104 factor arises when we contract the @xmath105 with generators coming from the operator . in the above
, the momentum integral is again evaluated using feynman parametrisation and then fourier transformed to position space to obtain @xmath106 together with the rest of the lines in fig .
[ fig:4 ] , which gives a factor of @xmath107 multiplying it , the total contribution of fig .
[ fig:4 ] is @xmath108 we can see that the coefficients of @xmath109 ( as well as that of @xmath110 ) in eqns.([d1 ] ) , ( [ d2 ] ) , ( [ d4 ] ) , ( [ d3 ] ) add up to zero .
in particular , the term involving @xmath86 appears only from the contributions from fig .
[ fig:1 ] and [ fig:2 ] and they cancel .
the term involving @xmath87 adds up to give a term proportional to @xmath111 , which is proportional to the beta function and hence vanishes in the conformal limit .
hence the only contribution to the required correlator comes from the f - term interaction as expected .
the computation of the anomalous dimension for quadratic operators such as @xmath112 and @xmath113 differs from all other values of @xmath114 . postponing the details for quadratic operators , in the following subsection we will exclude values of @xmath82 , @xmath115 , @xmath116 where @xmath117 .
the contribution from the f - term is obtained from a diagram similar to the one given in fig .
[ fig:2 ] .
the interaction vertices involved are @xmath118_q [ \bar{z_1},\bar{z_2}]_q \big)$ ] , @xmath119 and its cyclic permutations , where @xmath46_q = q z_1 z_2 - \bar{q } z_2 z_1 $ ] .
the combinatorics and integrals are the exactly as described earlier .
in addition , here , when @xmath120 , contributions involving the parameter @xmath45 appear .
the factor @xmath121 is introduced to take this into account .
the contribution to the above two - point correlator is @xmath122\nonumber \\ & \times & \frac{n^{k+l+m+1}}{256 \pi^6 |x|^{2(k+l+m ) } } \big ( \frac{1}{\epsilon } + 1 + 5 \gamma_e + 3 \log ( \pi ) + 3 \log|x|^2 \big ) \end{aligned}\ ] ] where @xmath123 the vanishing of the anomalous dimension now gives the condition @xmath124 before looking for solutions in full generality , let us first consider the @xmath0 limit , i.e. , @xmath49 and @xmath19 . in this limit , we obtain @xmath125 this has two non - trivial solutions : 1 .
@xmath126 , @xmath127 and permutations thereof ; 2 .
@xmath128 , @xmath129 , @xmath130 and permutations thereof .
\(i ) corresponds to operators of the form @xmath131 , and ( ii ) corresponds to operators of the form @xmath132 .
all these are the known @xmath133 4 chiral primary operators of the form we considered which are also chiral primaries . as mentioned earlier , the general formula given in eqn .
( [ anomcond ] ) is not valid for these operators since there is an extra contribution appearing in the deformed theory .
this will be discussed in the next subsection . ] .
we next consider the @xmath12-deformed theory which corresponds to keeping @xmath19 and restoring arbitrary values for @xmath45 .
the vanishing of the anomalous dimension is now @xmath134 among the two classes of solutions that we obtained in the @xmath0 limit , we see that those of type ( i ) continue to have vanishing anomalous dimension since @xmath121 and @xmath86 vanish separately for those values of @xmath114 .
however , this is no longer true for the operators of type ( ii ) .
these operators have the charges given in the list of chiral primaries given by lunin and maldacena for the @xmath12-deformed theory@xcite .
finally , we now consider the leigh - strassler theory , where @xmath135 as well .
none of the @xmath0 chiral primaries are protected in this theory . however , in the limit @xmath136 , the operator @xmath137 ( and also @xmath138 ) is found to be a solution of the eqn .
( [ anomcond ] ) , which is easy to understand as this operator can not get any contribution at one loop from the @xmath25 interaction .
we will now discuss the operators @xmath139 and @xmath140 that were not considered earlier .
we will see that both these operators are protected in the leigh - strassler theory . the anomalous dimension for dimension two operators obtains contributions from interactions involving double trace operators that appear in @xmath142 . in the @xmath12-deformed theory ,
this interaction only affects the @xmath112 operator , while in the general ls theory , the operator @xmath139 is affected by the @xmath25 dependent double trace operator . for all other operators , one finds that interactions involving double trace operators provide contributions that are suppressed by a factor of @xmath35 relative to the single trace interactions and thus can be ignored in the large @xmath4 limit .
the computation for these operators differs from the above due to a subtlety in taking the large @xmath4 limit .
the deformed theories have an extra interaction , as seen from eqn .
( [ beta1 ] ) , which is suppressed by a factor of @xmath4 relative to other interactions as it is a multi - trace operator and @xmath67 .
thus , the bosonic potential ends being a double trace which can be rewritten as a single trace using identities such as eqn .
( [ trform ] ) .
this term does * not * appear for @xmath50 as well .
note also that it vanishes in the @xmath0 limit . ] . for a dimension two operator ,
the trace algebra works out as follows , @xmath143^{2 } = \frac{(n^{2}-1)^{2}}{n } \sim n^{3}\end{aligned}\ ] ] the @xmath144 contribution is seen to be of the same order as the one from the single trace interaction piece in eqn .
( [ beta1 ] ) .
this can be ignored in the large @xmath4 limit while computing anomalous dimension for operators of dimension @xmath145 .
the important point to note is that this contribution is precisely the one that makes the anomalous dimension for @xmath146 vanish ( as has already been shown by others using different methods . ) we have seen that in the @xmath12-deformed theory , at one - loop , the family of operators of the form @xmath147 and @xmath148 are protected . for the leigh - strassler theory , we have only two kinds of operators which survive on including the @xmath25 deformation .
they are @xmath149 and @xmath140 . in order to further generalise the kinds of operators that one must consider , we revisit the chiral primaries of the @xmath0 theory .
we have found chiral primaries that involve only two flavors of the scalars .
what about those involving three flavours
? the simplest one will involve one of each flavor .
the @xmath0 chiral primary is @xmath150 more generally , chiral primaries of @xmath0 sym are obtained by considering linear combinations of all possible orderings of operators .
for instance , the above operator has @xmath151 possibilities .
the @xmath152 is the order of the permutation group in @xmath153 objects and the division by @xmath153 reflects the cyclic property of the trace . given a monomial , @xmath154 , the corresponding @xmath0 primary is given by the expression ( with @xmath155 ) @xmath156 where we sum over all permutations @xmath157 and @xmath158 is a symmetry factor@xcite .
thus , there is a one - to - one correspondence between monomials and @xmath0 chiral primaries .
thus , at dimension @xmath33 , the number of chiral primaries is @xmath159 which is the number of monomials at degree @xmath33 .
based on the form of f - term equations like @xmath160 , freedman and grsoy(fg ) have argued that one needs to associate a factor of @xmath161 for terms that are related by the exchange of @xmath162 and @xmath163 .
for instance , the chiral primary involving all three flavors , will become @xmath164 in the @xmath12-deformed theory by their prescription
. we will refer to this as the fg prescription . in the sequel
, we will verify the fg prescription works for operators involving up to six powers of the scalars only when they turn out to be chiral primaries .
chiral primaries in the @xmath12-deformed theory are classified by three charges corresponding to a @xmath22 subgroup of the @xmath165 r - symmetry in the @xmath0 theory .
the three scalars @xmath81 , @xmath162 and @xmath163 have charges @xmath166 , @xmath167 and @xmath168 respectively .
chiral primaries are thus labelled by their @xmath22 charges . here
we consider two - point functions of operators with charges @xmath169 : @xmath170 , @xmath171 , @xmath172 , @xmath173 and @xmath174 these are all the operators with @xmath175 with all @xmath176 non - vanishing . for a given choice of @xmath169 ,
there are several operators that carry this charge .
for example , there are three operators with charge @xmath170 as shown below .
we choose linear combinations of all such operators in two steps .
first , we obtain the corresponding @xmath0 primary .
second , we introduce powers of @xmath177 following the fg prescription@xcite . this is potentially a candidate for a chiral primary .
we then put in arbitrary coefficients in front of all operators to make our ansatz more general .
we then compute the anomalous dimensions of these operators at planar one - loop and obtain the condition for the vanishing of their anomalous dimensions . in the following
, we will see that the planar one - loop contribution to the anomalous dimension for all operators can be written as the sum of the absolute squares .
this enables us to solve the equations easily without any hidden assumptions .
we take the chiral primary to be of the following form @xmath179 the computation of the anomalous dimension of this composite operator proceeds as before .
the vanishing of the anomalous dimension is given by the condition matrix of anomalous dimensions from the following expression .
it may have some use in writing out the hamiltonian for the spin - chain but its use is limited due to the small length of the chain . ]
@xmath180 \big\ } = 0\ ] ] the condition can be rewritten as follows : @xmath181 this is a sum of three positive definite terms and is solved by @xmath182 and @xmath183 which makes it a chiral primary only for the @xmath184 theory . here
the chiral operator is taken to be @xmath186 there is an ambiguity in applying the fg prescription .
for instance , the operator with coefficient @xmath187 can be associated with either @xmath161 ( as we have chosen ) or @xmath188 .
this ambiguity disappears when @xmath189 .
the condition for the vanishing of the planar one - loop anomalous dimension is @xmath190 = 0\ .\end{aligned}\ ] ] the above expression can be written as follows : @xmath191 this has a solution only when @xmath192 and @xmath193 .
this is precisely the situation where the fg prescription works . when @xmath49 , this is @xmath0 chiral primary .
when @xmath194 , this operator is also a protected operator .
this is again an result expected from lunin and maldacena@xcite this is a chiral primary in the @xmath195 orbifold of @xmath0 theory . for all other values of @xmath12 ,
this operator is not a chiral primary . for the operator @xmath197
the condition for the vanishing of the one - loop anomalous dimension is @xmath198 \nonumber \\ & & + 4 |b|^2 + 3 |c|^2 + 5 |d|^2 + 5 |f|^2 + 4 |g|^2 + 3 = 0\ .\end{aligned}\ ] ] this can be written as the sum of squares as follows : @xmath199 the solution occurs only when @xmath200 and @xmath201 which is the known @xmath202 chiral primary .
thus , this is not a chiral primary for generic values of @xmath12 .
we next consider the operator with charge @xmath173 .
below the powers of @xmath45 have been assigned using the fg prescription
. however , there is an ambiguity in assigning the powers of @xmath177 .
for instance , the operator multiplying the coefficient @xmath204 can be assigned either @xmath205 or @xmath188 since it can be reached by two different set of exchanges .
this ambiguity however goes away when @xmath206 .
@xmath207 the vanishing of the anomalous dimension at one - loop is @xmath208 = 0 \ .\ ] ] again , this can be written as the sum of absolute squares .
@xmath209 clearly this has a solution @xmath210 only when @xmath206 .
note that this is precisely the value of @xmath45 , where the ambiguity in the fg prescription is removed . for the operator @xmath212 we obtain the following condition for the vanishing of the anomalous dimension @xmath213 = 0 \ .
\end{split}\ ] ] this is independent of @xmath214 and is solved by @xmath215 as can be clearly seen after rewriting the above expression in terms of sums of absolute squares .
@xmath216 hence this is a chiral primary for any value of @xmath12 on implementing the fg prescription . by studying the non - renormalisation properties of operators up to dimension six
, we see that for generic @xmath12 , chiral primaries appear only as operators of the form @xmath217 and @xmath218 ( other than the quadratic operators ) as expected from the refs .
further , the fg prescription works for these operators .
the absence of an ambiguity in implementing the fg prescription seems to be the key to the vanishing of the one - loop anomalous dimension .
this also picks out the special values of @xmath45 for which some operators are protected .
the general leigh - strassler deformation is invariant under the action of the trihedral group @xmath1 which is a finite non - abelian subgroup of @xmath219 .
the centre of this group is a @xmath20 which is sub - group of @xmath2 .
thus , the @xmath2 charge can be identified with the @xmath20 charge .
chiral primaries in this theory must appear as irreducible representations of @xmath1 . in appendix b ,
we have provided relevant details of the irreducible representations of @xmath1 . based on the representation theory ,
we obtain the following important and useful result : 1 . when the scaling dimension , @xmath220 mod @xmath153 , then chiral primaries _ must _ appear in any one of the nine one - dimensional representations , @xmath221 ( @xmath222 ) .
the representation @xmath223 corresponds to a singlet of @xmath1 .
we will label such operators @xmath224 to indicate the representation they belong to .
the charge @xmath225 for one - dimensional representations can be identified with the charge proposed in @xcite .
2 . when the scaling dimension , @xmath226 mod @xmath153 ( @xmath227 ) , then chiral primaries appear in the three - dimensional representation , @xmath228 and thus three operators form a triplet .
we label all such operators by @xmath229 .
given one operator of the triplet , the other two can be generated by the cyclic replacement @xmath230 .
this observation is useful in many ways .
there will be no mixing between operators which sit in distinct representations of @xmath1 .
this leads to a nine - fold reduction in the operators that one needs to consider for one - dimensional representations and a three - fold reduction for the three - dimensional representations .
since @xmath220 mod @xmath153 , one has to only consider the one - dimensional representations
. there are three operators with @xmath233 and we will consider the most general linear combination of them . @xmath234
the vanishing of the one - loop correction to the anomalous dimension is given by @xmath235 this can easily be seen as equivalent to @xmath236 this has two solutions : * @xmath237 , @xmath238 and @xmath239 .
this implies that the @xmath240 which was a chiral primary in the @xmath12-deformed theory is protected at one - loop in the ls theory as well . *
@xmath241 , @xmath242 , @xmath243 .
this is the operator @xmath244\ , \textrm { where } \
m\equiv \frac{\bar{h}'}{\bar{h}}\ .\ ] ] there are two other operators with @xmath29 and @xmath245 these are @xmath246\ .\ ] ] these are descendants that there are nine descendants ( with @xmath247 ) , one in each of the irreps , @xmath221 .
when , @xmath233 , there are three operators and one descendant while there are one operator in other sectors .
we obtain two protected operators in the @xmath247 sector which is consistent with this counting . ] and hence are not chiral primaries . for this operator @xmath249 the vanishing of the one - loop correction to the anomalous dimension is @xmath250 = 0\ ] ] when
@xmath251 , the above equation has no solution ( except in the @xmath0 limit ) implying that the operators are not chiral primaries .
they are known descendants@xcite . however , when @xmath252 , the condition becomes @xmath253 which has a solution only when @xmath254 . at all other points in the space of couplings ,
the operator is a descendant .
the operator that we will consider here is in the three - dimensional representation , @xmath256 , of @xmath1 .
below , we will consider only one operator in the triplet since the result is valid for all three operators .
the operator has charge @xmath257 .
further , it is a linear combination of several @xmath0 primaries
the unknown coefficients are labelled to remind the reader of this fact .
for instance , below terms with coefficient using the same letter of the alphabet are part of the same @xmath0 primary . @xmath258 requiring the vanishing of the one - loop correction to the anomalous dimension , we get , @xmath259 = \ 0\end{aligned}\ ] ] the above equation is rather hard to analyse and one may wonder if it has a solution .
we now make use of the fact that in the limit @xmath19 , these equations should provide us conditions that appeared in the @xmath12-deformed theory .
we have already seen that these can be written as the sum of squares . using this result as input ( and a check ! ) , we deform the @xmath19 term suitably such that all terms that appear as @xmath260 that appear above are accounted for .
this strategy works rather well and we obtain an expression ( given below ) that is easily analysed .
@xmath261 this equation has the following definite solution providing us the required chiral primary operator at one - loop planar level .
@xmath262 where @xmath263 .
we thus find only one protected operator that exists for generic values of the couplings .
we consider dimension six operators which are in the one - dimensional representation of @xmath1 .
this is an example where the condition that the operator be in an irrep of @xmath1 ( rather than an abelian subgroup as considered in @xcite , for instance ) leads to a simplification .
there is a three - fold reduction in the number of constants in the problem . @xmath265
given the complexity of the expression for the anomalous dimension , we used the same strategy that was employed for the @xmath266 operator .
this enabled us to re - express the anomalous dimension as the sum of absolute squares as given below .
operators take a similar form .
it is obtained by the the following replacements on the @xmath252 expressions : @xmath267 , @xmath268 , @xmath269 . and @xmath270 .
note that if any of the coefficients appears without being multiplied by @xmath116 , it is left unchanged .
it turns out these equations do not have a solution indicating that they are all descendants . ]
@xmath271 trying to solve the constraint equations arising from the above , we can see that there are no generic state satisfying them .
thus , there are no protected operators for generic values of couplings . however ,
at specific sub - loci in the coupling space there are solutions .
these belong to several branches which we list below : branches ( i ) and ( ii ) are connected to the operators that appear when @xmath189 in the @xmath12-deformed theory .
branch ( v ) degenerates to a linear combination of @xmath0 primaries when @xmath272 . branches ( iii ) and ( iv ) do not have such a limit .
* @xmath273 , with @xmath24 , @xmath25 arbitrary .
the solution is given by @xmath274 , @xmath275 , @xmath276 ; @xmath277 , @xmath278 , @xmath279 and @xmath280 .
* @xmath200 , @xmath281 .
the solution is given by the choices @xmath282 , @xmath283 , @xmath284 and @xmath285 .
* @xmath286 .
the solution is given by @xmath287 , @xmath288 , @xmath289 , @xmath290 , @xmath291 and @xmath292 .
* @xmath293 .
the solution is given by @xmath294 , @xmath295 , @xmath296 , @xmath297 , @xmath298 , @xmath299 , @xmath300 and @xmath301 .
* @xmath302 , @xmath303 and @xmath304 .
we now consider operators with @xmath3 and @xmath29 .
there are three operators in the representations , @xmath305 that we need to consider .
we first consider the operator in the representation @xmath306 it consists of @xmath30 terms however , the number of independent coefficients is reduced to @xmath61 due to the use of the trihedral symmetry .
we write the operator as the sum of four terms @xmath307 where @xmath308 and by @xmath309 we mean the operator obtained by the cyclic replacement @xmath310 in all the terms .
the operators in the representations @xmath305 ( @xmath251 ) are given by @xmath311 after some long and rather tedious algebra , one can express the vanishing of the one - loop anomalous dimension for the operator @xmath312 as the sum of absolute squares .
@xmath313 the above equations lead to @xmath314 equations in @xmath61 unknowns .
we find that there are precisely _ two _ solutions that exists for generic values of the parameters .
it is easy to see that there are some identifications amongst the @xmath315 .
they are @xmath316 , @xmath317 , @xmath318 and @xmath319 .
this reduces the number of unknowns to @xmath320 .
the explicit solutions are somewhat unilluminating and will not be presented here .
an important point is that the two solutions can be characterised by @xmath237 and @xmath227 . in the limit , @xmath321 ,
these two solutions reduce to the operators that exist in the @xmath12-deformed theory for general values of @xmath12 .
this is precisely what happened for the @xmath247 , @xmath29 operators as well .
this is a clear indication that the operators that are protected at one - loop in the @xmath12-deformed theory and are in the representation @xmath306 survive the @xmath25 deformation .
the operators in the representation @xmath305 with @xmath251 are however not protected operators .
we have shown that it is useful to organise chiral primaries in terms of representations of the discrete group @xmath1 .
in particular , we have seen that operators appearing only in one of the three representations , @xmath306 and @xmath32 are protected at planar one - loop level .
we conjecture that this result is true in general . the general pattern for operators protected at planar one - loop ( and possibly beyond ) organised in terms of the trihedral group is given in the table below when @xmath322 .
( we have excluded the quadratic operators since they have a somewhat different behaviour . ) [ cols="<,^,^,^",options="header " , ] the first column is only a reorganisation of the well - understood @xmath323 primaries into representations of @xmath1@xcite .
the second column follows from the lunin - maldacena prediction that chiral primaries in the @xmath12-deformed theory , for generic values of @xmath12 , arise only with charges @xmath217 and @xmath218 rewritten in terms of representations of @xmath1@xcite .
the last column is based on our computations in the ls theory and has been verified up to and including scaling dimension six .
further , we have seen that in other representations , operators are only protected in a submanifold in the space of couplings .
these submanifolds consist of several branches , some of which do not intersect the subspace of @xmath12-deformed theories .
in this paper , we have studied the planar one - loop contribution of operators in the ls theory for operators up to dimension six .
we have used the trihedral group to classify the operators and this has lead to a significant simplification to the problem .
we find that for generic values of couplings , the protected operators arise in the one - dimensional representation , @xmath306 when @xmath220 mod @xmath153 and in the three - dimensional representations @xmath32 when @xmath226 mod @xmath153 ( @xmath324 ) .
we conjecture that this is true in general .
it is interesting to see if there is a simple proof of this statement .
the leigh - strassler superpotential makes an interesting appearance in a different context .
a recent computation all - orders perturbative computation of the effective superpotential for the so - called long - branes on the cubic torus using the topological landau - ginzburg model turns out to be of the leigh - strassler form@xcite .
it is possible that this computation may be related to the quantum effective superpotential of the ls theory .-
deformed theory using a relationship with matrix models .
it is also of interest to see if these two effective potentials are related .
] we are pursuing the relationship of this work to the ls theory@xcite .
in particular , even if a direct map does nt exist , it suggests a re - ordering of the perturbative computation of the quantum effective superpotential for the ls theory and that the renormalised coefficients ( up to an overall normalisation ) should be expressible in terms of theta functions of characteristic three .
this statement is modulo the effect of the the chiral konishi anomaly which may modify the statement .
an open question is to find the gravity duals for ls theories .
a more limited question is to ask whether one can find special values of the couplings like the case of rational @xmath12 in the @xmath12-deformed theory . the crucial input in finding the gravity duals whenever @xmath12 was rational is the realisation that the effect of discrete torsion in abelian orbifolds is to @xmath45-deform the @xmath0 superpotential@xcite .
one may ask whether discrete torsion in non - abelian orbifolds could also produce the @xmath25 deformation .
the naive answer based on adapting the analysis of ref .
@xcite to include discrete torsion is that no such couplings can arise .
however , since those results are based on ` dimensional reduction ' , it would be interesting to actually carry out a cft computation in string theory to verify that such terms are not generated to come up with a no - go theorem
. one of us , km , would like to thank justin david for several helpful discussions throughout the work and hri , allahabad for hospitality during an extended visit during the initial part of the work .
we would also like to thank v. ravindran and n. dorey for useful discussions .
the work of km is supported by a senior research fellowship from the csir ( award no .
9/84(327)/2001-emr - i ) .
we also thank justin david , s. lakshmi bala and prasanta tripathy for a critical reading of a draft of this paper .
the lagrangian density of the leigh - strassler theory in terms of @xmath9 superfields is @xmath325 \nonumber \\\end{aligned}\ ] ] we denote the lowest component of the superfield @xmath326 by @xmath44 and its fermionic partner by @xmath327 .
the vector multiplet in the wess - zumino gauge has as components , the gauge field , @xmath328 and its superpartner , the gaugino , @xmath329 in addition to the auxiliary field @xmath330 .
all fields transform in the adjoint of @xmath36 . writing these in component fields we get the lagrangian @xmath331[z_j,\bar{z_j } ] + i \sqrt{2 } g\ \bar{\psi_i } [ z_i,\bar{\lambda } ] + i \sqrt{2 } g\ \psi_i [ \bar{z_i},\lambda ] \nonumber \\ & - i h\ \ \bar{\psi_3 } [ z_1,\psi_2 ] + \ i \bar{h}\ \bar{\psi_3 } [ \bar{z_1},\psi_2 ] - i h ' z_1 \psi_1 \psi_1 + i \bar{h ' } \bar{z}_1 \bar{\psi}_1 \bar{\psi}_1 + \textrm{cyc
} \big)\nonumber \\ & - v_f(z)\end{aligned}\ ] ] where @xmath332 $ ] and @xmath333 is as given in eqn . ( [ beta1 ] ) .
below , we provide the trace identities and normalisations that we have used in our paper . @xmath334
in this appendix , we discuss the representation theory of the trihedral group @xmath1.@xcite we expect chiral primaries of the leigh - strassler deformed theory to be in irreducible representations of this group .
trihedral groups are finite subgroups of @xmath219 of the form @xmath335 , where @xmath41 is a diagonal abelian group and @xmath56 is the cyclic @xmath20 generated by @xmath336 for the trihedral group , @xmath1 , the group @xmath41 is generated by @xmath337 and @xmath338 the matrix @xmath339 with @xmath340 , a non - trivial @xmath52-th root of unity . ] . in our example
, @xmath23 turns out to be the centre of @xmath1 and is a subgroup of @xmath2 .
the irreducible representations of @xmath1 consist of nine one - dimensional representations , @xmath221 ( @xmath222 ) and two three - dimensional representations @xmath32 ( @xmath324 ) .
the charge under @xmath23 can be clearly identified with @xmath2 charge . 1 . in the one - dimensional representations , one has the following action of the generators @xmath24 and @xmath341 @xmath342 2 . in the three - dimensional representation
, one has @xmath343 for @xmath324 .
note that when @xmath237 , the above representation is reducible to a direct sum , @xmath344 , of one - dimensional representations .
the chiral superfields @xmath345 are in the representation @xmath256 while their anti - chiral partners transform in the representation @xmath346 .
we will choose our chiral primaries to be in one of these representations .
since the interactions in the leigh - strassler theory are invariant under @xmath1 , it follows that chiral operators in distinct representations of @xmath1 can not mix .
however , operators in the same representation can and do mix . on taking the commutative limit
, all our chiral primary operators reduce to polynomials in three variables , @xmath347 , where we replaced the matrices @xmath44 by scalars @xmath348 ( using the lower - case to indicate the replacement ) .
first , the triplet @xmath349 transforms in the representation , @xmath256 .
second , we can organise the polynomials by degree the degree is the ( naive ) scaling dimension of the corresponding operator which we denote by @xmath33 .
thirdly , @xmath33 mod @xmath153 , is the @xmath20 charge of the polynomial under the centre of the group ( which is generated by @xmath23 defined above ) .
polynomials in these variables of a given degree , @xmath33 , can be further organised into irreducible representations of @xmath1 .
the precise representation is decided by the value of @xmath33 mod @xmath153 .
one has the following result : * [ @xmath350 * mod * @xmath351 all polynomials can be organised in one - dimensional representations of @xmath352 , i.e. , @xmath221 .
for example , when @xmath247 the polynomials @xmath353 and @xmath354 .
in particular , there is no polynomial whose degree is @xmath355 mod @xmath153 that is in the representation @xmath256 or @xmath346 . * [ @xmath356 * mod * @xmath351 all polynomials must necessarily arise in one of the the three - dimensional representations .
in fact , defining @xmath357 mod @xmath153 with @xmath241 or @xmath358 , the polynomials must be in the three - dimensional representation @xmath32 .
for example , consider the f - term equations ( @xmath359 ) which are of degree two .
a straightforward analysis shows that the three equations are in the representation @xmath346 .
the proof of the above statements goes as follows .
note that the generator @xmath23 can be realised in terms of @xmath24 and @xmath341 as @xmath360 . using this ,
notice that for a vector @xmath361 , @xmath362 .
thus , the @xmath20 charge associated with @xmath23 is zero implying that the @xmath2 charge is zero modulo three .
similarly , for a triplet @xmath363 , @xmath364 implying that the @xmath2 charge is @xmath365 modulo three .
this leads to the following conclusion : all chiral primaries with @xmath220 mod @xmath153 must be in any one of the one - dimensional representations of @xmath1 while those where @xmath226 mod @xmath153 must arise in the three dimensional representation @xmath32 .
we evaluate dimensionally regulated momentum integrals , with @xmath366 , using the feynman parametrisation @xmath367 } \int_0 ^ 1 dx \int_0 ^ 1 dy\ \delta(1-x - y)\ x^{\alpha_1 - 1 } y^{\alpha_2 - 1 } \nonumber \\
\times \int \frac{d^dk}{(2 \pi ) ^d } \frac{1}{(x k^2 + y p^2)^{\alpha_1 + \alpha_2 } } \hspace{1.5 in } \\ = \ \frac{1}{b[\alpha_1,\alpha_2 ] } \int_0 ^ 1 dx\ x^{\alpha_1 - 1 } ( 1-x)^{\alpha_2 - 1}\times \int \frac{d^dk}{(2 \pi ) ^d } \frac{1}{(x k^2 + ( 1-x ) p^2)^{\alpha_1 + \alpha_2 } } \nonumber\end{aligned}\ ] ] where @xmath368 $ ] is the beta - function .
the fourier transform to position space is done using the following integral .
@xmath369}{4^{\epsilon}\ \pi^{\frac{d}{2}}\gamma[s ] ( x^2)^{\frac{d}{2}-s}}\ ] ] * momentum integral ( i ) * @xmath370}{(4 \pi)^{2-\epsilon } p^{2\epsilon}}\end{aligned}\ ] ] using the formula ( [ frr ] ) above we obtain the integral in eqn.([anom1 ] ) in position space as @xmath371}{4^{2\epsilon}\ \pi^{2-\epsilon}\ \gamma[2\epsilon]\ ( |x|^2)^{2 - 3\epsilon}}\end{aligned}\ ] ] the integral in eqn.([anom1 ] ) gives @xmath372}{(4 \pi)^{2-\epsilon } p^{2\epsilon } } \big)^2 \frac{\gamma[2 - 3\epsilon]}{4^{2\epsilon}\ \pi^{2-\epsilon}\ \gamma[2\epsilon]\ ( |x|^2)^{2 - 3\epsilon } } \end{aligned}\ ] ] we expand this in powers of @xmath340 and obtain the answer in eqn .
( [ d1 ] ) . * momentum integral ( ii ) * the integral in eqn .
( [ d3 ] ) using feynman parametrisation @xmath373 } \int_0 ^ 1 dx \
x^{\epsilon}\int \frac{d^dp}{(2 \pi)^d } \frac{1 } { \big(x(p - q)^2 + xp^2 \big)^{2+\epsilon } } \nonumber \\ & = & \int_0 ^ 1 dx\ x^{\epsilon } \big(q^2 x ( 1-x)\big)^{- 2 \epsilon } \frac{\gamma[2 \epsilon]}{(4 \pi)^{2-\epsilon } \
b[1,1+\epsilon ] \gamma[2 + \epsilon]}\nonumber \\ & = & \frac{\gamma[2 \epsilon ] b[1 - 2\epsilon , 1-\epsilon]}{(4 \pi)^{2-\epsilon } \
b[1,1+\epsilon ] \gamma[2 + \epsilon ] ( q^2)^{- 2 \epsilon}}\end{aligned}\ ] ] again taking the fourier transform of the above expression we get the expression in eqn .
( [ d3 ] ) .
* gluon exchange contribution to anomalous dimensions * the contribution from interaction terms @xmath374_q)$ ] and @xmath375_q)$ ] to the anomalous dimension of the operator @xmath376 is computed here .
there are different contractions giving two distinct momentum integrals .
the diagram in fig .
[ fig:5 ] can be evaluated when both the interaction vertices involve @xmath377 ( or @xmath378 ) , when one vertex involves @xmath377 and the other @xmath378 .
we work out both the cases below . : from fig .
[ fig:5 ] given here we write down the feynman integral for this case .
@xmath379 considering only the part which is divergent we get , @xmath380 we consider the @xmath82-integration first .
take @xmath381 .
after dimensionally regulating and using feynman parametrisation @xmath382}{(4 \pi)^{2-\epsilon } \gamma[2 ] ( ( q - p)^2)^{\epsilon}}\int_0 ^ 1 dx x^{-\epsilon } ( 1-x)^{-\epsilon } = \frac{\gamma[\epsilon ] b[1-\epsilon,1-\epsilon]}{(4 \pi)^{2-\epsilon } ( q - p)^{2 \epsilon}}\end{aligned}\ ] ] doing the @xmath45-integration in the same way , @xmath383 b[1-\epsilon,1-\epsilon]}{(4 \pi)^{2-\epsilon } } \int \frac{d^dq}{(2 \pi)^{d } } \frac{1}{{q}^2 \big((q - p)^2\big)^{1+\epsilon } } \nonumber \\ & = & \frac{\gamma[\epsilon ] b[1-\epsilon,1-\epsilon]}{(4 \pi)^{2-\epsilon } } \frac{1}{b[1,1+\epsilon ] } \int_0 ^ 1 dy\ y^{\epsilon } \int \frac{d^dq}{(2 \pi)^{d } } \frac{1}{\big((q - yp)^2 + p^2 y(1-y ) \big)^{2+\epsilon } } \nonumber \\ & = & \frac{\gamma[\epsilon ] b[1-\epsilon,1-\epsilon]}{(4 \pi)^{2-\epsilon } } \frac{\gamma[2 \epsilon]}{(4 \pi)^{2-\epsilon } b[1,1+\epsilon ] \gamma[2 + \epsilon ] ( p^2)^{2 \epsilon } } \int_0 ^ 1 dy\ y^{-\epsilon } ( 1-y)^{-2 \epsilon } \nonumber \\ & = & \frac{\gamma[\epsilon ] b[1-\epsilon,1-\epsilon]}{(4 \pi)^{4 - 2\epsilon } } \
\frac{\gamma[2 \epsilon ] b[1-\epsilon,1 - 2 \epsilon]}{\gamma[1+\epsilon ] \
( p^2)^{2 \epsilon } } \end{aligned}\ ] ] again using formula ( [ frr ] ) we fourier transform and expand in powers of @xmath340 to obtain @xmath384 : @xmath385 the divergent part of this is @xmath386 \nonumber \\ & = & \int \int \frac{d^dk d^dq}{(2 \pi)^{2d } } \big [ \frac{1}{(k - q)^2 ( k - p)^2 q^2 ( q - p)^2 } - \frac{1}{2 k^2 ( k - q)^2 q^2 ( q - p)^2 } \big]\end{aligned}\ ] ] the two integrals in the final expression above are already evaluated .
b[1-\epsilon,1-\epsilon ] \gamma[2 \epsilon ] b[1-\epsilon,1 - 2 \epsilon]}{\gamma[1+\epsilon ] \
} - \frac{\gamma^4[1-\epsilon ] \gamma^2[\epsilon ] } { 2\ \gamma^2[2 - 2\epsilon ] } \big ) \frac{1}{(4 \pi)^{4 - 2\epsilon } ( p^2)^{2 \epsilon } } \ ] ] fourier transforming to position space using formula ( [ frr ] ) and expanding in powers of @xmath340 , we get the value of the above integral to be @xmath388 hence there is no contribution from this integral to the anomalous dimension .
the diagrams in fig . [ fig:6 ] are the one - loop contributions to the two point function @xmath389 .
the dashed lines are the fermionic propagators and the wiggly lines are gauge boson propagators .
gauge boson interactions are calculated in landau gauge .
* contribution from scalar tadpole * due to interaction terms in @xmath333 @xmath394 \nonumber \\ & & \ \ \ \ \ \
\times \int d^dk \frac{1}{p^2 \ k^2 \
p^2 } \nonumber \\ & = & -4 ( |h|^2 + |h'|^2/2 ) \
n tr(t^a t^b ) \ \int d^dk \ \frac{1}{p^2 \ k^2 \ p^2}\end{aligned}\ ] ] * contribution from yukawa interaction * vertices @xmath398_q \big ) - i h ' z_1 \psi_1 \psi_1 + \mathrm{c.c . } \big)$ ] @xmath399 \nonumber \\ \times \int d^dk \frac{\sigma^{\mu } k_{\mu }
\ \sigma^{\nu } ( p - k)_{\nu } } { p^2 k^2 ( p - k)^2 p^2 } \hspace{3 in } \\ = - 4 ( |h|^2 + |h'|^2/2)\ n tr(t^a t^b )
\ \int d^dk \
\frac{k\cdot p - k^2}{p^2 k^2 ( p - k)^2 p^2 } \hspace{1in}\nonumber\end{aligned}\ ] ] * contribution from scalar - gluon * interaction terms @xmath400)$ ] , @xmath401)$ ] @xmath402\nonumber \\ & & + 2 ( i)^2 g^2 \ \textrm{tr}\big ( t^d t^c t^a - t^d t^a t^c \big ) \ \textrm{tr}\big ( t^d t^c t^b - t^d t^b t^c \big ) \nonumber \\ & & \ \ \ \ \ \ \ \ \ \
\int d^dk \big [ \frac { ( p - k)^2}{2 p^2 k^2 ( p - k)^2 p^2 } \big ] \nonumber \\ & & + 4\ ( i)^2 g^2 \textrm{tr}\big ( t^d t^c t^a - t^d t^a t^c
\big ) \cdot \textrm{tr}\big ( t^b t^c t^d - t^b t^d t^c \big ) \nonumber \\ & & \ \ \ \ \ \ \ \ \ \
\int d^dk \big [ \frac { p^2 - p \cdot k}{2 p^2 k^2 ( p - k)^2 p^2 } \big ] \nonumber \\ & & = \ 2 g^2 n \textrm{tr}(t^a t^b ) \int d^dk \big [ \frac{1}{2 p^2 \ k^2 \ p^2 } + \frac{1}{p^2 k^2 ( k - p)^2 p^2}\big]\end{aligned}\ ] ] summing the contributions from each diagram , we see that quadratic divergences cancel .
the one - loop correction to @xmath405 is given as @xmath406 fourier transforming we get @xmath407 where @xmath408 .
in the large n limit we have @xmath409 .
hence we should get back the one - loop correction in @xmath410 = 4 theory . from eqn .
( [ betloop ] ) above , we get , @xmath411 this expression is similar in struture with the one obtained in @xcite .
j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv .
* 2 * ( 1998 )
231 [ int . j. theor .
* 38 * ( 1999 ) 1113 ] [ arxiv : hep - th/9711200 ] .
e. witten , `` anti - de sitter space and holography , '' adv .
theor . math .
* 2 * , 253 ( 1998 ) [ arxiv : hep - th/9802150 ] .
j. a. minahan and k. zarembo , `` the bethe - ansatz for n = 4 super yang - mills , '' jhep * 0303 * ( 2003 ) 013 [ arxiv : hep - th/0212208 ] .
n. beisert , c. kristjansen and m. staudacher , `` the dilatation operator of n = 4 super yang - mills theory , '' nucl .
b * 664 * ( 2003 ) 131 [ arxiv : hep - th/0303060 ] .
s. kachru and e. silverstein , `` 4d conformal theories and strings on orbifolds , '' phys .
* 80 * ( 1998 ) 4855 [ arxiv : hep - th/9802183 ] .
a. e. lawrence , n. nekrasov and c. vafa , `` on conformal field theories in four dimensions , '' nucl .
b * 533 * ( 1998 ) 199 [ arxiv : hep - th/9803015 ] .
d. martelli and j. sparks , `` toric geometry , sasaki - einstein manifolds and a new infinite class of ads / cft duals , '' commun .
* 262 * , 51 ( 2006 ) [ arxiv : hep - th/0411238 ] .
m. cvetic , h. lu , d. n. page and c. n. pope , `` new einstein - sasaki spaces in five and higher dimensions , '' phys . rev
. lett . * 95 * , 071101 ( 2005 ) [ arxiv : hep - th/0504225 ] .
r. g. leigh and m. j. strassler , `` exactly marginal operators and duality in four - dimensional n=1 supersymmetric gauge theory , '' nucl .
b * 447 * ( 1995 ) 95 [ arxiv : hep - th/9503121 ] .
m. kulaxizi , `` marginal deformations of n = 4 sym and open vs. closed string parameters , '' arxiv : hep - th/0612160 .
o. lunin and j. maldacena , `` deforming field theories with @xmath412 global symmetry and their gravity duals , '' jhep * 0505 * , 033 ( 2005 ) [ arxiv : hep - th/0502086 ]
. m. r. douglas , `` d - branes and discrete torsion , '' arxiv : hep - th/9807235 . m. r. douglas and b. fiol , `` d - branes and discrete torsion .
ii , '' jhep * 0509 * ( 2005 ) 053 [ arxiv : hep - th/9903031 ] .
d. berenstein and r. g. leigh , `` discrete torsion , ads / cft and duality , '' jhep * 0001 * , 038 ( 2000 ) [ arxiv : hep - th/0001055 ] .
d. berenstein , v. jejjala and r. g. leigh , `` marginal and relevant deformations of n = 4 field theories and non - commutative moduli spaces of vacua , '' nucl .
b * 589 * , 196 ( 2000 ) [ arxiv : hep - th/0005087 ] .
d. z. freedman and u. gursoy , `` comments on the beta - deformed n = 4 sym theory , '' jhep * 0511 * ( 2005 ) 042 [ arxiv : hep - th/0506128 ] . s. penati , a. santambrogio and d. zanon , `` two - point correlators in the beta - deformed n = 4 sym at the next - to - leading order , '' jhep * 0510 * ( 2005 ) 023 [ arxiv : hep - th/0506150 ] . g. c. rossi , e. sokatchev and y. s. stanev , `` new results in the deformed n = 4 sym theory , '' nucl . phys .
b * 729 * ( 2005 ) 581 [ arxiv : hep - th/0507113 ] .
a. mauri , s. penati , a. santambrogio and d. zanon , `` exact results in planar n = 1 superconformal yang - mills theory , '' jhep * 0511 * ( 2005 ) 024 [ arxiv : hep - th/0507282 ] .
a. mauri , s. penati , m. pirrone , a. santambrogio , d. zanon , `` on the perturbative chiral ring for marginally deformed n = 4 sym theories , '' jhep * 0608 * ( 2006 ) 072 [ arxiv : hep - th/0605145 ] .
v. v. khoze , `` amplitudes in the beta - deformed conformal yang - mills , '' jhep * 0602 * ( 2006 ) 040 [ arxiv : hep - th/0512194 ] .
g. c. rossi , e. sokatchev and y. s. stanev , `` on the all - order perturbative finiteness of the deformed n = 4 sym theory , '' nucl .
b * 754 * , 329 ( 2006 ) [ arxiv : hep - th/0606284 ] .
s. ananth , s. kovacs and h. shimada , `` proof of all - order finiteness for planar beta - deformed yang - mills , '' jhep * 0701 * , 046 ( 2007 ) [ arxiv : hep - th/0609149 ] .
d. bundzik and t. mansson , `` the general leigh - strassler deformation and integrability , '' jhep * 0601 * ( 2006 ) 116 [ arxiv : hep - th/0512093 ] .
n. beisert and r. roiban , `` beauty and the twist : the bethe ansatz for twisted n = 4 sym , '' jhep * 0508 * , 039 ( 2005 ) [ arxiv : hep - th/0505187 ] .
m. wijnholt , `` parameter space of quiver gauge theories , '' arxiv : hep - th/0512122 .
m. t. grisaru , w. siegel and m. rocek , `` improved methods for supergraphs , '' nucl .
b * 159 * ( 1979 ) 429 .
n. seiberg , `` naturalness versus supersymmetric nonrenormalization theorems , '' phys .
b * 318 * ( 1993 ) 469 [ arxiv : hep - ph/9309335 ] .
e. dhoker , d. z. freedman and w. skiba , `` field theory tests for correlators in the ads / cft correspondence , '' phys .
d * 59 * ( 1999 ) 045008 [ arxiv : hep - th/9807098 ] .
s. govindarajan and h. jockers , `` effective superpotentials for b - branes in landau - ginzburg models , '' jhep * 0610 * ( 2006 ) 060 [ arxiv : hep - th/0608027 ] .
n. dorey , t. j. hollowood and s. p. kumar , `` s - duality of the leigh - strassler deformation via matrix models , '' jhep * 0212 * ( 2002 ) 003 [ arxiv : hep - th/0210239 ] .
s. govindarajan and k. madhu ( work in progress ) .
t. klose , `` conformal dimensions of two - derivative bmn operators , '' jhep * 0303 * ( 2003 ) 012 [ arxiv : hep - th/0301150 ] .
leng , `` the mckay correspondence and orbifold riemann - roch , '' phd thesis ( u. warwick , 2002 ) available at the url : ` http://www.maths.warwick.ac.uk/%7emiles/doctors/leng/ ` w.m .
fairbairn , t. fulton and w.h .
klink , j. math .
phys . * 5 * ( 1964 ) 1038 ; ibid .
* 23 * ( 1982 ) 1747 . | we look for chiral primaries in the general leigh - strassler deformed @xmath0 super yang - mills theory by systematically computing the planar one - loop anomalous dimension for single trace operators up to dimension six .
the operators are organised into representations of the trihedral group , @xmath1 , which is a symmetry of the lagrangian .
we find an interesting relationship between the @xmath2-charge of chiral primaries and the representation of @xmath1 to which the operator belongs .
up to scaling dimension @xmath3 ( and conjecturally to all dimensions ) the following holds : the planar one - loop anomalous dimension vanishes only for operators that are in the singlet or three dimensional representations of @xmath1 . for other operators ,
the vanishing of the one - loop anomalous dimension occurs only in a sub - locus in the space of couplings .
hep - th/0703020 + iitm / ph / th/2006/17 + may 2007(v3 ) kallingalthodi madhu and suresh govindarajan + department of physics + indian institute of technology madras + chennai 600036 india + |
it is generally accepted that pulsars are produced from massive progenitor stars with a masses of @xmath2 and typical radii @xmath3 , collapsing at the end of their evolution to neutron stars in an explosion , which ejects the outer envelope forming a supernova remnant .
a pulsar is a rotating neutron star ( baade & zwicky 1934 ) with a mass of @xmath4 , a radius of @xmath5 km , and a rotational period typically of order @xmath6 s. pulsars are slowing down and loosing their rotational energy with the rate @xmath7 where @xmath8 is the measured rate of a pulsar slow - down , and @xmath9 is a generic moment of inertia of a pulsar . assuming that the pulsar releases its mechanical energy in form of magnetic dipole radiation ( ostriker & gunn 1969 ) one can estimate the pulsar surface magnetic field
@xmath10 which appears to be very strong and for a typical pulsar is of the order of @xmath11 . assuming a constant magnetic field throughout the pulsar lifetime , the characteristic age of a pulsar can be determined from its rotational period and slow - down rate @xmath12 the @xmath0-ray luminosity of a pulsar , @xmath13 , can be roughly estimated using the measured flux of high - energy @xmath0-rays , @xmath14 : @xmath15 where @xmath16 is the solid angle into which the pulsar beams and @xmath17 is the distance to the pulsar .
the @xmath0-ray luminosity is widely supposed to be proportional to the total loss of pulsar rotational energy @xmath18 where @xmath19 is the efficiency of converting the rotational energy of neutron star into @xmath0-rays . on the other side
the @xmath0-ray flux of a particular pulsar can be estimated as @xmath20 the conversion efficiency , @xmath19 , could be , in zero approximation , assumed as a phenomenological function of pulsar age , even though this function is not unique and not universal .
the observed @xmath0-ray flux depends at first on the size of the pulsar beam .
finally the spin - down powered pulsars can be characterized by a set of five parameters @xmath21 , which allow to predict approximate @xmath0-ray luminosity of a particular pulsar for a given @xmath22 and @xmath19 .
more accurate calculations of @xmath0-ray fluxes , light - curves and energy spectra can be performed using elaborated models of spinning down pulsars .
there are two major classes of models dealing with high - energy @xmath0-ray emission from the pulsars .
they differ primarily by the initial assumptions on where the particle acceleration and radiation take place , either near the surface of neutron star at its _ polar caps _ or in the _ outer gaps _ , which may occur in the outer magnetosphere of a pulsar at relatively large distance from the surface of the neutron star ( see figure 1 ) .
_ polar cap model_. a fast spinning , magnetized neutron star can generate a huge electric field , which induces an electrostatic potential @xmath23 right above the polar caps of a pulsar .
as suggested by goldreich & julian ( 1969 ) the corresponding electrostatic force may overcome gravity and pull out the charged particles from the surface of the neutron star , forming the current given by @xmath24 .
in addition charged particles may escape from the surface of the neutron star because its surface temperature @xmath25 ( becker & trmper 1997 ) may exceed the electron thermal emission temperature .
the electrons can be accelerated in the regions above the poles where the open magnetic field lines , where @xmath26 is an angular velocity of a pulsar . ]
originate ( sturrock 1971 ; ruderman & sutherland 1975 ) . those electrons may emit high - energy @xmath0-rays in three major processes : _
( i ) _ synchrotron emission induced by electrons spinning around the magnetic field lines ; _
( ii ) _ curvature radiation generated by the electrons moving along magnetic field lines ; _ ( iii ) _ inverse compton emission produced by high - energy electrons , which up - scatter low energy photons to @xmath0-ray energies . the @xmath0-rays interact with the pulsar magnetic field ( @xmath27 ) initiating the pair cascade ( harding 1981 ; daugherty & harding 1982 ) .
the pair formation fronts occur at 0.5 - 1 stellar radii above the surface of the neutron star where the magnetic field is rather strong ( for a typical pulsar @xmath28 ) . for such magnetic fields only the @xmath0-rays below a certain energy ( a few gev for typical pulsars ) may freely escape from the pair production region .
the super - exponential cut - off in the high - energy @xmath0-ray spectrum is an intrinsic feature the of polar cap model . _
outer gap model_. a vacuum gap may occur in the outer magnetosphere of a pulsar ( cheng , ho & ruderman 1986 ; romani 1996 ) , which is limited by the null charge surface @xmath29 and the light cylinder .
charged particles can flow out through the light cylinder , whereas they could not run through the null charge surface .
the @xmath0-rays ( curvature or inverse compton photons ) penetrating into the vacuum gap can produce a pair ( @xmath30 ) by interacting with non - thermal x - rays or infrared photons , which fill the gap .
the secondary electrons can be accelerated in the vacuum gap electric field and emit synchrotron and curvature @xmath0-ray photons .
the @xmath0-rays of energy at least up to 10 gev may escape from the vacuum gap .
this model is suggestive of pulsed tev @xmath0-ray emission induced by the inverse compton up - scattering of infrared photons by high - energy electrons ( romani 1996 ) .
@xmath31 r is a rank given according to the derived value of @xmath22 .
@xmath32 was seen only with comptel ( kuiper et al . 1999 ) up to an energy of about 30 mev , and not with egret .
even though both of the models can fit the variety of light - curves of radio pulsars as well as the photon spectra over the optical , radio , x - ray , and high - energy wavelengths , they inevitably predict very different spectra at energies above 10 gev .
the harder spectra of high - energy @xmath0-rays are more plausible for outer gap model , whereas such spectra are almost excluded in the polar cap model ( see harding 2000 ) except for millisecond pulsars .
future pulsar observations from the ground with erenkov telescopes will allow to measure the spectral shape of a pulsed emission in the relevant energy range ( above 10 gev ) , which will be tightly constraining the choice of an appropriate model of high - energy @xmath0-ray emission .
there are a few other models , which do not fit into the two classes mentioned above , for instance a _ neutron star wind model _ suggested recently by kirk , skjaeraasen , gallant ( 2002 ) .
the energy of the pulsar magnetic field may be converted into energy of accelerated particles within the magnetic field reconnections , which do occur at the null current sheet extending rather far from the surface of the neutron star beyond the light cylinder .
spectra of @xmath0-ray emission computed using such a model might be challenging future observations at high - energies from the ground .
@xmath33 see for details de jager et al .
( 2001 ) spectral references from macomb & gehrels ( 1999 )
the discovery of the first radio pulsar @xmath34 ( hewish et al .
1968 ) has triggered further large - scale search for radio pulsars , which brought us a lot more detections .
these days the catalogs of radio pulsars account for more than 1400 objects .
about 50 of those have been seen in x - rays , but only 7 at high energies ( @xmath35 mev ) ( kanbach 2002 ) , recently observed with egret - ray telescope on board _ compton gamma ray observatory ( cgro ) _ ] .
it is remarkable that most of the egret pulsars have got a top rank when they are listed according to the value of @xmath22 ( see table 1 ) .
most of the high - energy @xmath0-ray pulsars have been seen at radio wavelengths , except geminga ( @xmath36 ) , which appears to be radio - quiet .
the light - curves of pulsed high - energy @xmath0-ray emission generally do not resemble the light - curves of radio emission , even though they can be rather similar , e.g. for the crab pulsar ( @xmath37 ) .
three pulsars , crab , vela ( @xmath38 ) , and geminga , have a double - peaked light - curve with a substantial contribution of a bridge ( inter - peak ) emission
. the peculiar behavior of the light - curves reveals a complicated geometry of the pulsed emission , which is evidently not always similar to the light - curves at other wavelengths ( for detailed discussion see thompson 2000 ) .
six @xmath0-ray pulsars show a clear evidence of the pulsed emission above 5 gev .
spectral energy distributions ( sed ) of @xmath0-ray emission from the high - energy pulsars ( @xmath39 ) peak in the hard x - ray or @xmath0-ray range . the sed breaks at gev energies , which is an established distinctive feature of @xmath0-ray emission from the pulsars , which was predicted by the polar cap model .
the energy spectra of egret pulsars can be well fitted by power laws with spectral indices within the range @xmath40=1.42.1 plus a super - exponential cut - off at the energy about @xmath41 ( de jager et al .
2001 ) @xmath42 where @xmath43 is in units of gev .
parameters of the fit for six egret pulsars are summarized in table 2 .
the normalization constant @xmath44 in eqn.([fit ] ) corresponds to the photon flux at energy of 1 gev , and is an approximate measure of the pulsar luminosity in high - energy @xmath0-rays , @xmath45 .
one can expect that @xmath45 correlates with the total loss of the rotational energy of a pulsar ( @xmath22 ) . in figure 2 the scatter plot of @xmath44 over @xmath22 for six gev egret high - energy @xmath0-ray pulsars is shown .
even though the current sample of high - energy @xmath0-ray is very poor one can see an evident trend , which can be fitted as @xmath46 .
pulsar luminosity in high - energy @xmath0-rays is determined not only by the total loss of its rotational power but in addition by the efficiency of energy conversion from slow - down into high - energy @xmath0-rays ( see eqn.([epsi ] ) ) .
in fact , even a pulsar with a rather low value of @xmath22 and low rank ( see table 1 ) could appear to be a powerful @xmath0-ray emitter .
for instance @xmath0-ray pulsar @xmath47 has rather low rank 29 .
naturally the efficiency @xmath48 could be a function of pulsar age . the efficiency of energy conversion from rotational slow - down into high - energy @xmath0-rays can be determined at most by the increase of current flow out of the neutron star for elder pulsars ( harding 1981 ) , or by the corresponding increase in size of a particle acceleration region in the magnetosphere around the star ( ruderman , cheng 1988 ) .
the measure of conversion efficiency as a function of pulsar apparent age is shown in figure 3 .
the general tendency can be approximated by @xmath49 . for a better understanding of pulsar workings a substantially more representative sample of @xmath0-ray pulsars
is needed .
future observations from space with glast and from the ground with future erenkov instruments may significantly increase the sample of @xmath0-ray pulsars .
de jager et al ( 2001 ) have estimated the minimal exposure for a 5@xmath50 detection of egret high - energy @xmath0-ray pulsars .
pulsar @xmath51 appeared to be the best candidate for the future observations with h.e.s.s .
( for a description of the experiment , see hofmann 2001 ) .
a 10 hours exposure with h.e.s.s . will enable to resolve a pulsed high - energy @xmath0-ray emission above 30 gev from @xmath51 .
based on the outer gap model of @xmath0-ray emission ( see discussion above ) romani ( 1996 ) predicted a pulsed very high energy ( vhe ) tev emission through the inverse compton ( ic ) up - scattering of synchrotron photons on the primary electrons accelerated in the strong electric field of a pulsar .
it motivated numerous observations of pulsars with ground - based imaging and non - imaging atmospheric erenkov telescopes ( ong 1998 ; catanese & weekes 1999 ) . note that the ground - based instruments are approaching the energy range covered by satellite detectors and
already nowadays achieve an energy threshold as low as 60 gev .
basic results of these observations are summarized in table 2 . except for recent evidence for a detection of geminga with still low level of confidence
( neshpor & stepanian 2001 ) none of the observations have resolved a pulsed vhe @xmath0-ray emission .
the outer gap model could easily accommodate the non - detection of pulsed vhe @xmath0-ray emission by assuming higher fluxes of the microwave seed photons which cause the pair production of tev @xmath0-rays within the vacuum gap ( hirotani & shibata 2001 ) . predicted fluxes of pulsed ic tev @xmath0-rays rely heavily on the spectral energy distribution of the microwave seed photons , which is currently purely known .
@xmath52 u.l .
is an upper limit measured in @xmath53 .
neshpor & stepanian ( 2001 ) claim the detection of geminga pulsar at 4.4@xmath50 confidence level .
a detailed search for the ic component of vhe @xmath0-ray emission from pulsars will be performed with the ground - based imaging erenkov telescopes of the next generation with drastically improved flux sensitivity .
what sets the puzzle is that the crab nebula ( hillas et al .
1998 ; aharonian et al . 2000 ) , vela ( yoshikoshi et al . 1997 ; chadwick et al . 2000 ) , and @xmath51 ( kifune et al . 1995 ; chadwick et al . 1998 ) , are well established sources of dc gev - tev @xmath0-ray emission in the 60 gev - 50 tev energy range .
this steady flux emission might be associated possibly with the x - ray synchrotron nebula around the pulsar ( de jager & harding .
1992 ; de jager et al . 1996
; aharonian , atoyan , kifune 1997 ) .
however the spectral behavior of both , pulsed and unpulsed , components of high - energy @xmath0-rays within the energy range from 10 gev up to 100 gev remains unknown .
this energy range can be effectively studied from the ground with the forthcoming imaging atmospheric erenkov light telescopes ( see for a review catanese , weekes 1999 ) .
spectral measurements over this energy range will severely constrain the physics of particle injection and acceleration in a pulsar and within the surrounding pulsar nebula .
note that h.e.s.s . could detect a crab like source of dc vhe @xmath0-rays after a few minutes of observation ( konopelko 2000 ) .
most of the radio pulsars detected so far have their rotational periods within the range from 10 ms to 10 s , whereas a few of them form a separate sample of millisecond pulsars with rotational periods in the range from 1 ms to 10 ms .
@xmath54 is the first detected millisecond pulsar , and has a rotational period of 1.5 ms ( backer et al .
for such a short period it has a surprisingly small value of @xmath55 . using the relation given by eqn.([tau ] )
, one can estimate the pulsar age as @xmath56 years .
being rather old , the millisecond pulsars do not fit the general scheme of pulsar evolution , and most probably are re - accelerated by some external mechanism ( e.g. by accretion from a companion star in a binary system ) .
the millisecond pulsars should have smaller surface magnetic field ( see eqn.([field ] ) ) , @xmath57 , and that makes them particularly interesting with respect to the high - energy @xmath0-ray emission .
given such a low magnetic field , the cut - off energy of curvature photons due to the pair production ( see discussion in section 2 ) shifts to higher energies , and the maximum radiation of the millisecond pulsars should correspond to photons with energy @xmath58 ev ( usov 1983 ) .
the value of the photon flux above 100 gev , estimated for @xmath54 by usov ( 1983 ) , is approximately an order of magnitude higher than that of the crab pulsar , @xmath59 .
recent detailed calculations of the energy spectra of high - energy and very high energy @xmath0-ray emission from millisecond pulsars ( bulik , rudak , dyks 2000 ) in general have confirmed early expectations .
pulsar @xmath60 ( @xmath61ms , @xmath62 pc ) is a prime candidate for observations in the energy range above 100 gev .
its integral @xmath0-ray flux above 100 gev is expected to be as high as @xmath63 ( bulik , rudak , dyks 2000 ) , which is above the sensitivity limit of h.e.s.s .
( konopelko 2000 ) . assuming rather conservative @xmath0-ray flux for @xmath60
one needs a 50 hour exposure with h.e.s.s . in order to detect the pulsed @xmath0-rays .
another millisecond pulsar , @xmath64 , is a good candidate for pulsed high - energy @xmath0-ray emission seen by egret , and it is certainly a valuable target for future observations with h.e.s.s .
in addition to seven well established high - energy @xmath0-ray pulsars egret discovered a large number of unidentified @xmath0-ray sources at low galactic latitudes ( see kanbach 2002 ) . the recent parkes multi - beam radio surveys allowed to resolve a number of young energetic radio pulsars with some of those being in rather good positional coincidence with unidentified egret @xmath0-ray sources ( torres , butt , camilo 2001 ) .
such associations resemble in many ways the established egret @xmath0-ray pulsars . for instance
pulsar @xmath65 could be a counterpart of high - energy @xmath0-ray source @xmath66 ( kaspi et al . 2000 ) .
it has the ninth highest value of @xmath22 , and its @xmath0-ray light - curve morphology is similar to other known @xmath0-ray pulsars , namely , two narrow peaks separated by @xmath67 in the light - curve phase .
@xmath66 has a rather flat energy spectrum ( @xmath68 ) without clear evidence for a cut - off .
two other young radio pulsars @xmath69 ( @xmath70=68 ms ) and @xmath71 ( @xmath70=96 ms ) are located @xmath72 away from the position of the egret unidentified sources @xmath73 and @xmath74 , respectively ( damico et al . 2001 ) . in the energy range from 100 mev to 10 gev
, the spectra of these two sources can be fitted with power laws of photon spectral index @xmath75 2.02 and 1.82 , respectively , which are inside the spectral index range of the known @xmath0-ray pulsars .
a 20 hour exposure with h.e.s.s . for
this source may be sufficient for a clear detection of very high energy @xmath0-ray emission ( de jager et al .
2001 ) from both of these pulsars , assuming the energy cut - off at 40 gev .
about 30 young energetic radio pulsars are currently known as positional associations with unidentified egret @xmath0-ray sources ( damico 2001 ) .
in addition some of those pulsars are within the x - ray nebula , which confirms that they are plerions ( see roberts , romani , kawai 2001 ) , and may well be in addition the sources of dc high - energy @xmath0-ray emission coming from the x - ray nebula , surrounding the pulsar .
such sources could be effectively studied at very high energies from the ground with an instrument like h.e.s.s .
h.e.s.s . ( _ high energy stereoscopic system _ ) is one of the next generation ground - based instruments for very high energy @xmath0-ray astronomy ( hofmann 2001 ) . in its 1st phase
it will consist of four 12 m erenkov telescopes deployed in the khomas highland of namibia . h.e.s.s .
will have an energy threshold in the range of 50 - 100 gev ( konopelko 2000 ) , an angular resolution for individual photons of @xmath76 , and an energy resolution better than 20% . according to the current sensitivity estimate , h.e.s.s .
will need a 50 hour exposure to detect a @xmath0-ray point source with photon fluxes @xmath77 and @xmath78 above 100 gev and 1 tev , respectively .
the first observations with two telescopes are being currently planned in early 2003 . the complete system is due in 2004 . the visibility of h.e.s.s . along with a few selected pulsars as the potential targets for future observations
is shown in figure 4 .
pulsars are objects with an extreme physics environment .
they are offering physics processes to be studied by plasma electrodynamics , nuclear physics , cosmic ray physics , _
etc_. understanding the pulsar physics could substantially benefit from the enrichment of the presently very small sample of high - energy @xmath0-ray pulsars .
measurements of energy spectra for already established , as well as for newly discovered @xmath0-ray pulsars , will constrain the modeling of high - energy emission and in particular the location of the emitting regions with respect to the neutron star .
future observations at high - energies , which can be effectively done with the ground - based erenkov instruments , will stretch the existing models and will help to understand the physics of millisecond pulsars .
observations of associations of the energetic young radio pulsars at low galactic latitudes with unidentified egret @xmath0-ray sources will allow to prove that pulsars are in general powerful emitters of high - energy @xmath0-rays .
finally , a rich sample of @xmath0-ray pulsars can be used for better understanding the conversion of pulsar rotational energy into @xmath0-ray emission , which is determined by the size and geometry of emitting regions .
is one of the forthcoming ground - based erenkov detectors , such as cangaroo , magic , and veritas , and the present discussion is entirely relevant for the future observations with any of those instruments .
i would like to thank werner becker , alice harding , okkie de jager , gottfried kanbach , john kirk , yury lyubarsky , roger romani , bronislaw rudak , vladimir usov , and heinz vlk , for discussions on pulsar physics .
aharonian , f. , et al . 1999 , _
a&a _ , 346 , 913 aharonian , f. , et al .
2000 , _ apj _ , 539 , 317 aharonian , f. , atoyan , a. , kifune , t. 1997 , _ mnras _ , 291 , 162 akerlof , et al . , 1993 , _
a&a _ , 274 , l17 baade , w. , zwicky , f. 1934 , _ proc .
, 20 , 254 backer , d.c . , et al . 1982 , _ nature _ , 300 , 615 becker , w. , trmper , j. 1997 , _ a&a _ , 326 , 682 bulik , t. , rudak , b. , dyks , j. 2000 , _ mon . not .
r. astron .
_ , 317 , 97 catanese , m. , weekes , t. 1999 , _ pub .
astron . soc .
of pacific _ , 111 , 764 , 1193 chadwick , et al . , 1998 , _
_ , 9 , 131 chadwick , et al . , 2000 , _ apj _ , 537 , 414 cheng , k.s
, ho , c. , ruderman , m.a . 1986 , _ apj _ , 300 , 500 damico n. , et al . 2001 , _ apj _ , 552 , l45 damico , 2001 , _ talk at gamma ray astrophysics symposium _ , baltimore daugherty , j.k . , harding , a.k .
1982 , _ apj _ , 252 , 337 de jager o.c . ,
et al . 1992 , _ apj _ , 396 , 161 de jager o.c . , et al . 1996 , _ apj _ , 457 , 253 de jager o.c . ,
konopelko , a. , raubenheimer , b.c . , visser , b. 2001 , _ proc .
icrc , hamburg _ , vol . 6 , 2432 de naurois , et al .
2002 , _ apj _ , 566 , 343 goldreich , p. , julian , w.h .
1969 , _ apj _ , 157 , 869 harding , a.k .
1981 , _ apj _ , 245 , 267 harding , a. et al .
2000 , _ proc .
high energy gamma ray aston .
, heidelberg , eds .
aharonian , h.j .
vlk , aip conf .
558 , 115 hewish , a. , et al .
1968 , _ nature _ , 217 , 709 hillas , a.m. , et al .
1998 , _ apj _ , 503 , 744 hirotani , k. , shibata , s. 2001 , _ apj _ , 558 , 216 hofmann , w. 2001 , _ proc .
27th icrc , hamburg _ , vol . 7 , 2785 kanbach , g. 2002 , _ this proceedings _ kaspi , v.m . ,
2000 , _ apj _ , 528 , 445 kifune , t. , et al .
, 1995 , _ apj _ , 438 , l91 kirk , j. , skjaeraasen , o. , gallant , y. 2002 , _ submitted to a&a _ konopelko , a. , 2000 , _ proc .
high energy gamma ray aston .
, heidelberg , eds .
aharonian , h.j .
vlk , aip conf .
558 , 569 kuiper , l. , et al .
1999 , _ a&a _ , 351 , 119 lessard , et al . , 2000 , _ apj _ , 531 , 942 macomb , d.j .
, gehrels , n. 1999 , _ apjs _ , 120 , 335 neshpor , yu . , stepanian , a. , 2001 , _ astron . letters _ , 27 , 12 , 794 ong , r. 1998 , _ phys .
reports _ , 305 , 93 oser , et al . , 2001 , _
apj _ , 543 , 949 ostriker , j.p . ,
gunn , j.e .
1969 , _ apj _ , 157 , 1395 roberts , m. , romani , r. , kawai , n. 2001 , _ apjs _ , 133 , 451 romani , r. 1996 , _ apj _ , 470 , 469 ruderman , m. , cheng , k.s .
1988 , _ apj _ , 335 , 306 ruderman , m.a . , sutherland , p.g .
1975 , _ apj _ , 196 , 51 srinivasan , r. , et al .
_ apj _ , 489 , 170 sturrock , p.a . , 1971 , _ apj _ , 164 , 529 thompson , d.j .
2000 , _ proc .
high energy gamma ray aston .
, heidelberg , eds .
aharonian , h.j .
vlk , aip conf .
558 , 103 torres , d. , butt , y. , camilo , f. 2001 , _ apj _ , 560 , l155 usov , v.v .
1982 , _ nature _ , 305 , 409 yoshikoshi , et al . , 1997 , _
apj _ , 487 , l65 | since their discovery at radio wavelengths pulsars have been persistent targets for widespread multi - wave observations throughout optics , radio , x - rays , and high - energy @xmath0-rays .
observations with the egret @xmath0-ray telescope , on board compton gro satellite , confirmed the expectation of a pulsed high - energy emission up to a few gev . presently , at least seven objects are known as well established high - energy @xmath0-ray pulsars .
a few of those emit @xmath0-rays well above 1 gev .
forthcoming ground - based erenkov telescopes will enable observations of @xmath0-rays well below @xmath1 gev , finally reaching the yet unexplored energy gap at tens of gev .
h.e.s.s .
( _ high energy stereoscopic system _ ) is one of such instruments which is planned to be operational in 2004 . here
i summarize the basic scientific motivations , the h.e.s.s .
sensitivity , and the first targets for future pulsar observations at high energies from the ground . |
reactions with radioactive beams have attracted great experimental and theoretical interest during the last two decades @xcite .
progresses of this scientific endeavor were reported on measurements of nuclear sizes @xcite , use of secondary radioactive beams to elucidate reactions of astrophysical interest @xcite , fusion reactions with neutron - rich nuclei @xcite , tests of fundamental interactions @xcite , dependence of the equation of state of nuclear matter upon the asymmetry energy @xcite , and many other research topics . new research areas with nuclei
far from the stability line will become possible with newly proposed experimental facilities .
one of the projects at the future fair facility of the gsi laboratory / germany is the study of electron scattering off unstable nuclei in an electron - ion collider mode @xcite . a similar proposal exists for the riken laboratory facility in japan @xcite . by means of elastic electron scattering
, these facilities will become the main probe of charge distribution in unstable nuclei @xcite .
coulomb excitation has been very useful in determining the electromagnetic response in light nuclei @xcite . for neutron - rich isotopes
@xcite the resulting photo - neutron cross sections are characterized by a pronounced concentration of low - lying @xmath0 strength .
but it is well known that non - perturbative effects leading to distortion of the energy spectrum of the fragments interacting with the target ( see , e.g. ref .
@xcite ) is a problem of difficult nature .
the nuclear response probed with electrons is free from such effects .
the interpretation of the low - lying @xmath0 strength in neutron - rich nuclei engendered a debate : are these soft dipole modesjust a manifestation of the loosely - bound character of light neutron - rich nuclei , or are they the result of the excitation of a resonance ? @xcite . the electromagnetic response in light nuclei , leading to their dissociation , is related to the nuclear physics needed in several astrophysical sites @xcite .
the existence of pygmy resonances have important implications on theoretical predictions of radiative neutron capture rates in the r - process nucleosynthesis , and consequently on the calculated elemental abundance distribution in the universe @xcite .
here , @xmath1 @xmath2 @xmath3is the initial ( final ) angular momentum of the nucleus , @xmath4 and ( @xmath5 ) are the initial and final energy and momentum of the electron , and @xmath6 is the energy and momentum transfer in the reaction . for low energy excitations , @xmath7 , for electron energies @xmath8 mev and excitation energies @xmath9 mev . in the plane wave born approximation ( pwba ) and using the siegert s theorem , one can show that @xcite @xmath10 where @xmath11 , is the photo - nuclear cross section for the @xmath12-multipolarity , and @xmath13 .
the response function , @xmath14 is given by @xmath15 ^{2}\rho\left ( e_{\gamma}\right ) , \label{photo}\ ] ] where @xmath16 is the density of final states ( for nuclear excitations into the continuum ) with energy @xmath17 .
the geometric coefficient @xmath18 and the transition density @xmath19 will depend upon the nuclear model adopted .
one can also define a differential cross section integrated over angles so that @xmath20 and @xmath21 is the maximum electron scattering angle , which depends upon the experimental setup .
notice that the lowest limit in the above integral is @xmath22 , and not zero .
[ epa]-[epa3 ] show that under the conditions of the proposed electron - ion colliders , electron scattering will offer the same information obtained with real photons .
the reaction dynamics information is in the virtual photon spectrum , @xmath23 , while the nuclear response dynamics information is in eq .
[ photo ] .
@xmath24 is interpreted as the number of equivalent ( real ) photons incident on the nucleus per unit scattering angle @xmath25 and per unit photon energy @xmath26 for larger scattering angles the coulomb multipole matrix elements are in general larger than the electric ( @xmath12 ) multipoles , and monopole transitions become relevant @xcite .
. [ epa ] will not be valid under these conditions .
it is found that the spectrum @xmath27 increases rapidly with decreasing energies @xcite . for @xmath28 mev and excitation energies @xmath29 mev
, the spectrum yields the ratios @xmath30 and @xmath31 .
however , although @xmath32 increases with the multipolarity @xmath33 , the nuclear response decreases rapidly with @xmath33 , and @xmath0 excitations tend to dominate the cross sections . for larger electron energies the ratios @xmath34 and @xmath35 decrease rapidly .
a comparison between the @xmath0 virtual photon spectrum , @xmath36 , of 1 gev electrons with the spectrum generated by 1 gev / nucleon heavy ion projectiles was published in ref .
the virtual spectrum for the electron contains more hard photons , i.e. the spectrum decreases slower with photon energy than the heavy ion photon spectrum .
this is because , in both situations , the rate at which the spectrum decreases depends on the ratio of the projectile kinetic energy to its rest mass , @xmath37 which is much larger for the electron ( @xmath38 ) than for the heavy ion ( @xmath39 nuclear mass ) .
in a two - body model , the single - particle picture has been used previously to study coulomb excitation of halo nuclei with success @xcite .
the @xmath0 transition integral @xmath40 becomes @xmath41 , \label{isp}\ ] ] in terms of the effective range expansion of the phase shift , @xmath42 .
as shown in previous works @xcite , a peak is manifest in the response function , @xmath43 it appears centered at the energy @xcite
@xmath44 for a generic electric response of multipolarity @xmath33 . for @xmath0 excitations ,
the peak occurs at @xmath45 .
the second term inside brackets in eq . [ isp ] is a modification due to final state interactions .
it is important in many cases @xcite .
many weakly - bound nuclei , like @xmath46he or @xmath47li , require a three body treatment .
in one particular model , the bound state wavefunction in the center of mass system is written as an expansion over hyperspherical harmonics , see e.g. @xcite .
for weakly - bound systems having no bound subsystems the hyperradial functions behave asymptotically as @xcite @xmath48 as @xmath49 , where the two - nucleon separation energy is related to @xmath50 by @xmath51 .
( see also @xcite ) .
the @xmath0 transition matrix element is obtained by a sandwich of the e1 operator between @xmath52 and scattering wavefunctions .
following ref .
@xcite , but using distorted scattering states , @xmath53 where @xmath54 is the core - neutron asymptotic continuum wavefunction , assumed to be a @xmath55-wave , and @xmath56 is the neutron - neutron asymptotic continuum wavefunction , assumed to be an @xmath57-wave .
the @xmath0 strength function is proportional to the square of eq .
[ ie1 ] integrated over all momentum variables . as pointed out in ref .
@xcite , the @xmath0 three - body response function of @xmath47li can still be described by an expression similar to the two - body case , but with different factors .
explicitly , @xmath58 instead of @xmath59 , one has to use an effective @xmath60 , with @xmath61 . with this approximation ,
the peak of the strength function in the three - body case is situated at about three times higher energy than for the two - body case . in the three - body model ,
the maximum is thus predicted at @xmath62 , which fits the experimental peak position for the @xmath47li @xmath0 strength function very well @xcite .
it is thus apparent that the effect of three - body configurations is to widen and to shift the strength function @xmath63 to higher energies . as with giant dipole resonances ( gdr ) in stable nuclei
, one believes that pygmy resonances at energies close to threshold are present in halo , or neutron - rich , nuclei .
this was proposed by suzuki et al .
@xcite using the hydrodynamical model for collective vibrations .
we will use the method of myers et al .
@xcite , considering collective vibrations in nuclei as an admixture of goldhaber - teller ( gt ) and steinwedel - jensen ( sj ) modes . for light nuclei goldhaber - teller modes dominate . for spherically symmetric densities , the transition density , @xmath64 ,
can be calculated assuming a combination of sj and gt distributions @xcite
, @xmath65 where @xmath66 is the mean nuclear radius , and @xmath67 is the percent displacement of the center of mass of neutron and proton fluids in the gt ( @xmath68 ) and sj ( @xmath69 ) modes . for light , weakly - bound nuclei , it is appropriate to assume that the neutrons inside the core ( @xmath70 ) vibrate in phase with the protons .
the neutrons and protons in the core are tightly bound . calling the excess nucleons by @xmath71 the effective charge for the gt mode
is @xmath72 @xmath73 . in eq .
[ transdsjgt ] , @xmath74 is the spherical bessel function of first order , @xmath75 the percent displacement of the center of mass of the neutron and proton fluids in the sj mode , @xmath76 the hydrodynamical model can be further explored to obtain the energy and excitation strength of the collective excitations .
this can be achieved by finding the eigenvalues of the hamiltonian @xmath77 where @xmath78 is now a vector containing gt and sj contributions to the collective motion . @xmath79 and @xmath80 are the kinetic and potential energies @xmath81 matrices @xcite .
the kinetic term can be calculated from gt and sj velocity fields .
the potential term is related to the stiffness parameters of the liquid - drop model and adjusted to a best fit to the nuclear masses .
the stiffness of the system is due to the change in symmetry energy of the system as it goes out of the equilibrium position .
the last term in the hamiltonian is the rayleigh dissipation term , which is related to the fermi velocity of the nucleons @xcite and yields the width of the eigenstate . as shown by myers et al .
@xcite , the liquid drop model predicts an equal admixture of sj+gt oscillation modes for large nuclei .
the contribution of the sj oscillation mode decreases with decreasing mass number , i.e. @xmath82 as @xmath83 .
this is even more probable in the case of halo nuclei , where a special type of gt mode ( oscillations of the core against the halo nucleons ) is likely to be dominant .
for this special collective motion an approach different than those used in refs .
@xcite and @xcite has to be considered .
it is easy to make changes in the original goldhaber and teller @xcite formula to obtain the energy of the collective vibrations , yielding @xmath84 where @xmath85 and @xmath86 is the length within which the interaction between a neutron and a nucleus changes from a zero - value outside the nucleus to a high value inside , i.e. @xmath86 is the width of the nuclear surface .
@xmath87 is the energy needed to extract one neutron from the proton environment .
goldhaber and teller @xcite argued that in a heavy stable nucleus @xmath87 is not the binding energy of the nucleus , but the part of the potential energy due neutron - proton asymmetry . in the case of weakly - bound nuclei this picture changes and it is more reasonable to associate @xmath87 to the separation energy of the valence neutrons , @xmath88 .
i will use @xmath89 , with a parameter @xmath90 which is expected to be of order of one .
since for halo nuclei the product @xmath91 is proportional to @xmath92 we obtain the proportionality @xmath93 . using eq .
[ epgdr ] for @xmath47li , with @xmath94 fm , @xmath95 fm and @xmath96 mev , we get @xmath97 mev . considering that the pygmy resonance will most probably decay by particle emission
, one gets @xmath98 mev for the kinetic energy of the fragments .
this is about a factor 2 larger than what is obtained in a numerical calculation @xcite .
but it is within the right ballpark .
the hydrodynamical model is very unlikely to be an accurate model for light , loosely - bound , nuclei and is significant only in that a reasonable magnitude of the resonance energy is found . both the direct dissociation model and the hydrodynamical model yield a bump in the response function with position proportional to @xmath88 , the valence nucleon(s ) separation energy . in the direct dissociation model
the width of the response function depends on the separation energy and on the nature of the model , i.e. a two- or a three - body model . in the case of the pygmy resonance ,
this question is completely open .
the hydrodynamical model predicts @xcite for the width of the collective mode @xmath99 , where @xmath100 is the average velocity of the nucleons inside the nucleus .
this relation can be derived by assuming that the collective vibration is damped by the incoherent collisions of the nucleons with the walls of the nuclear potential well during the vibration cycles .
this approach mimics that used in the kinetic theory of gases for calculating the energy transfer of a moving piston to gas molecules in a container . using @xmath101 , where @xmath102 is the fermi velocity , with @xmath103 mev and @xmath104 fm
, one gets @xmath105 mev .
this is the typical energy width a giant dipole resonance state in a heavy nucleus .
but in the case of neutron - rich light nuclei @xmath100 is not well defined .
but the piston model is not able to describe the width of the response function properly .
microscopic models , such as those based on random phase approximation ( rpa ) calculations , are necessary to tackle this problem .
the halo nucleons have to be treated in an special way to get the response at the right energy position , and with approximately the right width .
right now , the problem remains if the experimentally observed peak in @xmath106 is due to a direct transition to the continuum , weighted by the phase space of the fragments , or if it proceeds sequentially via a soft dipole collective state . | pygmy resonances in light nuclei excited in electron scattering are discussed .
these collective modes will be explored in future electron - ion colliders such as elise / fair ( spokesperson : haik simon - gsi ) .
response functions for direct breakup are explored with few - body and hydrodynamical models , including the dependence upon final state interactions . |
the recent discovery of the top quark by the cdf and d0 collaborations at the fermilab tevatron has stimulated interest in top - quark physics @xcite .
the standard model of the strong and electroweak interactions predicts that the top quark should behave , both in its production and decay , as a heavy version of the five lighter quarks .
however , the large mass of the top quark relative to the five lighter quarks suggests that there may be something special about its properties .
it is therefore crucial that we study the properties of the top quark with every probe available to us . in this paper
we study a neglected probe of the top quark : the process @xmath4 via a virtual @xmath5-channel @xmath3 boson , as depicted in fig .
1(a ) @xcite .
this process is similar to the well - known @xmath3-gluon fusion process @xcite , depicted in fig .
1(b ) , in that it produces a single top quark rather than a @xmath6 pair .
however , the process @xmath7 probes the top quark with a timelike @xmath3 boson , @xmath8 , while the @xmath3-gluon - fusion process involves a spacelike @xmath3 boson , @xmath9 .
these processes are therefore complementary in that they probe the charged - current interaction of the top quark in different regions of @xmath10 .
, involves an on - shell @xmath3 boson , @xmath11 . ]
another important feature of @xmath4 is that the hadronic cross section can be reliably calculated .
the quark and antiquark distribution functions are evaluated at moderate values of @xmath12 , where they are well - known .
the qcd correction to this process is straightforward and can be carried out to at least @xmath13 .
furthermore , the initial quark - antiquark flux can be constrained by measuring @xmath14 , which also proceeds via a virtual @xmath5-channel @xmath3 boson . of the @xmath3 boson
can not be determined .
the process @xmath15 therefore provides a constraint on the quark - antiquark flux , not a direct measurement . ] the qcd correction to the initial state is automatically taken into account via this procedure ; only the qcd correction to the final state must be explicitly calculated , and this can be done reliably since it is free of collinear and infrared singularities .. ] in contrast , the @xmath3-gluon - fusion process involves the gluon distribution function , which is not well - known . since the most important issue to be resolved is the observability of @xmath4 above backgrounds , we concentrate on this process in the context of the standard electroweak model . in this context
, @xmath4 may provide the best direct measurement of the magnitude of the cabbibo - kobayashi - maskawa ( ckm ) matrix element @xmath1 .
if there are only three generations , unitarity of the ckm matrix implies that @xmath16 is very close to unity ( @xmath17 @xcite ) .
if there is a fourth generation , @xmath16 could be anywhere between ( almost ) zero and unity , depending on the amount of mixing between the third and fourth generations .
a measurement of @xmath16 less than unity would therefore be a major discovery .
including the top - quark decay , the process @xmath4 yields the final state @xmath18 .
there are several background processes which must be considered in order to establish whether the signal can be observed .
the dominant irreducible background , @xmath19 , was considered in ref .
we reconsider this background , as well as the other important irreducible and reducible backgrounds .
the full set of backgrounds we consider is : * @xmath18 * @xmath20 ( where @xmath21 denotes a light - quark or gluon jet ) * @xmath22 * @xmath23 ( @xmath3-gluon fusion ) * @xmath24 the raw cross sections for these processes at the tevatron ( @xmath25 tev @xmath26 collider ) and the cern large hadron collider ( lhc , @xmath27 tev @xmath28 collider ) are given in the first column of table 1 .
we use @xmath29 gev , and all cross sections are calculated at leading order . ) parton distribution functions @xcite , evaluated at @xmath30 , for all cross sections except @xmath3-gluon fusion , where we use @xmath31 .
we evaluate @xmath32 at the same scale as the parton distribution functions , with @xmath33 @xcite . ]
the cross sections include a factor of 2/9 for the branching ratio of @xmath34 ( @xmath35 ) .
the cross sections listed are for @xmath36 production only ; the inclusion of @xmath37 doubles the signals and backgrounds at the tevatron , and nearly doubles them at the lhc .
the @xmath20 cross section is infinite without cuts due to infrared and collinear singularities .
cccccc & & tevatron&2 tev @xmath26 & & + + & total@xmath38&acceptance&rejection&@xmath39 gev&@xmath40 tagging + + @xmath41 & 69 & 34 & 29 & 18 & 6.5 + @xmath18 & 3250 & 140 & 140 & 25 & 9.0 + @xmath20 & & 14000 & 14000 & 3500 & 0.35 + @xmath42 & 51 & 22 & 22 & 0.71 & 0.26 + @xmath43 & 116 & 31 & 5.7 & 2.3 & 0.83 + @xmath6 & 860 & 505 & 9.5 & 6.0 & 2.2 + + + & & lhc&14 tev @xmath28 & & + + & total@xmath38&acceptance&rejection&@xmath39 gev&@xmath40 tagging + + @xmath41 & 1100 & 290 & 190 & 130 & 47 + @xmath18 & 34500 & 925 & 925 & 235 & 85 + @xmath20 & & 340000&340000&140000 & 14 + @xmath42 & 785 & 156 & 156 & 6.6 & 2.4 + @xmath43 & 21000 & 4800 & 1100 & 615 & 220 + @xmath6 & 90500 & 40000 & 1200 & 810 & 290 + ll @xmath44 & @xmath45 gev + @xmath46 & @xmath47 gev ( 20 gev ) + @xmath48 & @xmath49 gev + @xmath50 & @xmath51 + @xmath52 gev & + we smear the jet energies with a gaussian function of width @xmath53 ( added in quadrature ) to simulate the resolution of the hadron calorimeter .
we do not smear the lepton energy , since this is a small effect compared with the smearing of the jet energies .
we show in fig .
2 the transverse - momentum spectra of the @xmath40 jet , the @xmath54 jet , and the lepton for the signal process . we list in table 2 the cuts imposed on the signal and backgrounds to simulate the acceptance of the detector .
the resulting cross sections are listed in the second column of table 1 .
the acceptance for the signal is @xmath55 at the tevatron and @xmath56 at the lhc . @xcite . ] in the third column of table 1 , the @xmath3-gluon fusion background ( denoted by @xmath57 ) is reduced by rejecting events in which a third jet is detected with @xmath58 gev and @xmath48 .
this provides a reduction factor of about 1/5 at both the tevatron and the lhc .
this also reduces the signal cross section , since the signal will sometimes be accompanied by a third jet due to qcd radiation .
we estimate plus an additional jet , and an estimate of the qcd correction to the inclusive cross section based on the correction to @xmath59 . ]
this reduction to be about @xmath60 at the tevatron and about @xmath61 at the lhc , and reduce the signal accordingly .-
gluon fusion is a small background at the tevatron , it may be advantageous to increase the jet - rejection threshold in order to preserve more of the signal . with a threshold of @xmath62 gev
, we estimate that almost none of the signal is lost , while the small @xmath3-gluon fusion background is increased by only a factor of 1.8 . ] the @xmath18 , @xmath20 , and @xmath42 backgrounds would also be reduced ; but , to be conservative , we we have not taken this into account . the @xmath6 background is reduced by rejecting events with an additional @xmath3 boson .
if the @xmath3 boson decays hadronically , the event is rejected if either jet has @xmath58 gev and @xmath48 .
if the @xmath3 boson decays leptonically , similar to @xmath63 for this @xmath3 decay .
this may underestimate the background for @xmath64 . ]
the event is rejected if @xmath47 gev and @xmath46 .
the overall reduction factor from rejecting these events is 1/50 at the tevatron and 1/30 at the lhc .
most of the remaining background events are from the leptonic decay of the extra @xmath3 boson ( @xmath65 at the tevatron , @xmath66 at the lhc ) .
this is unnecessary at the tevatron , but is desirable at the lhc , where @xmath6 is the largest background . ] in the fourth column of table 1 , we impose a cut on the @xmath67 invariant mass , @xmath39 gev .
this essentially eliminates the @xmath22 background , and reduces the @xmath18 and @xmath20 backgrounds .
the loss of signal is modest .
if there is a higgs boson of mass less than 120 gev , it too will contribute to the background via @xmath68 @xcite , and may require a slightly higher cut on @xmath69 . due to the enormous background from @xmath20 ( where @xmath21 denotes a jet from a light quark or gluon ) , it is necessary to tag one or both of the @xmath40 quarks from the @xmath70 signal .
the most efficient means of @xmath40 tagging is with a silicon vertex detector ( svx ) .
the cdf collaboration has achieved a tagging efficiency of roughly @xmath71 per `` fiducial '' @xmath40 jet ( @xmath72 gev and within the coverage of the svx ) , with a probability of less than @xmath73 for a light - quark or gluon jet to fake a @xmath40 jet @xcite .
the new svx for run ii , with three - dimensional vertex information , should increase the efficiency to as high as @xmath74 while maintaining or reducing the fake rate @xcite .
pixel detectors , such as those planned for the atlas @xcite and cms @xcite detectors at the lhc , could further increase the efficiency .
we assume a @xmath74 tagging efficiency - tagging efficiency . ] per fiducial @xmath40 jet ( @xmath72 gev , @xmath75 ) with a @xmath73 fake rate from fiducial light - quark and gluon jets .
the best signal significance is obtained by tagging both @xmath40 jets .
tagging , the @xmath3-gluon - fusion background is prohibitively large .
the @xmath54 jet is often below the @xmath76 gev threshold , and the @xmath40 jet from the top decay plus the spectator jet ( from the radiation of the virtual @xmath3 ) fake the signal . ]
the final column in table 1 lists the cross sections including the @xmath40-tagging efficiency . with
double @xmath40 tagging , the @xmath20 background is reduced to a negligible level at the tevatron . another piece of information at our disposal is the invariant mass of the @xmath3 boson plus @xmath40 jet , @xmath77 , which equals @xmath78 for the signal but has a continuous spectrum for the @xmath18 background .
unfortunately , this information can not be utilized directly .
one does not know _ a priori _ which @xmath40 jet is from the top - quark decay . furthermore , the @xmath3-boson momentum must be reconstructed from the lepton momentum plus the missing @xmath79 , and this yields two solutions .
thus each event has four possible assignments of momenta , leading to four different values of @xmath77 .
the most straightforward method to deal with this four - fold ambiguity is to assign four values of @xmath77 to each event , and plot the differential cross section as a function of this invariant mass ( weighting each entry by 1/4 of an event ) .
the resulting invariant - mass distributions for the signal and the largest backgrounds are shown in fig .
3(a ) at the tevatron and fig .
3(b ) at the lhc .
the signal is prominent at the tevatron , but is masked by the @xmath6 and @xmath3-gluon - fusion backgrounds at the lhc , which have the same shape as the signal .
these backgrounds are enhanced at the lhc relative to the @xmath7 signal because they are initiated by gluons . in @xmath26 collisions
, the @xmath40 jet from the process @xmath80 tends to go in the direction of the proton beam @xcite .
this can be used to select the correct @xmath40 jet to associate with the @xmath3 boson , with an efficiency of about @xmath81 .
furthermore , the correct neutrino solution tends to be the one with the smaller absolute longitudinal momentum . making these assignments ,
we show in fig .
4 the resulting invariant - mass distribution at the tevatron .
the signal is significantly narrower than in fig .
3(a ) , and clearly stands out above the @xmath18 and @xmath6 backgrounds . selecting a bin of width 40 gev centered on 175 gev in fig . 4
, the signal cross section is 4.4 @xmath82 and the background 3.4 @xmath82 . including the charge - conjugate process , @xmath83
, the total signal is 8.8 @xmath82 on a background of 7.2 @xmath82 .
the number of events acquired in 3 @xmath2 of integrated luminosity is enough for the signal to correspond to more than a @xmath84 fluctuation of the continuum background .
this integrated luminosity corresponds to about 1.5 years of running at the tevatron with the main injector and the recycler ( @xmath85 ) @xcite .
assuming @xmath16 is close to unity , 3 @xmath2 of integrated luminosity yields a @xmath86 measurement of @xmath16 .
this corresponds to a @xmath87 measurement of the partial width @xmath88 .
a @xmath89 measurement of @xmath16 requires 12 @xmath2 , which could be delivered in a little more than a year with a high - luminosity tevatron ( @xmath90 ) @xcite .
this corresponds to a @xmath86 measurement of @xmath91 .
the systematic error in the measurement is small , as discussed in the introduction .
if @xmath16 is significantly less than unity , the rate for @xmath0 is suppressed , and this process is therefore more difficult to detect . if no signal is observed with 3 @xmath2 of integrated luminosity , an upper bound of @xmath92 ( @xmath93 c. l. ) can be deduced .
the absence of a signal with 12 @xmath2 implies @xmath94 .
we have shown that the process @xmath4 is potentially observable at the tevatron with 3 @xmath2 of integrated luminosity .
the signal is straightforward to isolate from backgrounds , and requires tagging both @xmath40 jets .
in contrast , this process is masked at the lhc by backgrounds from @xmath6 and single - top - quark production via @xmath3-gluon fusion .
we hope this work will motivate a more detailed study of this process by the detector collaborations .
the process @xmath4 at the tevatron may be the best direct measurement of @xmath16 .
the @xmath3-gluon fusion process also provides a measurement of @xmath16 @xcite , and has the advantage of a larger cross section .
however , it suffers from the uncertainty in the gluon distribution function .
in contrast , the quark - antiquark flux for @xmath4 is well - known , and can be constrained by measuring @xmath14 . both @xmath4 and @xmath3-gluon fusion
should be pursued , since they probe the top - quark charged current in different regions of @xmath10 .
we are grateful for conversations with c. hill , s. parke , and t. liss .
t. s. was supported in part by a uk pparc post - doctoral fellowship .
s. w. was supported in part by department of energy grant de - fg02 - 91er40677 .
99 cdf collaboration , f. abe _ et al .
_ , 74 2626 1995 ; d0 collaboration , s. abachi _ et al . _ , 74 2632 1995 . s. cortese and r. petronzio , 253 494 1991 .
s. willenbrock and d. dicus , 34 155 1986 .
yuan , 41 42 1990 ; d. carlson and c .- p .
yuan , 306 386 1993 .
r. k. ellis and s. parke , 46 3785 1992 .
g. bordes and b. van eijk , 435 23 1995 . , 50 1173 1994 .
t. stelzer and w. f. long , comput .
phys . commun .
* 81 * , 357 ( 1994 ) .
a. martin , r. roberts , and w. j. stirling , ral-95 - 021 ( 1995 ) , to be published in phys .
b. atlas technical proposal , cern / lhcc/94 - 43 , lhcc / p2 ( 1994 ) .
cms technical proposal , cern / lhcc/94 - 38 , lhcc / p1 ( 1994 ) .
a. stange , w. marciano , and s. willenbrock , 49 1354 1994 ; d * 50 * , 4491 ( 1994 ) .
d. froidevaux and e. richter - was , cern - th-7459 - 94 ( 1994 ) .
s. mrenna and g. kane , calt-68 - 1938 ( 1994 ) .
t. liss , private communication .
d. amidei , private communication .
w. foster , private communication .
height0pt fig .
3 - differential cross section for @xmath95 versus the invariant mass of the @xmath3 boson and @xmath40 jet at ( a ) the tevatron and ( b ) the lhc . also shown
are the most important backgrounds ( @xmath96 denotes @xmath3-gluon fusion ) .
each event contributes four points to the distribution , each weighted by 1/4 , due to the ambiguity in the @xmath40 vs. @xmath54 and the two - fold ambiguity in the neutrino longitudinal momentum .
the signal is prominent at the tevatron , but is masked by the @xmath6 and @xmath3-gluon - fusion backgrounds at the lhc .
4 - same as fig .
3(a ) , except the @xmath40 jet with the highest rapidity with respect to the proton beam direction is selected as the @xmath40 from top decay , and the neutrino solution with the smaller absolute longitudinal momentum is selected .
this narrows the distribution from the signal in comparison with fig . | we consider single - top - quark production via the weak process @xmath0 at hadron colliders .
this process may provide the best measurement of the magnitude of the cabbibo - kobayashi - maskawa matrix element @xmath1 .
we show that a signal can potentially be observed at the fermilab tevatron with 3 @xmath2 of integrated luminosity .
in contrast , the signal is masked at the cern large hadron collider by top - quark pair production and single - top - quark production via @xmath3-gluon fusion .
-.5 in 6.5 in # 1 # 2 # 3 phys .
rev .
* # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys .
rev .
lett . *
# 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys .
rev .
d * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys .
lett .
* b#1 * , # 2 ( # 3 ) # 1 # 2 # 3 nucl .
phys .
* b#1 * , # 2 ( # 3 ) # 1 # 2 # 3 rev .
mod .
phys . * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 z. phys .
c * # 1 * , # 2 ( # 3 ) * single - top - quark production via @xmath4 * + + department of physics + university of durham + durham dh1 3le + england + + department of physics + university of illinois + 1110 west green street + urbana , il 61801 + |
hot x - ray emitting gas constitutes the dominant baryonic component in groups and clusters of galaxies .
the metals in this intracluster medium ( icm ) are believed to originate mainly in material ejected from group and cluster galaxies by supernovae @xcite , with a smaller portion driven out by galaxies through galaxy galaxy and galaxy icm interactions ( @xcite , but see also @xcite ) . besides galactic stellar winds ,
other non - gravitational processes have also played a role in redistributing the enriched gas , and more generally , in shaping the thermodynamic properties of the icm .
these processes include radiative cooling @xcite , and heating and mixing of gas by active galactic nuclei @xcite .
the spatial distribution of metals , and to some extent the temperature structure of the icm , reflect the action of such processes , thus offering insight into the importance and nature of the mechanisms , other than gravity , that have established the properties of the icm at the present epoch .
groups of galaxies are particularly interesting in this context because such processes are expected to be relatively more important in low - mass systems due to their shallower gravitational potentials .
iron and silicon comprise the major diagnostic elements for studies of the icm abundance distribution , as these elements give rise to the most prominent lines in the soft x - ray spectrum of thermal plasmas , and are among the elements for which there is a distinctively different yield in type ia and type ii supernova explosions according to standard supernova ( sn ) models ( see , e.g. , @xcite and references therein ) . whilst iron is predominantly produced by sn ia
, silicon is more evenly mixed between the two sn types .
the ratio of si to fe abundance therefore provides valuable information on the relative importance of sn ia vs. sn ii in enriching the icm .
abundance and temperature profiles of massive clusters have received considerable attention ( e.g. , @xcite ) .
it is now well established that clusters with cool central regions typically exhibit high central fe abundances ( e.g. , @xcite ) which can be attributed predominantly to sn ia enrichment by the stellar population of the central bright cluster galaxy @xcite .
clusters with no evidence for strong central cooling tend to display a more uniform fe distribution . at large radii ( @xmath3 , where @xmath4 is the radius enclosing a mean density of 200 times the critical density )
, @xmath5 typically drops to a value @xmath6 z@xmath1 @xcite . as suggested by , for example ,
@xcite , this could result from sn ii dominated enrichment at an early stage in the cluster formation via energetic starburst winds .
such a picture receives support from the observation that a substantial fraction of the icm metals were already in place by @xmath7 @xcite .
the situation in lower - mass systems is much less clear , despite their relatively larger importance for the cosmic baryon and metal budgets .
historically , this has reflected the difficulty in assembling large samples of groups bright enough to allow detailed , spatially resolved x - ray spectroscopy .
earlier works based on _ rosat _ and _ asca _ data typically featured only a handful of @xmath8 kev groups , often included as part of larger cluster - dominated samples ( e.g.@xcite ; @xcite ) . such studies suggested a @xmath9 ratio roughly solar in groups @xcite , about half the value found in early cluster studies @xcite , with an indication that si / fe rises above the solar ratio outside group cores ( e.g.@xcite ; @xcite ) .
however , the sensitivity and spatial resolution of _ rosat _ and _ asca _ imposed natural limits to the amount of detail one could obtain on the spatial variation of fe and si and hence the relative importance of the two supernova types at different group radii @xcite . for the same reasons , it is possible that the so - called fe and si biases @xcite may have affected some of these earlier results , at least for groups with thermally complex core regions . with the current generation of x - ray telescopes ,
the situation has improved dramatically .
recent _ xmm _ observations have shown that temperature profiles of cool - core groups display a striking similarity and typically peak at a radius @xmath10 @xcite , which corresponds to @xmath11 for the mass profiles derived for these groups .
another recent study , also based on _ xmm _ observations , is the work of @xcite , who investigated the distribution of temperature , density , metal abundance , and derived quantities using 2-d spectral mapping of a sample of 14 groups .
this work showed that , as for clusters , central metal excesses are present in groups with a cool core , a result already hinted at in earlier _ asca _ studies @xcite . while _ chandra _ lacks the sensitivity of _ xmm - newton _ , it has a superior spatial resolution , enabling a more robust removal of point sources , and also experiences a lower and more stable instrumental background .
_ chandra _ data can therefore complement _ xmm _ observations of groups , particularly in the group cores , but can also serve as an important validity check on _ xmm _ results for the x - ray faint outskirts of groups due to the lower background . here
we utilize this to derive radial profiles of temperature and the abundance of fe and si from _ chandra _ observations of a sample of 15 groups .
particular emphasis is given to combining the results obtained for individual groups to explore statistical trends within the sample .
one purpose of doing so is to obtain a clearer picture of the content and behaviour of metals in group outskirts , where most of the intragroup gas resides but where the low x - ray surface brightness often prohibits robust constraints for individual systems .
in addition , this approach allows an investigation into the signatures of feedback in group gas , including the origin of metals in sn and the redistribution of gas by active galactic nuclei ( agn ) , in more detail than is possible on a group - by - group basis .
furthermore , many numerical simulations of groups and clusters now include chemical evolution of the icm through the actions of galactic starburst winds ( e.g. , @xcite ) , agn outflows @xcite , and galaxy icm interactions @xcite .
therefore , it also seems timely to provide a firm observational baseline against which to test the input physics in such simulations , from the perspective of their predictions for the chemical ( and thermal ) properties of the icm in the group regime .
the paper is structured as follows . in section [ sec , sample ]
we describe the group sample and data analysis procedure , and section [ sec , results ] presents the main results for each group .
section [ sec , syst ] discusses the impact of systematic errors on our results . in section [ sec ,
combine ] we stack the results for individual groups , to investigate and quantify systematic trends across the sample .
section [ sec , conclude ] summarizes the results and presents our conclusions . in a companion paper ( rasmussen & ponman , in prep . ; hereafter paper ii ) , we discuss the implications of these results for the history and nature of galactic feedback and chemical enrichment within the sample .
@xmath12 km s@xmath13 mpc@xmath13 is assumed throughout , and all errors are given at the 68 per cent confidence level unless otherwise stated .
our sample is predominantly based on the 25 gems groups from the ` g'-sample of @xcite .
this subsample comprises those gems groups that display group - scale x - ray emission ( spatial extent of @xmath14 kpc ) , distinguished this way from systems that remain x - ray undetected and those in which the diffuse emission can be associated with the central galaxy rather than with the group as a whole . from these 25 groups , we effectively imposed a flux limit on the sample by only selecting systems with available _ chandra _ archival data containing more than 6,000 acis net counts from diffuse emission .
this was done in order to enable robust , spatially resolved spectroscopy in a number of regions for each group . in addition , only groups more distant than 20 mpc were included , to ensure that data extended well outside the central group galaxy . as one of our aims is to investigate signatures of galactic feedback in group gas , and since we wish to study the radial distributions of gas properties , this study requires systems which appear reasonably undisturbed by recent merger activity and associated gas mixing .
we therefore also excluded ngc 5171 from the sample , an obviously unrelaxed system already studied in detail by @xcite .
this left 11 groups , to which we have added a further four groups with data in the _ chandra _ archive that conform to the same criteria .
the resulting sample includes two hickson compact groups ( hcg 42 and 62 ) and a fossil group ( ngc 741 ) according to the definition of @xcite , and spans a temperature range @xmath152.1 kev , with a median temperature of @xmath16 kev and median redshift of @xmath17 .
it is worth emphasizing that the selected groups are all reasonably x - ray bright and that the sample is by no means statistically representative .
in fact , the invoked net counts criterion alone eliminated 13 of the 25 gems systems from which our sample was largely drawn , resulting in a sample lower limit of @xmath18 erg s@xmath13 .
three of our 15 groups are also included in the _ xmm _ sample of @xcite , with a further 6 included in the _ xmm _ sample of @xcite , enabling cross - comparison of the results obtained from the two instruments . for all our data sets , calibrated event lists
were regenerated using ciao v3.3 .
for very faint mode observations , the standard additional background screening was carried out .
bad pixels were screened out using the bad pixel map provided by the pipeline , and remaining events were grade filtered , excluding _ asca _
grades 1 , 5 , and 7 . periods of high background were filtered using 3@xmath19 clipping of full chip lightcurves , binned in time bins of length 259.8-s and extracted in off - source regions in the 2.57 kev band for back - illuminated chips and 0.312 kev for front - illuminated chips .
blank - sky background data from the calibration database were screened and filtered as for source data , and reprojected to match the aspect solution of the latter .
point source searches were carried out with the ciao task ` wavdetect ' using a range of scales and detection thresholds , and results were combined .
source extents were quantified using the @xmath20 detection ellipses from ` wavdetect ' , and these regions were masked out in all subsequent analysis .
diffuse emission from individual group galaxies other than the central early - type present in all groups was also masked out . to provide a rough idea about the spatial distribution of x - ray emission in each system ,
images of all groups are presented in fig .
[ fig , mosaic ] , showing the adaptively smoothed , background - subtracted and exposure - corrected 0.72.5 kev emission on the relevant central ccd ( the acis - i array for ngc 6338 and ngc 7619 , acis - s2 for ngc 2300 , and s3 for all others ) .
these images were generated following the procedure outlined in @xcite .
although small - scale substructure is visible in some of the systems , with the second brightest group galaxy also seen in x - rays in a few cases , most of the groups are evidently reasonably dynamically relaxed , supporting the use of one - dimensional radial profiles to describe the spatial variation in hot gas properties of the systems .
all regions used for spectral extraction were centred on the peak of the diffuse x - ray emission , which in most cases coincided to within 12 arcsec ( and always within 5 arcsec ) with the optical position of the brightest group galaxy as listed in the hyperleda database . for each ccd in each data set , the appropriate blank - sky data were scaled to match off - source count rates in the 1012 kev band and then used to generate background spectra .
all spectra were fitted in xspec v11.3 in the 0.75 kev band , with the low - energy cut - off chosen such as to reduce contamination from residual galactic soft x - ray emission not accounted for by our blank - sky background subtraction ( see section [ sec , syst ] for details ) .
indeed , no systematic soft residuals between source and background spectra were seen within this energy band for any of the groups . to put all groups on an equal footing ,
all radii were normalized to @xmath2 , using @xmath21 which is an empirical relation based on data outside the cluster core for a sample of 39 groups and clusters , including eight systems with @xmath22 kev @xcite . in order to obtain a representative mean temperature @xmath23 for use with equation ( [ eq , r500 ] ) , as well as a corresponding mean abundance @xmath24 , and to create spectral weights for subsequent fits in individual annuli ,
a source spectrum was first extracted within the radial range 0.10.3 @xmath2 .
this range was chosen because it extends to the maximum radius covered by acis for the more nearby systems , while in all cases being well outside the optical extent of the central galaxy .
the latter was quantified as @xmath25 , the ellipse outlining a @xmath26-band isophotal level of 25 mag arcsec@xmath27 .
the associated fit results should therefore remain largely uncontaminated by x - ray emission from the interstellar medium and any unresolved point sources in the central galaxy .
as we will see , this choice of radial range also excludes any cool core , in accord with our usage of equation ( [ eq , r500 ] ) .
the fits to these spectra all adopted an absorbed thermal apec model in xspec , assuming the abundance table from @xcite . using the derived value of @xmath23 with equation ( [ eq , r500 ] ) ,
@xmath2 was then re - evaluated iteratively , until the resulting values of @xmath23 varied by less than 1 per cent . throughout
, we will refer to the values of @xmath23 and @xmath24 obtained in this way as the mean temperature and abundance of each group .
spectra were then extracted in concentric annuli containing at least 2000 net counts in the 0.71.5 kev band , and accumulated in bins of at least 25 net counts .
spectral response products were weighted by a model of the ` mean ' spectrum extracted above .
again , an absorbed apec model was first fitted to the data , yielding a mean temperature and abundance for the relevant annulus .
a second fit was then performed using an absorbed vapec model , with all element abundances tied together except for fe and si .
for spectra extracted inside the region covered by the central early - type galaxy , a power - law of fixed index @xmath28 was added to the spectral model , to account for any unresolved contribution from low - mass x - ray binaries . to investigate the presence of multiple thermal components and address the importance of systematic errors associated with the fe bias @xcite
, we also fitted a 2-@xmath29 plasma model to all bins inside the optical extent of the central galaxy , with both the fe and si abundance tied for the two thermal components .
where either the resulting fe or si abundance was significantly different from the 1-@xmath29 result ( at 90 per cent confidence ) , we adopted the best - fitting 2-@xmath29 values for both metals , confirming that a 2-@xmath29 model provided a better fit in all cases .
as discussed in more detail in section [ sec , syst ] , this was relevant for only five groups .
we note that the adoption of the 2-@xmath29 results for the abundance in these five cases does not affect @xmath30 significantly , as this quantity has been derived within 0.10.3 @xmath2 , which is generally well outside the region for which 2-@xmath29 models provide superior fits in the relevant groups .
owing to the layout of the acis chips , we do not have full azimuthal coverage of the groups at large angular distances from the group centres .
our results at large radii could therefore be sensitive to azimuthal variations in fitted parameters . in an attempt to accommodate the systematic uncertainties
associated with this , we used azimuthal variations at smaller radii to gauge the variations further out . for each group , a number of spectra were extracted from an annulus outside @xmath25 extending to the maximum radius of full coverage . to mimic the situation at large radii , only data within a narrow strip across this annulus
were used , masking out the remaining regions to achieve , if possible , 2000 net counts within the strip , as for the full - coverage annuli .
this was repeated a number of times with varying strip position angles .
the resulting spectra were then each fitted with a vapec model as described above , and the resulting standard variations on relevant parameters ( @xmath29 , @xmath31 , @xmath32 ) adopted as the systematic uncertainties at large radii .
the mean uncertainties resulting from this approach were 8 , 34 , and 58 per cent on @xmath29 , @xmath31 , and @xmath33 , respectively .
these errors have been added in quadrature to the statistical errors obtained for all annuli with less than 75 per cent azimuthal coverage ( for most groups just one or two radial bins ) .
table [ tab , groups ] records the position of the x - ray peak of each group , group distance @xmath34 corrected for virgocentric infall , the useful acis exposure time @xmath35 , @xmath23 and @xmath30 from the ` mean ' spectrum , @xmath2 from equation ( [ eq , r500 ] ) , and the galactic value of the absorbing column density @xmath36 from @xcite as adopted in the fits .
note that the statistical uncertainty on @xmath2 based on that of @xmath23 alone is typically on the order of just a few per cent .
the actual uncertainties on @xmath2 are therefore likely to be dominated by systematics associated with the use of equation ( [ eq , r500 ] ) , but a quantification of this is beyond the scope of this paper . [
cols="<,^,^,^,^,^,^,^,^ " , ]
in this section we present the results for the projected radial profiles of temperature , fe and si abundance , and the ratio @xmath37 .
as one of our aims is a statistical investigation of radial trends across the sample , the profiles have not been deprojected , since this would remove the statistical independence of the individual radial bins .
however , we discuss the impact of deprojecting the data in section [ sec , syst ] , showing that this would not compromise our general conclusions . in order to compare observed abundances and their ratios to those expected from sn ia and sn ii enrichment , we adopted sn model yields taken from the literature . for sn ia
, the currently favoured delayed - detonation wdd2 scenario is assumed @xcite , while for core - collapse supernovae ( including types ib and ic ) , we have used the recent yields of @xcite .
the latter were mass - averaged using a salpeter initial mass function ( imf ) for the progenitors over the range 1050 m@xmath1 ( see @xcite and @xcite for a justification of this choice of imf ) , so that @xmath38 is the mass of the @xmath39-th element produced in a star of mass @xmath40 , with @xmath41 .
we have assumed a fractional metal mass of the sn ii progenitors of @xmath42 ( i.e. @xmath43 z@xmath1 ) , but the exact choice has no strong bearing on the resulting fe and si yields ( see , e.g. , @xcite ) .
the resulting average stellar yields are @xmath44 m@xmath1 and @xmath45 m@xmath1 for fe and si from sn ia , and @xmath46 m@xmath1 and @xmath47 m@xmath1 for core - collapse supernovae including sn ii .
we note that the yield uncertainties for sn ii are substantial , a factor @xmath48 for fe and @xmath49 per cent for si ; the yields adopted here are bracketed by the results of the different models considered by , for example , @xcite , and so can be viewed as representative of typical values in the literature . with the adopted abundance table ,
the nominal abundance ratios expected for pure sn ia and sn ii enrichment are @xmath50 z@xmath51/z@xmath52 ( sn ia ) and 2.62 z@xmath51/z@xmath53 ( sn ii ) .
note that this implies sn ii dominance ( in absolute numbers ) for @xmath54 , and that a solar abundance ratio requires 3.4 sn ii per sn ia . by adopting the sn ia yields of @xcite , we restrict ourselves to the ` standard ' picture of a reasonably homogeneous population of sn ia for simplicity .
we note , however , that recent developments indicate that the situation could be more complex , with evidence for a diversity in sn ia yields in cluster cores @xcite and for two distinct populations of sn ia @xcite .
radial temperature profiles for all groups derived from single - temperature models are presented in fig .
[ fig , t ] , with the radii of all annuli normalized to @xmath2 .
it is worth noting that emission is detected to fairly large radii in most of the groups , with the midpoint of our outermost radial bin corresponding to roughly 75 per cent of @xmath2 .
this can be ascribed to the relatively low and stable _ chandra _ background .
in general , the temperature profiles appear remarkably similar , with all but one system ( ngc 4125 ) showing evidence for cooler gas in the central regions .
the temperature profiles typically peak around @xmath55 and decline at large radii .
there is also an indication that the profiles flatten in the central regions , in general agremeent with the group results of @xcite .
compared to analogous profiles for relaxed clusters , this contrasts with the projected _ chandra _ temperature profiles of @xcite , the deprojected ones of @xcite , and the deprojected _ xmm _ profiles of @xcite , for which there is no clear evidence for a flattening at small radii . defining the gas entropy as @xmath56 , the deprojected cluster entropy profiles of @xcite _
do _ show a central flattening , but this was attributed to a central core in the density distribution rather than a flattening of the temperature profile .
the typical location of the temperature peaks in our sample is in excellent agreement with the _ xmm _ results of @xcite , who also found temperature peaks at @xmath57 for their cool - core groups and poor clusters .
their sample has four groups in common with ours ( ngc 533 , mkw 4 , ngc 4325 , and ngc 5044 ) , and has @xmath2 evaluated directly from the mass profiles derived for each group .
the good agreement with our results therefore also suggests that our estimates of @xmath2 are generally not seriously biased . as also noted by @xcite for their sample , the temperatures peak at larger overdensity than in clusters , where the profiles turn over at @xmath58 @xcite .
it is clear , however , even when normalized to @xmath2 , that the exact radius @xmath59 at which the peak occurs ( in other words , the radial extent of the cool core ) varies slightly from group to group .
naturally , this result will to some extent depend on the spatial resolution resulting from our binning criterion and thus the surface brightness of the group emission , but we note that the adopted spatial resolution does not depend systematically on , for example , group temperature .
we discuss the location of the temperature peak in more detail in section [ sec , temp2 ] .
the abundance profiles of fe and si are plotted in fig .
[ fig , abund ] . in all systems except ngc 2300 , and possibly the non cool core group ngc 4125
, there is clear evidence for a central excess of iron , as also seen for cool - core clusters .
this central excess often extends well beyond the optical extent of the central galaxy , and in some cases shows an off - centre peak inside @xmath25 , with a decline in abundance in the very centre .
ngc 533 is perhaps the clearest example of this , but the majority of groups with a well - resolved core ( roughly 8/11 ) shows some evidence of this feature .
this fe peak is typically accompanied by a similar feature in the si distribution , most clearly in hcg 42 , mkw 4 , and hcg 62 , with a hint of similar behaviour in a few more groups .
potentially , the peak could be ascribed to fe / si bias inside the central galaxy , lowering the derived abundance in the very core , but as discussed below , we find little evidence that this effect is important for our analysis in general .
in addition , such a peak is commonly seen also in well - resolved cluster cores where the fe bias is generally expected to be less important , e.g. , centaurus @xcite , perseus @xcite , a1795 @xcite , and a2199 @xcite .
alternative explanations for the peak include agn redistribution of enriched gas ( e.g. , @xcite ) , resonant scattering of emission line photons in group cores @xcite , or gravitational sedimentation of helium nuclei in the core @xcite .
we will discuss these possibilities in detail in paper ii . at large radii ,
a common feature of the fe distributions is a drop in @xmath31 to a value of @xmath60 z@xmath1 . in some groups ,
a rise in @xmath33 is also seen , largely independent of the behaviour of the fe profile .
we note that there is no indication that derived profiles for the two hickson groups or the fossil group ngc 741 differ significantly from those of the rest of the sample in this context . the silicon - to - iron ratio @xmath9 is plotted for all groups in fig . [ fig , ratio ] .
in many cases the ratio in the central bin is seen to be consistent with the solar ratio , with a contribution from sn ii formally required for about half the groups . for the other half ,
there is clear evidence for subsolar abundance ratios , consistent with enrichment by sn ia alone .
we do not believe there is any fundamental difference between the two subsamples if split according to this , as @xmath9 in the central bin clearly spans a continuous range of values rather than showing a bimodal distribution .
given the variation in spatial resolution among the groups and the uncertainties on @xmath9 , we caution further against such an interpretation at this stage .
for example , we find no evidence that the two subsamples cluster differently in the parameter space spanned by @xmath23 , @xmath24 , and @xmath61-band luminosity of the central galaxy .
nevertheless , we will revisit this issue in paper ii on the basis of integrated fe and si masses inside @xmath25 .
as discussed by @xcite , there is observational evidence for diversity in sn ia yields in the cool cores of more massive clusters , with solar si / fe ratios arising entirely from sn ia .
while our results allow for this possibility , we can not easily confirm that this result extends down into the group regime , as it is based on the abundances of additional elements such as s , ca , and ar . at the lower temperatures prevalent in groups , the abundances of these elements , with prominent emission lines in the range @xmath62 kev , are not in general robustly constrained . in the outer regions ,
the interpretation of our data remains unaffected by this ; here a contribution from sn ii is universally required within our sample , and the abundance ratios , given the typical uncertainties in sn ii yields of a factor @xmath48 ( e.g. @xcite ) , are consistent with pure sn ii enrichment in all cases .
many of the groups exhibit properties which warrant some additional comments .
below , we briefly discuss each group in the sample in turn , with emphasis on the derived abundances .
where relevant , we have also attempted a comparison of the derived fe profile to existing measurements , in particular the 2-@xmath34 _ xmm _ results of @xcite ( f06 and f07 in the following ) , while correcting for differences in adopted abundance tables .
it should be noted , however , that due to our incomplete azimuthal coverage at large radii , direct comparison is not always straightforward for sources with some degree of asymmetry in the fe distribution . + * ngc 383 * : the central galaxy hosts the bright radio source 3c 31 , representing one of the few examples of a central group galaxy showing a clearly identified large - scale radio jet with a well - constrained jet power @xcite . despite this , the @xmath29 and @xmath63-profiles of this group do not exhibit any clear distinguishing features compared to the rest of our sample .
for the temperature profile , this result is in accord with the group results of @xcite , which show that the slope of @xmath64 inside the cool core does not show any strong dependence on the radio power of the central radio source . +
* ngc 507 * : an example of a group in which radio lobes are present around the central galaxy @xcite and in which the highest - metallicity gas is displaced from the group core , perhaps as a consequence of radio outbursts lifting enriched gas out of the core .
if so , one might expect to see some degree of asymmetry in the fe distribution in the core .
the 2-@xmath34 fe map of f07 does reveal some asymmetry inside @xmath65 , showing a high - metallicity blob south - west of the group core , but it is not immediately obvious that this is necessarily associated with an agn outflow . to achieve a satisfactory fit in the outermost radial bin in our data , it was necessary to exclude the second brightest group galaxy , ngc 499 , out to 6 arcmin ( @xmath66 kpc ) .
comparison to the single-@xmath29 fe profile of f07 shows good overall agreement , including the off - centre fe peak at @xmath67 and the decline to @xmath68 z@xmath1 at @xmath69 .
+ * ngc 533 * : this group presents one of the clearest examples of an off - centre peak in the fe distribution within our sample . as for ngc 507 , which also hosts such an fe peak ,
1.4-ghz nvss data @xcite reveal a central radio source .
inside @xmath70 our fe results are consistent with the f07 value of @xmath71 z@xmath1 , as is our detection of the supersolar fe peak at @xmath72 .
the clear decline to 0.1 z@xmath1 at large @xmath73 is not obvious from the f07 results , however , which show considerable scatter outside @xmath74 .
we speculate that this relates to our assumption of circular symmetry and our incomplete azimuthal coverage at large radii .
the 2-@xmath34 fe map of f07 suggests an elongated abundance structure on large scales , with the acis - s array in our _ chandra _ data running very nearly perpendicular to the major axis of this structure .
this implies that our result includes low - metallicity gas at the largest radii , beyond those included in the f07 analysis , and that we are inevitably missing some of the higher - metallicity gas seen by f07 . +
* ngc 741 * : with a group x - ray luminosity of @xmath75 erg s@xmath13 @xcite , and a @xmath26-band magnitude difference between the first and second brightest galaxy ( the spiral ugc 1435 ) within @xmath2 of @xmath76 , this system meets the criteria for fossil groups introduced by @xcite .
its temperature and metallicity profiles appear similar to those of most other systems studied here , though the fe abundance at large radii is clearly lower than average .
it is also a group with a powerful central radio source and large - scale radio emission ( e.g. , @xcite ) , although the latter is probably a tailed radio source associated with the nearby group member ngc 742 rather than with the central elliptical itself ( jetha et al .
, in prep . ) .
+ * ngc 1407 * : this is probably a dynamically evolved group , with overall optical properties not too dissimilar to those of the fossil system ngc 741 @xcite .
its fe and si profiles appear fairly typical for the sample . + * ngc 2300 * : this is the only group within our sample with a spatially resolved fe profile outside @xmath25 that does not show clear evidence for a radial decline .
the si profile is also somewhat atypical of the sample , showing a clear rise with radius outside the core .
the relatively low and near constant fe abundance , as well as the radial rise in si , is in very good agreement with earlier _ rosat _ and _ asca _ results @xcite , although the _ xmm _ study of f06 seems to find slightly higher fe abundances inside @xmath77 than those seen here . with the central early - type galaxy ( ngc 2300 itself ) having a stellar mass @xmath78 times that of the second largest galaxy , ngc 2276 @xcite , the central galaxy
could potentially have been almost solely responsible for the chemical enrichment of the hot gas in this system .
this is also an example of a group in which ram - pressure stripping of ( presumably enriched ) galactic gas is taking place @xcite , illustrating that an sn driven outflow is not the only mechanism for enriching the hot gas in the group .
we note that there is also evidence of a recent merger in the central early - type @xcite , with further evidence for this seen in the icm pressure map of @xcite .
given that other features of this group seem typical of our sample , such as the presence of a cool core and a fairly undisturbed x - ray morphology , the question remains how to interpret the uniformly low fe abundance and the rise in @xmath33 with radius .
we note that @xmath31 at large radii is not atypical of the sample , nor is the radial profile of si / fe .
this suggests a suppression of both fe and si enrichment in the group core . in light of the evidence of a recent merger in the central early - type ,
one possible explanation is that a cool , low-@xmath63 system has dropped into the group core , thus diluting any fe excess but preserving the cool core .
another possibility , potentially induced by the very low central temperature , is that high-@xmath63 gas has cooled out in the group core .
based on the measured x - ray flux , the fitted spectral parameters , and the gas density profile of @xcite , we estimate a central cooling time of only @xmath79 yr . this would render such cool - out a viable scenario , in particular in light of the absence of any strong radio source in the central early - type ( 1.4 ghz radio power @xmath80 w hz@xmath13 based on the nvss flux ) which could indicate ongoing agn heating activity . in order to remove any fe excess and explain the ` suppressed ' fe level in the core , a total fe mass of @xmath81 m@xmath1 must have cooled out inside the typical radial extent of the fe excess of @xmath82 , assuming an initial fe abundance of @xmath71 z@xmath1 ( see section [ sec , combine ] for this choice of parameters ) .
this is equivalent to a total gas mass of @xmath83 m@xmath1 , which does not seem an unreasonably large amount .
in addition , cooling would also have raised the central entropy through the removal of low - entropy gas from the x - ray phase . among
the nine of our groups also included in the f06 and f07 samples , ngc 2300 does show the highest central entropy , lying @xmath84 above the sample mean derived inside @xmath74 ( f06 ) .
one concern for the cooling scenario , however , is that the cooled gas is likely to participate in star formation @xcite , but there is no indication that the central early - type is unusually blue , displaying a @xmath26@xmath61 colour of 4.2 . based on the x - ray data alone , we conclude that the cooling scenario provides a possible explanation for the lack of an fe excess in this group , but further tests of this would require constraints on the amount of cold gas and young stars in the central early - type galaxy . + * hcg 42 * : the observed range of @xmath85 z@xmath1 inside @xmath74 is consistent with the f07 results , as is the decline at large radii .
however , finding a low value of @xmath86 z@xmath1 outside the cool core , we do not confirm the value of @xmath87 z@xmath1 seen in the outermost bin of f07 , certainly in part because we do not have the statistics required to resolve this narrow region around @xmath69 .
+ * mkw 4 * : the presence of an off - centre fe peak in this group was already claimed by @xcite , who furthermore noted that no radio activity or merger evidence has been reported for the system . indeed , nvss data show no evidence of a central radio source , nor of any larger - scale radio emission surrounding the central galaxy , ngc 4073 @xcite .
this is in clear contrast to ngc 507 and ngc 533 , where this fe peak is accompanied by the presence of a central radio source .
we confirm the high central fe abundance of @xmath48 z@xmath1 seen by f07 , along with the sharp decline , starting at @xmath65 , towards a value of 0.1 z@xmath1 at large radii . + * ngc 4125 * : this group is the lowest - temperature system in the sample .
it shows no cool core , despite appearing reasonably relaxed in the x - ray ( fig .
[ fig , mosaic ] ) .
subdividing the innermost radial bin and fitting spectra for each sub - bin , each only @xmath88 pc wide , still does not reveal any temperature drop . as there appears to be a point - like x - ray source at the centre , detected above the local background at @xmath89 significance , one possibility is that the central temperature rise could be due to recent agn heating .
however , although a radio source is present in nvss data inside @xmath25 , this source is displaced from the optical and x - ray centre by 0.8 arcmin ( 4 kpc ) . to investigate the nature of the central x - ray source ,
its spectrum was fitted with a power - law absorbed by the galactic value of @xmath36 .
this provided an unacceptable fit , even when allowing for intrinsic absorption beyond the galactic value ( red .
@xmath90 for 9 degrees of freedom ) .
in contrast , an apec plus power - law model fits well ( red.@xmath91 ) , yielding @xmath92 and @xmath93 for the thermal plasma , and @xmath94 for the power - law , with a total , unabsorbed luminosity of @xmath95 erg s@xmath13 .
the plasma parameters are in good agreement with those found on larger scales in the group core , and the power - law with a contribution from low - mass x - ray binaries in the central elliptical .
detection of a central point source might therefore simply be attributed to a strongly peaked surface brightness profile , in line with the _ rosat _ finding of @xcite for this group .
we therefore conclude that there is no strong x - ray evidence for ongoing agn heating having raised the central icm temperature .
further support for this interpretation is provided by the low x - ray luminosity derived for the very central regions , which is well below typical agn luminosities .
an alternative explanation for the central temperature rise could be sought in the fact that the central elliptical is ` dynamically young ' @xcite , so a merger event within the past few gyr could have disturbed the central regions and mixed in hotter gas from larger radii .
the derived central cooling time is very short , however , @xmath96 yr as for ngc 2300 , but increasing to more than 3 gyr at @xmath25 as a consequence of the sharply peaked surface brightness , and hence density , profile @xcite . as this is a factor of three above the sound crossing time within that region
, it may suggest that the gas density distribution has largely recovered from such a merger disturbance , but significant central cooling has yet to commence .
another possibility is that the situation in ngc4125 is somewhat similar to that seen for the fossil group ngc6482 , one of the coolest known fossils , with a temperature of only @xmath97 kev outside the central 10 kpc @xcite . despite looking relaxed in the x - ray and displaying a central cooling time of @xmath98
yr , this system also shows no cool core .
a possible explanation is that any gas taking part in a steady - state cooling flow in this group is actually _ heated _ by gravitational @xmath99d@xmath100 work as it moves inwards , owing to the very cuspy total mass profile of the group .
@xcite find that , in such a scenario , the observed temperature profile can be explained by a cooling - flow mass accretion rate of just 2 m@xmath1 yr@xmath13 .
the sharply peaked density profile seen for ngc 4125 is consistent with the idea that a similar mechanism could be at work in this group . + * ngc 4325 * : this is a group in which the extent of the metal excess in the core substantially exceeds the optical extent of the central galaxy .
no central radio source is seen in nvss data @xcite , but there is evidence of a previous weak agn outburst , with an indication that another outburst may be about to occur @xcite .
the observed near - constancy of @xmath101 z@xmath1 inside @xmath74 agrees well with the f07 results , as does the very steep decline to a value @xmath102 z@xmath1 at @xmath103 . + * hcg 62 * : this group has been extensively studied with _
rosat _ , _ asca _ , _ chandra _ , and _ xmm _ @xcite .
it harbours one of the clearest examples of x - ray cavities in groups @xcite , indicating past radio activity . the older ciao task ` mkrmf '
was used to create spectral responses for this group , due to a lack of relevant calibration data for the newer task ` mkacisrmf ' otherwise used in this study . despite this and the complex x - ray morphology in the core of this system , the combined _ chandra _ and _ xmm _ study of @xcite finds a temperature profile very similar to the one derived here , and the roughly solar fe and si abundances seen in the core agrees well with the results of @xcite and @xcite . the sharp decline in the fe profile ,
beginning at @xmath67 and reaching a value of @xmath104 z@xmath1 at @xmath105 , is also in good agreement with f07 . +
* ngc 5044 * : this system has been studied extensively with both _
xmm _ and _ chandra _ @xcite . at @xmath82 ,
our results are consistent with the solar fe abundance found by f06 . in our outermost radial bin ,
we find @xmath106 z@xmath1 , confirming the low fe abundance of 0.10.15 z@xmath1 derived at large radii by @xcite and @xcite .
+ * ngc 5846 * : the fe map of f06 suggests a highly symmetric abundance structure , and supports our detection of an off - centre fe peak and a decline to 0.1 z@xmath1 at large radii . + * ngc 6338 * : reaching a peak temperature of @xmath107 kev , this is the hottest system in our sample . in fig .
[ fig , mosaic ] , extended emission is also seen from the second brightest galaxy .
this has been masked out to a radius of 1 arcmin in all spectral fits .
as for ngc 4325 , the central metal excess in this group extends well beyond the central galaxy , but is here also accompanied by a central radio source . +
* ngc 7619 * : ngc 7619 itself is one of a central pair of ellipticals of similar optical brightness in the group .
extended x - ray emission is also seen in fig .
[ fig , mosaic ] around the other pair member , ngc 7626 ; this was masked out to a radius of 2 arcmin in all spectral fits .
before proceeding to discuss the results in more detail , we address a number of analysis issues which could be affecting our results .
these include the presence of fe and si biases in the group core , and the reliability of our background estimates .
an immediate , although indirect , indication that such issues are not having a major impact on our results , is the fact , discussed below , that we detect the same overall radial trend in @xmath29 and @xmath63 for all systems , despite considerable variations in both source and background flux across the sample .
for the temperature profiles , the agreement with the _ xmm _ result of @xcite for the mean position of the temperature peak is encouraging . moreover , there is good overall agreement with derived fe profiles for the six groups overlapping with the _ xmm _ sample of @xcite , even at large radii where systematic uncertainties related to background subtraction could be important .
one issue which could call the robustness of our results into question is the fact that our profiles are projected ones , so all bins contain some contribution from emission further out . to test the importance of this effect , we deprojected the profiles for a few groups using the projct model in xspec . for simplicity
, a 1-@xmath29 apec model was fitted to the data .
the result for ngc 4325 is shown in fig .
[ fig , projct ] .
this group is typical of our sample in terms of the number of radial bins and the value of @xmath108 ; it also shows a near - solar si / fe ratio throughout , so the measurements should not be biased strongly by fixing the relative abundances .
the figure demonstrates reasonable overall agreement between projected and deprojected results .
as can be seen , projection tends to slightly reduce the temperature gradient in the core , as would be expected .
while the deprojected abundance profile recovers the increased fe level in the region inside @xmath109 , significant fluctuations between adjacent bins are evident in this region in disagreement with the projected profile .
these are most likely artefacts related to issues of fit stability , a problem commonly arising for spectral deprojections involving a large number of bins ( e.g. , @xcite ) .
results for less well - resolved systems support this suspicion and confirm the lack of systematic variations between projected and deprojected results .
we therefore conclude that our results are reasonably reliable representations of the three - dimensional profiles .
another issue is the fact that we have incomplete azimuthal coverage at large radii .
this could potentially play a role for individual groups in which a simple radial description of the abundance profile breaks down in the outskirts .
as mentioned , however , we have attempted to select systems with a reasonably regular x - ray morphology , and we have also tried to accommodate any remaining uncertainties related to this in our error analysis .
the good overall agreement with _
results at large radii indicates that this effect is not important , at least for the groups where a comparison is possible .
in addition , any bias would be strongly suppressed when averaging the results across the group sample , which is the approach taken in the next section .
we therefore believe our conclusions to be robust against this effect .
it is well known that high - quality observations of groups in many cases reveal disturbed central structures , showing that group cores can be thermally complex ( e.g. @xcite ) and require multi - component thermal models to properly fit their spectra . to investigate this
, we plotted the distribution of reduced @xmath110 values for all our fits , but found no clear trend with radius and no indication that multi - temperature models are generally _ required _ to describe the central regions .
nevertheless , particularly for systems such as ngc 383 , where the central cool regions are not well resolved because of our binning criterion , the presence of multi - phase gas could lead to artificially low abundance estimates in the innermost bin @xcite , and metallicities derived from 1-@xmath29 fits should here be regarded with some caution .
we tested for this effect by first radially subdividing the central radial bin and fitting spectra extracted within each of these two sub - bins .
eight of the 15 groups show evidence for @xmath29 variations between those two sub - bins . for all systems , the abundance values within the two sub - bins
were consistent with the full - bin value , and only one group displays evidence for @xmath63 variation between the two sub - bins ( ngc 383 , with the innermost sub - bin showing a lower value ) .
however , this does not in itself rule out the presence of multi - phase gas within each of these two sub - bins .
a further test therefore involved fitting two - temperature models to all bins within @xmath25 , as already described in section [ sec , specfit ] . only for five groups
do we find evidence that 2-@xmath29 models yield significantly higher abundances _ and _ provide superior fits , in some of the innermost radial bins ( fig .
[ fig , abund ] ) . for these , the 2-@xmath29 results for the abundances have been adopted , as plotted in fig .
[ fig , abund ] .
an example is ngc 507 , for which only the 2-@xmath29 fit recovers the high fe abundance of 23 z@xmath1 reported inside @xmath25 by @xcite on the basis of _ xmm _ data .
while 2-@xmath29 models certainly provide better fits in some cases , it is perhaps slightly surprising that they only yield significantly different abundances in a few of these , given the presence of central temperature variations in most of our groups .
a possible explanation is that we are typically sampling the core regions with rather finely spaced annuli .
the limited radial extent of each annulus may imply that mixing of temperature components may not be important within each bin .
for example , for the very core of mkw 4 , @xcite also find from the same _ chandra _ data that 2@xmath29 models do not produce significantly improved fits or higher @xmath31 , in agreement with our results .
a different conclusion was reached by @xcite based on _ xmm _ data , which @xcite attribute to the much larger spectral extraction region used by the latter authors .
support for this interpretation is provided by the work of @xcite , which shows that _ chandra _ spectra of elliptical galaxies can generally be described by 1-@xmath29 models , and that the multi-@xmath29 fits required to describe earlier _ asca _ spectra of the same objects were an artefact of sampling strong temperature gradients with poorer spatial resolution . the fact that we generally find 2@xmath29 models to yield better fits and higher abundances only for groups where our innermost bin covers a large fraction of @xmath25 , would seem to add support for this interpretation . indeed , as discussed in paper ii , for spectra covering the full region inside @xmath74 ( a region much larger than that covered by the central bin in all groups ) , we find that two - phase models are in most cases statistically required .
in addition , we note that the introduction of a @xmath111 power - law to account for unresolved x - ray binaries in the central early - type also mitigates the need for a 2-@xmath29 model in some cases .
it is finally worth noting that any remaining fe bias would only strengthen the result that an fe excess is present in the cores of most of our groups . qualitatively , therefore , the only significant feature potentially induced by the fe bias would be the off - centre fe peak seen in some systems .
however , some of our groups for which 2@xmath29 models provide better fits and significantly enhanced abundances in the central regions , such as ngc 507 , still show evidence for such fe and si peaks , suggesting this feature is reasonably robust to such a bias .
another important issue is the accuracy of our background subtraction .
this is mainly relevant at large radii , where the emission at all energies is typically dominated by the background .
as mentioned , we use blank - sky data to extract background spectra , scaled to match the 1012 kev count rates of our source data for each ccd . the resulting medium - to - hard background above @xmath112
kev seems properly accounted for in all cases , as there is no statistically significant residual emission above this energy within the spectral extraction regions employed for _ any _ data set . of potentially more concern
is the soft x - ray background below @xmath113 kev , whose intensity varies considerably across the sky . for 10 out of our 15 groups , the _ rosat _ all - sky survey @xmath114 ( 0.50.9 kev ) background count rates are within @xmath115 ( @xmath116 per cent deviation ) of the analogous value for the relevant blank - sky background data , lending support to our use of the latter for background estimation .
while most of our groups are thus not projected onto regions of significantly enhanced or depleted soft x - ray background , there are a few exceptions which all show enhanced background , most notably ngc 5044 and ngc 5846 which both lie close to the north polar spur . for most of our systems ,
intragroup emission covers the field of view ( @xmath2 is outside the field of view in most cases ) , so we can not in general use the outer portions of the observed fields to reliably assess the level of any residual soft background . instead , two independent tests were performed to assess the impact of varying the assumed background : we either simply rescaled the blank - sky background , or added a separate background model component in the fits to the group emission . in the first test , we repeated all fits for the three outermost radial bins of each group , with the blank - sky background normalization shifted by @xmath117 per cent .
the fits were performed in the 0.72.0 kev band only , to prevent the obvious mismatch resulting at higher energies from affecting the results .
we found that this in general induced negligible changes in the best - fitting temperature , which remained consistent with the value derived for the ` unperturbed ' background in all but two cases ( when increasing the background for ngc 4125 and ngc 6338 ) , for which the resulting fits were statistically unacceptable anyway . if anything , the best - fitting value of @xmath29 is slightly lower in all cases , so there is no indication that the temperature decline observed in all groups at large radii can be ascribed to inaccurate background subtraction .
we therefore conclude that the large - scale behaviour of the temperature profiles in fig .
[ fig , t ] is a robust result .
the resulting abundances are more sensitive to variations in the assumed background level .
for three of the groups ( hcg 42 , ngc 1407 , and ngc 2300 ) rescaling the background produced higher abundances in the outermost bin , inconsistent with the result obtained using our standard background level .
none of these cases yielded unacceptable fits , so we conclude that significantly higher values of @xmath63 in the outermost region probed for these systems are allowed , though not favoured , by the data . however , at least in the case of hcg 42 , there is good reason to believe that the existing abundance measurement is fairly reliable , as it compares well to the result derived independently from _ xmm _ data @xcite .
similar agreement is found for ngc 2300 when compared to earlier _ rosat _ and _ asca _ results ( see @xcite ) . as an additional test
, we also tried varying the background normalization by @xmath118 per cent for all groups . in the four cases where
the resulting fits did not become either significantly worse or statistically unacceptable , the derived @xmath29 and @xmath63 were consistent with the results obtained using the unperturbed background . as a second test , carried out for the five groups with significantly enhanced soft background , we included a spectral model to account for this excess background , and repeated all fits outside the cool core with this model added to that used to describe the group emission . following @xcite and @xcite , a fixed @xmath119 kev , @xmath120 z@xmath1 mekal background model was assumed , with only its normalization allowed to vary . only for ngc 5846 , which along with ngc 5044 shows the highest background excess ,
did we find significant changes in fitted parameters , with this ( improved ) fit yielding lower fe and si abundances , which we have consequently adopted . for ngc 5044 , no changes were seen , possibly because of the relatively high signal - to - noise ratio in the outermost bins for this group .
we confirmed that soft background excess is not an issue for the remainder of the sample .
for example , for relatively compact low surface - brightness sources such as hcg 42 and ngc 4125 , there is no statistically significant residual emission in the 0.72 kev band in a 2 arcmin wide annulus extending to the edge of the field of view . for the adopted soft background model , only 15 per cent of the total 0.50.9 kev flux ( used to determine the presence of a soft excess ) falls at energies above 0.7 kev .
the fact that energies below 0.7 kev are excluded in spectral fitting should therefore also reduce the problem .
finally , we do not see any systematic low - energy ( @xmath1211 kev ) residuals for any of the spectra .
this suggests that systematic uncertainties in our background subtraction is not a major issue at low energies either .
a radial profile of some quantity may not always provide a useful description for an individual system due to departures from spherical symmetry .
the average of many such profiles derived for different groups should be much less affected though . since our radial measurements for individual groups are statistically independent ( as they are based on projected profiles ) , it is possible to explore general trends in the radial distributions of @xmath29 and @xmath63 by superposing and averaging the profiles derived for individual groups . with the robustness of our results well established
, we can therefore proceed to discuss these in a statistical framework on the basis of correlation tests and orthogonal regression analyses @xcite performed on the combined measurements . throughout the section
, we have used the kendall rank - order correlation coefficient @xmath122 ( in the range @xmath123 $ ] , with @xmath124 for a strong , positive correlation ) , and the associated significance @xmath125 of @xmath122 being non - zero , to quantify the strength and nature of any linear relationship between various quantities . in order to examine the overall radial behaviour of @xmath29 for our sample , we plot in fig .
[ fig , tnorm ] all temperature measurements , normalized to the mean temperature @xmath126 of each group . in the plot ,
we have subdivided the data points according to whether they are inside or outside the @xmath29-peak located at @xmath127 , resulting in 134 and 58 measurements , respectively .
for ngc 4125 , which has a monotonically decreasing @xmath64 , we have grouped the data points inside @xmath25 with those at @xmath128 .
as discussed , temperature variations clearly exhibit a very similar overall behaviour .
similar to the case for clusters ( e.g.@xcite ) , and as discussed quantitatively below , there is a clear indication that the intrinsic scatter in @xmath129 is larger in the central regions , inwards of the temperature peak .
this is not surprising , as non - gravitational processes other than radiative cooling could be important in this region , thus helping to break any self - similarity otherwise induced by gravity . at first glance ,
figure [ fig , tnorm ] appears to be suggestive of a bimodality in @xmath64 in the core regions , but this could be largely driven by ngc 5044 , which is better sampled than the other groups and displays lower central values of @xmath130 than the majority . to investigate this in more detail
, we tested for any evidence of bimodality in @xmath130 by extracting the value for each group around a specific overdensity radius inside the cool core , excluding the non - cool core system ngc4125 . for @xmath131 , the smallest radius shown in fig .
[ fig , tnorm ] , a kolmogorov - smirnov test indicates a 65 per cent probability that the resulting values have been drawn from a single gaussian distribution with the given mean and variance . at @xmath132 , however , there is some evidence for bimodality ( see inset in fig .
[ fig , tnorm ] ) , with the corresponding probability being only 12 per cent , but increasing again to 45 and 91 per cent at 0.05 and @xmath133 , respectively .
intriguingly , we conclude that there is some evidence for a bimodal @xmath130 distribution around @xmath134 , but the result is marginal in the sense that a unimodal gaussian distribution can not be rejected at 90 per cent confidence , in part due to the relatively small sample size .
there is also no strong indications of bimodality at significantly smaller or larger radii . in order to investigate the potential role of agn feedback in establishing this tentative bimodality
, we searched for a central radio source in the brightest group galaxy following the procedure outlined in @xcite .
such a source was clearly identified in 11 of the groups ( with the off - centre source in ngc4125 as an additional candidate ) , and with hcg42 , mkw4 , and ngc4325 as the only clear exceptions .
for these 11 groups , testing for a linear correlation between @xmath130 at @xmath134 and the 1.4 ghz power of this central radio source yields a value of kendall s @xmath135 and an associated significance of @xmath136 .
thus , there is marginal statistical evidence for a correlation , but this is largely driven by the two most powerful sources ( in ngc383 and ngc741 ) , without which any correlation is absent ( @xmath137 ) . in conclusion , there appears to be no clear connection between this potential @xmath64 bimodality and the current radio output of a central agn .
a larger sample would be useful for a more decisive test of the presence of a bimodality , and for addressing its possible origin .
it is also worth noting that we are probing the cores of these groups down to much smaller fractions of @xmath2 than in typical cluster studies . interestingly , excluding the non cool core system ngc 4125 , the ratio @xmath138 of the temperature @xmath139 in the innermost radial bin to the mean temperature @xmath23 shows a mean and standard deviation of 0.58 and 0.14 , respectively ( 0.61 and 0.17 if ngc 4125 is included ) . this is higher than the typical cool - core cluster value of @xmath140 found , for example , for the projected @xmath29-profiles of @xcite or the deprojected ones of @xcite and @xcite .
it may reflect the fact that the temperature profiles of our groups flatten off at small radii ( inside @xmath141 ) , as suggested in section [ sec , temp ] , in contrast to what is seen in clusters .
[ fig , projct ] indicates that deprojecting our profiles would not change this conclusion , as already suggested by the internal consistency among projected and deprojected cluster results .
an orthogonal regression fit to the temperature profiles inside the cool core excluding the innermost @xmath142 because of the putative flattening in this region
yields @xmath143 the derived logarithmic slope of the profile in the core of @xmath144 is in excellent agreement with the value of @xmath145 seen for groups in the radio - quiet ( 1.4 ghz radio power of central early - type @xmath146 w hz@xmath13 ) subsample of @xcite .
it is , however , somewhat flatter than the slope of @xmath147 seen for their radio - loud subsample , or that of @xmath148 found for clusters by @xcite .
hence , not only do the temperature profiles appear to flatten off in the very core of these groups , their rise with radius is also somewhat shallower than in clusters .
we investigate these results from a slightly different perspective in figure [ fig , tctmax ] , plotting the ratio of @xmath139 to the peak temperature @xmath149 for each group .
while this ratio is subject to considerable scatter at a given mean system temperature @xmath23 , there is a weak indication ( @xmath150 , @xmath151 ) of a negative correlation between the quantities plotted ( the low - lying outlier in fig .
[ fig , tctmax ] is ngc 2300 ; if one were to exclude this data point , the significance of the correlation would increase to @xmath152 ) , thus confirming the notion that the cool core is less pronounced in groups than in clusters .
the implication seems to be that , relative to the maximum temperature of the system , cooling in clusters can generally progress further than in groups , and this group cluster discrepancy becomes increasingly pronounced towards lower mean system temperatures @xmath126 .
a potentially related issue is the radial extent of the cool cores . as mentioned in section [ sec , temp ] , the position @xmath59 of the temperature peak varies slightly among the groups . in order to investigate whether this might reflect the properties of the central galaxy or the group itself , we plot in fig .
[ fig , tmax ] @xmath127 as a function of the size and mass ( measured by @xmath25 and @xmath153 , respectively ) of the central galaxy and of the mass of the group ( measured by @xmath2 ) , with @xmath61-band luminosities @xmath153 taken from 2mass @xcite . testing for a linear correlation between @xmath59 and the three quantities plotted , we find @xmath154 and @xmath155 , for @xmath59 vs.@xmath2 .
the corresponding significances for a correlation with @xmath153 and @xmath25 are 2.8 and @xmath156 , respectively .
although not a strong result , it suggests that the central temperature structure of the group is more closely related to the properties of the group as a whole than to those of the central galaxy ( any correlation with central galaxy properties might naturally arise because @xmath25 and @xmath153 are themselves correlated with @xmath2 at 2.5 and @xmath157 significance , respectively ) . such an interpretation would be in qualitative agreement with the conclusions of @xcite and @xcite who found that the x - ray properties of gas surrounding central group galaxies are different from those of non - central galaxies and , indeed , appear more closely related to the group than to the central galaxy itself . even when excluding ngc 4125 , by far the coolest group in our sample , a fit to the data in fig .
[ fig , tmax]c gives the relation @xmath158 which , perhaps coincidentally , would predict the observed absence of a cool core in ngc 4125 .
if this relation can indeed be extended to @xmath159 kpc , the implication would be that systems with @xmath160 kpc , or , more precisely , with @xmath161 kev from equation ( [ eq , r500 ] ) , would generally not have cool cores .
this is perhaps not surprising , as the central gas properties in systems of such low temperatures could be strongly affected by galactic feedback .
measured relative to @xmath2 , fig .
[ fig , tmax]c also shows that while the cool core is generally smaller than in clusters , it approaches the typical cluster extent at the high end of the temperature range covered by our sample , with group and cluster results becoming statistically indistinguishable at @xmath162 kev according to equations ( [ eq , r500 ] ) and ( [ eq , cc ] ) .
outside the cool core , the temperature profiles in fig .
[ fig , tnorm ] clearly decline with @xmath73 , appearing to drop by almost a factor of two out to @xmath2 . a correlation test on the quantities plotted in fig .
[ fig , tnorm ] establishes a strong anticorrelation beyond @xmath163 , with @xmath164 and @xmath165 .
for the data in this region , we attempted three different parametrizations for this decline of the @xmath29-profile . from a _
analysis of the projected temperature profiles of a sample of 13 massive , relaxed clusters ( @xmath29-range 19 kev ) , @xcite find a relation @xmath166 outside the cool core ( assuming @xmath167 ) .
a similar linear fit for our groups yields @xmath168 for a reduced @xmath169 . taken at face value ,
this relation , shown as a dashed line in fig .
[ fig , tnorm ] , is in excellent agreement with that of the vikhlinin et al .
clusters , suggesting that temperature variations within groups are very similar to those of clusters at large radii , contrary to the case in the core .
the resulting fit quality is poor , however , so a relation of the form of equation ( [ eq , t1 ] ) is not particularly successful in describing our group data .
the main reason is that such a parametrization can not match the group data immediately outside the cool core without overpredicting observed temperatures at intermediate radii and underpredicting them further out . as a possible alternative
, we note that cosmological hydrodynamical simulations invoking gravity only have suggested a ` universal ' temperature profile for massive ( @xmath170 kev ) clusters , of the form @xmath171 @xcite , assuming @xmath172 .
this parametrization was shown by @xcite to be in good agreement with cluster results outside the core region , remaining a good description also when including galactic feedback and radiative cooling in the simulations .
motivated by this , we fitted a similar relation to our group data , yielding @xmath173 for @xmath174 . while resulting in a better fit than equation ( [ eq , t1 ] ) , this relation suffers from the same shortcomings in describing our data , albeit to a lesser degree .
nominally , the result , represented by a dash - dotted line in fig .
[ fig , tnorm ] , is broadly consistent with that found for the simulated clusters of @xcite , including the normalization factor of 1.33 found by the latter authors . prompted by the appearance of the data in fig . [
fig , tnorm ] , a third parametrization of the profiles was also investigated , in which @xmath29 is a linear function of @xmath175 .
the fit result , @xmath176 is statistically better ( @xmath177 ) than the results of equation ( [ eq , t1 ] ) and ( [ eq , t2 ] ) . the intrinsic scatter about this relation is negligible , whereas the corresponding scatter around equation ( [ eq , tcc ] ) is 12 per cent , thus confirming the indication of fig .
[ fig , tnorm ] that intrinsic scatter is larger in the core regions .
the derived relation of equation ( [ eq , t ] ) is plotted as a solid line in fig .
[ fig , tnorm ] .
the fact that equations ( [ eq , t1 ] ) and ( [ eq , t2 ] ) are good descriptions of both observed and simulated clusters , but are less representative of groups than equation ( [ eq , t ] ) , does not necessarily suggest the presence of intrinsic differences in the large - scale temperature structure of hot gas in these systems . the formal agreement with cluster results in terms of best - fitting parameters clearly
indicate that the temperature profiles of cool - core groups at large radii are similar to those of clusters .
we do not expect this conclusion to be affected by projection effects to an appreciable degree .
finally , we note that there is no indication that the possible bimodality in @xmath130 at @xmath178 is reflected in significantly different slopes at large @xmath73 ; assuming again a prescription of the form of equation ( [ eq , t ] ) outside @xmath59 , we find consistent slopes of @xmath179 and @xmath180 for the low- and high-@xmath130 subsamples seen in the inset of fig .
[ fig , tnorm ] , respectively .
summarizing , we find that cool cores in groups are less pronounced than those of clusters , in terms of both their spatial extent and relative temperature variation . at large radii
, however , the temperature profiles of our groups decline with radius in consistency with results for more massive systems . in fig .
[ fig , z_all ] , all measurements of @xmath31 and @xmath32 have been combined , demonstrating more clearly the central excess in iron abundance and the general decrease to a value of @xmath181 z@xmath1 at large radii suggested by fig .
[ fig , abund ] . for @xmath31 , there appears to be a clear trend with radius outside the cool core , and a test performed on the data points outside @xmath59 confirms the presence of a linear correlation between log @xmath31 and log @xmath182 at the @xmath183 level ( @xmath184 ) . inside the core ,
the data are consistent with a constant fe abundance ( @xmath185 ) , but this may mask a more complex behaviour as discussed below .
the si profiles are fairly similar to those of fe in the group cores , showing no systematic trend with radius ( @xmath186 ) .
si also appears from fig .
[ fig , z_all ] to show less systematic radial variation outside the core than fe .
indeed , outside @xmath127 no significant correlation ( @xmath187 , @xmath188 ) is found between log @xmath33 and log @xmath182 , though the value of @xmath122 suggests a mild overall decline with radius .
the situation is qualitatively similar to that seen for clusters , where si generally shows a more uniform distribution than fe @xcite .
as is evident from fig .
[ fig , abund ] , @xmath33 actually rises with radius in the outskirts for some of the groups , a fact which clearly contributes to obscuring any statistical evidence for a simple linear behaviour .
a similar feature has been noted in some relatively cool clusters but is seemingly absent in hotter systems @xcite .
in addition to the considerable statistical errors on the derived abundance data , particularly for si at large radii , there is clearly also substantial intrinsic scatter among the groups at any given radius . coupled with the lack of a clear radial correlation for si
, we have consequently refrained from fitting regression lines to the raw data in fig .
[ fig , z_all]a . in an attempt to suppress these intrinsic variations , we also plot in fig .
[ fig , z_all]b the abundance data normalized to the mean value @xmath30 .
as suggested by the figure , the resulting correlation is slightly stronger for @xmath31 ( @xmath189 , @xmath190 ) but is still not very significant for si ( @xmath191 ) .
the latter result is perhaps not surprising , given that the normalization factor , i.e.the mean abundance @xmath30 , has been derived assuming all abundances fixed relative to solar , so the fitted value of @xmath192 will be driven by the abundance as estimated from the prominent fe lines . the strong radial decline of @xmath31 and the hint of a more gentle decline of @xmath33 are also reflected in an increase in @xmath9 at large radii , as illustrated in fig .
[ fig , ratio_all ] .
this figure further suggests the presence of a dichotomy in the radial distribution of @xmath9 , with this quantity being roughly constant inside @xmath59 but rising further out .
a correlation test confirms that within the cool core there is only marginal evidence for a systematic variation in this quantity ( @xmath193 , @xmath194 ) , suggesting a mild radial decrease , with the data showing a mean and @xmath195 dispersion pf @xmath196 z@xmath197/z@xmath198 .
this mean value is thus slightly lower than , but consistent with , the local ( solar neighbourhood ) imf and sn mixture .
for the adopted sn yields , the value corresponds to a mixture of 70 per cent core - collapse sn and 30 per cent sn ia by number , in excellent agreement with recent results for the cores of clusters ( see @xcite and references therein ) , and consistent with the observed proportion of sn types at low redshift ( @xmath199 ) determined from the _ hubble _ and _ chandra _ deep fields @xcite . for the adopted sn yields ,
this implies that sn ia are the dominant fe contributor in group cores , being responsible for @xmath200 per cent of the fe in this region .
this agrees not only with results for cluster cores , as mentioned , but is also in accord with the corresponding fraction for the hot gas in elliptical galaxies of @xmath201 per cent @xcite .
outside @xmath59 , the correlation test confirms that @xmath37 rises with radius ( @xmath202 , @xmath203 ) , but the trend is probably suppressed due to large statistical scatter .
we note that many of the data points with very low @xmath204 have @xmath205 rather than high @xmath31 , but these have large fractional errors and most are consistent with the general trend .
similar to the case for clusters and as further illustrated in fig .
[ fig , ratio_all]b , the high central fe abundance seen in most of our systems can thus be ascribed predominantly to sn ia , with an increasing sn ii contribution towards larger radii and lower fe abundances .
we also note that the global mean and standard deviation of @xmath9 , derived from those data points for which the fractional errors ( at 90 per cent confidence ) on both @xmath33 and @xmath31 are less than 100 per cent , are @xmath206 and 0.8 in solar units , respectively , in agreement with the ( radially constant ) ratio of @xmath207 found for the @xmath208 kev subsample of @xcite .
the fact that our groups do _ not _ show a radially constant ratio is reflected in a large standard deviation around the global mean . ,
all measurements , in units of the solar ratio and as a function of ( a ) radius and ( b ) @xmath31 ( with error bars on @xmath31 omitted for clarity ) .
symbols are as in fig .
[ fig , tnorm ] . dashed lines mark the ratios expected from pure sn ia and sn ii enrichment , as in fig .
[ fig , ratio].,width=340 ] , all measurements , in units of the solar ratio and as a function of ( a ) radius and ( b ) @xmath31 ( with error bars on @xmath31 omitted for clarity ) .
symbols are as in fig .
[ fig , tnorm ] . dashed lines mark the ratios expected from pure sn ia and sn ii enrichment , as in fig .
[ fig , ratio].,width=340 ] to better investigate radial trends and further suppress intrinsic scatter among the groups , we plot profiles of all data points in fig .
[ fig , norm_bin ] , accumulated in radial bins of 20 measurements . in this representation , the temperature profiles are clearly seen in fig .
[ fig , norm_bin]a to peak at a radius @xmath209 , while flattening out in the group cores .
we re - iterate that this central flattening is not observed in clusters , even in cases where the cluster cores are well resolved on scales below @xmath210 ( e.g. , @xcite ) . for the fe abundance , fig .
[ fig , norm_bin]b reveals roughly solar abundances in the group cores , consistent with the results found for x - ray bright ellipticals by @xcite , five of which constitute the central group elliptical in our sample .
the figure also confirms the trends discussed above , indicating that outside the core , @xmath5 continuously declines out to the largest measured radii .
the binned results further enable a straightforward comparison to the corresponding fe profile found for cool - core clusters by @xcite .
this reveals similar values of fe enrichment within the central abundance excess in groups and clusters ( @xmath211 z@xmath1 ) , with some indication that the excess is slightly broader in clusters ; our binned fe profile declines from its central value by a factor of two around @xmath212 , whereas the corresponding radius for the @xcite clusters is @xmath213 .
the @xcite clusters display a flattening in the fe profile beyond @xmath77 .
contrary to what is observed for our groups .
note also that the off - centre abundance peaks hinted at in many of the groups with well - resolved cores , and the considerable intrinsic scatter these peaks cause in this representation , are visible for both the binned fe and si profiles , even when these are normalized to the mean abundances .
the binning of the abundance profiles clearly suppresses the considerable scatter seen in fig .
[ fig , z_all ] , resulting in a much more well - defined radial behaviour for both fe and si . we performed linear regression fits to these sample - averaged profiles in order to characterize the reasonably uniform behaviour outside the very group cores . excluding the innermost data point in fig .
[ fig , norm_bin]b , orthogonal regression on the binned fe profile yields @xmath214 relative to the corresponding solar abundances .
the fe abundance at @xmath2 suggested by this relation of @xmath215 z@xmath1 is lower than the typical value of @xmath216 z@xmath1 seen in the outskirts ( at @xmath217 ) of clusters ( e.g. , @xcite ; @xcite ; @xcite ) .
this is particularly true in light of the fact that we have adopted the abundance table of @xcite , for which the solar fe abundance is only 68 per cent of the @xcite value typically used in previous cluster studies .
the fe abundance in those previous studies must therefore be multiplied by 1.48 for a direct comparison to our results ( the si abundance , on the other hand , remains unaffected to within 1 per cent by this change in abundance table ) .
fig.[fig , norm_bin]c suggests that @xmath33 also declines outside the cool core , despite the fact that there is only marginal statistical evidence for a radial decline of @xmath33 outside @xmath59 in the unbinned data . as already indicated by the dichotomy in the si / fe ratio ( fig .
[ fig , ratio_all ] ) , the reason for this apparent inconsistency is that while si does decline immediately outside the cool core , it flattens out or even rises at the largest radii probed , thus suppressing any statistical evidence for a uniformly declining profile outside the core . beyond @xmath218 , roughly corresponding to the range covered by the outermost data point in fig .
[ fig , norm_bin]c , the unbinned si data show no clear correlation with @xmath73 ( @xmath219 ) . to capture this behaviour
, we performed a regression fit to all but the innermost data point in fig.[fig , norm_bin]c , and assumed @xmath220 to remain constant beyond the outermost data point at @xmath221 , yielding @xmath222 with @xmath223 and @xmath224 .
the value of 0.25 z@xmath1 at @xmath225 was chosen to ensure continuity , with its errors based on the fractional errors on the outermost data point .
an si abundance of @xmath226 0.20.25 z@xmath1 at @xmath2 is also in good agreement with earlier _ asca _ results for groups @xcite , whereas clusters tend to show slightly higher values of @xmath680.5 z@xmath1 @xcite at large radii .
this suggests a higher content of _ both _ fe and si in the outskirts of more massive systems . however , a direct comparison is somewhat impeded by the large statistical uncertainties on @xmath32 in groups , as well as the fact that @xmath220 does not show a uniform behaviour at large @xmath73 across our sample . for the @xmath9 ratio ,
a subsolar value is seen in the innermost bin in fig .
[ fig , norm_bin]f , with the ratio rising to @xmath227 within @xmath74 and further to @xmath228 well outside the central galaxy . as discussed in section [ sec , syst ] , we believe that the result for the outermost bin is not seriously affected by uncertainties in our background subtraction .
the result demonstrates more clearly what was already indicated by fig .
[ fig , ratio_all ] , that sn ia provide a dominant , but not exclusive , contribution to the enrichment within the cool core and hence inside the optical extent of the central galaxy .
well outside the group core the dominant contribution comes from sn ii , and at large radii the @xmath37 ratio rises to become consistent with the value of @xmath229 expected for pure sn ii enrichment . corrected for differences in adopted abundance tables , the value of the outermost bin in fig .
[ fig , norm_bin]f is at least as high those of the @xcite clusters .
our results , therefore , do not support the claim of the latter authors that groups on average have lower si / fe ratios at large radii than clusters . the results further indicate that , on average , both sn ia and sn ii enrichment is required at all radii within at least @xmath230 .
while this relies on the assumption that derived abundances in the central region are not seriously affected by the fe and si biases , similar central @xmath37 ratios of @xmath231 have also been found in _
asca _ and _ xmm _ observations of hotter clusters @xcite , although it should be mentioned that the cores of these clusters are generally less well resolved than those of our groups .
summarizing , we find that the distributions of both fe and si in the unbinned data ( fig . [ fig , z_all ] )
appear roughly uniform in the group cores , but the averaged profiles clearly confirm the evidence for off - centre peaks in some of the groups .
the enrichment in the cores contain a significant contribution from sn ia , probably hosted by the central , bright galaxy , an issue to which we will return in paper ii . on average , sn ii
dominate by _ number _ at all radii , in agreement with the results obtained by @xcite for a sample of three groups , and can on their own explain the observed abundances well outside the group core .
we believe this latter conclusion to be reasonably robust with respect to any remaining background issues .
a high si / fe ratio consistent with , or even exceeding , the expectation from our adopted sn ii yields is seen in the outermost bin of _ all _ of our groups , including groups such as ngc 383 , hcg 62 , and ngc 5044 , for which the data feature high signal - to - noise ratios even at large radii .
support for this result comes from the work of @xcite , in which a rise in @xmath9 to supersolar values at large radii is also seen for the groups in their sample .
the trend of si / fe increasing with radius has also been reproduced in cosmological simulations in which the galactic feedback implementation favours the ejection of sn ii products over those of sn ia @xcite .
we note that the interpretation that sn ii alone can explain the observed abundance pattern at large radii is reasonably robust to the choice of theoretical sn ii yields available in the literature . for example , for a standard salpeter imf , none of the sn ii model yields considered by @xcite would predict @xmath232 .
hence our conclusion regarding the sn ii contribution is only compromised if invoking a more ` top - heavy ' imf , with @xmath233 in equation ( [ eq , sn ] ) . in connection with this , it is worth pointing out , on the basis of the outermost three data points of fig .
[ fig , norm_bin]f , that the slope of @xmath234 for at least one of the two elements must change beyond @xmath2 if the observed si / fe ratio is to remain consistent with the imf and sn model yields adopted here .
deep , high s / n x - ray observations of group outskirts could shed light on this issue and thus help constrain theoretical sn yields , but such a study could well be beyond the capabilities of current x - ray telescopes for all reasonable exposure times . with their higher x - ray surface brightness at large radii , cool to intermediate - temperature clusters ( @xmath235 kev ) remain more attractive targets for this purpose , also because they show prominent emission lines from a wider range of relevant elements ( e.g. , @xcite ) .
we have conducted a homogeneous analysis of a sample of 15 x - ray bright galaxy groups observed by _
chandra _ , selected for their relatively undisturbed x - ray morphology and good photon statistics .
correlation tests and regression fits have been performed on the combined radial temperature and abundance profiles and derived quantities , both to investigate the nature of any relationship between various quantities , and to enable straightforward comparison to results of numerical simulations of group formation and evolution .
our derived radial temperature profiles show remarkable similarity when scaled to the mean system temperature @xmath23 derived outside the central region , and reveal that all but one group show clear evidence for a cool core .
we find that these cores are smaller than those seen in clusters , both in absolute physical extent and relative to a fixed overdensity radius , and that their sizes appear more closely related to the properties of the group as a whole than to those of the central early - type galaxy present in all our groups .
the absolute and relative cool core sizes approach those of clusters at the high temperature end of our sample .
the temperature profiles in the core are also shallower than those seen in clusters , equivalent to a less pronounced temperature decline in the core relative to the peak temperature displayed by the system .
this may indicate that radiative cooling can generally progress further in more massive systems .
the flatter group profiles inwards of the temperature peak relative to those of more massive systems certainly point to clear differences in the central icm heating history among systems of varying mass .
we find tentative evidence for a bimodality in @xmath130 inside the cool core , but this is confined to a narrow interval in radius and is not mirrored by differing slopes in the temperature profiles at large radii , which we find to be similar to those found for both observed and simulated clusters .
the fe and si abundance profiles show a number of distinct features common to most of the groups .
similar to the case for cool - core clusters , a central enhancement in fe abundance is clearly seen in all but two groups , one of which is ngc 4125 , the only non cool core group in our sample .
this fe excess can extend well beyond the optical extent of the central early - type galaxy present in all groups , and is in many cases accompanied by a similar feature in the si distribution .
in many of the groups with a well - resolved core , evidence for an off - centre abundance peak is also observed inside the central galaxy , with a decline in abundance at the very centre , similar to results for some well - resolved cluster cores . outside the central excess
, the fe profiles generally show a steep decline , approaching a value of @xmath236 solar at @xmath2 .
this is lower than the corresponding typical cluster value by a factor of two , suggesting fundamental differences in enrichment history for the outskirts of these low - mass systems compared to more massive clusters .
si shows similar overall features but declines less steeply with radius , and flattens out or even rises again at large radii for the majority of the groups . on average ,
si reaches a value of @xmath237 solar at @xmath2 , again slightly below typical cluster values . for the adopted sn yields , the ratio of si to fe abundance in the groups reveals a clear predominance of sn ia enrichment in the central regions , with @xmath200 per cent of the fe originating in sn ia
this is in line with results for both hot gas in elliptical galaxies and the cores of massive clusters , suggesting a very similar enrichment istory for the central regions in such systems across three orders of magnitude in total system mass .
the sn ii contribution in our groups increases with radius , however , and at the largest radii probed , or , equivalently , at the lowest fe abundances , the si / fe ratio is close to , and nearly always consistent with , the expectation for pure sn ii enrichment in all groups .
we find no indication that si / fe ratios at large radii are significantly different from the corresponding values in clusters .
we note , however , that the existence of two separate sn ia populations , as suggested by @xcite , could complicate the interpretation of these results .
the abundance distribution in the intragroup medium is expected to be tightly linked to the star formation history in groups . in a forthcoming companion paper
, we decompose the derived abundance profiles into contributions from the two major supernova types to explore this issue , along with the implications of our results for supernova and icm energetics , the role of agn and mergers in redistributing and reheating the enriched gas , and more generally for the enrichment history of the groups .
we thank stephen helsdon for his contribution to the early stages of this project .
we are grateful to alexis finoguenov and john osmond for useful discussions and for providing comparison data for their group samples prior to publication .
this work has made use of the nasa / ipac extragalactic database ( ned ) and the two micron all sky survey ( 2mass ) database .
jr acknowledges the support of the european community through a marie curie intra - european fellowship .
anders e. , grevesse n. , 1989 , geochim .
cosmochim acta , 53 , 197 arnaud m. , rothenflug r. , boulade o. , vigroux l. , vangioni - flam e. , 1992 , a&a , 254 , 49 balestra i. , tozzi p. , ettori s. , rosati p. , borgani s. , mainieri v. , norman c. , viola m. , 2007 , a&a 462 , 429 baumgartner w.h . , loewenstein m. , horner d.j . , mushotzky r.f . , 2005 , apj , 620 , 680 bhringer h. , matsushita k. , churazov e. , finoguenov a. , ikebe y. , 2004 , a&a , 416 , l21 buote d.a . , 2000a , mnras , 311 , 176 buote d.a . , 2000b , apj , 539 , 172 buote d.a . , lewis a.d
, brighenti f. , mathews w.g . , 2003a , apj , 594 , 741 buote d.a . , lewis a.d .
, brighenti f. , mathews w.g . , 2003b , apj , 595 , 151 buote d.a .
, brighenti f. , mathews w.g . , 2004 ,
apjl , 607 , l91 condon j.j . ,
cotton w.d . ,
greisen e.w . ,
yin q.f . , perley r.a . , taylor g.b . , broderick j.j . , 1998 , aj , 115 , 1693 cora s.a . , 2006 , mnras , 368 , 1540 croston j.h . , hardcastle m.j
, birkinshaw m. , 2005 , mnras , 357 , 279 dahlen t. , et al . , 2004 , apj , 613 , 189 davis d.s . , mulchaey j.s .
, mushotzky r.f . , burstein d. , 1996 , apj , 460 , 601 davis d.s . , mulchaey j.s .
, mushotzky r.f . , 1999 ,
apj , 511 , 34 de grandi s. , molendi s. , 2001 , apj , 551 , 153 de grandi s. , ettori s. , longhetti m. , molendi s. , 2004 , a&a , 419 , 7 de plaa j. , werner n. , bleeker j.a.m . , vink j. , kaastra j.s . , mendez m. , 2007 , a&a , 465 , 345 dickey j. m. , lockman f. j. , 1990 , ara&a , 28 , 215 domainko w. , et al . , 2006 , a&a , 452 , 795 donahue m. , horner d.j .
, cavagnolo k.w . ,
voit g.m . , 2006 ,
apj , 643 , 730 edwards l.o.v . ,
hudson , m.j . ,
balogh m.l . ,
smith r.j . , 2007 , mnras , in press ( arxiv:0704.3242
) ettori s. , fabian a.c .
, 2006 , mnras , 369 , l42 ettori s. , fabian a.c .
, allen s.w . ,
johnstone r.m . , 2002 , mnras , 331 , 635 fabbiano g. , schweizer f. , 1995 , apj , 447 , 572 finoguenov a. , ponman t.j . , 1999 , mnras , 305 , 325 finoguenov a. , david l.p . ,
ponman t.j . ,
2000 , apj , 544 , 188 finoguenov a. , arnaud m. , david l.p . , 2001a , apj , 555 , 191 finoguenov a. , reiprich t.h .
, bhringer h. , 2001b , a&a , 368 , 749 finoguenov a. , jones c. , bhringer h. , ponman t.j . , 2002 , apj , 578 , 74 finoguenov a. , burkert a. , bhringer h. , 2003 , apj , 594 , 136 finoguenov a. , davis d.s . , zimer m. , mulchaey j.s . , 2006 ,
apj , 646 , 143 ( f06 ) finoguenov a. , ponman t.j , osmond j.p.f . ,
zimer m. , 2007 , mnras , 374 , 737 ( f07 ) forbes d.a . , thomson r.c . , 1992 , mnras , 254 , 723 fukazawa y. , kawano n. , kawashima k. , 2004 , apjl , 606 , l109 fukazawa y. , makishima k. , tamura t. , ezawa h. , xu h. , ikebe y. , kikuchi k. , ohashi t. , 1998 , pasj , 50 , 187 fukazawa y. , et al . , 1996 , pasj , 48 , 395 gastaldello f. , buote d. , humphrey p. , zappacosta l. , bullock j. , brighenti f. , mathews w. , 2005 , in proceedings of the the x - ray universe 2005 , a. wilson , ed . , esa sp-604 , p. 743
gibson b.k . , loewenstein m. , mushotzky r.f . , 1997 , mnras , 290 , 623 gilfanov m.r .
, sunyaev r.a . ,
churazov e.m . , 1987 , sval , 13 , 3 grevesse n. , sauval a.j . , 1998 , space sci .
, 85 , 161 helsdon s.f . ,
ponman t.j . , 2003 , mnras , 340 , 485 helsdon s.f . , ponman t.j . , osullivan e. , forbes d.a . , 2001 ,
mnras , 325 , 693 humphrey p.j . ,
buote d.a . , 2006 ,
apj , 639 , 136 hwang u. , mushotzky r.f . , burns j.o .
, fukazawa y. , white r.a .
, 1999 , apj , 516 , 604 irwin j.a .
, bregman j.n . , 2001 , apj , 546 , 150 isobe t. , feigelson e.d .
, akritas m.g . ,
babu g.j . , 1990 , apj , 364 , 104 iwamoto k. , brachwitz f. , nomoto k. , kishimoto n. , umeda h. , hix , w.r .
, thielemann f .- k . , 1999 ,
apjs , 125 , 439 jetha n.n . ,
ponman t.j . ,
hardcastle m.j . ,
croston j.h . , 2007 , mnras , 376 , 193 johnstone r.m . , allen s.w . , fabian a.c . , sanders j.s .
, 2002 , mnras , 336 , 299 jones l.r .
, ponman t.j . ,
horton a. , babul a. , ebeling h. , burke d.j . , 2003 , mnras , 343 , 627 kim d .- w . , fabbiano g. , 2004 , apj , 613 , 933 khosroshahi h.g . , jones l.r . ,
ponman t.j . , 2004 , mnras , 349 , 1240 kraft r.p .
, forman w.r . , churazov e. , laslo n. , jones c. , markevitch m. , murray s.s . , vikhlinin a. , 2004 , apj , 601 , 221 laing r.a . ,
bridle a.h . , 2002 , mnras , 336 , 1161 loken c. , norman m.l . ,
nelson e. , burns j. , bryan g.l .
, motl p. , 2002
, apj , 579 , 571 mcnamara b.r . , nulsen p.e.j . , wise m.w . ,
rafferty d.a . , carilli c. , sarazin c.l .
, blanton e.l . , 2005 ,
nature , 433 , 45 mannucci f. , della valle m. , panagia n. , 2006 , mnras , 370 , 773 mathews w.g . , brighenti f. , buote d.a . ,
lewis a.d . , 2003 ,
apj , 596 , 159 moll r. , et al . , 2007 , a&a , 463 , 513 morita u. , ishisaki y. , yamasaki n.y .
, ota n. , kawano n. , fukazawa y. , ohashi t. , 2006 , pasj , 58 , 719 mulchaey j.s . ,
davis d.s . ,
mushotzky r.f .
, burstein d. , 2003 , apjs , 145 , 39 mushotzky r. , loewenstein m. , arnaud k.a . , tamura t. , fukazawa y. , matsushita k. , kikuchi k. , hatsukade i. , 1996 , apj , 466 , 686 nomoto k. , tominaga n. , umeda h. , kobayashi c. , maeda k. , 2006 , nucl .
a , 777 , 428 osmond j. p. f. , ponman t. j. , 2004 , mnras , 350 , 1511 osmond j.p.f . , ponman t.j .
, finoguenov a. , 2004 , mnras , 355 , 11 osullivan e. , vrtilek j.m . ,
read a.m. , david l.p . , ponman t.j .
, 2003 , mnras , 346 , 525 osullivan e. , vrtilek j.m .
, kempner j.c . , 2005 , apjl , 624 , l77
osullivan e. , vrtilek j.m . ,
harris d.e .
, ponman t.j , 2007 , apj , 658 , 299 piffaretti r. , jetzer p. , kaastra j.s . , tamura t. , 2005 , a&a , 433 , 101 pratt g.w . , bhringer h. , croston j.h . , arnaud m. , borgani s. , finoguenov a. , temple r.f . , 2007 , a&a , 461 , 71 rasmussen j. , ponman t.j .
, mulchaey j.s . , 2006 , mnras , 370 , 453 rebusco p. , churazov e. , bhringer h. , forman w. , 2006 , mnras , 372 , 1840 roediger e. , brggen m. , rebusco p. , bhringer h. , churazov e. , 2007 , mnras , 375 , 15 romeo a.d .
, sommer - larsen j. , portinari l. , antonuccio - delogu v. , 2006 , mnras , 371 , 548 russell p.a .
, ponman t.j . ,
sanderson a.j.r . , 2007 , mnras , in press ( astro - ph/0703010
) sanders j.s . , fabian a.c . , 2002 , mnras , 331 , 273 sanders j.s . , fabian a.c .
, dunn r.j.h . , 2005 , mnras , 360 , 133 sanderson a.j.r . , ponman t.j . , osullivan e. , 2006 , mnras , 372 , 1496 skrutskie m.f . , et al . , 2006 , aj , 131 , 1163 .
tamura t. , kaastra j.s .
, den herder j.w.a .
, bleeker j.a.m .
, peterson j.r . , 2004 , a&a , 420 , 135 tornatore l. , borgani s. , matteucci f. , recchi s. , tozzi p. , 2004 ,
mnras , 349 , l19 tozzi p. , rosati p. , ettori s. , borgani s. , mainieri v. , norman c. , 2003 , apj , 593 , 705 trentham n. , tully r.b . ,
mahdavi a. , 2006 , mnras , 369 , 1375 vikhlinin a. , markevitch m. , murray s.s . ,
jones c. , forman w. , van speybroeck l. , 2005 , apj , 628 , 655 vrtilek j.m .
, david l.p .
, grego l. , jerius d. , jones c. , forman w. , donnelly r.h . ,
ponman t.j .
, 2000 , in durret f. , gerbal d. , eds , constructing the universe with clusters of galaxies , linstitut dastrophysique de paris , paris zhang y .- y .
, bhringer h. , finoguenov a. , ikebe y. , matsushita k. , schuecker p. , guzzo l. , collins c.a . , 2006 , a&a , 456 , 55 | the distribution of metals in groups of galaxies holds important information about the chemical enrichment history of the universe . here
we present radial profiles of temperature and the abundance of iron and silicon of the hot intragroup medium for a sample of 15 nearby groups of galaxies observed by _
chandra _ , selected for their regular x - ray morphology .
all but one group display a cool core , the size of which is found to correlate with the mean temperature of the group derived outside this core . when scaled to this mean temperature ,
the temperature profiles are remarkably similar , being analogous to those of more massive clusters at large radii but significantly flatter inwards of the temperature peak .
the fe abundance generally shows a central excess followed by a radial decline , reaching a typical value of @xmath0 z@xmath1 within @xmath2 , a factor of two lower than corresponding results for clusters .
si shows less systematic radial variation , on average displaying a less pronounced decline than fe and showing evidence for a flattening at large radii .
off - centre abundance peaks are seen both for fe and si in a number of groups with well - resolved cores .
derived abundance ratios indicate that supernovae type ia are responsible for 80 per cent of the fe in the group core , but the type ii contribution increases with radius and completely dominates at @xmath2
. we present fitting formulae for the radial dependence of temperature and abundances , to facilitate comparison to results of numerical simulations of group formation and evolution . in a companion paper
, we discuss the implications of these results for feedback and enrichment in galaxy groups .
[ firstpage ] galaxies : clusters : general ( galaxies : ) intergalactic medium x - rays : galaxies : clusters |
transport coefficients have recently been computed to leading logarithmic order in weak coupling perturbation theory @xcite . however , at temeperatures of interest to heavy - ion collisions experiments , the gauge coupling @xmath0 and this computation is ineffective . at the other extreme of @xmath1 , the ads / cft correspondence has been exploited to bound the shear viscosity @xmath2 @xcite , where @xmath3 is the extropy density .
this result is for a highly supersymmetric version of qcd and needs to be corrected in order to connect with experiments .
there is clearly a case for a lattice computation , if one can have control over the process of analytic continuation from euclidean to minkowski space - time .
direct computations of the shear viscosity were tried @xcite , but were inconclusive because statistical problems beset even the euclidean computations of the correlator of the energy - momentum tensor , thus making it hard to check the analytic continuation . in this work
@xcite i shift the focus slightly and compute the electrical conductivity , @xmath4 , of the quark - gluon plasma .
this involves a lattice computation of the correlator of the electromagnetic current .
since this is no harder than the extraction of the @xmath5-meson mass , statistical errors in the euclidean computation are under control , and effort can be concentrated on the harder job of analytical continuation .
we are able to demonstrate control over this process , and thereby extract @xmath4 .
this is interesting in itself for heavy - ion collisions because several properties of soft photons in media can be inferred once @xmath4 is known .
since this is the first statistically reliable extraction of a transport coefficient from lattice qcd , one can use this to estimate many different quantities which depend on transport , and are important for heavy - ion physics . with @xmath4 in hand , and using two well - specified approximations , we are able to predict the value of @xmath6 , and compare the lattice results with others .
the lattice computations described in section 2 are from @xcite .
the material in section 3 is partly new .
all lattice computations of transport coefficients are performed in the context of linear response theory .
the response , @xmath7 , of a system to a time - dependent force @xmath8 is given , in the frequency domain by @xmath9 .
the response function @xmath10 can then be written in terms of the retarded correlation function of @xmath7 , and the transport coefficient can be extracted using kubo formul . for @xmath4 ,
the kubo formula is ( t ) = 16._i^i(,0,t)|_=0 , [ cond]where the sum is over spatial polarisations of the imaginary part of the retarded photon propagator , _
i.e. _ , the spectral density , @xmath11 , of the electromagnetic current correlator .
the usual qed ward identity gives @xmath12 , where @xmath13 is the usual charge susceptibility .
a lattice computation proceeds from the spectral representation for euclidean current correlators ( t , t ) = _ 0^ k(,t , t ) ( , t ) , [ spec]where @xmath14 is the free propagator , @xmath15 is the product of the zero ( spatial ) momentum vector correlator summed over all spatial polarisations , @xmath16 , and the em vertex factor @xmath17 , where @xmath18 is the charge of a quark of flavour @xmath19 . for later use
we note that @xmath20 for two flavours .
the main problem in this determination is that @xmath21 is needed at an infinite number of points and @xmath15 is known only at @xmath22 different values .
the solution is to constrain the function @xmath21 through an informed guess , and use a bayesian method to extract it .
we note a second complication .
a free field theory computation shows that @xmath23 increases as @xmath24 at large @xmath25 .
this means that a naive application of the kubo formul would violate causality @xcite .
we try to solve this problem by using @xmath26 which is the difference of the measured and free - field theory propagators to extract @xmath27 .
the propagator @xmath28 obtained through the fit of @xmath27 compared to the data for @xmath29 and @xmath30 ( circles ) , 10 ( squares ) and 12 ( pentagons ) with quark mass @xmath31 . ]
the propagator @xmath28 obtained through the fit of @xmath27 compared to the data for @xmath29 and @xmath30 ( circles ) , 10 ( squares ) and 12 ( pentagons ) with quark mass @xmath31 . ] using a variety of bayesian methods @xcite to extract @xmath26 we found a peak at @xmath32 @xcite and a smooth drop to zero at @xmath33 ( see figure [ fg.rho ] ) .
the checks which have been performed on this extraction are 1 .
the statistical errors are negligible , as seen in figures [ fg.rho ] and [ fg.prop ] .
the position of the peak seen in figure [ fg.rho ] is not strongly dependent on the lattice spacing , as checked by using @xmath30 , 10 and 12 .
3 . variations in the results are investigated due to algorithmic changes in the bayesian prior and the truncation and discreitzation of eq .
( [ spec ] ) .
the width of the band of systematic errors shown in figure [ fg.rho ] is thrice the width due to such variation
. it would be useful to have similiar tests of robustness on all uses of bayesian methods including the maximum entropy method ( mem ) .
the goodness of fit to the euclidean data can be judged from figure [ fg.prop ] .
we draw attention to the small error bars on this data when compared to the errors on typical measurements of euclidean correlators of energy momentum tensor .
having established the existence of the peak , further information is extracted using a parametrization of the peak in @xmath27 as the ratio of two polynomials , with that in the numerator being of lower order than the denonimator , so that the integrals used in the kramers - krnig relation converge .
the denominator has to be at least a quartic polynomial in order to smoothly connect to the weak - coupling picture of a pinch singularity giving rise to the peak along real @xmath25 @xcite .
( we note in this connection that earlier attempts to extract transport quantities have worked with the full @xmath21 rather than with @xmath27 .
this error can strongly suppress estimates of the transport coefficients . since @xmath21 increases with @xmath25 , this biases the fitted peak towards larger @xmath25 , thus decreasing the slope at the origin for any fit form with one peak . )
using these fits our estimate of the electrical conductivity is
t = [ result ]
in kinetic theory one can write the drude formula for the electrical conductivity = m , [ transport]where @xmath34 is the entropy density in quarks , @xmath35 their screening mass and @xmath36 their mean free time .
using the lattice result for @xmath4 one gets @xmath37 fm .
while charge is transported through quarks , momentum is transported mainly through gluons . due to colour factors
the mean free time for gluons is about half the mean free time for quarks .
then using a kinetic theory formula for the viscosity along with such an estimate of @xmath38 one gets s0.210.02
[ visco]only lattice errors are shown in this estimate .
errors on the kinetic theory are harder to extract .
the lattice + kinetic theory result is 2.5 times the ads / cft limit .
non - zero electrical conductivity leads to screening of photons with energy below the peak of the spectral function found earlier , , for @xmath39 mev .
this is a direct consequence of maxwell s equations .
this range of screened energies lies far below the excess which may have been seen by the cern na49 experiment .
it seems possible that direct photons at such low - energy could be distinguished from photons coming from decays of pions at rhic - ii .
one method for using an observation of photon dimming to extract the conductivity has been suggested @xcite .
the mean free path , @xmath40 , of these ultra - soft photons is given by the formula = 6 fm .
[ skin]since the fireball at rhic is about 7 fm in size , one could expect dimming by a factor of @xmath41 , , by a factor of 3.2 .
there is a self - consistency argument here involving only the data .
if the viscosity is small , then mean free times of quarks and gluons are small , and the mean free path of photons can only be about 20 ( @xmath42 ) times larger .
however , dimming is exponential in the mean free path .
for example , the mean free path as predicted in ads / cft would lead to dimming by a factor of 18.8 .
thus , the use of photon dimming to extract @xmath4 is an important cross check of values of @xmath6 extracted by other means .
the study of event - to - event fluctuations in limited rapidity bins is expected to give information on the phase diagram of qcd as well as make contact between experiments and basic lattice qcd computations .
however , fluctuations of conserved quantities in a limited region of space are expected to even out by transport processes such as diffusion .
the computation of @xmath4 is relevant not only to the diffusion of charge , but also of baryon number and isospin .
the diffusion constant can be obtained using the kinetic theory formula @xcite d=_qc_s^20.1 fm , [ einstein]where we have taken the square of the speed of sound @xmath43 at @xmath29 , as shown by recent lattice computations @xcite .
if the plasma is highly dissipative , then @xmath44 must be small .
when @xmath44 decreases , the fluctuation are more localized at the same time .
small viscosity implies smaller mean free path , and hence longer persistence of the fluctuation signal . in the approach of @xcite
this would mean that the rapidity width of a fluctuation is @xmath45 for fluctuations created at time @xmath46 fm .
when weak coupling was the only tool to investigate transport properties , it was felt that the qgp would be weakly dissipative .
now with other tools at our disposal , the question seems more open , and preliminary evidence is that dissipation is strong .
the first lattice results on transport @xcite pin down many aspects of transport , as outlined in this section , and lead to the hope that multiple experimental constraints on transport are feasible .
99 arnold p , moore g d and yaffe l g 2000 _ j. h. e. p. _ * 0011 * 001 arnold p , moore g d and yaffe l g 2001 _ j. h. e. p.
_ * 0111 * 057 kovtun p , son d t and starinets a o 2003 _ j. h. e. p. _ * 0310 * 064 karsch f and wyld h w 1987 _ phys .
_ d * 35 * 2518 nakamura a and sakai s 2005 _ phys .
lett . _ * 94 * 072305 gupta s 2004 _ phys
_ b * 597 * 57 hilgevoord j 1960 _ dispersion relations and causal description _
( amsterdam : north - holland ) aarts g and resco j m m ( 2002 ) _ j. h. e. p. _ * 0204 * 053 gupta s 2004 here s looking at you , fireball _ preprint _ hep - ph/0411355 aziz m a and gavin s 2004 _ phys .
_ c * 70 * 034905 gavai r v , gupta s and mukherjee s 2004 the speed of sound and specific heat in the qcd plasma : hydrodynamics , fluctuations and conformal symmetry _
hep - lat/0412036 | i describe the lattice determination of the electrical conductivity of the quark gluon plasma @xcite .
since this is the first extraction of a transport coefficient with a degree of control over errors , i next use this to make estimates of other transport related quantities using simple kinetic theory formul .
the resulting estimates are applied to fluctuations , ultra - soft photon spectra and the viscosity .
dimming of ultra - soft photons is exponential in the mean free path , and hence is a very sensitive probe of transport . . |
the structure of the @xmath6-nucleon scattering amplitude close to threshold ( @xmath12 = 1.486 gev ) is of much interest because it appears dominated by the presence of a baryon resonance slightly above threshold , the n*@xmath13(1535 ) .
the width of the n*(1535 ) is of the order of 150 mev .
its two main decay channels are the n@xmath14 ( 35 - 55 @xmath15 ) and the n@xmath6 ( 30 - 55 @xmath15 ) final states @xcite .
the @xmath6-nucleon scattering length , @xmath16 , characterizes the behaviour of the @xmath6-nucleon interaction at threshold .
it is a complex quantity whose real part is still poorly determined .
we refer to the recent work of green and wycech @xcite for a discussion of the range of values found in the literature ( from @xmath17 0.3 to @xmath171 fm ) .
the main reason for the spread in values is that @xmath16 is obtained indirectly and through model - dependent analyses of pion- and photon - induced @xmath6 production reactions .
the @xmath6-nucleon scattering length is a key quantity to assess the possibility of forming @xmath6-nuclear quasi - bound states @xcite . in this work
, we propose to study the threshold behaviour of the @xmath6-nucleon scattering amplitude through the @xmath18 reaction in the particular kinematics where the invariant mass of the @xmath19 pair is close to the n*(1535 ) mass .
we tune the incident laboratory photon energy in order to reach low momentum transfers , i.e. a sufficiently small @xmath20 . at the @xmath4 threshold ,
@xmath21=3 gev and @xmath20 is 1.2 gev@xmath22 .
we will consider values of @xmath21 ranging from 4 to 5 gev . at @xmath21=4 gev , @xmath20 is 0.38 gev@xmath22 . at 5 gev , it is 0.26 gev@xmath22 .
our main argument is that the @xmath18 process in these kinematics ( @xmath23 1 gev@xmath22 ) is dominated by the @xmath6 t - channel exchange and offers the possibility to test the @xmath6-nucleon scattering amplitude close to threshold .
both @xmath14- and @xmath6-exchanges can contribute .
the dominance of the @xmath6-exchange in the t - channel comes mainly from the property that the @xmath4-meson radiative decay probability to the @xmath24 channel is an order of magnitude larger than to the @xmath25 channel ( despite the larger phase space available for the latter decay ) .
this is ultimately related to the large @xmath26 content of the @xmath6-meson .
the s - wave @xmath27 amplitude is also significantly smaller than the @xmath28 amplitude .
our results on the @xmath18 reaction are based on @xmath27 and @xmath28 scattering amplitudes obtained in the unitary coupled - channel model of ref .
these amplitudes reproduce a large set of pion - nucleon and photon - nucleon scattering data in the energy range @xmath29 gev , in particular the pion- and photon - induced @xmath6-meson production cross sections .
the value obtained for @xmath16 is ( 1.03 + i 0.49 ) fm @xcite , in close agreement with the findings of ref .
@xcite .
our t - channel calculation of the @xmath18 reaction cross section in the n*(1535 ) region is described in section 2 .
we discuss the @xmath6- and @xmath5-exchanges and their interference .
the latter is constructive or destructive depending on the relative sign of the couplings constants @xmath30 and @xmath31 of the corresponding anomalous interaction lagrangians .
we display numerical results for the @xmath18 reaction cross section in section 3 .
we show the expected t - distributions at @xmath21=4 gev and 5 gev and emphasize the role of the double @xmath6-pole term and of the @xmath6-@xmath5 interference in these quantities . a few concluding remarks are given in section 4 .
the t - channel @xmath14- and @xmath6-exchange amplitudes contributing to the @xmath18 process in the n*(1535 ) region are shown in figs . 1 and 2 .
0.4true cm 0.4 true cm 0.4truecm 0.4true cm 0.4 true cm 0.4truecm we calculate the cross section for the @xmath18 reaction , assuming it is driven by the mechanisms displayed in figs . 1 and 2 .
we have three contributions , associated with the @xmath14-exchange , the @xmath6-exchange and their interference respectively .
the t - channel dominance is clearly an hypothesis which can but be confirmed by measurements of differential cross sections .
we note that recent data on the @xmath32 reaction at @xmath21=2.4 gev @xcite show that this process is dominated by t - channel exchanges from threshold onwards .
the exact nature of these exchanges is however not clear @xcite .
it is therefore of interest to study a @xmath4-photoproduction process in which the excitation of the target to the n*@xmath13(1535 ) is expected to favour specifically unnatural parity exchanges , i.e. the @xmath14 and @xmath6 t - channel contributions .
the 4-momenta of the photon , the proton , the @xmath4 , the @xmath6 and the final proton are denoted by @xmath33 , @xmath34 , @xmath35 , @xmath36 and @xmath37 .
the photon , initial proton and final proton polarizations are indicated by the symbols @xmath38 , @xmath39 and @xmath40 .
the total cross section reads @xmath41 the photon-@xmath4-pseudoscalar meson vertices are described by the anomalous interaction lagrangian @xmath42 where @xmath43 and @xmath44 denotes the chiral pseudoscalar mesons of interest , the pion or the @xmath6-meson . using this interaction lagrangian to calculate the @xmath45 and @xmath46 partial widths , we have @xmath47 identifying this expression with the experimental decay widths @xcite @xmath48 @xmath49 we obtain @xmath50 if we want to compute the interference between the @xmath14- and @xmath6-exchanges , we should know the relative sign of the coupling constants @xmath30 and @xmath31 .
it is generally assumed that the @xmath51 transition is dominated by the @xmath52 mixing .
the @xmath52 mixing angle is not easy to determine because it depends on the overlap of two narrow resonances far apart from each other .
it is therefore model - dependent and sensitive to the assumptions used to calculate the @xmath53-meson self - energy far from the @xmath53 pole .
the @xmath54 transition depends on the @xmath6-@xmath55 mixing , a phenomenon closely connected with the u(1 ) anomaly of qcd and still under analysis in different schemes @xcite .
based on recent analyses of the available data , it is most often found that both @xmath30 and @xmath31 have the same sign as @xmath56 , and hence a positive relative sign @xcite .
the opposite conclusion has also been reached @xcite . in view of the uncertainties in the su(3 ) symmetry breaking mechanisms
responsible for the mixing angles , we consider this relative sign as yet undetermined .
we will now establish the dominance of the @xmath6-exchange process in the @xmath18 reaction in the kinematics of interest ( @xmath57 4 - 5 gev and @xmath58 1 gev@xmath22 ) . using the interaction lagrangian ( 2 ) , the squared amplitudes corresponding to the @xmath14-exchange , the @xmath6-exchange and their interference are @xmath59 @xmath60 @xmath61 we work in the photon - proton center of mass reference frame where the total energy of the reaction is denoted by @xmath12 . in that reference frame ,
the photon 3-momentum is -@xmath62 and the 3-momentum of the @xmath63 pair is -@xmath64 .
we define the invariant mass @xmath65 of the final @xmath66 pair by @xmath67 and express the 4-momentum transfer @xmath68 as function of that variable , @xmath69 where @xmath70 is the angle between the initial photon and the produced @xmath4-meson .
we define the notation @xmath71 and @xmath72 . using the above variables , the differential cross section for the @xmath73 reaction with respect to @xmath74 and @xmath75
is given by @xmath76 the total cross section is obtained by integrating over @xmath74 from @xmath77 to @xmath78 and over @xmath75 from @xmath79 to @xmath80 . in all the results presented in the next section , we will restrict the interval of values of @xmath74 from @xmath81 gev@xmath22 to @xmath82 , the expected range of validity of our model . because @xmath77 extends to much lower values ( for @xmath831.54 gev , @xmath84 gev@xmath22 at @xmath855 gev for example )
, we will refrain from displaying integrated cross sections .
we note that eq .
( [ crosssectdiff ] ) is gauge - invariant because of the specific form of the interaction lagrangian ( [ anomalouslintpi ] ) .
it is of interest to study the pole structure of the three contributions to the cross section ( @xmath5-exchange , @xmath6-exchange , interference ) as it determines the shape of the t - distributions .
we have the following decompositions : @xmath86 simple effects can be understood from these expressions irrespectively of the dynamical aspects of the scattering amplitudes and of the strength of the @xmath87 vertex .
it is easy to see from eq .
( 14 ) that the double pion pole term ( proportional to @xmath88 ) dominates the pion - exchange contribution close to @xmath82 , becomes comparable to the single pion pole term ( proportional to @xmath89 ) around @xmath90 gev@xmath22 and smaller than the latter for larger values of @xmath91 . in the case of
the @xmath6-exchange displayed in eq .
( 15 ) , the single @xmath6 pole term is always dominant in the kinematic range under consideration .
close to @xmath82 , the double @xmath6 pole term represents typically 25 - 30@xmath15 of the total @xmath6-exchange contribution .
it is nevertheless instrumental in producing a rather sharp drop in the differential cross section . in the interference term ( 16 ) , the t - dependence is largely given by the single pion pole term at low @xmath91 .
to derive eq .
( 13 ) , we have assumed that the @xmath92 transition form factor is one .
we have no information on that quantity but it influences significantly the outcome of our calculation
. if the form factor is hard , its effect will be rather small .
if it is soft , it could affect substantially the t - dependence of the @xmath73 reaction .
the only argument we see in favour of a hard form factor is that there is no obvious intermediate state to build a form factor in the @xmath6 direction .
the @xmath6-meson couples dominantly to two photons and to three pions . to construct a form factor in the @xmath6-channel
, one would need a significant decay of the @xmath4-meson into a photon and two vector particles or into a photon and three pions .
the only available information is an upper limit of 5 10@xmath93 on the branching ratio of the @xmath4-meson to the @xmath94 channel @xcite . in the absence of more significant data ,
it is not possible to gain a reasonable understanding of the @xmath92 form factor for space - like @xmath6-mesons
. we will therefore set it to one , keeping this assumption in mind .
finally , if our model takes into account the @xmath6-nucleon final state interaction to all orders , it does not treat @xmath4-nucleon rescattering in the outgoing channel .
we do not expect this rescattering to be very important in the kinematics under consideration , i.e. with a large relative momentum between the @xmath4-meson emitted at small angles and the recoiling target products .
in particular , we are not in the threshold regime where cryptoexotic @xmath95 baryons or @xmath96 resonances could enlarge final state interactions @xcite .
we proceed to the calculation of the @xmath97 cross section as outlined in the previous section using the @xmath27 and @xmath28 scattering amplitudes obtained in the model of ref .
these amplitudes are displayed in fig .
1 true cm 0.4 true cm 0.7truecm 0.4true cm 0.4 true cm we notice that the @xmath14n@xmath98n amplitude is about two to three times smaller than the @xmath6n scattering amplitude in the region of the n*(1535 ) resonance . combined with the very unfavourable ratio @xmath991/29 of the anomalous coupling constants
, this effect suggests that the pion contribution to the @xmath97 cross section will be much smaller , typically by two orders of magnitude , than the @xmath6 contribution .
it is interesting to remark that the amplitude displayed in fig .
4 is very similar to the corresponding quantity obtained in ref .
@xcite , suggesting that our calculation does not depend too much on the specific model used to derive the amplitudes .
we emphasize also that the real part of the @xmath6n scattering amplitude is mostly visible in a narrow band of total center of mass energies , i.e. from threshold until 1.5 gev . beyond @xmath12=1.52 ,
the cross section will be dominated by the imaginary part of the amplitude .
consequently we will show results for two values of the invariant mass of the @xmath6n pair , @xmath100=1.49 gev ( very close to threshold ) and @xmath100=1.54 gev ( at the resonance peak ) .
we consider first @xmath100=1.49 gev and display @xmath101 at @xmath85 4 gev and @xmath85 5 gev .
this is shown in figs . 5 and 6 respectively .
0.4 true cm 0.4 true cm as discussed earlier , we do not know the relative sign of the @xmath102 and @xmath103 coupling constants .
we have therefore considered both signs .
if the coupling constants are of the same sign , the @xmath104 interference is destructive because of the opposite sign of the amplitudes ( solid curve ) .
if @xmath102 and @xmath103 are of opposite sign , the @xmath104 interference is constructive ( dashed curve ) . in order to evaluate the importance of this interference , we show in addition the @xmath6-exchange contribution ( dot - dashed line ) and the @xmath5-exchange contribution ( dotted line ) .
close to @xmath82 , the @xmath104 interference suppresses or enhances the differential cross section by about 30@xmath15 .
as anticipated , the @xmath5-exchange contribution is completely negligible .
we plot also the double pole contribution to the @xmath6-exchange ( dot - double - dashed line ) .
it represents typically a quarter of the @xmath6-exchange term but contributes very significantly to the t - dependence of the differential cross section .
if the @xmath104 interference is constructive , the pole structure of the @xmath6-exchange and of the interference leads to a rather sharp t - dependence close to @xmath82 .
this effect increases with increasing laboratory photon energy as a lower @xmath20 can be reached .
if the @xmath104 interference is destructive , the terms driving the increase of the differential cross section at low @xmath105 cancel significantly , leading to a rather flat behaviour .
we show in figs . 7 and 8 the t - distributions for the total @xmath6n center of mass energy @xmath100 = 1.54 gev , taken to be close to the n*(1535 ) mass
the comparison to the results obtained at @xmath100 = 1.49 gev for the same value of the incident photon laboratory energy shows that the cross section increases .
this is a consequence of the opening of the @xmath6n phase space and of the presence of the n*(1535 ) resonance ( implying a large imaginary part in the amplitude ) .
otherwise the features of the differential cross section @xmath101 are very similar for both values of @xmath100 .
we emphazise however that the differential cross section @xmath101 at @xmath100 = 1.49 gev and at @xmath100 = 1.54 gev tests different parts of the @xmath6n scattering amplitudes , the real and imaginary parts at threshold on the one hand and the imaginary part at the resonance peak on the other hand .
we remark also that the @xmath106 scattering amplitude is very much constrained by data on the @xmath107 cross section close to threshold @xcite . if our model is correct , accurate data on the @xmath101 reaction at low @xmath105 could therefore be interpreted in terms of the @xmath108 scattering amplitude .
we have underlined two uncertainties in such an analysis , the relative sign of the @xmath102 and @xmath103 coupling constants and a possible effect due to a @xmath92 transition form factor for space - like @xmath6 s .
we have studied the @xmath97 reaction with the idea of using future accurate data on this process to gain understanding of the threshold behaviour of the @xmath6-nucleon scattering amplitude .
we chose kinematic conditions where a t - channel meson - exchange description is expected to be valid ( i.e. low momentum transfers ) .
we showed that the process in which the initial photon dissociates into a @xmath4-meson and a virtual @xmath6-meson scattering from the proton target to produce an @xmath6 proton final state is dominant for ( @xmath109p ) pairs with invariant masses close to the n*(1535 ) mass .
it can however interfere with the analogous process where a @xmath5-meson is exchanged .
the @xmath5-exchange term is negligible .
the sign of the @xmath104 interference is at present not known .
we show how it changes both the t - dependence and the absolute value of the differential cross section . in view of the fact that the @xmath106 scattering amplitude in the n*(1535 ) resonance region is rather well - known
, accurate data on the @xmath97 reaction at @xmath85 4 - 5 gev and low t would put additional constraints on the still poorly controlled @xmath6-proton scattering amplitude close to threshold .
the expected cross sections are small but do not appear out of reach of present experimental facilities .
one of us ( m.s ) is indebted to slawomir wycech for a very useful discussion and to the hospitality of the gsi theory group , where part of this work was done .
we are thankful to our referees for comments which led us to significantly improve the contents of this paper .
we acknowledge the support of the european community - research infrastructure activity under the fp6 `` structuring the european research area '' programme ( hadron physics , contract number rii3-ct-2004 - 506078 ) . 99 s. eidelman et al .
, phys . lett .
b 592 ( 2004 ) 1 . a.m. green , s. wycech , phys .
c 71 ( 2005 ) 014001 .
j. niskanen et al . , int . j. mod
a 20 ( 2005 ) 634 and references therein .
lutz , gy .
wolf , b. friman , nucl .
phys . a 706 ( 2002 ) 431 , erratum - ibid a 765 ( 2006 ) 495 .
j. barth et al .
j. a. 17 ( 2003 ) 269 .
t. mibe et al . , phys .
95 ( 2005 ) 182001 .
r. escribano , j .-
frre , jhep06 ( 2005 ) 029 .
a. bramon , a. grau , g. pancheri , phys .
b 344 ( 1995 ) 240 . a.i .
titov et al . , phys .
c 60 ( 1999 ) 035205 .
m. benayoun , h.b .
oconnell , eur . phys . j. c 22 ( 2001 ) 503 .
a. sibirtsev , j. haidenbauer and u .- g .
meissner , nucl - th/0512055 | the @xmath0 reaction is studied in the kinematic region where the @xmath1 final state originates dominantly from the decay of the n*(1535 ) resonance .
the threshold laboratory photon energy for this reaction ( at the peak of the s11 resonance ) is @xmath23 gev .
we will discuss it somewhat above threshold , at @xmath3 gev , in order to reach lower ( absolute ) values of the squared 4-momentum transfer from the initial photon to the final @xmath4-meson . in these conditions , we expect the t - channel @xmath5- and @xmath6-meson exchanges to drive the dynamics underlying the @xmath0 process .
the initial photon dissociates into the final @xmath4-meson and a virtual pseudoscalar meson ( @xmath5 or @xmath6 ) .
the virtual pseudoscalar meson scatters from the proton target to produce the final @xmath1 state . the @xmath7 and @xmath8 amplitudes are derived in the framework of a coupled - channel effective field theory of meson - baryon scattering .
we found the @xmath6-meson exchange to be largely dominant .
the @xmath6-@xmath5 interference is of the order of @xmath9 .
the sign of this term is not known and has a significant influence on the results .
the @xmath10 amplitude being largely determined by data on the @xmath11 reaction , we found that the @xmath0 reaction cross section is rather directly related to the @xmath6-nucleon scattering amplitude in the n*(1535 ) resonance region .
accurate data on the @xmath0 process would therefore put additional constraints on this still poorly known amplitude .
0.3truecm _ key words _
: meson photoproduction ; eta - nucleon scattering ; n*(1535 ) _ pacs : _ 13.30.eg;13.60.le;13.60.rj;14.20.gk amssym.def amssym |
we consider the @xmath2 theory with a real scalar field @xmath3 in @xmath4 dimensions @xcite .
its dynamics determined by the lagrangian : @xmath5 where @xmath6 is a potential , defining the self - interaction of the field in the considered model @xcite : @xmath7 the plot of is shown at figure [ fig : sec2pic1 ] ( left panel ) .
we take up a model with a non - negative potential having two minima , so all static solutions with finite energy splits into disjoint classes , so called topological sectors , according to their asymptotic behavior at extremely large @xmath8 .
solutions with @xmath9 are called topological , while those with @xmath10 nontopological . both types of the solutions are of growing interest in physics . in particular , they arise in the questions of three- or two - dimensional domain walls . however
, a one - dimensional case also is curious and it was considered in different works for diverse models @xcite . in the @xmath2 model there is a soliton solution called a kink , in @xcite , @xcite a phenomenon of `` wobbling kink '' was studied .
moreover , a three- or two - dimensional domain wall presents a one - dimensional kink interpolating two different vacua of the model . in some cases for considering models the approximate methods
are offered @xcite .
the domain walls in the @xmath2 model can be applied in some cosmological models , for example , during discussions of the dark matter and dark energy themes @xcite .
the results of numerical simulations in other models @xcite can be applied in the solid body physics @xcite . the lagrangian with yields the equation of motion for @xmath3 after transition to dimensionless variables : @xmath11 at the next step
we find the analytical solutions of the equation and research them .
note , that the vacua of this model @xmath12 are stable solutions of .
moreover there is unstable permanent solution @xmath13 with infinite energy . except previous solutions
there is also a static non - trivial topological solution like solitary wave called a kink @xcite .
it can be easily found by solving static equation : @xmath14 the antikink @xmath15 is given as minus @xmath16 .
an energy functional for the lagrangian , in static case , is calling a mass of the kink @xmath17 .
the plot of is presented at figure [ fig : sec2pic1 ] ( right panel ) . in order to analyse the excitation spectrum of the static kink
, we add to it a small perturbation @xmath18 .
in other words we make such substitution to as @xmath19 taking the terms linear in @xmath18 , we obtain the next equation : @xmath20 the eigenvalue @xmath21 belongs to the discrete part of the excitation spectrum @xcite , but also there is one vibrational excitation : @xmath22 the above solutions are not a full set of solutions in the @xmath1 model .
let s consider a static wave solution with infinite energy .
we consider the static equation @xmath24 for an arbitrarily chosen value @xmath25 . ]
if one reduces the problem to newton s equation , will obtain @xmath26 figure [ fig : sec2lev4pic1 ] shows a plot of a @xmath27 . in this case
@xmath28 describes a trajectory of an oscillatory movement between points @xmath29 and @xmath30 . at @xmath31 oscillations
are periodic . in the limiting case
@xmath32 the fluctuations disappear , because time to return to the starting point @xmath32 reaches infinity . a transition to a dimensionless variable @xmath33 and following notations @xmath34 where @xmath35 and @xmath36 , leads us to : @xmath37 the last integral the elliptic integral of the first kind ( @xmath38 ) @xcite .
then the static periodic solution of equation can be written as @xmath39 where @xmath40 the elliptic sine @xcite . at small @xmath41 ( corresponding to @xmath42 )
there is a concordance @xmath43 . at @xmath44
the solution becomes a permanent unstable solution @xmath45 , previously noted .
the plots of are shown in the figure [ fig : sec2lev4pic2 ] for different values of parameter @xmath30 .
the elliptic sine period is calculated using the following formula @xcite : @xmath46
the long - living solutions with high amplitude are of growing interest in classical field theory .
this type of solutions , called bion or breather , was found early in the kink - antikink collisions in the @xmath1 model both in one- @xcite and three- dimensional cases @xcite . we propose a method `` cut and match '' to find a long - living field configuration with using previously found solutions @xmath13 , and of the equation . in this case a part of initial state is composed of the kink , which is divided for two equal pieces at @xmath47 .
these halves of the kink are fixed at @xmath48 .
then , one of the solutions ( @xmath13 or @xmath49 on the finite interval ) of is placed in the space between these two halves .
an initial state , which is composed in described manner , is shown by figure [ fig : sec3pic1 ] .
note , that if we take @xmath13 , the initial state will become unstable .
its energy linearly increases with growing of the distance @xmath50 .
in other case we take a solution in terms of elliptic solution for fixed value @xmath25 . for a smooth screed of selected solutions
one defines the value of @xmath51 as a half of the period @xmath52 of the elliptic function @xmath49 .
thus we obtain an initial configuration @xmath53 . in the following text
the definition @xmath54 means : in the area @xmath55 the initial state consists of a half of , in @xmath56 ( there @xmath57 , @xmath58 , @xmath59 ) , and in the area @xmath60 the solution consists of a second part of .
there @xmath52 is a period of elliptic function .
the profile of this type of initial state is shown in the figure [ fig : sec3pic1 ] ( for @xmath61 ) .
firstly , we consider a `` static initial state '' ( @xmath62 ) , but later we take into account some configurations with dynamics : @xmath63 for @xmath61 . ]
* numerical solution of the equation of motion*. we solve the partial differential equation with convergent difference scheme with non - fixed boundary conditions , while derivatives are approximated by finite differences .
the steps are taken as follows @xmath64 ( space step ) and @xmath65 ( time step ) , the equation is solved from @xmath66 to @xmath67 .
this choice of steps helps to optimize a ratio accuracy of the obtained results and the duration of computing . during the evolution
a check of law of conservation of energy is conducted , taking into account a flow of energy from fixed boundaries .
the initial states are compiled with the use of the computer algebra system mathematica 8 .
first of all , the initial condition consists of two halves of the kink , placed in @xmath48 , and unstable zero solution @xmath13 between them
. the energy increases linearly with growing value of @xmath51 .
two parts in the evolution are observed .
firstly , there is a convergence of both halves of kink with velocity equals the speed of light . when the halves finally meet each other , two processes alternate : a formation of loops and an emission of waves from the kink ( so called , `` wobbling kink '' ) .
the obtained solution @xmath69 is close to linearized solution of equation , where @xmath70 , which lives very long and is characterized by small emission of waves .
these waves carry off some energy from the area of localization .
let s explain this phenomenon .
at small values of @xmath71 the solution changes from @xmath72 to @xmath73 . as the @xmath74 is a periodic function , when @xmath75 the initial condition ( with loops ) comes not as an exciting of excited mode of elliptic function , but as an exciting of highly - amplitude vibration of a kink solution .
so the evolution of the initial state can be described qualitatively by @xmath76 the evolution can be described by as there are two modes in the kinks spectrum .
one of them , which correlates with , is responsible for small vibrations across the solution .
there is an idea , that the observed vibrations stop being small , but still can be described with similar periodic function as @xmath77 ( we take @xmath78 like in ) . in the parameter @xmath79 is taken as constant .
but in the numerical simulations it is nt a constant as there is a small emission from area of localisation of solution .
moreover , there is one precondition to describe qualitatively an obtained solution precisely with .
the function equals zero in @xmath47 one time , if @xmath80 , and three times , if @xmath81 .
note , that a quasi - periodic formation of the loops with period equals @xmath82 it is also one reason for using proposed phenomenological description .
this period correlates with @xmath83 : @xmath84 in @xcite is shown , that considering such substitution as @xmath85 in the equation in the quadratic approximation by @xmath86 gives us asymptotically stable solution .
its large amplitude vibrations are characterized by strong suppression . at figures [ fig : sec3lev1pic3 ] and [ fig : sec3lev1pic5 ]
two parts of evolution and a comparison with analytical solution are shown for two chosen moments of time . at high values of @xmath51
the observed loops in the evolution are characterized by not small amplitudes . in this case the final states of evolution can be identify with the elliptic solution @xmath49 between two halves of kink . in previous works
a long - living configuration has been found , so called bion @xcite .
however , an analytical description of the observed process has not been given . in our work
we take an initial condition @xmath89 , composed of one half - kink k and one half - antikink @xmath90 and a half of period of @xmath49 with fixed @xmath30 .
this initial state is dynamical ( @xmath91 ) .
we obtain long - living state with oscillation of an amplitude of @xmath92 at @xmath47 ( figure [ fig : sec3lev2pic1 ] ) .
the observing oscillations in terms of an amplitude @xmath93 are called a regular bion .
also , it can be considered as a new form of description of early found bion @xcite .
we take an initial condition @xmath53 for @xmath95 ( for smoother stitching ) , while the observed evolution does not qualitatively depend on @xmath30 .
further , for example , we show the results for the case @xmath61 .
we find two steps in the evolution : external ( a loop of high amplitude formation ) and internal ( a highly deformed kink ) . after some time the loops continue their forming with smaller amplitude .
after these 4 5 cycles the external step disappears and the solution starts to be presented by long - living excited kink with the wave packet emission from the area of localization . this phenomenon is called a `` wobbling kink '' . this state is a final step of the evolution , which is observed for other variants of initial states .
the profiles of @xmath3 for @xmath61 at some particular time are shown at figures [ fig : sec3lev2bpic1 ] and [ fig : sec3lev2bpic3 ] .
an addition of @xmath88 to the initial state of the configuration @xmath53 leads to a faster reduction of the amplitude . at low values of @xmath96 the loops
for the first time during the evolution the kink - antinkink pairs k@xmath15 turn up .
this phenomenon has a threshold .
the increase of @xmath97 gives us a qualitatively new type of the evolution ( @xmath98 for @xmath99 ) . in the system
the next transition happens : @xmath100 in the center of the configuration a topological number is changing .
a transition is shown at figure [ fig : sec3lev2dpic4 ] for @xmath101 .
the next increasing of @xmath97 ( @xmath102 for @xmath99 ) gives us the next transition @xmath103 now we observe a saving of topological number in the center .
these transitions are observed for different @xmath30 .
the transition is shown at figure [ fig : sec3lev2dpic5 ] for @xmath104 .
we expect that with increasing the value of @xmath97 the similar qualitative changes will be observed in the initial conditions of the evolution of initial state .
we study the new long - living solutions in classical field theory model @xmath2 in @xmath4 dimensions . in this work we use the special method for forming of initial state for numerical simulations
the method `` cut and match '' .
using this method gives us new long - living solutions both for vacuum solutions and solutions with nontrivial topological number . in previous work
@xcite a long - living configuration was observed in the kink - antikink scattering .
it was called a bion . in current work
the method `` cut and match '' gives us an opportunity to inject qualitative new way to describe a bion formation .
furthermore , the highly excited states of the kink are observed in a sector with nontrivial topological number .
we find a number of ways to reset this energy from this state . except for emission of wave packet with small amplitude , firstly , an arising of the kink - antikink pairs is observed .
this phenomenon can perceived as a way to energy reset . at the same time
there is a change of the topological number of the kink , located in the central zone in the area . at lower excitation energies
there is a long - living excited vibrational state of the kink .
the phenomenon of a `` wobbling kink '' is final state all considered means of exciting . after some time this excited state of a kink turns to a linearized one , which was formerly known a discrete mode of exciting kink . despite the large number of new results , the method `` cut and match ''
has a wide range of outstanding issues in the application of the @xmath2 model .
in particular , more detailed study of the dynamic of the initial conditions for the case of @xmath88 will be interesting , because in the last case there is a phenomenon of the birth of new kink - antikink pairs . in the conclusion
, we would like to notice that this research , fist of all , can be useful in the field theory .
but , also it can be useful in other area of physics .
the obtained results can be applied in elementary particle physics , cosmology , astrophysics and physics of condensed matter . with a high probability
we expect that the obtained solutions can be implemented in the early stages of the evolution of the universe .
we are very grateful to prof .
dr . i.l . bogolubsky for many numerous critical comments during a reading of the manuscript .
this work is supported by the russian federation government under the grant no . 945 from 18.11.2011 .
lizunova also acknowledges the support from the dynasty foundation and edward lozansky .
this work is part of the ditp consortium , a program of the netherlands organisation for scientific research ( nwo ) that is funded by the dutch ministry of education , culture and science ( ocw ) .
\a ) o.f . oxtoby and i.v .
barashenkov , teor . mat .
fiz . * 159 * , 527 ( 2009 ) [ theor .
159 * , 863 ( 2009 ) ] ; b ) o.f . oxtoby and i.v .
barashenkov , phys .
e * 80 * , 026608 ( 2009 ) ; c ) o.f .
oxtoby and i.v .
barashenkov , phys .
e * 80 * , 026609 ( 2009 ) .
rice and e.j .
mele , solid state comm . * 35 * , 487 ( 1980 ) ; m.j .
rice , phys .
b * 28 * , 3587 ( 1983 ) ; h. segur , j. math . phys . * 24 * , 1439 ( 1983 ) ; o.m .
kiselev , russ . j. math
. phys . * 5 * , 29 ( 1997 ) ; o.m .
kiselev , sib . math . j. * 41 * , 345 ( 2000 ) .
gornostyrev , m.i .
katsnelson , a.v .
kravtsov , and a.v .
trefilov , phys .
b * 60 * , 1013 ( 1999 ) ; yu.n .
gornostyrev , m.i .
katsnelson , a.v .
kravtsov , and a.v .
trefilov , phys .
e * 66 * , 027201 ( 2002 ) . | we look for long - living topological solutions of classical nonlinear @xmath0dimensional @xmath1 field theory . as for that the original method `` cut and match '' is offered . in the framework of this method
new long - living states are obtained in both topological sectors . in particular , a highly excited state of a kink
is found in one case .
we discover several ways of energy reset .
in addition to the expected emission wave packets ( with small amplitude ) in the case of some selected initial conditions a large energy reset becomes a result of the production of kink - antikink pairs . besides a topological number of a kink in the central region is changing in the contrast of saving full topological number . at lower excitation energies
there is a long - living excited vibrational state of the kink .
this phenomenon is the final stage of all considered initial states . over time this excited state of
the kink is changing to linearized well - known solution a discrete kinks excitation mode .
the proposed method yieldes a qualitatively new way of describing the large - amplitude bion , which was detected earlier in the kink scattering processes in the non - topological sector . |
the @xmath0-wave charmonium spectroscopy has been calculated by considering the hamiltonian with the non - relativistic qcd potential . for the next - to - leading order
qcd loop corrections to the potential , we expand the logarithmic factor @xmath1 about @xmath2 , where @xmath3 corresponds to the typical charmonium scale of order @xmath14 , so that the qcd potential can be modeled as the coulomb plus linear form , which is consistent with the cornell potential . in our approach , we have performed the best @xmath4 fit by comparing the current mass data of the @xmath0-wave charmonium states , @xmath5 , @xmath6 , @xmath7 , and @xmath8 , with their modified pad approximants . our results , in good agreement with the data , can successfully account for the hyperfine splitting for the @xmath9 state as well as for the @xmath10 state .
the fitted parameters are @xmath15 , and @xmath16 , consistent well with the ranges that one usually used . using then these three parameters
we have further predicted the @xmath0-wave mass spectrum with @xmath13 .
2.5 cm we are grateful to c.h . chen and c.w .
kao for useful comments .
we also would like to thank n. brambilla for correspondence .
this research was supported in part by the national center for theoretical sciences and the national science council of r.o.c . under grant
nsc96 - 2112-m-033 - 004-my3 and no .
nsc99 - 2112-m-003 - 005-my3 .
the eigenenergies and corresponding eigenfunctions of the hamiltonian operator @xmath17 can be calculated by using the rayleigh - schrdinger perturbation theory .
we decompose @xmath17 into two hermitian parts , @xmath18 and the rest , @xmath19 where @xmath20\delta^3(\vec{r})\,.\end{aligned}\ ] ] to solve the eigenvalue problem of @xmath17 , we instead consider the following hamiltonian function @xmath21 so that we have @xmath22 , where @xmath23 is a continuous real parameter .
[ [ the - corrections - due - to - the - spin - obit - and - tensor - interactions - in - s - wave - states - of - hamiltonian - h_c ] ] the corrections due to the spin - obit and tensor interactions in @xmath0-wave states of hamiltonian @xmath17 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ it is obviously that @xmath24 , where @xmath25 means the expectation value of the @xmath0-wave states , while @xmath26 is given by @xmath27 where @xmath28 \rangle \ , , \label{a6}\end{aligned}\ ] ] with @xmath29 being the spatial part of the wave function for the hamiltonian @xmath17 , given in eq .
( [ app : wf ] ) , which is spin - dependent but independent of @xmath30 since @xmath31 . in terms of the variables of spherical polar coordinates , @xmath32 , @xmath33 and @xmath30 : @xmath34 we have @xmath35 using the results of eqs .
( [ a1 ] ) and ( [ a5 ] ) to the angular integral in eq .
( [ a6 ] ) , we obtain @xmath36 and therefore @xmath37 since the spin - orbit and tensor interactions can be neglected for the @xmath0-wave states , following the standard approach , the eigenenergies @xmath38 of @xmath39 can be determined in terms of the perturbation expansion : @xmath40\delta^3(\vec{r})|\psi_{nlms}^{c(0 ) } \rangle \nonumber\\ & = & \int\psi^{c(0)}_{nlm}(\vec{r } ) \frac{4\pi c_f \alpha_s}{3m_c ^2}\left[s(s+1)-\frac{3}{2}\right]\delta^3(\vec{r})\psi^{c(0)}_{nlm}(\vec{r } ) r^2 \sin\theta dr d\theta d\phi \\
\biggl ( & = & \int^\infty _ 0 r^{(0)}_n(r ) \frac{4\pi c_f \alpha_s}{3m_c ^2}\left[s(s+1)-\frac{3}{2}\right](2l+1)\frac{\delta(r)}{4\pi r^2 } r^{(0)}_n(r)r^2 dr , \hskip0.3 cm { \rm for}\ m=0 \biggr)\ , , \nonumber\\ e_{nlms}^{(2)}&= & \sum_{k=1,\dots\atop k\neq n ~~}\frac{v_{nklms}v_{knlms}}{e^{(0)}_{nlms}-e^{(0)}_{klms}}\ , , \\ e_{nlms}^{(3)}&= & \sum_{k=1,\dots\atop k\neq n ~~ } \sum_{g=1,\dots\atop g\neq n ~~ } \frac{v_{nklms}v_{kglms}v_{gnlms } } { ( e^{(0)}_{nlms}-e^{(0)}_{klms})(e^{(0)}_{nlms}-e^{(0)}_{glms})}-\sum_{k=1,\dots\atop k\neq n ~~ } \frac{v_{nnlms}v_{nklms}v_{knlms}}{(e^{(0)}_{nlms}-e^{(0)}_{klms})^{2}},\end{aligned}\ ] ] where @xmath41 and @xmath42 are respectively the eigenenergies and eigenkets of @xmath43 , and @xmath44(2l+1 ) \frac{\delta(r)}{4\pi r^2 } r^{(0)}_k(r)r^2 dr , \hskip0.3 cm { \rm for}\ m=0 \biggr ) . ~~\end{aligned}\ ] ] the states @xmath42 read @xmath45 where @xmath46 } } \left(\frac{m_c\alpha r}{n b } \right)^l e^{-\frac{\alpha m_c}{2 n b}r}l^{2l+1}_{n - l-1 } \left(\frac{m_c\alpha r}{n b}\right ) . ~~\end{aligned}\ ] ] in the calculation , we introduce the transformation , called the pad approximation , to accelerate the convergence of the perturbative series of @xmath38 which are approximately presented as rational functions , @xmath47(\kappa)$ ] , @xmath48(\kappa)&= & \frac{p^c_0+p^c_1\kappa+p^c_2\kappa^2}{1+q^c_1 \kappa},\end{aligned}\ ] ] where @xmath49 and @xmath50 can be determined by the values of @xmath41 , @xmath51 , @xmath52 , and @xmath53 .
therefore , for the @xmath0-wave states , the eigenenergies of the hamiltonian @xmath17 are approximately to be @xmath54(1 ) & = & \frac{p^c_0+p^c_1+p^c_2}{1+q^c_1 } , \end{aligned}\ ] ] and the corresponding wave functions are @xmath55 where up to the second order @xmath56 \equiv\sqrt{\frac{1}{4\pi}}\bar{r}_{ns}(r)\ , , \label{app : wf}\end{aligned}\ ] ] with @xmath57 g. t. bodwin , e. braaten and g. p. lepage , phys .
d * 51 * , 1125 ( 1995 ) [ erratum - ibid .
d * 55 * , 5853 ( 1997 ) ] [ arxiv : hep - ph/9407339 ] .
e. braaten , arxiv : hep - ph/9702225 .
w. e. caswell and g. p. lepage , phys .
b * 167 * , 437 ( 1986 ) .
b. a. thacker and g. p. lepage , phys .
d * 43 * , 196 ( 1991 ) .
a. pineda and j. soto , nucl .
suppl . * 64 * , 428 ( 1998 ) [ arxiv : hep - ph/9707481 ] .
e. eichten , k. gottfried , t. kinoshita , k. d. lane and t. m. yan , phys .
d * 17 * , 3090 ( 1978 ) [ erratum - ibid .
d * 21 * , 313 ( 1980 ) ] .
e. eichten , k. gottfried , t. kinoshita , k. d. lane and t. m. yan , phys .
d * 21 * , 203 ( 1980 ) .
g. s. bali , phys .
b * 460 * , 170 ( 1999 ) [ arxiv : hep - ph/9905387 ] .
g. s. bali , arxiv : hep - ph/0010032 .
d. ebert , r. n. faustov and v. o. galkin , phys .
d * 67 * , 014027 ( 2003 ) [ arxiv : hep - ph/0210381 ] .
j. zeng , j. w. van orden and w. roberts , phys .
d * 52 * , 5229 ( 1995 ) [ arxiv : hep - ph/9412269 ] .
t. barnes , s. godfrey and e. s. swanson , phys .
d * 72 * , 054026 ( 2005 ) [ arxiv : hep - ph/0505002 ] .
b. q. li and k. t. chao , phys .
d * 79 * , 094004 ( 2009 ) [ arxiv:0903.5506 [ hep - ph ] ] .
s. l. olsen , arxiv:0801.1153 [ hep - ex ] .
s. godfrey and s. l. olsen , ann .
nucl . part .
sci . * 58 * , 51 ( 2008 ) [ arxiv:0801.3867 [ hep - ph ] ] .
m. b. voloshin , prog .
phys . * 61 * , 455 ( 2008 ) [ arxiv:0711.4556 [ hep - ph ] ] .
c. m. bender and s. a. orszag , _ advanced mathematical methods for scientists and engineers : asymptotic methods and perturbation theory _ , 2nd ed . ,
springer , new york , 1999 .
a. falkowski and m. perez - victoria , jhep * 0702 * , 086 ( 2007 ) [ arxiv : hep - ph/0610326 ] .
p. masjuan and s. peris , phys .
b * 686 * , 307 ( 2010 ) [ arxiv:0903.0294 [ hep - ph ] ] .
s. peris , phys .
d * 74 * , 054013 ( 2006 ) [ arxiv : hep - ph/0603190 ] .
a. h. mroue , l. e. kidder and s. a. teukolsky , phys .
d * 78 * , 044004 ( 2008 ) [ arxiv:0805.2390 [ gr - qc ] ] .
j. t. pantaleone , s. h. h. tye and y. j. ng , phys .
d * 33 * , 777 ( 1986 ) .
n. brambilla , y. sumino and a. vairo , phys .
b * 513 * , 381 ( 2001 ) [ arxiv : hep - ph/0101305 ] .
p. chen , phys .
d * 64 * , 034509 ( 2001 ) [ arxiv : hep - lat/0006019 ] .
x. liao and t. manke , arxiv : hep - lat/0210030 . t. barnes , arxiv : hep - ph/0406327 . | we calculate the @xmath0-wave charmonium spectroscopy using the hamiltonian with the non - relativistic qcd ( nrqcd ) potential .
the logarithmic factor @xmath1 , appearing in the next - to - leading order qcd loop corrections to the potential , is expanded about @xmath2 , where @xmath3 corresponds to the typical charmonium scale .
the resulting potential characterized by the coulombic and linear components is consistent with the form of the cornell potential .
we obtain @xmath4 fitting results for the masses of the @xmath0-wave charmonium states , @xmath5 , @xmath6 , @xmath7 , and @xmath8 in remarkable accordance with data
. our results successfully account for the hyperfine splitting for the @xmath9 state as well as for the @xmath10 state .
we further use the three best fit parameters : the charm quark mass @xmath11 , coupling constant @xmath12 and the corresponding scale @xmath3 to predict the @xmath0-wave mass spectrum with @xmath13 .
the hints for results are discussed .
[ cols= " < " , ] |
powerful high @xmath13 radio galaxies ( @xmath152 , hzrg ) are often surrounded by giant ly@xmath1 nebulae which can extend for more than 100 kpc ( e.g. mccarthy et al .
@xcite ; reuland et al .
@xcite , villar - martn et al .
@xcite ) and sometimes beyond the radio structures ( e.g. eales et al . @xcite ; kurk et al .
@xcite ; maxfield et al .
@xcite ) . these are clumpy , irregular ( with features such as filaments , plumes , ionization cones , e.g. reuland et al .
@xcite ) and often aligned with the radio axis ( mccarthy et al .
they are characterized by extreme kinematics , with measured fwhm @xmath141000 km s@xmath7 ( e.g. mccarthy , baum & spinrad @xcite ; villar - martn et al .
@xcite ) , compared with values of a few hundred in low - redshift radio galaxies ( e.g. tadhunter , fosbury & quinn @xcite ; baum , heckman & van breugel@xcite ) .
there is strong evidence that interactions between the radio structures and the ambient gas , produce an outflow responsible for such extreme kinematics ( e.g. humphrey et al .
@xcite , van ojik et al .
1997 ) .
the nebulae have typical values of ionized gas mass @xmath810@xmath16 m@xmath17 , ly@xmath1 luminosities @xmath8several@xmath510@xmath18 erg s@xmath7 and densities @xmath19 few to several hundred @xmath20 ( e.g. mccarthy @xcite , villar - martn et al .
they emit a rich emission line spectrum dominated in the optical ( uv rest frame ) by ly@xmath1 followed by civ@xmath211550 , heii@xmath211640 and ciii]@xmath211909 ( civ , heii and ciii ] hereafter ) .
such a line spectrum reveals high levels of metal enrichment and excitation mechanisms mostly related to the nuclear activity , at least in the direction along the radio structures ( e.g. vernet et al .
@xcite ) .
in addition to the highly perturbed gas , hzrg are often embedded in giant ( often @xmath22100 kpc ) , low surface brightness nebulae of metal rich , ionized gas with _ quiescent kinematics _
( villar - martn et al .
@xcite , reuland et al .
@xcite ) , i.e. , not perturbed by interactions with the radio structures .
we have recently shown that these quiescent nebulae are infalling towards the central region ( humphrey et al .
@xcite ) .
there are only several narrow band ly@xmath1 images of hzrg and most spectroscopic studies have been performed with the long - slit technique , with the slit aligned with the radio structures .
these studies have been seriously limited by , respectively , the lack of spectral information , or the lack of spatial information in directions other than the radio axis .
for this reason , we are carrying out an observational program of 3d integral field spectroscopy of powerful radio galaxies at @xmath22 - 3 with vimos on vlt and pmas / ppak on the 3.5 m telescope in calar alto observatory .
the main goal is to characterize the morphological , kinematic and ionization properties of the extended ionized gas in two spatial dimensions . in this paper
, we present results obtained for mrc 1558 - 003 ( @xmath13=2.53 ) , mrc 2025 - 218 ( @xmath13=2.63 ) and mrc 0140 - 257 ( @xmath13=2.64 ) , based on vimos - vlt data .
results on mrc 2104 - 242 ( @xmath13=2.49 ) can be found in @xcite .
a similar study of 4c40.36 ( @xmath13=2.27 ) and 4c48.48 ( @xmath13=2.34 ) based on pmas / ppak data ( 3.5 m telescope , calar alto observatory ) will be presented in snchez et al .
( 2007 , in prep . ) .
a @xmath23 0.73 , @xmath24 = 0.27 and @xmath25 = 62 km @xmath26 mpc@xmath7 cosmology is adopted in this paper ( sandage et al .
the observations ( program 075.b-0212(a ) ) were made on ut 2005 july 28 , 29 and 30 using the visible multiobject spectrograph ( vimos , le fvre et al .
@xcite ) , on the nasmyth focus of the ut3 vlt .
the instrument is equipped with an integral field unit with 6400 microlenses coupled to fibres . for the configuration selected by us ,
the number of fibers in use is 1600 , covering 27 x 27 on the sky with a 0.67 " sampling .
the hr@xmath27 grating was used , with an effective wavelength range @xmath84150 - 6200 , an instrumental profile of fwhm [email protected] and a pixel scale of 0.5 pixel@xmath7 .
the exposure time on each target was 7.3 h ( 22@xmath51200 s ) on mrc 1558 - 003 , 10 h ( 30@xmath51200 s ) on mrc 2025 - 218 and 8.3 h ( 25@xmath51200 s ) on mrc 0140 - 257 .
in addition , we had 4.5 h ( 9@xmath51800 s ) on mrc 1558 - 003 obtained in june 2004 ( program 073.b-0189(a ) , see villar - martn et al .
@xcite for a description of these observations ) .
the total integration time on this source was therefore 11.8 h. the seeing full width at half maximum during the observations was in the range @xmath80.44 - 1.32
( 1st night ) , 0.4 - 1.2 2nd night ) and 0.5-3.0 ( 3rd night ) for the 2005 observations , and 1.0 - 1.4@xmath29 for the 2004 data on mrc 1558 - 003
. for each galaxy a dithering pattern was applied , with a maximum offset of @xmath83 , and a range of dithering pointings between 3 and 7 .
the data were reduced using r3d ( snchez 2006 ) and iraf routines .
the data were bias subtracted .
the locations of the spectra were traced on a continuum - lamp exposure obtained before each target exposure .
the corresponding spectrum was then extracted for each fiber by coadding the flux intensity within a 5 pixel aperture , along the spectral dispersion axis , centered on the estimated location of the fiber centroid .
the wavelength calibration was performed using arc lamp spectra and the telluric emission lines in the science data .
the wavelength solution was stable within a fraction of 1 spectral pixel ( 0.5 ) across the whole spectral range and the whole field of view .
the wavelength calibration in the blue ( i.e. the ly@xmath1 spectral region ) was problematic due to the lack of bright emission lines in the arc lamp and sky spectra .
residual , artificial shifts in @xmath21 of up to 6 were measured for ly@xmath1 relative to civ and heii in mrc 1558 - 003 and mrc 2025 - 218 .
we have been able to correct for this effect by comparing with available keck long - slit spectroscopy ( villar - martn et al .
@xcite , humphrey @xcite ) , to an accuracy of @xmath31 . in any case
, such uncertainty will not affect the results presented here .
the fibre - to - fibre response at each wavelength was determined from a continuum - lamp exposure .
after these basic reduction steps , a data cube was created for each exposure .
the cubes for mrc 2025 - 218 were then recentered spatially at each wavelength by determining the centroid of a nearby star in the vimos field of view .
this recentering corrects for differential atmospheric refraction . for the other two objects , there are no stars in the vimos field of view .
we used the information on the spatial shifts contained in the image headers to apply the corresponding shifts .
however , we found that this technique left substantial spatial offsets .
for this reason , we used the ly@xmath1 peak of emission to recenter all cubes .
although this recentering is valid in the blue , it is not clear that it is also valid in the red . in particular , in the case of mrc 1558 - 003 , we found a residual spatial shift of @xmath81 spaxel in the direction of the radio structures between the blue ( ly@xmath1 ) and red ( civ , heii ) parts of the spectrum , which is much larger than found in previous works ( villar - martn et al .
@xcite ; humphrey et al .
@xcite ) . for this reason ,
an additional correction was applied .
uncertainties remain regarding the spatial centering in the direction perpendicular to the radio axis , although this is likely to be @xmath111 spaxel or 0.67@xmath29 .
this will not affect seriously the results presented here . for mrc 0140 - 257 , no useful previous works or additional data were available .
although the accuracy of the spatial centering between the red and the blue ends of the spectrum is uncertain , this will not affect the results presented here , due to the nature of our analysis and the clear separation between the spatial components of the ly@xmath1 nebula .
the cubes were then combined using iraf tasks , masking the broken and/or low sensitivity fibres .
a 3@xmath30 clipping algorithm removed cosmic rays .
the sky background was estimated before subtraction by selecting the spectra of object free areas and creating an interpolated datacube ( using e3d , snchez @xcite ) . a spatial median smoothing using a 4x4 spaxel box
was applied .
the nights were non - photometric .
observations of standard stars were used to perform a relative calibration from blue to red .
the absolute flux calibration for mrc 1558 - 003 and mrc 2025 - 218 was done using available keck long - slit spectroscopy along the radio axis .
we extracted a pseudo - slit from the vimos data as similar as possible to the slit aperture of the keck spectra .
the flux in different spectral windows was then compared and scaled .
the final flux calibration has an accuracy of @xmath820% . for mrc 0140 - 257 ,
this could not be done , due to the unavailability of a high quality long - slit spectrum , but as we explain in 5.3 , the agreement between the measured ly@xmath1 flux and published values is reasonably good ( @xmath820% ) .
cross - talk effects are estimated to be negligible .
the fibre to fibre contamination is expected to be @xmath115% for adjacent spectra in the ccd , dropping to less than 1% for the 3rd adjacent spectra . to overlay the radio maps ( carilli et al .
@xcite , pentericci et al .
@xcite ) on the ly@xmath1 images , a different method was used for each object depending on the information available in the data . for mrc 2025 - 218
, we placed the radio core at the position of the continuum centroid in the vimos data .
since this is a broad line object ( 3.2 ) , it is reasonable to expect that this marks the location of the active nucleus ( agn ) .
this is further supported by the fact that the uv continuum has an spatially unresolved component ( 3.2 ) . for mrc 1558 - 003
, we positioned the radio core at the spaxel with the maximum ly@xmath1 flux , which is expected to be shifted by a few tenths of a spaxel relative to the continuum centroid ( humphrey et al .
this method would be incorrect if the agn is spatially shifted relative to the continuum and/or line centroids .
however , we do not expect this to be the case since this is a broad line object and moreover , it shows little evidence for ly@xmath1 absorption ( see below ) .
the 1@xmath30 uncertainty in the ly@xmath1-radio registration is @xmath80.3 " in both cases . for mrc 0140 - 257
, we assumed that the radio core is placed between the two ly@xmath1 spatial components , although it is not necessarily the case .
this is the main source of uncertainty since the radio core could be shifted by 2@xmath29 if it is associated with one of the two ly@xmath1 blobs ( 4.3 ) .
the impact of this assumption on the interpretation of our results will be discussed when relevant .
the main scientific goal of our vimos - vlt observational program is to study the properties of the giant nebulae associated with a sample of hzrg . by isolating spectroscopically and/or spatially the emission from the perturbed and the quiescent gaseous components ( e.g. villar - martn et al .
@xcite ) , one of our priorities is to characterize the morphological , kinematic and ionization properties of the quiescent gas . in this way we can study the gas without the strong distortions that the radio structures can imprint on its observed properties .
such study has the potential to provide critical information on the star formation and chemical enrichment histories as well as the galaxy formation process ( villar - martn et al .
@xcite , humphrey et al .
@xcite , @xcite ) .
in order to achieve these goals , it is critical to detect the main uv lines ( very especially lya and heii ) with high signal / noise in the faint , outer regions of the objects , which are usually characterized by quiescent kinematics .
moreover , this would allow to study in detail the high surface brightness regions ( at least several times brighter ) .
although usually distorted by the radio structures , it should be possible to isolate spectroscopically the emission from the quiescent and the perturbed gas using especially the non resonant heii line , as villar - martn et al .
( @xcite ) .
the faintest regions of our interest have often ly@xmath1 surface brightness levels which are , within the errors , consistent with or even below the detection limit or our vimos data ( 3@xmath3110@xmath32 erg s@xmath7 @xmath33 arcsec@xmath34 ) for 8 - 10 hours exposure time . for comparison , van ojik et al .
( @xcite ) detected in 4 hours the giant , quiescent , low surface brightness nebula ( @xmath8l0@xmath32 erg s@xmath7 @xmath33 arcsec@xmath34 ) associated with a radio galaxy at @xmath103.6 using long slit spectroscopic data obtained with emmi ( eso multi - mode instrument ) on the 3.5 m ntt telescope ( la silla ) . with a 2.5@xmath29 wide slit , the spectral resolution was of 2.8 , not very different to that of our data . in the vimos data ,
not even ly@xmath1 could be definitively detected from the faintest regions of our interest , which at least in mrc 1558 - 003 and mrc 2025 - 218 we knew exist ( 3 ) . heii and civ are detected only from the highest surface brightness regions , but with not enough signal / noise ratio to perform an adequate kinematic and ionization analysis in two spatial dimensions .
thus , the investigation of the properties of the quiescent gas has been possible only for those objects where there is no signature of interactions between the radio structures and the gas ( mrc 1558 - 003 and mrc 0140 - 257 , see 4 ) or objects with no spectroscopic blend between the perturbed and the quiescent nebula ( mrc 2104 - 242 , villar - martn et al .
moreover , in all three cases , ly@xmath1 is not heavily absorbed .
for objecs where the emissions from the perturbed and the quiescent gas are blended and/or ly@xmath1 is absorbed ( i.e. an important fraction of hzrg ) , a more sensitive instrument / telescope combination is needed .
the three radio galaxies discussed in this paper , mrc 1558 - 003 , mrc 2025 - 218 , mrc 0140 - 257 , belong to the molonglo catalogue of high redshift radio galaxies ( large et al .
@xcite , mccarthy et al .
the objects were selected based on previous evidence for strong ly@xmath1 emission .
the two first objects were previously known to be associated with giant ( @xmath060 kpc ) ly@xmath1 nebulae ( see below ) . this object
is associated with radio structures which extend for @xmath89 or 84 kpc ( pentericci et al .
@xcite ) along a position angle ( @xmath35 ) of 75@xmath4 .
previous long - slit spectroscopic studies have shown that ly@xmath1 is extended for at least @xmath814@xmath29 ( @xmath8130 kpc ) along the slit with @xmath3672@xmath4 , well beyond the radio structures ( e.g. villar - martn et al .
@xcite ) . civ and heii are extended across @xmath86 , while nv , which is fainter , has a more compact appearance .
the optical continuum is also extended .
isaac near infrared spectroscopy revealed very broad h@xmath1 ( fwhm@xmath811700@xmath28900 km s@xmath7 , humphrey@xcite ; humphrey et al .
2007 , in prep . ) , evidence for an obscured broad line region ( blr ) .
according to the unified scheme of quasars and radio galaxies ( barthel @xcite ) , the detection of blr emission implies that the ionization cones axis is at a small angle to line of sight ( @xmath1145@xmath37 ) .
this is further suggested by the detection of a clear one - sided radio jet ( see fig . 1 )
. high resolution optical spectroscopy ( [email protected] ) showed no absorption features in the ly@xmath1 spectral profile integrated along @xmath35=85@xmath4 ( van ojik et al .
absorption features were not found either along @xmath35=75@xmath4 by villar - martn et al .
( @xcite ) , although the spectral resolution was rather low in this case ( fwhm@xmath811 ) .
the small impact of absorption effects is also suggested by the large ly@xmath1/heii values measured across the nebula ( humphrey @xcite , villar - martn et al .
@xcite ) for this object no hst or emission line images are available in the literature .
this radio galaxy is associated with a small radio source ( @xmath85 or 46 kpc , carilli et al .
@xcite ) .
there is strong evidence for an obscured blr in mrc 2025 - 218 .
the morphology of the object at optical and near infrared wavelengths is dominated by a point source ( pentericci et al . 1999 , 2001 ; mccarthy , person & west 1992 ) . broad civ and broad h@xmath1 ( fwhm@xmath86000 km s@xmath7 ) revealed by optical ( villar - martn et al .
@xcite ) and near infrared ( larkin et al . @xcite ; humphrey@xcite ; humphrey et al .
2007 , in prep . )
spectra confirm this interpretation .
the uv rest frame continuum emission shows also an extended ( @xmath86 ) diffuse component , which is well aligned with the radio axis ( pentericci et al .
1999 ) .
the ly@xmath1 emission is distributed bimodally ( mccarthy et al .
@xcite ) north - south with two peaks that correspond roughly to the two radio lobes .
the authors measured a total extension of @xmath84 or 37 kpc .
keck long - slit optical spectroscopy ( humphrey@xcite ) reveals that ly@xmath1 is extended up to @xmath89 or 83 kpc .
civ , heii and ciii ] are also spatially extended .
civ and ciii ] each have a spatially unresolved component centered at the continuum centroid and also an underlying extended component spanning @xmath85 along the radio axis .
nv is spatially unresolved .
@xcite detected absorption in the spectrum of mrc 2025 - 218 for civ , cii@xmath211335 , [email protected],1402.8 and , maybe , [email protected] + [email protected] .
the steep ly@xmath1 profile on the blue side of the line was also proposed to be a signature of absorption .
this object is associated with a small double radio source ( 4.2@xmath29 or 39 kpc , carilli et al .
@xcite ) . in the optical ,
it appears as a faint galaxy spatially extended and aligned with the radio source ( mccarthy , persson & west 1992 ) .
ly@xmath1 and h@xmath1 are the two lines detected in previous spectroscopic studies ( mccarthy et al .
@xcite ; eales & rawlings @xcite ) .
it has not been reported whether the lines are spatially extended . the f160w filter nicmos - hst image ( rest frame spectral window 3850 - 4950 , pentericci et al .
@xcite ) shows two peaks of emission with almost the same flux , closely aligned with the radio axis .
the authors suggest that this morphology could be due to a dust lane .
there are a few fainter clumps within 2 - 3 .
@xcite report ly@xmath1/h@xmath1=0.31 , which is @xmath830 times lower than standard case b photoionization model predictions . as the authors explain , this value is very uncertain , since they had to assume a fixed value of the [ nii]/h@xmath1 ratio ( 0.5 ) and no aperture corrections to the line fluxes were attempted . in spite of these uncertainties ,
the very large discrepancy with the model predictions suggest that ly@xmath1 absorption / extinction is present . for this object
no emission line images are available in the literature .
we present below the most relevant results obtained for the three radio galaxies investigated here . _ the ly@xmath1 nebula _
the vimos ly@xmath1+continuum image of mrc 1558 - 003 was created by adding the monochromatic images of the object within the [ 4285 - 4315 ] range .
a continuum image was subtracted to keep the line emission only .
this was created by combining the images extracted from two adjacent spectral windows at both sides of ly@xmath1 , with the same spectral width as the line+continuum image .
the resulting ly@xmath1 image is shown in fig . 1 ( see also fig
. 2 ) , with the 8.2ghz radio contours overlaid ( pentericci et al .
@xcite ) .
the nebula extends across @[email protected] or 84@xmath570 kpc@xmath39 ( but see 3.1 ) .
it is characterized by a strongly peaked rounded morphology .
emission from the nebula is detected only within the radio structures .
the faintest ly@xmath1 emission we detect on the vimos image has surface brightness @xmath810@xmath32 erg s@xmath7 @xmath33 arcsec@xmath34 ( 3@xmath30 values ) .
the angle between the nebular axis and the radio structures was measured using the position angle on the sky of the longest dimension of the ly@xmath1 nebula and the position angle of the line between the two brightest radio hot spots .
the nebula is misaligned by @xmath830@xmath285@xmath4 relative to the radio axis . as we found for mrc 2104 - 242 ( villar - martn et al .
@xcite ) , ly@xmath1 emission is detected outside any plausible ionization cones with opening angle @xmath4090@xmath4 ( barthel @xcite ) .
seeing effects are not likely to be responsible for this ly@xmath1 emission since the observations of this object were carried out under seeing conditions of fwhm@xmath11 1.5@xmath29 .
since this is a broad line object , it is possible that the broad rounded morphology of the ly@xmath1 nebula is due to orientation effects ( see 5.3 ) .
if the ionization cone axis lies close to the line of sight , a broader , rounder morphology is expected .
the total ly@xmath1 flux integrated over the nebula is 4.8@xmath510@xmath41 erg s@xmath7 @xmath33 corresponding to a luminosity of 3.4@xmath510@xmath6 erg s@xmath7 . _
civ and heii morphologies _
the civ and heii images were created by adding the monochromatic images of the object within the spectral windows [ 5455 - 5485 ] and [ 5780 - 5800 ] respectively .
continuum images adjacent in wavelength to each line and of the same spectral width were subtracted to keep the line emission only .
the resulting images are shown in fig . 2 .
the ly@xmath1 contours have been overplotted on the top panels for comparison .
the civ line is spatially extended with a maximum projected size of @xmath87.5@xmath29 .
the outer , low surface brightness regions extend in the same direction as ly@xmath1 .
heii , which is fainter , appears more rounded and compact , although it is spatially resolved with a maximum extension of @xmath83@xmath29 .
the inner ( @xmath82@xmath29 ) regions of the civ nebula seem to be extended along an axis ( roughly e - w ) which is shifted by @xmath815@xmath37 anti - clockwise relative to the axis defined by the outer regions and aligned within a few degrees with the radio structures .
the heii emission seems to be extended also in this direction .
there is some hint that this could also be the case for the inner ly@xmath1 nebula .
however , higher spatial resolution data would be necessary to investigate whether there is a real rotation of the nebular axis as we move outwards , which could be a consequence of an inhomogeneous distribution of material .
continuum is also detected , although the image is too noisy to characterize its morphology and determine its spatial centroid .
_ spatially extended emission line spectroscopy _
we show in fig .
3 the spatial maps of the fwhm ( corrected for instrumental profile ) and velocity shift @xmath42 of the ly@xmath1 line .
these values have been measured fitting gaussian profiles spaxel to spaxel .
@xmath42 has been computed relative to the ly@xmath1 emission at the spatial line centroid .
the errors on the fwhm and @xmath42 are estimated to be , in general , @xmath11100 km s@xmath7 and @xmath1140 km s@xmath7 respectively .
although the line profile is asymmetric at some spatial positions , this method allows a first order study of the spatial variation of the ly@xmath1 spectral profile across the nebula .
the line spectral profile shows strong spatial variations , as is obvious from fig .
the first striking characteristic is that ly@xmath1 presents an increasing redshift towards the nebular centroid , with a maximum shift in velocity of @xmath8400 km s@xmath7 at this position relative to the outer regions .
the fwhm varies between @xmath8450 and 1350 km s@xmath7 across the nebula .
three distinct regions ( @xmath43 , @xmath44 and @xmath45 in fig .
3 ) can be isolated in the fwhm map according to the line width .
the maximum values are measured in region @xmath44 ( in the range 1050 - 1340 km s@xmath7 ) , which runs approximately along the radio axis and contains the nebular centroid .
the heii fwhm measured from the integrated spectrum is 600@xmath28100 km s@xmath7 .
region @xmath43 shows narrower ly@xmath1 , although the line is still quite broad ( fwhm in the range 700 - 1030 km s@xmath7 ) while the heii fwhm is 650@xmath28200 km s@xmath7 in the integrated spectrum . at some positions
heii is as narrow as 450@xmath2850 km s@xmath7 ( consistent with villar - martn et al .
ly@xmath1 trends to be notably broader than heii ( this was also found by villar - martn et al .
@xcite along the radio axis ) .
region @xmath45 is clearly different both in line width ( fwhm(ly@xmath1)=650@xmath2850 km s@xmath7 , fig . 4 , bottom ) and velocity shift .
this region shows the largest blueshift relative to the ly@xmath1 centroid ( 350@xmath2820 km s@xmath7 for the integrated spectrum ) .
civ is detected in region @xmath45 , although noisy , and it has fwhm=600@xmath28100 km s@xmath7 .
_ ly@xmath1 absorption _
we do not find definitive ly@xmath1 absorption features in the vimos data across the nebula .
this is consistent with previous studies ( 3.1 ) .
_ the ly@xmath1 nebula _ the vimos ly@xmath1+continuum image of mrc 2025 - 218 was created by adding the monochromatic images of the object within the [ 4385 - 4445 ] range .
a continuum image extracted from an adjacent line - free region of the same spectral width was subtracted to keep the line emission only .
the resulting ly@xmath1 image is shown in fig . 4 , with the vla 8.2ghz radio contours overlaid ( carilli et al . @xcite ; see also fig .
the ly@xmath1 maximum extension is @xmath86.0 or @xmath855 kpc ( but see 3.2 ) , very similar to the radio source size .
the faintest ly@xmath1 emission we detect on the vimos image has surface brightness @xmath81.5@xmath510@xmath32 erg s@xmath7 @xmath33 arcsec@xmath34 ( 3@xmath30 values ) .
the maximum extension in the direction perpendicular to the radio structures as measured from the vimos data is @xmath84@xmath29 .
as @xcite already pointed out , the nebula is aligned with the radio structures within a few degrees .
the bimodal distribution described in 3.2 is clearly seen .
the total ly@xmath1 flux integrated over the nebula is 2.1@xmath510@xmath41 erg s@xmath7 @xmath33 corresponding to a luminosity of 1.7@xmath510@xmath6 erg s@xmath7 .
_ civ and continuum morphologies _
the civ image ( heii is too faint ) was created by adding the monochromatic images of the object within the spectral window [ 5610 - 5650 ] and subtracting the adjacent continuum .
the continuum image of the object was created by integrating across the [ 5670 - 5850 ] spectral window .
the resulting images are shown in fig.5 .
the ly@xmath1 contours have been overplotted ( top panels ) for comparison .
civ extends for @xmath84@xmath29 in the same direction as ly@xmath1 ( n - s ) .
the bimodal distribution shown by ly@xmath1 is distinguished in the civ image .
interestingly , the continuum is spatially unresolved in this direction , but is barely resolved in the e - w direction , with a fwhm of @[email protected]@xmath29 , compared with [email protected]@xmath29 for the star in the field .
the continuum centroid is located between the two ly@xmath1 spatial components , i.e. , where this line presents a minimum .
_ spatially extended emission line spectroscopy _
we show in fig .
6 the spatial maps of the fwhm and velocity shift @xmath42 of the ly@xmath1 line .
as before , these values have been measured by fitting gaussian profiles to the line spaxel by spaxel .
@xmath42 has been computed relative to the ly@xmath1 emission at the position of the continuum centroid .
the errors on the fwhm and @xmath42 are estimated to be , in general , @xmath11100 km s@xmath7 and @xmath1135 km s@xmath7 respectively .
this analysis reveals two clearly distinct regions ( north and south ) , which are different both in fwhm and @xmath42 .
these are coincident with the two ly@xmath1 spatial components discussed above .
the continuum centroid ( located between both regions ) shows intermediate fwhm values .
the ly@xmath1 spectra extracted from the apertures @xmath46 and @xmath47 ( north and south regions , fig .
6 ) and the region in between ( 2 spaxels , see aperture @xmath45 in fig .
6 ) are shown in fig .
7 ( left panels ) .
both regions emit also civ and heii ( fig.7 , top - right panel ) .
the civ - heii spectra ( fig . 7 , top right panel ) have been extracted using ony the highest flux spaxels in the n and s regions ( 5 and 4 spaxels respectively ) in order to maximize the signal to noise ratio for the lines .
the differences in fwhm and @xmath42 seen in fig .
6 are obvious also here .
the @xmath47 region shows the narrowest ly@xmath1 spectral profiles ( fwhm in the range 500 - 650 km s@xmath7 ) and the gas is blueshifted relative to the continuum centroid ( values in the range @xmath8 - 100 to -200 km s@xmath7 ) .
civ ( which is a doublet and , therefore , expected to be intrinsically broader ) and heii are also rather narrow , with fwhm=650@xmath2830 and 500@xmath2830 km s@xmath7 respectively , as measured from the spatially integrated spectrum
. the @xmath46 region is characterized by broader emission lines ( fwhm in the range @xmath8850 - 1070 km s@xmath7 for ly@xmath1 ) and the gas is redshifted relative to the continuum centroid ( @xmath42 in the range @xmath8 + 50 to + 250 km s@xmath7 ) .
the civ fwhm is 1220@xmath2840 km s@xmath7 .
heii is too noisy in the vimos spectrum to measure its fwhm .
however , keck long - slit spectroscopy ( humphrey @xcite ) of this region implies fwhm=1100@xmath28100 km s@xmath7 .
therefore , the three lines in the n region are very broad compared with the s region .
although absorption plays an important role in the ly@xmath1 spectral differences across the nebula ( see below ) , kinematics is also clearly having an effect , since the civ and heii lines show marked differences between the n and s regions .
the ly@xmath1 and civ spectra at the position of the continuum centroid are shown in fig .
7 ( bottom panels ) . very broad wings ( presumably from the blr , see 3.2 ) are seen in the line profiles of both ly@xmath1 and civ .
_ ly@xmath1 and civ absorption _
the ly@xmath1 and civ lines are clearly absorbed in this object ( fig .
7 ) , as was already discussed by other authors ( villar - martn et al .
since this paper is mostly focused on the properties of the emission line nebulae , we will present here a general description and defer a more detailed analysis of the absorbers for another publication ( humphrey et al .
2007 , in prep ) .
some peculiar properties of the ly@xmath1 spectral profile are a consequence of absorption rather than kinematics : the multiple peaks , the fact that the flux drops below the continuum level at some wavelengths , the sharp , almost vertical edge of the blue wing of the line , and the identical wavelength over the nebula at which this sharp edge is measured ( see below ) are most naturally explained by absorption .
several absorption features are detected in both ly@xmath1 and civ ( fig . 7 ) .
ly@xmath1 absorption is detected in more than 30 spaxels .
we set a lower limit to the size of the main absorber of @xmath84.7@[email protected]@xmath29 or @xmath843@xmath532 kpc@xmath39 .
it is possible that this absorber covers the ly@xmath1 nebula completely . since absorption is detected in civ as well
, this implies that the absorbing gas is highly ionized .
a striking characteristic is that the sharp edge of the ly@xmath1 blue wing ( see fig .
7 ) happens at almost exactly the same wavelength ( [email protected] ) in all spaxels where we have been able to measure it ( @xmath820 ) , revealing very little kinematic structure of the main absorber along the line of sight and accross its whole spatial extension .
_ the ly@xmath1 nebula _ the vimos ly@xmath1+continuum image of mrc 0140 - 257 was created by adding the monochromatic images of the object within the [ 4410 - 4430 ] range .
an adjacent continuum image was subtracted to keep the line emission only .
the resulting ly@xmath1 image is shown in fig . 8 with the vla 8.2ghz radio contours overlaid ( carilli et al .
@xcite ) .
two main spatial components ( @xmath48 and @xmath49 in the figure ) are identified in the image , which are aligned with the radio structures within a few degrees .
we have assumed that the radio core is located between @xmath48 and @xmath49 ( but see 2 ) .
the ly@xmath1 spatial centroids of @xmath48 and @xmath49 are separated by @xmath83.5@xmath29 or 32 kpc .
the maximum extension of the nebula ( @xmath48 and @xmath49 ) is 9@xmath29 or 83 kpc .
it , therefore , extends well beyond the radio structures .
this conclusion is independent of the location of the radio core along the @xmath48-@xmath49 line ( 2 ) .
the faintest ly@xmath1 emission we detect on the vimos data has surface brightness @xmath810@xmath32 erg s@xmath7 @xmath33 arcsec@xmath34 ( 3@xmath30 values ) .
the total ly@xmath1 flux integrated over the nebula is 3.6@xmath510@xmath50 erg s@xmath7 @xmath33 , which , taking aperture corrections into account , is consistent with @xcite .
it corresponds to a luminosity of 2.9@xmath510@xmath51 erg s@xmath7 .
_ continuum morphology _
a continuum image was created by collapsing the data cube across the continuum spectral windows [ 5270 - 5550 ] and [ 5640 - 5800 ] , i.e. rest frame range @xmath8[1450 - 1530 ] and [ 1555 - 1600 ] respectively . for comparison , the optical image of mccarthy , persson & west ( @xcite )
was obtained using a gunn - thuan filter , covering the rest frame range [ 1675 - 1920 ] .. we detect a faint source , which overlaps partially with component @xmath48 ( fig.9 ) . due to the large separation in wavelength between the continuum and ly@xmath1 images
, we can not discard that the spatial shift between the continuum centroid and @xmath48 is partially artificial ( see 2 ) .
although the source is apparently extended in the n - s direction , this is dubious .
the faintest structures to the north are consistent within the errors with the image detection limit . on the other hand ,
there are noise residuals at similar flux levels at different positions accross the vimos field of view .
_ spatially extended emission line spectroscopy _
we show in fig .
10 the spatial maps of the fwhm and @xmath42 of the ly@xmath1 line .
@xmath42 has been measured relative to the ly@xmath1 emission between @xmath48 and @xmath49 .
the erros on @xmath42 are estimated to be @xmath1140 km s@xmath7 .
the errors on the fwhm are in the range 60 - 100 km s@xmath7 .
the main reason for these large relative errors is the uncertainty on the continuum level , due to the noise ( this object is fainter ) and the presence of an underlying broad component ( see below ) .
the line is narrow , with fwhm@xmath3500 km s@xmath7 accross the whole nebula , compared with typical values of hzrg .
the ly@xmath1 spectra integrated over the highest signal to noise spaxels of @xmath48 ( 16 spaxels ) and @xmath49 ( 14 spaxels , see fig . 10 , green lines ) are shown in fig .
11 ( right panels ) . in both spatial components ,
the ly@xmath1 spectral profile is dominated by a strikingly narrow component , with fwhm=250@xmath2850 and @xmath3120 km s@xmath7 for @xmath48 and @xmath49 respectively .
an underlying broad component seems to be also present in both components.500 km s@xmath7 .
] this is most clearly detected when both spectra are added .
the fit to the line profile in the coadded spectrum of @xmath48 and @xmath49 is shown in fig .
12 together with the individual components isolated in the fit .
the underlying broad component has fwhm=1200@xmath28200 km s@xmath7 .
the velocity shift between @xmath48 and @xmath49 is 120@xmath2820 km s@xmath7 , which is also rather low compared with typical values in hzrg ( e.g. mccarthy et al .
@xcite ) . civ and heii are detected in @xmath49 ( fig .
11 , left ) .
two very narrow emission lines are identified at the expected wavelengths of the civ doublet .
we have fitted the lines with no constraints applied .
the resulting doublet consists of two emission lines separated by [email protected] ( consistent within the errors with the theoretical 9.2 at the redshift of the object ) and with flux ratio of @[email protected] , as expected in the optically thin case . taking the errors into account ,
both lines are unresolved ( observed [email protected] and [email protected] respectively ) , with fwhm@xmath3100 km s@xmath7 .
this is in excellent agreement with the width of the ly@xmath1 narrow component .
heii is also detected and is similarly narrow with fwhm=180@xmath2860 km s@xmath7 .
the small fwhm values of the civ and heii lines confirm that ly@xmath1 is intrinsically very narrow ( rather than absorbed ) .
_ ly@xmath1 absorption _
the ly@xmath1 spectral profile does not show clear evidence for absorption .
this is further supported by the large ly@xmath1 ratios .
in @xmath49 ly@xmath1/heii and ly@xmath1/civ are [email protected] and 7.4@xmath281 respectively .
these values are consistent with standard photoionization model predictions for the measured civ / [email protected] ( villar - martn et al . 2007 ) . ly@xmath1/heii and ly@xmath1/civ are @xmath14 6 and @xmath145 respectively for @xmath48 and it is not possible to say whether ly@xmath1 is absorbed .
the values for the integrated spectrum are @xmath2215 and 9@xmath282 respectively which for the measured civ / [email protected] do not imply absorption either .
this is in contradiction with @xcite ( see 3.3 ) .
the three radio galaxies investigated here are associated with giant ( @xmath060 kpc ) ly@xmath1 nebulae .
this was already known for mrc 1558 - 003 and mrc 2025 - 218 , but not for mrc 0140 - 257 .
the total ly@xmath1 luminosities are @xmath8(0.3 - 3.4)@xmath510@xmath6 erg s@xmath7 , within the range of typical values measured for hzrg . the morphologies are varied .
the nebula of one source has a centrally peaked , rounded appearance ( mrc 1558 - 003 ) . in the other two objects it consists of two spatial components .
this bimodal morphology could be a consequence of an obscuring dust structure ( e.g. knopp & chambers @xcite , reuland et al .
@xcite ) . for mrc 2025 - 218
this is supported by the fact that the continuum centroid coincides with a minimum in the ly@xmath1 flux .
@xcite also proposed the existence of a dust lane in mrc 0140 - 257 to explain the near infrared continuum morphology .
the alignment between the ly@xmath1 nebulae and the radio structures is remarkable in mrc 2025 - 217 and mrc 0140 - 257 ( @xmath310@xmath4 ) . in mrc 1558 - 003 ,
the nebula is misaligned by 30@xmath285@xmath4 .
similar values have been measured for other hzrg radio galaxies ( mccarthy , spinrad & van breugel @xcite ) .
based on previous , deeper spectroscopic studies and this work , we conclude that , although the high surface brightness emission tends to be confined within the radio lobes , ly@xmath1 emission is also detected beyond the radio structures in the three objects studied here ( see also villar - martn et al .
@xcite for other examples ) . in one case ( mrc 1558 - 003 )
, the new data reveal ly@xmath1 emission outside any plausible ionization cone and far from the radio structures ( see reuland et al .
@xcite , villar - martn et al .
@xcite for other examples ) .
projection effects might be responsible , since this is a broad line object and this is our favoured explanation ( see 5.3 ) .
alternatively , part of the ly@xmath1 emission might be resonantly scattered or powered by a mechanism not related to the active nucleus , such as young stars ( villar - martn et al .
@xcite ) . unfortunately
, the data are not deep enough to check whether lines other than ly@xmath1 are emitted in these regions .
cooling radiation ( haiman , spaans & quataert @xcite ) is an interesting possibility often discussed in the subject of radio quiet ( e.g. nilsson et al .
@xcite ) and radio loud ly@xmath1 nebulae .
however , the ly@xmath1 surface brightness we measure in regions outside the reach of the ionization cones ( @xmath1410@xmath32 erg s@xmath7 @xmath33 arcsec@xmath34 ) is too high compared with the model predictions ( see villar - martn et al .
@xcite for a more detailed discussion on this issue ; see also dijkstra , haiman & spaans @xcite ) .
interactions between the radio structures ( jet and radio lobes ) and the ambient gas are known to have a profound impact on the kinematic properties of the giant nebulae associated with many hzrg ( 1 ) .
for the three objects studied here , this is obvious only in mrc 2025 - 218 , in the northern region in particular , where the emission lines , ly@xmath1 , heii and civ have fwhm@xmath141000 km s@xmath7 ( 4.2 ) .
no evidence for interactions is found in the southern region , where the radio structures present a sharp bend ( see fig . 4 ) .
if this has been caused by the collision of the radio structures with a dense gaseous region , no clear signature has remained in the gas kinematic properties .
ly@xmath1 is also very broad across the mrc 1558 - 003 nebula .
however , fwhm@xmath01000 km s@xmath7 values are measured also in distant regions from the radio structures . since moreover heii is relatively narrow ( 450 - 650 km s@xmath7 across the nebula , see also villar - martn et al .
@xcite ) , we can not discard that resonance scattering effects are responsible for broadening the ly@xmath1 spectral profile . in mrc 0140 - 257
the nebular emission is dominated by quiescent gas ( 4.3 ) , i.e. , not perturbed by the radio structures .
we can not tell whether the faint ly@xmath1 underlying broad component ( fwhm=1200@xmath28200 km s@xmath7 ) is a consequence of radio / gas interactions or resonance scattering effects .
in fact , the most interesting result for this object is the strikingly quiescent kinematics revealed by the bulk of the emission lines across the nebula ( fwhm@xmath11300 km s@xmath7 and @xmath9120 km s@xmath7 ) .
this characteristic is unique among hzrg .
similar relaxed kinematics is rather extreme also among low @xmath13 radio galaxies , since only a few show such narrow line widths , usually measured in extranuclear regions ( e.g. baum , heckman & van breugel @xcite ) .
@xcite proposed that the extended _ quiescent _ ionized nebulae associated with numerous powerful radio galaxies at different redshifts are in the process of infall .
these authors searched for correlations between several radio and optical / uv asymmetries , and found that ( i ) the quiescent ionized gas has its highest redshift on the side of the nucleus with the brightest and more polarized radio hotspot , and that ( ii ) on the side where the lya emission is brightest relative to the other emission lines and continuum .
they concluded that orientation effects , with the quiescent gas in infall towards the nucleus , is the most natural scenario for explaining these correlations .
our study of mrc 2104 - 242 based on integral field vimos data also suggested that the giant quiescent ly@xmath1 nebula associated with this object could be in the process of infall ( villar - martn et al .
however , due to the uncertainty on the spatial gas distribution , a rotational pattern could not be discarded .
it was our later work ( humphrey et al .
@xcite ) on a larger sample what allowed us to discard rotation and favour the infall interpretation .
we investigate next whether the morphological and kinematic properties of the nebulae studied here are consistent with infall .
_ mrc 1558 - 003 _ [ fig : model ] according to @xcite the giant nebula associated with mrc 1558 - 003 is infalling towards the center . here
we describe an infall model that can explain the observations quite well .
an agn sits in the center of a halo of mass @xmath52 .
a dusty torus surrounding the agn causes it to photoionize a biconical region of the surrounding gas .
the opening angle of each cone is assumed to be @xmath53 .
the cone axis is not perfectly aligned with the line - of - sight ( l.o.s ) , but intersects it at an angle of @xmath54 .
furthermore , the cones are rotated by @xmath55 counterclockwise on the sky . in this scenario ,
one cone is pointing almost directly at the observer , while the other points in the opposite direction .
photoionization and subsequent recombination in the cones converts ionizing radiation emitted by the agn into ly@xmath1 , and thus results in spatially extended fluorescent ly@xmath1 emission @xcite . in this picture ,
infalling gas in the nearest cone will be emitted with a redshift relative to the systemic redshift .
we found that a model in which the gas density increases as @xmath56 , and the gas velocity increases @xmath57 can reproduce most observed properties of the nebula ( fig .
13 ) . in order to match the observations , it is crucial that the emission from the furthest cone is not observed ( see below ) . for a more detailed description of the model ,
the reader is referred to appendix a. according to fig .
13 , our infall model produces the following features : * a rounded and centrally peaked morphology ( _ left panel _ ) .
this is a consequence of the geometry of our model and the increase of the recombination rate ( and thus of fluorescent ly@xmath1 emission ) towards the agn . the innermost contour level in fig .
13 encloses a surface brightness that is @xmath58 times higher than that surrounded by the outermost contour .
these properties are similar to those observed ( fig .
2 ) , within a similar nebular size . in order to obtain a more symmetric appearance , as shown by the data , the axis cone should be closer to the line of sight . this would correspond to a quasar .
the fact that the agn continuum and broad line emission are detected at rest frame optical wavelengths , but are not obvious in the uv rest frame could be explained by the existence of dust obscuring the central regions . *
the largest redshift is observed at the ly@xmath1 centroid ( _ central panel _ ) , as observed in the data ( fig . 3 ,
because the infall velocity increases towards the agn , the redshift of the ly@xmath1 line decreases outwards .
labels in the figure denote the mean velocity shift of the emission line in km s@xmath7 with respect to he centroid , i.e. , the position of maximum ly@xmath1 flux . these values are consistent with the measured velocity offsets .
* the nebular centroid has the largest redshift w.r.t the systemic redshift .
it is not possible to determine the systemic redshift in this object .
however , it is interesting to note that in mrc 1558 - 003 ly@xmath1 is redshifted relative to the main rest - frame uv emission lines ( civ , heii , ciii ] ) both in the spatially integrated spectrum ( rttgering et al .
@xcite ) and at different spatial positions ( villar - martn et al .
@xcite ) . in our scenario ,
the ly@xmath1 emission from the far cone is negligible .
the other emission lines might also be fainter from the more distant cone because of extinction , but these effect would be less pronounced , since they are not affected by resonance scattering effects ( see also humphrey et al .
@xcite ) . as a result ,
the relative contribution from the far cone to the flux of these lines is larger than for ly@xmath1 both in the integrated spectrum and at different projected spatial positions .
since the emission from the more distant cone must be blueshifted relative to the near cone because of the infall pattern , the lines will be blueshifted relative to ly@xmath1 , which is consistent with the observations . to reproduce these two last features ,
the ly@xmath1 emission detected from the furthest cone must be negligible as the expected blueshift of these photons would eliminate the predicted increase of the ly@xmath1 redshift towards the nebular centroid and the blueshift relative to other emission lines .
although radiation blueward of the rest - frame ly@xmath1 frequency is subject to absorption in the igm , the igm at @xmath59 is not opaque enough to completely eliminate the flux from the far cone ( we used the model of dijkstra et al .
@xcite to calculate the impact of the igm on the ly@xmath1 line ) .
alternatively , the flux from the far side could be eliminated by a neutral and dusty spatially extended structure that lies between the two cones .
several studies suggest the existence of such structure ( e.g. humphrey et al .
@xcite , van ojik et al .
@xcite ) .
our model has greater difficulty reproducing the observed ly@xmath1 fwhm ( _ right panel _ , fig .
especially if purely radial infall is considered , the model produces a maximum fwhm of @xmath8500 km s@xmath7 , which is a factor of almost 3 short of what is observed for ly@xmath1 .
the model fwhm shown in fig . 13
is boosted to @xmath60 km s@xmath7 by convolving the spectrum at each pixel with a gaussian with a standard deviation of @xmath61 km s@xmath7 ( @xmath62 is the 1-d velocity dispersion of the halo ) .
this reflects that the infall probably does not occur purely along radial paths .
although inconsistent with the ly@xmath1 measurements , this model reproduces successfully the spatial variation of the heii fwhm observed by ( villar - martn et al .
2003 ) along the radio axis using long slit spectroscopic data . in this direction
, the nebula shows broader fwhm@xmath8750 km s@xmath7 at the spatial centroid and @xmath8500 km s@xmath7 in the outer parts ( villar - martn et al .
2003 ) , consistent with the models .
resonant scattering of ly@xmath1 photons has been ignored in our model because the gas within the cones is highly ionized and likely optically thin to ly@xmath1 ( see appendix a ) . although resonant scattering effects could explain the large ly@xmath1 fwhm values , and its broader profile compared with heii ( 4.1 ) , they would contradict other observational results .
@xcite have performed radiative transfer calculations through neutral collapsing gas clouds and found that resonant scattering does not only broaden the ly@xmath1 line , but that energy transfer from the infalling gas to the ly@xmath1 photons shifts the line to bluer wavelengths . on the contrary , the observations suggest that the ly@xmath1 line is redshifted with respect to other emission lines and probably relative to the system s redshift .
neither does resonant scattering through optically thick , collapsing , gas clouds reproduce the observation that the ly@xmath1 line is redshifted more towards the center .
if resonant scattering does not occur , then it remains to be explained why ly@xmath1 shows such large fwhm values and it is much broader than heii .
an alternative mechanism that can broaden ly@xmath1 relative to other emission lines and could produce a similar velocity pattern as that of infall is outflows ( dijkstra , haiman & spaans @xcite ) . in this scenario ,
a shell propagates outwards from the center , which is thick enough to be self - shielding ( with column densities of hydrogen of 10@xmath63 @xmath33 ) .
all material is highly ionized , except inside the shell that is moving outwards . in this case ,
ly@xmath1 photons basically bounce off the shell as if it were a mirror , scattering back and forth between the two expanding mirrors until they have been doppler boosted far enough for them to be able to simply propagate through the shell ( each time a lya photon scatters off the shell , it picks up a doppler shift from the shell ) .
this process will broaden ly@xmath1 , not affecting the other emission lines .
detailed modeling would be necessary to test whether outflows can reproduce all other observed properties of the nebula as infall does .
however , since outflows can not explain the radio / optical asymmetries seen in mrc 1558 - 003 ( humphrey et al .
@xcite ) , which on the other hand , are successfully explained with infall , here we will not explore the outflows scenario further . in conclusion
, our infall model can explain successfully the morphology , size , surface brightness distribution and the velocity field of the ly@xmath1 nebula associated with mrc 1558 - 003 .
it can also explain why ly@xmath1 is redshifted relative to other emission lines and the fwhm values of the non resonant heii line .
the infall scenario is also consistent with our previous results which imply that the _ quiescent _ nebulae in many hzrg ( and mrc 1558 - 003 in particular ) are infalling ( humphrey et al .
@xcite ) . on the other hand , this model fails to explain the large ly@xmath1 fwhm values .
a mechanism which might not be connected with the infall process could be responsible for the line broadening ) , while keeping the infall pattern and the heii fwhm intact . _
mrc 2025 - 218 _ it is not possible to investigate the kinematic pattern of the quiescent gas , since the data do not allow to isolate its emission in the ne component .
_ mrc 0140 - 257 _ it is not possible to disentangle how the quiescent gas is moving in this object .
the relaxed kinematics is not suggestive of outflows .
mrc 0140 - 257 does not follow the trend found by humphrey et al .
( @xcite ) in powerful radio galaxies at different @xmath13 , such that the quiescent gas has the highest redshift at the side of the brightest and more polarized radio hot spot .
the authors interpret these results as evidence for infall . on the contrary , the highest redshift in mrc 0140 - 257
is measured in @xmath48 , at the side of the fainter ( northern ) radio hotspot .
this discrepancy could be due to three different reasons : ( i ) the orientation diagnostics based on the radio properties is not valid ( ii ) the gas is not in infall ( iii ) our radio - optical registration ( 2 ) is incorrect . if the radio core was located on @xmath48 or @xmath49 ( rather than between the two ly@xmath1 components ) , we would have no information on the redshift on one side of the nucleus .
the interpretation of infall by humphrey et al .
( @xcite ) is based on the fact that the radio asymmetries in their sample provide a powerful diagnostic tool to constrain the orientation of the radio source ( and the ly@xmath1 nebula as a consequence ) .
one possible explanation for the mrc 0140 - 257 discrepancy is that in this particular case the radio properties are not reliable indicators of the object orientation .
this is actually suggested by the fact that the brightest ( southern ) hot spot shows the lowest polarization ( carilli et al .
@xcite ) , contrary to what we would expect if both flux and polarization asymmetries where mostly determined by orientation .
the orientation diagnostics based on the radio properties would fail , for instance , when the radio axis is very close to the plane of the sky . for mrc 0140 - 257 , on the contrary , the detection of a radio core ( carilli et al .
@xcite ) rather suggests that the angle between the radio axis and the plane of the sky is not negligible .
the diagnostics would also fail if the radio asymmetries are a consequence of environmental effects ( e.g. mccarthy , van breugel & kapahi @xcite ) , rather than orientation .
this is possibly the case in mrc 0140 - 257 . on one hand , the brightest
, less polarized hotspot is closest to the nucleus , and the ly@xmath64 emission on that side of the nucleus ( adopting our present radio - optical registration ) is also brightest .
this is what we expect from mccarthy , van breugel & kapahi ( @xcite ) environmental scenario . on the other hand ,
the southern ( brighter ) radio structure consists of two radio hot spots ( fig.8 ) . if this is due to a interactions with the gas ( e.g. carilli et al .
@xcite ) , this process would enhance the radio emission and the higher brightness of the southern radio structure would therefore be a consequence of environmental effects .
alternatively , it is possible that the ly@xmath1 nebula is not infalling towards the center . @xmath48 and @xmath49 could be two objects rotating around a common center of mass ( e.g. de breuck et al .
if this is located between @xmath48 and @xmath49 , we estimate a dynamical mass of @xmath03@xmath65 m@xmath17 for a radius @xmath66=16 kpc and a projected rotation velocity of 60 km s@xmath7 ( i.e. half of the @xmath42 value between @xmath48 and @xmath49 ) .
this mass value is at least a factor of 100 smaller than the stellar masses inferred for other hzrg ( seymour et al .
@xcite , villar - martn et al .
@xcite ) or for massive , early - type galaxies at @xmath22 in deep , ir surveys ( e.g. daddi et al .
mrc 0140 - 257 could be an unusually small mass hzrg , a possibility that is also suggested by the small fwhm values measured in @xmath48 and @xmath49 .
it would be essential to obtain rest frame h and k magnitudes ( e.g. seymour et al .
@xcite , villar - martn et al .
@xcite ) to make a proper estimation of the stellar mass .
if , on the contrary , this is progenitor of a massive elliptical galaxy , as one expects for a powerful hzrg ( e.g. mclure et al .
@xcite ) , the rotation plane of @xmath48 and @xmath49 must almost coincide with the plane of the sky .
ly@xmath1 absorption has definitively been detected in mrc 2025 - 218 , spatially extended for @xmath1443@xmath532 kpc@xmath39 .
the absorber could therefore be larger in size than the optical size of the host ( if already formed ) underlying galaxy .
civ is also absorbed and , thus , the absorbing gas is highly ionized . as in this object ,
very little kinematic structure along the line of sight was also found by wilman et al .
( @xcite ) across the giant ly@xmath1 radio quiet nebula lab-2 discovered by steidel et al .
( @xcite ) . a more detailed analysis of the absorbers associated with mrc 2025 - 218 will be presented in humphrey et al .
( 2007 , in prep . )
the morphological and spectroscopic properties of the giant ( @xmath060 kpc ) ly@xmath1 nebulae associated with three radio galaxies ( mrc 1558 - 003 , mrc 2025 - 218 and mrc
0140 - 217 ) at @xmath22.5 have been investigated using vimos integral field spectroscopic data on vlt .
the three objects are associated with giant ( @xmath060 kpc ) ly@xmath1 nebulae .
this was already known for mrc 1558 - 003 and mrc 2025 - 218 , but not for mrc 0140 - 257 .
the morphologies are varied . in one source
the nebula has a centrally peaked , rounded appearance ( mrc 1558 - 003 ) , while in two objects it consists of two spatial components ( mrc 2025 - 218 , mrc 0140 - 257 ) .
the total ly@xmath1 luminosities are in the range ( 0.3 - 3.4)@xmath510@xmath6 erg s@xmath7 .
the ly@xmath1 spectral profile shows strong variation through the nebulae , with fwhm values in the range @xmath8400 - 1500 km s@xmath7 and velocity shifts @xmath9120 - 600 km s@xmath7 .
kinematic disturbance induced by the radio structures plays a clear important role in mrc 2025 - 218 .
most spectroscopic and morphological properties of the giant nebula associated with mrc 1558 - 003 are successfully explained by a scenario such that the giant nebula is collapsing towards the center in the potential well of a 5@xmath510@xmath67 m@xmath17 halo .
this is consistent with our previous conclusion that the giant nebulae associated with this and other hzrg are in infall ( humphrey et al .
@xcite ) . on the other hand
, this model has problems to reproduce the large ly@xmath1 fwhm values ( @xmath01000 km s@xmath7 ) , which might be a consequence of a mechanism unrelated to the infall process .
we have discovered a giant ( @xmath875 kpc ) ly@xmath1 nebulae associated with a radio galaxy at @xmath102.64 .
it is very closely aligned with the radio axis and it extends well beyond the radio structures .
it is characterized by a strikingly relaxed kinematics ( fwhm@xmath11300 km s@xmath7 and @xmath12120 km s@xmath7 ) , unique among hzrg .
this object might have an unusually small dynamical mass ( @xmath143@xmath510@xmath68 m@xmath69 ) for a hzrg , although this needs further investigation with other type of data .
ly@xmath1 absorption is definitively detected in mrc 2025 - 218 .
the absorbing screen has a spatial extension of 43@xmath532 kpc@xmath39 at least .
it is highly ionized ( civ is also absorbed ) and shows a remarkable velocity coherence ( @xmath11100 km s@xmath7 ) across its spatial extension .
thanks to an anonymous referee for useful comments on the paper .
we thank chris carilli and laura pentericci for providing the radio maps presented in this paper .
the work of mv - m and rgd has been supported by the spanish ministerio de educacin y ciencia and the junta de andaluca through the grants aya2004 - 02703 and tic-114 .
sfs thanks the spanish plan nacional de astronoma , program aya2005 - 09413-c02 - 02 ( mec ) and the plan andaluz de investigacin of junta de andaluca , research group fqm322 .
ah acknowledges support from a unam postdoctoral research fellowship .
thanks to joel vernet for producing figures 3 and 8 .
barthel p. , 1989 ,
apj , 336 , 606 baum s. , heckman t. , van breugel w. , 1990 , apjs , 74 , 389 carilli c. , rttgering h. , van ojik r. , miley g. , van breugel w. , 1997 , apjss , 109 , 1 daddi e. et al .
2004 , apj , 600l , 127 de breuck c. , downes d. , neri r. , van breugel w. , reuland m. , omont a. , ivison r. , 2005 , a&al,430l,1 dijkstra m. , haiman z. , spaans m. , 2006 , apj , 649 , 14 dijkstra , m. , lidz , a. , wyithe , s , mnras , 2007 , in press ( astro - ph/0701667 ) eales s. , rawlings s. , dickinson m. , spinrad h. , hill g. , lacy m. , 1993 , apj , 409 , 578 eales s. , rawlings s. , 1996 , apj , 460 , 68 haiman z. , rees m. j. 2001 , apj , 556 , 87 haiman z. , spaans m. , quataert e. , 2000 , apjl , 537 , 5 humphrey a. , 2004 , phd thesis , univ . of hertfordshire humphrey a. , villar - martn m. , fosbury r. vernet j. , di serego alighieri s. , 2006 , mnras , 369 , 1103 humphrey a. , villar - martn m. , fosbury r. vernet j. , di serego alighieri s. , 2007 , mnras , 375 , 705 knopp g. , chambers k. , 1997 , apjs , 109 , 367 kurk j. , pentericci l. , rttgering h , miley g. , 2002 , revista mexicana de astronoma y astrofsica ( serie de conferencias ) , vol . 13 , pp . 191 - 195 large m. , mills b. , little a. , crawford d. , sutton j. , 1981 , mnras , 194 , 693 larkin j. et al . 2000 , apjl , 533 , 61 le fvre o. et al . , 2003 , proc .
of the spie , 4841 , p. 1670
, masanori & moorwood eds .
maxfield l. , spinrad h. , stern d. , dey a. , dickinson m. , 2002 , aj , 123 , 2321 mccarthy p.j .
, spinrad h. , dickinson m. , van breugel w. , liebert j. , djorgovski s. , eisenhardt p. , 1990a
, apj , 365 , 487 mccarthy p.j . , kapahi v. , van breugel w. , subrahmanya , c. , 1990b , aj , 100 , 1014 mccarthy p.j .
, van breugel w. , kapahi v. , subrahmanya , c. , 1991 , aj , 102 , 522 mccarthy p.j .
, van breugel w. , kapahi v. , 1991 , apj , 371 , 478 mccarthy p. , persson s. & west s. , 1992 , apj , 382 , 52 mccarthy p. , 1993 ,
ara&a , 31 , 639 mccarthy p.j . , spinrad h. , van breugel w. , 1995 , apjss , 99 , 27 mccarthy p. baum s. , spinrad h. , 1996 , apjs , 106 , 281 mclure r. , kukula m. , dunlop j. , baum s. , odea c. , hughes d. , 1999 , mnras , 308 , 377 nilsson k. , fynbo j. , mller p. , sommer - larsen j. , ledoux c. , 2006 , a&al , 452 , 23 pentericci l. , rttgering h. , miley g. , mccarthy p. , spinrad h. , van breugel w. , macchetto f. , 1999 , a&a , 341 , 329 pentericci l. , van reeven w. , carilli c. , rttgering h. , miley g. , 2000 , a&as , 145 , 121 pentericci l. , mccarthy p. , rttgering h. , miley g.,van breugel w. , fosbury r. , 2001 , apjss , 135 , 63 reuland m. et al .
, 2003 , apj , 592 , 755 reuland m. et al .
, 2007 , aj , in press ( astro - ph/0702753 ) rttgering h. , van ojik , r. , miley g. , chambers k. , van breugel w. , de koff s. , 1997 , a&a , 326 , 505 snchez s.f .
, 2004 , an , 325 , 167 snchez s.f . , 2006 , an , 327 , 850 sandage a. , tamman g. , saha a. , reindl b. , macchetto f. , panagia n. , 2006 , apj , 653 , 843 seymour n. , stern d. , de breuck d. , vernet j. , rettura a. , dickinson m. , dey a. , eisenhardt p. et al . , 2007 ,
apjs , in press ( astro - ph/0703224 ) steidel c. , adelberer k. , shapley a. , pettini m. , dickinson m. , giavalisco m. , 2000 , apj , 532 , 170 tadhunter c. , fosbury r. , quinn p. , 1989
, mnras , 240 , 225 van ojik r. , rttgering h. , carilli c. , miley g. , bremer m. , macchetto f. , 1996 , a&a , 313 , 25 van ojik r. , rttgering h. , miley g. , hunstead r. , 1997 , a&a , 317 , 358 vernet j. , fosbury r. , villar - martn m. , cohen m. , cimatti a. , di serego alighieri , goodrich r. , 2001 , a&a , 366 , 7 villar - martn m. , fosbury r. , binette l. , tadhunter c. , rocca - volmerange b. , 1999 , mnras , 351 , 47 villar - martn m. , vernet j. , di serego alighieri s. , fosbury r. , humphrey a. , pentericci l. , 2003 , mnras , 346 , 273 villar - martn m. et al .
2006 , mnras , 359 , 5 villar - martn m. , humphrey a. de breuck c. , fosbury r. , binette l. , vernet j. , 2007 , mnras , 375 , 1299 wilman r. , gerssen j. , bower r. , morris r. , bacon r. , de zeeuw p. , davies r. , 2005 , nature , 436 , 227 [ app ]
for each position @xmath70 the spectrum was obtained by adding the emission from each cell @xmath71 inside the cone along that line of sight ( l.o.s ) .
the emissivity for a cell was calculated by assuming photoionization equilibrium .
this implies the volume emissivity at radius @xmath66 scales as @xmath72 , where @xmath73 is the number density of hydrogen nuclei at radius @xmath66 , and we used that the gas is ( almost ) fully ionized . in the frame of the cell ( which itself is falling toward the center )
, the ly@xmath1 line is broadened by thermal motions of the atoms in the gas , which results in a voigt profile .
the bulk motions of the cell then cause a doppler shift of the ly@xmath1 line .
the total spectrum at @xmath70 on the sky , denoted by @xmath74 , is obtained by adding the contributions from all cells . to be more precise ,
the spectrum at @xmath70 is given by @xmath75 where @xmath76 if the cell is inside the cone , and otherwise @xmath77 .
furthermore , @xmath78 is the voigt function , where @xmath79 @xmath80 @xmath81 . here
@xmath82 is the einstein a - coefficient for the transition , @xmath83 is the thermal velocity of the hydrogen atoms in the gas , given by @xmath84 , where @xmath85 is the boltzmann constant , @xmath86 the gas temperature , and @xmath87 the proton mass . ] .
the gas density and velocity profiles were assumed to be of the forms @xmath88 , @xmath89 .
the infall velocity was assumed to be equal to the circular velocity of the halo at the virial radius to prevent both fields from diverging , both @xmath90 and @xmath91 were kept constant at radii less than @xmath92 times the virial radius ( which corresponds to @xmath93 kpc ) .
these fields were chosen to give the best agreement with the observations .
note that these parameters were chosen to produce a good fit to the observations . to enhance the fwhm at each location
@xmath70 were convolved with a gaussian of width @xmath61 km s@xmath7 .
equation [ eq : spectrum ] is valid in the limit where the cloud is transparent to all ly@xmath1 photons . in our model
the ly@xmath1 from the furthest cone is not observed and we only need to consider ly@xmath1 emission from the nearest cone . for these photons
we find that the optically - thin approximation is very reasonable : all ly@xmath1 photons are emitted in the cone where the gas is highly ionized .
for the cone orientation of our model , each ly@xmath1 photon never leaves the cone on its way to the observer . the total optical depth for a given ly@xmath1 photon
is reduced even further by velocity gradients along a l.o.s . since all these photons
are emitted predominantly redward of the ly@xmath1 resonance , scattering in the igm is not important . | the morphological and spectroscopic properties of the giant ( @xmath060 kpc ) ly@xmath1 nebulae associated with three radio galaxies at @xmath22.5 ( mrc 1558 - 003 , mrc 2025 - 218 and mrc 0140 - 257 ) have been investigated using integral field spectroscopic data obtained with vimos on vlt .
the morphologies are varied .
the nebula of one source has a centrally peaked , rounded appearance . in the other two objects , it consists of two spatial components .
the three nebulae are aligned with the radio axis within @xmath330@xmath4 .
the total ly@xmath1 luminosities are in the range ( 0.3 - 3.4)@xmath510@xmath6 erg s@xmath7 .
the ly@xmath1 spectral profile shows strong variation through the nebulae , with fwhm values in the range @xmath8400 - 1500 km s@xmath7 and velocity shifts @xmath9120 - 600 km s@xmath7 .
we present an infall model that can explain successfully the morphology , size , surface brightness distribution and the velocity field of the ly@xmath1 nebula associated with mrc 1558 - 003 .
it can also explain why ly@xmath1 is redshifted relative to other emission lines and the fwhm values of the non resonant heii line .
this adds further support to our previous conclusion that the _ quiescent _ giant nebulae associated with this and other high redshift powerful radio galaxies are in infall .
a problem for this model is the difficulty to reproduce the large ly@xmath1 fwhm values , which might be consequence of a different mechanism .
we have discovered a giant ( @xmath885 kpc ) ly@xmath1 nebula associated with the radio galaxy mrc 0140 - 257 at @xmath102.64 .
it shows strikingly relaxed kinematics ( fwhm@xmath11300 km s@xmath7 and @xmath12120 km s@xmath7 ) , unique among high @xmath13 ( @xmath142 ) radio galaxies .
[ firstpage ] galaxies : active ; galaxies : high redshift ; galaxies : individual : mrc 1558 - 003 , mrc 2025 - 218 , mrc 0140 - 257 |
in this paper , we present a number of results concerning a dynamical version of coalescing random walks , which was recently introduced in @xcite .
our results concern times of hausdorff dimension less than one where the system of coalescing walks behaves exceptionally .
the results are analogous to and were motivated by the model of dynamical percolation and its exceptional times @xcite . in this section ,
we define the basic model treated in this paper , which we call the dynamical discrete web ( ddw ) , recall some facts about dynamical percolation , and then briefly describe our main results .
the justification for calling this model a discrete web is that there is a natural scaling limit , which is one of our main motivations for analyzing the discrete web ( as it is in @xcite ) ; we also discuss in this section that scaling limit , which is a dynamical version of the brownian web ( see @xcite ) .
a paper is in preparation @xcite on the construction of that model , which is closely related to the brownian net of sun and swart @xcite .
we note that conjectures conerning ways to construct scaling limits of dynamical percolation ( in two - dimensional space ) appear in @xcite .
we further note that exceptional times for dynamical versions of random walks in various spatial dimensions have been studied in @xcite and elsewhere , but these are quite different from the random walks of the ddw , as we note in subsection [ mainresults ] below .
let @xmath4 for @xmath5 denote a simple symmetric random walk on @xmath6 starting at @xmath7 , i.e. at @xmath8 at @xmath9 .
( for real @xmath10 , we set @xmath11)$ ] , where @xmath12 $ ] denotes the integer part of @xmath13 . ) if we also consider other simple symmetric random walks starting from arbitrary points on the even space - time sublattice @xmath14 , which are independent of each other except that they coalesce when they meet , that is the system of ( one - dimensional ) coalescing random walks that is closely related to the one - dimensional ( discrete time ) voter model ( see @xcite ) and may be thought of as a one plus one dimensional directed percolation model .
the percolation structure is highlighted by defining @xmath15 for @xmath16 to be the increment between times @xmath17 and @xmath18 of the random walker at location @xmath19 at time @xmath17 .
these bernoulli variables are symmetric and independent and the paths of all the coalescing random walks can be reconstructed by assigning to any point @xmath20 an arrow pointing from @xmath20 to @xmath21 and considering all the paths starting from arbitrary points in @xmath22 and following the arrows .
we note that there is also a set of dual ( or backward ) paths defined by the same @xmath15 s with arrows from @xmath23 to @xmath24 .
the collection of all dual paths is a system of backward ( in time ) coalescing random walks that do not cross any of the forward paths . the ddw is a very simple stochastic process @xmath25 in a new dynamical time parameter @xmath0 whose distribution at any deterministic @xmath0 is exactly that of the static coalescing random walk model just described .
specifically , let @xmath26 be a family of independent continuous time cadlag markov processes with state space @xmath27 and rate @xmath28 for changing state in either direction , with the initial condition that @xmath29 is a family of independent bernoulli random variables with @xmath30 .
although this dynamical version of coalescing random walks sounds quite trivial at first hearing , it turns out that it can have interesting behavior at exceptional values of the dynamical time parameter s. this is a feature that it shares in common with dynamical percolation .
static percolation models are defined also in terms of independent bernoulli variables @xmath31 , indexed by points @xmath32 in some @xmath33-dimensional lattice , which in general are asymmetric with parameter @xmath34 .
there is a critical value @xmath35 when the system has a transition from having infinite clusters ( connected components ) with probability zero to having them with probability one .
it is expected that at @xmath36 there are no infinite clusters and this is proved for @xmath37 and for high @xmath33 ( see , e.g. , @xcite ) . in dynamical percolation one extends @xmath31 to time varying functions @xmath38 , as in the case of coalescing walks , except that the transition rates for the jump processes @xmath38 are chosen to have the critical asymmetric @xmath39 distribution to be invariant .
the question raised in @xcite was whether there were exceptional times when an infinite cluster ( say , one containing the origin ) occurs , even though this does not occur at deterministic times .
this was answered negatively in @xcite for large @xmath33 and , more remarkably , was answered positively by schramm and steiff for @xmath37 in @xcite , where they further obtained upper and lower bounds on the hausdorff dimension ( as a subset of the dynamical time axis ) of these exceptional times .
we apply in this paper the approaches used for dynamical percolation to the dynamical discrete web .
although we restrict attention to one - dimensional random walks whose paths are in two - dimensional space - time and hence analogous to @xmath37 dynamical percolation , by considering different possible exceptional phenomena , we use both the high @xmath33 and @xmath37 approaches of @xcite .
a natural initial question was whether there might be exceptional dynamical times @xmath0 for which the walk from the origin @xmath40 is transient ( say to @xmath41 ) .
our first main result ( see theorem [ th1 ] in section [ tameness ] below ) , modeled after the high-@xmath33 dynamical percolation results , is that there are no such exceptional times .
as we explain in remark [ remtame ] in section [ tameness ] , a small modification of the proof of theorem [ th1 ] shows that there are also no exceptional times where some pair of walks avoids eventually coalescing .
our other two main results are modelled after the @xmath37 dynamical percolation results .
one of them ( see theorem [ violation ] below ) concerns a kind of violation of the central limit theorem , or more accurately a kind of weak subdiffusivity , by the random walk @xmath40 for exceptional dynamical times @xmath0 ; namely , that @xmath42 for all @xmath43 .
the other ( see theorem [ dimension ] ) gives upper and lower bounds on the hausdorff dimension of these exceptional times , that depend nontrivially on the constant @xmath3 so that the dimension tends to zero ( respectively , one ) as @xmath44 ( respectively , @xmath45 ) .
this is strikingly in contrast with the dynamical random walks of @xcite where there are no exceptional times for which the law of the iterated logarithm fails . to explain why the walks of @xcite can behave so differently from those of the discrete web , we note that a single switch in the former case affects only a single increment of the walk while some switches in the discrete web change the path of the walker by a `` macroscopic '' amount , as discussed in the next subsection on scaling limits see also figure 1 where switching has changed one of the paths macroscopically . by an obvious symmetry argument
, there are also exceptional dynamical times @xmath0 for which @xmath46 .
one may ask whether there are exceptional @xmath0 for which @xmath47 .
as discussed in remark [ twosided ] below , it can be shown , at least for small @xmath3 , that there are no such exceptional times .
the case of large @xmath3 is unresolved .
[ fixeds ] be the random walk at dynamical time @xmath0 starting from @xmath48 .
this graph , with @xmath13 the horizontal and @xmath49 the vertical coordinate , represents simultaneously the family of functions @xmath50 ( @xmath51 ) .
the lowest path , say for @xmath13 greater than about @xmath52 , differs `` macroscopically '' from the others.,title="fig : " ] there is a natural scaling limit of the ( static ) coalescing random walks model , the brownian web ( see @xcite ) .
here one does a usual diffusive scaling in which the random walk time @xmath13 is scaled by @xmath53 , and space by @xmath54 so that the random walk path starting from @xmath55 $ ] at time @xmath56 $ ] scales to a brownian motion starting from @xmath57 at time @xmath58 .
the collection of all random walk paths from all space - time starting points scales to a collection of coalescing brownian motion paths starting from all points of continuum space - time . now taking the rate of switching to be of order @xmath59 , rescaling time and space respectively by @xmath53 and @xmath54 , and then letting @xmath60 go to @xmath8 leads to a nontrivial limit @xmath61 , the dynamical brownian web .
the idea of taking a scaling limit of the dynamical discrete web to obtain a dynamical continuum model is a natural one , which is at the heart of @xcite , although their approach appears to be somewhat different than the one we had already been taking .
both approaches are closely related to the brownian net construction of sun and swart @xcite as will be extensively explored in @xcite .
as we shall discuss in the next subsection , our approach is based on the construction of a certain poissonian marking of special space - time points of the ( static ) brownian web .
these are the so - called @xmath62 points where a single brownian web path enters the point from earlier times and then two paths leave to later times , one to the left and one to the right with exactly one of the those two paths the continuation of the path from earlier time and the other one `` newly - born '' ; see figure 2 .
neither the idea of doing a poissonian marking of special points for the brownian web nor the idea of using those marked points to construct a scaling limit of a dynamical discrete model is completely new .
in particular , we note that a different type of marking ( of @xmath63 points ) was used in @xcite to study the scaling limits of noisy voter models . also the idea of using
marked double points of @xmath64 to construct the scaling limit of two - dimensional dynamical percolation is discussed in @xcite .
one motivation for the proposed marking in the @xmath64 context was the analogy with markings of @xmath62 as well as of @xmath63 points of the brownian web . in the dynamical brownian web @xmath65 , one can also consider exceptional dynamical times @xmath0 where the path @xmath66 starting from the origin at continuous time @xmath9 behaves differently than an ordinary brownian motion path .
the results of @xcite are very similar to those of this paper for the discrete web .
indeed , in some respects , the proofs are simpler since calculations with brownian motions are often easier than those with random walks .
there is however one substantial complication , which is the main focus of @xcite and the reason we do not present the dynamical brownian web exceptional time results already in this paper .
that complication is the actual construction of the dynamical brownian web a construction that is considerably less trivial than that of the dynamical discrete web , as we explain in the next subsection .
[ changedirections ] is the vertical and @xmath49 the horizontal coordinate .
if a @xmath67 ( left ) point of the original web ( left side of the figure ) is marked to switch at some @xmath68 $ ] , then the direction of that @xmath62 point in the web @xmath69 is changed at @xmath70 so that it becomes a @xmath71 ( right ) point as on the right side of the figure with the incoming path joined to the rightmost of the two paths starting from that point.,title="fig : " ] at the discrete level the scaling is chosen in such a way that between the dynamical times @xmath8 and @xmath0 , in a macroscopic box ( i.e. , one with size of order @xmath72 in the original lattice ) , the number of arrows that change direction will be of order @xmath53 .
the situation can be simplified by focusing on switchings with `` macroscopic '' effects ( i.e. , switchings that will lead to a macroscopic alteration of a walker s trajectory in the initial web @xmath73 ) .
a priori , one should also consider combinations of switchings that have macroscopic effects , but it turns out ( this will be proved rigorously in @xcite ) that the probability of macroscopic effects from switching two or more arrows is negligible compared to switching single arrows , and can be neglected .
there is a natural way of characterizing those critical switchings .
for example , let us consider the forward ( rescaled ) path @xmath74 starting from the origin in @xmath73 and assume that the arrow located at some @xmath75 is orginally oriented to the left .
now we ask whether a switching of this single arrow will alter the path in such a way that the altered path will be to the right of @xmath76 , where @xmath77 are both positive macroscopic quantities .
this will happen if and only if the backward path @xmath78 starting from @xmath79 hits @xmath74 at time @xmath13 ( more precisely , hits @xmath80 at time @xmath81 ) .
more generally , the critical arrows leading to similar alterations are the `` contact '' points between @xmath74 and the backward path @xmath78 at which a switching occurs on @xmath82 $ ] .
but it is now fairly easy to see what the statistics of such a set of points are .
in fact , let @xmath83 be the random variable counting the number of such switchings up to the macroscopic time @xmath13 .
the distribution of @xmath83 is simply given by : @xmath84 where @xmath85 are i.i.d .
bernoulli random variables with @xmath86 , which is @xmath87 as @xmath88 . as @xmath88
, @xmath89 converges to the `` local time '' @xmath90 of the forward brownian path @xmath91 starting from the origin along a backward brownian path @xmath92 starting on the right of the path @xmath93 ( the joint distribution of @xmath93 and @xmath92 is analysed in @xcite and this `` local time '' will be defined precisely in @xcite ) .
further , it is a standard fact that @xmath94 converges to a poisson process .
hence , @xmath83 will converge to a poisson process run by the random clock @xmath95 .
in other words , this set of points will consist of a two - dimensional ( @xmath13 and @xmath0 ) poisson point process with intensity measure @xmath96 , where @xmath97 is lebesgue measure and @xmath98 is the local time measure ( note that @xmath99 will be a locally finite measure so that the poisson process is well defined ) . so far , we have only selected the critical switchings inducing a specific type of macroscopic effect .
namely , the ones altering the path @xmath93 in such a way that a point originally on one side of @xmath93 will be on the opposite side after switching occurs . but in order to select all the critical arrows leading to any kind of macroscopic changes , we should not only consider a poisson process run by the local time of a single forward path @xmath93 against a single backward path @xmath92 , but rather a poisson process run by the `` local time of the entire forward web along the entire backward web '' multiplied by the intensity @xmath100 .
in other words , the set of marked points will be a three - dimensional poisson point process with intensity measure @xmath101 , where @xmath97 is lebesgue measure ( in the variable @xmath0 ) and @xmath102 is the local time measure of the forward web along the backward web . would not be a locally finite measure i.e. , the set of marked points in space - time is actually dense in @xmath103 .
however , like what is presented in @xcite ( see p. 11
there ) , one can add an extra coordinate and lift @xmath102 to be a @xmath104-finite measure , or equivalently approximate @xmath102 by a sequence of locally finite measures @xmath105 , do the markings using @xmath105 , and then let @xmath106 . ]
since the @xmath62 points of the continuum web are precisely those at which a forward and a backward path meet ( see , e.g. , @xcite ) , the measure @xmath107 will be supported by this set of points . from our previous description of them , it should be clear that each @xmath62 point has a preferred left or right `` direction '' .
for example , a @xmath71 ( right ) point is one for which the continuing path ( coming in from earlier time ) is to the right of the `` newly - born '' path .
hence , at the continuum level , the analog of an arrow switching will simply be a change of direction of all marked @xmath62 points ( see figure 2 ) .
the web @xmath69 at time @xmath108 will be `` simply '' deduced from @xmath109 by switching the direction of all marked @xmath62 points whose @xmath0-coordinate is in @xmath110 $ ] .
a last comment concerns the nature of the dependence of the two continuum paths @xmath111 and @xmath112 .
these turn out to be a pair of `` sticky '' brownian motions , which are independent except when they touch each other .
this is one of the major observations in @xcite ; we give a brief derivation of this fact in section [ pairsofpaths ] by analyzing pairs of paths in the discrete setting to see what must occur in the continuum scaling limit . in section [ tameness ] , we state our main theorem about tameness ; i.e. , that there are no exceptional dynamical times when the random walkers are transient .
we also give there some other results about tameness in two extended remarks one about non - coalescence and the other about two - sided bounds of order @xmath113 . then in section [ existence ] , we show that there are exceptional dynamical times when the walkers are ( weakly ) subdiffusive i.e. , have one - sided bounds of order @xmath113 . in section [ hausdorff ]
we derive upper and lower bounds on the hausdorff dimension of the set of such exceptional dynamical times .
some estimates for random walks that are needed for our arguments are given in appendix [ appendix ] .
the dynamics can be described , equivalently to the definition in section [ intro ] , in the following manner . the initial configuration is set to @xmath115 , but now we place independent poisson clocks at each site @xmath20 that ring at rate @xmath116 . every time clocks ring we toss independent fair coins to decide on the values of @xmath117 after the ring .
statistically the two descriptions are equivalent .
the motivation for this second description is that it leads to a useful representation of the interaction between the discrete webs at different dynamical times @xmath0 and @xmath118 . in particular , let @xmath119 and @xmath120 be the walks starting at @xmath7 and defined for times @xmath121 , belonging to @xmath25 and @xmath114 . if @xmath122 , then @xmath123 and @xmath124 are independent since the directions of the arrows at two distinct sites are independent . on the other hand , if @xmath125 , then the next steps of the two walks are now correlated .
if the clock at the site @xmath126 did not ring on @xmath127 , then the two paths will coincide at time @xmath128 .
if it rang at least once , then with probability @xmath129 they will coincide , and with probabily @xmath129 they wo nt .
let us now define inductively a sequence of pairs of stopping times @xmath130 with @xmath131 and : @xmath132 on the interval of integer time @xmath133 $ ] , the paths @xmath134 coincide and at time @xmath135 they decide to separate with probability @xmath129 . in other words , from time @xmath136 , the walkers @xmath137 move independently until the next meeting time @xmath138 .
hence , if we skip the intervals of time @xmath139 , @xmath140 behave as two independent random walks @xmath141 , while if we skip the intervals @xmath142 , the two walks coincide with a single random walk @xmath143 . furthermore , since @xmath143 is constucted from the arrow configuration at different sites than the ones used to construct @xmath141 , it is independent of @xmath141 ; and @xmath144 are i.i.d .
random variables with @xmath145 .
now , skipping the intervals @xmath139 corresponds to making the random time change @xmath146 with @xmath147 , and 1 .
@xmath148 , 2 .
@xmath149 , while skipping @xmath142 corresponds to making the time change @xmath150 ; i.e. , @xmath151 where @xmath152 are three independent standard random walks . in the following @xmath153 ,
distributed as @xmath137 , will be referred to as a pair of _ sticky random walks_. [ varyings ] fixed , this graph represents @xmath154 starting from @xmath48 ( with @xmath155 and @xmath156).,title="fig : " ] time and space are respectively rescaled by @xmath53 and @xmath157 and the rate of switching is taken as @xmath158 . from the previous section ,
the pair of rescaled processes @xmath159 is statistically equivalent to @xmath160 @xmath161 where @xmath162 are three independent rescaled random walks , and @xmath163 with 1 .
@xmath164 2 .
@xmath165 3 .
@xmath166 are i.i.d random variables taking values in @xmath167 , with @xmath168 . as @xmath88
, @xmath169 converges in distribution to three independent brownian motions @xmath170 ; and @xmath171 converges in distribution to the local time @xmath172 at the origin of @xmath173 .
moreover , as a consequence of the law of large numbers , if we take @xmath174 , with @xmath175 of order @xmath176 , then @xmath177 converges to @xmath178 .
hence , it should come as no surprise that @xmath179 converges to @xmath180 where now @xmath181 , which is identical in distribution to a pair of sticky brownian motions with stickiness parameter @xmath182 ( see , e.g. , [ ss06 ] ) .
we note that for small @xmath60 , the location @xmath183 , of the path starting from some @xmath184 is , for fixed @xmath13 , quite discontinuous in @xmath0 see figure 3 .
[ th1 ] almost surely , all the paths are recurrent for every @xmath0 . in section 3 of @xcite , it is proved that for any homogeneous graph with critical probability @xmath35 for percolation and such that @xmath185 , the probability that ( with parameter @xmath34 ) the origin belongs to an infinite cluster , satisfies @xmath186 for @xmath187 , there is almost surely no dynamical time @xmath0 at which percolation occurs . in our setting
, an entirely parallel argument can be used to show tameness of the dynamical discrete web with respect to recurrence .
we discuss this briefly below , pointing to the relevant parts of @xcite .
we consider the event @xmath188 that the walker starting from @xmath20 does not visit the site to the left of its starting position , that is , that the path @xmath189 started at @xmath20 does not contain any @xmath190 with @xmath191 .
let @xmath192 be the probability of that event under @xmath34 i.e. , when the random walk increments are @xmath193 ( resp .
, @xmath194 ) with probability @xmath34 ( resp . ,
@xmath195 ) . under the usual coupled construction of the model for @xmath196
$ ] , this event is increasing with @xmath34 in @xmath197 $ ] . for @xmath198 ( resp .
@xmath199 ) , @xmath189 is distributed as a right ( resp . left ) drifting random walk . in particular
, it is well known that for @xmath200 $ ] @xmath201 we now describe the parallel argument alluded to above .
let @xmath202 denote the cardinality of the set @xmath203:\,a_{i , j}\mbox { occurs in } ( \xi)_s\}$ ] .
@xmath192 and @xmath202 are the analogues of @xmath204 and @xmath205 in section 3 of @xcite .
an analogue of lemma 3.1 there also holds here with the same proof , where here @xmath206 is the analogue of @xmath35 there , and the analogue of @xmath207 is the number @xmath208 of @xmath209 such that @xmath188 occurs in @xmath210 } \xi^s_{i',j'},\,(i',j')\in{\mathbb z}_{even}\right\},\ ] ] and we conclude from ( [ tame2 ] ) that @xmath211 .
analogues of lemmas 3.2 and 3.4 also hold with the same proofs for the analogue quantities , and with @xmath188 replacing the event \{_v percolates_}. we then have that @xmath212 , and thus almost surely for every @xmath0 every walker eventually visits the site to the left of its starting position .
the same is of course true of the site to the right of the starting position by symmetry .
we conclude that almost surely for every @xmath13 every walker eventually visits every site in @xmath213 ( infinitely often ) .
[ remtame ] another property of the static discrete web with respect to which the dynamical one is tame is the almost sure coalescence of all of its paths .
it is enough to consider the case of two paths . for those ,
a similar argument as that for recurrence holds .
the analogue objects to be considered in this case are as follows . given @xmath214 ( let us assume that @xmath215 ) , let @xmath216 be the event that the paths starting from @xmath217 do not eventually coalesce , and let @xmath218 be the cardinality of the set @xmath203:\,c_{v , v'}\mbox { occurs in } ( \xi)_s\}$ ] . for @xmath219 , let also @xmath220 be the event that the path of @xmath221 starting from @xmath222 and that of @xmath223 starting from @xmath224 do not coalesce eventually , where @xmath225 } \xi^s_{i , j},\,(i , j)\in{\mathbb z}_{even}\right\}.\ ] ] let now @xmath226 be the number of @xmath209 such that @xmath220 occurs . to show that @xmath227 , we analyze @xmath226 and its related quantities analogously to the analysis of @xmath208 and its related quantities to show that @xmath228 . in particular , the fact that @xmath229 follows from standard random walk estimates , using the fact that the difference of two random walk paths is another random walk path .
[ twosided ] a main result of this paper is the existence of exceptional @xmath0 such that for all @xmath13 , @xmath230 ( see theorem [ violation ] in section [ existence ] ) , and of course there are then also exceptional @xmath0 such that @xmath231 .
however , it can be shown that there are no exceptional @xmath0 for the two - sided bound @xmath47 , at least for small enough @xmath3 .
the precise condition on @xmath3 under which we can prove this result is that @xmath232 , where @xmath233 is defined in proposition [ upper ] below .
note that this condition implies according to proposition [ upper ] that the hausdorff dimension of the set of exceptional @xmath0 for either of the corresponding one - sided bounds does not exceed @xmath206 .
the proof of this tameness claim combines arguments like those of theorem [ th1 ] and remark [ remtame ] with the estimates of lemma [ cont ] and proposition [ principal ] and with an application of the fkg inequalities .
the specific fkg inequality , for the two events @xmath234 that for some @xmath235 $ ] and all @xmath13 , @xmath236 , is that @xmath237 .
this is so because @xmath238 ( resp . , @xmath239 )
is an increasing ( resp . , decreasing ) event with respect to the basic @xmath240 processes
see , e.g. , lemma 3.3 of @xcite for more details .
we finally note that by essentially the same arguments one obtains tameness for two - sided bounds of the form @xmath241 provided that @xmath242 are small enough that @xmath243 .
let @xmath244 be a sequence of positive integers divisible by @xmath245 .
we construct inductively a sequence of `` diffusive '' boxes @xmath246 in the following manner : * @xmath247 is the rectangle with vertices @xmath248 , @xmath249 , @xmath250 and @xmath251 .
* let @xmath252 and @xmath253 be respectively the middle point of the lower edge and the upper right vertex of @xmath254 .
@xmath255 is the rectangle of height @xmath256 and width @xmath257 such that @xmath258 equals @xmath253 ( see figure 4 ) .
let @xmath259 be the event that the path of @xmath25 starting at @xmath260 is at or to the right of @xmath261 at time @xmath262 and that it is never to the left of the left edge @xmath263 of the box @xmath246 .
we would like to prove that for a certain choice of @xmath264 , there exist some exceptional times @xmath0 at which @xmath265 occurs for every @xmath266 .
at those times , this would imply that the path starting from the origin stays to the right of the graphs obtained by patching together the left edges , @xmath263 , ( see figure 4 ) . by the same kind of reasoning used in dynamical percolation @xcite , to prove that ordinary diffusive behavior does not occur at certain exceptional dynamical times @xmath0
, it suffices to derive the following lemma , which we do later in this section of the paper . with @xmath13 the vertical and @xmath49 the horizontal coordinate .
the solid curves represent segments of the paths starting from @xmath267 , @xmath268 and @xmath269 for which the events @xmath270 , @xmath271 and @xmath272 occur .
the leftmost curve represents @xmath273 .
] [ propp ] there exists @xmath274 such that if @xmath275 + 1)$ ] for @xmath276 ( where @xmath277 $ ] is the integer part of @xmath49 ) with @xmath278 , then @xmath279 where @xmath34 is bounded away from @xmath8 when @xmath116 is bounded away from @xmath280 .
let @xmath281 be the set of times @xmath0 on @xmath282 $ ] such that @xmath283 occurs .
the previous lemma implies that @xmath284 . since @xmath285 is obviously decreasing in @xmath286 ,
if the @xmath281 were closed subsets of @xmath282 $ ] it would follow that @xmath287 . as explained in the proof of the next theorem , for @xmath288 , @xmath289 for some @xmath290 ( depending on @xmath291 ) and all @xmath292 . unfortunately , the set of times at which one arrow is ( or any finitely many are ) oriented to the right ( resp .
, to the left ) is not a closed subset of @xmath282 $ ] since we have a right continuous process , and thus @xmath281 is not a closed set .
this extra technicality is handled like in lemma 5.1 in @xcite , as follows .
let @xmath293 denote the ( random ) set of all switching times for all @xmath294 s . by modifying every @xmath294 so that for @xmath295 , @xmath296 ( rather than being right - continuous ) , each @xmath281 is replaced by a closed @xmath297 .
on the other hand , @xmath298 as a consequence of the fact that @xmath293 is countable and by independence of the @xmath117 s no @xmath295 can be exceptional .
the uniformity with respect to small @xmath116 in lemma [ propp ] means that once space and time are diffusively rescaled by @xmath299 and @xmath53 and @xmath116 is rescaled by @xmath300 , the inequality ( [ esd ] ) is still valid , with @xmath34 fixed as @xmath88 . as a consequence of lemma [ propp ]
, we will obtain the following .
( * violation of the clt * ) [ violation ] for @xmath301 , let @xmath302)$ ] where @xmath303 is the path starting at @xmath7 of the dynamical discrete web with switching rate @xmath304 .
there exists @xmath305 such that @xmath306 , the probability to have a nonempty set of exceptional times s in @xmath282 $ ] for which @xmath307 for all @xmath308 satisfies the following : 1 . for any @xmath309
, there exist @xmath310 such that @xmath311 2 .
similarly , for any @xmath312 , @xmath313 , there exists @xmath314 such that ( [ gg ] ) is valid .
3 . for any fixed @xmath60 , @xmath315 , @xmath316 . in
the unrescaled coordinates , we take boxes @xmath246 as in lemma [ propp ] with @xmath317 + 1 ) \in ( \gamma^k/\lambda,4+\gamma^k/\lambda]$ ] . then in rescaled coordinates we have boxes @xmath318 with ( spatial ) width @xmath319 $ ] and ( temporal ) height @xmath320 .
let @xmath321 denote the right - continuous function obtained by joining together the left boundaries @xmath322 of @xmath318 .
on @xmath323 with @xmath324 , we have @xmath325 @xmath326
. if @xmath327 are such that @xmath328 then we will have @xmath329 for all @xmath10 as desired .
the inequality ( [ boundary ] ) can be rewritten as @xmath330^{1/2 } \ , .\ ] ] using the bound @xmath331 on the left - hand side of ( [ gggg ] ) and the bounds @xmath332 on the right - hand side , it follows that in order to verify ( [ gggg ] ) it suffices to have , for @xmath333 , @xmath334 using the elementary bound @xmath335 ( for @xmath336 ) , we see that in order to verify ( [ anotherineq ] ) , it suffices to have , for @xmath333 , @xmath337 choosing @xmath338 yields this inequality provided @xmath339 .
it is easy to see that for any @xmath340 , this will be valid provided @xmath341 is small enough and @xmath312 is large enough so that @xmath342 .
this and lemma [ propp ] prove the first claim of the theorem ; the second claim , in which @xmath312 and @xmath341 are given , follows similarly .
we now turn to the proof of the final claim .
we set @xmath343 since essentially the same proof works for any @xmath344 .
let @xmath345}$ ] denote the set of @xmath346 $ ] such that @xmath347 for @xmath348 and let @xmath17 be an integer so large that ( by the second claim of the theorem ) @xmath349 .
first , @xmath350 } \supset \hat t_{j}^{[0,1]}\bigcap \{s\in [ 0,1]:\xi_{m , m}^s = + 1 \ \ \textrm{for } \ \ m < j\}$ ] where @xmath351}$ ] is the set of @xmath346 $ ] such that @xmath352 for @xmath353 . furthermore , @xmath351 } \supset \bar t_{j}^{[0,1]}$ ] , where @xmath354}$ ] is the set of @xmath346 $ ] such that @xmath355 for @xmath353 .
but @xmath354}$ ] is just the translation ( from @xmath7 to @xmath356 ) of @xmath357}$ ] .
since @xmath358 : \forall \ k < j , \ \xi_{k , k}^s=+1\}$ ] and @xmath359}$ ] are independent , it follows that @xmath360 , \
\forall \
k < n , \ \
\xi_{k , k}^s=+1)>0 \ , .\ ] ] in particular , if we assume that @xmath361 converges to the dynamical brownian web ( see subsection [ continuum ] ) in some appropriate sense as @xmath88 , this shows that the analogue of theorem [ tameness ] ( except for the final claim with @xmath60 fixed and @xmath362 ) will be valid for the continuum model as well .
we now turn to : _ proof of lemma [ propp ] .
_ by the cauchy - schwarz inequality , @xmath363\right)^2 } { \mathbb{e}\left[\left(\int_0 ^ 1 \prod_{k=0}^n 1_{a_k^s } \ ds \right)^2\right ] } \\ & = & \left(\left[\int_0 ^ 1 \int_0 ^ 1 \prod_{k=0}^n \frac{\mathbb{p}(a_k^s \bigcap a_k^{s'})}{\mathbb{p}(a_k)^2 } \ ds \ ds ' \right ] \right)^{-1 } \label{integrand}\end{aligned}\ ] ] where @xmath364 and the equality is a consequence of the stationarity of @xmath365 and the independence between the different boxes @xmath246 .
it is enough to show that the integrand in the last expression of ( [ integrand ] ) is bounded above by a integrable function on @xmath282\times[0,1]$ ] , uniformly in @xmath286 .
the rest of the proof will verify this property .
now , for fixed @xmath266 and two deterministic times @xmath366 , let us rescale space and time respectively by @xmath367 and @xmath368 . also , let @xmath369 be the paths starting at @xmath206 at time @xmath8 , defined as the rescaled and translated version of the paths @xmath370 starting at @xmath260 , the middle point of the lower segment of the box @xmath246 .
@xmath179 is a pair of sticky rescaled random walks starting at @xmath206 at time @xmath9 whose statistics ( up to a translation of starting point ) are described in equations ( [ stat1])-([stat2 ] ) . by definition , @xmath371 } \ \tilde s_i^\delta ( t ) > 0)$ ] . to complete the proof of lemma [ propp ] , we will use the following lemma , in which @xmath157 may be taken as an integer divisible by @xmath245 .
[ inequality ] let @xmath369 be a pair of sticky random walks starting from @xmath206 at time @xmath9 as defined in ( [ stat1])-([stat2 ] ) .
let @xmath372 be the event that for @xmath373 , @xmath374 and @xmath375 }
\ \tilde s_i^\delta ( t ) \geq0 $ ] . if @xmath376 , then for @xmath377 small enough , @xmath378 where @xmath379 and @xmath380 are positive constants ( independent of @xmath381 and @xmath60 ) . in this proof we set @xmath382 .
for any positive @xmath383 , let @xmath384 be as in ( [ stat1])-([stat2 ] ) and let @xmath385 } s_i^\delta(t)$ ] .
then @xmath386 we start by dealing with the first term of the right - hand side of ( [ equation1 ] ) .
first , @xmath387 ) + 2 \mathbb{p}(\inf s_1^\delta\in[-\delta^\alpha,0])\end{aligned}\ ] ] using the independence of the walks @xmath388 and the equidistribution of @xmath389 .
the last two terms can be dealt with in a number of ways .
for example , in @xcite , it is proved that a sequence of rescaled standard random walks @xmath390 and a brownian motion @xmath92 can be constructed on the same probability space in such way that for for any @xmath391 the quantity @xmath392 goes to @xmath8 faster than any power of @xmath60 . on this probability space , @xmath393 )
\leq \mathbb{p } ( { \hat b}(1 ) \in [ 1- 2 \delta^\alpha,1+\delta^\alpha])+ \mathbb{p}(\sup|\hat b - s_i^\delta|>\delta^\alpha ) \ , , \\
\mathbb{p}(\inf s_i^\delta\in[-\delta^\alpha,0])\leq \mathbb{p } ( \inf { \hat b } \in [ - 2 \delta^\alpha,\delta^\alpha])+ \mathbb{p}(\sup|\hat b - s_i^\delta|>\delta^\alpha ) \ , , \label{es2}\end{aligned}\ ] ] where the sup ( and inf ) are over @xmath394 $ ] . since @xmath395
, we have @xmath396 , implying that for @xmath397 and @xmath60 small enough the last terms on the right - hand side of ( [ 91 ] ) and ( [ es2 ] ) are bounded by @xmath398 , and consequently by @xmath399 . finally , ( [ 90 ] ) , ( [ 91 ] ) and ( [ es2 ] ) yield : @xmath400 where @xmath379 is a positive constant and @xmath397 .
it only remains to deal with the rest of the terms on the right - hand side of equation ( [ equation1 ] ) .
we will prove that @xmath401 ; the other terms can be treated in a similar fashion . for any @xmath402
, we have @xmath403 lemma [ lemm1 ] below takes care of the second term on the right - hand side of the inequality when @xmath404 . on the other hand , since @xmath405 we have that @xmath406 now , on the event @xmath407 , by definition of @xmath408 , we have for any @xmath409 $ ] : @xmath410 since @xmath411 and @xmath412 is an increasing function , it follows that @xmath413 implying @xmath414}|s_3^\delta ( t ) |\geq
\frac{\delta^\alpha}{2 } ) + \mathbb{p}(\sup_{t \in [ 1-\delta^{\beta},1]}|s_1^\delta(1)-s_1^\delta(t)| \geq\frac{\delta^{\alpha}}{2 } ) \
, .\end{aligned}\ ] ] by ( the @xmath415 version of ) doob s inequality , we then have @xmath416 therefore , taking @xmath417 with @xmath418 gives the desired bound for the second term of the right - hand side of inequality ( [ equation1 ] ) .
for the first term of this inequality , we only needed @xmath419 and the conclusion follows .
[ lemm1 ] for any @xmath420 and @xmath421 small enough @xmath422 where @xmath423 . by the markov inequality , @xmath424 since @xmath425 converges in distribution to the local time of @xmath426 , where @xmath93 is a standard brownian motion , @xmath427 is uniformly bounded in @xmath60 .
furthermore , @xmath428 , implying that @xmath429 * completion of proof of lemma [ propp ] . * + recall that @xmath430 + 1 ) \geq \gamma^k / \lambda$ ] , where @xmath431 with @xmath432 to be fixed later .
by lemma [ inequality ] , there exists @xmath433 small enough such that ( [ lemmainequality ] ) is valid for @xmath434 . we define @xmath435 $ ] so that for @xmath436 , ( [ lemmainequality ] ) is valid for @xmath259 and @xmath437 .
@xmath438 is independent of @xmath116 and since @xmath439 , @xmath440 where @xmath380 and @xmath379 are as in lemma [ inequality ] .
the right - hand side of ( [ final1 ] ) is independent of @xmath116 and @xmath441 and is finite .
indeed , @xmath442 since the boxes @xmath246 have diffusively scaled sizes and therefore @xmath443 as @xmath444 , where @xmath445 is the event that a brownian motion @xmath446 starting at @xmath129 at time 0 has @xmath447 and @xmath448 } { \hat b}(t)>0 $ ] . on the other hand , @xmath449 where @xmath450 . taking @xmath451 , by ( [ final1 ] ) and ( [ final2 ] ) we have that for every @xmath286 @xmath452 with @xmath453 and @xmath454 . since @xmath455\times[0,1])$ ] ) and ( [ energy ] ) is uniform in @xmath286 , this concludes the proof of lemma [ propp ] .
in this section , we derive some lower and upper bounds for the set of exceptional dynamical times @xmath456 . to simplify notation , the rate of switching @xmath116 and the scaling parameter @xmath60 will both be taken equal to @xmath176 from now on .
however , as in the previous section , it can easily be checked that essentially all the results stated below are again uniform in @xmath457 once space and time and @xmath116 are properly rescaled according to @xmath60 ( see subsection [ scalinglimits ] ) .
the result that is not uniform as stated is proposition [ constant ] ; to have uniformity , @xmath458 should be replaced by @xmath459 for any @xmath460 .
[ def1 ] we say that @xmath0 is a @xmath3-exceptional time if the path @xmath461 in @xmath25 starting from the origin at time @xmath9 does not cross the moving boundary @xmath462 .
@xmath463 is then defined as the set of all @xmath3-exceptional times @xmath464 .
clearly , the set consisting of all the k - exceptional times in @xmath465 is a non - decreasing function of @xmath3 .
note that in definition [ def1 ] , the constant term for the moving boundary is fixed at @xmath176 .
the next propostion asserts that for fixed @xmath3 the hausdorff dimension @xmath466 of the set of exceptional times is unchanged if @xmath176 is replaced by any @xmath467 .
( the remark following the proof of the proposition points out that more can be proved by essentially the same arguments . ) we note that as in dynamical percolation ( see sec .
6 of @xcite ) , @xmath468 is a.s .
a constant by the ergodicity in @xmath0 of the dynamical discrete web .
[ constant ] the hausdorff dimension @xmath466 of the set @xmath469 of exceptional times @xmath470 such that @xmath119 does not cross the moving boundary @xmath471 does not depend on @xmath472 ( for fixed @xmath3 ) . by monotonicity in @xmath266 ,
it is enough to prove that @xmath473 for @xmath266 any positive integer . by the same reasoning
used to prove the last claim of theorem [ violation ] , the hausdorff dimension of @xmath474 is @xmath475 the hausdorff dimension of the set of times @xmath476 where @xmath477 is the translation ( from @xmath7 to @xmath478 ) of @xmath479 . by ergodicity in @xmath0 , the a.s .
constant @xmath480 is the essential supremum of the random variable @xmath481)$ ] . on the other hand , since @xmath482 $ ] and @xmath203 : \forall \ m < k , \ \xi_{m , m}^s=+1\}$ ] are independent and the probability to have @xmath483 , \ \forall \
m < k , \ \xi_{m , m}^s=+1\}$ ] is strictly positive , it follows that @xmath484 : \forall \ m < k , \
\xi_{m , m}^s=+1\ } \bigcap \bar t_k(k))$ ] has the same essential sup as @xmath485)$ ] .
hence @xmath486 and the conclusion follows .
define @xmath487 to be the set of @xmath0 such that for some @xmath488 and some @xmath489 , the infimum over @xmath348 of @xmath490 ( or equivalently of @xmath491 ) is @xmath492 .
it is not hard to see by arguments like those of proposition [ constant ] that @xmath493 .
[ lower ] @xmath494 converges to @xmath176 as @xmath45 .
let @xmath495 be fixed .
since @xmath463 increases with @xmath3 it is enough to show that for @xmath3 large enough the hausdorff dimension is at least @xmath383 .
consider the random measure @xmath496 , defined as @xmath497 , for any borel set @xmath498 in @xmath282 $ ] ( where @xmath265 are defined as in section [ existence ] ) .
we define the @xmath383-energy of @xmath496 as @xmath499 by identical arguments as in section 6 of [ ss05 ] , if the expected value of @xmath500 is bounded above as @xmath106 , then the hausdorff dimension of the set of exceptional @xmath0 ( in @xmath465 ) for which @xmath501 occurs is at least @xmath383 . by fubini s theorem , @xmath502 and by ( [ energy ] ) , we have that @xmath503 \right ) \ ] ] with @xmath504 .
in particular , taking @xmath505 as in the proof of theorem [ violation ] with @xmath291 large enough , @xmath506 can be made smaller than @xmath176 , and the right - hand side of ( [ haus ] ) is finite . combining the proofs of lemma [ propp ] and proposition [ lower ] with the remark following proposition [ constant ] , we can obtain a more explicit lower bound on @xmath507 as follows .
set @xmath508 , where @xmath445 is the event that a brownian motion @xmath446 starting at @xmath206 at time @xmath9 has @xmath447 and @xmath509 , and define @xmath510 as the solution in @xmath511 of @xmath512 for @xmath513 . letting @xmath514 , we then have @xmath515 we will prove the following proposition .
[ upper ] for any @xmath516 , @xmath517 where @xmath518 is the real solution @xmath519 of the equation @xmath520 furthermore , @xmath521 and more significantly @xmath522 . as a consequence of propositions
[ constant ] , [ lower ] and [ upper ] , we immediately have the following .
[ dimension ] the limits as @xmath523 and @xmath524 of @xmath468 are @xmath525 for any continuous function @xmath526 starting at @xmath527 , @xmath528 such that @xmath529 , the set of exceptional times for which the path starting from the origin at time @xmath8 does not cross @xmath526 has hausdorff dimension zero . to prove proposition [ upper ] we need the two following lemmas .
( sato [ s77 ] ) [ samere ] let @xmath530 , where @xmath531 are both positive , and @xmath93 is a standard brownian motion .
then there exists @xmath532 such that @xmath533 where @xmath534 is the real solution in @xmath535 of ( [ sato ] ) .
[ cont ] for any @xmath536 , let @xmath537 where @xmath538 and @xmath93 is a standard brownian motion .
then for some @xmath539 and @xmath540 , @xmath541 let @xmath542 be the density of @xmath543 . by the girsanov theorem , @xmath544 where @xmath545 is the density corresponding to a standard brownian motion .
therefore , since @xmath546 , we have @xmath547 integrating by parts , we get that @xmath548 on the one hand , by lemma [ samere ] , @xmath549 on the other hand , @xmath550 the last three displayed equations together easily imply ( [ exitbound ] ) .
_ proof of proposition [ upper ] .
_ we are now ready to obtain an upper bound for the hausdorff dimension of the set of k - exceptional times @xmath463 .
let us partition @xmath282 $ ] into intervals of equal length @xmath551 , and select the intervals containing a k - exceptional time .
the union of those is a cover of @xmath463 and we now estimate the number @xmath552 of intervals in the cover .
let @xmath553 be the event that there is a time @xmath0 in @xmath554 $ ] such that @xmath555 .
from the full dynamical arrow configuration for all @xmath556 $ ] , we construct a static arrow configuration as follows .
we declare the static arrow at @xmath20 to be right oriented if and only the dynamical arrow is right oriented ( i.e. , @xmath557 ) at some @xmath235 $ ] ( a similar construction was used in section [ tameness ] ) . in this configuration ,
the path @xmath558 starting from the origin and following the arrows is a slightly right - drifting random walk with @xmath559 .
clearly , @xmath560 proposition [ principal ] of appendix [ appendix ] implies that for any @xmath561 @xmath562 and by lemma [ cont ] it follows that @xmath563 hence @xmath564 so that @xmath565 by fatou s lemma , @xmath566 is almost surely bounded , which implies that the hausdorff dimension of @xmath463 is bounded above by @xmath567 and completes the proof of proposition [ upper ] .
we will prove the following proposition .
[ principal ] let @xmath568 , where @xmath93 is a standard brownian motion , and let @xmath558 be a discrete time simple random walk with drift given by @xmath569 for @xmath513 there exists @xmath570 such that for any @xmath516 , @xmath571 we consider @xmath572 the discrete time random walk embedded in the drifting brownian motion @xmath573 .
namely , we define inductively a sequence of stopping times @xmath574 and their increments @xmath575 , with @xmath576 and @xmath577 and then we define @xmath578 . the proof of proposition [ principal ] will be done by coupling @xmath572 and @xmath573 in this particular way .
note that @xmath572 has a different drift than @xmath579 since @xmath580 is not @xmath581 , but rather is @xmath582 . but proving ( [ theineq ] ) with @xmath558 replaced by @xmath583 suffices since @xmath583 has a larger positive drift than @xmath558 .
now , let us consider some variants of @xmath573 and @xmath583 .
define @xmath584 , where @xmath585 , as explained later , and @xmath586 is defined as @xmath587 this is the diffusion obtained by turning on a small drift of magnitude @xmath588 after the stopping time @xmath589 . in exactly the same way , we define @xmath590 and @xmath591 for @xmath586 ( but without the @xmath551 superscript ) and define @xmath592 . in particular @xmath593 and @xmath594
coincide for @xmath595 . 1 .
the first part is to show that except on a set of probability @xmath606 , @xmath607 can be replaced by @xmath608 in the last expression of ( [ fff ] ) , with @xmath609 so that @xmath610 as @xmath611 .
this will be done essentially by an application of the law of large numbers .
once the above replacement has been made , the desired conclusion follows directly from the correlation inequality of lemma [ third ] and the inequality , @xmath612 where the @xmath613 were defined in ( [ tautau ] ) , with @xmath614 . noting that @xmath615 and hence @xmath616
, the last displayed inequality is a standard fact about sums of i.i.d .
positive mean random variables . in gambling terms
, it says that a gambler with a slight advantage has a strictly positive probability of never falling behind .
it remains to justify the first part , for which it is enough to prove that , up to an error of at most @xmath606 , on the event @xmath617 , the inequality @xmath618 is actually valid for all @xmath286 . for @xmath619 ,
the @xmath591 are the exit times @xmath620 of brownian motion with a small drift @xmath551 .
clearly , @xmath621 as @xmath622 . by the law of large numbers and standard large deviation estimates
, we can assume that @xmath552 is in @xmath623 $ ] and show that the event @xmath624 occurs , except on a set of probability @xmath606 .
hence , up to this error , on the event @xmath625 , the inequality @xmath626 can be extended from all @xmath627 also to all @xmath628 .
hence , it only remains to control the indices @xmath286 in @xmath629\}$ ] . since @xmath630 , and
since we can assume that @xmath631 , we get that for any such @xmath286 , @xmath632 this completes the proof of lemma [ firstlemma ] _ proof of lemma [ third ] .
_ this result is a consequence of the fkg inequality for independent random variables .
the variables @xmath634 ( see ( [ tautau ] ) ) and @xmath635 are completely independent , since in the case of a standard brownian motion , knowing the exit time from the interval @xmath636 $ ] does not give any information about the exit location .
hence , given @xmath552 , @xmath637 behaves as a usual symmetric simple random walk for @xmath627 .
thereafter the walk has a small positive drift to the right .
this suggests that the indicator of the event @xmath638 can be expressed as a nondecreasing function of @xmath639 ( and some other variables to be determined ) since the larger @xmath639 is , the smaller @xmath552 will be , inducing more drift for @xmath596 . to make this more precise , we will couple @xmath640 and @xmath641 .
the coupling involves the mutually independent @xmath642-valued @xmath639 , @xmath27-valued @xmath643 and @xmath644-valued @xmath645 , with @xmath646 the coupling is not via a brownian motion but rather is given in terms of our independent variables by @xmath647 and @xmath648 it is clear now that the above suggestion about the nondecreasing nature of the event in question is indeed valid . since the other event
, @xmath649 is clearly nondecreasing with @xmath650 , the claim of lemma [ third ] follows by the fkg inequality .
_ proof of lemma [ secondlemma ] .
let us condition on @xmath552 .
we use the coupling of @xmath651 and @xmath583 just discussed . now @xmath652 and we need to prove that the last term is of order @xmath551 for a suitable choice of the exponent @xmath380 ( where @xmath653 is the time threshold at which @xmath596 starts drifting ) . first , @xmath654 denoting by @xmath655 the solution of @xmath656 , we will show by induction on @xmath19 that as @xmath622 , @xmath657 , s_{\epsilon}'(n)-\bar s_{\epsilon}(n ) \geq k(1-l)\sqrt{n } ) \ , = \ , \nonumber \\
o(\epsilon k_1 + \epsilon^2 ( k_2-k_1)^2 + ... + \epsilon^2 ( k_{i}-k_{i-1})^2).\end{aligned}\ ] ] assuming this has been proved , we let @xmath658 be such that @xmath659 , and ( [ ps ] ) then implies that @xmath660 , \ , s_{\epsilon}'(n)- \bar s_{\epsilon}(n)\geq k(1-l ) n^\frac{1}{2 } ) \ , = \nonumber \\ o(\epsilon k_1 + \epsilon^2 ( k_2-k_1)^2 + ... + \epsilon^2 ( k_{n}-k_{n-1})^2 ) \ , .\end{aligned}\ ] ] for large @xmath286 , @xmath661 and therefore @xmath662 , implying that @xmath663 , \ , s_{\epsilon}'(n)- \bar s_{\epsilon}(n)\geq k(1-l ) n^\frac{1}{2})= o(\epsilon^2 n(\epsilon)^{\frac{3}{2 } } ) \ , .\ ] ] since we can assume by the law of large numbers that @xmath664 , taking @xmath665 implies that @xmath666 , and we get that the last term of ( [ reduce ] ) is @xmath606 . note that everything was done independently of @xmath552 ( except that @xmath667 ) .
therefore , summing over the possible values of @xmath552 would finish the proof .
it remains to prove ( [ ps ] ) , which we do by induction .
first , for @xmath668 , since @xmath669 , @xmath670 , \ , s_{\epsilon}'(n)-\bar s_{\epsilon}(n ) \geq k(1-l ) \sqrt{n } ) & = & \mathbb{p}(s_{\epsilon}'(\cdot)-\bar s_{\epsilon}(\cdot )
\ \textrm{jumps on $ [ 0,k_1]$ } ) \\ & = & k_1 o(\epsilon ) .\end{aligned}\ ] ] next , assuming that ( [ ps ] ) is valid up to @xmath19 , we have @xmath671 , \ , \ s_{\epsilon}'(n)-\bar s_{\epsilon}(n ) \geq k(1-l)\sqrt{n } ) & \\ & \leq \mathbb{p } ( { \textrm { for some}}\ n\in(0,k_i ] , \ s_{\epsilon}'(n)- \bar s_{\epsilon}(n)\geq k(1-l)\sqrt{n } ) & \\ & + \mathbb{p}(\forall n \in ( 0,k_i ] , \ s_{\epsilon}'(n)- \bar s_{\epsilon}(n)<k(1-l)\sqrt{n } & \\ & \ { \textrm { and for some } } \ n \in ( k_{i},k_{i+1 } ] , \ s_{\epsilon}'(n)- \bar s_{\epsilon}(n)\geq k(1-l)\sqrt{n } ) \ , . & \end{aligned}\ ] ] we need to bound the last term of this inequality .
since on @xmath672 $ ] we have @xmath673 $ ] ( by the definition of @xmath655 ) , and since @xmath674 only takes integer value , if @xmath675 , then @xmath676 . on the other hand , @xmath677 , and
then our process has to jump at least twice on @xmath672 $ ] . but
this probability is bounded by a term of order @xmath678 and ( [ ps ] ) follows .
this completes the proof of lemma [ secondlemma ] . _
acknowledgements . _ the research of l.r.g .
fontes was supported in part by fapesp grant 2004/07276 - 2 and cnpq grants 307978/2004 - 4 and 484351/2006 - 0 ; the research of the other authors was supported in part by n.s.f .
grants dms-01 - 04278 and dms-06 - 06696 . | the dynamical discrete web ( ddw ) , introduced in recent work of howitt and warren , is a system of coalescing simple symmetric one - dimensional random walks which evolve in an extra continuous dynamical time parameter @xmath0 .
the evolution is by independent updating of the underlying bernoulli variables indexed by discrete space - time that define the discrete web at any fixed @xmath0 . in this paper
, we study the existence of exceptional ( random ) values of @xmath0 where the paths of the web do not behave like usual random walks and the hausdorff dimension of the set of exceptional such @xmath0 .
our results are motivated by those about exceptional times for dynamical percolation in high dimension by hggstrom , peres and steif , and in dimension two by schramm and steif .
the exceptional behavior of the walks in the ddw is rather different from the situation for the dynamical random walks of benjamini , hggstrom , peres and steif . in particular
, we prove that there are exceptional values of @xmath0 for which the walk from the origin @xmath1 has @xmath2 with a nontrivial dependence of the hausdorff dimension on @xmath3 .
we also discuss how these and other results extend to the dynamical brownian web , a natural scaling limit of the ddw .
the scaling limit is the focus of a paper in preparation ; it was also studied by howitt and warren and is related to the brownian net of sun and swart . |
frequency division multiplexing ( ofdm ) has been widely adopted in both cellular systems , such as long - term evolution ( lte ) and wi - fi systems @xcite .
the main advantage of ofdm modulation is to convert an intersymbol interference ( isi ) channel into multiple isi - free subchannels and thus reduces the demodulation complexity at the receiver @xcite .
however , since each symbol is only transmitted over a parallel flat fading subchannel , the conventional ofdm technique may not collect multipath diversity , it thus performs worse than single carrier transmission @xcite . furthermore , ofdm has high peak - to - average power ratio ( papr ) of the transmitted signals , which may affect its applications in broadband wireless communications .
single - carrier frequency domain equalization ( sc - fde ) is an alternative approach to deal with isi with low transmission papr @xcite .
however , induced by both fast fourier transform ( fft ) and inverse fft ( ifft ) operations at the receiver , sc - fde suffers from the drawback that transmitter and receiver have unbalanced complexities @xcite . as a result , ofdm is more suitable for downlink with high transmission speed , whereas sc - fde can be applied for uplink that reduces papr and transmitter complexity as in lte @xcite .
vector ofdm ( v - ofdm ) for single transmit antenna systems first proposed in @xcite converts an isi channel to multiple vector subchannels where the vector size is a pre - designed parameter and flexible . for each vector subchannel
, the information symbols of a vector may be ( are ) isi together . since the vector size is flexible , when it is @xmath0 , v - ofdm coincides with the conventional ofdm .
when the vector size is @xmath1 , each vector subchannel may have two information symbols in isi . when the vector size is large enough , say , the same as the ifft size , then the maximal number of information symbols are in isi and it is then equivalent to sc - fde .
therefore , v - ofdm naturally builds a bridge between ofdm and sc - fde in terms of both isi level and receiver complexity @xcite , and it has attracted recent interests . for v - ofdm ,
an adaptive vector channel allocation scheme was proposed for v - ofdm systems @xcite .
some key techniques , such as carrier / sampling - frequency synchronization and guard - band configuration in v - ofdm system were designed and made comparisons with the conventional ofdm systems @xcite .
iterative demodulation and decoding under turbo principle is an efficient way for v - ofdm receiver @xcite .
constellation - rotated v - ofdm was proposed with improved multipath diversity @xcite .
linear receivers and the corresponding diversity order analyses are recently given in @xcite and phase noise influence is investigate in @xcite .
for a very broadband channel , the ifft size of an ofdm system needs to be very large , which may cause practical implementation problems , such as high papr and high complexity .
in contrast , for a v - ofdm , its ifft size can be fixed and independent of a bandwidth , while its vector size can be increased to accommodate the increased bandwidth . in this paper , we are interested in v - ofdm over a broadband sparse channel in the sense that it has a large time delay spread but only a few of nonzero taps @xcite . for sparse channels ,
there have been many studies in the literature , see for example @xcite .
recently , sparse fft ( sfft ) theory was proposed by the computer science & artificial intelligence lab , massachusetts institute of technology @xcite .
if a signal has a small number @xmath2 of nonzero fourier coefficients , the output of the fourier transform can be represented succinctly using only @xmath2 coefficients . for such signals ,
the runtime is sublinear in the signal size @xmath3 rather than @xmath4 .
furthermore , several new algorithms for sfft are presented , i.e. , an @xmath5-time algorithm for the exactly @xmath2-sparse case , an @xmath6-time algorithm for the general case @xcite . in this paper , a sparse channel estimation and decoding scheme for v - ofdm systems is proposed .
inspired by the idea of sfft under the condition of signals with only a few nonzero fourier coefficients , we first use pilot symbols to obtain channel frequency response ( cfr ) , and then estimate channel impulse response ( cir ) by using sparse ifft ( sifft ) .
based on the estimation of nonzero channel coefficients and their corresponding coordinates , an efficient partial intersection sphere ( pis ) decoding is investigated and it achieves the same diversity order as the maximum likelihood ( ml ) decoding .
the main contributions of the paper are summarized as follows . * we find a connection between the exactly and approximately sparse channel models in the estimation of a v - ofdm sparse multipath channel . for a multipath channel with only a few nonzero taps ,
if there is no noise during the transmission , then the sparse channel can be estimated by the exactly sparse multipath channel algorithm that corresponds to algorithm 3.1 for exactly sparse fft in @xcite .
when there is additive white gaussian noise ( awgn ) during the transmission , the sparse multipath channel can be estimated by the approximately sparse multipath channel algorithm that corresponds to algorithms [email protected] for generally sparse fft in @xcite .
* by using the sifft - based algorithms , one can directly recover the nonzero channel coefficients and their corresponding coordinates , which is significant to the pis decoding process .
* for the pis decoding in v - ofdm systems , the bit error rate ( ber ) is dependent of @xmath8 nonzero taps in a sparse channel and the vector size @xmath9 to some extent , but roughly independent of the maximum delay @xmath10 .
* for any given small sphere radius , the proposed pis decoding and ml decoding are of the same diversity order , which is equal to the cardinality of the set of reminder coordinates after mod @xmath9 , but the pis decoding can substantially reduce the computational complexity with probability @xmath0 .
the reminder of the paper is organized as follows . in section
ii , the system model of v - ofdm is reviewed . in section iii ,
sifft - based channel estimation schemes for the exactly sparse case and the approximately sparse case are introduced . in section
iv , a pis decoding for v - ofdm systems is proposed and analyzed . in section
v , simulation results are presented and discussed . in section
vi , this paper is concluded .
we first briefly recall a v - ofdm system for single transmit antenna , which is shown in fig .
1 . the description of system model follows the notations in @xcite below . in v - ofdm systems , @xmath11 symbols @xmath12 are blocked into @xmath13 vectors ( called vector blocks ( vb ) ) of size @xmath9 .
denote the @xmath14th transmitted vb in @xmath15 as @xmath16^{\mathrm t},~l=0,1,\ldots , l-1\ ] ] where @xmath17 denotes the transpose .
assume the average power is normalized , i.e. , @xmath18 , where @xmath19 denotes the mathematical expectation .
accordingly , @xmath20 is defined as the normalized vb - based ifft of size @xmath13 , i.e. , @xmath21 here , @xmath20 is a column vector of size @xmath9 represented as @xmath22^{\mathrm t}$ ] .
after parallel to series ( p / s ) conversion , the transmitted signal sequence @xmath23 is obtained as @xmath24^{\mathrm t}$ ] . in order to avoid the interblock interference ( ibi ) ,
the length of cp denoted by @xmath25 should not be shorter than the maximum time delay of a multipath channel .
note that @xmath25 does not need to be divisible by @xmath9 . at the transmitter ,
the signal sequence @xmath26 inserted by cp , is transmitted serially through the channel with the order @xmath27^{\mathrm t}$ ] . at the receiver
, the received sequence is modeled by the transmitted signal through a frequency selective fading channel with complex awgn .
an inverse process as in transmitter is performed to recover the original symbols . after removing cp
, the received sequence @xmath28 is equal to the circular convolution of the transmitted signal and cir with awgn @xmath29 where @xmath30 denotes the circular convolution , cir @xmath31 and @xmath10 denotes the maximum time delay spread of the multipath channel .
after zero padding of @xmath32 , cfr @xmath33 is the @xmath11-point fft ( without normalization ) of @xmath32 . assume the additive noise @xmath34 is independent and identically distributed ( i.i.d . )
random sequence whose entry @xmath35 .
accordingly , define the transmitted signal - to - noise ratio ( snr ) as @xmath36 .
@xmath37 is then blocked into @xmath13 column vectors of size @xmath9 .
denote the @xmath2th vector in @xmath37 as @xmath38^{\mathrm t},~k=0,1,\ldots , l-1.\ ] ] take the normalized component - wise vector fft operation of size @xmath13 as @xmath39 the @xmath14th received vb in @xmath40 is also a column vector of size @xmath9 represented as @xmath41^{\mathrm t}$ ] .
it is derived from @xcite that the relationship between the @xmath14th transmitted vb @xmath42 and received vb @xmath43 as @xmath44 where @xmath45 is a blocked channel matrix of the original isi channel @xmath46 as @xmath47 where @xmath48 is the @xmath49th polyphase component of @xmath46 , @xmath50 .
the additive noise @xmath51 in ( 6 ) is the blocked version of @xmath52 whose entries have the same power spectral density as in @xmath34 that are i.i.d . complex gaussian random variables .
note that @xmath53 can be diagonalized as @xmath54 where @xmath55 is a unitary matrix whose entries @xmath56_{r , c}=\frac{1}{\sqrt{m}}\mathrm e^{-\mathrm j\frac{2\mathrm\pi}{n}(l+rl)c},~r , c=0,1,\ldots , m-1 $ ] , and @xmath57 denotes the conjugate transpose , @xmath58 is an @xmath59 diagonal matrix defined as @xmath60 furthermore , the unitary matrix @xmath55 can be rewritten as @xmath61 where @xmath62 denotes the @xmath59 normalized discrete fourier transform ( dft ) matrix whose entries @xmath63_{r , c}=\frac{1}{\sqrt{m}}\mathrm e^{-\mathrm
j\frac{2\mathrm\pi}{m}rc},~r , c=0,1,\ldots , m-1 $ ] , and @xmath64 is a diagonal matrix defined as @xmath65 it can be seen from ( 6 ) that the original isi channel @xmath46 of @xmath66 symbols interfered together is converted to @xmath13 vector subchannels , each of which may have @xmath9 symbols interfered together .
note that @xmath9 is the vector size and can be flexibly designed .
when @xmath67 , ( 6 ) is back to the original ofdm , i.e. , no isi occurs in each subchannel . when @xmath68 , all @xmath66 symbols are interfered together and it is back to the sc - fde .
now , we rewrite the relationship of inputs and outputs in ( 6 ) for the better understanding of channel transmission structure @xmath69 it is straightforward to show that after the unitary transformation , the @xmath14th vb @xmath42 is transmitted parallel over the subchannels @xmath70 .
mathematically , @xmath55 is a kind of rotation matrix , @xmath58 can be thus viewed as the equivalent channel fourier coefficients .
denote @xmath71 as the number of pilot channels and assume @xmath13 is divisible by @xmath71 .
if the @xmath72th vb @xmath73^{\mathrm t}$ ] is allocated to transmit pilot symbols , then the pilot symbols are transmitted parallel over the equivalent channels @xmath74 .
if @xmath71 pilot channels are evenly distributed over @xmath13 subchannels , i.e. , @xmath75 then the equivalent channels allocated to transmit pilot symbols are aligned as @xmath76^{\mathrm t}\nonumber\\ = & \big[h_0,h_{\frac{l}{p}},\ldots , h_{n-\frac{l}{p}}\big]^{\mathrm t}\end{aligned}\ ] ] therefore , it is not difficult to find that pilot symbols are also evenly distributed over the equivalent channels . fig . 2 shows the pilot pattern for v - ofdm systems with parameters @xmath77 . due to the even distribution
, @xmath78 can be regarded as the downsampling of the cfr @xmath79 such that for most cases @xmath80 , we usually only need to perform ifft with size @xmath81 rather than @xmath11 to exactly estimate the cir @xmath32 .
with the increase of communication bandwidth , the number of equivalent channels @xmath11 needs to be increased and a signal sequence can be transmitted over more parallel channels simultaneously . accordingly , either the number of vbs @xmath13 or the vb size @xmath9 increases proportionally .
more parallel channels means higher rate for transmission , which however , increases the computational complexity of both channel estimation and decoding . in this section
, an sifft - based approach is proposed for sparse multipath channel estimation that can directly obtain the nonzero channel coefficients and their corresponding coordinates .
denote @xmath82 , @xmath83 , and @xmath84 .
then ( 12 ) is further rewritten as @xmath85 note that after the unitary transformation , the entries of @xmath86 are also i.i.d .
complex gaussian random variables .
@xmath87 is defined as a column vector @xmath88^{\mathrm t}$ ] and @xmath89 is the estimator of @xmath87 .
it is convenient to estimate @xmath87 by the least squares approach such that @xmath90^{-1}\widetilde{\bm y}_l$ ] , @xmath91 .
consider the pilot channels are evenly distributed over @xmath13 channels as ( 13 ) .
denote @xmath92 and @xmath93 as the estimators of @xmath78 and @xmath32 , respectively . according to ( 14 ) ,
@xmath94 is comprised of @xmath95 , and can be estimated by @xmath96^{-1}\widetilde{\bm y}_{l_p}$ ] , @xmath97 . with the knowledge of column vector @xmath78 , we can further obtain column vector @xmath32 by implementing the ifft operation without normalization . since the additive noise @xmath98 is an @xmath99 i.i.d .
random sequence whose entries @xmath100_m\thicksim\mathcal{cn}\left(0,\sigma^2\right)$ ] , @xmath50 , it is not difficult to check that @xmath101 is an unbiased estimator of @xmath102 , i.e. , @xmath103 , @xmath97 . besides , the mean squared error ( mse ) of @xmath101 , denoted by @xmath104 , can be derived as @xmath105\nonumber\\ & = \sigma^2\big[\mathrm{diag}\big\{\big|\widetilde{\bm x}_{l_p}\big|^2\big\}\big]^{-1}\end{aligned}\ ] ] since pilot signals are transmitted through a multipath channel with awgn , our goal is to design the pilot symbols to minimize the mse of estimator @xmath93 , i.e. , @xmath106\big\}\ ] ] where @xmath107 denotes the constellation of the transmitted symbol @xmath108 , @xmath109 is the trace of square matrix defined as the sum of the main diagonal . obviously , @xmath93 is also an unbiased estimator of @xmath110 due to the linear transformation @xmath111 where @xmath112 denotes the @xmath113 normalized inverse dft ( idft ) matrix whose entries @xmath114_{r , c}=\frac{1}{\sqrt{mp}}\mathrm e^{\mathrm j\frac{2\mathrm\pi}{mp}rc},~r , c=0,1,\ldots , mp-1 $ ] .
now , we check the mse of estimator @xmath93 as @xmath115=\frac{\mathbf f_{mp } \mathbf\sigma\mathbf f_{mp}^{-1}}{mp}\ ] ] where @xmath116 is a diagonal matrix whose diagonal entries @xmath117_{p+mp}=\left[\mathbf \sigma_{l_p}\right]_m,~m=0,1,\ldots , m-1,~p=0,1,\ldots , p-1 $ ] .
it is not difficult to find that the diagonal entries in ( 19 ) are the same , hence the entries of @xmath118 are identically distributed with complex gaussian random variables , but may not be independent of each other .
the trace of ( 19 ) can also be simplified as @xmath119\big\}=\frac{\mathrm{tr}\left\{\mathbf\sigma\right\}}{mp}\ ] ] note that ( 20 ) is only dependent on the sum of the mse of each pilot channel
. therefore , ( 17 ) can be optimized by designing the pilot symbols for each channel , respectively , such that @xmath120 , @xmath97 .
consider the total number of transmitted signals @xmath121 is fixed , and @xmath71 pilot channels are evenly distributed over @xmath13 channels with @xmath122 .
assume the transmitted sequence @xmath123 is binary phase - shift keying ( bpsk ) signals in v - ofdm system .
for the @xmath72th pilot channel , the normalized expectation of mse over @xmath9 symbols , defined as @xmath124 , should be minimized . according to parseval s theorem , @xmath125 .
note that such pilot symbol design is not unique .
table i lists a type of pilot symbol design for different @xmath13 and @xmath9 .
it is shown that such design can keep @xmath126 within @xmath127 , and does not increase with @xmath9 . in section iii b
, we will find that with such pilot symbol design , the snr and the power ratio of the dominant entries to the rest entries in @xmath93 are in the same order of magnitude , regardless of the parameters @xmath9 and @xmath71 , which is accordance with approximately sparse channel . [ cols="^,^,^,^,^,^",options="header " , ] for the conventional ofdm signal , the mses of the cfr estimator for all subchannels are the same .
for the v - ofdm , however , the entries in @xmath128 may no longer be constant modulus since unitary transformation @xmath129 is evolved in the original pilot symbols , and @xmath104 varies from subchannel to subchannel .
similar to the channel spectral nulls in the ofdm systems , if symbol spectral nulls are existed in @xmath128 , the overall estimation error in the @xmath14th subchannel may be very large . for the sparse channel
that the pilot symbol are not well designed , the noise component may be comparable to the dominant component in the estimator @xmath93 , and in what follows , @xmath93 may be obtained as a non - sparse channel . for a very broadband channel
, @xmath130 may become very large . in this case
, it becomes expensive to implement the large size ifft directly .
inspired by the idea of sfft proposed in @xcite , for a sparse multipath channel , we can perform sifft to estimate cir from cfr . in what follows
, we will focus on such sparse multipath channel estimation with and without awgn in the transmission , respectively .
we first consider a simple noiseless sparse multipath channel , i.e. , @xmath131 , which is called an exactly sparse multipath channel .
the pilot signals are transmitted through a sparse channel with only @xmath8 nonzero taps spread but without additive noise .
if ifft is performed to estimate cir from cfr , in this case , it is no doubt that the estimator @xmath132 is also a column vector of size @xmath81 with only @xmath8 nonzero entries .
the sifft - based channel estimation for exactly sparse multipath channel is illustrated in algorithm 1 .
the input @xmath92 is first permuted by @xmath133 , then multiplied by flat window filtering @xmath134 . after the permutation and filtering
, the nonzero channel taps can be sampled at the interval @xmath135 .
substituting the permutation operator @xmath136 for @xmath133 and repeating the above process , the nonzero coordinates can be recovered from the phase difference between these two permutations , and their corresponding values are obtained by permutation @xmath133 . after repeating @xmath137 times , one can eventually recover @xmath93 with exact @xmath8-sparse .
the algorithm includes three functions : * exactlysparseifft : iterate coordinatevalue and update @xmath93 , repeat @xmath137 times and eventually find @xmath93 with exact @xmath8-sparse . *
coordinatevalue : access to hashtobins and obtain nonzero coordinates and their corresponding values .
it can find more than half of nonzero entries in @xmath93 each time . * hashtobins : permute @xmath92 and guarantee that nonzero entries in @xmath93 are separated into different bins and then compute @xmath138-dimensional ifft in @xmath139 , where @xmath138 denotes the number of bins and is set proportional to @xmath8 .
similar to @xcite , the proposed algorithm has two fundamental steps , i.e. , permutation and flat window filtering .
the purpose of permutation is to separate nonzero coefficients into different bins randomly .
the design of filtering is a tradeoff between the filter flatness and the support of the window . rather than the exactly sparse algorithm presented in @xcite that recovers the sparse signal with only a few nonzero
fourier coefficients , algorithm 1 is aimed to estimate the cir from the cfr by using the pilot symbols .
the main differences between them are listed as follows . * instead of the permutation operator presented in @xcite , in this subsection , we redefine the permutation operator @xmath140 as @xmath141 where @xmath142 is a discrete sequence in frequency domain with size @xmath3 , and @xmath143 .
denote @xmath144 as the idft of @xmath145 .
it is not hard to derive that @xmath146 where @xmath147 , and @xmath148 is the idft of @xmath149 . compared with the definition of permutation in @xcite , when computing the coordinates of nonzero channel coefficients in algorithm 1 , the proposed permutation can be recovered directly without the required dictionary . *
different from the exactly sparse algorithm in @xcite being only suitable for integers , we further expand the application to complex domain since any nonzero tap in sparse channel is a complex number . more specifically ,
denote the resolution @xmath150 as the minimum value that nonzero entries can be detected , if @xmath150 is set less than or equal to the minimum magnitude of nonzero channel taps , then all nonzero channel coefficients can be recovered with high probability . for the flat window with gaussian filtering
, the sample sequence should be collected with the length at least @xmath151 , where @xmath152 , it may thus increase the sample sequence length if the sparse channel exists small nonzero taps .
therefore , the proposed algorithm can be suitable for the complex channel at the expense of a potential higher complexity .
the exactly sparse case in @xcite is the case when a signal has only a few nonzero fourier coefficients .
accordingly , algorithm 1 is suitable for the case that the estimator @xmath93 has only a few nonzero taps .
therefore , it is straightforward to employ algorithm 1 to recover @xmath8 nonzero entries in @xmath93 .
it has been proved in @xcite that the complexity of exactly sparse algorithm is @xmath153 .
furthermore , if the length of symbol sequence @xmath81 is sufficiently large such that @xmath154 is satisfied , algorithm 1 can recover the correct coordinates and their corresponding values with high probability .
@xmath155 @xmath93 * initialization : * @xmath156 @xmath157 @xmath158 @xmath159 + @xmath93 @xmath160 @xmath161 , for @xmath162 choose @xmath163 randomly from @xmath164 choose @xmath165 randomly from @xmath166 @xmath167 @xmath168 * initialization : * @xmath169 @xmath170 @xmath171 @xmath172 + @xmath93 @xmath173 @xmath174 @xmath175 @xmath176_{\frac{n}{b}j}$ ] + @xmath177 now , we consider a more practical scenario that the pilot signals are transmitted through a sparse channel with only @xmath8 nonzero taps spread and awgn , which is called an approximately sparse multipath channel .
since awgn is induced during the transmission , the estimator @xmath93 is no longer with only @xmath8 nonzero entries . in fact
, the estimator @xmath132 has @xmath8 dominant entries and the rest entries are small , when the snr is not low .
for the approximately sparse vector @xmath93 , define the parameter @xmath178 as the maximum expectation power ratio of the @xmath8 selected entries to the rest entries such that @xmath179 where @xmath180 denotes the @xmath181 norm of a vector .
@xmath178 reflects how approximately the sparse multipath channel is and determines the root - mean - square error ( rmse ) of sifft algorithm . in particular ,
exactly sparse is an extreme case for @xmath182 .
the sifft - based channel estimation for approximately sparse multipath channel is shown in algorithm 2 that has the following basic idea . to deal with noise , the algorithm estimates the nonzero coordinates and their corresponding values separately . for the coordinate estimation ,
all the coordinates are first divided into small regions .
the input @xmath92 is permutated randomly by @xmath140 and @xmath183 , respectively , then multiplied by flat window filtering @xmath134 .
the phase difference between these two permutations determines the circular distance to each region .
select the appropriate regions with the nearest circular distance and get one vote . after repeating the above process @xmath184 times ,
choose the final regions with more than @xmath185 votes . by narrowing the regions of nonzero
coordinates in each iteration , the algorithm eventually obtains the nonzero coordinates . for the value estimation , after the permutation and filtering , the nonzero values corresponding to the coordinates estimated before are obtained by permutation @xmath140 . repeating @xmath186 times and choose the median as the estimations of the values such that the estimation error decreases exponentially with @xmath186 .
repeat the above process @xmath187 times and ultimately recover @xmath93 with @xmath8 dominant taps .
the algorithm includes five functions , in which hashtobins is defined the same as in algorithm 1 .
* approximatelysparseifft : iterate coordinate and value , then update @xmath93 . in each iteration , reduce @xmath2-sparse to @xmath188-sparse , repeat @xmath187 times and eventually find @xmath93 with @xmath8 dominant entries .
* coordinate : access to range and narrow the range of dominant coordinates , repeat @xmath189 times until the dominant coordinates are uniquely determined . *
range : permute @xmath92 randomly with @xmath184 times , divide all the coordinates into several regions , find the appropriate regions with the nearest circular distance and then gets one vote .
after repeating @xmath184 times , choose the final regions with more than @xmath185 votes .
* value : access to hashtobins and obtain the estimations of the values , repeat @xmath186 times and take the median of such values with real and imaginary parts , respectively .
* initialization : * @xmath190 @xmath191 @xmath192 @xmath193 @xmath194 @xmath195
@xmath159 + @xmath93 choose @xmath163 randomly from @xmath164 choose @xmath196 randomly from @xmath166 @xmath197 * initialization : * @xmath169 @xmath198 @xmath199 + @xmath93 @xmath173 @xmath174 @xmath175 @xmath176_{\frac{n}{b}j}$ ] + @xmath177 * initialization : * @xmath200 for @xmath201 choose @xmath163 randomly from @xmath164 choose @xmath165 randomly from @xmath166 @xmath202 @xmath203 @xmath204 + @xmath205 * initialization : * @xmath206 for @xmath207 @xmath208 choose @xmath209 randomly from @xmath166 choose random variable @xmath210 evenly distributed from @xmath211 @xmath212 @xmath213 @xmath214 @xmath215 @xmath216 @xmath217
@xmath218 + @xmath219 similar to @xcite , the permutation operator @xmath140 in this subsection is defined as @xmath220 where @xmath221 .
accordingly , the idft of @xmath140 is derived as @xmath222 where @xmath223 . as we will see from the proof in appendix a , algorithm 2 is suitable for the case when the estimator @xmath93 has a few dominant taps and thus it can be applied to recover @xmath8 dominant taps in @xmath93 . for a sparse multipath channel with awgn ,
the complexity of the approximately sparse algorithm is @xmath224 , which is more complicated than the exactly case , but still much simpler than the ifft with @xmath225 operations .
if the condition @xmath154 holds , algorithm 2 can estimate the coordinates and their corresponding values with low rmse . in section
v , we will present some simulation results to show that the rmse of channel estimation is influenced by @xmath178 and ultimately determined by both @xmath226 and the design of pilot symbols . as a result ,
the sifft - based channel estimation algorithm can not only reduce computational complexity , but also return the nonzero channel coefficients and their corresponding coordinates directly , which is significant to the following pis decoding process .
in v - ofdm systems , the performances of several common decoding approaches were analyzed in @xcite .
it was proved in @xcite that the diversity order of the ml decoding is @xmath227 . in @xcite , it was shown that the diversity order of the minimum mean square error ( mmse ) decoding can achieve @xmath228 , where @xmath229 represents the spectrum efficiency in bits / symbol , while for the zero - forcing ( zf ) decoding , the diversity order is @xmath0 .
for all the demodulations of the ml , zf and mmse , they need to obtain @xmath33 which is computed by the @xmath11-point fft of the zero padded @xmath110 .
if @xmath11 is very large , the computational load is high . for a sparse channel ,
most entries in @xmath230 are zero and the nonzero entries are regularly placed .
based on this observation , it may be better to extract nonzero entries over each row and search all possible symbol sequences lying in a certain sphere of radius around the received signal .
hence , the complexity of searching such possible sequences is exponential to the number of nonzero entries in each row of @xmath230 , which is much less than @xmath9 when @xmath9 is large as what is studied in this paper . in this section ,
a partial intersection sphere ( pis ) decoding algorithm is proposed for a sparse multipath channel . here
, partial intersection means the intersection of the existed and the current nonzero coordinate sets . in each iteration , the algorithm only needs to compare the current symbol sequences corresponding to the coordinates belonging to the partial intersection with the existed ones .
the proposed pis decoding algorithm is illustrated as algorithm 3 and explained below in detail .
assume the sparse channel @xmath110 has only @xmath8 nonzero taps with the maximum delay @xmath10 .
denote @xmath231 as the set of coordinates of nonzero taps for the sparse multipath channel @xmath110 , i.e. , @xmath232 and the cardinality of set @xmath231 is equal to @xmath8 , i.e. , @xmath233 .
considering the special structure of @xmath234 , denote @xmath235 as the set of the reminders of the nonzero channel coefficient coordinates modulo @xmath9 , i.e. , @xmath236 denote @xmath237 as the cardinality of @xmath235 , i.e. , @xmath238 .
suppose @xmath239 are the @xmath237 entries in @xmath235 with the ascending order @xmath240 .
for the case in this paper , we have @xmath241 .
in what follows , it will be found that the diversity order for the pis decoding is only related to @xmath237 . in the @xmath49th iteration with @xmath242 ,
denote @xmath243 and @xmath244 as the sets of the existed coordinates of the nonzero entries in the first @xmath245 rows and the current coordinates of the nonzero entries in the @xmath49th row of @xmath230 , respectively .
@xmath246 ( called partial intersection ) is defined as the intersection of @xmath243 and @xmath244 , i.e. , @xmath247 .
note that @xmath230 is from a pseudo - circulant matrix ( 7 ) where the number of nonzero entries in each row is equal to @xmath237 .
recall that @xmath107 is the constellation of the transmitted symbol @xmath248 . for the initialization , the set of the existed coordinates of the nonzero entries @xmath249 and the set of entire symbol sequences @xmath250 are empty sets , respectively , i.e. , @xmath251 , @xmath252 .
then , we describe the updating process of pis decoding in the @xmath49th iteration with @xmath242 as follows . 1 .
extract @xmath237 entries from the @xmath49th row and @xmath253th , @xmath254th,@xmath255 , @xmath256th columns of @xmath230 , generate @xmath257 as @xmath258 vector @xmath259_{m,(m - i_0)\bmod m},\left[\bm{\mathcal h}_l\right]_{m,(m - i_1)\bmod m},\ldots$ ] @xmath260_{m,(m - i_{\kappa-1})\bmod m}\big]$ ] .
2 . search all possible symbol sequences @xmath261^{\mathrm
t},~\bm s\in\mathbb x^{\kappa}$ ] , that lie in the certain sphere of radius @xmath262 around the received signal @xmath263 and generate the set of symbol sequences @xmath264 as @xmath265 where @xmath263 is the @xmath49th entry of the column vector @xmath266 .
3 . for each @xmath267
, construct an injective mapping of coordinates @xmath268 . for each symbol
sequence @xmath269 , where @xmath270 is the set of entire symbol sequences generated from the previous iteration , compare the current symbol sequence @xmath271 for the coordinates belonging to the partial intersection @xmath246 with the existed symbol sequence @xmath272 , namely , if @xmath273 holds for all @xmath274 , where @xmath275 stands for the @xmath276th entry in @xmath272 , then @xmath272 is put into the set of symbol sequences @xmath277 , which can be expressed as @xmath278 .
then , insert the symbols whose coordinates belong to the complement of the partial intersection @xmath246 to each symbol sequence @xmath272 , i.e. , @xmath279 , where @xmath280 stands for the complement of @xmath246 in @xmath244 , set @xmath281 , insert @xmath282 to each symbol sequence @xmath272 in @xmath277 and generate @xmath283 , the new set of symbol sequences is thus updated as @xmath284 .
repeat step 3 by enumerating all @xmath267 .
then the set of entire symbol sequences @xmath285 is obtained by the union of all @xmath286 , i.e. , @xmath287 .
@xmath288 is updated to @xmath289 as the existed coordinates of nonzero entries for the next iteration .
after @xmath9 iterations , the set of possible vb sequences @xmath290 can be ultimately obtained , then choose the symbol sequence @xmath291 with the minimum @xmath181 distance of @xmath292 as the estimation of the transmitted vb @xmath293 .
@xmath294 @xmath295 * initialization : * @xmath296 calculate @xmath297 according to ( 7 ) @xmath298 @xmath299 @xmath239 are @xmath237 entries in @xmath235 with the ascending order @xmath240 @xmath263 is the @xmath49th entry of column vector @xmath266 @xmath300 is @xmath258 vector aligned as the @xmath49th row and @xmath253th , @xmath254th@xmath301th columns of @xmath53 @xmath302 @xmath303 @xmath304 @xmath268 @xmath305 @xmath306 @xmath307 @xmath308 @xmath309 + @xmath295 for a v - ofdm system with the pis decoding , assume the cir @xmath110 and the average power of complex awgn @xmath310 are known at the receiver . denote @xmath311 as the correct symbol sequence corresponding to the transmitted symbols , i.e. , @xmath311 is extracted from the @xmath253th , @xmath254th,@xmath255 , @xmath256th entries of @xmath293 and generated as @xmath312_{(m - i_0)\bmod m},\left[\bm x_l\right]_{(m - i_1)\bmod m},\ldots$ ] @xmath313_{(m - i_{\kappa-1})\bmod m}\big]$ ] .
hence , the distance between @xmath263 and @xmath314 , i.e. , @xmath315 , is rayleigh distributed with mean @xmath316 and variance @xmath317 . according to the cumulative distribution function of rayleigh distribution , the probability that the transmitted symbol @xmath311 lies in the sphere radius @xmath262 in the @xmath49th iteration is @xmath318 from the previous analysis , the additive noise @xmath319 in ( 6 ) is an @xmath320 vector whose entries are i.i.d .
complex gaussian random variables . hence , for @xmath50 , the events that the transmitted symbol @xmath311 lies in the sphere are independent and the probabilities that each event occurs are the same .
after @xmath9 iterations , the occurrence of event @xmath321 is equivalent to the occurrence of all the @xmath9 events @xmath322 , then we have @xmath323 note that @xmath324 is a set of possible vb sequences whose corresponding @xmath237-dimensional symbol sequences lie in the sphere radius such that ( 28 ) holds for all @xmath9 rows in ( 6 ) .
in fact , the choice of sphere radius @xmath262 is a tradeoff between the symbol error rate ( ser ) performance and the computational complexity . with the increase of @xmath262 ,
@xmath325 also increases which consequently improves the ser performance .
however , this means that more possible symbol sequences need to be compared in each iteration .
the ser of the proposed pis decoding @xmath326 is a function of sphere radius @xmath262 and calculated by the law of total probability @xmath327 note that for @xmath328 , there is no doubt that @xmath329 since @xmath330 is chosen from @xmath290 .
denote @xmath331 as the probability that symbol error occurs conditioned on @xmath321 , i.e. , @xmath332 substituting ( 32 ) into ( 31 ) , @xmath326 can be further simplified as @xmath333 for a v - ofdm system , the signal vector @xmath293 needs to be specifically rotated / transformed to achieve full diversity for the ml decoding as done in @xcite . in section ii b , it was analyzed that if @xmath71 subchannels with the indices @xmath334 are allocated to transmit pilot symbols , the ifft / sifft - based channel estimation can be applied to recover cir @xmath110 . for the remaining subchannels allocated to transmit data symbols , it is proved in appendix b that the diversity order of the ml decoding for a sparse multipath channel is @xmath237 .
we describe the diversity order by the exponential equality @xmath335 , which is mathematically defined as @xmath336 @xcite , where the ser of the ml decoding @xmath337 can be found in @xcite . instead of the ml decoding that enumerates symbol constellation and estimates the transmitted symbols with the minimum distance , the pis decoding first generates the set of possible transmitted symbols @xmath290 and then chooses the transmitted symbols only from @xmath290 with the minimum distance .
it is proved in appendix c that @xmath338 .
accordingly , for sufficiently large @xmath226 , @xmath331 is exponentially less than or equal to @xmath339 , which can be expressed as @xmath340 , i.e. , @xmath341 .
therefore , from ( 33 ) , @xmath326 is exponentially less than or equal to @xmath339 if and only if @xmath342 is exponentially less than or equal to @xmath339 , i.e. , @xmath343 we say the sphere radius @xmath262 is the asymptotically greater than or equal to the sphere radius @xmath344 , denoted by @xmath345 , when @xmath346 for a sufficiently large @xmath36 , the infinitesimal @xmath342 approximates to @xmath347 . substituting ( 34 ) into ( 35 ) and supposing @xmath226 is sufficiently large , we have @xmath348 . furthermore , for a sufficiently large @xmath226 , the term @xmath349 can be neglected compared with @xmath350 , then the necessary and sufficient condition of @xmath351 is @xmath352 it is well known that the ser of the proposed pis decoding can not be better than that of the ml decoding . according to ( 36 ) , we have the following lemma that gives the criterion of sphere radius satisfying @xmath353 .
the ser of the proposed pis decoding is exponentially equal to that of the ml decoding in the choice of sphere radius @xmath354 for a sufficiently large @xmath226 . for the v - ofdm system
, it is known that the complexities with respect to complex multiplication operation of mmse decoding and ml decoding are @xmath355 and @xmath356 , respectively .
the pis decoding only needs @xmath357 trials with @xmath237 complex multiplication operation in each trial .
hence , the complexity with respect to complex multiplication operation of the pis decoding is @xmath358 .
besides , the evaluation and comparison operations should be taken into account in pis decoding , which in fact , may vary from @xmath359 to @xmath360 and are related to the cardinality of @xmath270 and ultimately dependent of sphere radius @xmath262 . as illustrated in algorithm 3 ,
the evaluation operation is an operator used for assignment where the source @xmath361 is a complex number and the destination @xmath362 is the @xmath363th entry in the symbol sequence @xmath364 , i.e. , @xmath365 , while the comparison operation is one of relational operator used to check the equality of two complex numbers @xmath275 and @xmath366 , i.e. , @xmath367 , if the equality holds return @xmath0 , otherwise return @xmath368 . in assembly language , an evaluation operation or a comparison operation usually executes @xmath0 instruction cycle , whereas a real multiplication operation executes @xmath369 instruction cycles or slightly more due to hardware .
although different operations may have different execute time , the number of instruction cycles for any operation is fixed and can be seen as a constant .
then , the total complexity depends ultimately on the number of operations executed in the program . with the increase of @xmath226 ,
it is wise to decrease the sphere radius @xmath262 such that the computational complexities with respect to the evaluation and comparison operations can reach the lower bound @xmath359 with probability @xmath0 .
note that @xmath370 , then we have the following theorem .
for any given small sphere radius , the pis decoding can achieve the diversity order @xmath237 which is the same as the ml decoding for a sparse multipath channel , but the computational complexity decreases from @xmath356 to @xmath358 with probability @xmath0 .
see appendix d for the proof .
therefore , by choosing @xmath262 asymptotically equal to @xmath371 , the proposed pis decoding algorithm can balance the tradeoff between the ser performance and the computational complexity . since the diversity order is @xmath237 that depends on the set of the reminders of the @xmath8 nonzero channel coefficient coordinates modulo @xmath9 , in practice , for a given channel model , i.e. , for a given set of coordinates of nonzero channel coefficients , one may properly choose @xmath9 such that @xmath237 is maximized .
in this section , we provide simulation results to verify the previous analysis .
the bpsk modulation is employed in the v - ofdm system .
sparse multipath channel @xmath110 is modelled as @xmath8 i.i.d .
complex gaussian distributed nonzero taps @xmath372 randomly distributed within the maximum delay @xmath10 .
we first employ the rmse to evaluate the performances of the sifft - based sparse channel estimation .
then , we give an example of @xmath373 different channels with deterministic nonzero coordinates to make a comparison of the diversity order .
besides , we investigate the relationship between the ber performance of pis decoding and the parameters @xmath10 , @xmath8 , @xmath9 , respectively .
furthermore , the pis decoding is compared with the conventional zf , mmse , ml decoding schemes in the v - ofdm system .
finally , channel estimation and decoding algorithm are jointly considered to show the ber performances in both ofdm and v - ofdm systems .
3 and 4 show the rmse performances of the sifft - based sparse channel estimation with and without noise , respectively . in the simulation of the sifft - based exactly sparse multipath channel estimation ,
the parameters @xmath374 and @xmath138 are set to @xmath375 such that @xmath376 is a constant regardless of @xmath8 .
it can be seen from fig .
3 that the rmse of the channel estimation is below @xmath377 but reduces the complexity to @xmath153 , where @xmath71 is pilot channel number .
the estimation error is mainly caused by the imperfect permutation that the nonzero entries are not separated into different bins .
for the sifft - based approximately sparse multipath channel estimation , suppose the parameters @xmath378 and @xmath379 that can keep the collision at a relatively low level .
4 indicates that with the increase of @xmath226 , @xmath8 dominant entries are slightly influenced by the rest entries @xmath178 , which consequently , reduces the rmse of channel estimation .
for instance ,
when @xmath380 , the rmse of sparse channel estimation is below @xmath381 .
whereas when @xmath382 , however , there is a sharp decrease for @xmath383 since the condition @xmath154 does not hold in this case .
the complexity for the sifft - based approximately sparse multipath channel estimation is @xmath224 . in fig . 5
, we give an example of @xmath373 different channels with deterministic nonzero coordinates and for each channel , the nonzero channel coefficients are i.i.d .
complex gaussian distribution .
suppose @xmath384 , @xmath385 , @xmath386 , and the nonzero coordinates for channel a : @xmath387 , channel b : @xmath388 , channel c : @xmath389 , channel d : @xmath390 , channel e : @xmath391 , channel f : @xmath392 .
accordingly , the reminders of the nonzero coordinates modulo @xmath9 for channel a : @xmath393 , channel b : @xmath394 , channel c : @xmath395 , channel d : @xmath396 , channel e : @xmath397 , channel f : @xmath398 .
it can be seen from fig .
5 that the diversity order of channel a is @xmath0 , channel b and channel c are @xmath1 , channel d and channel e are @xmath399 , channel f is @xmath369 .
it is pointed out that although @xmath400 and @xmath401 , their corresponding diversity orders are different . as a result
, we can verify the previous analysis that the diversity order of sparse multipath channel is determined by the cardinality of the set of reminder coordinates after mod @xmath9 , rather than the cardinality of coordinate set itself .
6@xmath78 show how the parameters @xmath10 , @xmath8 , @xmath9 influence on the ber performance of pis decoding , respectively .
suppose the transmitted snr @xmath402 . in fig 6 ,
we compare the ber with respect to the maximum delay @xmath10 .
it can be seen that the ber is roughly irrelevant to the variation of @xmath10 , since @xmath8 nonzero taps are randomly distributed within @xmath66 taps .
7 investigates the relationship between the number of nonzero taps @xmath8 and the ber performance .
simulation result indicates that with the increase of @xmath8 , the ber decreases almost linearly in the logarithmic scale .
the reason is that the diversity order can be directly determined by @xmath237 and increases with @xmath8 with large probability .
8 shows the ber performance with respect to vb size @xmath9 in the v - ofdm system . the ber first decreases with the increase of @xmath9 for @xmath403 , whereas for @xmath404 , the ber increases with @xmath9 instead . on one hand
, a larger @xmath9 can avoid @xmath8 nonzero taps interacting with each other better after mod @xmath9 , in this case , @xmath405 with high probability which may improve the ber performance . on the other hand , for a given sphere radius @xmath262 , with the increase of @xmath9 , the probability that the transmitted symbols lie in the certain sphere decreases exponentially according to ( 30 ) , and thus diminishes the advantage of the pis decoding .
therefore , one can improve the ber performance of the pis decoding by choosing an appropriate vb size @xmath9 .
9 compares different decoding approaches in the v - ofdm system .
suppose the parameters @xmath406 , @xmath380 , @xmath384 , @xmath407 , @xmath408 .
since the condition @xmath409 does not hold , @xmath410 with not a small probability which may diminish the multipath diversity orders of the mmse decoding , ml decoding and pis decoding .
in fact , when @xmath10 is sufficiently large such that the reminders of the nonzero coordinates modulo @xmath9 can be regarded randomly distributed at the coordinates @xmath411 , for the given @xmath8 and @xmath9 , the probability mass function of @xmath237 is @xmath412 . after averaging over the random nonzero coordinates of channel , the diversity orders of the zf decoding , mmse decoding , ml decoding and pis decoding are corresponding to the minimum of @xmath237 and thus equal to @xmath0 . for the different decoding approaches , denote the bers of zf decoding , mmse decoding , ml decoding , pis decoding as @xmath413 , @xmath414 , @xmath337 , @xmath415 , respectively .
it is well known that @xmath416 .
simulation result indicates that the pis decoding loses certain ber performance since ( 36 ) does not hold if @xmath226 is not large enough , while for @xmath417 , the proposed pis decoding outperforms the zf decoding and mmse decoding and gradually approximates to the ml decoding with the increase of @xmath226 .
furthermore , the complexity of the pis decoding decreases with @xmath226 and is much less than the ml decoding . and
@xmath385.,width=336 ] with @xmath402 and @xmath408.,width=336 ] with @xmath402 and @xmath408.,width=336 ] with @xmath402 and @xmath408.,width=336 ] in fig .
10 , we consider the channel estimation and decoding algorithms jointly .
the ber performance is not only dependent of the decoding approaches , but also influenced by the channel estimation accuracy .
suppose the parameters @xmath418 , @xmath380 .
for the ofdm system , the receiver employs the fft - based interpolation for channel estimation and symbol - by - symbol decoding with parameters @xmath419 , pilot channel number @xmath420 . for the v - ofdm system with linear receivers
, we estimate channel by the conventional ifft - based approach , and employ the zf decoding and the mmse decoding with parameters @xmath421 , @xmath385 , @xmath422 , respectively . for the v - ofdm system with the ml decoding
, we estimate the channel by the conventional ifft - based approach as well with the parameters @xmath423 , @xmath407 , @xmath424 . for the v - ofdm system with the pis decoding ,
the sifft - based algorithm is employed for sparse multipath channel estimation with parameters @xmath421 , @xmath385 , @xmath422 .
if a slight bias is induced during the process of channel estimation , the sphere radius should not be extremely small since a robust sphere radius is needed to guarantee that the probability of the transmitted symbols lying in the sphere does not decrease . an empirical method to balance
the tradeoff between the estimation error and the complexity is to choose sphere radius @xmath425 .
it can be seen from fig .
10 that v - ofdm system outperforms the conventional ofdm system . compared with the zf decoding and the mmse decoding , the proposed sifft - based channel estimation and pis decoding reduces the ber significantly . and
in this paper , we investigate sparse multipath channel estimation and decoding for broadband v - ofdm systems . from the system model ,
if the pilot channels are evenly allocated over multiple subcarriers , the pilot symbols are evenly distributed over the equivalent channels . for the sparse multipath channel estimation , we first design a type of pilot symbols that can minimize the mse of an estimator
. then , we give sifft - based algorithms for exactly and approximately sparse multipath channel estimations corresponding to the cases with and without awgn induced during the transmission , respectively .
the remarkable significance of the sifft - based approach is to estimate the nonzero channel coefficients and their corresponding coordinates directly . for the pis decoding algorithm ,
the diversity order is determined by not only the number of nonzero taps , but also the coordinates of nonzero taps .
simulation results indicate that the ber performance of the pis decoding is comparable to that of the ml decoding with certain sphere radius for a sufficiently large snr , but reduces the complexity substantially .
for the pilot signals being transmitted through multipath channel with awgn , it was derived in ( 18 ) that @xmath93 is an unbiased estimator of @xmath110 .
the diagonal entries of ( 19 ) corresponding to the variance of @xmath93 , which are all equal to @xmath426 .
hence , @xmath93 is a random vector and can be regarded as @xmath110 with an additive noise whose entries are identically distributed with complex gaussian noise , i.e. , @xmath427 , but may not be white . for the sparse channel with only @xmath8 nonzero taps , @xmath178 approximates the expectation power ratio of @xmath8 dominant entries in @xmath93 to the rest entries such that @xmath428 considering the sparse channel that @xmath429 , ( 37 ) can thus be further simplified as @xmath430 it was listed in table i that the designed pilot symbols can reduce the interference efficiently .
suppose @xmath181 norm of the sparse channel is normalized , i.e. , @xmath431 .
since @xmath432 is proportional to @xmath310 , we have @xmath433 where the typical value of @xmath434 is @xmath435 and roughly independent of @xmath9 .
therefore , it can be naturally concluded that if the designed pilot symbols are transmitted through the normalized sparse multipath channel with awgn , then the estimator @xmath93 is an approximately sparse vector with the parameter @xmath436 , where @xmath226 denotes the transmitted snr , the typical value of @xmath434 is @xmath127 and roughly independent of @xmath9 .
assume the ideal channel state information is known at the receiver .
denote @xmath437 as the estimation of @xmath293 with the ml decoding , then the ser of the ml decoding conditioned on @xmath230 is written as @xmath438 . according to the ml decoding , for the transmitted symbol @xmath293 and a distinct symbol @xmath439 from the symbol constellation ,
if @xmath440 holds , then the symbol error occurs .
since the event @xmath441 is equivalent to @xmath442 @xcite , it is not difficult to derive a lower bound of the ser as @xmath443 and an upper bound of the ser as @xmath444 furthermore , the total number of elements in the symbol constellation @xmath445 is @xmath446 , then the upper bound ( 41 ) can be further simplified as @xmath447 therefore , the upper and low bounds of the ser have the same tendency and only differ by a constant multiplier .
denote @xmath448 , then we have @xmath449 where the @xmath450-function is define as @xmath451 . substituting ( 8) into ( 43 ) and
note that the @xmath181 distance does not change after the unitary transformation , ( 43 ) can be simplified as @xmath452 consider the sparse channel @xmath110 has only @xmath8 i.i.d .
nonzero taps and each nonzero entry is a complex gaussian random variable with zero mean and unit variance . recall that @xmath231 is the set of coordinates of the nonzero taps .
suppose @xmath453 are the @xmath8 entries in @xmath231 with the ascending order @xmath454 .
we construct a @xmath455 vector @xmath456^{\mathrm t}$ ] and all the @xmath8 entries in @xmath457 are i.i.d .
gaussian variables .
denote @xmath458 as an @xmath459 matrix constructed by extracting the @xmath14th , @xmath460th , @xmath461 , @xmath462$]th rows , @xmath463th , @xmath464th , @xmath461 , @xmath465th columns of the @xmath11-point fft matrix without normalization , i.e. , @xmath466_{r , c}=\mathrm e^{-\mathrm j\frac{2\mathrm\pi}{n}(l+rl)j_c}$ ] .
then , we have @xmath467 .
therefore , ( 44 ) can be further rewritten as @xmath468 where @xmath469 .
denote @xmath470 as the rank of @xmath471 , i.e. , @xmath472 , and @xmath473 are the @xmath470 nonzero eigenvalues corresponding to such positive semidefinite hermitan matrices @xmath471 with the descending order @xmath474 .
denote the constellation of pairwise error @xmath475 . after averaging over the complex gaussian random channel @xmath457
, an upper bound of the ser ( 42 ) can be calculated by the chernoff bound as @xcite @xmath476^{-r_{l\min}}\end{aligned}\ ] ] where @xmath477 .
denote @xmath478 and @xmath479 .
hence , the snr of the ml decoding is exponentially less than or equal to the minimum rank of @xmath471 , i.e. , @xmath480 . to derive a lower bound of the ser , ( 40 )
can be further calculated by the recursion of the integration by parts @xmath481 where the factorial @xmath482 , and the double factorial @xmath483 .
then , the snr of the ml decoding is exponentially greater than or equal to the minimum rank of @xmath471 , i.e. , @xmath484 . according to ( 46 ) and ( 47 ) ,
we conclude that the diversity order of the ml decoding is equal to the minimum rank of @xmath471 among all non - all - zero @xmath485 and all @xmath486 , which is also denoted by @xmath487 .
note that @xmath488 now , we consider two matrices @xmath489 and @xmath458 separately .
note that @xmath490 and @xmath64 is a rotation matrix evolved by v - ofdm modulation itself .
according to theorem 2 in @xcite , for the pulse - amplitude modulation ( pam ) or bpsk modulation , @xmath491 for @xmath492 , while for the quadrature amplitude modulation ( qam ) or quadrature phase - shift keying ( qpsk ) modulation , @xmath491 for @xmath493 .
in other words , for the conventional modulation , @xmath489 has full rank in most subchannels .
furthermore , from the previous analysis in section ii b , the @xmath71 subchannels with the indices @xmath334 are allocated to transmit pilot symbols .
hence , @xmath489 always has full rank for the channels allocated to transmit data symbols . on the other hand ,
@xmath458 can be constructed as @xmath494 $ ] , where the column vector @xmath495j_q}\big]^{\mathrm t}$ ] .
recall that @xmath235 is the coordinates of the nonzero taps modulo @xmath9 and @xmath496 are the @xmath237 entries in @xmath235 with the ascending order @xmath497 . according to ( 27 ) ,
@xmath498 , @xmath499 and an integer @xmath2 such that @xmath500 .
then , @xmath501 , where @xmath502i_p}\big]^{\mathrm t}$ ] . if there exists a distinct @xmath503 and an integer @xmath504 such that @xmath505 still holds , then @xmath506 .
hence , the vectors @xmath507 and @xmath508 are linearly dependent of @xmath509 . since there are only @xmath237 such @xmath510 in total ,
if let @xmath511 $ ] , we have @xmath512 . since the vectors in each column of a dft matrix are linearly independent , the vector columns @xmath513 which are equal to the @xmath514th columns of the @xmath9-point fft matrix without normalization while multiplied by the factor @xmath515 are also linearly independent
then , the maximal number of linearly independent columns of @xmath516 is @xmath237 .
thus , we have @xmath517 . since @xmath489 is an invertible matrix , @xmath518 . as a result
, it can be concluded that the diversity order of the ml decoding for a sparse multipath channel is @xmath237 .
denote the estimation symbol sequence @xmath519 as the conventional ml decoding of @xmath293 .
in contrast to the pis decoding , the ser of the ml decoding can be written as @xmath520 since @xmath521 , we have @xmath522
assume sparse channel @xmath110 has only @xmath8 i.i.d .
nonzero taps and each channel coefficient follows complex gaussian random distribution .
it is not difficult to find that the nonzero entries in @xmath230 are also complex gaussian random variables but the variances may not be equal . recall that @xmath264 was defined in ( 28 ) that all possible symbol sequences lying in the certain sphere of radius @xmath262 around the received signal @xmath263 .
@xmath311 is the correct symbol sequence corresponding to the transmitted symbols .
@xmath523 is the distance between @xmath263 and @xmath314 , i.e. , @xmath524 . since the noise @xmath51 in ( 6 ) is complex awgn
, @xmath523 is rayleigh distributed with mean @xmath316 and variance @xmath317 .
according to the triangle inequality , @xmath525 , we have @xmath526 compared with @xmath264 defined in ( 28 ) , it can be found that @xmath531 . if @xmath262 is chosen exponentially equal to @xmath371 such that lemma 1 is satisfied , then @xmath532 as @xmath533 . due to rayleigh distribution , @xmath534 with probability @xmath0 as @xmath533 . since the probability density function of each nonzero entry in @xmath230 is a complex gaussian function , for any given @xmath535 ,
if @xmath536 , we have @xmath537 it is found that for @xmath533 , @xmath538 such that ( 53 ) approximates to @xmath368 . since @xmath527 has only a bounded finite elements , we have @xmath539 , thus , @xmath527 has only one entry @xmath311 with probability @xmath0 when @xmath533 . obviously , @xmath540 with probability 1 as @xmath533 . since @xmath264 is a subset of @xmath527 , @xmath264 also has only one entry @xmath311 with probability @xmath0 . for any given small sphere radius @xmath262 ,
lemma 1 is satisfied as @xmath533 such that the pis decoding achieves the same multipath diversity order as the ml decoding , which is equal to @xmath237 .
furthermore , the cardinality @xmath541 of the set @xmath270 of possible symbol sequences in algorithm 3 remains @xmath0 in each iteration that the complexity for updating @xmath277 can be reduced significantly . for @xmath542 , the evaluation and comparison operations
are performed @xmath543 times in the @xmath49th iteration .
consider @xmath9 iterations and @xmath13 subchannels , the complexities of the evaluation and comparison operations are @xmath359 with probability @xmath0 . in the previous analysis of the pis decoding ,
the complexity of complex multiplication operation is @xmath358 .
since the evaluation and comparison operations for a complex number are faster than a complex multiplication operation , the total complexity of the pis decoding is @xmath358 with probability @xmath0 .
d. falconer , s. l. ariyavisitakul , a. benyamin - seeyar , and b. eidson , `` frequency domain equalization for single - carrier broadband wireless systems , '' _ ieee commun . mag .
4 , pp . 58@xmath766 , apr . 2002 . x .- g .
xia , `` precoded and vector ofdm robust to channel spectral nulls and with reduced cyclic prefix length in single transmit antenna systems , '' _ ieee trans .
49 , no . 8 , pp . 1363@xmath71374 , aug .
2001 .
h. zhang , x .-
xia , q. zhang , and w. zhu , `` precoded ofdm with adaptive vector channel allocation for scalable video transmission over frequency selective fading channels , '' _ ieee trans . mobile comput .
_ , vol . 1 , no
. 2 , pp . 132@xmath7141 , apr . 2002 .
h. zhang , x .-
xia , l. j. cimini , and p. c. ching , `` synchronization techniques and guard - band - configuration scheme for single - antenna vector - ofdm systems , '' _ ieee trans .
wireless commun .
5 , pp . 2454@xmath72464 , sep . 2005 . c. han , t. hashimoto , and n. suehiro , `` constellation - rotated vector ofdm and its performance analysis over rayleigh fading channels , '' _ ieee trans . commun .
828@xmath7838 , mar .
2010 . c. han and t. hashimoto , `` tight pep lower bound for constellation - rotated vector - ofdm under carrier frequency offset and fast fading , '' _ ieee trans .
62 , no . 6 , pp . 1931@xmath71943 , june 2014 .
i. ngebani , y. li , x .-
xia , s. a. haider , a. huang , and m. zhao , `` analysis and compensation of phase noise in vector ofdm systems , '' _ ieee trans . signal process .
6143@xmath76157 , dec . 2014 .
m. z. win and r. a. scholtz , `` characterization of an ultra - wideband wireless indoor channel : a communication - theoric view , '' _ ieee j. select areas commun .
1613@xmath71627 , dec . 2002 . c. r. berger , s. zhou , j. c. preisig , and p. willett , `` sparse channel estimation for multicarrier underwater acounstic communication : from subspace methods to compressed sensing , '' _ ieee trans .
signal process .
_ , vol 58 , no .
3 , pp . 1708@xmath71721 , mar . 2010 .
h. vikalo and b. hassibi , `` on the sphere - decoding algorithm ii .
generalizations , second - order statistics , and applications to communications , '' _ ieee trans . signal process .
2819@xmath72834 , aug .
2005 .
r. prasad , c. r. murthy , and b. d. rao , `` joint approximately sparse channel estimation and data detection in ofdm systems using sparse bayesian learning , '' _ ieee trans .
signal process .
3591@xmath73603 , july 2014 .
v. tarokh , n. seshadri , and a. r. calderbank , `` space - time codes for high data rate wireless communication : performance criterion and code construction , '' _ ieee trans .
inf . theory _
2 , pp . 744@xmath7765 , mar . 1998 . | vector orthogonal frequency division multiplexing ( v - ofdm ) is a general system that builds a bridge between ofdm and single - carrier frequency domain equalization in terms of intersymbol interference and receiver complexity . in this paper
, we investigate the sparse multipath channel estimation and decoding for broadband v - ofdm systems .
unlike the non - sparse channel estimation , sparse channel estimation only needs to recover the nonzero taps with reduced complexity .
we first consider a simple noiseless case that the pilot signals are transmitted through a sparse channel with only a few nonzero taps , and then consider a more practical scenario that the pilot signals are transmitted through a sparse channel with additive white gaussian noise interference . the exactly and approximately sparse inverse fast fourier transform ( sifft )
can be employed for these two cases .
the sifft - based algorithm recovers the nonzero channel coefficients and their corresponding coordinates directly , which is significant to the proposed partial intersection sphere ( pis ) decoding approach . unlike the maximum likelihood ( ml ) decoding that enumerates symbol constellation and estimates the transmitted symbols with the minimum distance , the pis decoding first generates the set of possible transmitted symbols and then chooses the transmitted symbols only from this set with the minimum distance .
the diversity order of the pis decoding is determined by not only the number of nonzero taps , but also the coordinates of nonzero taps , and the bit error rate ( ber ) is also influenced by vector block size to some extent but roughly independent of the maximum time delay .
simulation results indicate that by choosing appropriate sphere radius , the ber performance of the pis decoding outperforms the conventional zero - forcing decoding and minimum mean square error decoding , and approximates to the ml decoding with the increase of signal - to - noise ratio , but reduces the computational complexity significantly . vector orthogonal frequency division multiplexing , sparse multipath channel , sparse inverse fast fourier transform , partial intersection sphere decoding , diversity order . |
semiconductor physics and magnetism are established subfields of condensed matter physics that continue to reveal a rich variety of new phenomena , often in new types of solid state materials .
the properties of semiconductors are extraordinarily sensitive to impurity atoms , defects , and charges on external gates .
magnetism is a collective electronic phenomenon with an ordered state that is often stable to exceptionally high temperatures .
magnetic order , when it is present , has a large impact on other material properties including transport and optical properties . in both semiconductor and magnetic cases , sophisticated and economically important technologies have been developed to exploit the unique electronic properties , mainly for information processing in the case of semiconductors and for information storage and retrieval in the case of magnetism .
the realization of materials that combine semiconducting behavior with robust magnetism has long been a dream of material physics .
one strategy for creating systems that are simultaneously semiconducting and magnetic , initiated in the late 1970 s @xcite , is to introduce local moments into well - understood semiconductors .
the result is a new class of materials now known as diluted magnetic semiconductors ( dmss ) . over the past fifteen years , building on a series of pioneering publications in the 1990 s @xcite , it has been established that several ( iii , v ) compound semiconductors become ferromagnetic when heavily doped with mn , and that the ferromagnetic transition temperatures can be well above 100 k. in semiconductors like gaas and inas , mn has been shown to act both as an acceptor and as a source of local moments .
these ( iii , mn)v materials are examples of ferromagnetic semiconductors , a phrase we reserve for magnetic systems in which ferromagnetism is due primarily to coupling between magnetic element moments that is mediated by conduction - band electrons or valence - band holes .
this definition implies that in ferromagnetic semiconductors , magnetic properties can be influenced by the same assortment of engineering variables that are available for other more conventional semiconductor electronic properties . in the best understood arsenide dmss ,
semiconductor valence - band carriers participate in the magnetic order .
the materials require participation of valence band holes for the formation of a ferromagnetic state .
efforts to increase their critical temperatures further run into incompletely understood fundamental limits on the ratio of the magnetic transition temperature to the fermi temperature of the free - carrier systems and the role of disorder in these heavily doped materials .
the tension between achieving high curie temperatures and the desire for low , and therefore gateable , carrier densities , is among the major issues in the study of these materials . in this article
we review the considerable theoretical progress that has been made in understanding the very broad range of properties that occur in ( iii , mn)v ferromagnetic semiconductor epilayers in different regimes of mn content and defect density .
the main focus of this article is on the extensively studied ( ga , mn)as ferromagnetic semiconductor , but we also make frequent comments on other ( iii , mn)v dmss . comparisons to experimental data are made throughout the article . in section
i we review progress that has been achieved in the effort to realize useful dms materials for spintronics ( or magnetoelectronics ) . in section
ii we discuss the properties of dilute mn atoms in a ( iii , v ) crystal , and the various mechanisms that can couple the orientations of distinct moments and lead to ferromagnetism . in section
iii we discuss several different strategies that can be used to elevate material modeling from a qualitative to a more quantitative level .
sections iv - vii address a variety of different characteristics of ( iii , mn)v layers , including their structural , magnetic , magneto - transport , and magneto - optical properties . finally in section viii
we discuss the ferromagnetic ordering physics in ( iii , mn)v dmss in the broad context of magnetic interactions in systems with coupled local and itinerant moments , and then extrapolate from ( iii , mn)v materials to comment on the effort to find high temperature ferromagnetism in other dms materials .
we conclude in section ix with a brief summary .
to partially remedy omissions in the bibliography that originate from our incomplete coverage of this topic , we refer to an extended database of published work and preprints maintained at http://unix12.fzu.cz/ms .
the structure of the database is similar to the structure of this review and we encourage the reader in need of a more detailed bibliography to use this resource .
a number of review articles on various aspects of the physics of dmss have been published previously and may help the reader who seeks a broader scope than we are able to supply in this review .
the extensive body of research on dmss in the 1980 s , focused mostly on ( ii , mn)vi alloys , is reviewed in @xcite .
several extended papers cover the experimental properties of ( iii , mn)v dmss , particularly ( ga , mn)as and ( in , mn)as , interpreted within the carrier - mediated ferromagnetism model @xcite .
theoretical predictions based on this model for a number of properties of bulk dmss and heterostructures are reviewed in @xcite .
a detailed description of wide band gap and oxide dmss can be found in @xcite .
we also mention here several specialized theoretical reviews focusing on the predictions of density functional first principles calculations for ( iii , mn)v dmss @xcite , on mn - doped ii - vi and iii - v dmss in the low carrier density regime @xcite , and on effects of disorder in ( ga , mn)as @xcite . iii - v materials are among the most widely used semiconductors .
there is little doubt that ferromagnetism in these materials would enable a host of new microelectronics device applications if the following criteria were met : i ) the ferromagnetic transition temperature should safely exceed room temperature , ii ) the mobile charge carriers should respond strongly to changes in the ordered magnetic state , and iii ) the material should retain fundamental semiconductor characteristics , including sensitivity to doping , electric fields produced by gate charges , and sensitivity to light . for more than a decade these three key issues have been the focus of intense experimental , and theoretical research into the material properties of mn - doped iii - v compounds . at first sight ,
fundamental obstacles appear to make the simultaneous achievement of these objectives unlikely .
nevertheless , interest in this quest remains high because of the surprising progress that has been achieved .
highlights of this scientific endeavor are briefly reviewed in this introductory chapter . under equilibrium growth conditions the incorporation of magnetic mn ions into iii - as
semiconductor crystals is limited to approximately 0.1% . beyond this doping level ,
surface segregation and phase separation occur . to circumvent the solubility problem a non - equilibrium ,
low - temperature molecular - beam - epitaxy ( lt - mbe ) technique was applied and led to first successful growth of ( in , mn)as and ( ga , mn)as dms ternary alloys with more than 1% mn .
since the first report in 1992 of a ferromagnetic transition in @xmath0-type ( in , mn)as at a critical temperature @xmath1 k @xcite , the unfolding story of critical temperature limits in ( iii , mn)v dmss has gone through different stages .
initial experiments in ( in , mn)as suggested an intimate relation between the ferromagnetic transition and carrier localization , reminiscent of the behavior of manganites ( perovskite ( la , a)mno@xmath2 with a = ca , sr , or ba ) in which ferromagnetism arises from a zener double exchange process associated with d - electron hopping between mn ions @xcite .
( we comment at greater depth on qualitative pictures of the ferromagnetic coupling in sections [ micro - general ] and [ discussion - general ] . )
this scenario was corroborated by a pioneering theoretical ab - initio study of the ( in , mn)as ferromagnet @xcite and the mechanism was also held responsible for mediating ferromagnetic mn - mn coupling in some of the first ferromagnetic ( ga , mn)as samples with @xmath3 s close to 50 k @xcite . in 1998
the tohoku university group announced a jump of @xmath3 in @xmath0-type ( ga , mn)as to 110 k @xcite and pointed out that the critical temperature value was consistent with the kinetic - exchange mechanism for ferromagnetic coupling , also first proposed by zener ( see section [ micro - general ] ) . in its simplest form , ferromagnetism in this picture
follows @xcite from ruderman - kittel - kasuya - yosida ( rkky ) indirect coupling between mn @xmath4-shell moments mediated by induced spin - polarization in a free - hole itinerant carrier system ( see sections [ micro - general ] , [ bulk - mag - tc - sw ] and [ discussion - general ] ) .
zener proposed this mechanism originally for transition metal ferromagnets for which the applicability of this picture is now known to be doubtful because of the itinerant character of transition metal @xmath4-electrons .
the model of mn(@xmath5 ) local moments that are exchange - coupled to itinerant @xmath6-band carriers does however provide a good description of mn - doped iv - vi and ii - vi dmss @xcite . the key difference between ( iii , mn)v materials like ( ga ,
mn)as and iv - vi and ii - vi dmss is that mn substituting for the trivalent cation ( ga ) is simultaneously an acceptor and a source of magnetic moments .
theoretical critical temperature calculations based on the kinetic - exchange model predict room temperature ferromagnetism in ( ga , mn)as with 10% mn content . in spite of these optimistic predictions , the goal of breaking the 110 k record in ( ga , mn)as remained elusive for nearly four years . only recently has progress in mbe growth and in the development of post - growth annealing techniques @xcite made it possible to suppress extrinsic effects , pushing @xmath3 in ( ga , mn)as up to 173 k @xcite .
@xmath3 trends in current high quality ( ga , mn)as epilayers are consistent with the zener kinetic - exchange model @xcite . the current @xmath3 record should be broken if dms material with a higher concentration of substitutional mn ions can be grown .
based on the few experimental and theoretical studies reported to date , ( iii , mn)sb dmss are expected to fall into the same category as ( ga , mn)as and ( in , mn)as dms .
the kinetic - exchange model calculations predict @xmath3 s that are small compared to their arsenide counterparts @xcite .
this difference , confirmed by experiment @xcite , is caused by the weaker @xmath7 exchange and smaller magnetic susceptibility ( smaller effective mass ) of itinerant holes in the larger unit - cell antimonides .
also consistent with the kinetic - exchange model is the remarkable observation of an increase of @xmath3 in ( in , mn)sb by 25% induced by the applied hydrostatic pressure @xcite . moving in the opposite direction in the periodic table towards ( iii , mn)p and ( iii
, mn)n appears to be the natural route to high @xmath3 ferromagnetic semiconductors .
the kinetic - exchange model predicts @xmath3 s far above room temperature in these smaller lattice constant materials , in particular in ( ga , mn)n @xcite . also the solubility limit of mn is much larger than in arsenides , making it possible in principle to grow highly mn - doped dmss under or close to equilibrium conditions .
however , the nature of the magnetic interactions in mn - doped phosphides and nitrides is not completely understood either theoretically or experimentally @xcite .
as the valence band edge moves closer to the mn @xmath4-level and the @xmath7 hybridization increases with increasing semiconductor gap width and decreasing lattice constant , charge fluctuations of the @xmath4-states may become large @xcite . with increasing ionicity of the host crystal , the mn impurity may also undergo a transition from a @xmath5 divalent acceptor to a @xmath8 trivalent neutral impurity @xcite . in either case
, the picture of ferromagnetism based on the zener kinetic - exchange model needs to be revisited in these materials .
experimental critical temperatures close to 1000 k have been reported in some ( ga , mn)n samples @xcite .
it is still unclear , however , whether the high - temperature ferromagnetic phase should be attributed to a ( ga , mn)n ternary alloy or to the presence of ferromagnetic metal precipitates embedded in the host gan lattice .
reports of ( ga , mn)n epilayers synthesized in cubic and hexagonal crystal structures , of p - type and n - type ferromagnetic ( ga , mn)n , and of multiple ferromagnetic phases in one material all add to the complex phenomenology of these wide - gap dmss @xcite .
uncertainties apply also to the interpretation of ferromagnetism seen in the ( ga , mn)p samples studied to date , which have been prepared by post mbe ion - implantation of mn followed by rapid thermal @xcite or pulse laser melting annealing @xcite . experiments in these materials have not yet established unambiguously the nature of magnetic interactions in the ( iii , mn)p compounds . however ,
a comparative study of ( ga , mn)p and ( ga , mn)as prepared by the post mbe ion - implantation and pulse laser melting annealing suggests carrier mediated origin of ferromagnetism in the ( ga , mn)p material @xcite . our current understanding of the material physics of ( iii , mn)v dms epilayers suggests that synthesis of room temperature ferromagnetic semiconductor will require a level of doping and defect control comparable to what has now been achieved in high quality ( ga , mn)as samples , mn densities of order 10% , and may require the use of wider gap iii - v alloys .
finally , we note that efforts to enhance curie temperature in mn - doped ( iii , v ) semiconductors have also led to material research in more complex semiconductor heterostructures with highly mn - doped monolayers ( @xmath9-doped layers ) , showing promising results @xcite .
spintronic devices exploit the electron spin to manipulate the flow of electrons and therefore require materials in which the charge and spin degrees of freedom of carriers are strongly coupled @xcite .
the most robust , and currently the most useful , spintronic devices rely on the collective behavior of many spins in ferromagnetic materials to amplify the coupling of external magnetic fields to electronic spins , a coupling that is very weak for individual electrons .
the intrinsically large spin - orbit interaction in iii - sb and iii - as valence band states makes these hosts ideal candidates for exploring various spintronic functionalities . in ( ga ,
mn)as dms epilayers , for example , the measured anisotropic magneto - resistance ( amr ) effect ( the relative difference between longitudinal resistivities for different magnetization orientations ) can reach @xmath10% @xcite .
a particularly strong manifestation of valence - band spin - orbit coupling occurs in the antisymmetric off - diagonal element of the resistivity tensor .
the anomalous hall effect ( ahe ) shown in fig .
[ ahe_nott ] , which completely dominates the low - field hall response in ( ga , mn)as and some other iii - v dmss , has become one of the key tools used to detect the paramagnetic / ferromagnetic transition @xcite . its large value is due to the spin - polarization of holes and provides strong evidence for the participation of mobile charge carriers in the ordered magnetic state of these dmss . in metals ,
the current response to changes in the magnetic state is strongly enhanced in layered structures consisting of alternating ferromagnetic and non - magnetic materials .
the giant magneto - resistance effect @xcite which is widely exploited in current technology , for example in field - sensors and magnetic random - access memories , reflects the large difference between resistivities in configurations with parallel and antiparallel polarizations of ferromagnetic layers in magnetic superlattices or trilayers like spin - valves and magnetic tunnel junctions @xcite .
the effect relies on transporting spin information between layers and therefore is sensitive to spin - coherence times in the system . despite strong spin - orbit coupling which reduces spin coherence in dmss ,
functional spintronic trilayer devices can be built , as demonstrated by the measured large mr effects in ( ga , mn)as based tunneling structures @xcite .
the coercivities of individual dms layers can be tuned via exchange - biasing to an antiferromagnet @xcite which is a standard technique used in metal giant - magnetoresistance devices @xcite .
dms ferromagnets possess all the properties that are exploited in conventional spintronics .
they qualify as ferromagnetic semiconductors to the extent that their magnetic and other properties can be altered by the usual semiconductor electronics engineering variables .
the achievement of ferromagnetism in an ordinary iii - v semiconductor that includes several per cent of mn demonstrates on its own the sensitivity of magnetic properties to doping .
remarkably , doping profiles and , correspondingly magnetic properties can be grossly changed , even after growth , by annealing .
early studies of ( ga , mn)as indicated that annealing at temperatures above the growth temperature leads to a reduction of magnetically and electrically active mn ions and , at high enough annealing temperatures , to the formation of mnas clusters @xcite . on the other hand ,
annealing at temperatures below the growth temperature can substantially improve magnetic and transport properties of the thin dms layers due to the out - diffusion of charge and moment compensating defects , now identified as interstitial mn @xcite .
( in , mn)as based field effect transistors were built to study electric field control of ferromagnetism in dmss .
it has been demonstrated that changes in the carrier density and distribution in thin film dms systems due to an applied bias voltage can reversibly induce the ferromagnetic / paramagnetic transition @xcite .
another remarkable effect observed in this magnetic transistor is electric field assisted magnetization reversal @xcite .
this novel functionality is based on the dependence of the width of the hysteresis loop on bias voltage , again through the modified charge density profile in the ferromagnetic semiconductor thin film .
experiments in which ferromagnetism in a ( iii , mn)v dms system is turned on and off optically add to the list of functionalities that result from the realization of carrier - induced ferromagnetism in a semiconductor host material @xcite .
the observed emission of circularly polarized light from a semiconductor heterostructure , in which electrons ( holes ) injected from one side of the structure recombine with spin - polarized holes ( electrons ) emitted from a dms layer @xcite , is an example of phenomena that may lead to novel magneto - optics applications .
tunneling anisotropic magnetoresistance ( tamr ) is another novel spintronic effect observed in ( ga , mn)as @xcite .
tamr , like amr , arises from spin - orbit coupling and reflects the dependence of the tunneling density of states of the ferromagnetic layer on the orientation of the magnetization with respect to the current direction or the crystallographic axes .
the larger characteristic electronic length scales in dmss compared to ferromagnetic metals make it possible to lithographically define lateral structures with independent magnetic areas coupled through depleted regions that act as tunnel barriers and magnetic weak links .
the electrical response to magnetization reversals in these spintronic nanodevices can lead to mr effects with magnitudes of order 1000% @xcite , as shown in fig .
[ point_cont ] , and with a rich phenomenology @xcite .
wider lateral constrictions have been used to demonstrate controlled domain - wall nucleation and propagation in dms stripes @xcite , a prerequisite for developing semiconductor logic gates based on magnetic domain manipulation @xcite .
( ga , mn)as nanoconstrictions with lateral side gates have revealed a new effect , coulomb blockade anisotropic magnetoresistance , which reflects the magnetization orientation dependence of the single - electron charging energy @xcite .
these spintronic single - electron transistors offer a route to non - volatile , low - field , and highly electro- and magneto - sensitive operation .
the magnetic dipole - dipole interaction strength between two discrete moments separated by a lattice constant in a typical solid is only @xmath11 1 kelvin , relegating direct magnetic interactions to a minor role in the physics of condensed matter magnetic order .
relativistic effects that lead to spin - orbit coupling terms in the hamiltonian provide more plausible source of phenomena that are potentially useful for spintronics .
although these terms are critical for specific properties like magnetic anisotropy , they are rarely if ever crucial for the onset of the magnetic order itself .
instead the universal ultimate origin of ferromagnetism is almost always the interplay between the electronic spin degree of freedom , repulsive coulomb interactions between electrons , and the fermionic quantum statistics of electrons .
the pauli exclusion principle correlates spin and orbital parts of the many - electron wavefunction by requiring the total wavefunction to be antisymmetric under particle exchange .
whenever groups of electrons share the same spin state , the orbital part of the many - body wavefunction is locally antisymmetric , lowering the probability of finding electrons close together and hence the interaction energy of the system . because magnetic order is associated with the strong repulsive coulomb interactions between electrons
, it can persist to very high temperatures , often to temperatures comparable to those at which crystalline order occurs .
ferromagnetism can be as strong as chemical bonds .
very often the quantum ground state of a many electron system has non - zero local spin - density , either aligned in the same direction in space at every point in the system as in simple ferromagnets , or in non - collinear , ferrimagnetic , or antiferromagnetic materials in configurations in which the spin direction varies spatially .
although this statement on the origin of magnetic order has very general validity , its consequences for a system of nuclei with a particular spatial arrangement , are extraordinarily difficult to judge . because ferromagnetism is a strong - coupling phenomenon , rigorous theoretical analyses are usually not possible .
there is no useful universal theory of magnetism .
understanding magnetic order in a particular system or class of systems can be among the most challenging of solid state physics problems @xcite .
for most systems , it is necessary to proceed in a partially phenomenological way , by identifying the local spins that order , and determining the magnitude and sign of the exchange interactions that couple them by comparing the properties of simplified ( often ` spin - only ' ) model hamiltonians with experimental observations .
one approach that is totally free of phenomenological parameters is density functional theory ( dft ) , including its spin - density functional ( sdf ) generalizations in which energy functionals depend on charge and spin densities .
although dft theory is exact in principle , its application requires that the formalism s exchange - correlation energy functional be approximated .
approximate forms for this functional can be partially phenomenological ( making a pragmatic retreat from the _ ab initio _ aspiration of this approach ) and are normally based in part on microscopic calculations of correlation effects in the electron gas model system .
this is the case for the often - used local ( spin ) density approximation ( l(s)da ) @xcite . for many magnetic metals ,
in which correlations are somewhat similar to those in the electron gas model system , _ ab initio _ lsda theory provides a practical and sufficiently accurate solution of the magnetic many - body problem @xcite .
this is particularly true for elemental transition metal ferromagnets fe , co , and ni and their alloys @xcite . in practice lsda theory functions as a mean - field theory in which the exchange - energy at each point in space increases with the self - consistently determined local spin density . with increasing computer power ,
lsda theory has been applied to more complex materials , including dmss . as we discuss below , both phenomenological and dft approaches provide valuable insight into ( iii , mn)v ferromagnetism .
model hamiltonian theories are likely to remain indispensable because , when applicable , they provide more transparent physical pictures of ferromagnetism and often enable predictions of thermodynamic , transport , and other properties that are sometimes ( depending on material complexity ) , beyond the reach of _ ab initio _ theory techniques .
of particular importance for dmss is the capability of model hamiltonians to describe localized electronic levels coincident with an itinerant - electron band which , strictly speaking , is beyond the reach of the effectively one - particle band theories of solids that emerge from lsda theory @xcite .
interpreting experiments with model hamiltonian approaches can , on the other hand , be misleading if the model is too simplified and important aspects of the physics are absent from the model .
what s more , even simplified models usually leave complex many - body problems that can not be completely solved .
magnetism in ( ga , mn)as and some other ( iii , mn)v ferromagnets originates from the mn local moments .
( as already pointed out , mn - doped phosphides and nitrides dmss are less well understood , however , local mn moments are likely to play an important role in these materials as well . ) the dependence of the energy of the system on the relative orientation of mn moments is generally referred to as an _ exchange _ interaction .
this terminology is part of the jargon of magnetism and recognizes that fermi statistics is the ultimate origin .
several types of qualitative effects that lead to exchange interactions can be separately identified when addressing magnetic order in ( iii , mn)v dmss ; the applicability of each and the relative importance of different effects may depend on the doping regime and on the host semiconductor material . in this section
we first introduce the jargon that is commonly used in the magnetism literature , by briefly reviewing some of the effects that can lead to magnetic coupling , most of which have been recognized since near the dawn of quantum mechanics . for spins carried by itinerant electrons ,
exchange interactions are often most simply viewed from a momentum space rather than a real space point of view .
stoner s _ itinerant - exchange _
@xcite favors spontaneous spin - polarization of the entire electron gas because electrons are less likely to be close together and have strongly repulsive interactions when they are more likely to have the same spin .
because the band energy is minimized by double - occupation of each bloch state , the stoner ferromagnetic instability occurs in systems with a large density of states at the fermi energy .
this helps to explain , for example , why ferromagnetism occurs in the late _ 3d _ transition elements .
a large density of states makes it possible to gain exchange energy by moving electrons from one spin - band to the other while keeping the kinetic energy cost sufficiently low .
since the key spins in many ( iii , mn)v dms materials are localized the stoner mechanism does not drive ferromagnetism , although we will see in section [ bulk - mag - tc ] that it still plays a minor supporting role . in many systems , including ( iii , mn)v dmss , both the local nature of the moments and strong local coulomb interactions that suppress charge ( valence ) fluctuations play a key role and have to be included even in a minimal model .
many mechanisms have been identified that couple localized spins in a solid .
the origin of heisenberg s _ direct - exchange _ @xcite between two local spins is the difference between the coulomb energy of a symmetric orbital wavefunction ( antisymmetric singlet spin wavefunction ) state and an antisymmetric orbital wavefunction ( symmetric triplet spin wavefunction ) state .
kramer s _ super - exchange _
interaction @xcite , applies to local moments that are separated by a non - magnetic atom . in a crystal environment
, an electron can be transferred from the non - magnetic atom to an empty shell of the magnetic atom and interact , via direct exchange , with electrons forming its local moment .
the non - magnetic atom is polarized and is coupled via direct - exchange with all its magnetic neighbors . whether the resulting super - exchange interaction between the local moments is ferromagnetic or antiferromagnetic depends on the relative sign of the two direct - exchange interactions @xcite . in ( iii , mn)v materials , superexchange gives an antiferromagnetic contribution to the interaction between mn moments located on neighboring cation sites .
zener s _ double - exchange _
mechanism @xcite also assumes an intermediate non - magnetic atom . in its usual form , this interaction occurs when the two isolated magnetic atoms have a different number of electrons in the magnetic shell and hopping through the intermediate non - magnetic atom involves magnetic - shell electrons .
combined with the on - shell hund s rule , double - exchange couples the magnetic moments ferromagnetically .
parallel spin alignment is favored because it increases the hopping probability and therefore decreases the kinetic energy of the spin - polarized electrons .
a version of double - exchange , in which mn acceptor states form an impurity band with a mixed @xmath12-character , has often been referred to in the ( iii , mn)v literature . in this picture electrical conduction and mn - mn exchange coupling
are both realized through hopping within an impurity band .
the potential importance of double - exchange is greater at lower mn doping and in wider - gap ( iii , mn)v materials . finally , we identify zener s _ kinetic - exchange _ @xcite or indirect exchange interaction .
it arises in models with local , usually @xmath4-shell or @xmath13-shell , moments whose coupling is mediated by s- or p - band itinerant carriers .
the local moments can have either a ferromagnetic direct exchange interaction with band electrons on the same site and/or an antiferromagnetic interaction due to hybridization between the local moment and band electrons on neighboring sites @xcite .
polarization of band electrons due to the interaction at one site is propagated to neighboring sites . when the coupling is weak ( the band carrier polarization is weak , _
e.g. _ , at temperatures near curie temperature ) , the effect is described by rkky theory which was originally applied to carrier mediated indirect coupling between nuclear moments @xcite and between local @xmath4-shell moments in metals @xcite .
the range of this interaction can be long and the interactions between separate local moments can be either ferromagnetic or antiferromagnetic and tend to vary in space on the length scale of the itinerant band s fermi wavelength . unlike the double - exchange case , magnetic order in this case does not lead to a significant change in the width of the conducting band .
this type of mechanism certainly does play a role in ( iii , mn)v ferromagnetism , likely dominating in the case of strongly metallic ( ga , mn)as , ( in , mn)as , and mn - doped antimonides .
there is no sharp distinction between impurity band double - exchange and kinetic - exchange interactions ; the former is simply a strong coupling , narrow band limit of the latter . the starting point for developing a useful predictive model of ( iii , mn)v ferromagnetism is achieving a full understanding of the electronic state of a single mn impurity in the host lattice .
we need to fully understand the character of the isolated local moments before we can critically discuss how they are coupled .
the character of the local moment need not be the same in all ( iii , mn)v materials .
the remaining subsections will focus on properties of a mn impurity in gaas and on the nature of ferromagnetic coupling in ( ga , mn)as and related arsenide and antimonide dmss . at the end of this section
, we comment on how things might change in wider - gap hosts like gap and gan . among all ( iii , v ) hosts ,
mn impurity has been studied most extensively in gaas .
the elements in the ( ga , mn)as compound have nominal atomic structures [ ar]@xmath14 for ga , [ ar]@xmath15 for mn , and [ ar]@xmath16 for as .
this circumstance correctly suggests that the most stable and , therefore , most common position of mn in the gaas host lattice is on the ga site where its two @xmath17-electrons can participate in crystal bonding in much the same way as the two ga @xmath17-electrons .
the substitutional mn@xmath18 , and the less common interstitial mn@xmath19 , positions are illustrated in fig . [ mn_ga ] . because of the missing valence @xmath20-electron , the mn@xmath18 impurity acts as an acceptor . in the electrically neutral state , labeled as @xmath21 , mn@xmath18 has the character of a local moment with zero angular momentum and spin @xmath22 ( lnde g - factor @xmath23 ) and a moderately bound hole .
the local moment is formed by three occupied @xmath24 bonding states with dominant @xmath25 ( @xmath26 , @xmath27 , @xmath28 ) character and by two occupied @xmath29 ( @xmath30 , @xmath31 ) orbitals that are split from the @xmath25 states by the tetrahedral crystal field and do not strongly hybridize with the @xmath6-orbitals .
all occupied @xmath4-orbitals have the same spin orientation and together comprise the @xmath22 local moment .
the weakly bound hole occupies one of the three antibonding @xmath24 levels with dominant as 4@xmath0 character .
the charge @xmath32 ionized mn@xmath18 acceptor center , labeled as @xmath33 , has just the @xmath22 local spin character .
electron paramagnetic resonance ( epr ) and ferromagnetic resonance ( fmr ) experiments confirm the presence of the @xmath33 center through the entire range of mn concentrations in both bulk and epilayer ( ga , mn)as @xcite . the @xmath22 local moment on mn was detected through a resonance line centered at @xmath23 and , in low mn - density samples , through a sextet splitting of the line due to the hyperfine interaction with the @xmath34 @xmath35mn nuclear spin .
the neutral mn@xmath18 centers are more elusive because of nearly full compensation by unintentional donor impurities at low mn concentrations and because of the metal - insulator transition at high mn concentrations .
nevertheless , a multitude of experimental techniques , including epr @xcite , infrared ( ir ) spectroscopy @xcite , and magnetization measurements @xcite , have detected the @xmath21 center in ( ga , mn)as . strikingly direct evidence was given by an stm experiment @xcite , shown in fig .
[ mn_stm ] , in which the state of a single impurity atom was switched between the ionized @xmath33 and the neutral @xmath21 by applying a bias voltage that corresponded to a binding energy @xmath36 ev .
the binding energy @xmath37 mev inferred from ir spectroscopy @xcite is consistent with the above stm measurement and with inferences based on photoluminescence experiments @xcite .
these observations identify mn as a moderately shallow acceptor in gaas whose band gap is @xmath38 ev . the binding energy , which governs the electrical behavior of the mn impurity . has contributions from both coulomb attraction between the hole and the @xmath33 core ( and a central cell correction ) and spin - dependent @xmath7 hybridization .
the latter effect is responsible for the exchange interaction on which this review centers .
we now discuss this parameter in more detail .
the top of the gaas valence band is dominated by @xmath20-levels which are more heavily weighted on as than on ga sites .
direct exchange between the holes near the top of the band and the localized mn @xmath4-electrons is weak since mn@xmath18 and as belong to different sublattices .
this fact allows @xmath7 hybridization to dominate , explaining the antiferromagnetic sign of this interaction @xcite seen in experiment @xcite .
there is a simple physical picture of the @xmath7 exchange interaction which applies when interactions are treated in a mean - field way , and therefore also applies as an interpretation of lsda calculations .
given that the filled , say spin - down , mn @xmath4-shell level is deep in the valence band and that the empty spin - up @xmath4-level is above the fermi level and high in the conduction band , hybridization ( level repulsion of like - spin states ) pushes the energy of spin - down valence band states up relative to the energy of spin - up valence band states .
the resulting antiferromagnetic coupling between valence band states and local mn spins is illustrated schematically in fig .
[ hybrid ] .
the same basic picture applies for itinerant valence band states in a heavily - doped metallic dms and for the acceptor state of an isolated mn@xmath18 impurity .
note that the cartoon band structure in fig .
[ hybrid ] is plotted in the electron picture while the dms literature usually refers to the antiferromagnetic @xmath7 coupling between _
holes _ and local mn moments .
we comment in detail on the equivalent notions of @xmath7 exchange in the physically direct electron - picture and the computationally more convenient hole - picture for these @xmath0-type dmss in section [ bulk - mag - magnet ] .
we have already mentioned in section [ micro - general ] the conceptional inadequacy of effective single - particle theories , including lsda , in dealing with local moment levels coincident with itinerant electron bands .
anderson suggested a many - body model hamiltonian that circumvents this problem by including explicitly the coulomb correlation integral of the localized electron states in the hamiltonian @xcite .
the problem is that the change in the effective potential when the number of occupied localized orbitals changes by one , the hubbard constant @xmath39 , can be comparable to or larger than other band parameters , invalidating any mean - field - like approach .
the consequences of this fact can be captured at a qualitative level in models that include the hubbard @xmath39 @xcite . in these phenomenological models ,
the localized orbital part of the hamiltonian generally has an additional parameter , the hund s rule constant @xmath40 .
this parameter captures the local direct exchange physics which favors spin - polarized open shell atomic states . for the case of the mn(@xmath41 ) configuration ,
@xmath40 forces all five singly - occupied @xmath4-orbitals to align their spins in the ground state .
recently considerable effort has been devoted to developing approaches that combine the local correlation effects induced by the @xmath39 and @xmath40 terms in phenomenological models , with sdf theory @xcite .
we comment on these _ ab initio _ techniques in section [ theory - first ] .
when hybridization @xcite between the local moment and band electron states is weak it can be treated perturbatively .
the schrieffer - wolff transformation of the anderson hamiltonian , @xmath42 removes the hybridization ( last term in eq .
( [ anderson ] ) and leads to a model in which the local moment spin interacts with the valence band via a spin - spin interaction only , @xmath43 , with the number of electrons in each band fixed @xcite . here
we assume for simplicity a single localized orbital and a single itinerant band , we use @xmath44 to represent band states and @xmath4 to represent the localized impurity state , @xmath45 labels spin , @xmath46 a single particle energy , and @xmath47 and @xmath48 are the standard second quantization operators .
this procedure is normally useful only if the hybridization is relatively weak , in which case it is not usually a problem to _ forget _ that the canonical transformation should also be applied to operators representing observables . strictly speaking
, the schrieffer - wolf transformation also leads to a spin - independent interaction @xcite which is normally neglected in comparison with the stronger spin - independent long - range part of the coulomb potential .
since the valence band states of interest in gaas , even in heavily doped samples , are near the brillouin - zone center @xmath49 , the single phenomenological constant extracted from experiment for this interaction should be thought of as its value when both initial and final states are at the @xmath49-point .
the quantity @xmath50 is then approximated by @xmath51 , where @xmath52 @xmath53 represents the as @xmath0-orbital
mn @xmath4-orbital hybridization potential ( neglecting again the multiple orbital nature of the @xmath0 and @xmath4-levels for simplicity ) , @xmath54 is the single - particle atomic - level energy of the occupied mn @xmath4-state measured from the top of the valence band , and @xmath55 is the energy cost of adding a second electron to this orbital @xcite . when the hole density is large or the holes are more strongly localized near mn acceptors , the crystal - momentum dependence of this interaction parameter can not be entirely neglected @xcite .
since @xmath56 originates from hopping between mn and the neighboring as atoms , the @xmath7 exchange potential , @xmath57 , produced by mn impurity at site @xmath58 has a range of order one lattice constant , and @xmath59 in eq .
( [ j0jr ] ) we assumed that the perfect crystal bloch function , @xmath60 , is composed of a slowly varying envelope function and a periodic function @xmath61 with the normalizations @xmath62 and @xmath63 . here @xmath64 is the crystal volume and @xmath65 is the unit cell volume . in gaas , @xmath66=0.045 nm@xmath67 and the lattice constant @xmath68 nm . these wavefunctions can be obtained from @xmath69 theory which treats the band hamiltonian of the system perturbatively expanding around the @xmath49-point @xcite .
@xmath56 in eq .
( [ j0jr ] ) corresponds to the average value of @xmath57 seen by the @xmath70 bloch state over the @xmath71-th unit cell .
the slowly varying envelope function sees an effective zero - range @xmath7 exchange potential , since @xmath72 j({\bf r}_i -{\bf r } ) u^{\ast}_{n,{\bf k}}({\bf r } ) u_{n^{\prime},{\bf k}^{\prime}}({\bf r})\nonumber \\ & \approx & \frac{1}{v}\sum_{j=1}^{n_{u.c . }
} \exp\left[i({\bf k}- { \bf k^{\prime}}){\bf r}_j\right ] \int_{j - u.c.}^{\ , } d{\bf r } j({\bf r}_i -{\bf r } ) u^{\ast}_n({\bf r } ) u_n({\bf r})\nonumber \\ & = & \frac{1}{v}\sum_{j=1}^{n_{u.c . } } \exp\left[i({\bf k}- { \bf k^{\prime}}){\bf r}_j\right ] j_0\omega_{u.c.}\delta_{i , j}\nonumber \\ & \approx & \frac{1}{v}\int d{\bf r}\exp\left[i({\bf k}- { \bf k^{\prime}}){\bf r}\right ] j_{pd}\delta({\bf r}_i-{\bf r})\ ; , \end{aligned}\ ] ] where @xmath73 is the number of unit cells in the crystal volume @xmath64 and @xmath74 .
note that the assumption of zero - range in the @xmath69 treatment corresponds to the neglect of momentum dependence of this interaction for wavevectors close to the @xmath49-point in a microscopic picture .
positive sign of the @xmath75 ( or @xmath56 ) constant corresponds to an antiferromagnetic @xmath7 coupling .
we also note that in the ii - vi dms literature a symbol @xmath76 is often used instead of @xmath75 and @xmath77 instead of @xmath56 , where @xmath78 is the number of unit cells ( or cation sites ) per volume @xcite . the value of the @xmath75 constant is often considered to be independent of the host semiconductor @xcite .
indeed the increase of @xmath79 @xcite in eq .
( [ anderson ] ) with decreasing lattice constant is partly compensated by smaller @xmath65 ( @xmath80 ) , and the increase of @xmath81 in a wider gap host is partly compensated by the decrease of the term @xmath82 .
although it may have similar values in many materials , @xmath75 will tend to be larger in larger gap , smaller lattice constant hosts . in a virtual crystal mean - field approximation
, the @xmath7 exchange potential due the mn impurities in a ga@xmath83mn@xmath84as dms , @xmath85 , has the periodicity of the host crystal .
( here @xmath86 is the mean - field mn spin . )
the valence band states in this approximation experience an effective single - particle kinetic - exchange field , @xmath87 , where @xmath88 is the mn@xmath18 density . finally , we discuss the relationship between the exchange constant @xmath75 and the exchange constant , @xmath89 , used to provide a @xmath69 interpretation of spectroscopic studies of the neutral @xmath21 center .
we emphasize that the use of a @xmath69 approach assumes that the bound hole is spread over at least several lattice constants in each direction .
the fact that it is possible to achieve a reasonably consistent interpretation of the very detailed spectroscopic data in this way is in itself strong support for the validity of this assumption .
the coupling between the weakly bound hole moment @xmath90 and the local spin @xmath91 of the mn@xmath18 core is expressed in the form @xmath92 , where @xmath93 , @xmath94 is the ( atomic scale ) total angular momentum operator of the band - hole at the @xmath49-point ( @xmath95 or 1/2 for the as @xmath20-orbitals forming the band states near @xmath70 ) , and @xmath96 is the additional ( hole binding radius scale ) angular momentum acquired by the hole upon binding to the mn@xmath18 impurity . the ir spectroscopy data @xcite have been analyzed @xcite within a spherical approximation , _
i.e. _ , considering only the @xmath97 , @xmath98-like bound state .
( note that a sizable anisotropic @xmath4-like component in the bound - hole ground state has been identified in the analysis of the stm data @xcite . )
further simplification is achieved by the neglect of the admixture of the two @xmath99 ( @xmath100 ) @xmath49-point states which is justified by the large spin - orbit splitting , @xmath101 mev , of these states from the two heavy - hole states ( @xmath95 , @xmath102 ) and the two light - hole states ( @xmath95 , @xmath100 ) .
writing the ground - state wavefunction in the form @xmath103 with a spherically symmetric envelope function @xmath104 and for @xmath105 ( @xmath95 ) , the expectation value of the exchange potential reads @xmath106 where @xmath107 is the mean value of the square of the slowly varying envelope function within the unit cell containing the mn@xmath18 impurity .
( [ expjpd ] ) implies that @xmath108 _
i.e. _ , the ratio between @xmath89 and @xmath75 is determined by the strength of the binding of the hole to the @xmath33 mn@xmath18 core and is larger for more localized holes .
a combination of ir data and theoretical calculations has been used to analyze this in more detail @xcite .
first of all , the value of the g - factor @xmath109 of the neutral @xmath21 mn@xmath110 complex obtained from ir spectroscopy measurements is in agreement with the theoretical value expected for the total angular momentum state @xmath111 of the complex , confirming the antiferromagnetic character of the @xmath7 coupling between the hole and the local mn spin .
the contribution of the @xmath7 potential to the binding energy is then given by @xmath112/2 $ ] which for the @xmath113 ground state gives @xmath114 .
the ir spectroscopy measurement of the splitting @xmath115 between the @xmath113 and @xmath116 states gives @xmath117 mev , _
i.e. _ , the contribution to the binding energy from the @xmath7 interaction is approximately 26.25 mev . the remaining binding energy , 112.4 - 26.25=86.15 mev , is due to the central field potential of the impurity . bhattacharjee and
benoit @xcite used the hydrogenic - impurity model with a screened coulomb potential and a central cell correction , whose strength was tuned to reproduce the value 86.15 mev , to obtain a theoretical estimate for @xmath118 nm@xmath119 . from eq .
( [ epsjpd ] ) they then obtained @xmath120 mev nm@xmath67 . given the level of approximation used in the theoretical description of the @xmath21 state , this value is in a reasonably good agreement with the exchange constant value @xmath121 mev nm@xmath67 ( @xmath122 ev ) inferred from photoemission data @xcite .
we note here that photoemission spectroscopy @xcite has represented one of the key experimental tools to study the properties of mn impurity in dmss , in particular , the position of the mn @xmath4-level and the strength and sign of the @xmath7 coupling .
an indirect measurement of the @xmath75 constant is performed by fitting the photoemission data to a theoretical spectrum of an isolated mnas@xmath123 cluster @xcite .
this procedure is justified by the short - range character of the @xmath7 exchange interaction .
most of the single mn - impurity spectroscopic studies mentioned above were performed in samples with doping levels @xmath124 for which the ga@xmath83mn@xmath84as random alloy can be grown under equilibrium conditions . in these materials
mn can be expected to occupy almost exclusively the low energy ga - substitutional position .
ferromagnetism , however , is observed only for @xmath125 which is well above the equilibrium mn solubility limit in gaas and , therefore , requires a non - equilibrium growth technique ( in practice low - temperature mbe ) to avoid mn precipitation .
the price paid for this is the occurrence of a large number of metastable impurity states .
the most important additional defects are interstitial mn ions and as atoms on cation sites ( antisite defects ) . both act as donors and can have a severe impact on the electric and magnetic properties of the dms epilayers .
more unintended defects form at higher mn doping because of the tendency of the material , even under non - equilibrium growth conditions , toward self - compensation .
direct experimental evidence for mn impurities occupying interstitial ( mn@xmath19 ) rather than substitutional positions was uncovered by combined channeling rutherford backscattering and particle induced x - ray emission measurements @xcite .
this technique can distinguish between mn@xmath126 and mn@xmath18 by counting the relative number of exposed mn atoms and the ones shadowed by lattice site host atoms at different channeling angles . in
highly doped as - grown samples , the experiment identified nearly 20% of mn as residing on interstitial positions .
the metastable nature of these impurities is manifested by the substantial decrease in their density upon post - growth annealing at temperatures very close to the growth temperatures @xcite .
detailed resistance - monitored annealing studies combined with auger surface analysis established the out - diffusion of mn@xmath19 impurities towards the free dms epilayer surface during annealing @xcite .
the characteristic energy barrier of this diffusion process is estimated to be 1.4 ev .
( note that a factor of 2 was omitted in the original estimate of this energy in @xcite . )
isolated mn@xmath19 spectroscopy data are not available in ( ga , mn)as , underlying the importance of theoretical work on the electric and magnetic nature of this impurity .
density functional calculations @xcite suggest , _
e.g. _ , that minority - spin mn@xmath19 @xmath4-states form a weakly dispersive band at @xmath127 ev below fermi energy , a feature which is absent in the theoretical mn@xmath18 spectra .
_ ab initio _ total energy calculations @xcite showed that mn can occupy two metastable interstitial positions , both with a comparable energy , one surrounded by four ga atoms ( see fig . [ mn_ga ] ) and the other surrounded by four as atoms .
the two mn@xmath19 states have similar local magnetic moments and electro - negativity .
the calculations have confirmed that mn@xmath19 acts as a double - donor , as expected for a divalent metal atom occupying an interstitial position .
each interstitial mn therefore compensates two substitutional mn acceptors .
it seems likely that because of the strong coulombic attraction between positively charged mn@xmath19 and negatively charged mn@xmath18 defects , the mobile interstitials pair - up with substitutional mn during the growth , as illustrated in fig .
[ mn_ga ] @xcite .
the total spin of a mn@xmath18-mn@xmath19 pair , inferred from _ ab initio _ calculations is much smaller than the local spin @xmath22 of the isolated mn@xmath18 acceptor .
this property , interpreted as a consequence of short - range antiferromagnetic interactions between two local moment defects that have comparable local moments , has been confirmed experimentally @xcite . according to theory
, the strength of this magnetic coupling contribution to the mn@xmath18-mn@xmath19 binding energy is 26 mev @xcite .
_ ab - initio _ calculations of the spin - splitting of the valence band indicate that the @xmath75 coupling constants of interstitial and substitutional mn are comparable @xcite .
this would suggest a negligible net @xmath7 coupling between the antiferromagnetically coupled mn@xmath18-mn@xmath19 pair and the valence band holes .
we do note , however , that these lsda calculations give @xmath128 mev nm@xmath67 which is more than twice as large as the experimental value , so this conclusion must be regarded somewhat cautiously .
it is generally accepted that this discrepancy reflects the general tendency of the dft - lsda and similar theories to systematically underestimate the splitting between the occupied mn @xmath4 states in the valence band continuum and the empty mn @xmath4-states .
( [ anderson ] ) illustrates how this deficiency of sdf band theories translates into an overestimated strength of the @xmath7 exchange interaction .
apart from this quantitative inaccuracy , the relative strength of the @xmath7 interaction of mn@xmath19 compared to mn@xmath18 is still a somewhat controversial issue in the theoretical literature @xcite , which has not yet been settled experimentally .
this property , and many others related to magnetism are sensitive to details of the sdf implementation , to the way in which disorder is accounted for , and even to technical details associated with the way in which the bloch - schrdinger equation is solved numerically .
for example , the positive charge of the mn@xmath18-mn@xmath19 ( single - acceptor double donor ) pair , which will tend to reduce its exchange coupling with valence band holes due to coulomb repulsion , is not included in all approaches .
in addition to the direct hole and local moment compensation effects of mn@xmath19 defects on ferromagnetism in ( ga , mn)as , the structural changes they induce in the crystal are indirectly related to important magnetic properties , particularly to the various magnetic and transport anisotropies . _
ab initio _ theory predicts that the separation of the four nearest as neighbors surrounding the mn@xmath19 in a relaxed lattice is increased by 1.5% compared to the clean gaas lattice @xcite . because of the coulomb repulsion between ga cations and mn@xmath19 defects , an even larger lattice expansion ( @xmath129 ) is found for the mn@xmath19 in the four ga tetrahedral position .
when grown on a gaas substrate , this effect of interstitial mn leads to a lattice - matching compressive strain in the ( ga , mn)as thin layers that induces a large uniaxial magnetic anisotropy , as we discuss in detail in section [ bulk - mag - micromag ] .
low - temperature growth of gaas is known to lead to the incorporation of high levels of as antisite defects .
this property is a combined consequence of the non - equilibrium growth conditions and the as overpressure often used in the mbe process to assure the 2d growth mode .
these double - donor defects are likely to be also present in the ( ga , mn)as epilayers and may contribute to hole compensation . unlike mn@xmath19 impurities , as antisites are stable up to @xmath11450 @xmath130c @xcite .
this is well above the temperature at which mn precipitation starts to dominate the properties of ( ga , mn)as and , therefore , the as antisites can not be removed from the epilayer by a post - growth annealing treatment .
experimental studies suggest that the degradation of ( ga , mn)as magnetic properties due to hole compensation by as antisites can be reduced by using as@xmath131 dimers instead of as@xmath123 tetramers and by maintainting a strictly stoichiometric growth mode @xcite . the following elements of the qualitative picture of ferromagnetism in ( ga , mn)as emerge from the experimental data and theoretical interpretations discussed in this section ( see also @xcite ) .
the low - energy degrees of freedom in ( ga , mn)as materials are the orientations of mn local moments and the occupation numbers of acceptor levels near the top of the valence band .
the number of local moments participating in the ordered state and the number of holes may differ from the number of mn@xmath18 impurities in the iii - v host due to the presence of charge and moment compensating defects .
hybridization between mn @xmath4-orbitals and valence band orbitals , mainly on neighboring as sites , leads to an antiferromagnetic interaction between the spins that they carry . at low concentrations of substitutional mn ,
the average distance between mn impurities ( or between holes bound to mn ions ) , @xmath132 , is much larger than the size of the bound hole characterized approximately by the impurity effective bohr radius , @xmath133 . here
@xmath134 is the number of mn impurities per unit volume , @xmath135 and @xmath136 are the semiconductor dielectric function and lattice constant , respectively , and @xmath137 is the effective mass near the top of the valence band . for this very dilute insulating limit , a theoretical concept was introduced in the late 1970 s in which a ferromagnetic exchange interaction between mn local moments is mediated by thermally activated band carriers @xcite .
experimentally , ferromagnetism in ( ga , mn)as is observed when mn doping reaches approximately 1% @xcite and the system is near the mott insulator - to - metal transition , _
i.e. _ , @xmath138 @xcite . at these larger mn concentrations , the localization length of the impurity band states
is extended to a degree that allows them to mediate ferromagnetic exchange interaction between mn moments , even though the moments are dilute .
several approaches have been used to address ferromagnetism in dmss near the metal - insulator transition ( for a review see @xcite , including finite - size exact diagonalization studies of hole - hole and hole - impurity coulomb interaction effects @xcite , and the picture of interacting bound magnetic polarons or holes hopping within the impurity band @xcite . because of the hopping nature of conduction and the mixed @xmath12 character of impurity band states , the regime is sometimes regarded as an example of double - exchange ferromagnetism . at even higher mn concentrations ,
the impurity band gradually merges with the valence band @xcite and the impurity states delocalize . in these metallic ( ga , mn)as ferromagnets , which we focus on in the following sections ,
the coupling between mn local moments is mediated by the @xmath7 kinetic - exchange mechanism @xcite .
a qualitatively similar picture applies for ( in , mn)as and mn - doped antimonides . in the metallic limit
the influence of coulomb and exchange disorder on perfect crystal valence - band states can be treated as perturbatively .
the crossover from impurity - band mediated to bloch valence - band mediated interactions between mn moments is a gradual one . in the middle of the crossover regime , it is not obvious which picture to use for a qualitative analysis and quantitative calculations are not possible within either picture .
strongly localized impurity - band states away from fermi energy may play a role in spectroscopic properties @xcite , even when they play a weaker role in magnetic and transport properties .
the crossover is controlled not only by the mn density but ( because of the importance of coulomb interaction screening ) also by the carrier density .
there is a stark distinction between the compensation dependence predicted by impurity - band and bloch valence - band pictures .
when the impurity - band picture applies , ferromagnetism does not occur in the absence of compensation@xcite , because the impurity band is filled .
given this , we can conclude from experiment that the impurity band picture does not apply to optimally annealed ( weakly - compensated ) samples which exhibit robust ferromagnetism .
the phenomenology of mn - doped phosphide or nitride dmss is more complex , with many aspects that probably can not be captured by the zener kinetic - exchange model @xcite .
experimentally , the nature of the mn impurity is very sensitive to the presence of other impurities or defects in the lattice @xcite .
this substantially complicates the development of a consistent picture of ferromagnetism in these materials .
larger @xmath7 coupling in the wide gap dmss ( reported _ e.g. _ in the photoemission experiment @xcite ) , and stronger bonding of the hole to the mn ion , might shift the metal - insulator transition to higher mn densities . at typical dopings of several per cent of mn
, the impurity band is still detached from the valence band @xcite and ferromagnetic mn - mn coupling is mediated by holes hopping within the impurity band .
recent experiments indicate that this scenario may apply to ( ga , mn)p with 6% mn doping @xcite .
lsda calculations suggest @xcite that the @xmath7 hybridization can be so strong that the admixture of mn @xmath139 spectral weight at the fermi energy reaches a level at which the system effectively turns into a @xmath4-band metal . to illustrate this trend we show in fig .
[ lsda_total_dos ] the lsda and lda+u calculations of the spin - split total density of states ( dos ) , and in fig . [ lsda_mn_dos ] the results for the mn @xmath4-states projected dos in ( ga , mn)as and ( ga , mn)n @xcite . indeed , in the wider gap ( ga , mn)n the spectral weight of mn @xmath4-orbitals at the fermi energy is large and is not significantly suppressed even if strong on - site correlations are accounted for by introducing the phenomenological hubbard parameter in the lda+u method ( see next section for details on combined hubbard model and sdf techniques ) . but for partial mn 3@xmath4 dos . in lda+u the @xmath7 hybridization in ( ga , mn)as is weak for both majority band and minority band states.,width=192 ] another possible scenario for these more ionic iii - p and iii - n semiconductors , supported by epr and optical absorption measurements and _ ab initio _
calculations @xcite , involves a transition of the substitutional mn from a divalent ( @xmath5 ) impurity to a trivalent ( @xmath8 ) impurity .
this strongly correlated @xmath8 center , with four occupied @xmath4-orbitals and a non - degenerate empty @xmath4-level shifted deep into the host band gap , may form as a result of a spontaneous ( jahn - teller ) lowering of the cubic symmetry near the mn site . if the energy difference between divalent and trivalent mn impurity states is small , the dms will have a mixed mn - valence which evokes the conventional double - exchange mechanism .
systems with dominant @xmath8 character of mn impurities , reminiscent of a charge transfer insulator , will inevitably require additional charge co - doping to provide for ferromagnetic coupling between dilute mn moments @xcite .
our focus is on the theory of ( iii , mn)v ferromagnets and we therefore present in this section an overview of the different approaches that can be used to interpret the existing experimental literature and to predict the properties of materials that might be realized in the future .
since the electronic and magnetic properties of ferromagnetic semiconductors are extremely sensitive to defects that are difficult to control in real materials and may not be completely characterized , the ability to make reasonably reliable theoretical predictions that are informed by as many relevant considerations as possible can be extremely valuable to the effort to discover useful new materials .
for example we would like to be able to make confident predictions of the ferromagnetic transition temperature of a ( iii , mn)v material as a function of the density of substitutional mn , interstitial mn , co - dopants , antisite defects , and any other defects whose importance might be appreciated in the future .
this ability is developing , we believe , although there is no simple _ silver bullet _ that solves all difficulties for all host materials , and there may still be some considerations that are important for less studied host materials and are not yet part of the discussion . in this section
we address in a general way the strengths and weaknesses of some of the different theoretical approaches .
the following sections of the review compare predictions made with different types of theoretical approaches with experimental data on a variety of important electronic and magnetic properties . in sdf theory
@xcite all many - body effects are buried in a complex exchange - correlation energy functional . once an approximation is made for this functional
, predictions for electronic and magnetic properties depend only on the particular arrangement of atomic nuclei under consideration . in principle nuclear positions
can be relaxed to make sure that the spatial distribution of nuclei is metastable and therefore realizable .
the exchange - correlation energy functional leads to a self - consistently determined spin - dependent exchange - correlation potential that appears in an effective independent - particle hamiltonian .
the main technical challenge in dft theory applications is the development of numerically efficient methods that provide accurate solutions of single - body schrdinger equations ( see review articles @xcite ) .
dft theory is established as a flexible and valuable tool for studying the microscopic origins of magnetism and for predicting electronic , magnetic , and ground - state structural properties in a wide - variety of materials @xcite .
it has the advantage that it is a first principles approach without any phenomenological parameters .
dft theory falls short of being a complete and general solution to the many - electron problem only because the exact form of the exchange - correlation energy functional is unknown .
a simple and widely successful approximation is the lsda @xcite .
the problem of solving lsda equations with adequate accuracy remains a challenge even in perfectly ordered crystals . in dmss
the degrees of freedom that are important for ferromagnetism , the orientations of the mn local moments , typically reside on approximately 1/40 of the atomic sites which further complicates numerical implementation of the lsda technique .
other length scales that are characteristic for the physics of interest in these materials , like the fermi wavelengths of the valence band carriers , are also longer than the atomic length scale on which dft theory interrogates matter .
this property limits the number of independent magnetic degrees of freedom that can be included in a dft simulation of dms materials .
the problem is exacerbated by the alloy disorder in ( ga , mn)as . even if all mn atoms substitute for randomly chosen lattice sites , it is necessary to find some way to average over microrealizations of the alloy .
disorder - averaging coherent - potential approximation ( cpa ) @xcite and supercell approaches , have been used successfully in combination with dft calculations to address those physical parameters of ( iii , mn)v dmss that are derived from total - energy calculations , such as the lattice constants , formation and binding energies of various defects , and the type of the magnetic order ( see _ e.g. _ @xcite ) .
supercell calculations have usually studied interactions between mn moment orientations by comparing the energies of parallel spin and opposite spin orientation states in supercells that contain two mn atoms .
an effective spin hamiltonian can be extracted from this approach if it is assumed that the interactions are pairwise and of heisenberg form . even when these assumptions are valid , the interaction extracted from these calculations is the sum of the interactions at a set of separations connected by the supercell lattice vector .
if the mn - mn spin interaction has a range larger than a couple of lattice constants , this poses a problem for the supercell approach .
longer range interactions can , however , be estimated using a spin wave approach which allows spin - orientation variations that are incommensurate with the supercell .
effective interactions extracted in this way lead to a classical heisenberg model from which the critical temperature can usually be calculated without substantial further approximation @xcite .
the cpa approach can estimate the energy cost of flipping a single spin in the ferromagnetic ground state , which is proportional to the mean - field - approximation for the critical temperature of the effective heisenberg model @xcite , and in this sense is limited in its predictive powers when mean - field theory is not reliable .
alternatively , a more detailed picture of magnetic interactions is obtained by direct mapping of the cpa total energy to the heisenberg hamiltonian @xcite .
lsda predictions for spectral properties , like the local dos , are less reliable than predictions for total energy related properties .
this is especially true for states above the fermi energy , and is manifested by a notorious inaccuracy in predicted semiconductor band gaps . from a dft theory point of view , this inconsistency arises from attempting to address the physics of quasiparticle excitations using ground - state dft . in mn - doped dmss
, the lsda also fails to account for strong correlations that suppress fluctuations in the number of electrons in the @xmath4-shell .
one generally accepted consequence is that the energy splitting between the occupied and empty @xmath4-states is underestimated in sdf theory , leading to an unrealistically large @xmath4-state local dos near the top of the valence band and to an overestimate of the strength of the @xmath7 exchange .
lda+u @xcite and self - interaction corrected ( sic ) lsda @xcite schemes have been used to obtain more realistic energy spectra and help to establish theoretically the microscopic origins of ferromagnetism in ( iii , mn)v semiconductor alloys @xcite .
lda+u schemes used in studies of ( iii , mn)v dmss combine sdf theory with the hubbard description of strongly correlated localized orbitals .
additional parameters from the hubbard model are added to the energy functional which are obtained by fitting to experiment or , in principle , can be calculated self - consistently @xcite .
the sic - lsda method is based on realizing that spurious self - interactions present in the sdf lead to unphysically large energy penalties for occupying localized states .
subtracting these interactions of a particle with itself from the density functional suppresses the tendency of the lsda to delocalize strongly correlated atomic orbitals .
a practical approach that circumvents some of the complexities of this strongly - correlated many - body problem is based on the anderson many - body hamiltonian theory @xcite and a tight - binding - approximation ( tba ) band - structure theory @xcite .
the tba hamiltonian includes the 8@xmath1408 @xmath141 term with second - neighbor - interaction integrals describing the host semiconductor @xcite and terms describing hybridization with non - magnetic impurities and mn .
effective single - particle tba theory is obtained from the anderson hamiltonian by replacing the density operators in the hubbard term in eq .
( [ anderson ] ) with their mean values @xcite . in the tba model ,
local changes of the crystal potential at mn and other impurity sites are represented by shifted atomic levels .
the parameters chosen for the atomic level shifts and the hopping amplitudes between atoms can be inferred from experiment in a manner which corrects for some of the limitations of lsda theory .
the parameterization , summarized in @xcite , provides the correct band gap for the host crystal and the appropriate exchange splitting of the mn @xmath4-states . in the calculations ,
the hole density can be varied independently of mn doping by adding non - magnetic donors ( _ e.g. _ si or se in gaas ) or acceptors ( _ e.g. _ c or be in gaas ) . although the tba model is a semi - phenomenological theory , it shares with first principles theories the advantage of treating disorder microscopically .
a disadvantage of the tight - binding model approach , which is often combined with the cpa , is that it normally neglects coulomb interaction effects which influence the charge and spin densities over several lattice constants surrounding the mn ion positions .
curie temperatures , magnetizations , the lifetimes of bloch quasiparticle states , the effects of doping and disorder on the strength of the @xmath7 exchange coupling , and the effective mn - mn magnetic interaction are among the problems that have been analyzed using this tool @xcite . the highest critical temperatures in ( ga
, mn)as dmss are achieved in optimally annealed samples and at mn doping levels above 1.5% for which the band holes are itinerant , as evidenced by the metallic conductivities @xcite .
in this regime , semi - phenomenological models that are built on crystal bloch states rather than localized basis states for the band quasiparticles might be expected to provide more useful insights into magnetic and magneto - transport properties .
a practical approach to this type of modeling starts from recognizing that the length scales associated with holes in the dms compounds are still long enough that a @xmath69 , envelope function description of the semiconductor valence bands is appropriate . since for many properties it is necessary to incorporate spin - orbit coupling in a realistic way , six- or eight - band kohn - luttinger ( kl ) @xmath69 hamiltonians that include the spin - orbit split - off band are desirable @xcite .
the kinetic - exchange effective hamiltonian approach @xcite asserts the localized character of the five mn@xmath142 @xmath4-orbitals forming a moment @xmath22 and describes hole states in the valence band using the kl hamiltonian and assuming the @xmath7 exchange interaction between mn@xmath142 and hole spins .
as discussed in section [ micro - submn_pd ] , the exchange interaction follows from hybridization between mn @xmath4-orbitals and valence band @xmath0-orbitals .
the approach implicitly assumes that a canonical transformation has been performed which eliminated the hybridization @xcite .
the @xmath69 approximation applies when all relevant wavevectors are near the brillouin - zone center and the model also assumes from the outset that the states near the fermi energy mainly have the character of the host semiconductor valence band , even in the neighborhood of a substitutional mn . when these assumptions are valid it follows from symmetry considerations that the spin - dependent part of the effective coupling between mn and band spins is an isotropic heisenberg interaction characterized by a single parameter .
if the kl hamiltonian parameters are taken from the known values for the host iii - v compound @xcite , the strength of this exchange interaction @xmath75 can be extracted from one set of data , for example from spectroscopic studies of isolated mn acceptors as explained in section [ micro - submn_pd ] , and used to predict all other properties @xcite .
since the value of @xmath75 can be obtained from experiments in a paramagnetic state the approach uses no free parameters to model ferromagnetism in these systems . in the absence of an external magnetic field
the kl kinetic - exchange hamiltonian has the general form : @xmath143 where @xmath144 includes the @xmath69 kl hamiltonian and the interactions of holes with the random disorder potential and with other holes . the second term in eq . ( [ coupling ] ) represents the @xmath7 exchange interaction between local mn spins @xmath145 and hole spins @xmath146 .
the @xmath69 approach has the advantage that it focuses strongly on the magnetic degrees of freedom introduced by the dilute moments , which can simplify analysis of the model s properties .
disorder can be treated in the model by introducing born approximation lifetimes for the bloch states or by more sophisticated , exact - diagonalization or monte - carlo methods @xcite .
this approach makes it possible to use standard electron - gas theory tools to account for hole - hole coulomb interactions @xcite .
the envelope function approximation is simply extended to model magnetic semiconductor heterostructures , like superlattices or quantum wells @xcite .
this strategy will fail , however , if the @xmath7 exchange is too strong and the mn acceptor level is correspondingly too spatially localized or too deep in the gap . for example , mn - doped gap and gan compounds are likely less favorable for this approach than ( iii , mn)as and ( iii , mn)sb compounds . generally speaking the advantages of a fully microscopic approach have increasing importance for more localized acceptors , and hence shorter range mn - mn interactions , while the advantages of the @xmath69 approach are more clear when the acceptors are more shallow and the mn - mn interactions have longer range .
there has also been theoretical work on ( iii , mn)v dms materials based on still simpler models in which holes are assumed to hop between mn acceptor sites , where they interact with the mn moments via phenomenological exchange interactions @xcite .
hamiltonians used in these studies have a form of ( or similar to ) , @xmath147 where @xmath148 creates a hole at site @xmath71 with spin @xmath149 , the hole spin operator @xmath150 , and @xmath151 are the pauli matrices .
the models apply at least qualitatively in the low mn density limit and are able to directly attack the complex and intriguing physics of these unusual insulating ferromagnets .
insulating ferromagnetism persists in this limit even when the carrier density is not strongly compensated @xcite .
the dependence of ferromagnetism in this regime on the degree of compensation has not yet been systematically studied experimentally and seems certain to pose challenging theoretical questions .
the models that have been used to study ferromagnetism in this regime can easily be adapted to include holes that are localized on ionized defects that may occur in addition to mn acceptors .
other related models assume that the mn acceptors are strongly compensated so that the density of localized holes is much smaller than the density of mn ions , leading to a polaronic picture in which a single hole polarizes a cloud of mn spins @xcite .
the free - parameter nature of these phenomenological models means that they have only qualitative predictive power .
they are not appropriate for the high @xmath3 ( ga , mn)as materials which are heavily doped by weakly compensated mn acceptors and are metallic . on the other hand , the impurity band models may represent a useful approach to address experimental magnetic and transport properties of ferromagnetic ( ga , mn)p where holes are more strongly localized @xcite .
experimental efforts to increase mn doping in ( ga , mn)as dmss beyond the solubility limit of 0.1% have been assisted by modern _ ab initio _ theoretical studies of impurity formation energies and effects related to the growth kinetics @xcite .
substitution of ga by mn is expected , based on these studies , to be enhanced when the ga chemical potential is kept low relative to the mn chemical potential , _
i.e. _ , under ga - poor , mn - rich growth conditions @xcite .
calculations also suggest that one of the major drawbacks of bulk growth techniques is that they allow phase separated precipitates , such as mnas , to attain their most stable free - standing lattice geometry , leading to relatively low formation energies for these unwanted phases @xcite . in thin film epitaxy ,
competing phases are forced to adopt the crystal structure of the substrate , which can significantly increase their formation energies @xcite .
the non - equilibrium lt - mbe has been a particularly successful growth technique which allows a synthesis of single - phase ( ga , mn)as dmss with mn concentrations up to @xmath152 . if the growth temperature is precisely controlled , a 2d growth mode of uniform dmss can be maintained and , at the same time , a large number of mn atoms is incorporated in mn@xmath18 positions @xcite . in ( ga , mn)as dmss a significant fraction of the mn atoms is also incorporated in interstitial positions @xcite .
adsorption pathways that can funnel mn to interstitial sites have been identified theoretically ( see fig .
[ erwin ] ) using _ ab initio _ calculations of the potential energy surface of mn adsorbed on gaas(001 ) @xcite .
first principles calculations have also confirmed that interstitial mn@xmath19 impurities are metastable in gaas , showing that the three distinct positions they can occupy are two tetrahedral t(as@xmath153 or ga@xmath153 ) positions surrounded by four near - neighbor as or ga atoms , and one hexagonal position with three ga and three as nearest neighbors . among the three interstitial sites the hexagonal position is clearly less favorable , especially so in an overall p - type ( ga , mn)as material @xcite .
the typical energy barrier for mn diffusion between interstitial sites is approximately 1 ev @xcite . on the other hand , diffusion of mn between ga substitutional positions
involves a kick - out mechanism of mn@xmath19 + ga@xmath18 @xmath154 mn@xmath18 + ga@xmath19 for which the typical barrier is about 3 ev @xcite .
interstitial mn@xmath19 is therefore much more mobile than substitutional mn@xmath18 . ; the corresponding surface geometry is shown ( light gray for as , dark gray for ga , yellow for mn ) .
typical adsorption pathways funnel mn adatoms to this interstitial site ( heavy curves ) or to a cave site , @xmath155 ( light curves ) .
inset : binding energy of a mn adatom centered on the as - dimer ; for comparison , results are also shown for a ga adatom .
when additional as is deposited , the metastable mn site , @xmath156 , becomes more favorable and leads to partial incorporation of substitutional mn . from @xcite .
, width=326 ] as mentioned in section [ micro - other_imp - int_mn ] , mn@xmath19 donors are likely to form near - neighbor pairs with mn@xmath18 acceptors in as - grown materials due to the strong coulomb attraction .
the net magnetic moment of such a pair is close to zero @xcite .
although mn@xmath19 can be removed by low - temperature annealing , the number of substitutional mn@xmath18 impurities will remain smaller than the total nominal mn doping .
the mn@xmath18 doping efficiency is , therefore , one of the key parameters that may limit the maximum @xmath3 that can be achieved in ( ga , mn)as epilayers , as we discuss in detail in section [ bulk - mag - tc ] . _ ab initio _ calculations of the formation energies can be used to estimate the dependence of mn@xmath18 and mn@xmath19 partial concentrations on total mn doping in as - grown materials @xcite .
similarly , correlated doping effects can be studied for other defects that occur frequently in lt - mbe ( ga , mn)as materials , such as the arsenic antisites as@xmath18 @xcite .
correlations between acceptors ( mn@xmath18 ) and donors ( mn@xmath19 or as@xmath18 ) in iii - v semiconductors like gaas are strong due to the nearly covalent nature of bonding in these crystals .
the cohesion energy of the covalent networks has a maximum if the fermi energy @xmath157 lies within a band gap .
whenever @xmath157 is shifted to the valence band or the conduction band the strength of the bonds is reduced because of the occurrence of unfilled bonding states or occupied antibonding states , respectively . in case of weak doping , small changes in the impurity concentration can easily move @xmath158 across the band gap with a negligible influence on the energy spectrum .
the dependence of the formation energy , _
i.e. _ the energy cost for incorporating a particular impurity in a crystal , on the number of electrically active impurities can then be represented by the corresponding change in @xmath158 multiplied by the charge state of the impurity @xcite . in case of
strongly doped and mixed crystals , the redistribution of electronic states in the valence band due to the impurities may play a more important role and should therefore be included in the microscopic calculations . in general , the formation energy as a function of impurity concentrations can be obtained from the composition - dependent cohesion energy @xmath159 of the crystal . assuming a sample consisting of @xmath160 unit cells of the impure ( mixed ) crystal , the formation energy @xmath161 of an acceptor a replacing atom x is defined as the reaction energy of the substitution process , sample + a @xmath162 sample with one extra acceptor + x. the corresponding reaction energy is @xmath163 where @xmath164 is the total energy of the doped crystal normalized to a unit cell , @xmath165 and @xmath166 are the acceptor and donor concentrations , and the last two terms represent total energies of free - standing atoms x and a. with increasing size of the sample , @xmath167 , the first term in eq .
( [ reaction_en_a ] ) approaches the derivative of @xmath164 with respect to @xmath165 and the atomic energies can be absorbed by using the relation between the total and cohesion energy . as a result , the cohesion energy represents a generating functional for the formation energies , _
i.e. _ , @xmath168 the formation energy of a donor d substituting for an atom y has the same form with @xmath169 and @xmath170 .
having defined @xmath161 and @xmath171 , we note that in the low concentration regime where they depend linearly on @xmath172 and @xmath173 , @xmath174 _ i.e. _ , that the mutual influence of the two kinds of impurities is symmetric .
@xmath175 in eq .
( [ corr_energy ] ) plays the role of a correlation energy characterizing the co - doping process . for positive @xmath175 ,
the formation energy of one impurity increases in the presence of the other . in this case
the material tends to be either n - type or p - type rather than a compensated semiconductor . on the other hand
, negative correlation energy indicates that the presence of impurities of one kind makes the incorporation of the other dopants easier . in this case compensation
is favored .
doping correlations over a wide and continuous range of impurity concentrations have been studied using the cpa , combined with either the parameterized tba model or the _ ab initio _ linearized - muffin - tin - orbital ( lmto ) dft method @xcite .
results of these calculations are summarized in figs .
[ formen1 ] and [ formen2 ] .
the zero of energy is set to correspond to the formation energy of a reference ga@xmath176mn@xmath177as system with all mn atoms occupying substitutional ga positions .
four representative examples are considered here .
se@xmath178 and si@xmath110 are typical single donors in gaas which occupy the anion and cation sublattice , respectively . the as antisite defects as@xmath110 and the mn@xmath19 interstitials , are the most important native defects in ( ga , mn)as , both acting as double donors . fig . [ formen1 ] shows that the formation energy of mn@xmath18 decreases with an increasing number of donors .
the curves are grouped into pairs according to the charge state of the donors , with only a minor influence of the particular chemical origin of the defect .
the dependence is almost linear for low concentrations and the slope of the function is roughly proportional to the charge state of the donor .
all this indicates that the variation of the formation energy of mn@xmath18 is mostly determined by the above mentioned fermi - level effect , and that the redistribution of the density of states induced by donor defects plays a minor role .
formation energies of the interstitial mn@xmath19 in the tetrahedral t(as@xmath153 ) position are shown in fig .
[ formen2 ] . for the donor mn@xmath19 impurity , the formation energy increases with the density of other donors .
this means that the creation of mn@xmath19 is efficiently inhibited in the presence of as@xmath18 .
analogous results are obtained for as@xmath18 antisite defects . of substitutional mn@xmath18 in ga@xmath176mn@xmath177as as a function of the concentration of various donors .
from @xcite .
, width=204,height=328 ] of the interstitial mn impurity in ga@xmath176mn@xmath177as a function of the concentration of various donors . from @xcite .
, width=204,height=328 ] in fig .
[ formen4 ] we show the change of formation energies of as@xmath18 and mn@xmath19 as a function of the number of mn@xmath18 . in both cases , the formation energy is a decreasing function of the density of mn@xmath18 .
this self - compensation tendency is an important mechanism controlling the properties of as - grown ( ga , mn)as mixed crystals .
it explains the observed charge compensation in as - grown materials and is responsible to a large degree for the lattice expansion of highly mn - doped ( ga , mn)as dmss , as we discuss in section [ bulk - struct - lc ] . of the formation energies of the interstitial mn impurity and as@xmath18 antisite defect in ga@xmath83mn@xmath179as a function of the concentration @xmath180 of the mn atoms substituted in the ga sublattice .
the variations are measured from the values for the reference material with 4 % of mn@xmath18 from @xcite .
, width=204,height=328 ] the formation energies can be used to theoretically estimate partial concentrations of substitutional mn@xmath18 , @xmath181 , and interstitial mn@xmath19 , @xmath182 , in as - grown ( ga , mn)as materials .
an assumption is made in these calculations , whose validity is tested by a comparison with experimental data , that the probabilities of mn atoms to occupy substitutional or interstitial positions are determined by the respective formation energies , @xmath183 and @xmath184 , even in the non - equilibrium lt - mbe grown materials .
the balanced distribution of mn@xmath18 and mn@xmath19 is reached when @xcite , @xmath185 as also expected from the growth point of view .
partial concentrations @xmath186 of mn are obtained by solving eq .
( [ cohesion_cond ] ) together with the condition @xmath187 , @xmath188 . in fig .
[ partial ] we show results of tba / cpa calculations @xcite ; for @xmath189 @xmath190 and @xmath191 .
lmto / cpa theory calculations give very similar predictions @xcite .
the linear relations between @xmath192 , @xmath193 , and @xmath180 reflect the fact that the difference of the formation energies of mn@xmath18 and mn@xmath19 impurities ( see inset of fig . [ partial ] ) can be , up to @xmath194 , approximated by a linear function of @xmath192 and @xmath193 , @xmath195 this relation indicates that for @xmath196 , mn@xmath142 has a lower formation energy than mn@xmath19 and mn atoms tend to occupy substitutional positions . at and above @xmath197 , @xmath198 approaches zero and both mn@xmath110 and mn@xmath19 are formed .
the theoretical results are in a very good agreement with experimental data , as shown in fig .
[ partial ] .
the balance considerations , confirmed experimentally in samples with mn@xmath18 concentrations up to 6.8% , suggest that there is no fundamental physics barrier to increasing mn@xmath18 concentration up to 10% and beyond .
very precise control over the growth temperature and stoichiometry is , however , required for maintaining the 2d growth mode of the uniform ( ga , mn)as materials at these high doping levels .
finally we note that during growth the formation energies control incorporation of mn atoms , assuming that the total amount of mn in the material is related to a sufficiently high chemical potential in the mn source .
the annealing processes , on the other hand , do not depend on formation energies but rather on energy barriers surrounding individual metastable positions of mn in the lattice .
the barriers are larger for mn@xmath18 @xcite so that post - growth low temperature annealing can be used to remove mn@xmath19 without changing the number of mn@xmath18 significantly .
( red line ) and interstitial mn@xmath19 ( blue line ) impurities . triangles and circles represent experimental mn@xmath18 and mn@xmath19 partial concentrations , respectively .
inset : formation energies of mn@xmath18 and mn@xmath19 as a function of total mn concentration . from @xcite .
, width=364 ] changes in the lattice constant of ( ga , mn)as dmss , relative to the lattice constant of undoped gaas , are too small to significantly suppress or enhance @xmath7 kinetic exchange or other magnetic coupling mechanisms .
direct effects of doping - induced lattice distortion on the onset of ferromagnetism are therefore negligibly small .
nevertheless , variations in the lattice parameter provide a measure of impurity concentrations in the dms material .
the impurities do , of course , control ferromagnetism through their doping properties . also , because thin film ( ga , mn)as epilayers are not relaxed , lattice constant mismatch between the dms layer and the substrate induces strains that in many cases determine magnetocrystalline and magnetotransport anisotropies , as we discuss in detail in sections [ bulk - mag - micromag - aniso ] and [ bulk - magtransp - amr ] .
theoretical calculations @xcite of the dependence of the lattice constant on the density of the most common impurities in dmss represent , therefore , another piece to the mosaic of our understanding of ferromagnetism in these complex systems . considering the values of the atomic radii of mn ( @xmath199 = 1.17 ) and ga ( @xmath200 = 1.25 ) , the substitutional mn@xmath18 impurity may be expected to lead to only very small changes ( reductions ) in the lattice constant .
this expectation is consistent with the calculated @xcite lattice constant of a hypothetical zinc - blende mnas crystal whose value is comparable to that of gaas . on the other hand ,
as antisites produce an expansion of the gaas lattice @xcite and a similar trend can be expected for interstitial mn .
modern density - functional techniques allow one to move beyond intuitive theoretical considerations and discuss the dependence of the lattice constant on impurity concentrations on a more quantitative level @xcite .
the cpa is again a useful tool here for studying ( ga , mn)as properties over a wide range of impurity concentrations .
some quantitative inaccuracies in theoretical results due to the limitations of the lmto / cpa approach ( _ e.g. _ neglect of local lattice relaxations ) have been corrected by using the full - potential linearized- augmented - plane - wave supercell method @xcite . starting with an ideal ( ga , mn)as mixed crystal with all mn atoms occupying substitutional ga positions , these calculations give the following vegard s law expression for the doping dependence of the lattice constant : @xmath201 with the expansion coefficient @xmath202 ranging from -0.05 to 0.02 depending on the method used in the calculation @xcite . as expected
, @xmath203 changes only weakly with the mn@xmath18 density @xmath192 .
a similar linear dependence is obtained for hypothetical crystals where as@xmath18 ( or mn@xmath19 ) is the only impurity present in the material , as shown in fig .
[ alatt1 ] . according to the more reliable full - potential supercell calculations @xcite ,
the composition - dependent lattice constant is found to obey : @xmath204 where @xmath193 and @xmath205 are the densities of mn@xmath19 and as@xmath18 , respectively , and @xmath206 is the lattice constant of pure gaas .
( circles ) , mn@xmath19 ( full triangles ) , and as@xmath18 ( empty triangles ) . from @xcite.,width=253,height=325 ] recently , several experimental works @xcite have studied the dependence of lattice constants in ( ga , mn)as materials on disorder , based on the comparison between as - grown and annealed samples .
the measurements confirmed that both as@xmath18 and mn@xmath19 defects lead to a significant expansion of the lattice . in samples grown under as - rich condition ,
which is expected to inhibit formation of mn@xmath19 impurities , annealing has virtually no effect on the measured lattice constant .
this is consistent with the stability of as@xmath18 defects up to temperatures that are far above the annealing temperatures .
mn@xmath19 impurities , on the other hand , can be efficiently removed by low - temperature annealing .
consistently , annealing leads to a significant reduction of the lattice constant in materials which contain a large number of these defects in the as - grown form , as shown in fig .
[ lattice_const_nott ] . on a quantitative level ,
experimental data suggest a stronger lattice expansion due to mn@xmath18 and a weaker expansion due to as@xmath18 and mn@xmath19 , compared to the theoretical predictions of eq .
( [ a_all ] ) . the quantitative disagreement can be attributed , in part , to the simplified description of the system within the theoretical model .
also , the presence of other lattice imperfections or inaccuracies in the determination of experimental doping values may have partly obscured the direct quantitative comparison between experiment and theory .
curie temperatures in metallic ( ga , mn)as have been studied theoretically starting from the * k*@xmath207*p * kinetic - exchange effective hamiltonian @xcite and from microscopic tba or sdf band structure calculations @xcite . for a more detail description of these theoretical approaches see sections [ theory - first]-[theory - kp ] .
the advantage of the * k*@xmath207*p * kinetic - exchange model is that it uses the experimental value for the @xmath7 coupling constant @xmath75 , _
i.e. _ , it correctly captures the strength of the magnetic interaction that has been established to play the key role in ferromagnetism in ( ga , mn)as . the model also accounts for strong spin - orbit interaction present in the host valence band which splits the three @xmath0-bands into a heavy - hole , light - hole , and a split - off band with different dispersions .
the spin - orbit coupling was shown @xcite to play an important role in suppressing magnetization fluctuation effects and , therefore , stabilizing the ferromagnetic state up to high temperatures .
on the other hand , describing the potentially complex behavior of mn@xmath142 in gaas by a single parameter may oversimplify the problem .
the calculations omit , for example , the suppression of @xmath3 in low hole - density ( ga , mn)as materials due to the direct antiferromagnetic superexchange contribution to the coupling of near - neighbor mn pairs .
the whole model inevitably breaks down in dms systems with holes strongly bound to mn acceptors or with large charge fluctuations on mn@xmath142 @xmath4-shells .
the advantage of microscopic approaches to curie temperature calculations is that they make no assumption about the character of mn@xmath18 impurities in gaas and their magnetic coupling .
they are therefore useful for studying material trends in @xmath3 as a function of mn doping or the density of other intentional or unintentional impurities and defects present in real systems . because spin - orbit interactions add to the numerical complexity of calculations that are already challenging , they have normally been neglected in this approach .
another shortcoming , discussed already in section [ theory - first ] , of the lsda approaches is an overestimated strength of the @xmath7 exchange as compared to experiment . within the mean - field approximation , which considers thermodynamics of an isolated mn moment in an effective field and neglects correlated mn - mn fluctuations ,
microscopic calculations typically yield larger @xmath3 s than the effective hamiltonian model that uses the experimental value for @xmath75 .
stronger @xmath7 exchange and the omission of spin - orbit coupling effects in the dft theories , however , also leads to a larger suppression of the curie temperature due to fluctuation effects . ( a closer agreement on the character of the @xmath3 versus mn - doping curves , calculated within the two formalisms , is obtained when the deficiencies of lsda theories are partly eliminated by using , _
e.g. _ , the lda+u . ) despite the above weaknesses of semi - phenomenological and microscopic calculations , a qualitatively consistent picture is clearly emerging from these complementary theoretical approaches that , as we discuss below , provides a useful framework for analyzing measured @xmath3 s .
our review of theoretical @xmath3 trends in ( ga , mn)as starts with the results of the kl kinetic - exchange hamiltonian and mean - field approximation to set up a scale of expected curie temperatures .
these estimates which are not accurate in all regimes , are simplified by assuming a homogeneous distribution of mn@xmath18 ions and neglecting the role of other defects , apart from their potential contribution to hole or moment compensation @xcite . in the case of microscopic models
this assumption is equivalent to the virtual crystal approximation .
microscopic tba calculations have shown very little effect of positional disorder on the strength of magnetic couplings in ( ga , mn)as epilayers with metallic conductivities , partly justifying the virtual - crystal approach @xcite .
in addition , detailed theoretical studies confirm the absence of any significant magnetic frustration associated with the random positions of mn@xmath18 moments in the lattice in the more metallic ferromagnetic semiconductors @xcite .
in the very dilute limit , however , @xmath208 becomes sensitive to the distribution of the mn moments in the lattice @xcite . in the mean - field approximation @xcite , each local mn@xmath142 moment
is described by a hamiltonian @xmath209 where @xmath210 is the mn@xmath142 local spin operator , @xmath211 , and @xmath212 is the mean spin density of the valence band holes ( for the definition of the @xmath75 field see section [ micro - submn_pd ] ) .
@xmath213 is an effective field seen by the local moments due to spin - polarization of the band holes , analogous to the nuclear knight shift .
similarly @xmath214 is an effective magnetic field experienced by the valence band holes which is proportional to the density and the mean spin polarization of the mn@xmath142 local moments .
the dependence of @xmath215 on temperature and field @xmath213 is given @xcite by the brillouin function @xcite : @xmath216 the curie temperature is found by linearizing @xmath213 and @xmath217 around @xmath218 : @xmath219 here @xmath220 is the itinerant hole spin susceptibility given by @xmath221 and @xmath222 is the total energy per volume of the holes .
( [ bs ] ) and ( [ linear ] ) give @xmath223 the qualitative implications of this @xmath3-equation can be understood within a model itinerant hole system with a single spin - split band and an effective mass @xmath137 .
the kinetic energy contribution , @xmath224 , to the total energy of the band holes gives a susceptibility : @xmath225 where @xmath226 is the fermi wavevector . within this approximation
@xmath3 is proportional to the mn@xmath142 local moment density , to the hole fermi wavevector , _
i.e. _ to @xmath227 where @xmath0 is the hole density , and to the hole effective mass @xmath137
. a more quantitative prediction for the curie temperature is obtained by evaluating the itinerant hole susceptibility using a realistic band hamiltonian , @xmath228 where @xmath229 is the six - band kl hamiltonian of the gaas host band @xcite and @xmath230 is the hole spin operator @xcite .
the results , represented by the solid black line in fig
. [ tc_mf_sw ] , are consistent with the qualitative analysis based on the parabolic band model , _
i.e. _ , @xmath3 roughly follows the @xmath231 dependence .
based on these calculations , room temperature ferromagnetism in ( ga , mn)as is expected for 10% mn@xmath18 doping in weakly compensated samples @xcite .
hole - hole coulomb interaction effects can be included , in the lowest order of perturbation theory by adding the hole exchange contribution to the total energy @xcite .
the red line in fig .
[ tc_mf_sw ] shows this stoner @xmath3 enhancement calculated numerically using the kinetic - exchange effective model with the six - band kl hamiltonian .
@xmath3 stays roughly proportional to @xmath232 even if hole - hole exchange interactions are included , and the enhancement of @xmath3 due to interactions is of the order @xmath233% @xcite .
( thin red line ) , @xmath3 after correcting for correlated mn orientation fluctuations using a spin - wave approximation ( blue symbols ) . from @xcite . , width=316 ] the mean - field effective hamiltonian analysis above neglects discreteness in the random mn@xmath142 positions in the lattice and magnetic coupling mechanisms additional to the kinetic exchange contribution , particularly the near neighbor superexchange .
the former point can be expected to influence @xmath3 at large hole densities , _
i.e. _ , when the hole fermi wavelength approaches inter - atomic distances .
of course , the entire phenomenological scheme fails on many fronts when the fermi wavelength approaches atomic length scales since it is motivated by the assumption that all relevant length scales are long ; the * k*@xmath207*p * band structure , the use of the host material band parameters , and the neglect of momentum dependence in the @xmath75 parameter all become less reliable as the hole density increases to very large values .
the approximations are apparently not fatal , however , even for @xmath234% and any degree of compensation . in the opposite limit of strongly compensated systems , where the overall magnitude of the hole - mediated exchange is weaker ,
antiferromagnetic superexchange can dominate the near - neighbor mn@xmath142-mn@xmath142 coupling @xcite , leading to a reduced curie temperature @xcite .
we emphasize that the * k*@xmath207*p * kinetic - exchange model can not be applied consistently when nearest neighbor interactions dominate , since it implicitly assumes that all length scales are longer than a lattice constant .
we also note that net antiferromagnetic coupling of near - neighbor mn@xmath18-mn@xmath18 pairs is expected only in systems with large charge compensations . in weakly compensated (
ga , mn)as the ferromagnetic contribution takes over @xcite .
in addition to the above effects related to random mn distribution , mn positional disorder can directly modify @xmath7 interactions when the coherence of bloch states becomes significantly disturbed .
microscopic theories , such as the tba / cpa @xcite or _ ab initio _ approaches based on either cpa or supercell band structures @xcite , capture all these effects on an equal footing and can be used to estimate trends in mean - field @xmath3 beyond the virtual crystal approximation .
the mean - field cpa curie temperatures are obtained by evaluating the energy cost of flipping one mn@xmath18 moment with all other moments held fixed in the ferromagnetic ground state .
it can be evaluated for any given chemical composition @xcite and defines an effective exchange field @xmath235 acting on the local moment .
this energy change corresponds to @xmath213 in the kinetic - exchange model used in the previous section , _
i.e. _ , @xmath236 results based on microscopic tba band structure calculations are shown in fig .
[ tc_tba ] as a function of hole density for several mn@xmath18 local moment concentrations @xcite .
( the hole density is varied independently of mn@xmath18 doping in these calculations by adding non - magnetic donors or acceptors . )
comparison with fig .
[ tc_mf_sw ] identifies the main physical origins of the deviations from the @xmath237 trend .
black dots in the left panel of fig .
[ tc_tba ] correspond to a relatively low local mn@xmath18 moment concentration ( @xmath238 ) and hole densities ranging up to @xmath239 and show the expected suppression of @xmath3 at large @xmath0 .
the effect of superexchange in the opposite limit is clearly seen when inspecting , _
e.g. _ , the @xmath240 data for @xmath241 nm@xmath119 .
the mean - field tba / cpa curie temperature is largely suppressed here or even negative , meaning that the ferromagnetic state becomes unstable due to the short - range antiferromagnetic coupling .
note , that the neglect of coulomb interactions in these tba / cpa calculations likely leads to an overestimated strength of the antiferromagnetic superexchange .
the inhomogeneity of the carrier distribution in the disordered mixed crystal may also contribute to the steep decrease of @xmath3 with increasing compensation seen in fig .
[ tc_tba ] .
although the curie temperatures in the left panel of fig . [ tc_tba ] appear to depart strongly for the @xmath237 dependence , the linearity with @xmath180 is almost fully recovered when @xmath3 is plotted as a function of the number of holes per mn@xmath18 local moment , @xmath242 ( see right panel of fig .
[ tc_tba ] ) .
note that for compensations ( @xmath243 ) reaching 100% this property of the superexchange coupling is reminiscent of the behavior of ( ii , mn)vi diluted magnetic semiconductors @xcite in which mn acts as an isovalent magnetic impurity .
the dependence on @xmath0 in ( ga , mn)as is expected to become very weak , however , when approaching the uncompensated state .
similarly , the prospects for substantial increases in @xmath3 by non - magnetic acceptor co - doping of weakly compensated material appear to be quite limited . in the left panel of fig .
[ lda_tc ] we show mean - field cpa curie temperatures in uncompensated ( @xmath244 ) ( ga , mn)as dmss as a function of mn doping calculated using lda and lda+u _ ab initio _ methods .
the lda+u calculations , which give more realistic values of the @xmath7 exchange coupling , confirm the linear dependence of @xmath3 on @xmath180 , showing no signs of saturation even at the largest doping @xmath245% considered in these calculations . the potential influence of correlated mn - moment fluctuations ( corrections to mean field theory ) on ferromagnetic ordering in ( ga
, mn)as can be recognized by considering , within a simple parabolic band model , the rkky oscillations effect which occurs as a consequence of the @xmath246 anomaly in the wavevector dependent susceptibility of the hole system @xcite . in this theory , which treats the hole system perturbatively around the paramagnetic state ,
the sign of the hole - mediated mn@xmath142-mn@xmath142 coupling varies as @xmath247 , where @xmath4 is the distance between mn@xmath142 moments , and its amplitude decays as @xmath248 . estimating the average mn@xmath142-mn@xmath142 separation in a ( ga , mn)as random alloy as @xmath249 for uncompensated ( ga , mn)as systems and neglecting spin - orbit coupling and band warping , @xmath250 , which means that the role of the rkky oscillations can not be generally discarded . in realistic valence bands the oscillations are suppressed due to non - parabolic and anisotropic dispersions of the heavy- and light - hole bands and due to strong spin - orbit coupling @xcite .
more quantitatively , the range of reliability and corrections to the mean - field approximation in ( ga , mn)as can be estimated by accounting for the suppression of the curie temperature using the quantum theory of long - wavelength spin - waves or using monte carlo simulations which treat mn - moments as classical variables .
for weak @xmath7 exchange coupling , @xmath251 where @xmath157 is the hole fermi energy , the spin - polarization of the hole system is small and the rkky and spin - wave approximations treat collective mn - moment fluctuations on a similar level .
the advantage of the spin - wave theory is that it can be used to explore the robustness of ferromagnetic states over a wider range of @xmath7 couplings , including the more strongly exchange - coupled as - grown materials with large mn density and large hole compensation ( see section [ discussion - general ] for a general discussion of magnetic interactions in the two coupling - strength limits ) .
calculations in metallic systems have been performed starting from the * k*@xmath207*p * kinetic - exchange effective hamiltonian @xcite or from the sdf band structure calculations @xcite .
( monte carlo studies of @xmath3 in the regime near the metal - insulator are reported in @xcite . ) within a non - interacting spin - wave approximation , the magnetization vanishes at the temperature where the number of excited spin waves equals the total spin of the ground state @xcite : @xmath252 where @xmath253 is the debye cutoff and @xmath254 is proportional to the zero - temperature spin - wave stiffness parameter @xmath255 ( @xmath256 ) and the mean - field temperature - dependent average spin on mn , @xmath257 @xcite .
( for a detail discussion of the micromagnetic parameter @xmath255 see section [ bulk - mag - micromag ] ) . comparing the spin - stiffness results obtained using the * k*@xmath207*p * kinetic - exchange model with a simple parabolic band and with the more realistic , spin - orbit coupled kl hamiltonian , the spin stiffness is observed to always be much larger in the kl model @xcite . for ( ga , mn)as , the parabolic - band model underestimates @xmath258 by a factor of @xmath1110 - 30 for typical hole densities .
this larger spin stiffness in the spin - orbit coupled valence bands is due to the heavy - hole light - hole mixing .
crudely , the large mass heavy - hole band dominates the spin susceptibility and enables local ( mean - field ) magnetic order at high temperatures , while the dispersive light - hole band dominates the spin stiffness and enables long - range magnetic order .
the analysis highlights that the multi - band character of the semiconductor valence band plays an important role in the ferromagnetism of ( ga , mn)as . critical temperature estimates based on eq . ( [ collective_tc ] ) , the kl kinetic - exchange model , and including also the stoner enhancement of @xmath3 are summarized in fig .
[ tc_mf_sw ] by the blue symbols .
these @xmath3 estimates indicate that @xmath3 will remain roughly proportional to @xmath180 even at large dopings .
the suppression of @xmath3 due to spin - waves increases with increasing hole density relative to the local moment concentration , resulting in saturation of the critical temperature with increasing @xmath0 at about 50% compensation .
the suppression of @xmath3 due to correlated mn - moment fluctuations is also observed in the lsda calculations @xcite .
the trend is illustrated in fig .
[ lda_tc_sw ] where collective fluctuations are accounted for using the spin - wave theory or the monte carlo approach @xcite ; similar trends of suppressed mean - field @xmath3 due to collective mn - moment fluctuations have been predicted by a spin - wave theory using a more elaborate , self - consistent rpa technique on random lattice @xcite .
a larger suppression of the mean - field @xmath3 seen in the _ ab initio _ calculations , compared to the kl kinetic - exchange model results , can be attributed partly to the simpler , three - fold degenerate lsda valence band structure in theories that neglect spin - orbit coupling .
also , the stronger @xmath7 exchange in the lsda theories may result in a weaker spin stiffness of the magnetic system , as the holes are more strongly bound to the mn acceptors and the hole mediated mn - mn coupling has a more short - range character . the enhancement of fluctuation effects in stronger @xmath7 coupled systems is clearly seen in fig . [ lda_tc_sw ] when comparing the lsda results for narrower gap ( weaker @xmath7 exchange ) ( ga , mn)as and wider gap ( stronger @xmath7 exchange ) ( ga , mn)n .
the quantitative discrepancies between kl kinetic - exchange model and lsda results for the mean - field @xmath3 and for the suppression of ferromagnetism due to collective mn - moment fluctuations partly cancel out , leading to similar overall predictions of the curie temperatures in ( ga , mn)as .
based on the @xmath3 analysis alone it is therefore difficult to determine whether the magnetic interactions have a short - range or long - range character in ( ga , mn)as dmss .
theoretically , the localization of the hole around the mn impurity and the range of magnetic mn - mn interactions can be studied using microscopic tba or _
ab initio _ calculations of the charge and moment distribution in the lattice or by mapping the total energy of the dms crystal to the heisenberg hamiltonian @xcite . moving down the anion column in the periodic table from the nitride dmss to antimonides ,
the holes become more delocalized @xcite and , consequently , the mn - mn interactions are more long - range in these microscopic calculations . in ( ga ,
mn)as the lsda theory predicts short - range magnetic coupling while the lda+u or sic - lsda results suggest that the holes which mediate the mn - mn exchange interaction are more delocalized @xcite . combined theoretical and experimental studies of remanent magnetization , micromagnetic parameters , and magneto - transport coefficients , discussed in detail in sections [ bulk - mag - magnet]-[bulk - ac_magtransp ] ,
indicate that in high quality ( ga , mn)as ferromagnets with metallic conductivities ( conductivity increases with decreasing temperature ) the holes are sufficiently delocalized to make the kinetic - exchange model approach applicable .
it is natural to expect that the free - carrier - mediated ferromagnetism picture will also apply in narrower gap antimonide dmss , such as ( in , mn)sb , with even larger conductivities due to the smaller hole effective mass @xcite .
smaller effective mass and larger unit cell volume , as compared to arsenide dmss , explain the smaller curie temperatures in ( iii , mn)sb .
this trend is illustrated in table [ tc_mf_as_sb ] by comparing the respective mean - field kl kinetic - exchange model @xmath3 s . in mn - doped phosphides and
nitrides the suppression of @xmath3 due to effects beyond the mean - field virtual - crystal approximation , seen in the lsda calculations and related to the short - range nature of magnetic interactions , may prevail @xcite .
we note also that the above lsda curie temperature studies do not capture the possible transition of the mn impurity state in the wide gap iii - v s to the highly correlated tri - valent ( @xmath8 ) center with four strongly localized @xmath4-electrons and an empty @xmath4-state deep in the gap , in which case the hole mediated ferromagnetism picture these calculations imply is not applicable @xcite . .
mean - field ( @xmath259 ) , stoner enhanced ( @xmath260 ) , spin - wave suppressed ( @xmath261 ) , and estimated including both stoner enhancement and spin - wave suppression ( @xmath262 ) ferromagnetic transition temperatures in iii - v host semiconductors doped with 5% of mn and with itinerant hole densities @xmath263 and @xmath264 nm@xmath119 . from @xcite . [ cols="^,^,^,^,^,^",options="header " , ] [ mag_xmcd ] a small set of parameters is often sufficient to phenomenologically describe the long - wavelength properties of ferromagnets .
the description , which usually captures all properties that are relevant for applications of magnetic materials , starts from the micromagnetic energy functional @xmath265 $ ] of the spatially dependent magnetization orientation @xmath266 @xcite .
the micromagnetic energy functional treats the long - ranged magnetic dipole interactions explicitly and uses a gradient expansion for other terms in the energy .
the magnitude @xmath267 of the magnetization is one of the micromagnetic parameters characterizing the material .
zeroth order terms in the energy functional are the magnetic anisotropy energy , @xmath268 $ ] , and the zeeman coupling to an external magnetic field if present , @xmath269 .
the leading gradient term , @xmath270 $ ] , referred to as exchange energy in micromagnetic theory , represents the reduction in magnetic condensation energy when the magnetization orientation is not spatially constant .
micromagnetic parameters used to characterize the magnetic anisotropy energy depend on the symmetry of the system .
for example , in a ferromagnet which possesses uniaxial anisotropy with the easy - axis aligned along the @xmath271-direction , @xmath272 where @xmath273 is the uniaxial anisotropy constant . as we explain below
, the magnetic anisotropy of ( ga , mn)as ferromagnets is a combination of the cubic term , the in - plane uniaxial anisotropy , and the uniaxial term induced by the growth - direction lattice - matching strain which often dominates in ( ga , mn)as epilayers . the large anisotropies and relatively small magnetic moments of these dilute magnetic systems rank ( ga , mn)as dmss among hard ferromagnets ( magnetic hardness parameter @xmath274 ) with outstanding micromagnetic properties , including frequently observed single - domain - like characteristics of field - induced magnetization reversals @xcite .
the anisotropy of the exchange term in the micromagnetic functional is often neglected in which case it can be written as , @xmath275 , where @xmath255 is the spin stiffness constant .
collective magnetization dynamics is described by the landau - lifshitz - gilbert equation , @xmath276 where @xmath277 and @xmath278 is the gilbert damping micromagnetic parameter . in this section
we review kl kinetic - exchange model calculations of the micromagnetic parameters of metallic ( ga , mn)as ferromagnets @xcite and predictions based on the microscopic values of these parameters for anisotropy fields , the characteristic size of domains , and the critical current in current - induced magnetization switching @xcite .
magnetocrystalline anisotropy , which is the dependence of the energy of a ferromagnet on the magnetization orientation with respect to crystallographic axes , is a spin - orbit coupling effect often associated with localized electrons in magnetic @xmath4 or @xmath13-shells .
local mn moments in ( ga , mn)as , however , are treated in the kl kinetic exchange model as pure spins @xmath22 with angular momentum @xmath279 and , therefore do not contribute to the anisotropy .
the physical origin of the anisotropy energy in this model is spin - orbit coupling in the valence band @xcite . even within the mean - field approximation to the kl kinetic - exchange model
, magnetic anisotropy has a rich phenomenology which has explained a number of experimental observations in the ( iii , mn)v dmss , including easy axis reorientations as a function of hole density , temperature , or strains in the lattice @xcite .
the remarkable tunability of the magnetic properties of ( ga , mn)as dmss through lattice matching strains is an important byproduct of lt - mbe growth of ferromagnetic films with lattices locked to those of their substrates in the plane perpendicular to the growth axis .
x - ray diffraction studies have established that the resulting strains are not relaxed by dislocations or other defects even for @xmath280 m thick epilayers @xcite .
we mentioned in section [ bulk - struct - lc ] that the lattice constant of relaxed ( ga , mn)as is larger than the lattice constant of gaas , especially so if interstitial mn@xmath19 or as@xmath18 antisites are present in the dms crystal .
( ga , mn)as grown on the gaas substrate is therefore under compressive strain .
tensile strained ( ga , mn)as dmss have been produced by using ( ga , in)as substrates @xcite .
[ 001 ] growth direction strain breaks the cubic symmetry of the ( ga , mn)as resulting in a combined cubic and uniaxial anisotropy form of the energy functional , @xmath281 & = & k_{c1 } \left ( n_x^2 n_y^2 + n_x^2 n_z^2 + n_y^2 n_z^2 \right ) \nonumber \\ & + & k_{c2 } \left ( n_x n_y n_z \right ) ^2+k_un_z^2\ ; .\end{aligned}\ ] ] here @xmath282 and @xmath283 are the two lowest order cubic anisotropy constants @xcite .
strain must be included in the * k*@xmath207*p * description in order to evaluate the uniaxial anisotropy constant @xmath284 . for small strains this
is done by expressing the positional vector @xmath285 in the strained lattice in terms of @xmath286 in the unstrained lattice as @xmath287 , and expanding the kl hamiltonian in lowest order of the strain constants @xmath288 @xcite . in ( ga , mn)as epilayers grown along the [ 001 ] direction ,
the strain constant @xmath289 can be tuned from approximately @xmath290% to @xmath291% . for the larger strain values
, the uniaxial term dominates the total anisotropy energy @xcite . in fig .
[ aniso_dietl ] we show mean - field kl kinetic - exchange model calculations of the uniaxial anisotropy field @xmath292 relative to the @xmath293 magnetization @xmath294 .
@xmath295 corresponds to the minimum external magnetic field necessary to align magnetization @xmath267 along the hard axis .
the figure illustrates the dependence of the easy axis orientation on the hole density and on the sign of @xmath296 .
in particular , the easy - axis is in - plane for compressive strain ( @xmath297 ) and out - of - plane for tensile strain ( @xmath298 ) , consistent with experiment @xcite .
the ability to manipulate the easy axis orientation from in - plane to out - of - plane has many implications for fundamental research on dms materials and is very attractive also from the point of view of potential applications in the magnetic recording technologies .
( divided by @xmath267 ) necessary to align magnetization @xmath267 along the hard axis for compressive ( a ) and tensile ( b ) biaxial strain in ga@xmath83mn@xmath179as film for various values of the spin - splitting parameter @xmath299 .
the easy axis is along [ 001 ] direction and in the ( 001 ) plane at low and high hole concentrations for compressive strain , respectively ( a ) .
the opposite behavior is observed for tensile strain ( b ) .
the symbol [ 100 ] @xmath154 [ 001 ] means that the easy axis is along [ 100 ] , so that @xmath300 is applied along [ 001 ] ( @xmath301 mev corresponds to the saturation value of @xmath267 for ga@xmath302mn@xmath303as ) . from @xcite , width=316 ] (
ga , mn)as epilayers have also a relatively strong in - plane uniaxial anisotropy component @xcite .
the in - plane easy axis is consistently associated with particular crystallographic directions and can be rotated from the [ @xmath30410 ] direction to the [ 110 ] direction by low - temperature annealing . although the origin of the in - plane uniaxial anisotropy has not been established , its dependence on hole concentration ( varied by annealing ) and temperature was modeled successfully within the kl kinetic - exchange model by assuming that it is associated with a small shear strain @xmath305% @xcite .
thermal and quantum magnetization fluctuation effects in metallic ( ga , mn)as dmss have been described by holstein - primakoff boson representation of the mn local spin operators . assuming that fluctuations around the mean - field orientation are small , the relationship between spin - raising and lowering operators and boson creation and annihilation operators is , @xmath306 and @xmath307 ( @xmath308 , @xmath309 @xcite . after integrating out the itinerant hole degrees of freedom in the coherent - state path - integral formalism of many - body problem , the partition function ,
@xmath310 \exp ( -s_{eff}[\bar z z ] ) , \label{part_func}\ ] ] depends only on the bosonic degrees of freedom ( represented by the complex numbers @xmath271 and @xmath311 in the effective action @xmath312 ) .
the independent spin - wave theory is obtained by expanding @xmath313 $ ] up to quadratic order in @xmath271 and @xmath311 , _ i.e. _ , spin excitations are treated as noninteracting bosons .
the spin stiffness parameter @xmath255 is then calculated by fitting the microscopic spin - wave dispersion at long wavelength to the form , @xmath314 to match the conventions used for exchange and anisotropy constants in micromagnetic theory .
typical values of @xmath255 in ( ga , mn)as derived from the above many - body formalism and the kl kinetic - exchange description of the hole bands are shown in fig .
[ stiffness ] .
the values are consistent with the experimental spin stiffness parameter in a ga@xmath315mn@xmath316as measured by ferromagnetic resonance @xcite .
mn@xmath303as as a function of the hole density calculated using the holstein - primakoff representation of fluctuating mn local spins and the kl kinetic - exchange description of the hole bands .
, width=316 ] the damping of small cone - angle magnetization precession in a ferromagnet is parameterized by the gilbert coefficient . for small fluctuations of the mn magnetization orientation in ( ga , mn)as around the easy axis , eq .
( [ llg ] ) can be used to derive an expression for the linear response of a magnetic system to weak transverse fields in terms of the phenomenological constants of micromagnetic theory . for zero external static magnetic field and zero wavevector ( uniform rotation ) the corresponding inverse susceptibility
reads : @xmath317 where @xmath318 and @xmath319 is the frequency of the external rf field perturbation .
microscopically , gilbert damping in ( ga , mn)as dmss was attributed to the @xmath7 exchange - coupling between local mn moments and itinerant holes @xcite .
the elementary process for this damping mechanism is one in which a local - moment magnon is annihilated by exchange interaction with a band hole that suffers a spin - flip .
this process can not by itself change the total magnetic moment since the exchange hamiltonian commutes with the total spin @xmath320 .
net relaxation of the magnetization requires another independent process in which the itinerant hole spin relaxes through spin - orbit interactions .
a fully microscopic theory of the kinetic - exchange contribution to the gilbert coefficient was derived by comparing eq .
( [ classicalsusc ] ) with microscopic linear response theory and by identifying the gilbert coefficient with the dissipative part of the quantum - mechanical susceptibility @xcite , @xmath321 \rangle\theta(t - t^{\prime } ) .
\label{kubo_susc}\ ] ] here @xmath322 and @xmath323 are the mn transverse spin - density operators . as in the microscopic theory of the spin - stiffness , the correlation function ( [ kubo_susc ] ) for the uniform ( @xmath70 ) precession mode was evaluated using the the long - wavelength non - interacting spin - wave form of the partition function ( [ part_func ] ) @xcite .
gilbert coefficients in ( ga , mn)as dmss obtained within this formalism agree quantitatively with experimental values of @xmath278 in homogeneous ( annealed ) systems measured from the width of the ferromagnetic resonance curves , as shown in figs .
[ gilbert ] and [ gilbert_exp ] .
mn@xmath324as samples measured as a function of temperature for [ 001 ] and [ 110 ] dc magnetic - field orientations ( main plot ) and as a function of the field angle at 4k ( inset ) . from @xcite .
, width=316 ] theoretical values of magnetic anisotropy and spin stiffness have been combined to estimate the typical domain size in tensile strained ( ga , mn)as epilayers with out - of - plane easy axis @xcite .
the calculated low - temperature width of a single domain stripe of 1.1@xmath325 m compares favorably with the experimental value of 1.5@xmath325 m seen in the micro - hall probe experiments @xcite . near @xmath3
the discrepancy between theoretical and experimental domain sizes becomes large , however , and has been attributed to critical fluctuations effects not included in the mean - field theory .
the calculated gilbert damping coefficient and magnetic anisotropy constants were used to predict critical currents for spin - transfer magnetization switching @xcite in ( ga , mn)as based tunneling structures @xcite .
spin - polarized perpendicular - to - plane currents in magnetic multilayers with non - parallel spin configurations can transfer spin between magnetic layers and exert current - dependent torques @xcite .
the critical current for magnetization switching is obtained by adding the torque term to the landau - lifshitz - gilbert eq .
( [ llg ] ) @xcite .
critical currents @xmath326 a @xmath327 obtained in these calculations and confirmed in experiment @xcite are two orders of magnitude smaller than those observed typically in metals .
the small moment densities explain a large part of the orders of magnitude reduction in the critical current .
this finding suggests that dms materials have the potential to be particularly useful for exploiting the current - induced magnetization reversal effect in magnetic tunnel junctions .
studies of the temperature dependent resistivity , anisotropic magnetoresistance , and anomalous and ordinary hall effects have been used to characterize dms materials and to test different theoretical models describing these ferromagnets . in this section
we review dc magneto - transport properties of ( ga , mn)as focusing mainly on the metallic regime .
( ga , mn)as materials can exhibit insulating or metallic behavior depending on the doping and post - growth annealing procedures . ( strictly speaking , a material is defined as being metallic if its resistivity is finite in the limit @xmath328 , although for practical reasons this adjective is often used to describe a material whose resistivity decreases with temperature over most or all of the range of temperatures studied in a particular series of experiments . ) in optimally annealed samples with low density of unintentional defects , metallic behavior is observed for mn doping larger than approximately 1.5% @xcite . in section [ micro_picture ]
we already discussed this observation as a consequence of the mott metal - insulator transition due to doping with substitutional mn@xmath18 acceptors .
we also introduced in sections [ micro_picture ] and [ theory - lattice ] some of the theoretical work that has qualitatively addressed magnetic and transport properties of dms systems near the metal - insulator transition . to our knowledge
, a systematic experimental analysis has not yet been performed which would allow a reliable assessment of the theory predictions in this complex and intriguing regime . on the other hand , the dc transport in metallic ( ga , mn)as dmss has a rich phenomenology which has been explored in a number of experimental works and microscopic understanding of many of these effects is now well established .
lsda / cpa band structure calculations combined with kubo linear response theory were used to study correlations between the low temperature conductivity and the density of various defects in the lattice , the hole density , and @xmath3 in metallic ( ga , mn)as @xcite . the theory tends to overestimate the conductivity in low - compensation materials but the overall range of values @xmath329
@xmath330@xmath331 for typical material parameters is consistent with measured data . as illustrated in fig .
[ dc_cond_turek ] and [ dc_cond_tc_nott ] , the theory models capture at a qualitative level the correlation between large conductivities and high curie temperatures seen in experiment @xcite .
comparable absolute values of the @xmath293 conductance and similar trends in the dependence of the conductance on the density of impurities were obtained using the kl kinetic - exchange model and the semiclassical boltzmann description of the dc transport @xcite .
we note , however , that there are differences in the detailed microscopic mechanisms limiting the conductivity in the two theoretical approaches .
scattering in the _ ab initio _ theory is dominated by the @xmath7 exchange potential on randomly distributed mn atoms , and by the local changes of the crystal potential on the impurity sites .
long - range coulomb potentials produced by mn@xmath18 acceptors and other charged defects are omitted in the cpa approach . in the kinetic - exchange effective hamiltonian model
the elastic scattering effects were included using the first order born approximation @xcite .
the corresponding transport weighted scattering rate from the mn@xmath18 impurity has a contribution from both the @xmath7 exchange potential and the long - range coulomb potential , @xmath332 with the scattering matrix elements , @xmath333 here @xmath334 is the host semiconductor dielectric constant , @xmath335 is the multi - component eigenspinor of the kl hamiltonian ( [ kl - pd - ham ] ) and @xmath336 is the corresponding eigenenergy , and the thomas - fermi screening wavevector @xmath337 , where @xmath338 is the density of states at the fermi energy @xcite .
( analogous expressions apply to scattering rates due to other defects . )
the relative strengths of scattering off the @xmath7 exchange and coulomb potentials can be estimated by assuming a simple parabolic - band model of the valence band characterized by the heavy - hole effective mass @xmath339 .
the @xmath7 kinetic - exchange contribution in this approximation is @xmath340 and the scattering rate due to the screened coulomb potential , @xmath341 , is given by the standard brooks - herring formula @xcite . in ( ga , mn)as with @xmath342 nm@xmath119 and @xmath343% , these estimates give @xmath344 mev and @xmath345 mev . for lower compensation ( @xmath346 nm@xmath119 ) , the screening of the coulomb potential is more efficient resulting in values of @xmath341 below 100 mev , but still several times larger than @xmath347 .
( note that these elastic scattering rates are smaller , although by less than a factor of 10 , than other characteristic energy scales such as the fermi energy and the spin - orbit coupling strength in the gaas valence band which partly establishes the consistency of this theoretical approach . )
the dominance of the coulomb potential in born approximation scattering rates for typical chemical compositions is confirmed by calculations based on the six - band kl hamiltonian @xcite .
the good agreement between @xmath293 conductivity values obtained using the _ ab initio _ and the kinetic - exchange model theories should therefore be taken with caution as it may originate , to some extent , from the stronger local @xmath7 exchange in the lsda / cpa theory which partly compensates the neglect of long - range coulomb potentials in this _ ab initio _ approach .
boltzmann transport theory combined with the kl kinetic - exchange model of the ( ga , mn)as band structure is a practical approach for studying magneto - transport effects that originate from the spin - orbit coupling . in section [ bulk - mag - micromag ]
we reviewed the predictions of the model for the role of spin - orbit interaction in magnetic properties of ( ga , mn)as . here
we focus on the anisotropic magnetoresistance ( amr ) effect which is the transport analogue of the magnetocrystalline anisotropy @xcite .
the amr effect can be regarded as the first spintronic functionality implemented in microelectronic devices .
amr magnetic sensors replaced simple horse - shoe magnets in hard - drive read heads in the early 1990 s . with the introduction of giant magnetoresistance based devices in 1997 ,
magnetoresistive sensing launched a new era era in the magnetic memory industry .
in ferromagnetic metals , amr has been known for well over a century .
however , the role of the various mechanisms held responsible for the effect has not been fully sorted out despite renewed interest motivated by practical applications @xcite .
the difficulty in metal ferromagnets partly stems from the relatively weak spin - orbit coupling compared to other relevant energy scales and complex band structure . in ( ga , mn)as ferromagnets with strongly spin - orbit coupled holes occupying the top of the valence band , modeling of the amr effect
can be accomplished on a semiquantitative level without using free parameters , as we review below .
( ga , mn)as epilayers with broken cubic symmetry due to growth - direction lattice - matching strains are usually characterized by two amr coefficients , @xmath348 where @xmath349 is the growth direction and the electrical current @xmath350 .
the three different experimental configurations used to determine @xmath351 and @xmath352 are illustrated in fig .
[ amr_schem ] .
the coefficient @xmath352 is given by combined effects of spin - orbit coupling and the current induced broken symmetry between the two in - plane cubic axes .
the difference between @xmath352 and @xmath351 is a consequence of the lowered symmetry in strained samples .
theoretical predictions for @xmath352 and @xmath351 were compared with measurements in a series of ( ga , mn)as epilayers grown under compressive strain @xcite . the calculated sign of the amr coefficients and the magnitude which varies form @xmath353% in highly mn doped materials to @xmath10% in samples with low mn concentration are consistent with experiment .
reminiscent of the magnetocrystalline anisotropy behavior , the theory predicts rotations as a function of strain of the magnetization direction corresponding to the high ( or low ) resistance state .
calculations illustrating this effect are shown in fig .
[ amr ] and the experimental demonstration in ( ga , mn)as epilayers with compressive and tensile strains is presented in fig .
[ amr_dietl ] . in the top panel of fig .
[ amr_dietl ] , corresponding to ( ga , mn)as material with compressive strain , @xmath354 ( @xmath355 ) is the high resistance configuration , the intermediate resistance state is realized for in - plane @xmath356 ( @xmath357 ) , and the low resistance state is measured when @xmath358 ( @xmath359 ) . in the sample with tensile strain ,
the in - plane @xmath356 and @xmath354 curves switch places , as seen in the bottom panel of fig .
[ amr_dietl ] .
consistent with these experimental observations the theoretical @xmath352 and @xmath351 coefficients are negative and @xmath360 for compressive strain and @xmath361 for tensile strain . and compensation due to as - antisites or total mn doping @xmath362 and compensation due to mn - interstitials .
main panel : out - of - plane amr coefficient as a function of strain for several mn dopings .
lower inset : out - of - plane and in - plane amr coefficients as a function of strain .
upper inset : out - of - plane amr coefficient as a function of the hole density . from @xcite .
, width=316 ] mn@xmath303as grown on gaas substrate under compressive strain for current along the [ 110 ] crystal direction ( upper panel ) and [ 100 ] crystal direction ( middle panel ) for three different orientations of the magnetization as indicated in fig .
[ amr_schem ] .
lower panel : experimental magnetoresistance curves in ga@xmath363mn@xmath364as grown on ( in , ga)as substrate under tensile strain ( current along the [ 110 ] crystal direction ) . from @xcite .
, width=316 ] the anomalous hall effect ( ahe ) is another transport phenomenon originating from spin - orbit coupling which has been used to study and characterize ferromagnetic films for more than one hundred years @xcite .
the difficulties that have accompanied attempts at accurate microscopic description of the effect in metal ferromagnets are reminiscent of those in the amr .
the success of ahe modeling in ( ga , mn)as materials @xcite , reviewed in this section , has had implications also beyond the field of dmss .
it helped to motivate a reexamination of the ahe in transition metals and and in a series of more complex ferromagnetic compounds which has led to a significant progress in resolving the microscopic mechanism responsible for the effect in these materials @xcite . the hall resistance @xmath365 of a magnetic thin film of thickness @xmath4 is empirically observed to contain two distinct contributions , @xmath366 @xcite .
the first contribution arises from the ordinary hall effect ( ohe ) which is proportional to the applied magnetic field @xmath367 , and the second term is the ahe which may remain finite at @xmath368 and depends instead on the magnetization . @xmath369 and @xmath370 are the ohe and ahe coefficients respectively . to put the ahe studies in dmss in a broader perspective ,
the following paragraphs offer a brief excursion through the history of ahe theory and an overview of microscopic mechanisms discussed in this context in the literature ( a more detailed survey is given , _
e.g. _ , in @xcite ) .
the first detailed theoretical analysis of the ahe was given by karplus and luttinger @xcite , where they considered the problem from a perturbative point of view ( with respect to an applied electric field ) and obtained a contribution to the anomalous hall conductivity in systems with spin - orbit coupled bloch states given by the expression : @xmath371 where @xmath372 is the fermi occupation number of the bloch state @xmath373 .
this contribution is purely a property of the perfect crystal and has become known in recent years as the intrinsic ahe .
it leads to an ahe coefficient @xmath370 proportional to @xmath374 and therefore can dominate in metallic ferromagnets that have a relatively large resistivity .
the intrinsic ahe is related to bloch state berry phases in momentum space and depends non - perturbatively on spin - orbit interaction strength when degeneracies in momentum space are lifted by spin - orbit coupling @xcite .
this point is particularly relevant for ferromagnetic semiconductors because all carriers that contribute to transport are located near particular points in the brillouin zone , often high symmetry points at which degeneracies occur . shortly after the seminal work of karplus and luttinger , smit proposed a different interpretation of the ahe based on a picture of asymmetric spin - dependent skew scattering off impurity potential involving spin - orbit coupling @xcite .
analytically , the skew scattering appears in the second order born approximation applied to the collision term of the boltzmann transport equation .
this mechanism gives a contribution to @xmath375 , _
i.e. _ , proportional to the density of scatterers and dependent on the type and range of the impurity potential .
the ahe conductivity has a number of contributions in addition to the skew scattering and intrinsic contributions , that can originate either from spin - orbit coupling in the disorder scattering or spin - orbit coupling in the bloch bands . among these
, side jump scattering has been identified has an important contribution @xcite .
side jump due to spin - orbit coupling in the bloch bands appears as a ladder diagram vertex correction to the intrinsic anomalous hall effect and its importance depends on the nature of that coupling @xcite .
the ratio of intrinsic and skew - scattering contributions to the ahe conductivity can be approximated , assuming a single spin - orbit coupled band and scattering off ionized impurities , by the expression @xcite : @xmath376 where @xmath377 is the ratio of the density of ionized impurities and the carrier density , @xmath378 is the average distance between carriers in units of bohr radius , @xmath379 is the mean free path , and @xmath380 , varying slightly with scattering length . for
the short range scattering potential considered by luttinger @xcite and nozieres and lewiner @xcite , @xmath381 : @xmath382 the estimate is a useful first guess at which mechanism dominates in different materials but one must keep in mind the simplicity of the models used to derive these expressions . in ( ga , mn)as , eq . ( [ sjtoss_ion ] ) gives a ratio of intrinsic to skew scattering contribution of the order of 50 and the intrinsic ahe is therefore likely to dominate @xcite .
consistently , the experimental @xmath370 in metallic ( ga , mn)as dmss is proportional to @xmath374 @xcite .
microscopic calculations of the intrinsic low - temperature ahe conductivity in ( ga , mn)as were performed by taking into account the berry phase anomalous velocity term in the semiclassical boltzmann equation leading to eq .
( [ sigaheberry ] ) @xcite , and by applying the fully quantum mechanical kubo formalism @xcite . in both approaches ,
the kl kinetic - exchange model was used to obtain the hole band structure . in the kubo formula ,
the dc hall conductivity for non - interacting quasiparticles at zero external magnetic field is given by @xcite @xmath383 the real part of eq .
( [ sigahekubo ] ) in the limit of zero scattering rate ( @xmath384 ) can be written as , @xmath385 } { ( e_{n\vec{k}}-e_{n'\vec{k}})^2}. \label{sigahekubo2}\end{aligned}\ ] ] realizing that the momentum matrix elements @xmath386 , eq .
( [ sigahekubo2 ] ) can be shown to be equivalent to eq .
( [ sigaheberry ] ) .
the advantage of the kubo formalism is that it makes it possible to include finite lifetime broadening of the quasiparticles in the simulations .
( see section [ bulk - magtransp - amr ] for the discussion of quasiparticle scattering rates in ( ga , mn)as . whether or not the lifetime broadening is included , the theoretical anomalous hall conductivities are of order 10 @xmath330 @xmath331 for typical ( ga , mn)as dms parameters . on a quantitative level , non - zero @xmath49 tends to enhance @xmath387 at low mn doping and suppresses @xmath387 at high mn concentrations where quasiparticle broadening due to disorder becomes comparable to the strength of the kinetic - exchange field . a systematic comparison between theoretical and experimental ahe data is shown in fig . [ ahe_teor_exp ] @xcite .
the results are plotted versus nominal mn concentration @xmath180 while other parameters of the measured samples are listed in the figure legend .
experimental @xmath387 values are indicated by filled squares and empty triangles correspond to theoretical data obtained for @xmath388 .
results shown in half - open triangles were obtained by solving the kubo formula for @xmath387 with non - zero @xmath49 due to scattering off mn@xmath18 and as - antisites , or mn@xmath18 and mn@xmath19 impurities .
the calculations explain much of the measured low - temperature ahe in metallic ( ga , mn)as dmss , especially so for the mn@xmath19 compensation scenario .
the largest quantitative discrepancy between theory and experiment is for the @xmath389% material which can be partly explained by a non mean - field - like magnetic behavior of this specific , more disordered sample . in dmss
the ahe has played a key role in establishing ferromagnetism and in providing evidence for the hole - mediated coupling between mn local moments @xcite .
the dominance of the ahe in weak field measurements ( see fig . [ ahe_nott ] ) allows the hall resistance to serve also as a convenient proxy for the magnetization . on the other hand
, the same property can obscure hall measurements of the hole density .
if the magnetization is not fully saturated at low fields , for example , then @xmath390 will increase with increasing external field @xmath367 through the dependence of @xmath387 on @xmath391 , and hole densities derived from the slope of @xmath392 will be too low .
accurate determination of hole densities in dmss is essential , however , and the hall effect is arguably the most common and accurate non - destructive tool for measuring the level of doping in semiconductors .
hall experiments performed in high magnetic fields to guarantee magnetization saturation seem a practical way for separating ahe contributions , especially in samples showing weak longitudinal magnetoresistance @xcite .
hole density measurements performed using this technique assume that the hall factor , @xmath393 with @xmath394 , is close to 1 despite the multi - band spin - orbit coupled nature of hole dispersion in ( ga , mn)as ferromagnets . in the following paragraphs
we briefly review a theoretical analysis of this assumption @xcite .
microscopic calculations in non - magnetic p - type gaas with hole densities @xmath395 cm@xmath119 have shown that @xmath397 can vary between 0.87 and 1.75 , depending on doping , scattering mechanisms , and on details of the model used for the gaas valence band @xcite .
an estimate of the influence on @xmath397 from the spin - splitting of the valence band and from the anomalous hall term is based on the kl kinetic - exchange model description of the hole band structure .
the hall conductivity has been obtained by evaluating the kubo formula at finite magnetic fields that includes both intra - band and inter - band transitions .
the approach captures the anomalous and ordinary hall terms on equal footing @xcite .
many of the qualitative aspects of the numerical calculations , shown in fig .
[ rh8 ] , can be explained using a simple model of a conductor with two parabolic uncoupled bands .
note that the typical scattering rate in ( ga , mn)as epilayers is @xmath398 mev and that the cyclotron energy at @xmath399 t is @xmath400 mev , _
i.e. _ , the system is in the strong scattering limit , @xmath401 . in this limit
, the two band model gives resistivities : @xmath402 where the indices 1 and 2 correspond to the 1st and 2nd band respectively , the total density @xmath403 , and the zero - field conductivity @xmath404 .
( [ rho_2band ] ) suggests that in the strong scattering limit the multi - band nature of the hole states in ( ga , mn)as should not result in a strong longitudinal magnetoresistance .
this observation is consistent with the measured weak dependence of @xmath405 on @xmath367 for magnetic fields at which magnetization in the ( ga , mn)as ferromagnet is saturated @xcite .
the simple two - band model also suggests that the hall factor , @xmath397 , is larger than one in multi - band systems with different dispersions of individual bands .
indeed , for uncoupled valence bands , _
i.e. _ when accounting for intra - band transitions only , the numerical hall factors in the top panels of fig .
[ rh8 ] are larger than 1 and independent of @xmath406 as also suggested by eq .
( [ hall_two_band ] ) . the suppression of @xmath397 when spin - orbit coupling is turned on , shown in the same graphs , results partly from depopulation of the angular momentum @xmath99 split - off bands .
in addition to this `` two - band model '' like effect , the inter - landau - level matrix elements are reduced due to spin - orbit coupling since the spinor part of the eigenfunctions now varies with the landau level index . in ferromagnetic ga@xmath83mn@xmath84as
the bands are spin - split and higher bands depopulated as @xmath180 increases . in terms of @xmath397 , this effect competes with the increase of the inter - landau - level matrix elements since the spinors are now more closely aligned within a band due to the exchange field produced by the polarized mn moments . increasing @xmath180 can therefore lead to either an increase or a decrease in @xmath397 depending on other parameters , such as the hole density ( compare top right panels of fig .
[ rh8 ] ) .
inter - band transitions result in a more single - band like character of the system , _
i.e. _ @xmath397 is reduced , and the slope of the @xmath392 curve now depends more strongly on @xmath406 .
although the ahe and ohe contributions to @xmath407 can not be simply decoupled , the comparison of numerical data in the four panels confirms the usual assumption that the ahe produces a field - independent off - set proportional to magnetization and @xmath374 .
the comparison also suggests that after subtracting @xmath408 , @xmath397 can be used to determine the hole density in ( ga , mn)as with accuracy that is better than in non - magnetic gaas with comparable hole densities . for typical hole and mn densities in experimental ( ga , mn)as epilayers the error of the hall measurement of @xmath0 is estimated to be less than @xmath409 @xcite .
typical fermi temperatures , @xmath410 , in ferromagnetic ( ga , mn)as are much larger than the curie temperature relegating direct fermi distribution effects of finite temperature to a minor role in transport
. the carrier - mediated nature of ferromagnetism implies , however , strong indirect effects through the temperature dependence of the magnetization .
a prime example is the ahe which from the early studies of ( iii , mn)v dmss has served as a practical tool to accurately measure curie temperatures @xcite . a rough estimate of @xmath3 can be inferred also from the temperature dependent longitudinal resistivity which exhibits a shoulder in the more metallic ( optimally annealed ) samples and a peak in the less metallic ( as - grown ) materials near the ferromagnetic transition @xcite .
an example of this behavior is shown in fig .
[ rho ] for a ( ga , mn)as material with 8% nominal mn - doping @xcite .
the shoulder in @xmath411 has been qualitatively modeled using the mean - field , kl kinetic - exchange hamiltonian given by eq .
( [ kl - pd - ham ] ) .
solutions to the boltzmann equation @xcite are shown in fig .
the temperature dependence of the longitudinal conductivity follows in this theory from variations in the parameters derived from the spin - polarized hole band structure ( _ e.g. _ fermi wavevector ) and from variations in screening of impurity coulomb potentials .
the peak in resistance near @xmath3 has been discussed in terms of scattering effects beyond the lowest order born approximation and by using a network resistor model @xcite .
it has also been suggested that this transport anomaly in more highly resistive dmss is a consequence of the change in localization length caused by the ferromagnetic transition @xcite .
, vs. t for a mn doping of 5% and hole densities @xmath412 and @xmath412 from @xcite.,width=316 ] above the curie temperature , measurements of @xmath405 have been used to estimate the value of @xmath75 .
assuming scattering off the @xmath7 exchange potential at randomly distributed paramagnetic mn@xmath18 impurities and parabolic hole bands , the corresponding contribution to the resistivity is approximated by , @xmath413 , \label{high_t_rho_xx}\ ] ] where @xmath414 and @xmath415 are the transverse and longitudinal magnetic susceptibilities @xcite .
this theory overestimates critical scattering , particularly near the curie temperature where the susceptibility diverges .
far from the transition on the paramagnetic side , however , fitting eq .
[ high_t_rho_xx ] to experimental magnetoresistance data gives an estimate of the @xmath75 @xcite which is consistent with values inferred from spectroscopical measurements @xcite .
the prospects for new technologies based for example on materials in which the ferromagnetic transition can be controlled by light or on ( iii , mn)v faraday isolators monolithically integrated with existing semiconductor lasers , have motivated research in magneto - optical properties of dmss @xcite .
apart from these applied physics interests , ac probes have been used to study dms materials by a wide range of experimental techniques . in sections
[ micro - submn_pd ] and [ bulk - mag - magnet - manymn ] we mentioned x - ray spectroscopies ( core - level photoemission and xmcd ) used to characterize mn 3@xmath4 states and detect the sign and magnitude of the @xmath7 exchange coupling .
dispersions of hole bands in dmss have been studied by angle - resolved photoemission with ultraviolet excitations @xcite and infrared - to - ultraviolet spectroscopic ellipsometry @xcite .
raman scattering induced by excitations in the visible range was used as an alternative means of estimating hole densities @xcite .
spectroscopic studies of isolated mn(@xmath5+hole ) impurities in the infrared region provided key information on the valence of mn in ( ga , mn)as , as discussed in section [ micro - submn ] , and cyclotron resonance measurements were used to study highly mn - doped dms materials in this frequency range @xcite .
the microwave epr and fmr experiments , mentioned in section [ micro - submn ] and [ bulk - mag - micromag ] , have been invaluable for understanding the nature of mn in iii - v hosts at low and high dopings , and for characterizing magnetocrystalline anisotropies and magnetization dynamics in ferromagnetic materials . in this section
we review studies of some of the magneto - optical responses @xcite , particularly magnetic circular dichroism ( mcd ) , in the visible range and infrared absorption @xcite .
optical absorption due to electron excitations across the band gap is a standard characterization technique in semiconductors . in ( ga , mn)as
, the absorption occurs in the visible range and the position of its edge on the frequency axis depends on the circular polarization of the incident light .
analysis of this magneto - optical effect provides information on the @xmath7 exchange induced band splitting and on doping in the dms material @xcite .
the schematic diagram in fig . [ vis_mcd_schem ] shows that for a given sign of the exchange coupling , the order of absorption edges corresponding to the two circular photon polarizations can reverse in p - type materials , compared to systems with a completely filled valence band .
calculations for ( ga , mn)as that include this moss - burnstein effect were carried out using the mean - field kl kinetic - exchange model @xcite .
the resulting absorptions @xmath416 of the @xmath417 circularly polarized light , and mcd , defined as @xcite @xmath418 } { { \rm re}[\sigma_{xx}(\omega)]},\ ] ] are shown in fig .
[ dietl01 ] . as suggested in the cartoon , the sign of the mcd signal in ( ga
, mn)as is opposite to the one obtained in bulk ( ii , mn)vi dmss where the sense of band splittings is the same as in ( ga , mn)as but mn substituting for the group - ii element is an isovalent neutral impurity @xcite .
( the moss - burnstein sign change in mcd was also observed in co - doped p - type ( ii , mn)vi quantum well @xcite . )
mev ( corresponding to @xmath293 mean - field from local mn moments of density @xmath419% ) and hole concentration @xmath420 cm@xmath119 .
inset shows how the fermi sea of the holes reverses the relative positions of the edges corresponding to @xmath421 and @xmath422 polarizations in agreement with experimental findings .
bottom panel : spectral dependence of magnetic circular dichroism in the optical range in ( ga , mn)as computed for hole concentration @xmath420 cm@xmath119 and various spin - splitting parameters @xmath299 .
the magnitudes of mcd at given @xmath299 are normalized by its value at 1.78 ev . from @xcite.,width=307 ] experimentally , the incorporation of several per cent of mn in gaas strongly enhances mcd , as shown in fig .
[ matsukura_mcd ] @xcite , and the sign of the signal near band - gap frequencies is consistent with the above theory which assumes antiferromagnetic @xmath7 exchange coupling @xcite .
mn@xmath84as films at t = 5 k and b = 1 t. the spectrum of gaas is magnified ten times because the signal is weaker than that of ga@xmath83mn@xmath84as . from @xcite.,width=326 ] in ( iii , mn)v dmss , light absorption can occur at sub - band - gap frequencies due to valence - band to mn impurity band excitations @xcite in the more insulating materials and due to intra - valence - band excitations @xcite in the more metallic systems , as illustrated schematically in fig .
[ infra_schem ] . the infrared absorption associated with substitutional mn impurities
is spectrally resolved from higher energy excitations to donor levels of the most common unintentional defects , such as the mn@xmath19 interstitials and as@xmath18 antisites in ( ga , mn)as , and therefore represents another valuable probe into intrinsic properties of these systems . since infrared wavelengths are much larger than typical ( sub - micron ) dms epilayer widths , the absorption is related to the real part of the conductivity by @xcite @xmath423}{y+y_0}\,\,,\ ] ] where @xmath424 and @xmath425 are the admittances of the substrate and free space , respectively . model hamiltonians ( [ spinlatham ] ) ( see section [ theory - lattice ] ) combined with dynamical mean - field - theory or monte carlo simulations were used to study the role of the impurity band in infrared absorption @xcite .
the impurity band forms in this theory when the strength of the model effective exchange interaction @xmath90 is comparable to the width of the main band , characterized by the hopping parameter @xmath426 .
a non drude peak is observed in the frequency - dependent conductivity , associated with transitions from the main band to the impurity band . the behavior is illustrated in fig . [ alvarez ] together with the predicted temperature dependence of the absorption spectra obtained by the monte carlo technique . as discussed in section [ theory - lattice
] , these model calculations are expected to apply to systems with strong @xmath7 exchange coupling , like ( ga , mn)p and possibly also to ( iii , mn)v dmss which are strongly compensated due to the presence of unintentional donor defects .
( impurity - band mediated ferromagnetism does not occur in uncompensated samples . ) vs. @xmath319 for a 10@xmath14010 periodic system with 26 spins ( @xmath427 ) , @xmath428 , @xmath429 , and for different temperatures , as indicated .
inset : drude weight , @xmath258 , vs. temperature @xmath430 @xcite .
[ alvarez],width=307 ] theoretical infrared absorption spectra calculated using the @xmath431 model for ( ga , mn)as dmss with delocalized holes in the semiconductor valence band , plotted in fig .
[ infra_sinova ] , show a similar non - drude characteristics with a peak near the excitation energy of 220 mev .
the underlying physics is qualitatively different , however , as the peak in these kl kinetic - exchange model calculations originates from heavy - hole to light - hole intra - valence - band transitions @xcite .
these results were obtained by evaluating the kubo formula for ac conductivity assuming non - interacting holes and modeling disorder within the first order born approximation ( see eq .
( [ gamma ] ) in section [ bulk - magtransp - dc_cond ] ) . in fig .
[ yang2 ] we show theoretical predictions of exact diagonalization studies based on the kl kinetic - exchange hamiltonian but treating disorder effects exactly in a finite size system .
the results correct for the overestimated dc conductivity in the former model , which is a quantitative deficiency of the born approximation as already mentioned in section [ bulk - magtransp - dc_cond ] . at finite frequencies , the theoretical absorption in these metallic ( ga , mn)as dmss is almost insensitive to the way disorder is treated in the simulations , as see from figs .
[ infra_sinova ] and [ yang2 ] .
$ ] and absorption coefficient @xmath432 for carrier densities from @xmath433 to @xmath434 in the direction indicated by the arrow , for ga@xmath302mn@xmath303as @xcite .
disorder is treated within the first order born approximation .
the dot - dashed line is the experimental absorption curve for a sample with 4% mn doping from @xcite .
, width=307 ] nm@xmath119 , @xmath435 nm@xmath119 ( @xmath436% ) . from @xcite.,width=326 ] experimental infrared absorption studies in ferromagnetic ( ga , mn)as epilayers exhibit several common features summarized in fig .
[ singlet ] @xcite .
ferromagnetic materials ( @xmath437% curves in fig . [ singlet ] ) show a non - drude behavior in which the conductivity increases with increasing frequency in the interval between @xmath438 and @xmath439 mev , a broad absorption peak near @xmath440 mev that becomes stronger as the sample is cooled , and a featureless absorption up to approximately 1 ev . as seen in fig .
[ singlet ] , the peak is absent in the reference , lt - mbe - grown gaas sample confirming that the infrared absorption in ferromagnetic ( ga , mn)as is related to changes in the band structure near the fermi energy induced by mn impurities . % sample ( dashed line ) , a ferromagnetic @xmath437% sample ( thick solid line ) , and the lt gaas film ( thin black line ) .
bottom panel : temperature dependence of conductivity for @xmath437% sample on a log scale for @xmath441 and @xmath442 . from @xcite .
, width=307 ] the presence of a finite - frequency peak in both impurity band and kl kinetic - exchange models for the infrared conductivity leads to an ambiguity in the interpretation of existing data which has for the most part been taken in as grown , presumably heavily compensated material .
the metallic behavior of the @xmath437% material below @xmath3 , seen in the lower panel of fig .
[ singlet ] , favors the inter - valence - band absorption scenario . on the other hand , the large compensation likely present in as - grown low-@xmath3 ( ga
, mn)as suggests that many holes may be strongly localized and that both absorption mechanisms may contribute to the measured absorption peak .
experiments in a series of samples interpolating between as grown and optimally annealed limits , analogous to the resistance - monitored annealing studies @xcite , should enable a clear interpretation of infrared absorption spectra in ( ga , mn)as dmss .
these studies will hint towards necessary refinements of the simplified theories used so far , _
e.g. _ inclusion of the energy dependence of @xmath75 and a more quantitative theory of the impurity band model .
the support for either scenario by these experiments has to be considered in conjunction with other available data in a self - consistent picture , _
e.g. _ in an impurity band picture @xmath3 is predicted to approach zero as the system reaches zero compensation whereas the kl kinetic - exchange model has an opposite trend .
systems with local moments coupled to itinerant electrons are common in condensed matter physics and exhibit a wide variety of behaviors .
ferromagnets are far from the most common low - temperature states .
for that reason it is useful to ask how ( iii , mn)v materials , and ( ga , mn)as with its robust ferromagnetic order in particular , fit in this larger context . this general qualitative analysis can help to identify some of the key factors that might limit the strength of ferromagnetic interactions in the highly doped and strongly @xmath7 coupled dms ferromagnets for which mean - field theory predicts the highest curie temperatures . an important class of materials that has been very extensively studied is heavy fermions , in which f - electron local moments are exchange coupled to band electrons @xcite .
kondo lattice models , which are believed to qualitatively describe heavy fermion systems , assume that local moments exist at each lattice site .
models of dms systems which make a virtual crystal approximation ( see section [ micro - submn_pd]and [ bulk - mag - tc - mf ] ) place moments on all lattice sites and are therefore kondo lattice models , often with specific details that attempt to capture some of the peculiarities of specific dms materials . theories of kondo lattice model often start from the comparison of the rkky ( see section [ micro - general ] ) and kondo temperature scales @xcite .
the characteristic rkky temperature refers to the strength of interactions between local moments mediated by a weakly disturbed carrier system and is proportional to the mean - field @xmath3 given by eq .
( [ tc ] ) .
the kondo scale refers to the temperature below which strong correlations are established between an isolated local moment and the carrier system with which it interacts .
standard scale estimate formulas @xcite applied to the case of dms ferromagnets imply that the kondo scale is larger than the rkky scale when the mean - field exchange coupling , @xmath443 , is larger than the fermi energy @xmath157 of the hole system , in other words in the strong coupling regime .
the kondo scale falls rapidly to small values at weaker coupling .
( in heavy - fermion materials the kondo temperature scale is larger than the rkky temperature scale . ) , while short - range order may still exists above the curie temperature . , width=240 ] optimally annealed metallic ( ga , mn)as materials are on the weak - coupling side of this boundary , but ( ga , mn)n and possibly ( ga , mn)p may be starting to reach toward the strong coupling limit ( if the simple s=5/2 local moment model still applies in these materials ) . as the strong coupling limit
is approached , quantum fluctuations ( see discussion below eq .
( [ two - spin - ham ] ) ) will play a greater role , reducing the saturation moment per mn and eventually driving down the ferromagnetic transition temperature .
when the kondo scale is much larger than the rkky scale , the local moments are screened out by strongly correlated band electron spin - fluctuations and effectively disappear before they have the opportunity to couple . on the weak coupling side ,
rkky interactions in kondo lattice models tend to lead to ferromagnetism only when the number of itinerant electrons per moment is small , _
i.e. _ , only when at least the near - neighbor rkky interaction is ferromagnetic @xcite .
one of the surprising features of ( ga , mn)as is the property that ferromagnetism still occurs when the number of itinerant electrons per moment is @xmath353 . as we have mentioned in section [ bulk - mag - tc - sw ] , this property follows from the specific multi - band electronic structure and spin - orbit coupling at the top of the valence band .
nevertheless , frustrating antiferromagnetic rkky interactions , and exchange interactions that promote non - collinear magnetic states ( see section [ bulk - mag - magnet ] ) , will eventually become important for sufficiently large carrier densities .
as this regime is approached from the ferromagnetic side , the transition temperature will be suppressed .
these tendencies are summarized schematically in fig .
[ discussion ] .
similar considerations apply in assessing the robustness of ferromagnetism in the ordered state as characterized by the spin - stiffness micromagnetic parameter @xcite . we have used this approach in section [ bulk - mag - tc - mf ] when analyzing the limitations of the mean - field theory in ( ga , mn)as . starting from the ferromagnetic state
, long - wavelength spin - orientation modulation will tend to lower the energy of some mn - mn interactions for sufficiently high carrier densities .
the spin - stiffness will weaken as the frustrated magnetism regime is approached , until finally the energy of ferromagnetic spin - waves will become negative signaling the instability of this state @xcite .
similarly for sufficiently strong coupling , the band system will be ( nearly when spin - orbit coupling is included ) fully spin - polarized and the cost of spin - orientation spatial variation will be borne mainly by the kinetic energy of the hole system and no longer increase with exchange coupling . as shown schematically in the lower panel of fig .
[ discussion ] , hole spins in this regime are locally antiferromagnetically locked to the fluctuating mn moment orientations . for relatively small hole fermi energies the kinetic energy cost of these fluctuations is small , resulting in soft spin - wave excitations of the magnetic system @xcite . in this regime
long - range ferromagnetic order disappears at temperatures smaller than the mean - field @xmath3 , _ i.e. _ , short - range order still exists above the curie temperature .
the white bottom - left area in the panels of fig . [ discussion ] qualitatively depicts the parameter range in which the ferromagnetic rkky mean - field state applies . here
@xmath3 increases with the carrier density , local moment density and ( quadratically ) with the strength of the @xmath7 exchange coupling . for a fixed ratio of @xmath242 ,
increasing @xmath444 corresponds to moving only slowly ( as @xmath445 ) upwards in the diagrams .
this may explain why ( ga , mn)as materials with larger mn doping and with similar hole compensations as in the ferromagnetic low local moment density systems do not show any marked weakening of the ferromagnetic state .
with @xmath242 still fixed , attempts to increase @xmath3 in ( iii , mn)v dmss by increasing the @xmath75 constant , in _
e.g. _ ternary host alloys of ga(as , p ) , might at some point reach the boundary of the soft spin - wave ( kondo screened ) state .
similarly , the kondo lattice model allows only a limited space for enhancing the robustness of the ferromagnetic state by tweaking the carrier and local moment densities independently . in this case
moving horizontally from the boundary of the frustrated rkky ( ferromagnetically unstable ) state is accompanied by approaching vertically the soft spin - wave ( kondo screened ) regime , and vice versa . viewed from the opposite perspective , however , it is astonishing that a window in this parameter space has been found by the material research of ( iii , mn)v compounds for robust dms ferromagnets with curie temperatures close to 200 k. the diagrams do not imply any general physical mechanism that limits @xmath3 in these materials below room temperature .
our remarks on the cartoons in fig . [ discussion ] refer to the properties of kondo lattice models which have been adjusted to reflect peculiarities of the zinc - blende semiconductor valence band .
we have so far neglected the importance of disorder and of coulomb scattering in dmss , and these can modify some parts of the simple qualitative picture depicted in fig [ discussion ] .
this is particularly true in the very low - density isolated mn limit , _
i.e. _ , very strong exchange coupling limit .
because of coulomb attraction between valence band holes and the charged mn ion which carries the local moment , a total angular momentum @xmath113 isolated bound state is created , as we have explained in section [ micro - submn_pd ] , instead of the strongly correlated kondo singlet .
the importance of coulomb interactions and disorder are lessened by screening and pauli exclusion principle effects when both the mn density and the carrier density are high . in this section
we will narrow down the discussion of magnetism in local moment systems to semiconducting compounds , focusing on the phenomenology of ferromagnetic dmss other than the ( iii , mn)v materials .
almost any semiconducting or insulating compound that contains elements with partially filled @xmath4 or @xmath13 shells ( local moments ) will order magnetically at a sufficiently low temperature .
semiconductors and insulators with high density of magnetic moments usually order antiferromagnetically however , although ferromagnetism does occur in some cases .
a famous example of a ferromagnetic system that can be regarded as a doped semiconductor is provided by the manganite family ( _ e.g. _ la@xmath83sr@xmath179mno@xmath2 ) whose ferromagnetism is favored by the double - exchange mechanism and occurs over a wide range of transition temperatures from below 100 k to nearly 400 k. the onset of magnetic order in these systems is accompanied by a very large increase in conductivity . for a review ,
see for example @xcite .
other well known dense moment ( of order one moment per atom ) ferromagnetic semiconductors with strong exchange interaction between itinerant and local spins include eu- and cr - chalcogenides @xcite , such as rock - salt euo and spinel cdcr@xmath131se@xmath123 with curie temperatures 70 k and 130 k , respectively .
dms systems in which magnetic atoms are introduced as impurities have moments on only a small fraction of all atomic sites .
the mechanisms that control magnetic order are therefore necessarily associated with the properties of these impurities .
the coupling between moments will generally depend on the locations of the dilute moments in the host lattice , on the doping properties of the magnetic impurities , and on other dopants and defects present in the material .
it seems plausible therefore that when dms systems are ferromagnetic , their magnetic and magnetotransport properties will be more sensitive to engineerable material properties .
this review has concentrated on ( ga , mn)as and related materials in which , as we have explained , substitutional mn acts both as an acceptor and as a source of local moments .
ferromagnetism is carrier mediated and it has been demonstrated that it persists to surprisingly high temperatures .
more may be achieved in the future by tweaking these materials . on the other hand
there is a vast array of alternate dms materials that could be contemplated .
research to date has only scratched the surface of the volume of possibilities - we are truly still at the beginning of the road in studying diluted moment magnetism in semiconductors .
each system brings its new challenges .
the interpretation of simple magnetic and transport characterization measurements is often not immediately obvious , in particular because of the possibility that the moments will segregate into crystallites of one of a variety of available dense moment minority phases which are often thermodynamically more stable .
in addition , magnetic properties will certainly depend in general not only on the dilute moment density , which normally is well controlled and variable , but also on the partitioning of local moments among many available sites in the host crystal which is not always known and is usually much harder to control .
the search for promising dms materials would be simplified if _
ab initio _ dft methods had reliable predictive power .
unfortunately this luxury appears to be absent in many cases because of extreme sensitivity of magnetic properties to details of the electronic structure and because of strong correlation effects that are often present in these systems .
we mention briefly in the following paragraphs some of the other classes of diluted magnetic semiconductors which have been studied .
the class of dms ferromagnets that is closest to ( iii , mn)v materials is ( ii , mn)vi compounds co - doped with group v element acceptors .
examples include p-(zn , mn)te : n @xcite and p-(be , mn)te : n @xcite .
these materials differ from ( ga , mn)as mainly because the local moments and holes are provided by different types of impurities and can be controlled independently .
although the physics behind ferromagnetism seems to be very similar in the two classes of materials , the highest ferromagnetic transition temperatures that have been achieved are much smaller in the case of co - doped ( ii , mn)vi materials , @xmath446 rather than @xmath447 . the difference is explained partly by difficulty in achieving the same extremely high hole doping ( @xmath448 ) in ( ii , mn)vi that has been achieved in ( ga , mn)as and partly by a favorable interplay between electrostatic and magnetic effects in the ( iii , mn)v materials . in ( iii , mn)v materials , unlike ( ii , mn)vi materials ,
the mn moment is charged and attracts holes .
the tendency of holes to have a higher density near mn sites tends to increase the effective strength of the @xmath7 exchange interaction .
this effect is magnified when two mn moments are on neighboring cation sites . in ( ga
, mn)as the interaction between mn moments on neighboring cation positions is ferromagnetic , compared to the strongly antiferromagnetic interactions seen in ( ii , mn)vi materials .
( for a discussion of the interplay between electrostatic and magnetic interactions see @xcite . ) in p - doped ( ii , mn)vi materials , competition between antiferromagnetic near neighbor interactions and the longer range carrier - mediated ferromagnetic interactions suppresses the magnetic ordering temperature .
this competition apparently does not occur in ( iii , mn)v ferromagnets with large hole densities .
( zn , mn)o is an interesting ii - vi counterpart of the nitride iii - v dms @xcite . with advances in oxide growth techniques (
zn , mn)o , can be considered to be much like other ( ii , mn)vi dms materials and its investigation was originally motivated by theoretical work @xcite that extrapolated from experience with ( iii , mn)v dms ferromagnets and predicted large @xmath3 s .
studies of this material have provided clear evidence of strong @xmath7 exchange but , so far , have led to inconsistent conclusions about the occurrences of long range magnetic order @xcite .
tetrahedral dms materials doped with transition metal atoms other than mn have shown promising results . for instance
( zn , cr)te is apparently homogeneous , has the required coupling between local moments and carriers @xcite and curie temperature as high as 300 k has been reported for this material .
it may be , though , that the ferromagnetism is due to superexchange interactions rather than being carrier mediated since it occurs at very small ratios of the carrier density to the moment density @xcite .
another interesting material with cr moments is ( ga , cr)n which exhibits ferromagnetism at @xmath449 @xcite .
the question still at issue in this material is the possible role of dense moment precipitates .
traditional groups of ferromagnetic dmss also include ( iv , mn)vi solid solutions with the rock - salt structure @xcite .
although the band structures of iv - vi and iii - v semiconductors are quite different , these dms ferromagnets ( _ e.g. _ ( pb , sn , mn)te ) appear to have a carrier - mediated mechanism quite similar to that of ( ga , mn)as . holes with densities up to 10@xmath450 cm@xmath119 are supplied in these materials by cation vacancies , rather than by mn substitution for the divalent cations .
the reported curie temperatures in ( pb , sn , mn)te are below 40 k @xcite .
dms ferromagnetism with si or ge as the host semiconductor is obviously attractive because of their greater compatibility with existing silicon based technology . in si , mn impurities favor interstitial position which significantly complicates the synthesis of a uniform dms system .
mn in ge , on the other hand , is a substitutional impurity and ferromagnetism has been reported in mbe grown @xmath451 thin film dmss @xcite .
careful studies of @xmath451 @xcite have demonstrated that slow low - temperature growth is required to avoid the formation of thermodynamically stable dense - moment ferromagnetic precipitates ; it is likely that the high - temperature ferromagnetism sometimes found in these materials is due to precipitates .
the latest studies @xcite appear to indicate that true long - range order in @xmath452 emerges only at low temperatures @xmath453 and that weak coupling between remote moments is mediated by holes which are tightly bound to mn acceptors .
further work is necessary to determine whether this picture of magnetism , reminiscent of the polaronic physics discussed in the context of ( ga , mn)p or low - carrier - density ( iii , mn)v systems @xcite , applies to @xmath451 dmss .
the possible presence of dense - moment thermodynamically stable precipitates has also confused studies of oxide semiconductor dms systems .
more consistent evidence of above room temperature ferromagnetism has been reported in co - doped tio@xmath131 although the origin of ferromagnetism in this material is still under debate ( see , _
e.g. _ , @xcite and recent reviews @xcite ) .
mn - doped indium - tin - oxide ( ito ) is another promising candidate for a transparent ferromagnetic semiconductor which could be easily integrated into magneto - optical devices .
particularly encouraging is the observation of a large anomalous hall effect showing that charge transport and magnetism are intimately connected in this oxide dms @xcite .
other interesting related materials are the mn - doped ii - iv - v@xmath131 chalcopyrites surveyed theoretically in a first principles calculation study @xcite .
three of these compounds , cdgep@xmath131 , zngep@xmath131 and znsnas@xmath131 , have shown ferromagnetism experimentally .
the origin of this ferromagnetic behavior has not been explored extensively yet .
finally we mention recent observations of ferromagnetic order up to @xmath454 k in a layered semiconductor sb@xmath131te@xmath2 doped with v or cr @xcite .
these highly anisotropic materials combine dms behavior with strong thermoelectric effects .
the character of the ferromagnetic coupling in these compounds is unclear at present .
this article is a review of theoretical progress that has been achieved in understanding ferromagnetism and related electronic properties in ( iii , mn)v dmss .
the materials we have focused on have randomly located mn(@xmath5 ) local moments which interact via approximately isotropic exchange interactions with itinerant carriers in the semiconductor valence band .
( ga , mn)as is by far the most thoroughly studied and the best understood system in this class .
some ( iii , mn)v materials may exhibit fluctuations in the mn valence between mn(@xmath5 ) and mn(@xmath8 ) configurations or have dominant mn(@xmath8 ) character , possibly ( ga , mn)n for example . magnetic and other properties of materials in the latter class will differ qualitatively from those of ( ga , mn)as and this review makes no attempt to discuss the theory that would describe them . when we refer to ( iii , mn)v ferromagnetism below , it should be understood that any materials that prove to be in the latter class are excluded .
interest in dms ferromagnetism is motivated by the vision that it should be possible to engineer systems that combine many of the technologically useful features of ferromagnetic and semiconducting materials .
this goal has been achieved to an impressive degree in ( iii , mn)v dmss , and further progress can be anticipated in the future .
the goal of high temperature semiconductor ferromagnetism flies in the face of fundamental physical limits , and the fact that so much progress has nevertheless been achieved , is due to a serendipitous combination of attributes of ( iii , mn)v materials .
we have reserved the term ferromagnetic semiconductor for materials in which the coupling between local moments is mediated by carriers in the host semiconductor valence or conduction bands . then magnetic properties can be adjusted over a broad range simply by modifying the carrier system by doping , photo - doping , gating , heterojunction band - structure engineering , or any technique that can be used to alter other semiconductor electronic properties .
most of these tuning knobs have already been established in ( iii , mn)v ferromagnets .
the progress that has been made in achieving ( ga , mn)as ferromagnetism and in understanding its phenomenology has a few lessons .
the analysis of any dms should start with understanding the properties of isolated defects associated with the magnetic element . in the case of ( ga , mn)as materials the desirable magnetic defect is substitutional mn@xmath18 , because mn then both introduces a local moment and acts as an acceptor .
the holes doped in the system by mn@xmath18 impurities provide the glue that couples the moments together . understanding the role of other defects that are present in real material is also crucial .
substitutional mn@xmath19 is particularly important in ( ga , mn)as because it reduces the number of free moments and reduces the density of the hole gas that mediates ferromagnetism .
learning how to remove defects that are detrimental to strong magnetic order is key to creating useful materials .
these two steps have been largely achieved in ( ga , mn)as . there is every reason to believe that if the same progress can be made in other dms materials , some will be even more magnetically robust . we have reviewed in this article a number of theoretical approaches that shed light on what controls key properties of ferromagnetic ( iii , mn)v semiconductors .
first principles electronic structure calculations give a good overview of fundamental material trends across the series and explain many of the structural characteristics of these alloys .
semi - phenomenological microscopic tight - binding models provide a convenient way to use experimental information to improve the quantitative accuracy of the description .
another phenomenological description that successfully models magnetic , magneto - transport , and magneto - optical properties is a single parameter theory that adopts a @xmath431 description for the host semiconductor valence bands , and assumes that the exchange interaction between local moments and band electrons is short - ranged and isotropic .
the single exchange parameter that appears in this theory can be determined by fitting to known properties of an isolated mn local moment , leading to parameter - free predictions for ferromagnetism .
qualitative models which focus on what kind of physics can occur generically for randomly located local moments that are exchange coupled to either localized or itinerant band electrons , also provide useful insights for interpreting experiments . the most important properties of ( iii , mn)v materials are their curie temperature and ferromagnetic moment which reflect both the strength of the coupling between mn local spins and its range . the highest ferromagnetic transition temperatures in ( ga
, mn)as epilayers have so far been achieved with substitutional mn@xmath18 fractions in the neighborhood of 5% by post - growth annealing which eliminates most of the interstitial mn ions .
achieving @xmath3 values close to room temperature in ( ga , mn)as , which is expected to occur for 10% mn@xmath18 doping , appears to be essentially a material growth issue , albeit a very challenging one . in
optimally annealed samples , experiment and theoretical considerations indicate that the mn - mn exchange interactions are sufficiently long range to produce a magnetic state that is nearly collinear and insensitive to micro - realization of the mn@xmath18 spatial distribution .
magnetization and curie temperatures in these systems are well described by mean - field theory .
the magnetic and transport properties of high - quality ( ga , mn)as materials are those of a low - moment - density , low - carrier - density metallic ferromagnet , with a few special twists . because of the strong spin - orbit interactions in the valence band , metallic ( ga , mn)as shows a large anomalous contribution to the hall effect and the source of magnetic and transport anisotropies is the more itinerant electrons , unlike the transition metal case in which anisotropies originate primarily in @xmath4-electron spin - orbit interactions . the small moment densities lead to a large magnetic hardness reflected in a single - domain - like behavior of many ( ga , mn)as thin films . they also explain a large part of the orders of magnitude reduction in the current densities required for transport manipulation of the magnetic state through spin - momentum - transfer effects .
the low - carrier density of the itinerant holes responsible for magnetic coupling means that they are concentrated around a particular portion of the brillouin - zone in the valence band which has a large oscillator strength for optical transitions to the conduction band .
this property opens up opportunities for optical manipulation of the magnetic state that do not exist in transition metal ferromagnets and have not yet been fully explored .
the research which we have reviewed here that is aimed at an understanding of the optical properties of ( ga , mn)as ferromagnets is still incomplete , particularly for the most interesting ideal annealed materials , and will be important in setting the ground work for the exploration of new effects .
these conclusions do not necessarily apply to all ( iii , mn)v ferromagnets .
for example , material trends suggest that wider band - gap hosts would have stronger exchange scattering that would lower the conductivity , shorten the range of mn - mn exchange interactions , and increase the importance of quantum fluctuations in mn and band hole spin orientations .
this could eventually lead to curie temperatures significantly below the mean - field estimates . in the opposite limit , when the exchange interaction is weak enough to be treated perturbatively , sign variations in the rkky mn - mn interaction are expected to lead to frustration and weaken ferromagnetism at large carrier densities .
one of the important miracles of ( ga , mn)as ferromagnetism is that this effect is much weaker than would naively be expected because of the complex valence band structure . in ( ga , mn)as , incipient frustration that limits magnetic stiffness may be responsible for the weak dependence of ferromagnetic transition temperature on carrier density .
this property of ( ga , mn)as suggests that little progress on the @xmath3 front is likely to be gained by non - magnetic acceptor co - doping .
disorder is an inevitable part of the physics of all dms ferromagnets because of the random substitution of elements possessing moments for host semiconductor elements . even in metallic , ideal annealed samples
that have only substitutional mn@xmath18 impurities , randomness in the mn microstructure leads to both coulomb and spin - dependent exchange potential scattering . for ( ga , mn)as
, coulomb scattering dominates over exchange potential scattering , limiting the conductivity to @xmath455 .
frustration and disorder are certainly very important near the onset of ferromagnetism at low mn density , where the network of exchange interactions that lead to long range order is still tenuous .
studies of well characterized materials with low mn fraction near the metal - insulator transition are now possible because of progress in understanding the role of defects and are likely to exhibit complex interplay between disorder , and coulomb and exchange interactions .
we thank all colleagues who have given us the permission to show their results in this review . among the many stimulating discussions we would like to especially acknowledge our interactions with experimentalists at the university of nottingham .
work on the article was supported by by the grant agency of the czech republic through grant no .
202/05/0575 , by the academy of sciences of the czech republic through institutional support no .
av0z10100521 , by the ministry of education of the czech republic center for fundamental research lc510 , by the uk epsrc through grant gr / s81407/01 , by the welch foundation , by the us department of energy under grant de - fg03 - 02er45958 , and by the us office of naval research under grant onr - n000140610122 . | the body of research on ( iii , mn)v diluted magnetic semiconductors initiated during the 1990 s has concentrated on three major fronts : i ) the microscopic origins and fundamental physics of the ferromagnetism that occurs in these systems , ii ) the materials science of growth and defects and iii ) the development of spintronic devices with new functionalities .
this article reviews the current status of the field , concentrating on the first two , more mature research directions . from the fundamental point of view , ( ga , mn)as and several other ( iii , mn)v dmss are now regarded as textbook examples of a rare class of robust ferromagnets with dilute magnetic moments coupled by delocalized charge carriers .
both local moments and itinerant holes are provided by mn , which makes the systems particularly favorable for realizing this unusual ordered state .
advances in growth and post - growth treatment techniques have played a central role in the field , often pushing the limits of dilute mn moment densities and the uniformity and purity of materials far beyond those allowed by equilibrium thermodynamics . in (
iii , mn)v compounds , material quality and magnetic properties are intimately connected . in the review we focus on the theoretical understanding of the origins of ferromagnetism and basic structural , magnetic , magneto - transport , and magneto - optical characteristics of simple ( iii , mn)v epilayers , with the main emphasis on ( ga , mn)as . the conclusions we arrive at are based on an extensive literature covering results of complementary _ ab initio _ and effective hamiltonian computational techniques , and on comparisons between theory and experiment .
the applicability of ferromagnetic semiconductors in microelectronic technologies requires increasing curie temperatures from the current record of 173 k in ( ga , mn)as epilayers to beyond room temperature .
the issue of whether or not this is a realistic expectation for ( iii , mn)v dmss is a central question in the field and motivates many of the analyses presented in this review . |
sparnaay s 1958 experiments [ 1 ] exposed the existence of zero - point radiation , which nernst had considered as a possibility in 1916 .
however , sparnaay was not looking for it , since he was only trying to check casimir s hypothesis about the mutual attraction of two uncharged conductor plates placed very close together [ 2 ] .
this attraction should have disappeared when the temperature approached absolute zero , but sparnaay found that at near that temperature there was still some attraction not accounted for by casimir s hypothesis , being independent of temperature and obeying a very simple law ; it is directly proportional to the surface of the plates , and inversely proportional to the 4@xmath7 power of the distance @xmath8 between them .
sparnaay observed this force when the plates were placed in a very complete vacuum at near zero - absolute temperature . for a distance of @xmath9 cm between the plates
, he was able to measure a force of @xmath10 , and deduced the formula @xmath11 in a near - perfect vacuum and at a temperature very near absolute zero , which implies the absence of any photon gas " , the phenomenon observed by sparnaay could only be produced by a radiation inherent to space .
this could be only the case if its spectrum is relativistically invariant , which could only happen if its spectral distribution is inversely proportional to the cubes of the wavelengths in other words , if the number of photons of wavelength @xmath12 which strike a given area within a given time is inversely proportional to @xmath13 .
a function of spectral distribution which is inversely proportional to the cubes of the wavelengths , implies a distribution of energies which is inversely proportional to the 4@xmath7 power of the wavelengths , because the energies of the photons are inversely proportional to the wavelengths . in 1969 , timothy h. boyer [ 4 ] showed that the spectral density function of zero - point radiation is @xmath14 where @xmath15 is the number giving the measurement of wavelength @xmath12 .
this function produces the next , for the corresponding energies @xmath16 for @xmath17 , @xmath18 .
there must therefore be a threshold for @xmath12 , which will hereafter be designated by the symbol @xmath1 . in the sparnaay effect , the presence of a force which is inversely proportional to the 4@xmath7 power of the distance @xmath8 between the plates , while the distribution of energies is also inversely proportional to the 4@xmath7 power of the wavelengths , leads us to infer that the cause of the apparent attraction must lie in some factor which varies inversely with the wavelength , and which causes a behaviour different from the photons of wavelengths equal to or greater than @xmath8 , since : @xmath19 the factor in question turns out to be the proportion of photons which are reflected , not absorbed ; that is the coefficient of reflection @xmath20 , while the different behaviour is caused by the obstacle which the presence of each one of the two plates offers to the reflection from the inner face of the other one , of photons of wavelength equal or greater than @xmath8 " . as we know , the energy which is transferred by a reflected photon is double that transferred by one which is absorbed . to simplify the following arguments , it is convenient to use the @xmath21 system of measurements in which the basic magnitudes are the quantum of electric charge , the mass of the electron and the speed of light . in this system ,
the units of length and time are respectively @xmath22 and @xmath23 .
the results of sparnaay s experiments provide a link with reality which may be enough to show the intensity of the energy flow belonging to zero - point radiation . a recent paper [ 5 ]
has shown that this flow is one which corresponds to the incidence in an area @xmath24 of one photon of wavelength @xmath1 every @xmath25 , plus one photon of wavelength @xmath26 every @xmath27 , etc . up to one photon of wavelength @xmath28 every @xmath29 , where @xmath30 [ 6 ] .
this radiation implies the energy flow per @xmath24 and @xmath25 which is given by @xmath31 where @xmath32 [ 7 ] .
this energy flow produces a force per @xmath33 which is given by @xmath34 which when expressed in the c.g.s .
system and per @xmath35 is given by @xmath36 in cosmic rays , photons have been observed with energies of the order of more than @xmath37 ev , [ 8 ] which implies that the value of @xmath38 must be greater than @xmath39 .
the value of @xmath40 is therefore inmense . in consequence
the analysis of the possible interactions of zero - point radiation with the elementary particles assumes enormous interest .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if we have a charged sphere , all the electrical forces repel , and the electron will tend to fly apart ... the charge must be mantained over the sphere by something which stops it from flying off .
poincar was the first to point out that this ` something ' must be allowed for in the calculation of energy and momentum " .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if the electron s mass and charge were distributed according to some spatial configuration , the repulsion of the charge against itself would tend to cause the particle to fly apart . seeing this
, poincar suggested that something must exist which can counteract this repulsion and this ` something ' has therefore been named ` poincar s tensions ' .
the possible configuration must be one which has no favoured direction , which implies that it must have spherical symmetry .
electrostatic repulsion would appear in the form of a centrifugal force . to meet this
, poincar s tensions would have to be arranged as centripetal forces able to counteract it .
when submitted to these fields of opposing forces , the charge of the electron would tend to distribute itself equally over a spherical surface .
zero - point radiation , which arrives equally from all directions of space , provides centripetal energy flows towards every imaginable spherical surface , and from these flows there could be derived equally centripetal flows , to fill the role of poincar s tensions .
sparnaay s experiments have allowed us to measure the intensity of the energy flows inherent to zero - point radiation ; compton s experiments discovered the laws which govern the phenomena of dispersal and energy transfer produced by the interaction of photons with free particles and matter .
finally the differential cross section for these phenomena follows the klein - nishima - tann formula . with this knowledge we can attempt to calculate the centripetal force which could be produced by the zero - point radiation , on an electron with its charge distributed equally over a spherical surface of radius @xmath41 . the collision of a photon with a free particle of matter , shown in fig .
1 , produces the compton effect .
the colliding photon of energy @xmath42 and wavelength @xmath12 , loses part of its energy to the particle , and is diverted at an angle @xmath43 to its previous trajectory .
when losing energy , it increases its wavelength by an amount which is given by compton s equation @xmath44 where @xmath45 , where @xmath46 is the said particle mass . for the electron , compton s wavelength @xmath47 . in the case of zero - point radiation , the free particle of matter , which is subjected to equal forces from all directions , does not move , but suffers compression towards its centre .
the encounter of a photon with an electron does not always produce the compton effect ; the differential cross section for this @xmath48 is given by the klein - nishima - tann formula for dispersal with electrons , which is @xmath49 where @xmath43 is the angle of dispersal , @xmath50 the frequency of the arriving photon , @xmath51 that of the dispersed photon , @xmath52 the classical radius of the electron , @xmath48 is the differential cross section for the compton effect , and @xmath53 . by bringing in this relation and remembering that @xmath54
, we can write @xmath55 d\te\eqno{(4)}\ ] ] taking all the foregoing into consideration , we also know that : * the number of photons of zero - point radiation with wavelength @xmath28 , which converge from all directions in an area @xmath24 is @xmath56 every @xmath25 . * the energy transferred to the electron by a photon of wavelength @xmath57 , which after interacting is diverted at an angle @xmath43 from its trajectory , is @xmath58= hc\left[\que1{nq_\la}-\que1{nq_\la+\que{2\pi}\al l_e(1-\cos\te)}\right].\ ] ] therefore , the energy transferred every @xmath25 by these photons of point - zero radiation converging on an area @xmath24 is given by @xmath59 where @xmath60 @xmath61,\ ] ] @xmath62 whence we derive the following definite integrals : @xmath63,\ ] ] @xmath64 @xmath65,\ ] ] @xmath66\ ] ] @xmath67 @xmath68,\ ] ] where @xmath69 .
the addition of this integrals gives @xmath70 \left.\disp+\que1{a(1+a)}-\que1{2a(1+a)^2}- \que{(3 + 2a)}{a^3}l(1+a)\right]\end{array}\eqno{(6)}\ ] ] therefore we have : @xmath71 $ ] ; where @xmath72=\left[\que{l(1+a)}a-\que5{6(1+a)}+\que1{6(1+a)^2}+\cdots\right]\ ] ] as in ( 6 ) .
therefore @xmath73 $ ] ; obviously @xmath74 and @xmath75 .
since in the @xmath21 system @xmath76 , this factor can be ignored and we can write @xmath77,\eqno{(7)}\ ] ] and @xmath78.\eqno{(8)}\ ] ] a first attempt to solve the problem showed us that the energy flow ( 3 ) would be excessive .
sparnaay s measurements allow us to deduce that , in the case of photons with wavelength equal to or greater than @xmath79 cm , the energy flow of the zero - point radiation agrees with the energy flow corresponding to the incidence , in an area @xmath24 , of the photon of wavelength @xmath1 every @xmath25 , plus another of wavelength @xmath26 every @xmath27 , up to a photon of wavelength @xmath28 every @xmath29 .
however this does not mean that , at present , the photon with the shortest wavelength in the zero - point radiation , has to be the photon of wavelength @xmath4 .
if we suppose that its wavelength is @xmath80 , the flow of energy to be taken into account is : @xmath81 instead of @xmath82 as given by ( 3 ) .
the condition @xmath75 implies @xmath83 @xmath84<1\ ] ] it is reasonable to suppose that the wavelength @xmath80 of the most energetic photon in the actual zero - point radiation is much greater than the wavelength of the photon whose energy is @xmath85 , which is @xmath86 .
if this assumption does not cause any future contradiction and leads to a value of @xmath87 , which means that the energy of the photon with wavelength @xmath4 is greater than @xmath37 ev , it must be considered as reasonable .
accordingly @xmath88 , and we can write : @xmath89 besides , we have for any value of @xmath90 : @xmath91 ( 1+a)^{-2}=1 - 2a+3a^2 - 4a^3+\cdots+(-1)^m(m+1)a^m,\\[+4pt ] ( 1+a)^{-3}=1 - 3a+6a^2 - 10a^3+\cdots+(-1)^m\que{(m+1)(m+2)}2a^m . \end{array}\ ] ] by introducing these series into @xmath92 $ ] , as in ( 6 ) , we obtain : @xmath93=\que7{12}a-\que{11}{10}a^2+\que{37}{20}a^3+\cdots+\\ \\[-4pt ] \disp + ( -1)^m \left[\que1{m+1}+\que2{m+2}-\que3{m+3}-1-\que{m(m-1)}6\right]a^m \end{array}\eqno{(10)}\ ] ] to obtain the total energy transferred each @xmath94 to an area @xmath24 , we can write : @xmath95 from ( 8) we have @xmath96=\que{2\pi^2}\al\que{k_\la}n[a]m_ec^2\ ] ] whence @xmath97}{n^4}\right)m_ec^2\ ] ] each @xmath25 .
therefore the energy flow per @xmath24 each @xmath98 is : @xmath99 a^m\right\ } \end{array}\eqno{(11)}\ ] ] to an area @xmath24 . since @xmath69 we can write : @xmath100 \end{array}\!\!\!\eqno{(12)}\ ] ] to an area @xmath24 , whence @xmath101 @xmath102\ ] ] to an area @xmath24 ; where @xmath103\que1{m+3}\ ] ] by introducing @xmath104 we can write : @xmath105 to an area @xmath24 , in such a way that @xmath106_m,\eqno{(14)}\ ] ] where we used @xmath107 fig .
2 shows an electron configured as a spherical surface as a result of the equilibrium between centrifugal forces of electrostatical repulsion and centripetal forces derived from the interaction with zero - point radiation .
if this interaction happens with zero - point radiation falling on an area @xmath24 situated on @xmath108 on the surface of the electron , and coming from all directions of the half space defined by the tangent plan at @xmath108 and opposite to @xmath109 , the total energy flow over the said area will be @xmath110 , and the total energy flow transferred to the electron will be @xmath111 .
this energy flow produces a force @xmath112 .
if we imagine a half sphere whose center is at @xmath108 and whose radius has a measure equal to @xmath113 the components according the direction @xmath114 of the forces @xmath113 coming fron the referred half space have the same measurements as the distances of the points of the surface of the said sphere to the tangent plant at @xmath108 and the number which expresses their sum is that which measures the volume of the said half sphere ; i.e. @xmath115 .
now , in the present case we must divide this number by the area of the said half sphere , because the total force @xmath116 is the sum of those which come from all directions .
to each point of this area there corresponds one of those directions .
therefore the sum of the components according the radius at the point of arrival of the forces which strike in an area @xmath24 is given by @xmath117 which gives @xmath118_m\que{m_ec}{t_e}\eqno{(15)}\ ] ] the electrostatic repulsion for the fraction @xmath119 of the charge @xmath120 which corresponds to a fraction @xmath24 of a sphere with radius @xmath121 is @xmath122 by equalising ( 15 ) and ( 16 ) we obtain @xmath123_m\eqno{(17)}\ ] ]
figure 3 shows two electrons whose centers are @xmath124 apart .
a fraction of the zero - point radiation flow which reaches the electron whose center is at @xmath125 is spent to equalize the poincar s tensions at its surface . according to ( 15 ) this fraction implies the force .
@xmath126_m \que{m_ec}{t_e}.\ ] ] therefore , the zero - point flows which emerge after interacting with this particle imply forces towards @xmath125 whose intensities at the distance @xmath127 from this point would be equal to @xmath128 .
the same occurs with the zero - point radiation which interacts with the electron at @xmath125 after having interacted with the electron at @xmath129 .
both reductions produce an apparent attraction given by @xmath130_m \que{m_ec}{t_e}\eqno{(18)}\ ] ] the expression of @xmath131 in the @xmath21 system is @xmath132 , where @xmath133 .
the attraction between two electrons which are @xmath124 distant is : @xmath134 by equalising ( 18 ) and ( 19 ) we obtain @xmath135_m}{g_e}\eqno{(20)}\ ] ] from ( 17 ) and ( 20 ) we obtain @xmath136^{1/2},\ ] ] where @xmath137 is the length of the radius of the electron and @xmath138 the number of @xmath1 by one @xmath139 .
the quantum character of @xmath140 requires that if @xmath141 , @xmath142 , where @xmath108 is a digit .
for @xmath143 we have @xmath144 the planck length @xmath145 , expressed in the @xmath21 system is @xmath146 .
therefore @xmath147 .
@xmath6 is bigger than @xmath1 , on the same way that @xmath148 is bigger than @xmath149 . notwithstanding
the real wavelength of the photon of maximum energy is @xmath1 and not @xmath6 .
the quantum of the magnitude @xmath150 is bigger than the easy combination @xmath149 .
the quantum @xmath1 can be expressed as @xmath151 , which is minor than the easy combination @xmath152 .
the minimum value of @xmath41 is 1 , and for it : @xmath5 and @xmath153 @xmath154 @xmath155 the wavelengths of the most energetic photons observed in cosmic rays are more than @xmath156 .
the relation between the force of gravitation on the electrostatic force is : @xmath157
[ 1 ] m. j. sparnaay , physica * 24 * , 51 ( 1958 ) . | knowing the magnitude of the energy flow inherent to zero - point radiation allows us to approach the question of its possible interaction with particles of matter .
its photons are not different from the rest , and must in principle be subject to the compton effect and the klein - nishima - tann formula for its cross section . on this assumption , it is shown here that zero - point radiation may be powerful enough to explain poincar s tensions and to supply an efficient cause for gravitation
. this could be only the case if the classic radius of the electron measures @xmath0 , where @xmath1 is the minimum wavelength for electromagnetic radiation , and if the wavelength of the most energetic photon in the actual zero - point radiation is @xmath2 .
to the first of these numbers there corresponds the energy @xmath3 mev for the photon whose wavelength is @xmath4 .
this gives also the relation @xmath5 , where @xmath6 is the planck length .
finally the relation between the force of gravity and the electrostatic force is explained by the equations obtained in this paper .
= = = = = = = = = |
images collected with ground based telescopes show temporal variation in image quality , primarily due to fluctuations in the index of refraction of the air along the light path .
these index of refraction fluctuations are caused by differences in air temperature arising from turbulent mixing .
these turbulent motions happen on smaller scales than the gross motions of the winds .
still it is known that the wind has a part in the strength of these fluctuations .
hufnagel @xcite derived a model for the vertical profile of strength of turbulence @xmath7 which depended on a wind speed profile centered at a particular altitude .
furthermore , it has been observed that the index of refraction fluctuations are enhanced where there are shears in winds @xcite , so that direction as well as speed can be a factor , although this is difficult to quantify @xcite . aside from the effects of wind on turbulence strength ,
the movement of turbulent regions past an observer induces temporal variations in seeing .
if wind speeds are sufficiently high that turbulence does not evolve significantly over a measurement period , we apply taylor s hypothesis and treat the temperature fluctuations as quantities that are advected by the wind @xcite . the following wind model is often used in the modeling of adaptive optics and other imaging systems ( _ e.g. _ @xcite ) , @xmath8 ^ 2\right),\ ] ] where @xmath9 is the wind speed in m / s , @xmath10 is the height in m , and @xmath11 is the angle from the zenith .
bufton @xcite is often cited as the source for this model ( _ e.g. _ @xcite ) , but that paper does not present an explicit wind model .
the basic form of the model was presented by greenwood @xcite .
readers often over look the statement in the body of the original paper that the height @xmath12 corresponds to a mean sea level altitude of 3048 m. this is the height of the maui space surveillance system on haleakala , where the analysis was being used .
so the height of the tropopause term , 9400 m , should really be 12448 m when the model is used for sites other than haleakala .
this caveat has been overlooked in a number of references , leading to misleading results . the model from greenwood @xcite
was derived by averaging radiosonde data collected from balloons launched from lihue on the island of kauai from 1950 - 1970 and from hilo on the island of hawaii from 1950 - 1974 .
this was more fully detailed in a technical report , which did not explicitly generalize the equation , but the authors clearly understood what the various terms signified @xcite .
the wind model is generalized as , @xmath13 ^ 2\right)\ ] ] where @xmath14 is the wind speed , @xmath10 is the altitude , @xmath15 is the wind speed at the ground or low altitude , @xmath16 is the wind speed at tropopause , @xmath17 is the altitude of the tropopause , @xmath18 is the thickness of the tropopause layer and @xmath19 is the zenith angle of the observation @xcite .
there are several uses for a wind profile .
it can be used in conjunction with a turbulence model to predict the performance of instrumentation .
most commonly this is applied to the servo bandwidth of closed loop systems @xcite .
the minimum bandwidth of the system ( to limit servo lag ) is related to the greenwood frequency . because the greenwood frequency depends on a product of @xmath4 and a power of the wind speed @xmath20 , it is important to get the wind profile and turbulence profiles as a function of altitude .
a wind profile can also be used to simulate turbulence in conjunction with phase screens .
currently , there is an increased emphasis on modeling effects of turbulence on telescopes with diameters of 20 - 50 meters .
significantly more spatial variation of wavefront optical path difference is captured with such large telescopes in a short exposure than on a smaller telescope .
some of that variation is due to the presence of multiple layers of turbulence , moving at different speeds and in different directions .
the patterns may not simply translate across the telescope aperture as implied by the taylor hypothesis , but are more likely toboil " with a complicated wind profile .
this can have an impact on the design and performance of an adaptive optics system . in order to
better model atmospheric turbulence , we used archival radiosonde data to compute the wind profiles for a number of astronomical observatories . we also show that the wind profile has a strong seasonal variation .
we also extend the wind profiles , by including wind direction .
unfortunately , the wind directions do not always lend themselves to an analytic expression , so we only show them in graphs . using the same radiosonde data
, we examined the variation of the gradient richardson number @xcite ) , which is an indicator of the stability of parcels of air .
since it depends upon strength of wind shear , as opposed to wind speed , it also serves to inform the modeler that wind speed alone is not sufficient .
as will be seen in our plots , the wind speed can be modeled with a few simple terms , but the form of the richardson number plots indicates the true nature of atmospheric turbulence : layered , with sharp boundaries ( although these are muted by the resolution of our data ) .
worldwide there are over 900 sites that launch radiosondes on a routine basis .
radiosondes are small instrument packages mounted on weather balloons .
they record atmospheric pressure , temperature , potential temperature , mixing ratio , dew point , and relative humidity as a function of height .
data are taken at about 6 second intervals , but databases may only have the data recorded at 60 second intervals .
readings at certain specific standard barometric pressures are made regardless of when they occur .
since the balloons rise at about 5 m / s , the best vertical resolution would be approximately 30 meters , while the usual recorded resolution would be 300 meters and frequently the recorded data are at irregular intervals .
wind speed and wind direction are obtained from either radar observation or a navigation system such as long range navigation ( loran ) and newer systems may use the global positioning system .
almost all of these sites launch balloons at 0 ut and 12 ut .
the information is used as entries to weather prediction and simulation programs .
we downloaded atmospheric sounding data for 1973 january till 2006 september for a number of stations from the university of wyoming s weather web ( http://weather.uwyo.edu/upperair/sounding.html ) .
the stations were chosen based upon their proximity to an astronomical observatory .
not all the sites had data for the entire time period ; the dates for the available data are also listed in the table . in addition
there are gaps of varying length in each data set , which are most likely due to equipment problems .
these data drop outs are not a significant problem due to the large number of data points that are available .
the names of stations and their details along with the nearby observatories are listed in table [ station_table ] .
the name of the station is given , along with its latitude , longitude and altitude , the dates when data were recorded and the names of nearby observatories . in parenthesis
are the distance in km between the observatory and the radiosonde launch site and the altitude of the observatory in km . *
* .[station_table]details of the radiosonde launch sites used [ cols="<,^,^,^,^,<,^,^ " , ] the results are shown in figs .
[ hilo_wind][anto_wind ] respectively .
each figure is composed of 12 graphs corresponding to the 12 months of the year .
the first letter of the corresponding month is in the upper right corner .
each graph has the the wind speed plotted with a solid line and the fitted wind speed computed from the coefficients in tables [ hilo_coeffs][anto_coeffs ] is plotted as a dashed line .
the wind direction is plotted as a solid red line .
the wind speed is marked along the bottom of the graph , while the wind direction is marked on the top of the graph .
the wind directions often show sharp spikes at the edges of jets .
those are an artificial result of the averaging of directions and can be ignored . the wind speed and direction for hilo , hawaii .
each of the graphs in this figure has the the measured wind speed as a function of height plotted with a solid black line and the wind speed computed from the model coefficients in table [ hilo_coeffs ] is shown as a black dashed line .
the wind direction is plotted as a solid red line .
the wind speed is marked along the bottom of the graph , and the wind direction is marked on the top of the graph .
the first letter of the corresponding month is in the upper right corner .
, height=283 ] the wind speed and direction for oakland , california .
the layout of the figure is the same as in fig .
[ hilo_wind ] , except the fitted wind model comes from table [ oakland_coeffs ] .
, height=283 ] the wind speed and direction for san diego , california .
the layout of the figure is the same as in fig .
[ hilo_wind ] , except the fitted wind model comes from table [ sandiego_coeffs].,height=283 ] the wind speed and direction for tucson , arizona .
the layout of the figure is the same as in fig .
[ hilo_wind ] , except the fitted wind model comes from table [ tucson_coeffs ] . ,
height=283 ] the wind speed and direction for flagstaff , arizona .
the layout of the figure is the same as in fig .
[ hilo_wind ] , except the fitted wind model comes from table [ flagstaff_coeffs ] .
, height=283 ] the wind speed and direction for tenerife , canary islands , spain , which is useful for the observatories located in the canary islands .
the layout of the figure is the same as in fig .
[ hilo_wind ] , except the fitted wind model comes from table [ tenerife_coeffs ] .
, height=283 ] the wind speed and direction for antofagasta , chile .
the layout of the figure is the same as in fig .
[ hilo_wind ] , except the fitted wind model comes from table [ anto_coeffs ] . ,
height=283 ] for both hemispheres , the tropospheric winds increase in the winter months and decrease in the summer months .
stratospheric winds increase in the summer months , the opposite of the tropospheric winds . during the summer months
, tenerife has a low level ( @xmath21 km ) jet form .
this is most obvious in july and august , but first appears in june and persists to october . in the greenwood model , @xmath22 is the low altitude wind speed and that is true for our model for hawaii .
it is not always true for other sites , as these values are sometimes negative .
the reason for this is that the wind profiles for those sites has a slightly different shape than the hawaiian wind profile .
this can be seen in figs .
[ oakland_wind ] and[flagstaff_wind ] .
greenwood @xcite only provides a model for the wind speed , though @xcite does show the annual mean wind direction for hilo and lihue . as shown by figs .
[ hilo_wind]-[anto_wind ] the wind direction is not an easily modeled function .
it can most easily be modeled as a square wave . since stability of flows
does depend upon the amount of shear present , an analysis that computed the wind shear at the upper and lower boundaries would tend to overestimate the instability . looking at figs .
[ hilo_wind][anto_wind ] , the wind direction does not change instantaneously , but it does change very quickly , often changing direction by @xmath23 in a km . some regions of slowly changing direction , most commonly above 25 km , are not very well modeled by a square wave . a square wave model will suffice for modeling where phase screens at only a few altitudes are used to model the distribution of turbulence .
note that the reversal of wind direction at altitudes above 20 km is a well known feature of the zonal wind flow .
it is due to a higher altitude jet whose center is north of the equator in winter , and south of the equator in summer .
there are also northern and southern jets at higher altitudes which move in the same direction as the jets in the lower stratosphere .
these are seasonal and impact the wind speed and directions in our data @xcite .
the gradient richardson number is a dimensionless ratio , @xmath24 , related to the buoyant production or consumption of turbulence divided by the shear production of turbulence @xcite .
it is defined as [ eq_richardson ] r_i = , where @xmath25 is the acceleration of gravity , @xmath26 is the potential temperature , @xmath10 is the altitude , @xmath27 and @xmath14 are the horizontal components of the wind vector .
normally a gradient richardson number with a value less than 0.25 is considered to be turbulent and a value greater than 0.25 to be non - turbulent .
it is used to indicate dynamic stability and the formation of turbulence . in order to see at which altitudes turbulence forms , we used data from the radiosondes to compute the monthly binned richardson number for each site in table [ station_table ] using eq .
[ eq_richardson ] at each of the fixed grid points used in [ data_analysis ] .
the grid heights points are every 400 m and some of the finer atmospheric layers may not be captured with our analysis .
the derivatives were computed using a three - point lagrangian interpolation routine .
the computed richardson numbers were then binned into two bins : @xmath28 , and @xmath29 . for a given site , at each altitude the numbers in each bin were normalized by the total number of data points at that altitude .
this gives the richardson number as a function of altitude .
these are shown in figs .
[ hilo_rich]-[anto_rich ] .
as expected , the richardson numbers above 20 km are very high and have little impact on astronomical seeing ; as a result we only plotted the values below 20 km .
it is important to note that these values are for radiosondes launched from sites at lower altitudes than astronomical observatories .
the highest richardson number values occur at these low altitudes and are not a concern to most astronomical observatories . the binned richardson number for hilo , hawaii the figure shows the fraction of richardson numbers falling into two bins .
yellow signifies turbulent conditions ( @xmath28 ) and blue signifies non - turbulent conditions ( @xmath30).,height=283 ] the binned richardson number for oakland , california .
yellow signifies turbulent conditions ( @xmath28 ) and blue signifies non - turbulent conditions ( @xmath30).,height=283 ] the binned richardson number for san diego , california .
yellow signifies turbulent conditions ( @xmath28 ) and blue signifies non - turbulent conditions ( @xmath30 ) .
, height=283 ] the binned richardson number for tucson , arizona .
yellow signifies turbulent conditions ( @xmath28 ) and blue signifies non - turbulent conditions ( @xmath30 ) .
, height=283 ] the binned richardson number for flagstaff , arizona .
yellow signifies turbulent conditions ( @xmath28 ) and blue signifies non - turbulent conditions ( @xmath30 ) .
, height=283 ] the binned richardson number for tenerife , canary islands , spain .
yellow signifies turbulent conditions ( @xmath28 ) and blue signifies non - turbulent conditions ( @xmath30 ) .
, height=283 ] the binned richardson number for antofagasta , chile .
yellow signifies turbulent conditions ( @xmath28 ) and blue signifies non - turbulent conditions ( @xmath30).,height=283 ] examining the individual months reveals that the figures are not as smooth as the wind speed figures , but instead have many fine layers .
the layers are significant , even if they do not seem to explicitly show turbulence ( ie their value of @xmath24 is > 0.25 ) .
because the @xmath24 calculation has a height increment of 400 m , and because the precise positions and altitudes of jets vary over periods of a day or less , the effective values of the velocity derivatives in equation [ eq_richardson ] are lower in our results than they might be at any given time .
this averaging leads to relatively large calculated @xmath24 values .
it also means that our data will not show many instances of @xmath28 .
but where is does occur it is significant .
moreover , any drop in the plotted @xmath24 number indicates a persistent region of instability , and it is likely that some turbulence is present there for extended periods , since low gradient richardson numbers indicate the presence of atmospheric turbulence , they can be an indicator of seeing . in fig .
[ hilo_rich ] , the winter months for hawaii have the highest proportion of richardson numbers lower than 0.25 ( yellow values ) , and the summer months have the lowest . in other words
the winter months have more turbulent conditions than the summer months .
studies of seeing from mauna kea show that this seasonal behavior of seeing correlates with this very well @xcite .
there is relatively little seasonal variation in for antofagasta , chile ( fig .
[ anto_rich ] ) and this correlates well with seeing measurements carried at nearby cerro armazones @xcite which show no clear evidence for any seasonal dependency .
the trend in fig .
[ tenerife_rich ] agrees with seeing measurements for la palma , canary islands , showing that seeing is best may - september @xcite .
flagstaff , arizona ( fig . [ flagstaff_rich ] has more turbulent values than oakland , california , which is agreement with site comparison done between anderson mesa ( outside of flagstaff ) and lick observatory ( near oakland ) @xcite .
this shows that richardson numbers computed from radiosonde data can be used to differentiate between astronomical sites at a macro level which can be of use for narrowing site selection for new telescopes .
it is important to remember that the richardson numbers computed from radiosonde data do not include local effects such as the turbulent surface boundary layer that can have a profound impact on seeing .
small differences in altitudes and location can alter the observed seeing significantly @xcite . while the richardson number does not indicate the strength of turbulence , knowing that there are persistent turbulent layers may be useful to site planners .
for example , knowing roughly where and when one might expect upper layers of turbulence to be present could be useful for measurements , such as laser communications , which are strongly influenced by scintillation .
on the other hand , we have seen that some layers closer to the ground have @xmath24<0.25 .
these must also be persistent to show up over the averaging .
knowing that these layers exist is useful for large telescopes that plan to compensate for near ground effects .
there have been a few other studies of wind speed for different sites , though they have been done for only a limited number of sites and with different methodology .
sarazin & tokovinin @xcite suggested that the wind speed at the 200 mbar pressure isobar ( ) is a good indication of atmospheric turbulence at a given site ( is well monitored for meteorology ) .
the advantage of this is that there are models of for the entire planet , while there are only a fixed number of radiosondes launched each day .
several groups have used this idea to characterize and compare several astronomical sites @xcite including several sites which are included in this study . during the year
, the altitude at which occurs varies , this makes it difficult to use the values in eq .
[ greenwood_fit ] .
it is useful for site comparisons , but it supplements rather than replaces the knowledge of the vertical wind profile .
garca - lorenzo et al .
@xcite studied for paranal and la silla in chile , san pedro mrtir in mexico , mauna kea in hawaii and la palma in the canary islands .
the last two observatories are also characterized in this paper .
we show a seasonal variation in @xmath31 in tables [ hilo_coeffs ] and [ tenerife_coeffs ] which has the highest tropospheric wind speed in the spring and the lowest tropospheric winds in the summer .
garca - lorenzo et al .
showed that the wind speeds at mauna kea and la palma have this same seasonal behavior .
we have presented upper - level wind models computed from archival radiosonde data suitable for the major astronomical observatories in the united states , chile and the canary islands .
we find that the commonly used greenwood model is not suitable for sites other than hawaii .
in addition it produced a single wind profile for the entire year , when it is more correct to use different wind models over the course of the year .
we find that , as may be expected from geophysical fluid dynamics , that sites with similar latitudes share similar wind profiles .
in addition we have computed richardson number profiles from the same datasets .
those results indicate the presence of turbulence at different altitudes with seasonal variations that seem to agree with variations in seeing . these models and results will be of use to modelers of atmospheric turbulence and instrument developers .
this research was funded by the air force office of scientific research and by the air force research laboratory s directed energy directorate .
a portion of the research in this paper was carried out at the jet propulsion laboratory , california institute of technology , under a contract with the national aeronautics and space administration .
radiosonde data were obtained from the wyoming weather web , maintained by the department of atmospheric science of the university of wyoming . | an understanding of wind speed and direction as a function of height are critical to the proper modeling of atmospheric turbulence
. we have used radiosonde data from launch sites near significant astronomical observatories and created mean profiles of wind speed and direction and have also computed richardson number profiles . using data from the last 30 years ,
we extend the 1977 greenwood wind profile to include parameters that show seasonal variations and differences in location .
the added information from our models is useful for the design of adaptive optics systems and other imaging systems .
our analysis of the richardson number suggests that persistent turbulent layers may be inferred when low values are present in our long term averaged data .
knowledge of the presence of these layers may help with planning for adaptive optics and laser communications .
= @xmath0 @xmath1 @xmath2 @xmath3 _*[email protected]_ accepted to optics express r.e .
hufnagel,``variations of atmospheric turbulence , '' in _ topical meeting on optical propagation through turbulence _ , ( optical society of america , 1974 ) , paper wa1 - 1 .
r.r .
beland , `` propagation through atmospheric optical turbulence , '' in _ the infrared & electro - optical systems handbook _ , vol .
2 , f.g .
smith , ed .
( spie , bellingham , 1993 ) , 158 - 232 . e.m .
dewan , r.e . ,
good , r. beland , & j. brown , `` a model for @xmath4 ( optical turbulence ) profiles using radiosonde data , '' pl - tr-93 - 2043 , ( phillips laboratory , albuquerque 1993 ) . g.i .
taylor , `` the spectrum of turbulence , '' proc .
r. soc .
london , ser .
a * 164 * , 476490 ( 1938 ) .
j.l .
bufton , `` comparison of vertical profile turbulence structure with stellar observations , '' * 12 * , 17851793 ( 1973 ) .
r.r .
parenti , & r.j .
sasiela , `` laser guide - star systems for astronomical applications , '' * 11 * , 288309 ( 1994 ) .
l.c .
andrews , & r.l .
phillips , _ laser beam propagation through random media _
( spie , 1998 ) .
r.k .
tyson , _ principles of adaptive optics , _ ( academic , san diego , 1998 ) .
greenwood , d.p .
`` bandwidth specification for adaptive optics systems , '' 1977 , * 67 , * 390 . d.p .
greenwood , & d.l .
fried , `` power spectra requirements for wavefront - compensative systems , '' radc - tr-75 - 227 , ( rome air development center , rome , 1975 ) .
j.w .
hardy , _ adaptive optics for astronomical telescopes _ , ( oxford univ .
, new york , 1998 ) .
p.k .
kundu , _ fluid mechanics _ , ( academic , san diego , 1990 ) . r.a .
fesen , `` a high - altitude station - keeping astronomical platform , '' * 6267 * , 62670 t , ( 2006 ) .
r.e .
newell , j.w .
kidson , d.g .
vincent , & g.j .
boer , _ the general circulation of the tropical atmosphere and interactions with extratropical latitudes , _ vol 1 , ( mit , cambridge , 1972 ) .
bely , `` weather and seeing on mauna kea , '' pub .
astro .
soc .
pacific * 99 * , 560570 ( 1987 ) .
m.s .
seigar , a.j .
adamson , n.p .
rees , t.g .
hawarden , m.j .
currie , & t.c .
chuter , `` seeing statistics at the upgraded 3.8 m uk infrared telescope '' , * 4844 * , 366 , ( 2002 ) .
w. skidmore , s. els , t. travouillon , r. riddle , m. schck , e. bustos , j. seguel , & d. walker , `` thirty meter telescope site testing v : seeing and isoplanatic angle , '' pub .
astro .
soc .
pacific * 121 * , 1151 - 1166 , ( 2009 ) . c. muoz - tun , a.m. varela , & t. mahoney , `` homogeneity of image quality at the roque de los muchachos observatory , '' new astronomy reviews * 42 * , 409416 , ( 1998 ) .
r.w .
wilson , n. omahony , c. packham , & m. azzaro , `` the seeing at the william herschel telescope , '' 1999 , mon . not .
royal astro .
soc . * 309 * , 379387 ( 1999 ) .
d.l .
walters , & l.w .
bradford , `` measurements of @xmath5 and @xmath6 : two decades and 18 sites , '' * 36 * , 78767886 ( 1997 ) .
m. sarazin , & a. tokovinin , `` the statistics of isoplanatic angle and adaptive optics time constant derived from dimm data , '' , in _ beyond conventional adaptive optics : a conference devoted to the development of adaptive optics for extremely large telescopes_. e. vernet , r. ragazzoni , s. esposito , & n. hubin , eds ( european southern observatory , garching , 2002 ) , 321328 .
s. chueca , b. garca - lorenzo , e.g. mendizbal , t. varela , j.j .
fuensalida , & c. muonz - tun , `` input parameters of the hv model above canarian observatories . ''
* 5237 * , 159166 ( 2004 ) .
e. carrasco , r. avila , & a. carramiana , `` high - altitude wind velocity at sierra negra and san pedro mrtir , '' pub .
astro . soc .
pacific * 117 * , 104110 ( 2005 ) .
b. garca - lorenzo , j.j .
fuensalida , c. muonz - tun , & e. mendizabal , `` astronomical site ranking based on tropospheric wind statistics , '' mon . not .
royal astro .
soc . * 356 * , 849858 ( 2005 ) . |
one of the distinctive features of the phase diagram is the critical point at which the first - order transition between hadron matter and qgp ends .
we currently do not have a systematic way of locating this point from first principles as model and lattice calculations face many challenges and much work still needs to be done in order to overcome them .
( for review , see e.g. , @xcite . ) in the meantime , if the critical point is located in a region accessible to heavy - ion collision experiments , it can be discovered experimentally in current and future experiments at rhic , bnl and sps .
it is therefore important to define and select experimental observables that will allow us to locate the critical point , if it is located in an experimentally accessible region .
upon scanning in the center of mass energy @xmath0 and thus in @xmath1 , one should then be able to locate ( or rule out the presence of ) the critical point by using observables that are sensitive to the proximity of the freeze - out point to the critical point @xcite , such as pion and proton multiplicity fluctuations .
as we vary @xmath0 , if the freeze - out point approaches the critical point , we would see an increase and then a decrease , as we move away from it , in the fluctuations in the number of these particles . in this paper
we describe how to use this non - monotonic behavior in the fluctuations of particle numbers near the critical point as a probe to determine its location .
the critical mode , @xmath2 , is the mode which develops large long wavelength fluctuations at the critical point .
its screening mass is related to the correlation length @xmath3 by @xmath4 . in the case of an infinite system , @xmath3 diverges at the critical point . in reality
, @xmath3 reaches a maximum value at the critical point but does not diverge because as it cools the system spends only a finite time in the vicinity of the critical point .
estimates of the rate of growth of @xmath3 as the collision cools past the critical point suggest that the maximal value of @xmath3 that can be reached is around @xmath5 fm @xcite , compared to the natural @xmath6 fm away from the critical point . near the critical point , the @xmath7 and @xmath8 interaction couplings are given by @xmath9 where the dimensionless couplings @xmath10 and @xmath11 are universal and they have been determined for the ising universality class @xcite . throughout this paper we shall use @xmath12 and @xmath13 as benchmark values , because these are the midpoints of the ranges of values known for these constants @xcite . the long wavelength fluctuations in the @xmath2-field manifest themselves in observable quantities in so far as they affect the fluctuations of the occupation numbers of particles that couple to the @xmath2-field , like protons and pions , which interact with @xmath2 through the effective lagrangian : @xmath14 . throughout this paper
we will use @xmath15 mev ( see ref .
@xcite ) and @xmath16 ( see , e.g. , @xcite ) as benchmark values .
it is important to bear in mind that both these parameters and @xmath17 and @xmath18 are all uncertain at the factor of 2 level .
the contribution of critical fluctuations to the 2- , 3- and 4- particle correlators due to @xmath2-exchanges is given by ( see refs .
@xcite , @xcite and @xcite ) @xmath19 @xmath20 equations ( [ eq:2corr ] ) - ( [ eq:4corr ] ) apply to both protons ( with @xmath21 ) and pions ( with @xmath22 ) .
the degeneracy factor @xmath23 is 2 for both protons and pions since the coupling to the @xmath2-field is both spin and charge `` blind '' .
the variance of the fluctuating occupation number distribution is denoted by @xmath24 and @xmath25 is the relativistic gamma - factor of the particle with a given momentum @xmath26 . in order to obtain the critical contribution to the cumulants of the particle multiplicity distributions using the correlators given above
, each momentum index needs to be integrated over using @xmath28 a cumulant with @xmath29 protons and @xmath30 pions will be denoted by @xmath31 . in order to remove the @xmath32 dependence of the cumulants we choose to normalize them as @xmath33 we can see from expressions ( [ eq:2corr ] ) - ( [ eq:4corr ] ) that higher cumulants are proportional to higher powers of @xmath3 and
thus increase by a larger factor near the critical point where @xmath3 becomes large .
in the case of free particles in the classical boltzmann regime , with no critical fluctuations , the fluctuations of any particle number obey poisson statistics .
the poisson contribution to @xmath34 and @xmath35 is 1 , and zero for the mixed cumulants @xmath36 . in reality , this 1 of poisson statistics gets few percent contributions from bose - einstein statistics , from system size fluctuations , from initial state correlations that are incompletely washed out , and from interactions other than those with the fluctuations that are enhanced near the critical point .
we are ignoring all of these non - critical corrections to the poissonian 1 .
existing data on the ratio of the fourth to second net proton cumulants , @xmath37 , at @xmath38 , 62.4 and 200 gev @xcite confirm that the non - critical corrections to the poissonian 1 are indeed small .
we now illustrate _
possible _ experimental outcomes of measurements of the cumulants , assuming that the matter produced at the freezeout point of the fireball evolution for some collision energy @xmath0 is near the critical point . to start ,
let us assume that the critical point occurs at @xmath39 mev and that @xmath40 fm .
the simplest ansatz for @xmath41 that we have found that incorporates the above physics and has roughly the correct scaling behavior from universality of critical phenomena is @xmath42^{1/3}}\ , \ ] ] with @xmath43 chosen in order to reflect the asymmetry in @xmath41 when approaching the critical point from the left ( smaller @xmath1 ) or from the right .
we shall define the width @xmath44 as the width in @xmath1 within which @xmath45 fm .
we also use our benchmark values of the four nonuniversal parameters that determine the @xmath36 for a given @xmath3 , namely @xmath46 , @xmath16 , @xmath47 and @xmath48 and we have allowed for the fact that the chemical freeze - out temperature @xmath49 decreases somewhat with increasing @xmath1 by using an empirical parametrization of heavy ion collision data from ref .
@xcite .
numerical results for some examples of normalized cumulants are shown in fig .
[ fig : om4pa ] .
we can clearly see the peak in all the normalized cumulants near the critical point .
the results indicate that the more protons are involved in the observation measure , the easier it is to identify the critical contribution .
it is readily apparent that the measurement of these observables in heavy ion collisions at a series of collision energies is very well suited to ruling out ( or discovering ) the presence of the qcd critical point .
-dependence of selected normalized cumulants .
we only include the poisson and critical contributions to the cumulants . in the left panel
we show @xmath50 and vary @xmath44 . in the right panel
we show other examples of @xmath51 s with @xmath52 mev .
see text for details .
, title="fig : " ] -dependence of selected normalized cumulants .
we only include the poisson and critical contributions to the cumulants . in the left panel
we show @xmath50 and vary @xmath44 . in the right panel
we show other examples of @xmath51 s with @xmath52 mev .
see text for details .
, title="fig : " ] in order to locate the critical point , experimental results on multiplicity cumulants will need to be compared to the theoretical predictions of the critical contribution to these cumulants .
but , recall that we had to choose benchmark values for four parameters : @xmath53 , @xmath54 , @xmath10 , @xmath11 and also use an ansatz for @xmath41 .
these parameters are not known reliably or accurately enough to permit a quantitative prediction for the effect of the critical point on any one of the cumulants we have described . in this section ,
we suppose that at some @xmath0 there is experimental data showing several of the cumulants significantly exceeding their poisson values .
we ask how ratios of cumulants can be used to extract information on @xmath3 and the values of the four parameters . and
, we construct ratios of cumulants that are independent of @xmath3 and all the parameters , allowing us to make robust predictions for the contribution of critical fluctuations to these ratios . in table
[ table : ratios ] we present the parameter dependence of various cumulant ratios ( with poisson subtracted before taking the ratios ) , where we defined @xmath55 .
@xmath58 & - & @xmath59 & -&-&@xmath60 + @xmath61 & @xmath60&2 & - & - & - + @xmath62 & @xmath63 & -&1 & - & - + @xmath64 & - & -&@xmath60&1 & - + @xmath65 & - & - & - & - & - + @xmath66 & - & - & - & - & - + @xmath67 & - & - & - & - & - + @xmath68 & - & - & - & - & - + @xmath69 & - & - & - & - & - + the correlation length @xmath3 and the four nonuniversal parameters always appear in certain combinations in the multiplicity cumulants and it turns out that we can only constrain four independent combinations .
the first four rows of table [ table : ratios ] show ratios that can be used to constrain one example of four such combinations .
the last five entries in the table show parameter - independent ratios .
they have no @xmath3-dependence , no dependence on the four poorly known parameters , and no @xmath1-dependence .
we find that these five ratios are all precisely 1 .
now let us see how we can use these five ratios in order to locate the critical point .
suppose that as you change the center of mass energy @xmath70 of the collisions , experimental evidence for peaks shown in fig .
[ fig : om4pa ] begin to emerge .
how do you check in a parameter - independent fashion whether the behavior seen in experimental data is consistent with the hypothesis that it is due to critical fluctuations ?
you first subtract the poisson contributions , and then construct the last five ratios in table i. if the fluctuations seen in this hypothetical data are in fact due to the proximity of the critical point , all five of these ratios will be equal to 1 , with no theoretical uncertainties arising from uncertainty in the values of the parameters
. this would be strong evidence indeed for the discovery of the qcd critical point .
m. a. stephanov , k. rajagopal and e. v. shuryak , phys .
lett . * 81 * , 4816 ( 1998 ) [ arxiv : hep - ph/9806219 ] .
m. a. stephanov , k. rajagopal and e. v. shuryak , phys .
d * 60 * , 114028 ( 1999 ) [ arxiv : hep - ph/9903292 ] .
b. berdnikov and k. rajagopal , phys .
d * 61 * , 105017 ( 2000 ) [ arxiv : hep - ph/9912274 ] .
d. t. son and m. a. stephanov , phys .
d * 70 * , 056001 ( 2004 ) [ arxiv : hep - ph/0401052 ] . c. nonaka and m. asakawa , phys .
c * 71 * , 044904 ( 2005 ) [ arxiv : nucl - th/0410078 ] . | the qcd critical point can be found in heavy ion collision experiments via the non - monotonic behavior of many fluctuation observables as a function of the collision energy .
the event - by - event fluctuations of various particle multiplicities are enhanced in those collisions that freeze out near the critical point .
higher , non - gaussian , moments of the event - by - event distributions of such observables are particularly sensitive to critical fluctuations , since their magnitude depends on the critical correlation length to a high power .
we present quantitative estimates of the contribution of critical fluctuations to the third and fourth moments of the pion and proton , as well as estimates of various measures of pion - proton correlations , all as a function of the same five non - universal parameters .
we show how to use nontrivial but parameter independent ratios among these more than a dozen fluctuation observables to discover the critical point .
we also construct ratios that , if the critical point is found , can be used to overconstrain the values of the non - universal parameters . |
let @xmath0 be a complex manifold and @xmath1 be a holomorphic line bundle .
in complex geometry ( @xcite ) , @xmath2 is * positive * if it admits a metric @xmath3 such that the curvature @xmath4 ( with respect to the natural unique hermitian connection @xmath5 ) is a kahler form .
when @xmath6 is a projective variety , positivity of @xmath2 is equivalent to * ampleness*. a line bundle @xmath2 is ample if some positive tensorial power @xmath7 is very ample , i.e. @xmath7 is generated by its holomorphic sections . let @xmath8 be a holomorphic section of @xmath2 .
the zero set @xmath9 of @xmath10 is a divisor of complex codimension 1 in @xmath6 .
the complement @xmath11 inherits a * weinstein structure * @xmath12 from @xmath13 ( definition [ wmanifold ] ) : @xmath14 the weinstein structure @xmath15 satisfies @xmath16 notice that the 1-form @xmath17 blows up at @xmath18 although its exterior derivative @xmath19 extends to @xmath6 .
in fact , since @xmath20 is integrable , this structure has a classical name in complex geometry world : stein structure .
it is important to stress that the weinstein structure is an object in the symplectic world because the integrable complex structure is not part of the data .
the complement @xmath21 topologically depends on @xmath10 . for those sections
@xmath10 whose zero sets @xmath18 are smooth divisors , @xmath21 are all diffeomorphic .
in fact , the weinstein structures @xmath22 are all weinstein homotopic ( definition [ whomotopy ] ) .
* question : * suppose @xmath23 is a family of sections of @xmath2 such that @xmath24 is smooth except when @xmath25 , @xmath26 is singular .
how do we describe the modification of the weinstein structure of @xmath27 in term of @xmath28 this paper studies the case @xmath26 is a * normal crossing divisor * with no triple intersection . by definition ,
a divisor is normal crossing if there exists local holomorphic coordinates @xmath29 near each point @xmath30 such that @xmath26 is given by the equation @xmath31 for some @xmath32 .
@xmath26 has no triple intersection if and only if @xmath33 . in symplectic topology , there are parallel notions of symplectic divisor and symplectic normal crossing divisor ( @xcite , @xcite ) .
although we make some use of the integrable complex structure in the proof of the main result , there should be similar results . in the case
@xmath2 is very ample , bertini s theorem @xcite guarantees the existence of @xmath34 near @xmath35 . without loss of generality ,
assume @xmath36 where @xmath37 the singular set @xmath38 of @xmath26 is a smooth complex codimension 2 submanifold .
let @xmath39 be the base locus of the pencil generated by @xmath40 and denote @xmath41 the ( imprecise ) main result is the following ( for precise statements , see theorem [ 1 ] and theorem [ 2 ] ) . [ mainresult ] up to homotopy of weinstein structures , there is a weinstein ( domain ) embedding of @xmath42 the complement @xmath43 is decomposed into elementary weinstein cobordisms , whose critical points are corresponding to critical points of @xmath44 , with a shift of indices by 2 . in the case
the critical point in a cobordism has index @xmath45 , the ( lagrangian ) unstable submanifold can be obtained from the fibration @xmath46 and the ( lagrangian ) unstable submanifold of the corresponding critical point in @xmath44 .
we depart to discuss an application of theorem [ mainresult ] .
there is a large class of weinstein manifolds which appear from algebraic geometry : * affine varieties*. given an affine variety @xmath47 , @xmath48 inherits a weinstein structure from the distance function @xmath49 hironaka s theorem implies that @xmath48 can be compactified to a projective variety @xmath6 by adding a normal crossing divisor @xmath50 at infinity .
such compactification is generally very useful in computing symplectic invariants @xmath48 , such as its * symplectic homology*. symplectic homology , originally defined and extended in @xcite , @xcite , @xcite , is hamiltonian - floer theory applied to the non - compact setting . in the case of weinstein manifold @xmath51 , roughly speaking , symplectic homology is the homology of a chain complex generated by critical points of @xmath52 ( which record the topology of @xmath48 ) and reeb orbits of the contact manifold @xmath53 . by choosing appropriate boundary @xmath54 of @xmath48
, the compactification @xmath55 gives a nice description of the reeb dynamics @xmath54 as follows .
there exists an open subset @xmath56 such that @xmath57 is a circle bundle over the complement of a neighborhood of the singular set of @xmath38 in @xmath50 , whose circle fiber is a reeb orbit of @xmath54 .
the homology classes of the reeb orbits in @xmath58 can be written as linear combinations of the fibers in @xmath57 .
the differential of the symplectic homology chain complex can then be computed via some energy or topological arguments or via reducing the solutions of cauchy - riemann equations to some holomorphic curve counts ( @xcite , @xcite ) . back to our situation ,
the first statement in theorem [ mainresult ] implies that there is a viterbo transfer chain map @xmath59 in our forthcoming paper ( @xcite ) , following the bourgeois - ekholm - eliashberg s approach ( @xcite ) , we show that the mapping cone of @xmath60 is quasi - isomorphic to a chain complex @xmath61 constructed from the lagrangian unstable submanifolds .
this chain complex is parallel to the hochschild chain complex from the @xmath62-structure of these lagrangian unstable submanifolds in the wrapped fukaya category of @xmath63 ( @xcite , @xcite ) . since the lagrangian unstable submanifolds is obtained from the fibration @xmath64 and some lagrangian balls in @xmath38 , it is reasonable to expect that one may reduce computing @xmath61 to analyzing the lagrangian balls in @xmath38 and the fibration @xmath64 .
this may be useful because @xmath38 has complex codimension 2 in @xmath6 making it easier to work with . * acknowledgement * : this paper is the first part of the author s ph.d thesis .
the author would like to thank his advisor yakov eliashberg for many enlightening discussions throughout the years .
the author also thank mohammed abouzaid , denis auroux and tobias ekholm for very helpful conversations about symplectic homology .
we recall from @xcite some basic notions of weinstein structure .
[ wmanifold ] a _ weinstein structure _ on an open manifold @xmath48 consists of @xmath51 , where * @xmath65 is a symplectic form on @xmath48 , * @xmath66 is an exhausting generalized morse function , * the vector field @xmath67 given by @xmath68 is complete and gradient - like for @xmath52 . in the case
@xmath52 has only finitely many critical points , @xmath51 is said to be of _ finite type_. by an exhausting generalized morse function , we mean it is a proper and bounded from below function .
moreover , its critical points are either nondegenerate or embryonic .
the latter means that in a neighborhood of the critical point @xmath69 with coordinates @xmath70 , @xmath71 is in the birth - death family : @xmath72 the vector @xmath67 is gradient - like with respect @xmath52 if there is some riemannian metric on @xmath48 and a positive function @xmath73 such that @xmath74 a _ weinstein cobordism _
@xmath51 is a compact manifold @xmath48 with @xmath75 , which are regular level sets of the minimal and maximal values of @xmath52 , and such that @xmath76 satisfies all the conditions in definition [ wmanifold ] except that @xmath67 is inwardly and outwardly transversal to @xmath77 and @xmath78 respectively . if @xmath79 , then @xmath51 is called a _
weinstein domain_. given a weinstein domain @xmath80 we can attach a positive half of the symplectization @xmath81 ( with an appropriate strictly increasing function @xmath82 ) to obtain a weinstein manifold .
this is called the completion of @xmath48 , also denoted by @xmath48 .
any finite type weinstein manifold is the completion of some weinstein domain .
our main source of examples comes from affine varieties .
if @xmath47 is an affine variety , @xmath48 admits a weinstein structure @xmath83 and @xmath52 some @xmath84-small perturbation of @xmath85 .
suppose @xmath6 is a projective variety with a positive line bundle @xmath86 the complement @xmath21 of a normal crossing divisor @xmath87 of a holomorphic section @xmath10 of @xmath88 admits a morse function ( after some perturbation ) @xmath89 define @xmath90 any sublevel set @xmath91 with regular value @xmath92 is a weinstein domain .
the set of critical points of @xmath52 forms a compact subset in @xmath21 ( @xcite ) .
when @xmath92 is sufficiently large so that its complement in @xmath21 contains no critical point of @xmath93 , denote by @xmath21 the weinstein manifold obtained from completing @xmath91 .
[ whomotopy ] a _ weinstein homotopy _ of finite type weinstein structures @xmath94}$ ] is a family such that @xmath95}\mbox{crit}(\phi_t)$ ] is compact in @xmath48 and the underlying @xmath96 is a morse homotopy .
recall that @xmath96 ( with @xmath95}\mbox{crit}(\phi_t)$ ] is compact in @xmath48 ) is a morse homotopy if there exists a finite set @xmath97 such that for each @xmath98 , the function @xmath99 has a unique birth - death type critical point such that it does not lie on any other critical levels . for @xmath100 ,
@xmath99 is morse .
recall the setting of our problem : there is a family of holomorphic sections @xmath101 such that the zero sets @xmath102 are smooth except at @xmath103 , where @xmath26 is a normal crossing divisor with no triple intersection .
denote by @xmath104 the base locus of this family .
the restriction of @xmath105 to @xmath106 , where @xmath107 is the singular set of @xmath26 , is independent of @xmath108 up to constant addition .
the differential @xmath109 is thus independent of @xmath108 and we define @xmath110 [ 1 ] up to homotopy of weinstein structures , there is a weinstein embedding of @xmath42 for @xmath112 sufficiently small , there is a natural bijection between the set of critical points of @xmath105 in the complement @xmath43 and @xmath113 moreover , the index of a critical point equals the index of the corresponding element in @xmath111 plus 2 .
let @xmath114 be a sublevel of @xmath115 such that @xmath116 . for sufficiently small @xmath108 ,
the weinstein vector field @xmath117 is @xmath118-close to @xmath119 .
hence , by choosing a sublevel @xmath120 such that @xmath121 , we obtain an embedding @xmath122 proving the first half of the theorem .
the critical points of @xmath105 in @xmath123 converge to critical points of @xmath115 as @xmath124 .
let s analyze other critical points of @xmath105 .
we now show that given any neighborhood @xmath58 of @xmath38 , the other critical points of @xmath105 lie in @xmath58 for sufficiently small @xmath108 .
assume the contrary that there is a sequence @xmath125 such that as @xmath126 , @xmath127 then @xmath128 .
there is a local holomorphic coordinates @xmath129 near @xmath69 and a local trivialization @xmath130 of @xmath2 such that @xmath131 denote @xmath132 we have at @xmath133 , @xmath134 as @xmath126 , it contradicts finiteness of @xmath135 .
moreover , the other critical points of @xmath105 can not wander off to @xmath39 .
indeed , assume there is a sequence @xmath125 such that as @xmath126 , @xmath136 there is a local holomorphic coordinates @xmath137 and a local holomorphic trivialization @xmath130 of @xmath2 such that @xmath138 again , denote @xmath139 we have at @xmath140 @xmath141 @xmath142 @xmath143 the first two equations imply that there exists @xmath144 such that for sufficient small @xmath145 , at @xmath140 , @xmath146 thus @xmath147 contradicting finiteness of @xmath148 .
having checked that the other critical points of @xmath105 must stay near @xmath44 and away from @xmath39 , we proceed to prove the second half of the theorem .
it is obvious that if a sequence @xmath125 converges to @xmath149 , then @xmath150 .
conversely , we will show that for each @xmath151 any sufficiently small neighborhood @xmath58 of @xmath69 satisfies @xmath152 has a unique critical point for sufficiently small @xmath153 let @xmath154 be local holomorphic coordinates near @xmath69 and local holomorphic trivialization @xmath155 of @xmath2 such that @xmath156 .
denote by @xmath157 let s consider the function @xmath158 since the derivative @xmath159 is nondegenerate , inverse function theorem implies that @xmath160 is a local diffeomorphism . in particular
, for each sufficiently small @xmath108 , there exists a unique solution @xmath161 near @xmath162 satisfying @xmath163 , i.e. @xmath164 is a critical point of @xmath105 .
finally , we need to show that @xmath165 . computing the second derivatives of @xmath105 , we obtain all second derivatives of @xmath105 at @xmath164 are bounded except for @xmath166 which have order of @xmath167 .
this implies that as @xmath124 , there are two eigenvalues of @xmath168 diverging to @xmath169 , two eigenvalues diverging to @xmath170 , and the other eigenvalues converging to the eigenvalues of @xmath171 ( which are independent of @xmath108).therefore , @xmath172 [ ex1 ] let @xmath173 with coordinates @xmath174 $ ] and @xmath175 $ ] is the hyperplane class . take @xmath176 and equip @xmath2 with the following metric : on the affine part @xmath177 , the norm of the section @xmath178 in @xmath179 is @xmath180 .
consider @xmath181 where @xmath182 are generic .
then @xmath63 and @xmath27 are weinstein equivalent to @xmath183 and @xmath184 to see the first , consider the hypersurface @xmath185 . the weinstein structure on @xmath186
can be identified with a weinstein structure on @xmath187 , which is homotopic to the standard weinstein structure on @xmath188 ( @xcite ) .
there is a free @xmath189-action on @xmath186 : @xmath190 { \rightarrow}[z_0 : ... : z_n : -z_{n+1}],\ ] ] the quotient of which can be identified with @xmath63 . the corresponding @xmath189-action on @xmath188 gives the standard weinstein structure on @xmath191 the critical point set of @xmath115 is @xmath192 | |z_0| = |z_1|\}.$ ] the critical point set of @xmath105 is the union of the critical point set of @xmath115 and @xmath193 , ... , [ 0 : 0 : 0 : ... : 1]\}.\ ] ] [ ex2 ] let @xmath194 with first and second factors
coordinates @xmath174 $ ] and @xmath195 $ ] and @xmath175 $ ] and @xmath196 $ ] are the hyperplane classes of the first and second factors .
take @xmath197 and equip it with a similar metric described in the previous example .
consider @xmath198 where @xmath199 are generic .
then @xmath63 and @xmath27 are weinstein equivalent to @xmath200 and @xmath201 to see the first , consider the hypersurface @xmath202 the weinstein structure on @xmath203 can be identified with a weinstein structure on @xmath204 which is homotopic to the standard weinstein structure on @xmath205 .
there is a free @xmath206-action on @xmath160 : @xmath207 , [ z'_0 : ... : z'_n : z'_{n+1 } ] ) = ( [ tz_0 : ... : tz_n : z_{n+1 } ] , [ t^{-1}z'_0 : ... : t^{-1}z'_n : z'_{n+1}]),\ ] ] the quotient of which can be identified with @xmath63 .
the corresponding @xmath208-action on @xmath205 gives the standard weinstein structure on @xmath209 the function @xmath115 has only one critical point @xmath210 , [ 1 : 0 : ... : 0])$ ] , which has index 0 .
the set of critical points of @xmath105 consists of the critical point of @xmath115 and @xmath211 points @xmath212 \big| 1 \leq i \leq n\},\ ] ] with critical indices @xmath213 consider two holomorphic sections @xmath214 and @xmath215 of some positive line bundle@xmath216 such that @xmath217 is smooth . define @xmath218 with line bundle @xmath219 define @xmath220 and @xmath221 and assume that @xmath102 is smooth except for @xmath103 .
notice that @xmath222 $ ] is a normal crossing divisor with no triple intersection .
its singular set @xmath38 is @xmath223 $ ] .
the divisor @xmath102 is the blow up of @xmath6 along the base locus @xmath224 of the pencil generated by @xmath214 and @xmath215 .
the complement @xmath27 is @xmath225 with subcritical weinstein structure ( i.e. the index of each critical point is strictly less than @xmath226 ) .
all of the interesting symplectic topology lies in a neighborhood of @xmath227 in @xmath228 [ prop1 ] let @xmath229 be the @xmath19-orthogonal complement of @xmath230 .
assume that the derivative of @xmath231 vanish in @xmath232 . then @xmath233 moreover , the vector field @xmath117 is tangent to @xmath44 . as a consequence ,
the unstable submanifold of a critical point @xmath69 of @xmath105 which lies in @xmath111 contains the unstable submanifold ( lying in @xmath44 ) of @xmath69 of @xmath234 the proof is straightforward from the following computation . in @xmath44 , @xmath235 thus since @xmath236 , @xmath150 implies that @xmath237 i.e. @xmath69 is a critical point of @xmath105 .
since @xmath238 vanish in @xmath232 , @xmath117 must be @xmath239orthogonal to @xmath232 .
hence @xmath117 is tangent to @xmath44 .
the rest of the paper discusses about the unstable submanifold of @xmath240 which corresponds to @xmath241 under special assumption , proposition [ prop1 ] implies that the unstable submanifold @xmath242 of @xmath164 is a disc bundle over the unstable submanifold @xmath243 lying in @xmath44 of @xmath69 .
consider @xmath244 with coordinates @xmath154 and @xmath245 a pluri - subharmonic function .
if @xmath69 is a critical point of @xmath52 , then @xmath246 and @xmath69 is a critical point of @xmath247 .
denote by @xmath248 the unstable submanifold of @xmath69 of @xmath247 .
then the unstable submanifold @xmath242 of @xmath69 of @xmath52 is @xmath249 since we assume @xmath256 is smooth for small @xmath108 , the fibration induces a connection in @xmath257 as follows : at each point @xmath258 , define @xmath259 the subspace that is @xmath260orthogonal to the fiber @xmath261 .
consider @xmath262 { \rightarrow}\mathbb{c } : \gamma(t ) = t$ ] .
the connection induces a characteristic foliation on @xmath263)$ ] . for any compact subset @xmath264 ,
we denote by @xmath265 the set of points in @xmath266)$ ] which under the characteristic flow converge to points in @xmath267 . notice that @xmath268 is a nested sequence :
if @xmath269 @xmath270 this is a normal crossing generalization of the standard lefschetz fibration picture in symplectic topology ( @xcite ) . since @xmath273 @xmath274 is tangent to the level sets @xmath275 . in particular , @xmath276 is tangent to the circle @xmath277 since @xmath64 is holomorphic , @xmath278 is perpendicular to the circle @xmath279 .
thus , @xmath254 is tangent to @xmath272 .
[ 2 ] suppose @xmath69 has critical index , i.e. @xmath282 is a lagrangian in @xmath44 .
let @xmath283 be a compact ball containing @xmath69 .
given a small neighborhood @xmath284 of @xmath267 , then for each sufficiently small @xmath108 , there exists a weinstein homotopy that starts with @xmath285 and stays constant outside @xmath286 such that the resulting weinstein structure has in @xmath286 a unique critical point coinciding with @xmath69 and its unstable submanifold in @xmath267 is @xmath265 . in complex two dimensional case , @xmath38 is a set of finitely many points .
let @xmath288 and @xmath289 .
denote by @xmath290 , also called a thimble in the lefschetz fibration @xmath64 .
fix an @xmath291 ( we will later make a choice of sufficiently small @xmath292 ) .
since @xmath293 is a lagrangian disc , weinstein neighborhood theorem implies its neighborhood is symplectomorphic to @xmath294 this gives coordinate @xmath295 in a neighborhood of @xmath293 such that notice that lemma [ lefschetz ] implies that @xmath307 is the unstable submanifold of @xmath304 .
in fact , for sufficiently small @xmath308 , the neighborhood @xmath309 satisfies @xmath310 is a weinstein structure for all sufficiently small @xmath108 .
this follows from @xmath311 and @xmath312 on @xmath293 . indeed , it is sufficient to check that @xmath331 is @xmath292-uniformly bounded . since @xmath332 , there is @xmath144 such that @xmath333 also @xmath334 can be choosen so that @xmath335 hence on @xmath336 , we obtain @xmath337 outside this region , by definition @xmath338 .
thus , @xmath331 is @xmath292-uniformly bounded .
lemma [ bound ] implies that the liouville vector field @xmath339 of @xmath326 ( i.e. @xmath340 ) satisfies @xmath341 is @xmath292-uniformly bounded .
denote @xmath342 we will show that @xmath343 is a weinstein structure when @xmath292 and @xmath108 are sufficiently small . in @xmath344
it is the weinstein structure @xmath345 while in @xmath346 , it is the weinstein structure @xmath347 . consider the region @xmath348 .
write @xmath349 since @xmath350 is a finite function in @xmath315 with value @xmath351 at @xmath69 , a similar arguments as in lemma [ bound ] implies that @xmath352 is @xmath292-uniformly bounded .
hence , @xmath353 is @xmath292-uniformly bounded .
now , @xmath354 where the first term has order @xmath355 and the later has order @xmath356 if @xmath357 is sufficiently small .
therefore by choosing @xmath292 sufficiently small and @xmath108 such that @xmath357 sufficiently small , we obtain that @xmath339 is gradient - like with respect to @xmath358 .
to construct a homotopy of weinstein structure from @xmath285 to @xmath343 , consider the path of 1-forms @xmath359 and functions @xmath360 , where @xmath361 $ ] .
a similar argument as before show that they form a weinstein structure in @xmath348 . in the region @xmath362
, unfortunately they are not .
for example , the zero of the vector field may not be a critical point of the function .
however , a similar argument as in theorem [ 1 ] implies that for sufficiently small @xmath108 , there is a unique zero of @xmath359 and a unique critical point of @xmath360 in @xmath363 . since @xmath311 and @xmath312 on @xmath293
, we also obtain @xmath364 = ( 1-t)\lambda_{\epsilon}+ t\tilde{\lambda}_{\epsilon},\ ] ] @xmath365 = d((1-t)\lambda_{\epsilon}+ t\tilde{\lambda}_{\epsilon } ) = \omega,\ ] ] on @xmath293 .
choosing @xmath292 sufficiently small , we can modify @xmath360 to make it a weinstein structure with @xmath359 in @xmath362 . in higher dimensional case ,
the proof is similar .
let @xmath267 be a compact ball in @xmath44 containing @xmath69 . since @xmath267 is a lagrangian in @xmath44 and @xmath254 preserves @xmath19 , @xmath368 is lagrangian in @xmath6 .
for sufficiently small @xmath292 , @xmath368 is a two - disc bundles over @xmath267 .
the coordinate defines a projection @xmath376 from the neighborhood to @xmath44 .
define @xmath377 and @xmath378 .
notice that @xmath303 .
the liouville vector field @xmath304 of @xmath305 is given by @xmath379 where @xmath380 is the weinstein vector field of @xmath381 and is independent of @xmath108 .
denote @xmath382 since @xmath315 is contractible , we can similarly construct @xmath318 such that @xmath383 such that @xmath384 let @xmath385 such that @xmath82 has compact support in a neighborhood of @xmath267 .
let @xmath386 be the cutoff function as before except that we also multiply it with @xmath387 .
the 1-form @xmath338 along @xmath388 and hence can be arbitrarily small along the horizontal directions , ( i.e. the @xmath389 directions ) if @xmath390 is sufficiently small . also , as is proven in the lower dimensional case
, @xmath331 is @xmath292-uniformly bounded along the vertical directions ( i.e. the @xmath391 directions ) .
this enables us to adapt the proof of the case @xmath392 to higher dimensional case .
given @xmath283 some sublevel in @xmath44 , for sufficiently small @xmath108 , there is a weinstein homotopy of @xmath285 to @xmath343 such that the unstable submanifold of @xmath69 in some regular sublevel @xmath393 is @xmath394 .
moreover , one can choose a boundary @xmath398 of @xmath343 ( not necessarily a level set of @xmath358 , but the vector field @xmath339 is outwardly transversal to @xmath398 ) so that the contact manifold @xmath399 is foliated by reeb orbits which form a circle bundle over @xmath102 .
the intersection @xmath400 is a legendrian knot , a lift of a lagrangian @xmath401 in @xmath102 , which is hamiltonian isotopic to @xmath397 , to the unit circle normal bundle of @xmath102 in @xmath6 .
the first half of the statement is obvious from the construction .
consider the normal bundle @xmath402 and equip it a connection @xmath403form @xmath404 so that the curvature of @xmath404 is @xmath19 in @xmath102 .
notice that @xmath404 is a 1-form on @xmath405 , invariant under radial scaling and @xmath406 extends and equals @xmath19 in @xmath102 .
symplectic neighborhood theorem yields a symplectomorphism @xmath60 from a neighborhood of the zero section of @xmath407 ( equipped with symplectic form @xmath408 ) to a neighborhood of @xmath102 in @xmath6 such that @xmath409 on @xmath410 .
let @xmath398 be the image of the circle bundle of radius @xmath411 under @xmath60 .
outside @xmath315 ( where the weinstein homotopy stays constant ) , it is well - known that @xmath117 is outwardly transversal to @xmath398 . inside @xmath315 , from the construction , it is sufficient to check that the vector field @xmath254 is outwardly transversal to @xmath398 . recall that @xmath254 lies in the @xmath19-orthogonal connection induced by the holomorphic map @xmath412 .
the vector field @xmath278 is perpendicular to circles with center @xmath108 in @xmath413 .
the property of @xmath60 implies that when @xmath411 is sufficiently small , @xmath278 is transversal to the image of each fiber of @xmath398 under @xmath414 hence @xmath254 is outwardly transversal to @xmath398 and consequently @xmath339 is outwardly transversal to @xmath398 .
since @xmath307 is lagrangian and @xmath339 is transversal to @xmath398 , the intersection @xmath415 is a legendrian knot .
this legendrian intersects transversally each circle fiber at at most one point because the image @xmath416 is the segment @xmath417 $ ] .
flowing @xmath418 along @xmath419 gives an embedded lagrangian circle @xmath401 .
similarly , for each @xmath361 $ ] , flowing @xmath418 along the vector field @xmath420 gives an embedded lagrangian circle @xmath421 .
we claim that @xmath422 are all hamiltonian isotopic .
choose a bijective parametrization @xmath423 , we then get a parametrization @xmath424 .
denote by @xmath425 \times s^1 { \rightarrow}d_{\epsilon},$ ] @xmath426 . to show @xmath427 are all hamiltonian isotopic ,
it is sufficient to check @xmath428 \times s^1 } \tilde{\theta}^*\omega = 0.\ ] ] let @xmath429 be the lagrangian cylinder obtained from flowing @xmath418 along the vector field @xmath420 ( in particular , @xmath430 ) .
we cap @xmath431 \times s^1)$ ] with @xmath432 and @xmath429 to obtain a ( torus ) cycle @xmath433 in @xmath6 .
in fact @xmath433 lies in the complement of @xmath26 , where @xmath19 is exact .
it follows that @xmath434 and thus @xmath435 \times s^1 } \tilde{\theta}^*\omega = 0 $ ] holds .
therefore , @xmath401 is hamiltonian isotopic to @xmath397 . | we describe how the weinstein structure of the complement is modified when a positive divisor degenerates to a normal crossing divisor with no triple intersection . |
in this section we consider a vortex in the order parameter similar to the @xmath5 phase of @xmath40he .
namely , we assume that the condensate wave function has the form @xmath41 ( k_x \pm { i}k_y ) \ ; , \label{a - phase}\ ] ] and that the vector @xmath42 defining the orientation of the triplet is _ fixed _ ( without loss of generality , take it to be in the @xmath43 direction .
we also assume a fixed chirality of the order parameter and take positive sign in place of @xmath44 in ( [ a - phase ] ) . then the only allowed vortices are those involving a rotation of the phase of the order parameter , and the magnetic flux in such vortices is quantized in conventional superconducting flux quanta @xmath45 .
below we analyze the low - lying energy levels obtained as solutions to bogoliubov
degennes equations for the mean - field hamiltonian @xmath46 \psi_\alpha + \psi_\uparrow^\dagger \left ( \delta_x * { p_x\over k_f } + \delta_y * { p_y\over k_f } \right ) \psi_\downarrow^\dagger + { \rm h.c . }
\ ; , \label{hamiltonian}\ ] ] where @xmath47 are the electron operators ( @xmath11 is the spin index ) , @xmath48 is the external potential of impurities , @xmath49 is the electromagnetic vector potential , @xmath50 and @xmath51 are the coordinate - dependent components of the superconducting gap .
[ in the bulk , the preferred superconducting order is one of the two chiral components @xmath52 , but in inhomogeneous systems , such as a vortex core , an admixture of the opposite component may be self - consistently generated @xcite
. we account for this effect by allowing the two independent order parameters @xmath53 and @xmath54 . ]
star ( @xmath55 ) denotes the symmetrized ordering of the gradients @xmath56 and the order parameters @xmath57 [ definition : @xmath58 . at infinity
, the order parameter imposes the vortex boundary conditions : @xmath59 where @xmath60 and @xmath61 are polar coordinates .
plus or minus signs in the exponent correspond to a positive or a negative single - quantum vortex . for an axially - symmetric vortex with the chirality of the order parameter non - self - consistently fixed ( @xmath62 ) , without the vector - potential @xmath49 and without disorder @xmath48 ,
the low - lying eigenstates of the hamiltonian ( [ hamiltonian ] ) have been found by kopnin and salomaa @xcite .
the spectrum is @xmath63 with @xmath64 .
this result should be compared with the spectrum of the vortex core in a @xmath65-wave superconductor @xcite : @xmath66 the common feature of the spectra in the @xmath65-wave and @xmath0-wave cases is the symmetry with respect to zero energy . if we interpret holes in the negative - energy levels as excitations with positive energies ( and with opposite spin )
, then this symmetry implies that the excitations are doubly degenerate in spin : to each spin - up excitation there corresponds a spin - down excitation at the same energy . for the @xmath65-wave vortex , this degeneracy is due to the full spin - rotation @xmath67 symmetry .
the @xmath0-wave hamiltonian ( [ hamiltonian ] ) has a reduced spin symmetry .
namely , it has the symmetry group @xmath68 generated by rotations about the @xmath69-axis ( @xmath70 , @xmath71 ) and by the spin flip @xmath72 , @xmath73 .
this non - abelian group causes the two - fold degeneracy of all levels , except for the zero - energy level(s ) where the symmetry @xmath68 may mix the creation and annihilation operators for the same state .
this symmetry is crucial for our discussion .
note that we have not included in the hamiltonian neither the spin - orbit term @xmath74)$ ] , nor the zeeman splitting @xmath75 .
either of these terms would break the spin symmetry @xmath68 , which would eventually result in a different universality class of the disordered system ( type @xmath2 with non - degenerate levels ) , if these terms are sufficiently strong .
the difference between the @xmath65- and @xmath0-wave vortices is the zero - energy level in the @xmath0-wave case .
it has been shown by volovik that this level has a topological nature @xcite . indeed , suppose we gradually increase disorder in the hamiltonian ( [ hamiltonian ] ) . the levels shift and mix , but
the degeneracy of the levels remains the same as long as the symmetry @xmath68 is preserved .
the total number of levels remains _ odd _ , and therefore the zero - energy level _ can not shift _ if the final hamiltonian is a continuous deformation of the original one ( without disorder ) , i.e. if the topological class of the boundary conditions ( [ boundary - conditions ] ) remains the same .
now we can identify the symmetry class of the bogoliubov
degennes hamiltonian .
the only symmetry of the mean - field hamiltonian ( [ hamiltonian ] ) is the spin symmetry @xmath68 .
the time - reversal symmetry is already broken by the vortex and by the pairing , and therefore neither the vector potential @xmath76 nor local deformations of @xmath57 can reduce the symmetry of the hamiltonian . when projected onto spin - up excitations @xmath77 d^2 { { \mathbf r}}$ ] , the bogoliubov
degennes hamiltonian for the two - component vector @xmath78 takes the form : @xmath79 & \left[{\delta_x\over k_f}*(-{i}\nabla_x)+ { \delta_y\over k_f}*(-{i}\nabla_y)\right ] \cr \left[{\delta^*_x\over k_f}*(-{i}\nabla_x)+ { \delta^*_y\over k_f}*(-{i}\nabla_y)\right ] & - \left [ { ( -{i}\nabla + e { { \mathbf a}})^2 \over 2 m } + v({{\mathbf r } } ) -\varepsilon_f \right ] } \ ; . \label{bdgequation}\ ] ] in an arbitrary orthonormal basis of electronic states , this hamiltonian may be written as a matrix @xmath80 from the hermiticity of the hamiltonian , it follows that @xmath81 . from the explicit form of the @xmath0-wave pairing , @xmath82
( it is here that the @xmath0-wave structure of the pairing is important ; for @xmath65-wave pairing we would have @xmath83 instead ) .
these are the only restrictions on the hamiltonian ( [ ham - short ] ) .
if we define @xmath84 the restrictions on the hamiltonian ( [ ham - short ] ) are equivalent to the condition that the rotated matrix @xmath85 is real antisymmetric , i.e. it belongs to the lie algebra @xmath86 , where @xmath87 is the dimension of the hilbert space ( the same rotation of the hamiltonian was used in @xcite to identify the @xmath2 universality class ) .
the last step in our argument is to note that , under the vortex boundary conditions , the dimension of the hamiltonian ( [ ham - short ] ) is odd , not even ( this may be difficult to visualize from the particle - hole representation ( [ ham - short ] ) , but easier from the rotated hamiltonian @xmath88 ) .
thus for a single - quantum vortex , we identify the space of the hamiltonians as @xmath3 ( class @xmath1 in cartan s notation , see table 1 ) .
now let us consider a slightly different order parameter : suppose that the condensate wave function is described by the same expression ( [ a - phase ] ) , but now the vector @xmath42 is able to rotate ( either in plane or in all three dimensions ) .
the order parameter does not change under simultaneous change of sign of the vector @xmath42 and phase shift by @xmath89 of the phase @xmath90 : @xmath91 .
this makes possible a half - quantum vortex with magnetic flux @xmath92 @xcite . on going around such a vortex , both the vector @xmath42 and the phase @xmath90 rotate by @xmath89 ( fig . [ fig1]a ) ( see also discussion of such vortices in @xcite ) . .
( * b * ) spin vortex .
the angle @xmath93 is shown , scaledwidth=75.0% ] consider first the case when the vector @xmath42 rotates in a plane ( without loss of generality , let @xmath42 rotate in the @xmath94@xmath95 plane ) .
then the condensate wave function ( [ a - phase ] ) is the sum of two decoupled components containing @xmath96 and @xmath97 pairings : @xmath98 ( k_x + { i}k_y)\ ] ] [ here @xmath60 and @xmath93 are the polar coordinates ] . in a half - quantum vortex , one of the pairing components @xmath99 and @xmath100 has an odd winding number of the phase around the vortex , while the other has an even winding number of the phase . in the simplest vortex of this type ,
the even winding number is zero , the corresponding component of the order parameter is nearly uniform and does not produce states near fermi energy .
the component with an odd winding number ( without loss of generality , let it be @xmath99 , and the winding number be one ) has vortex boundary conditions .
the mean - field hamiltonian decouples and the vortex part of the hamiltonian ( spin - up sector ) takes the form [ we have also assumed an axially symmetric order parameter ] : @xmath101\ ; .
\label{ham - halfquantum}\ ] ] in this form the half - quantum vortex is equivalent to a single - quantum vortex in a superconductor of spinless fermions ( such a superconductor must have odd pairing ) .
the quasiparticles do not have a definite spin projection , but mix particles and holes in the same spin - up sector : @xmath102 { d}^2 { { \mathbf r}}$ ] . the bogoliubov degennes equations for @xmath103 and @xmath104 turn out to be identical to those for a single - quantum vortex ( [ bdgequation ] ) , and the solution of kopnin and salomaa @xcite is equally applicable to this half - quantum vortex resulting in the same spectrum ( [ p - spectrum ] ) .
however , because of the relation @xmath105 , the number of subgap states in the half - quantum vortex is one half of those in a single - quantum vortex : particles in negative - energy levels are identical to holes in positive - energy levels .
the zero - energy state is a majorana fermion @xcite .
now add a disorder term to the hamiltonian of a clean vortex ( [ ham - halfquantum ] ) .
consider the following perturbations : * potential scattering : @xmath106 , * zeeman field : @xmath107 , * vector potential : @xmath108 , * spin - orbit scattering : @xmath109 \right ) \psi_\beta$ ] , * deformation of order parameter : + @xmath110 we assume that the perturbation may be switched on adiabatically in such a way that subgap levels stay localized .
one possible way to fulfill this condition is to require that the perturbation is introduced only in a finite region around the vortex core .
the symmetry classification proceeds slightly differently depending on whether the perturbation preserves the decoupling of the hamiltonian into spin - up and spin - down sectors .
the hamiltonian remains decoupled under the perturbations @xmath111 , @xmath112 , @xmath113 with @xmath114 , @xmath115 with @xmath116 , and @xmath117 with diagonal @xmath118 . with these perturbations ,
the `` spin - up '' sector of the hamiltonian in the basis @xmath78 preserves its form ( [ ham - short ] ) with the same symmetries as for a single - flux vortex . the symmetry class of the hamiltonian is therefore identified as @xmath3 ( class @xmath1 ) .
if other perturbations are present , so that the hamiltonian no longer splits into `` spin - up '' and `` spin - down '' parts , the full matrix of the hamiltonian needs to be considered . without any spin structure ,
the hamiltonian breaking time - reversal and spin symmetries belongs to the @xmath86 symmetry class , as shown by altland and zirnbauer @xcite .
their argument also works in our case , with the reservation that the number of levels @xmath87 is odd ( with each level counted twice : once as a particle and the other time as a hole ) .
it follows from the usual continuity consideration ( the majorana fermion at zero energy survives any local perturbation ) .
therefore the symmetry class is again @xmath3 ( class @xmath1 ) .
[ in the examples considered by altland and zirnbauer , the number of levels was even , which lead to the symmetry class @xmath2 or @xmath119 .
it is important however , that to reach the rmt limit in the latter case , such a system must have sufficiently strong disorder to bring the quasiparticles from the `` spin - down '' sector ( gapped in the clean vortex ) to the fermi energy and to mix them strongly with the quasiparticles in the `` spin - up '' sector . in the present paper
, we do not discuss conditions for reaching this limit .
in this section we consider a hypothetical phase of a triplet superconductor analogous to the phase @xmath1 of @xmath40he .
namely , the condensate wave function is assumed to be @xmath120\ ; , \qquad { { \mathbf d}}({{\mathbf k}})=r_{\theta}({{\mathbf k}})\ ; , \label{b - phase}\ ] ] and @xmath121 is a rotation in the plane by angle @xmath93 ( like in the previous sections , we consider a two - dimensional problem , and @xmath122 is a two - dimensional vector ) . in this phase , the vector @xmath42 rotates a full turn as the vector @xmath122 goes around the circular fermi surface .
the parameter @xmath93 denotes the angle between vectors @xmath122 and @xmath42 .
we neglect spin - orbit interactions and assume that all values of @xmath93 are allowed .
it is important that this phase preserves the time - reversal symmetry .
moreover , a vortex structure is possible without time - reversal symmetry breaking .
it is realized by rotating the angle @xmath93 a full turn of @xmath123 when going around the vortex center and keeping the phase @xmath90 constant ( fig .
[ fig1]b ) .
this vortex preserves the time - reversal symmetry and is a topological defect in the spin structure of the triplet pairing , thus we shall call it `` spin vortex '' .
the hamiltonian of the clean axially symmetric spin vortex splits into `` spin - up '' and `` spin - down '' sectors , similarly to the case of the half - quantum vortex considered in the previous section . in the spin vortex , both the `` spin - up '' and the `` spin - down '' components have vortex structure .
the `` spin - up '' hamiltonian is identical to the hamiltonian @xmath124 of ( [ ham - halfquantum ] ) , while the `` spin - down '' component is its complex conjugate @xmath125 ( with spin reversed ) or vice versa . in a clean spin vortex , the spectrum of each of the two vortices ( in the `` spin - up '' and the `` spin down '' sectors ) are identical to that in the half - quantum vortex of the previous section , and
therefore the total spectrum coincides with that of the single - flux vortex of section i ( with the two majorana fermions combining into a single zero - energy level ) .
in contrast to the single - quantum vortex , it is now not the spin symmetry which is responsible for the double degeneracy of the levels with non - zero energy , but the time - reversal symmetry ( the degeneracy is due to kramers theorem ) .
now suppose that the hamiltonian is perturbed by a disorder term preserving the time - reversal symmetry .
the allowed perturbations include potential scattering @xmath111 , spin - orbit scattering @xmath115 , as well as deformation of the order parameter @xmath117 in a time - reversal invariant way .
then the levels with non - zero energy stay doubly degenerate , and the zero - energy level can not shift ( because it is non - degenerate ) .
the symmetry class of the perturbed hamiltonian may be identified from the argument of altland and zirnbauer as @xmath2iii ( time - reversal symmetry present , spin symmetry broken ) @xcite . in the @xmath0-wave vortex ,
the number of subgap levels is odd , and therefore the symmetry class is @xmath2iii - odd , in contrast to class @xmath2iii - even studied by altland and zirnbauer [ similarly to distinguishing between classes @xmath1 and @xmath2 , we find it useful to distinguish between classes @xmath2iii - even and @xmath2iii - odd : they have different parameters @xmath11 , as a consequence of the zero - energy level in odd dimensions ] .
in this section we discuss whether the rmt limit may be realized in disordered vortices .
we suppose that the superconductor contains a finite concentration @xmath126 of spinless impurities with strong electron scattering ( born parameter is of order one ) . then the effect of impurities is dual
: they directly contribute a potential term @xmath111 in the hamiltonian and they also suppress the superconducting order parameter thus inducing an inhomogeneous pairing perturbation @xmath117 .
we shall see that the latter effect has a much stronger influence on shifting the energy levels of the subgap states .
consider first the potential term @xmath111 with the impurity potential @xmath127 .
the matrix element of the impurity potential between the two localized states is equal to @xmath128 from explicitly solving bogoliubov
degennes equations for a clean @xmath0-wave vortex @xcite , one finds that @xmath129 and @xmath130 are proportional to each other , to the leading order in @xmath131 ( where @xmath132 is the energy of the @xmath133-th level )
. therefore the matrix element @xmath134 between any two low - lying states nearly vanishes ( this was noticed by volovik in @xcite ) .
a more accurate calculation shows that matrix elements of the charge density between states at energy @xmath35 is of the order of @xmath135 , which is by the factor @xmath136 smaller than that in the @xmath65-wave vortex . as a consequence , even for strong impurities ( with @xmath137 ) , an extremely high impurity concentration ( @xmath138 ) is required to mix low - energy levels a much higher disorder than that destroying @xmath0-wave superconductivity ( @xmath139 ) @xcite .
therefore , for a uniformly disordered @xmath0-wave superconductor , the @xmath111 term is insufficient for the rmt regime in the vortex core .
the suppression of the order parameter @xmath117 turns out to be a more important effect . in unconventional superconductors ,
spinless impurities suppress pairing in the region of size @xmath140 which is of the same order as the size of the vortex core . for a homogeneous order parameter ( without vortex ) ,
the suppression of the order parameter @xmath141 decreases as @xmath142 ( in our two - dimensional problem ) at distances up to @xmath140 , and is negligible at distances much larger than @xmath140 @xcite , and the integral suppression may be estimated as @xmath143 .
also , in a chiral superconductor , the component of the order parameter of opposite chirality is admixed in the vicinity of impurity @xcite .
the two effects gives comparable contributions to the matrix elements between subgap states .
an accurate calculation of the impurity influence on the states in the vortex core requires a full self - consistent solution for the order parameter taking into account both the vortex and the impurities .
this goes beyond the scope of this paper , and we only estimate the order of magnitude of the corresponding matrix elements .
the deformation of the order parameter @xmath117 has the matrix elements between subgap states @xmath144 and @xmath133 : @xmath145 u_n({{\mathbf r } } ) \nonumber\\ & & + v_m^*({{\mathbf r } } ) \left[\delta{\delta^{(\alpha)}}^*({{\mathbf r } } ) * ( { i}\nabla_\alpha)\right ] v_n({{\mathbf r } } ) \bigg)\ ; .\end{aligned}\ ] ] unlike the case of the matrix elements of @xmath111 , there is no cancellation in @xmath146 , and we may estimate , for one impurity in the vortex core @xmath147 thus , a single impurity introduces matrix elements of the order of interlevel spacing .
these estimates suggest that the rmt regime may be achieved when the number of impurities in the vortex core is much greater than one ( which corresponds to the moderately clean regime @xmath148 ) .
we have shown that symmetries of vortex hamiltonians in @xmath0-wave superconductors correspond to one of the two random - matrix ensembles : @xmath1 or @xmath2iii - odd .
these two rmt ensembles are distinguished from the rest of ensembles by the presence of zero - energy modes ( zero - energy modes are also present in chiral ensembles at @xmath149 , see table [ table1 ] ) .
zero - energy levels appear in @xmath0-wave vortices as a consequence of odd pairing symmetry combined with the vortex ( topologically nontrivial ) boundary conditions .
we have checked the symmetries of the hamiltonian explicitly taking three vortices as examples : a single - quantum vortex with spin symmetry , a half - quantum vortex , and a flux - less ( spin ) vortex without time - reversal symmetry breaking .
based on these examples , we may conjecture that classes @xmath1 and @xmath2iii - odd appear in any vortex - like structure with odd pairing , whenever a zero - energy mode is present . in cases with time reversal symmetry
, the vortex belongs to the class @xmath2iii - odd ; when the time - reversal symmetry is broken to the class @xmath1 . of course
, this can also be checked explicitly in any particular case .
the two classes @xmath1 and @xmath2iii - odd considered in this paper are odd - dimensional counterparts of the classes @xmath2 and @xmath2iii - even , respectively ( see table 1 ) , with the even - dimensional classes realized in conventional ( singlet - pairing ) normal - superconducting structures @xcite .
far from zero energy , the statistics of energy levels is not affected by the zero mode . in the immediate vicinity of zero energy ,
repulsion from the zero mode in classes @xmath1 and @xmath2iii - odd increases the exponent @xmath11 ( see table 1 ) , suppressing the average density of states @xmath150 as @xmath151 for class @xmath1 , and @xmath152 for class @xmath2iii - odd . because of the zero modes , the classes @xmath1 and @xmath2iii - odd stand somewhat separately from the rest of rmt ensembles .
in particular , they do not appear in the table of correspondence between symmetries of the hamiltonian and those of the transfer matrix @xcite ( for rmt ensembles without zero - energy levels , this table establishes a one - to - one correspondence ) .
neither these two ensembles are known to be derived from a supersymmetric sigma - model as other classes are @xcite .
understanding the latter fact remains an interesting problem , as well as a possible microscopic derivation of the random - matrix theory for @xmath0-wave vortices with the supersymmetric method , analogously to that for the @xmath65-wave vortex @xcite .
several more comments can be made on comparing our results to the level statistics in @xmath65-wave vortices .
the latter is known to belong to the symmetry class @xmath7 @xcite having spin - rotational symmetry and broken time - reversal symmetry @xcite .
in contrast with the @xmath67 spin symmetry in conventional superconductors , the single - quantum vortex considered in our paper has the @xmath68 spin symmetry instead , which leads to a somewhat different symmetry classification . the resulting symmetry is @xmath153 , which leads to classes @xmath2 ( in even dimensions , no vortex ) or @xmath1 ( odd dimensions , vortex with zero mode ) .
the joint probability distributions of the energy levels ( [ probability - alpha - beta ] ) for classes @xmath1 ( @xmath0-wave vortex ) and @xmath7 ( @xmath65-wave vortex ) coincide ( with @xmath154 ) , and only the zero - energy level distinguishes the two ensembles . in particular , the average density of states in the class @xmath1 random - matrix ensemble is @xmath155 ( for class @xmath7 it is the same but without the @xmath156 term ) . in @xmath65-wave vortices , koulakov and
larkin found that in a wide range of impurity concentrations , the level mixing does not lead to a class @xmath7 , but has more symmetries @xcite .
these symmetries arise from the fact that the impurities are local scatterers in a chiral quasi - one - dimensional two - channel system , resulting in an ensemble of @xmath157 random matrices , instead of @xmath158 matrices with large @xmath12 .
such effect seems improbable in @xmath0-wave vortices where the main contribution to level mixing comes from the distortion of the order parameter around impurities which is extended in space to distances of the order of the core size . in this paper
we performed only an order - of - magnitude estimate suggesting that the rmt limit may be reached in moderately clean vortices ( with the number of impurities per vortex core much greater than one ) .
a more accurate self - consistent treatment of impurities in the vortex core is required for a quantitative study of the effects of disorder .
the author wishes to thank m. v. feigelman for suggesting this problem , discussions , and comments on the manuscript . at different stages of work ,
the author benefited from discussions with g. blatter , v. geshkenbein , d. gorokhov , r. heeb , l. ioffe , c. mudry , m. skvortsov , g. e. volovik , and m. zhitomirsky .
the author thanks swiss national foundation for financial support . p. w. brouwer , c. mudry , a. furusaki : phys .
* 84 * , 2913 ( 2000 ) [ cond - mat/9904200 ] ; + p. w. brouwer , a. furusaki , i. a. gruzberg , c. mudry : localization and delocalization in dirty superconducting wires , cond - mat/0002016 k. ueda , t. m. rice : ` heavy electron superconductors some consequences of the @xmath0-wave pairing ' . in : _ theory of heavy fermions and valence fluctuations _ , ed . by t.
kasuya , t. saso ( springer , berlin 1985 ) | in disordered vortices in @xmath0-wave superconductors the two new random - matrix ensembles may be realized : @xmath1 and @xmath2iii - odd ( of @xmath3 and @xmath4 matrices respectively ) .
we predict these ensembles from an explicit analysis of the symmetries of bogoliubov
degennes equations in three examples of vortices with different @xmath0-wave order parameters .
a characteristic feature of the novel symmetry classes is a quasiparticle level at zero energy .
class @xmath1 is realized when the time - reversal symmetry is broken , and class @xmath2iii - odd when the time - reversal symmetry is preserved .
we also suggest that the main contribution to disordering the vortex spectrum comes from the distortion of the order parameter around impurities . since wigner s modeling hamiltonians of complex nuclei by random matrices @xcite
, the random - matrix theory ( rmt ) has played an important role in studying mesoscopic systems . in many cases ,
a chaotic ( non - integrable ) mesoscopic system may be accurately described by rmt . the supersymmetric technique by efetov @xcite provides a microscopic explanation of the rmt approximation for disordered systems .
according to the rmt approximation , the only characteristics of the system affecting the eigenvalue correlations at small energy scales are its symmetries .
therefore , a problem arises of classifying symmetries of random - matrix ensembles .
it has been suggested by several authors @xcite that random - matrix ensembles may be classified as corresponding symmetric spaces .
the symmetric spaces may be divided into twelve infinite series reviewed in table 1 ( we split class @xmath2iii in two subclasses : @xmath2iii - even and @xmath2iii - odd ) @xcite . each of the symmetry classes may occur in one of the three forms : positive - curvature , negative - curvature and flat @xcite .
the corresponding jacobians in the matrix space are expressed in terms of trigonometric , hyperbolic , and polynomial functions respectively . in the present paper
we shall only discuss the rmt for hamiltonians forming a linear space and therefore described by zero - curvature ( flat ) versions of rmt .
cccccc .symmetric spaces and universality classes of random - matrix ensembles : ( * a * ) the three wigner - dyson classes @xmath5 ( unitary ) , @xmath5i ( orthogonal ) , @xmath5ii ( symplectic ) @xcite ; ( * b * ) chiral classes @xmath5iii ( unitary ) , @xmath6i ( orthogonal ) , @xmath7ii ( symplectic ) , with dimensions @xmath8 have @xmath9 zero modes @xcite ; ( * c * ) superconducting classes @xmath7 , @xmath2 , @xmath7i , @xmath2iii - even @xcite ; ( * d * ) @xmath0-wave vortex classes @xmath2iii - odd , @xmath1 have one zero mode [ cols="^ " , ] & dimension & rank & @xmath10 & @xmath11 + [ 10pt ] @xmath5 [ gue ] & su(@xmath12 ) & @xmath13 & @xmath14 & 2 & @xmath15 + @xmath5i [ goe ] & su(@xmath12)/so(@xmath12 ) & @xmath16 & @xmath14 & 1 & @xmath15 + @xmath5ii [ gse ] & su(@xmath17)/sp(@xmath12 ) & @xmath18 & @xmath14 & 4 & @xmath15 + @xmath5iii [ chgue ] & su(@xmath0+@xmath19)/s(u(@xmath0)@xmath20u(@xmath19 ) ) & @xmath21 & @xmath0 & 2 & @xmath22 + @xmath6i [ chgoe ] & so(@xmath0+@xmath19)/so(@xmath0)@xmath20so(@xmath19 ) & @xmath23 & @xmath0 & 1 & @xmath24 + @xmath7ii [ chgse ] & sp(@xmath0+@xmath19)/sp(@xmath0)@xmath20sp(@xmath19 ) & @xmath25 & @xmath0 & 4 & @xmath26 + @xmath7 & sp(@xmath12 ) & @xmath27 & @xmath12 & 2 & 2 + @xmath2 & so(@xmath17 ) & @xmath28 & @xmath12 & 2 & 0 + @xmath7i & sp(@xmath12)/u(@xmath12 ) & @xmath29 & @xmath12 & 1 & 1 + @xmath2iii - even & so(@xmath30)/u(@xmath17 ) & @xmath31 & @xmath12 & 4 & 1 + @xmath2iii - odd & so(@xmath30 + 2)/u(@xmath17 + 1 ) & @xmath32 & @xmath12 & 4 & 5 + @xmath1 & so(@xmath17 + 1 ) & @xmath27 & @xmath12 & 2 & 2 + [ table1 ] the simplest examples of the rmt symmetry classes are the three wigner - dyson classes : unitary , orthogonal and symplectic ( @xmath5 , @xmath5i , and @xmath5ii , respectively , in cartan s notation ) @xcite . in these classes ,
the energy level correlations are invariant under translations in energies ( at energy scales much smaller than the spectrum width ) , and the joint probability distribution of the energy levels @xmath33 is @xmath34 the parameter @xmath10 determines the strength of level repulsion and takes values 2 , 1 , and 4 for the unitary , orthogonal , and symplectic ensembles , respectively .
other symmetry classes appear when there exists an additional symmetry relating energies @xmath35 and @xmath36 . in the corresponding random - matrix ensembles , the levels
@xmath33 repel not only each other , but also their mirror images @xmath37 , and the jacobian has the form @xmath38 it is now characterized by the two parameters @xmath10 and @xmath11 ( the latter responsible for suppressing the density of states near @xmath39 ) . the three chiral classes ( @xmath5iii , @xmath6i , and @xmath7ii ) appear in systems with a chiral symmetry anticommuting with the hamiltonian @xcite . four more classes ( @xmath7 , @xmath2 , @xmath7i , and @xmath2iii ) describe mesoscopic superconducting systems ( with or without spin - rotational and time - reversal symmetries ) @xcite .
the present paper is devoted to the last two lines in table 1 : classes @xmath2iii - odd and @xmath1 .
we shall demonstrate that these two classes correspond to the symmetries of a @xmath0-wave vortex with or without time - reversal symmetry , respectively ( the symmetry class @xmath1 in @xmath0-wave vortices has also been predicted in @xcite ; a partial account of the present work appeared as a preprint @xcite ) .
below we consider three particular @xmath0-wave vortices and describe their symmetries : the single - quantum vortex in the @xmath5-phase with a fixed spin orientation of the order parameter , the half - quantum vortex in the @xmath5-phase with rotating orientation of the order parameter , and the spin ( fluxless ) vortex in the @xmath1-phase .
we explicitly demonstrate that in the first two examples the symmetries of the hamiltonian are of type @xmath1 , and in the last example of type @xmath2iii - odd .
then we briefly discuss microscopic requirements for the rmt limit and show that the main contribution to the level mixing comes from inhomogeneous suppression of the order parameter by impurities .
our estimates suggest that the rmt limit may be achieved in a moderately clean superconductor . |
the sheer volume of today s digital data has made _ distributed storage systems _ @xmath13dss@xmath14 not only massive in scale but also critical in importance . every day
, people knowingly or unknowingly connect to various private and public distributed storage systems , include large data centers ( such as the google data centers and amazon clouds ) and peer - to - peer storage systems ( such as oceanstore @xcite , total recall @xcite , and dhash++ @xcite ) . in a distributed storage system ,
a data file is stored at a distributed collection of storage devices / nodes in a network .
since any storage device is individually unreliable and subject to failure ( i.e. erasure ) , redundancy must be introduced to provide the much - needed system - level protection against data loss due to device / node failure .
the simplest form of redundancy is _
replication_. by storing @xmath15 identical copies of a file at @xmath15 distributed nodes , one copy per node , a @xmath15-replication system can guarantee the data availability as long as no more than @xmath16 nodes fail .
such systems are very easy to implement , but extremely inefficient in storage space utilization , incurring tremendous waste in devices and equipment , building space , and cost for powering and cooling .
more sophisticated systems employing _ erasure coding _
@xcite can expect to considerably improve the storage efficiency .
consider a file that is divided into @xmath12 equal - size fragments .
a judiciously - designed @xmath17 $ ] erasure ( systematic ) code can be employed to encode the @xmath12 data fragments ( terms _ systematic symbols _ in the coding jargon ) into @xmath11 fragments ( termed _ coded symbols _ ) stored in @xmath11 different nodes . if the @xmath0 $ ] code reaches the singleton bound such that the minimum hamming distance satisfies @xmath18 , then the code is _
maximum distance separable _
( mds ) and offers redundancy - reliability optimality . with an @xmath17 $ ] mds erasure code , the original file can be recovered from any set of @xmath12 encoded fragments , regardless of whether they are systematic or parity .
in other words , the system can tolerate up to @xmath19 concurrent device / node failure without jeopardizing the data availability . despite the huge potentials of mds erasure codes ,
however , practical application of these codes in massive storage networks have been difficult .
not only are simple ( i.e. requires very little computational complexity ) mds codes very difficult to construct , but data repair would in general require the access of @xmath12 other encoded fragments @xcite , causing considerable input / output ( i / o ) bandwidth that would pose huge challenges to a typical storage network . is first split into five equal packets @xmath20 and then is encoded into @xmath21 packets , using a @xmath22 linear code .
these @xmath21 encoded packets are stored at @xmath21 nodes @xmath23 , which are divided into three groups @xmath24 , @xmath25 and @xmath26 .
each group can perform local repair of up to two node - failures .
for example , if node @xmath27 fails , it can be repaired by any two packets among @xmath28 and @xmath29 .
moreover , the entire file @xmath30 can be recovered by five packets from any five nodes @xmath31 which intersect each group with at most two packets .
for example , @xmath30 can be recovered from five packets stored at @xmath32 and @xmath33 . ]
motivated by the desire to reduce repair cost in the design of erasure codes for distributed storage systems , gopalan _ et al . _
@xcite introduced the interesting notion of _ symbol locality _ in linear codes .
the @xmath34th coded symbol of an @xmath17 $ ] linear code @xmath35 is said to have locality @xmath36 if it can be recovered by accessing at most @xmath9 other symbols in @xmath37 .
the concept was further generalized to @xmath1 locality by prakash _
@xcite , to address the situation of multiple device failures .
according to @xcite , the @xmath34th code symbol @xmath38 , in an @xmath17 $ ] linear code @xmath37 is said to have locality @xmath1 if there exists an index set @xmath39 $ ] containing @xmath34 such that @xmath40 and each symbol @xmath41 , can be reconstructed by any @xmath42 symbols in @xmath43 , where @xmath44 is an integer .
thus , when @xmath45 , the notion of locality in @xcite reduces to the notion of locality in @xcite .
two cases of @xmath1 codes are introduced in the literature : an @xmath46 code is a systematic linear code whose _
information symbols _ all have locality @xmath1 ; and an @xmath2 code is a linear code all of whose _ symbols _ have locality @xmath1 .
hence , an @xmath2 code is also referred to as having _ all - symbol locality _ @xmath1 , and an @xmath46 code is also referred to as having _ information locality _ @xmath1 . a symbol with @xmath1 locality
given that at the most @xmath47 symbols are erased can be deduced by reading at most @xmath9 other unerased symbols .
clearly , codes with a low symbol locality , such as @xmath48 , impose a low i / o bandwidth and repair cost in a distributed storage system . in a dss system
, one can use `` group '' to describe storage nodes situated in the same physical location which enjoy a higher communication bandwidth and a shorter communication distance than storage nodes belonging to different groups . in the case of node failure , a _ locally repairable code _ makes it possible to efficiently recover data stored in the failed node by downloading information from nodes in the same group ( or in a minimal number of other groups ) .
[ fig - dss ] provides a simple example of how an @xmath2 code is used to construct a distributed storage system . in this example , @xmath37 is a @xmath22 linear code of length @xmath21 and dimension @xmath49 .
note that a failed node can be reconstructed by accessing only two other existing nodes , while it takes five existing nodes to repair a failed node if a @xmath50 $ ] mds code is used .
locality was identified as a repair cost metric for distributed storage systems independently by oggier _
@xcite , gopalan _ et al .
_ @xcite and papailiopoulos _ et al .
_ @xcite using different terms .
in @xcite , gopalan _ et al .
_ introduced the concept of symbol locality of linear codes and established a tight bound for the redundancy in terms of the message length , the distance , and the locality of information coordinates .
a generalized concept , i.e. , @xmath1 locality , was addressed by prakash _
it was proved in @xcite that the minimum distance @xmath51 of an @xmath46 linear code @xmath37 is upper bounded by @xmath52 where @xmath11 and @xmath12 are the length and dimension of @xmath37 respectively .
it was also proved that a class of codes known as pyramid codes @xcite achieve this bound . since
an @xmath2 code is also an @xmath46 code , ( [ eqn:1 ] ) also presents an upper bound for the minimum distance of @xmath2 codes .
locality of general codes ( linear or nonlinear ) and bounds on the minimum distance for a given locality were presented in parallel and subsequent works @xcite .
an @xmath2 code ( systematic or not ) is also termed a _ locally repairable code ( lrc ) _ , and @xmath2 codes that achieve the minimum distance bound are called _
optimal_. it was proved in @xcite that there exists optimal locally repairable linear codes when @xmath53 and @xmath54 . under the condition that @xmath53 , a construction method of optimal locally repairable vector codes was proposed in @xcite , where maximal rank distance ( mrd ) codes were used along with mds array codes .
for the special case of @xmath4 , tamo _ et al . _
@xcite proposed an explicit construction of optimal lrcs when @xmath55 or @xmath56 except for the special case that @xmath57 , no results are known about whether there exists optimal @xmath2 code when @xmath58 . up to now , designing lrcs with optimal distance
remains an intriguing open problem for most coding parameters @xmath59 and @xmath10 .
since large fields involve rather complicated and expensive computation , a related interesting open problem asks how to limit the design ( of optimal lrcs ) over relatively smaller fields . in this paper , we investigate the structure properties and the construction of optimal @xmath2 linear codes of length @xmath11 and dimension @xmath12 .
a simple property of optimal @xmath2 linear codes is proved in lemma [ low - bound ] , which shows that @xmath60 for any optimal @xmath2 linear code .
hence we impose this condition of @xmath60 throughout our discussion of optimal @xmath2 codes .
the main results of this paper include : \(1 ) we prove a structure theorem for the optimal @xmath2 linear codes for @xmath61 .
this structure theorem indicates that it is possible for optimal @xmath2 linear codes , a sub - class of optimal @xmath46 linear code , to have a simpler structure than otherwise .
\(2 ) we prove that there exist no optimal @xmath2 linear codes for @xmath62 or @xmath63 where @xmath64 and @xmath65 such that @xmath66 and @xmath67 ( theorems [ non - exst ] and [ non - exst-1 ] ) .
\(3 ) we propose a deterministic algorithm for constructing optimal @xmath2 linear codes over any field of size @xmath68 when @xmath69 or @xmath70 where @xmath64 and @xmath65 such that @xmath66 and @xmath67 ( theorem [ opt - ext-1 ] and [ opt - ext-2 ] ) .
\(4 ) we propose another deterministic algorithm for constructing optimal @xmath2 linear codes over any field of size @xmath68 when @xmath71 or @xmath72 where @xmath64 and @xmath65 such that @xmath66 and @xmath67 ( theorem [ opt - ext-3 ] and [ opt - ext-4 ] ) . a summary of our results
is given in fig [ sumy ] .
note that if none of the conditions in ( [ eqn:2])-([eqn:5 ] ) holds , it then follows that @xmath73 in that case , if condition ( [ eqn:6 ] ) does not hold , we have @xmath74 ; and if condition ( [ eqn:7 ] ) does not hold , we have @xmath75 , i.e. , @xmath76 . hence ,
if , neither condition ( [ eqn:6 ] ) nor condition ( [ eqn:7 ] ) holds ( in addition to ( [ eqn:2])-([eqn:5 ] ) ) , then one of the following two conditions must be satisfied : @xmath77 or @xmath78 in other words , if none of the conditions ( [ eqn:2])-([eqn:7 ] ) holds , then either ( [ eqn:8 ] ) or ( [ eqn:9 ] ) will hold . from our existence proof and/or constructive results ,
the existence of optimal @xmath2 linear code is not known only for a limited scope with parameters described by ( [ eqn:8 ] ) and ( [ eqn:9 ] ) .
the remainder of the paper is organized as follows . in section
, we present the notions used in the paper as well as some preliminary results about @xmath2 linear codes . in section
, we investigate the structure of optimal @xmath2 linear codes when @xmath61 ( should they exist ) . in section , we consider the non - existence conditions for optimal @xmath2 linear codes under conditions ( [ eqn:2 ] ) and ( [ eqn:3 ] ) .
a construction of optimal @xmath2 linear codes for conditions ( [ eqn:4 ] ) and ( [ eqn:5 ] ) is presented in section , and a construction of optimal @xmath2 linear codes for conditions ( [ eqn:6 ] ) and ( [ eqn:7 ] ) is presented in section .
finally , we conclude the paper in section .
for two positive integers @xmath79 and @xmath80 , we denote @xmath81=\{t_1,t_1 + 1,\cdots , t_2\}$ ] and @xmath82=\{1,2,\cdots , t_2\}$ ] . for any set @xmath83
, the size @xmath13cardinality@xmath14 of @xmath83 is denoted by @xmath84 .
if @xmath85 is a subset of @xmath83 and @xmath86 , then we say that @xmath85 is an @xmath9-subset of @xmath83 .
let @xmath87 be the @xmath12-dimensional vector space over the @xmath88-ary field @xmath89 .
for any subset @xmath90 , we use @xmath91 to denote the subspace of @xmath87 spanned by @xmath92 . in the sequel ,
whenever we speak of an @xmath2 or @xmath46 code , we will by default assume it is an @xmath0 $ ] linear code ( i.e. , its length , dimension and minimum distance are @xmath93 and @xmath51 respectively ) .
suppose @xmath37 is an @xmath0 $ ] linear code over @xmath94 , and @xmath95 is a generating matrix of @xmath96 , where @xmath97,$ ] is the @xmath34th column of @xmath98 .
we denote by @xmath99 the collection of columns of @xmath98 .
it is well known that the distance property is captured by the following condition ( e.g. @xcite ) .
[ fact ] an @xmath17 $ ] code @xmath37 has a minimum distance @xmath51 , if and only if @xmath100 for every @xmath101 having @xmath102 .
equivalently , @xmath103 for every @xmath104 of size @xmath105 . for any subset
@xmath106 $ ] , let @xmath107 denote the punctured code of @xmath37 associated with the coordinate set @xmath83 .
that is , @xmath107 is obtained from @xmath37 by deleting all symbols @xmath108\backslash s$ ] , in each codeword @xmath109 .
[ def - locality ] suppose @xmath110 and @xmath44 .
the @xmath34th code symbol @xmath111 , in an @xmath0 $ ] linear code @xmath37 is said to have locality @xmath1 if there exists a subset @xmath39 $ ] such that * @xmath112 ; * the minimum distance of the punctured code @xmath113 is at least @xmath10 .
[ rem - loty ] let @xmath95 be a generating matrix of @xmath37 . by lemma [ fact ] , it is easy to see that the second condition in definition [ def - locality ] is equivalent to the following condition * @xmath114 for any subset @xmath115 of size @xmath116 , where @xmath117 ; moreover , by conditions ( 1 ) and ( 2@xmath118 ) , we have @xmath119 that is , @xmath120 and @xmath121 of size @xmath116 , @xmath122 is an @xmath89-linear combination of @xmath123 .
this means that the symbol @xmath124 can be reconstructed by the @xmath42 symbols in @xmath125 .
an @xmath2 code @xmath37 is said to be _ optimal _ if the minimum distance @xmath51 of @xmath37 achieves the bound in ( [ eqn:1 ] ) .
the following remark follows naturally from definition [ def - locality ] and remark [ rem - loty ] .
[ rem - locality ] if @xmath37 is an @xmath2 code and @xmath95 is a generating matrix of @xmath37 , then we can always find a collection @xmath126 , where @xmath39 , i=1,\cdots , t$ ] , such that * @xmath127 ; * @xmath128 $ ] and @xmath115 of size @xmath116 , where @xmath117 ; * @xmath129}s_i=[n]$ ] and @xmath129\backslash\{j\ } } s_i\neq[n],\forall j\in[t]$ ] .
we call the set @xmath130 an @xmath1-_cover set _ of @xmath37 .
the following lemma presents a simple property of @xmath2 codes .
[ low - bound ] an @xmath2 code @xmath37 satisfies * the minimum distance @xmath131 . *
if @xmath37 is an optimal @xmath2 code , then @xmath132 .
\1 ) let @xmath130 be an @xmath1-cover set of @xmath37 . for any @xmath133 , since @xmath129}s_i=[n]$
] , there is an @xmath134 $ ] such that the punctured codeword @xmath135 is nonzero in @xmath136 . by the second condition of definition [ def - locality ] , the hamming weight of @xmath135 is at least @xmath10 .
thus , the hamming weight of @xmath137 is at least @xmath10 . since @xmath133 is arbitrary , the minimum distance @xmath131 .
\2 ) since @xmath37 is an optimal @xmath2 code , from the minimum distance bound in ( [ eqn:1 ] ) , @xmath138 from claim 1 ) , @xmath131 ; which leads to @xmath139 hence , @xmath140 which implies that @xmath132 .
in this section , we prove a structure theorem for optimal @xmath2 codes under the condition of @xmath61 . throughout this section , we assume that @xmath37 is an @xmath2 code over the field @xmath89
, @xmath141 is an @xmath1-cover set of @xmath96 , where @xmath39 $ ] , @xmath142 , and @xmath95 is a generating matrix of @xmath37 .
we denote @xmath99 and @xmath143 and @xmath144 are viewed as vectors of @xmath87 , it is possible for @xmath145 where @xmath146 .
however , when treating them as two different columns of @xmath98 , we shall view @xmath147 and @xmath144 as two separate elements in @xmath148 ( even though they may be identical ) . ] .
then for any @xmath149 $ ] , we have @xmath150 and by remark [ rem - locality ] , we get @xmath151}\mathcal g_i=\mathcal g ~\text{and } ~ \cup_{i\in[t]\backslash\{j\}}\mathcal g_i\neq\mathcal g , \forall j\in[t].\label{eqn:11}\end{aligned}\ ] ] we first give some lemmas to help prove our main results . [ rank - sum ] consider three sets @xmath152 . if @xmath153 is a subset of @xmath92 satisfies : @xmath154 then @xmath155 since @xmath156 and @xmath157 , we have @xmath158 therefore , @xmath159 .
[ cup - rank ] suppose @xmath160 $ ] such that @xmath161 , @xmath162 .
then @xmath163 we prove this lemma by induction . from remark [ rem - loty ] , @xmath164 .
hence the claim holds for @xmath165 .
now consider @xmath166 .
we assume that the claim holds for @xmath167 , i.e. , @xmath168 we shall prove that the claim is true for @xmath169 .
first , we point out that @xmath170 in fact , if @xmath171 , then @xmath172 . from condition ( 2 ) of remark [ rem - locality ] , @xmath173 , which presents a contradiction to the assumption that @xmath174 .
thus , @xmath175 let @xmath176 and @xmath177 .
let @xmath178 be a fixed @xmath47-subset of @xmath179 and @xmath180 . from condition
( 2 ) of remark [ rem - locality ] , @xmath181 then , from lemma [ rank - sum ] , we get @xmath182 i.e. , @xmath183 clearly , @xmath184 is a disjoint union of @xmath185 and @xmath186 .
hence , @xmath187 and from ( [ eqn:10 ] ) , we get @xmath188 combining ( [ eqn:12])-([eqn:14 ] ) , we have @xmath189 which completes the proof .
[ not - ctn ] suppose @xmath37 is an optimal @xmath2 code
. then * @xmath190 . *
if @xmath191 $ ] and @xmath192 , then @xmath193 and @xmath194\backslash j$ ] .
* if @xmath191 $ ] and @xmath195 , then @xmath196 and @xmath197 .
\1 ) ( proof by contradiction ) suppose @xmath198 .
then from remark [ rem - locality ] , @xmath199 hence , @xmath200}s_i|\\ & \leq&t(r+\delta-1)\\ & \leq&(\lceil\frac{n}{r+\delta-1}\rceil-1)(r+\delta-1)\\&<&n\end{aligned}\ ] ] which presents a contradiction .
hence , it must hold that @xmath201 . moreover , from claim 2 ) of lemma [ low - bound ] , @xmath132 .
@xmath202 \2 ) from remark [ rem - loty ] , @xmath203 $ ] . hence ,
if @xmath192 , then @xmath204 i.e. , @xmath205 now , suppose @xmath206 , and we will see a contradiction results .
first , we can find a subset @xmath207 such that @xmath208 and @xmath209 for any proper subset @xmath210 of @xmath211 .
in particular , we have @xmath212 note that @xmath213 next , we can find a sequence @xmath214 such that @xmath215 and @xmath216 . in particular , @xmath217 .
therefore , there exists a @xmath218 such that @xmath219 .
denote @xmath220 . then @xmath221 and @xmath222 from lemma [ cup - rank ] , @xmath223 then by equations ( [ eqn:15 ] ) and ( [ eqn:16 ] ) , @xmath224 since @xmath225\backslash j$ ] , @xmath226 .
moreover , since @xmath208 and @xmath216 , so @xmath227 . from equation ( [ eqn:11 ] )
, we have @xmath228 .
then , from equation ( [ eqn:17 ] ) , we get @xmath229 since we assumed @xmath230 , then @xmath231 . by lemma
[ fact ] , we have @xmath232 which contradicts the assumption that @xmath37 is an optimal @xmath2 code .
hence , it must be that @xmath233 .
code @xmath37 , we obtain a subset @xmath101 such that @xmath234 and rank@xmath235 .
then by lemma [ fact ] , the minimum distance of @xmath37 is @xmath236 , which also provides a proof of the minimum distance bound in ( [ eqn:1 ] ) . ]
\3 ) suppose @xmath237 , where @xmath238 .
by claim 2 ) , @xmath239 first , we have @xmath196 . otherwise , as in the proof of claim 2 ) , we can find a sequence @xmath240 , @xmath241 and a set @xmath242 such that @xmath243 by lemma [ fact ] , @xmath244 which contradicts the assumption that @xmath37 is an optimal @xmath2 code .
therefore , we have @xmath245 .
now , by lemma [ cup - rank ] , @xmath246 this completes the proof .
we now present our main theorem of this section .
[ stru - opt ] suppose @xmath37 is an optimal @xmath2 linear code .
if @xmath61 and @xmath48 , then the following conditions hold : * @xmath247 are mutually disjoint ; * @xmath248 $ ] , and the punctured code @xmath113 is an @xmath249 $ ] mds code . in particular , we have @xmath250 . since @xmath61 and @xmath48 , then @xmath251 for some @xmath252 .
by 1 ) of lemma [ not - ctn ] , @xmath253 .
let @xmath254 $ ] be arbitrarily chosen .
let @xmath255 be an @xmath169-subset of @xmath256 $ ] such that @xmath257 .
then by 3 ) of lemma [ not - ctn ] , @xmath258 and @xmath259 since @xmath260 and by remark [ rem - locality ] , @xmath261 , then equations ( [ eqn:18 ] ) and ( [ eqn:19 ] ) imply that @xmath262 , @xmath263 , and @xmath264 are mutually disjoint . in particular , @xmath265 , @xmath266 and @xmath267 .
since @xmath268 and @xmath269 are arbitrarily chosen , we have proved that @xmath270 , @xmath263 , and @xmath264 are mutually disjoint .
hence , @xmath250 .
moreover , by lemma [ fact ] and remark [ rem - loty ] , @xmath113 is an @xmath249 $ ] mds code . in @xcite
, it was proved that if @xmath37 is an optimal @xmath46 code , then there exists a collection @xmath271 which has the same properties in theorem [ stru - opt ] , where @xmath272 is a properly - defined value .
thus , theorem [ stru - opt ] shows that as a sub - class of optimal @xmath46 codes , optimal @xmath2 codes tend to have a simpler structure than otherwise .
in this section , we derive two sets of conditions under which there exists no optimal @xmath2 linear codes . from the minimum distance bound in ( [ eqn:1 ] ) , we know that when @xmath273 , optimal @xmath2 linear codes are exactly mds codes .
hence , in this section , we focus on the case of @xmath48 . the first result is obtained directly from theorem [ stru - opt ] .
[ non - exst ] if @xmath58 and @xmath274 , then there exist no optimal @xmath2 linear codes .
if @xmath37 is an optimal @xmath2 linear code and @xmath61 , then by theorem [ stru - opt ] , @xmath53 , which contradicts the condition that @xmath58 . hence , there exist no optimal @xmath2 linear codes when @xmath58 and @xmath61 .
when @xmath58 and @xmath275 , we provide in the below a set of conditions under which no optimal @xmath2 code exists .
[ non - exst-1 ] suppose @xmath64 and @xmath65 , where @xmath67 and @xmath66 .
if @xmath276 and @xmath277 , then there exist no optimal @xmath2 codes .
we prove this theorem by contradiction .
suppose @xmath37 is an optimal @xmath2 code over the field @xmath89 and @xmath130 is an @xmath1-cover set of @xmath37 .
then by claim 1 ) of lemma [ not - ctn ] , we have @xmath278 moreover , by 3 ) of lemma [ not - ctn ] , for any @xmath279-subset @xmath255 of @xmath256 $ ] , @xmath280 for each @xmath134 $ ] , if @xmath281 , let @xmath282 $ ] be such that @xmath283 and @xmath284 ; if @xmath285 , let @xmath286 .
then clearly , @xmath287}t_i=\cup_{i\in[t]}s_i=[n]\ ] ] and for any @xmath279-subset @xmath255 of @xmath256 $ ] , @xmath288 let @xmath289 be a @xmath290 matrix such that @xmath291 if @xmath292 , and @xmath293 otherwise .
for each @xmath294 $ ] , let @xmath295;m_{i , j}=1\}.\ ] ] then @xmath296 is the number of @xmath297)$ ] satisfying @xmath292 , and this number equals the number of @xmath298s in the @xmath299th column of @xmath300 . since @xmath129}t_i=[n]$ ] , then @xmath301 $ ] . on the other hand , by the construction of @xmath300 , for each @xmath134 $ ] , @xmath302;m_{i , j}=1\}.$ ] thus , the number of the @xmath298s in each row of @xmath300 is @xmath303 .
it then follows that the total number of the @xmath298s in @xmath300 is @xmath304 combining ( [ eqn:20 ] ) and ( [ eqn:22 ] ) , we have @xmath305 since @xmath276 , then @xmath306 hence from ( [ eqn:23 ] ) , we have @xmath307 let @xmath308;|a_j|>1\}$ ]
. from ( [ eqn:24 ] ) , @xmath309 and @xmath310 without loss of generality , assume @xmath311 .
since @xmath312 , we can find a number @xmath313 such that @xmath314 and @xmath315 .
this means that we can find a subset @xmath316 such that @xmath317 and @xmath318 also note that @xmath319 because otherwise , @xmath320 , which contradicts ( [ eqn:25 ] ) .
let @xmath321 . then from ( [ eqn:25 ] ) , @xmath322 since @xmath277 , then @xmath323 , we get @xmath324 let @xmath255 be a @xmath279-subset of @xmath256 $ ] such that @xmath325 . by the construction of @xmath300 and @xmath326 , for each @xmath327 , there are at least @xmath296 subsets in @xmath328 containing @xmath299 , and there are at least @xmath329 subsets in @xmath328 containing @xmath330 .
hence , @xmath331 ) and ( [ eqn:27 ] ) , we have @xmath332 which contradicts ( [ eqn:21 ] ) . thus , we can conclude that there exist no optimal @xmath2 linear codes when @xmath276 and @xmath277 .
_ example : _ we now provide an example to help illustrate the method used in the proof of theorem [ non - exst-1 ] .
let @xmath333 and @xmath334 .
suppose @xmath335 , @xmath336 , @xmath337 and @xmath338 .
following the notations in the proof of theorem [ non - exst-1 ] , we have therefore , @xmath339 , and @xmath340 .
note that @xmath341 .
let @xmath342 and @xmath343 ; then @xmath344 .
let @xmath345 , then @xmath346 hence , @xmath347 .
( see the illustration of @xmath300 below . ) more generally , in this example , for any @xmath348 and @xmath349 such that @xmath350 and @xmath351=\{1,\cdots,13\}$ ] , we can always find a @xmath191 $ ] such that @xmath352 . in general , since @xmath66 , then @xmath353 .
if @xmath354 , then we have @xmath355 . hence , when @xmath356 and @xmath354 , then by theorem [ non - exst-1 ] , there exist no optimal @xmath2 codes .
in this section , we propose a deterministic algorithm for constructing optimal @xmath2 linear codes over the field of size @xmath68 , when @xmath53 or @xmath357 , where @xmath64 and @xmath65 satisfying @xmath66 and @xmath67 .
recall that when @xmath53 , it was proved in @xcite that optimal @xmath2 linear codes exist over the field of size @xmath54 .
note that our method requires a much smaller field than what s shown in @xcite , and hence it also has a lower complexity for implementation . to present our method
, we will use the following definitions and notations , most of which follow from @xcite .
[ core ] let @xmath126 be a partition of @xmath358 $ ] and @xmath359 $ ] .
a subset @xmath106 $ ] is called an @xmath360-_core _ if @xmath361 $ ] .
if @xmath83 is an @xmath360-core and @xmath362 , then @xmath83 is called an @xmath363-_core_. clearly , if @xmath106 $ ] is an @xmath360-core and @xmath364 , then @xmath365 is also an @xmath360-core . in particular ,
if @xmath106 $ ] is an @xmath360-core and @xmath365 is a @xmath12-subset of @xmath83 , then @xmath365 is an @xmath366-core . before presenting our construction method
, we first give a lemma , which will take an important role in our discussion .
[ sub - space ] let @xmath367 and @xmath92 be @xmath368 subspaces of @xmath369 and @xmath370 $ ] . if @xmath371 , then @xmath372 .
we prove this lemma by induction . clearly , the claim is true when @xmath165 .
now , we suppose that the claim is true for @xmath167 , i.e. , @xmath373 then there exists an @xmath374 such that @xmath375 .
if @xmath376 , then @xmath377 and @xmath372 .
so we assume @xmath378 .
since @xmath379 , there exists a @xmath380 such that @xmath381 .
then for any @xmath382 and @xmath383 , @xmath384 @xmath13otherwise , @xmath385 , which contradicts to the assumption that @xmath386 since @xmath371 , we can pick a subset @xmath387 .
then @xmath388otherwise , by the pigeonhole principle , there is a subset @xmath389 and a @xmath390 such that @xmath391 , which contradicts to the proven result that for any @xmath382 and @xmath383 , @xmath392 without loss of generality , assume @xmath393 .
note that @xmath378 and @xmath381 , then @xmath394 .
hence , @xmath395 . on the other hand , since @xmath396
, then @xmath397 .
so @xmath398 , which completes the proof .
we present our construction method in the following theorem .
[ opt - suf-1 ] let @xmath130 be a partition of @xmath358 $ ] and @xmath359 $ ] .
suppose @xmath399 and for any @xmath279-subset @xmath255 of @xmath256 $ ] , @xmath400 .
if @xmath68 , then there exists an optimal @xmath2 linear code over @xmath89 . for each @xmath134 $ ] , let @xmath401 be an @xmath402-subset of @xmath403 .
let @xmath404}u_i$ ] and @xmath405 .
let @xmath255 be a @xmath279-subset of @xmath256 $ ] . since @xmath406 , from the assumptions of this theorem ,
@xmath407 the construction of an optimal @xmath2 code consists of the following two steps : _ step 1 _ : construct an @xmath408 $ ] mds code @xmath409 over @xmath89 . since @xmath410 , such an mds code exists over @xmath89 .
let @xmath411 be a generating matrix of @xmath409 .
we index the columns of @xmath411 by @xmath412 , i.e. , @xmath413 , where @xmath414 is a column of @xmath411 for each @xmath415 .
_ step 2 _ : extend @xmath409 to an optimal @xmath2 code @xmath37 over @xmath89 .
this can be achieved by the following algorithm .
* algorithm 1 : * \1 .
let @xmath416 .
@xmath34 runs from @xmath298 to @xmath417 .
\3 . while @xmath418 : \4 .
pick a @xmath419 and let @xmath420 be such that for any @xmath363-core @xmath421 , @xmath422 is linearly independent .
\5 .
@xmath423 .
let @xmath37 be the linear code generated by the matrix @xmath424 @xmath425 . to complete the proof of theorem [ opt - suf-1 ]
, we need to prove three claims : in claim 1 and claim 2 below we show that the code @xmath426 output by algorithm 1 is indeed an optimal @xmath2 linear code over @xmath89 ; in claim 3 , we prove that the vector @xmath427 described in line 4 of algorithm 1 can always be found , hence the algorithm does terminate successfully .
+ * claim 1 : * the code @xmath426 output by algorithm 1 is an @xmath2 linear code over @xmath89 . by definition [ def - locality ] and remark [ rem - loty ]
, we aim to show that for every @xmath428 $ ] and for every subset @xmath429 with @xmath116 , it holds that @xmath430 since in line 4 of algorithm 1 , we choose @xmath431 , we have @xmath432 by induction , @xmath433 suppose @xmath134 $ ] and @xmath115 such that @xmath116
. then @xmath434 .
since @xmath399 , we can find a @xmath279-subset @xmath210 of @xmath256 $ ] such that @xmath435 . for each @xmath436 ,
let @xmath437 be an @xmath438-subset of @xmath439 such that @xmath440 .
clearly , @xmath441 is an @xmath360-core . from the assumption of this lemma , @xmath442
hence @xmath443 let @xmath83 be a @xmath12-subset of @xmath441 such that @xmath444 , then @xmath83 is an @xmath363-core .
therefore , @xmath445 is linearly independent , which in turn implies that @xmath446 is also linearly independent .
therefore , @xmath447 ) .
+ * claim 2 : * the code @xmath426 output by algorithm 1 has minimum distance achieving the upper bound ( [ eqn:1 ] ) , and hence is an optimal @xmath2 linear code . according to lemma [ fact ] and ( [ eqn:1 ] )
, it suffices to prove that for any subset @xmath448 $ ] of size @xmath449 , @xmath450 let @xmath451;|t\cap s_j|\geq|s_j|-\delta+1\}.\ ] ] for each @xmath452 , let @xmath437 be an @xmath438-subset of @xmath453 ; for each @xmath454\backslash j$ ] , let @xmath455 .
then @xmath456}w_j$ ] is an @xmath360-core .
we consider the following two cases : case 1 : @xmath457 . without loss of generality , assume that @xmath195 , then pick a @xmath279-subset @xmath211 of @xmath255 , and replace @xmath255 by @xmath211 in our discussion . ] .
since @xmath458 , then @xmath459}w_j|\geq|\cup_{j\in j}w_j|\geq k.\ ] ] case 2 : @xmath192 . in that case , @xmath460}w_j|\geq |t|-|j|(\delta-1)\geq in both cases , @xmath461}w_j|\geq k$ ] .
let @xmath83 be a @xmath12-subset of @xmath462 , then @xmath83 is an @xmath366-core .
therefore , @xmath445 are linearly independent and @xmath463 from equation ( [ eqn:1 ] ) and lemma [ fact ] , we get @xmath464 where @xmath51 is the minimum distance of @xmath37 .
thus , @xmath37 is an optimal @xmath2 code .
+ * claim 3 : * the vector @xmath427 in line 4 of algorithm 1 can always be found .
the proof of this claim is based on a classical technique in network coding @xmath13e.g .
, @xcite@xmath14 .
since @xmath413 is a generating matrix of the mds code @xmath409 , then for any @xmath363-core @xmath465 , @xmath445 is linearly independent . by induction
, we can assume that for any @xmath366-core @xmath466 , @xmath445 are linearly independent .
let @xmath467 be the set of all @xmath468 such that @xmath469 is an @xmath363-core . by definition [ core ] , for any @xmath470 , @xmath471 @xmath472\backslash\{i\},\ ] ] and @xmath473 note that @xmath474 hence @xmath475 thus , there is an @xmath476 . since @xmath477 are mutually disjoint , @xmath478\backslash\{i\}$ ] .
therefore , @xmath479 then @xmath480 is an @xmath363-core . by assumption , @xmath481 is linearly independent .
hence @xmath482 and @xmath483 since @xmath484 , by lemma [ sub - space ] , @xmath485 let @xmath427 be a vector in @xmath486 . then for any @xmath470
, @xmath487 are linearly independent .
suppose @xmath488 is an @xmath366-core . if @xmath489 , then @xmath466 and by assumption , @xmath445 is linearly independent . if @xmath490 , then @xmath491 and by the selection of @xmath427 , @xmath445 is linearly independent .
hence we always have that @xmath445 is linearly independent .
thus , the vector @xmath427 satisfies the requirement of algorithm 1 . from the proof of theorem [ opt - suf-1 ]
, we can see that @xmath130 is in fact an @xmath1-cover set of the code @xmath37 , where @xmath37 is the output of algorithm 1 .
the following example demonstrates how does algorithm 1 work .
_ example : _ we now construct an optimal @xmath2 linear code with @xmath492 and @xmath493 .
let @xmath494 and @xmath495 .
let @xmath496 and @xmath497 .
our construct involves the following two steps .
step 1 : construct a @xmath498 $ ] mds code , where @xmath499 .
let @xmath500 be a generating matrix of such code .
step 2 : extend @xmath500 to a matrix @xmath501 such that @xmath98 is a generating matrix of an optimal @xmath502 linear code .
it remains to determine @xmath503 and @xmath504 via two iterations .
1 . @xmath505 : @xmath506 and @xmath507 .
we can verify that @xmath508 , @xmath509 , @xmath510 and @xmath511 are all subsets of @xmath512 which is an @xmath363-core and contains the index @xmath513 .
let @xmath514 , @xmath515 , @xmath516 .
since @xmath500 generates an mds code , then @xmath517 and @xmath518 are linearly independent .
so @xmath519 .
similarly , @xmath520 . by lemma [ sub - space ] , if @xmath521 , then @xmath522 .
note that @xmath523
. therefore , let @xmath524 then for any @xmath363-core @xmath525 , @xmath526 is linearly independent .
@xmath527 : @xmath528 and @xmath529 .
similarly , we can verify that @xmath530 , @xmath531 , @xmath532 , @xmath533 and @xmath534 are all subsets which is an @xmath363-core and contains the index @xmath535 .
let @xmath536 , @xmath537 , @xmath538 , @xmath515 , @xmath539 . clearly , @xmath540 .
by lemma [ sub - space ] , if @xmath541 , then @xmath542 . as @xmath543 , let @xmath544 then for any @xmath363-core @xmath83 , @xmath526 is linearly independent .
thus , we can obtain a matrix @xmath501 such that for any @xmath363-core @xmath83 , @xmath526 is linearly independent .
let @xmath37 be the linear code generated by @xmath98 .
then @xmath37 is an optimal @xmath502 linear code .
we can in fact employ a smaller field than @xmath545 .
the following is a generating matrix of an optimal @xmath502 linear code : @xmath546 over the field @xmath547 , where @xmath548 . in the rest of this section
, we shall use theorem [ opt - suf-1 ] to prove that optimal @xmath2 linear codes exist over a field of size @xmath68 when @xmath53 or @xmath549 , where @xmath64 and @xmath65 satisfying @xmath66 and @xmath67 . by claim 2 ) of lemma
[ low - bound ] , @xmath132 is a necessary condition for the existence of optimal @xmath2 linear codes .
for this reason , we assume @xmath132 holds in both cases .
[ opt - ext-1 ] suppose @xmath53 .
if @xmath68 , then there exists an optimal @xmath2 linear code over @xmath89 .
let @xmath550 .
note that we have assumed that @xmath132 . then @xmath551 let @xmath552 be a partition of @xmath553 such that @xmath554 . for any @xmath191 $ ] of size @xmath195 , @xmath555 by theorem [ opt - suf-1 ] , if @xmath68 , then there exists an optimal @xmath2 code over @xmath89 .
if @xmath53 and @xmath556 , then following a similar line of proof in @xcite , we can show that @xmath557 .
under these two conditions , it was proved in @xcite that there exists an optimal @xmath2 code over the field @xmath89 of size @xmath54 .
our method requires a field of size only @xmath558 , which is at the largest a fraction @xmath559 of @xmath560 .
[ opt - ext-2 ] suppose @xmath64 and @xmath65 , where @xmath67 and @xmath66 .
suppose @xmath549 and @xmath131 .
if @xmath68 , then there exists an optimal @xmath2 linear code over @xmath89 .
let @xmath561 . since we have assumed that @xmath132 , we get @xmath562 note that @xmath563 .
let @xmath564 be a partition of @xmath565 and @xmath566 $ ] . for any @xmath191
$ ] of size @xmath195 , we have the following two cases : case 1 : @xmath567 . then @xmath568 case 2 : @xmath569 . since @xmath549 , then @xmath570 hence , for any @xmath279-subset @xmath255 of @xmath256 $ ] , @xmath197 . by theorem [ opt - suf-1 ] , if @xmath68 , there exists an optimal @xmath2 code over @xmath89 . when @xmath4 , the conditions of theorem [ opt - ext-1 ] and theorem [ opt - ext-2 ] become @xmath571 and @xmath57 respectively . for this special case , tamo _
@xcite introduced a different construction method which is very easy to implement .
however , the method in @xcite requires the field size @xmath572 , which is larger than the field size @xmath573 of our method .
in this section , we present yet another method for constructing optimal @xmath2 codes .
this constructive method also points out two other sets of coding parameters where optimal @xmath2 codes exist . as the method in section , this method construct an optimal @xmath2 code which has a given set @xmath574 as its @xmath1-cover set .
the difference is that the set @xmath574 used by this method has a more complicated structure .
we again borrow the notion of _ core _ from @xcite .
[ frame ] let @xmath130 be a collection of @xmath575-subsets of @xmath358 $ ] , @xmath576 be a partition of @xmath256 $ ] and @xmath577 $ ] .
we say that @xmath574 is an @xmath578-_frame _ over the set @xmath358 $ ] , if the following two conditions are satisfied : * for each @xmath579 $ ] , @xmath580 and @xmath581 are mutually disjoint ; * @xmath582\}\cup\{s_j;j\in b\}$ ] is a partition of @xmath358 $ ] .
[ eg - core ] let @xmath583 be what s shown in fig [ fig - core ] .
clearly @xmath574 is an @xmath578-frame over @xmath358 $ ] , where the subsets @xmath584 have a common element @xmath585 , and the subsets @xmath586 have a common element @xmath587 .
-frame , where @xmath588 , @xmath589 and @xmath590.,width=188 ] [ g - core ] a subset @xmath106 $ ] is said to be an @xmath591-_core _ if the following three conditions hold : * if @xmath579 $ ] and @xmath592 , then @xmath593 ; * if @xmath579 $ ] and @xmath594 , then there is an @xmath595 such that @xmath596 and @xmath597 ; * if @xmath598 , then @xmath599 .
additionally , if @xmath106 $ ] is an @xmath360-core and @xmath362 , then @xmath83 is called an @xmath363-_core_. clearly , if @xmath106 $ ] is an @xmath360-core and @xmath364 , then @xmath365 is also an @xmath360-core .
in particular , if @xmath106 $ ] is an @xmath360-core and @xmath365 is a @xmath12-subset of @xmath83 , then @xmath365 is an @xmath366-core .
_ example [ eg - core ] continued : _ in example [ eg - core ] , let @xmath334 . then @xmath600 and @xmath601 are both @xmath363-core .
however , @xmath602 and @xmath603 are not @xmath360-core , because @xmath83 does not satisfy condition ( 2 ) and @xmath365 does not satisfy condition ( 3 ) of definition [ g - core ] .
[ core - form ] let @xmath574 be an @xmath578-frame as in definition [ frame ] .
suppose @xmath399 and for any @xmath279-subset @xmath255 of @xmath256 $ ] , @xmath604 .
then the following hold : * if @xmath448 $ ] has size @xmath605 , then there is an @xmath606 such that @xmath83 is an @xmath363-core . * for any @xmath134 $ ] and @xmath115 of size @xmath86 ,
there is an @xmath363-core @xmath83 such that @xmath444 .
\1 ) let @xmath607 ; |t\cap s_\ell|\geq r\}.\ ] ] for each @xmath579 $ ] and @xmath608 , we pick a subset @xmath609 as follows : \1 ) if @xmath610 , then let @xmath611 for each @xmath612 .
\2 ) if @xmath613 and @xmath614 , then for each @xmath615 , let @xmath616 be an @xmath9-subset of @xmath617 satisfying @xmath618 , and for each @xmath619 , let @xmath611 .
\3 ) if @xmath613 and @xmath620 , then fix an @xmath621 , and let @xmath622 be an @xmath9-subset of @xmath623 , let @xmath616 be an @xmath624-subset of @xmath617 for each @xmath625 , and let @xmath611 for each @xmath619 .
moreover , for each @xmath626 , let @xmath616 be an @xmath9-subset of @xmath617 , and for each @xmath627 , let @xmath611 .
let @xmath628}w_\ell$ ] , then by definition [ g - core ] , @xmath629 is an @xmath360-core .
we now prove that @xmath630 .
let @xmath631 ; j\cap a_j\neq\emptyset\}.\ ] ] we need to consider the following two cases : case 1 : @xmath457 .
without loss of generality , assume @xmath195 , then pick a @xmath279-subset @xmath211 of @xmath255 , and replace @xmath255 by @xmath211 in our discussion . ] .
then from the assumption of this lemma , @xmath632 , @xmath633 since @xmath634 is a partition of @xmath256 $ ] , @xmath635 is a partition of @xmath255 and @xmath636 combining ( [ eqn:28])@xmath637([eqn:30 ] ) , we have @xmath638 by the construction of @xmath629 , we have @xmath639 ) and ( [ eqn:32 ] ) imply that @xmath640 case 2 : @xmath641 . by the construction of @xmath629 , for each @xmath579 $ ] and @xmath642 , @xmath616 is obtained by deleting at most @xmath47 elements from @xmath643 .
we thus have @xmath644 moreover , @xmath645 then @xmath646}w_\ell| \geq|t|-|j|(\delta-1).\ ] ] note that @xmath647 and @xmath641 .
therefore @xmath648 gathering both cases , we always have @xmath630 .
let @xmath83 be a @xmath12-subset of @xmath629 .
note that @xmath629 is an @xmath360-core .
so @xmath649 is an @xmath363-core .
\2 ) to prove the second claim of lemma [ core - form ] , note that @xmath399 , and hence we can always find a @xmath279-subset @xmath255 of @xmath256 $ ] such that @xmath650 .
similar to the proof of 1 ) , for each @xmath651 , we can pick a @xmath616 such that @xmath440 , @xmath652 is an @xmath360-core and @xmath653 . let @xmath83 be a @xmath12-subset of @xmath652 such that @xmath444 .
then @xmath83 is an @xmath363-core and @xmath444 .
_ example [ eg - core ] further continued _ : consider the @xmath578-frame @xmath574 in example [ eg - core ] .
let @xmath334 .
then @xmath574 satisfies the conditions of lemma [ core - form ] .
we consider the following two instances : instance 1 : @xmath654 . as in the proof of lemma [ core - form ] , @xmath655 and @xmath656 .
let @xmath657 , @xmath658 , @xmath659 , @xmath660 , @xmath661 , @xmath662 and @xmath663 for @xmath664
. then @xmath665 .
instance 2 : @xmath666 . then @xmath667 and @xmath641 .
let @xmath668 , @xmath669 and @xmath670 . then @xmath671 @xmath672 .
[ nor - core ] let @xmath574 be an @xmath578-frame as in definition [ frame ] .
for each @xmath579 $ ] and @xmath673 , let @xmath401 be an @xmath9-subset of @xmath403 such that @xmath674 . for each @xmath675 ,
let @xmath401 be an @xmath9-subset of @xmath403 .
let @xmath676}u_i.\ ] ] then by definition [ g - core ] , @xmath412 is an @xmath591-core .
clearly , @xmath677 in example [ eg - core ] , let @xmath334 , then @xmath678 @xmath679 @xmath680 is an @xmath360-core obtained by the process of remark [ nor - core ] .
[ core - exist ] let @xmath574 be an @xmath578-frame as defined in definition [ frame ] and @xmath412 be what s described in remark [ nor - core ] .
suppose @xmath681 , s_0\subseteq\omega$ ] and @xmath134 $ ] .
if @xmath682 and @xmath469 is an @xmath363-core , then there exists an @xmath683 such that @xmath684 is an @xmath363-core . by the construction of @xmath412 , @xmath685 .
since @xmath686 , so @xmath687 since @xmath469 is an @xmath363-core , by definition [ g - core ] , @xmath688 and @xmath689 thus , we can find an @xmath476 . if @xmath598 , then by definition [ frame ] , @xmath690\backslash\{i\}$ ] .
then @xmath684 is an @xmath363-core . now
, suppose @xmath673 for some @xmath579 $ ] .
we need to consider the following two cases .
case 1 : @xmath691 . since @xmath692 , then @xmath693 and @xmath694\backslash\{i\}$ ] .
then @xmath684 is an @xmath366-core .
case 2 : @xmath695 .
since @xmath469 is an @xmath363-core , from definition [ g - core ] , we differentiate the following two sub - cases : subcase 2.1 : @xmath696 . in that case , it is clear that @xmath684 is an @xmath363-core .
subcase 2.2 : there is an @xmath697 such that @xmath698 , @xmath699 and @xmath700 . in that case , we have @xmath701 let @xmath702 , then @xmath693 and @xmath694\backslash\{i\}$ ] .
it then follows that @xmath684 is an @xmath363-core .
_ example [ eg - core ] and [ eg - omg0 ] continued : _ consider again example [ eg - core ] .
let @xmath334 , @xmath703 and @xmath704 , where @xmath412 is as in example [ eg - omg0 ] .
we can easily verify the following : let @xmath705 , @xmath706 ; then @xmath707 is an @xmath363-core .
if we further let @xmath708 , then @xmath684 is also an @xmath366-core .
let @xmath709 ; then @xmath710 is an @xmath363-core .
if we further let @xmath711 , then @xmath712 is also an @xmath363-core .
let @xmath713 ; then @xmath714 is an @xmath363-core .
if we further let @xmath715 , then @xmath716 is also an @xmath363-core .
[ code - extd ] let @xmath574 be an @xmath578-frame defined in definition [ frame ] and @xmath412 be what s defined in remark [ nor - core ] .
let @xmath681 $ ] and @xmath717 such that for any @xmath363-core @xmath466 , the vectors in @xmath445 are linearly independent .
suppose @xmath134 $ ] and @xmath718 .
if @xmath68 , then for any @xmath682 , there is a @xmath719 such that for any @xmath363-core @xmath488 , the vectors in @xmath445 are linearly independent . let @xmath467 be the set of all @xmath468 such that @xmath469 is an @xmath363-core .
for any @xmath470 , by lemma [ core - exist ] , there is an @xmath720 such that @xmath684 is an @xmath366-core .
from the assumptions , @xmath721 is linearly independent .
hence @xmath722 thus , @xmath483 since @xmath484 , by lemma [ sub - space ] , @xmath485 let @xmath723 .
then for any @xmath363-core @xmath488 , the vectors in @xmath724 are linearly independent .
the second construction method for optimal @xmath2 codes is illustrated in the proof of the following theorem .
[ opt - suf ] let @xmath574 be an @xmath578-frame in definition [ frame ] .
suppose @xmath399 and for any @xmath279-subset @xmath255 of @xmath256 $ ] , @xmath604 .
if @xmath68 , then there exists an optimal @xmath2 linear code over @xmath89 .
let @xmath412 be what s described in remark [ nor - core ] and @xmath405 . clearly , @xmath725 since @xmath399 , let @xmath255 be a @xmath279-subset of @xmath256 $ ] ;
then from the assumptions , @xmath726 by remark [ nor - core ] , @xmath727 . hence latexmath:[\[l=|\omega_0|\geq|\cup_{i\in
j}u_i|=\cup_{i\in j}s_i|- the construction of an optimal @xmath2 code consists of the following two steps . _ step 1 _ : construct an @xmath408 $ ] mds code @xmath409 over @xmath89 .
such an mds code exists when @xmath410 .
let @xmath411 be a generating matrix of @xmath409 .
we index the columns of @xmath411 by @xmath412 , i.e. , @xmath413 , where @xmath414 is a column of @xmath729 .
_ step 2 _ : extend the code @xmath409 to an optimal @xmath2 code @xmath37 .
this can be achieved by the following algorithm , which appears similar to algorithm 1 ( on the surface ) but is actually different ( in details ) .
* algorithm 2 : * \1 .
let @xmath416 .
@xmath34 runs from @xmath298 to @xmath417 .
\3 . while @xmath418 : \4 .
pick a @xmath419 and let @xmath420 be such that for any @xmath363-core @xmath421 , @xmath422 is linearly independent .
\5 .
@xmath423 .
let @xmath37 be the linear code generated by the matrix @xmath424 @xmath425 .
since @xmath413 is a generating matrix of the mds code @xmath409 , so for any @xmath363-core @xmath465 , @xmath724 is linearly independent .
then in algorithm 2 , by induction , we can assume that for any @xmath363-core @xmath466 , @xmath445 is linearly independent . by lemma [ code - extd ] , in line 4 of algorithm 2 ,
we can always find a @xmath427 satisfying the requirement .
hence , by induction , the collection @xmath730\}$ ] satisfies the condition that for any @xmath363-core @xmath106 $ ] , @xmath445 is linearly independent .
moreover , since in line 4 of algorithm 2 , we can choose a @xmath431 , which satisfies @xmath432 by induction , @xmath731 for any @xmath134 $ ] and @xmath115 of size @xmath86 , by claim 2 ) of lemma [ core - form ] , there is an @xmath363-core @xmath83 such that @xmath444 . hence @xmath445 is linearly independent .
thus , @xmath732 therefore , by definition [ def - locality ] and remark [ rem - loty ] , @xmath37 is an @xmath2 code . finally , we prove that the minimum distance of @xmath37 is @xmath733 .
suppose @xmath448 $ ] and @xmath449 . by 1 ) of lemma [ core - form
] , there is an @xmath606 which is an @xmath366-core .
therefore , @xmath734 by the minimum distance bound in ( [ eqn:1 ] ) and lemma [ fact ] , the minimum distance of @xmath37 is @xmath735 hence @xmath37 is an optimal @xmath2 code .
_ example [ eg - core ] continued : _ consider the @xmath578-frame @xmath574 in example [ eg - core ] .
let @xmath334 .
then it is obvious @xmath574 satisfies the conditions of theorem [ opt - suf ] .
thus , we can use algorithm 2 to construct an optimal @xmath2 linear code over the field of size @xmath736 .
note that @xmath737 .
hence , @xmath58 and this is a new optimal @xmath2 code . as applications of theorem [ opt - suf ] , in the following ,
we show that optimal @xmath2 codes exist for two other sets of coding parameters . from claim 2 ) of lemma
[ low - bound ] , we know that @xmath132 is a necessary condition for the existence of optimal @xmath2 linear codes .
thus we will assume @xmath132 in the following discussion .
[ opt - ext-3 ] suppose @xmath64 and @xmath65 , where @xmath67 and @xmath66 .
suppose @xmath738 and @xmath739 .
if @xmath68 , then there exists an optimal @xmath2 linear code over @xmath89 .
let @xmath561 .
note that we have assumed that @xmath132 . then @xmath740 let @xmath741 and @xmath742 then from the assumptions , @xmath743 .
therefore @xmath744 and @xmath745 from equation ( [ eqn:33 ] ) , @xmath746 .
the set @xmath747 $ ] can be partitioned into @xmath368 mutually disjoint subsets , say , @xmath748 , each of size @xmath749 .
let @xmath750 moreover , from equation ( [ eqn:34 ] ) , the set @xmath751 $ ] can be partitioned into @xmath752 mutually disjoint subsets , say , @xmath753 , each of size @xmath303 .
let @xmath754 and @xmath755 , @xmath756 , and @xmath757 .
then @xmath141 is an @xmath578-frame . for any @xmath279-subset @xmath255 of
@xmath256 $ ] , since @xmath739 , then @xmath758 let @xmath759 and @xmath760 . by the construction of @xmath574
, we have @xmath761 by theorem [ opt - suf ] , if @xmath68 , then there exists an optimal @xmath2 code over @xmath89 .
[ opt - ext-4 ] suppose @xmath64 and @xmath65 , where @xmath67 and @xmath66 .
suppose @xmath762 and @xmath763 .
if @xmath68 , then there exists an optimal @xmath2 linear code over @xmath89 .
let @xmath561 .
note that we have assumed that @xmath132 . then @xmath764 let @xmath765 and @xmath766 then by assumption , @xmath767 .
it then follows that @xmath768 from equation ( [ eqn:35 ] ) , the set @xmath769 $ ] can be partitioned into @xmath169 mutually disjoint subsets , say , @xmath770 , each of size @xmath771 . for each @xmath772
, we can find two subsets @xmath773 of @xmath774 such that @xmath775 and @xmath776 then @xmath777 let @xmath778 and @xmath779 .
moreover , from equation ( [ eqn:36 ] ) , the set @xmath751 $ ] can be partitioned into @xmath780 mutually disjoint subsets , say @xmath781 , each of size @xmath303 .
let @xmath782 , @xmath783 $ ] and @xmath784 .
then @xmath141 is an @xmath578-frame . for any @xmath279-subset @xmath255 of @xmath256 $ ] .
since @xmath785 , then @xmath786 let @xmath787;a_j\subseteq j\}$ ] . then @xmath788 ) an ( [ eqn:38 ] ) , we have @xmath789 since @xmath790 is an integer , then @xmath791 by the construction of @xmath574 , we have @xmath792 by theorem [ opt - suf ] , if @xmath68 , then there exists an optimal @xmath2 code over @xmath89 . we now provide some discussions of theorem [ opt - ext-4 ] . since @xmath67 , then @xmath793 . given
@xmath794 and @xmath10 , let @xmath795 .
then the conditions @xmath762 and @xmath796 can always be satisfied when @xmath797 . on the other hand , when @xmath798 and @xmath799 , then @xmath800 or @xmath801 and @xmath802 , which leads to @xmath763 . by theorem [ opt - ext-4 ] , there exist optimal @xmath2 codes when @xmath797 , @xmath798 and @xmath799 . [
cols="<,<,<,<,<,<,<,<,<,<,<",options="header " , ]
we have investigated the structure properties and construction methods of optimal @xmath2 linear codes , whose length and dimension are @xmath11 and @xmath12 respectively . a structure theorem for optimal @xmath2 code with @xmath61
is first obtained .
we next derived two sets of parameters where no optimal @xmath2 linear codes could exist ( over any field ) , as well as identified four sets of parameters where optimal @xmath2 linear codes exist over any field of size @xmath68 .
some of these existence conditions were reported in the literature before , but the minimum field size we derived is ( considerably ) smaller than those derived in the previous works .
our results have considerably substantiated the results in terms of constructing optimal @xmath2 codes , and there are now only two small holes ( two subcases with specific parameters ) where the existence results are unknown . except for these two small subcases , for all the other cases , given each tuple of @xmath803 , either an optimal @xmath2 linear code does not exist or an optimal @xmath2 linear code can be constructed using a deterministic algorithm . as an illustrative summary of our results , we also provide in table 1 an example of the existence of optimal @xmath2 linear codes for the parameters of @xmath804 , @xmath805 and @xmath806 . in this table , e@xmath807 means that optimal @xmath2 linear codes can be constructed by the method in @xcite or by our theorem [ opt - ext-1 ] and algorithm 1 ( which requires a substantially smaller field ) ; e@xmath808resp .
e@xmath809 , e@xmath810 means optimal @xmath2 linear codes can be constructed by theorem [ opt - ext-2 ] @xmath13resp .
theorem [ opt - ext-3 ] , theorem [ opt - ext-4]@xmath14 ; n@xmath811resp .
n@xmath812 means optimal @xmath2 linear codes do not exist according to theorem [ non - exst ] @xmath13resp .
theorem [ non - exst-1]@xmath14 ; and @xmath813 means we do not yet know whether an optimal @xmath2 linear code exists or not .
r. bhagwan , k. tati , y .- c .
cheng , s. savage , and g. m. voelker , `` total recall : system support for automated availability management , '' in _ symp .
networked systems design and implementation ( nsdi ) _ , 2004 .
c. huang , m. chen , and j. li , `` pyramid codes : flexible schemes to trade space for access efficiency in reliable data storage systems , '' in _ network computing and applications , 2007 .
sixth ieee international symposium on_. ieee , 2007 , pp .
79 - 86 .
n. prakash , g. m. kamath , v. lalitha , and p. v. kumar , `` optimal linear codes with a local - error - correction property , '' in _ information theory proceedings ( isit ) , 2012 ieee international symposium on _ , pp .
2776 - 2780 , ieee , 2012 .
s. jaggi , p. sanders , p. a. chou , m. effros , s. egner , k. jain , and l. tolhuizen , `` polynomial time algorithms for multicast network code construction , '' _ ieee trans .
inf . theory _ ,
51 , no . 6 , pp .
1973 - 1982 , jun . 2005 | linear erasure codes with local repairability are desirable for distributed data storage systems .
an @xmath0 $ ] code having all - symbol @xmath1-locality , denoted as @xmath2 , is considered optimal if it also meets the minimum hamming distance bound .
the existing results on the existence and the construction of optimal @xmath3 codes are limited to only the special case of @xmath4 , and to only two small regions within this special case , namely , @xmath5 or @xmath6 , where @xmath7 and @xmath8 .
this paper investigates the existence conditions and presents deterministic constructive algorithms for optimal @xmath2 codes with general @xmath9 and @xmath10 .
first , a structure theorem is derived for general optimal @xmath2 codes which helps illuminate some of their structure properties .
next , the entire problem space with arbitrary @xmath11 , @xmath12 , @xmath9 and @xmath10 is divided into eight different cases ( regions ) with regard to the specific relations of these parameters . for two cases ,
it is rigorously proved that no optimal @xmath2 could exist . for four other cases
the optimal @xmath2 codes are shown to exist , deterministic constructions are proposed and the lower bound on the required field size for these algorithms to work is provided .
our new constructive algorithms not only cover more cases , but for the same cases where previous algorithms exist , the new constructions require a considerably smaller field , which translates to potentially lower computational complexity .
our findings substantially enriches the knowledge on @xmath2 codes , leaving only two cases in which the existence of optimal codes are yet to be determined . |
extremal dynamics has been employed extensively in modelling far from equilibrium systems such as biological evolution @xcite and driven interfaces @xcite .
although processes with extremal dynamics do not have a phase transition ( there is no control parameter ) they exhibit scaling properties reminiscent of those observed at continuous phase transitions @xcite .
indeed , it was suggested some time ago that the appearance of self - organized " scaling properties in extremal dynamics and in sandpile models corresponds to forcing the order parameter ( associated with an underlying phase transition ) to zero from above @xcite . the purpose of this work is to explore the connection between these scaling properties and those observed at a phase transition to an absorbing state .
the connection between extremal dynamics and directed percolation ( dp ) , the prime example of an absorbing - state phase transition , was first suggested by paczuski , maslov and bak @xcite and investigated in detail by these authors in the context of the bak - sneppen ( bs ) model and related processes @xcite .
( the latter work , as well as ref .
@xcite , clearly demonstrated that the critical exponents of the bs model are _ not _ those of dp . ) sornette and dornic @xcite and grassberger and zhang @xcite have shown how a variant of directed percolation may be transformed via extremal dynamics to display soc .
these studies indicate that self - organized criticality ( soc ) @xcite under extremal dynamics arises because the system is driven to a critical point associated with a phase transition to an absorbing state @xcite , as is also the case for sandpiles @xcite . in the present work we are particularly interested in the modifications needed to transform a ( non - extremal , non - soc ) model having an absorbing state to one exhibiting soc under extremal dynamics .
we develop a general scheme relating the two classes of models via an intermediate ,
extremal - absorbing " process whose absorbing - state critical point corresponds exactly to the critical behavior observed in the corresponding soc model .
two examples ( the bs model and an extremal contact process ) are studied in detail , yielding refined estimates for critical properties , and evidence of a new universality class associated with absorbing phase transitions under extremal dynamics .
the prime example of extremal dynamics is the bs model @xcite , proposed to explain mass extinctions observed in the fossil record . while its application in the evolutionary context is debated @xcite , it remains an intriguing and incompletely understood example of scaling behavior far from equilibrium .
the contact process ( cp ) @xcite is the most familiar example of a markov process exhibiting a phase transition to an absorbing state .
we focus on the absorbing version of the bs model , and the extremal version of the cp , to illustrate the relations between extremal dynamics and absorbing phase transitions .
our analysis of the spread of activity leads to new or refined values for the exponents @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 , and for the critical threshold of the bs model .
( for extremal dynamics , the avalanche exponent is @xmath8 . )
finite - size scaling analysis of stationary properties at the critical point yields estimates of the exponent ratios @xmath9 and @xmath10 .
studies of modified bs models have shown that scaling properties are insensitive to changes that preserve its basic symmetries ( that is , invariance under translation and reflection ) @xcite , pointing to the existence of a bs universality class .
nonextremal models that exhibit a phase transition to an absorbing state , and that possess these same symmetries , and no additional ones , belong generically to the directed percolation ( dp ) universality class @xcite . here
we show that such models fall in a new extremal - dp " universality class when modified to follow extremal dynamics .
the absorbing phase transition corresponding to the scaling behavior of the bs and other extremal models belongs to the extremal - dp class , not that of ordinary directed percolation . in other words ,
extremal dynamics is a relevant perturbation for absorbing - state transition , just as was shown by sneppen in the context of interface depinning @xcite .
the balance of this paper is organized as follows .
section ii presents a general scheme linking ordinary absorbing - state phase transitions and extremal dynamics via an intermediary extremal - absorbing " model . in sec .
iii we describe our simulation method and results .
our findings regarding scaling and universality are discussed in sec .
mean - field analyses are presented in the appendix .
in this section we examine how a stochastic model with an absorbing - state phase transition may be transformed to exhibit scale invariance under extremal dynamics .
we begin , for generality , by defining a rather abstract scheme , and then discuss specific examples .
a large class of models exhibiting an absorbing - state phase transition may be formulated as follows @xcite .
consider a stochastic process @xmath11 defined on a connected graph @xmath12 of @xmath13 sites .
( @xmath12 consists of a set of sites with links between certain pairs of sites .
typical examples are a ring of @xmath13 sites , and the @xmath14-dimensional hypercubic lattice @xmath15 , with links between nearest neighbors . )
the _ state _ @xmath16 of site @xmath17 is 0 or 1 , the latter value denoting an active site , the former an inactive one .
for each site @xmath17 in @xmath12 we define a neighborhood @xmath18 , or , more generally , a set of neighborhoods @xmath19 , @xmath20, ... ,@xmath21 .
the dynamics of @xmath11 proceeds in steps .
each step involves choosing an active site @xmath17 ( the _ central site _ for this step ) , at random , and changing the states of the sites in @xmath22 according to a certain rule @xmath0 . in case there are two or more neighborhoods , one of them ,
@xmath23 say , is chosen at random from the collection , with probability @xmath24 , and a rule @xmath25 is applied to the site or sites in @xmath26 . in general @xmath0 ( or @xmath25 ) is a probabilistic rule . at each step
the number of active sites may change , and if at any moment there are no active sites ( @xmath27 ) , the process has fallen into an absorbing state and there is no further evolution .
otherwise the dynamics proceeds to the next step . using @xmath28 to denote the entire set of activity variables @xmath29 at step @xmath30 ,
the dynamics generates a sequence @xmath31 starting from the initial configuration @xmath32 .
it is frequently of interest to associate a continuous time variable @xmath33 with the process .
this is usually done by associating a time increment @xmath34 with each step , where @xmath35 is the number of active sites just before the step is realized .
we define the _ order parameter _ as @xmath36 $ ] , i.e. , the fraction of active sites at time @xmath33 .
( the event space here is the set of all realizations of the process up to time @xmath33 , starting from a given initial probability distribution on configuration space . )
if the stationary order parameter , defined by @xmath37 , vanishes , the process is said to be in the _ absorbing phase _ ; otherwise it is in the _ active _ phase .
a simple model exhibiting a phase transition to an absorbing state is the _ contact process _ (
cp ) @xcite .
here we consider the one - dimensional version .
there are three sets @xmath23 , conveniently denoted as @xmath38 and @xmath39 .
the associated probabilitites are @xmath40 and @xmath41 . in terms of the usual parametrisation @xcite , @xmath42 , where @xmath43 represents the rate of spread of activity .
( in the epidemic " interpretation of the cp , active sites represent infected organisms , inactive sites susceptibles , and @xmath44 is the infection rate . )
the updating rules are : @xmath45 , @xmath46 . in other words ,
an active site has a probability per unit time of @xmath47 to become inactive , while an inactive site @xmath48 becomes active at rate @xmath49 $ ] , where @xmath50 is the number of active neighbors of site @xmath48 .
as is well known , the one - dimensional cp exhibits a continuous phase transition between an absorbing phase and an active one at @xmath51 @xcite .
it is convenient to associate the control parameter with the updating rule @xmath0 rather than with the probabilities @xmath24 .
we therefore reformulate the cp as follows . with the sets
@xmath52 and @xmath53 defined above , we take @xmath54 and @xmath55 , and define @xmath56 .
the updating functions are : @xmath57 ( ` w.p . '
denotes with probability " ) , and @xmath58 it is easy to verify that the transition rates satisfy @xmath59 , just as in the original formulation . the critical value @xmath60
. the following three - site contact process ( cp3 ) will play an important role in our analysis @xcite .
for each site we define the set @xmath61 ( the central site and its nearest neighbors ) .
the updating function @xmath0 takes values of 1 and 0 with probabilities @xmath62 and @xmath63 respectively , independently at each of the three sites in @xmath22 .
@xcite this is called model 3. ) simulations of the cp3 show that it exhibits a continuous phase transition at @xmath64 .
further interacting particle systems , such as the pair contact process @xcite and the diffusive cp @xcite , can be accomodated within the scheme set out above .
we shall assume that the process @xmath11 is defined so that the control parameters ( such as @xmath62 ) are associated with the updating rule @xmath0 .
each time a site is updated , the value of @xmath0 may be determined by comparing a random number @xmath65 with the parameter in question .
( this is of course the usual procedure in simulations . ) in the cp3 , for example , we take @xmath66 if @xmath67 , and zero otherwise , where @xmath65 is uniformly distributed on the interval [ 0,1 ] .
call the random number associated with the most recent updating of site @xmath17 , @xmath68 , so that @xmath69 with @xmath70 the unit step function .
( the initial values of the @xmath68 are assigned according to the state variables @xmath16 .
for example , if all sites are initially active , we draw the initial @xmath68 from the distribution uniform on @xmath71 $ ] . ) for the cp , eq .
( [ f0cp ] ) requires that we update the central site @xmath17 with a number chosen uniformly from [ 0,1 ] .
according to eq .
( [ fpmcp ] ) , the same applies when updating an _ inactive _ neighbor ( @xmath72 ) , but when updating an _ active _ neighbor , the random number @xmath65 is drawn from the interval @xmath71 $ ] , since an active neighbor remains active . summarizing , we have shown how a particle system @xmath11 may be formulated using a set of random variables @xmath68 , such that site @xmath17 is active if @xmath68 is smaller than a certain parameter @xmath62 .
@xmath11 suffers an absorbing - state phase transition at @xmath73 .
we now define two related processes , @xmath74 and @xmath75 .
the former , _ extremal absorbing _ process , is obtained by modifying how the central site is selected .
instead of choosing it at random from among the currently active sites , it is taken to be _ the active site having the smallest _ @xmath68 .
as in the original process @xmath11 , if there are no active sites ( i.e. , @xmath76 , @xmath77 ) , the process has reached an absorbing configuration and the evolution ceases .
thus @xmath74 possesses an absorbing state , and since the original process @xmath11 exhibits a phase transition between an active and an absorbing phase as the control parameter @xmath62 is varied , we expect @xmath74 to as well , at some value @xmath78 .
( the reason is that the relative likelihood of generating and destroying active sites varies with @xmath62 , just as in @xmath11 . )
mean - field theory ( see appendix ) yields @xmath79 . due to the different correlations generated under extremal dynamics , however , the critical value @xmath78 of @xmath74 is in general different from @xmath80 .
we define the _ extremal _ process @xmath75 by relaxing the condition in @xmath74 , that the extremal site @xmath17 must be active ( i.e. , have @xmath81 ) for the dynamics to proceed . if @xmath11 is the original contact process , then @xmath75 is a process in which either the minimal site or one of its nearest neighbors is updated at each step .
if @xmath11 is the cp3 , @xmath75 is the familiar bak - sneppen model .
note that @xmath75 has no absorbing state , hence no phase transition to such a state .
its stationary properties are nevertheless intimately connected with the critical - point properties of @xmath74 , as we now explain . of particular interest
is the stationary probability density @xmath82 of site variables under extremal dynamics . as is well known , @xmath83 in the bak - sneppen model , in the infinite - size limit .
( @xmath84 is the normalization factor . )
we expect @xmath82 to exhibit a step - function singularity in any extremal model @xmath75 @xcite .
this feature is in fact already present in the original absorbing - state model @xmath11 at its critical point , because at the critical point @xmath73 , the stationary density of active sites ( having @xmath85 ) tends to zero as the system size @xmath13 goes to infinity .
the distribution on the allowed " region @xmath86 is uniform , since the @xmath87 are drawn from a uniform distribution .
thus @xmath82 jumps from zero to a finite value at @xmath88 . in the _
supercritical _ regime ( @xmath89 ) , @xmath82 is equal to a constant @xmath90 for @xmath67 , such that @xmath91 ( the order parameter ) , and takes a different constant value , @xmath92 , on the interval @xmath93 $ ] .
once again , the stationary density is discontinuous at @xmath94 .
what holds for @xmath11 also holds qualitatively for @xmath74 .
the critical value @xmath78 may , as noted , differ from @xmath80 , but since @xmath74 exhibits an absorbing - state phase transition , its stationary distribution @xmath82 also has a step - function singularity . just at @xmath95 , the order parameter @xmath96 , but as @xmath97 the survival time of the process tends to infinity .
this means that the process can survive indefinitely , with the choice of the central site _ restricted to the set having _ @xmath98 .
the presence of active sites in the range @xmath99 is then irrelevant , since a site with @xmath98 is always available .
thus for @xmath100 , the distribution @xmath82 exhibits a step - function singularity at @xmath78 .
extremal dynamics effectively pins " the singularity at @xmath78 .
the foregoing remarks on @xmath74 obtain in the infinite - size limit ; for finite @xmath13 there is a nonzero probability ( for @xmath101 ) , that all sites have @xmath86 so that the system eventually becomes trapped in the absorbing state .
( the mean lifetime , however , is expected to grow exponentially with @xmath13 , for @xmath102 . ) in summary , if @xmath11 exhibits an absorbing - state phase transition , then @xmath74 should as well , although not necessarily at the same value of @xmath62 .
the distribution @xmath82 possesses a step function singularity in both cases . near the critical point ( @xmath103 ) of the original process
@xmath11 we expect @xmath104 , with @xmath105 the critical exponent associated with the order parameter .
below the upper critical dimension ( @xmath106 for dp @xcite ) , @xmath107 . in the supercritical regime of @xmath74 ,
on the other hand , @xmath108 since @xmath82 jumps from zero to a finite value at @xmath109 .
thus the order parameter exponent @xmath105 is _ unity _ in @xmath74 .
this is but one example showing that the extremal condition changes the value of a critical exponent .
another example , established some time ago by paczuski and coworkers @xcite , is that the spreading exponent @xmath4 ( defined sec .
iii.c ) is generically zero for extremal dynamics , whereas @xmath110 for directed percolation .
further evidence that extremal dynamics modifies critical exponent values will be given below .
in @xmath111 , the central site is always chosen ( in the @xmath112 limit ) from the set @xmath113 , just as in @xmath74 at its critical point .
since the order parameter is zero in the latter case , we may assert that extremal dynamics effectively tunes the order parameter to zero .
( @xmath114 approaches zero from above as @xmath112 . )
as we have seen , the existence of sites with @xmath115 becomes irrelevant in @xmath74 , in the infinite - size limit .
in other words , @xmath111 is identical to the critical process @xmath74 in this limit .
( starting from the same initial configuration , and using the same set of random numbers , the same sequence of sites will be updated in the two processes . )
since the extremal version of the cp3 is the familiar bak - sneppen model , we shall refer to the cp3@xmath116 as the _ absorbing bak - sneppen _ ( abs ) model .
our objective is to characterize the behavior of the abs model and of the extremal and ea versions of the contact process .
the abs model is closely related to the @xmath0-_avalanche _ process studied in ref .
an @xmath0-avalanche ( in the present notation , @xmath62-avalanche ) , begins when the minimal site variable @xmath117 , the minimum having been _ greater _ than @xmath62 at the preceding step or steps , and continues until the minimum is once again @xmath118 .
( as @xmath62 approaches @xmath78 from below , the mean avalanche duration diverges . )
the dynamics of the bs model continues , regardless of whether a given avalanche has terminated or not .
but in the abs model @xmath119 represents an absorbing state and the dynamics ceases . in the bs model , it is common to analyze the properties of @xmath62-avalanches in the stationary state .
it is similarly of interest to study stationary properties of the abs model , attained once the system has relaxed , after an initial transient period .
we may also study the mean lifetime of the active state as a function of system size .
another approach to studying absorbing - state phase transitions consists in following the spread of activity starting from a single active site .
this spreading phenomenon in the bs model was studied in refs .
@xcite , where it is called the bs branching process , and in ref .
@xcite under the name of the bs(@xmath120 ) model .
the assertions regarding extremal and extremal - absorbing models are supported by mean - field theory ( mft ) , which is presented in the appendix .
in particular , for the abs model @xmath121 , just as in the mft of the original bs model . in the extremal and the extremal - absorbing contact process , the stationary probability densities exhibit ( for @xmath122 ) , _ two _ discontinuities , one at @xmath123 , the other at @xmath124 . [ these coalesce at @xmath125 . note that the parameter @xmath62 continues to influence the form of @xmath126 in the extremal contact process , due to the nature of the updating rule , eqs .
( [ f0cp ] ) and ( [ fpmcp ] ) . ]
these predictions are in qualitative agreement with simulation .
before discussing our results we insert an observation on simulations of the bs model .
since the site with the smallest variable , @xmath127 , must be identified at each step , it becomes important to devise an effective search strategy .
an efficient general - purpose search algorithm uses a binary tree structure to identify @xmath127 .
one approach @xcite utilizes a lattice of @xmath128 sites . at the first level of selection ,
each site is compared with one of its neighbors and the minimum of the pair selected . at the next level
the minimum between each neighboring pair is selected , and so on , so that at the @xmath30-th level the global minimum is identified . a second binary scheme @xcite is formulated as follows .
site 0 is placed at the apex of the tree .
site 1 is placed on the level below the apex , to the left of 0 if @xmath129 , to the right if @xmath130 .
a site @xmath17 is added to the tree in the following way : we go down the tree comparing @xmath68 with the variables @xmath131 , turning left or right depending on whether @xmath68 is smaller or larger than @xmath87 , until we find an empty site . building the tree in this way , @xmath127 will occupy the leftmost position in the tree . in these schemes , maintaining the tree structure , once constructed from the initial set of variables @xmath68 , requires a small number of operations at each step , and is many times more efficient that a repeated global search for the minimim .
we find , nonetheless , that a suitably _ restricted _ search requires less cpu time in the stationary state .
a special property of the bs model ( shared by its absorbing version , and by other extremal models ) , is that the minimal site falls , with a probability approaching unity as the system size grows , in the interval @xmath132 $ ] . at the same time
the density of sites with values in this interval approaches zero as @xmath112 .
this suggests maintaining a list of sites having @xmath133 @xcite .
then the search for @xmath127 may be restricted to the list , except for the rare instances in which the latter is empty .
( for the abs we must in any case restrict the search to sites with @xmath134 . )
if the system is large , so that the typical number of sites with @xmath67 is not small , it becomes advantageous to introduce a _
second _ list , of sites having @xmath135 .
when this relatively short list is nonempty ( as is usually the case ) the search for @xmath127 is restricted to it . in studies of the bs model
, we obtained the greatest efficiency using @xmath136 , while the criterion for the first list was @xmath137 , that is , slightly _ below _
( the occasional need to perform a global search , in the rare instances when both lists are empty , is more than compensated by their reduced sizes when using these values . )
compared with the binary tree method , our approach results in threefold reduction in cpu time , in the stationary state , for a system of 1000 sites .
( the binary tree approach may prove more efficient for studying transients , since initially the lists will not be short , if the @xmath68 values are chosen uniformly on [ 0,1 ] . ) using the simulation method described above , we conducted extensive studies of the one - dimensional abs model .
we initialize the system with all sites active and allow it to relax until mean properties fluctuate about stationary values .
the stationary properties are then obtained from temporal averages over the set of surviving realizations .
each step corresponds to a time interval of @xmath138 , with @xmath35 the number of active sites just before the updating is performed .
results for the stationary order parameter ( i.e. , the density of sites with @xmath67 ) , on lattices confirm that in the supercritical regime ( @xmath102 ) , the order parameter grows linearly with @xmath139 , as anticipated in sec .
we study the finite - size scaling behavior of the stationary order parameter @xmath140 and of the lifetime @xmath141 at the critical point .
( @xmath141 is obtained from an exponential fit to the survival probability @xmath142 . ) the expected finite - size scaling behaviors at the critical point are : @xmath143 and @xmath144 .
we performed simulations at @xmath145 and @xmath146 , the latter being the preferred literature value for the threshold in the one - dimensional bs model , while the former is favored by the results discussed in the following subsection .
we studied systems of 1000 , 2000 , 4000, ... ,32000 , sites in simulations of @xmath147 to @xmath148 time steps . for @xmath149 - 32000 , the results for the order parameter follow a power law with @xmath150 . the data for smaller system sizes , however ,
show systematic deviations from a pure power law , leading us to seek a correction to scaling term ; a correction decaying @xmath151 leads to a good fit .
we fit the expression @xmath152 to the data for @xmath153 , allowing @xmath9 and @xmath154 as adjustable parameters ; the best - fit values are @xmath155 and @xmath156 .
the simulation data , and the difference from best fit of eq .
( [ fitrho ] ) are plotted in fig .
1 , showing the high quality of fit .
the data for the lifetime , using system sizes of 125 , 250, ... 8000 yield @xmath157 , with no obvious correction term ( see fig .
we also determined the stationary moment ratio @xmath158 at the critical point .
the estimates for @xmath159 decrease slowly with @xmath160 , and appear , when plotted versus @xmath161 , to approach a limiting ( @xmath162 ) value of 1.030(5 ) ( see fig .
( for the one - dimensional cp , @xmath163 at the critical point . )
essentially the same results are obtained regardless of whether we use @xmath164 or 0.66702 in these simulations . scaling properties at an absorbing state phase transition
are also reflected in the spread of activity from an initially localized region @xcite . in spreading simulations of the abs model
we start the system with a single site @xmath165 and all others above this value .
( this is completely equivalent to the bs branching process studied in @xcite . ) at @xmath166 , the process generates a scale - invariant pattern of activity that may be characterized by power - laws for the survival probability @xmath142 , the mean number of active sites @xmath167 and the mean - square distance @xmath168^{-1 } \langle \sum_j r_j^2 ( t ) \rangle$ ] .
( @xmath169 denotes the position of the @xmath170 active site at time @xmath33 .
note that @xmath167 is taken over all realizations , including those that have become trapped in the absorbing state at or before time @xmath33 . )
the scaling laws are typically written in the form @xmath171 relations that have been verified to high precision for various examples @xcite .
( we use @xmath7 to denote the spreading exponent , to avoid confusion with the _ dynamical _ exponent @xmath172 . ) the appearance of power laws is commonly used to locate the critical point @xcite .
the spreading exponent @xmath3 is related to the avalanche size exponent @xmath141 , defined ( in the bs model ) via @xmath173 , where @xmath174 is the probability of an avalanche having a duration between @xmath175 and @xmath176 .
thus the survival probability @xmath177 , implying @xmath178 .
we performed spreading simulations of the abs model at @xmath62 = 0.66699 , 0.66700 , 0.66701 , 0.66702 and 0.66703 .
each realization was followed up to a maximum time of about @xmath179 ; the total number of realizations ranged from @xmath180 to @xmath181 , depending on the value of @xmath62 . to locate the critical point we study the local slope @xmath182 , plotted versus @xmath183 . for @xmath184 the local slope
is expected to veer downward at large times , and vice - versa .
( numerically , @xmath185 is given by the slope of a least - squares linear fit to the data in an interval @xmath186 $ ] , with geometric mean @xmath33 . ) on the basis of the local slope data ( see fig .
4 ) we conclude that @xmath187 .
this is consistent with previous estimates , which place the threshold at 0.66702(8 ) @xcite and 0.66702(3 ) @xcite .
( we did not find analyses of the local slopes @xmath188 or @xmath189 , defined analogously to @xmath190 , useful in locating the critical point . ) analyzing the data at the critical point , we are unable to obtain good fits to @xmath191 , @xmath30 and @xmath192 using simple power - law expressions . including a subdominant power - law correction in the relations of eq .
( [ sprexps ] ) greatly improves the quality of fit . in particular
, the survival probability can be fit quite accurately using @xmath193 where @xmath194 is a constant and the best - fit values are @xmath195 and @xmath196 .
the same value for @xmath3 is found using the data for @xmath197 .
( the choice of a correction term decaying as @xmath198 is motivated by the fact that the local slopes @xmath199 and @xmath200 are essentially linear when plotted versus @xmath198 , as seen in the inset of fig .
4 . ) in fig .
5 we plot @xmath191 and the ratio of @xmath191 to the fitting function , eq .
( [ ppwc ] ) ; the ratio is seen to be essentially constant for @xmath201 . the mean - square displacement may also be fit using an asymptotic power law and correction term .
we find @xmath202 with @xmath203 and @xmath204 .
it has been argued that @xmath205 quite generally for extremal dynamics @xcite .
our data for the one - dimensional abs model support this conclusion : on a double - logarithmic plot , @xmath167 clearly grows more slowly than a power law . while @xmath205 is compatable with @xmath167 growing without limit as @xmath206 ( for example , @xmath207 , as suggested in ref .
@xcite ) , our results support the conclusion that @xmath167 saturates at a _ finite value _
@xmath208 at long times . specifically , we are unable to fit the long - time behavior using an expression of the form @xmath209 . on the other hand , we find @xmath210 , with @xmath211 , suggesting that @xmath212 $ ] .
in fact an excellent fit is obtained using @xmath213 and @xmath214 , as can be seen in fig .
( saturation of @xmath167 does not occur on the time scale of the simulation ; for the anistropic case , shown in the inset of fig .
6 , saturation is in fact evident . ) in the absorbing phase ( @xmath215 ) , the survival probability must vanish as @xmath206 .
our data follow @xmath216 , where @xmath217 , with @xmath218 . on the other hand , for @xmath219
, the survival probability tends to a finite value as @xmath206 .
we obtain @xmath220 , with @xmath221 .
( in dp and allied models @xmath222 @xcite , but this need not hold for models in other universality classes . ) in the cp and other nonextremal models , the spread of activity in the supercritical regime follows a simple pattern : the size @xmath223 of the active region grows linearly with time , and the number of active sites @xmath30 grows @xmath224 . our observation of subdiffusive spreading ( and saturation in @xmath30 ) at the critical point lead us to investigate supercritical spreading in the abs model .
we find that spreading is indeed _
sublinear_. for example , using @xmath225 in a study extending to @xmath226 to avoid transient effects , we obtain @xmath227 with @xmath228 and @xmath229 with @xmath230 .
( the exponent governing @xmath192 should be twice that for @xmath30 , since active regions have a finite activity density in the supercritical regime . )
once again , extremal dynamics is seen to limit the growth of activity .
we performed spreading simulations of the cp3 model , using the approach described in the previous subsection .
each realization is followed up to a maximum time of @xmath231 . using the power - law behavior of @xmath142 and @xmath167 as the criterion for criticality
, we find @xmath232 for the cp3 .
( note that this is some 5% smaller than the critical value of the corresponding extremal model . ) analyzing the local slopes , we obtain @xmath233 , @xmath234 and @xmath235 .
these values are fully consistent with those for directed percolation ( see table [ tab1 ] ) , confirming that the cp3 model belongs to the same universality class as the original contact process .
a striking difference between extremal and nonextremal models with an absorbing state is that the spread of activity in the critical process is much slower in the former .
this is of course reflected in the value @xmath205 for extremal models , ( while for example @xmath236 for dp in one dimension ) , and in the subdiffusive growth in @xmath192 in the abs model . in figs .
7 and 8 we compare typical evolutions in the abs model and its nonextremal analog , the cp3 , at their respective critical points . it is evident that the activity spreads much more slowly in the abs than in the cp3 .
a further notable difference is that in the abs a site can remain active for a very long time , i.e. , while the site is not the minimum site or a neighbor of it .
thus the rates of both addition and loss of active sites are much smaller in the critical extremal process than in the corresponding nonextremal one . in light of the discussion of sec .
ii , it is of interest to study the behavior of other absorbing - state models under extremal dynamics . as a first step we report simulation results for the extremal - absorbing contact process ( cp@xmath116 ) .
we performed spreading simulations to determine @xmath78 and the exponents @xmath3 , @xmath4 and @xmath7 , using simulations running to a maximum time of @xmath231 in @xmath237 independent realizations .
we find @xmath238 for the extremal cp , compared with 0.76733 for the original ( non - extremal ) process .
( note that , as in the comparison between the cp3 and abs models , @xmath239 .
this may reflect the slower spread of activity under extremal dynamics . ) as in the case of the abs model , the decay of the survival probability at the critical point follows an expression of the form of eq .
( [ ppwc ] ) , here with best - fit parameters @xmath240 and @xmath241 .
the exponent @xmath3 is essentially the same as found for the abs model , while the correction term is about twice as large . at the critical point
the derivative @xmath242 , again indicating a behavior of the form @xmath212 $ ] , here with @xmath243 .
the mean - square distance of active sites from the origin grows in a manner similar to that in the abs model .
we are again able to fit the data for @xmath192 using an expression of the form of eq .
( [ r2pwc ] ) , with @xmath244 and @xmath245 .
these results strongly suggest that the cp@xmath116 belongs to the same universality class as the abs model .
we turn now to the rather surprising behavior of the stationary probability density @xmath82 in the extremal cp .
recall that mean - field theory predicts @xmath246 for @xmath247 , while for @xmath248 there are two steps , one at @xmath123 , the other at @xmath249 . in simulations of the cp@xmath250 on a ring
we find a single step discontinuity for @xmath251 , and , for @xmath219 , a pair of steps , one at @xmath249 , the other at @xmath252 .
the positions of the singularities as obtained in simulation ( using data for system sizes @xmath253 , 200, ... ,1600 to extrapolate the position in the infinite - size limit ) , are shown in fig .
the lower singularity @xmath254 is seen to bifurcate from @xmath166 just at the critical point , in qualitative agreement with mft .
note however that the position of the singularity is not constant for @xmath255 , as predicted by mft . the density @xmath82 is shown for @xmath256 and @xmath257 in fig .
10 . in the latter case
it is evident that the step at @xmath249 is sharp ( this is true even for small systems ) and derives from the singular nature of the updating rule .
the step at @xmath258 , by contrast , is subject to finite - size rounding , and becomes sharper with increasing system size , as is characteristic of a critical singularity .
the finite width of the peak at @xmath259 ( in the process with @xmath260 ) , appears to be a finite size effect as well : it becomes sharper with increasing @xmath160 , suggesting that the singularities merge in the limit @xmath162 .
the scaling behavior of the bak - sneppen model changes when the updating rule is asymmetric @xcite . the same critical exponents are found for a highly anisotropic version in which at each step , only the minimal site and its neighbor on the right are updated @xcite , and for weak anisotropy @xcite , so that one may identify an anisotropic bs universality class . in this section
we report results of spreading simulations of the anisotropic absorbing bs model . to obtain these results
, we simulated the anisotropic abs ( in the highly anisotropic version ) in studies extending to a maximum time of 1.6@xmath261 , using 3@xmath261 realizations . analysing the local slope @xmath182 , we determined the threshold of this model to be @xmath262 .
( the best previous estimate is @xmath263 @xcite . )
a typical evolution of the critical spreading process is shown in fig .
the local slope @xmath185 yields the estimate @xmath264 . for anisotropic models we define @xmath265 as the mean - square _ radius of gyration _ ,
i.e. , the distance is measured relative to the current center of mass of the set of active sites , rather than to a fixed origin .
this is done to eliminate a spurious contribution due simply to the overall drift in the active region . for
the anisotropic abs model @xmath265 may again be fit with an expression of the form of eq .
( [ r2pwc ] ) , with @xmath266 and @xmath267 .
the exponents @xmath3 and @xmath7 are quite different from those of the isotropic model . despite these differences
, we again find @xmath205 for the anisotropic model .
as before , the mean number of active sites @xmath167 saturates at long times , more rapidly in fact than in the isotropic abs model ( see fig . 6 , inset ) .
we are able to fit the data well using @xmath268 $ ] with parameters @xmath269 and @xmath270 .
the nonextremal model corresponding to the anisotropic abs model is a two - site contact process , cp2 , which is simply the cp3 with updating restricted to the central site and its neighbor on the right .
we have verified that the spreading exponents of the cp2 model are those of directed percolation .
( here again , we define @xmath192 as the mean - square radius of gyration . )
this leads to the interesting conclusion that a perturbation ( asymmetric updating ) that is _ irrelevant _ for a nonextremal model is relevant for the corresponding extremal system .
( we note that , because the two sites in the cp2 are updated in the same manner , the model does not fall in the so - called anisotropic - dp class , for which bonds along different axes are present with different probabilities @xcite . )
we investigate the relation between extremal dynamics , exemplified by the bak - sneppen model , and nonextremal models exhibiting a phase transition to an absorbing state , using general arguments , mean - field theory and simulation .
the relation between the bs model and directed percolation was already suggested some time ago @xcite .
here we clarify this connection by showing how a generic absorbing - state model can be transformed to an extremal one via the associated extremal - absorbing model .
the nonextremal ` precursor ' of the bs model is a three - site contact process @xcite , cp3 , which , like the original cp , belongs to the directed percolation universality class .
the bs model and the extremal version of the cp belong to a common universality class that may be dubbed ` extremal dp ' ( edp ) .
a number of extremal dynamics classes distinct from edp are discussed in @xcite ; another example is the anisotropic bs model . we expect that further extremal dynamics universality classes exist , for example an extremal parity - conserving class @xcite , although examples of the latter have yet to be studied .
our results for the critical exponents of the edp class , which includes the bs model and the extremal cp , are compared against those of ordinary dp ( in one spatial dimension ) in table i. ( here we have taken the values @xmath205 and @xmath271 to be _ exact _ for edp . ) the differences between the two sets of exponent values are evident .
our results @xmath272 , and @xmath273 are in agreement with the earlier estimates @xcite of 1.07(1 ) and 2.10(5 ) , respectively .
our result is however somewhat higher than grassberger s result @xmath274 @xcite .
certain scaling relations are expected to hold among the critical exponents @xcite . in spreading processes
one expects @xmath275 ; our data are nearly consistent with this , yielding @xmath276 .
the relation @xmath277 is also satisfied : our data yield @xmath278 . finally , we consider the generalized hyperscaling relation @xcite @xmath279 in @xmath14 dimensions . using our data , we find the difference between the two sides of this relation to be 0.09(6 ) .
our results are marred by another inconsistency that may reflect corrections to scaling or finite size effects : the product @xmath280 , with the first two factors determined from finite - size scaling at the critical point , and the final factor obtained from the decay of the survival probability in the subcritical regime , should equal @xmath271 ; our data yield 1.08(3 ) .
these minor inconsistencies suggest that one or more of the exponents may be in error by 5% or so . refining their values
will require accumulating larger data sets in simulations of larger systems , a task we leave for future work .
( the studies reported here were quite demanding computationally , representing approximately 6 months cpu time on an alpha workstation . )
in the course of our study we revisit a three - site contact process ( cp3 ) that is the nonextremal analog of the bs model @xcite .
we verify that the cp3 belongs to the universality class of directed percolation , as expected @xcite .
similarly , we define extremal and extremal - absorbing versions of the original contact process ( cp@xmath250 and cp@xmath116 , respectively ) and verify that their scaling properties are the same as those of the bs model . the stationary probability density for the cp@xmath250 follows , in general terms , the predictions of mean - field theory , but certain interesting differences exist , as detailed in sec .
it is clear that when an absorbing - state model is modified to follow extremal dynamics , its critical exponents are altered . in general extremal dynamics tens to slow the spread of activity in the critical and supercritical regimes .
one may nevertheless inquire whether any more general features of the original model are preserved under ` extremalization ' .
a candidate for such a conserved property is the critical dimension @xmath281 . in critical phenomena , various universality classes ( differing in the symmetry group of the order parameter , or the presence of conserved quantities ) may share the same @xmath281 if the algebraic structure of their continuum description ( in particular , the power of the lowest - order nonlinear term in the order parameter , in a landau - ginzburg - wilson effective hamiltonian ) is the same .
thus @xmath106 for all models in the @xmath30-vector family .
extending this idea to extremal models is questionable , since there is no continuum description at hand .
( at first glance , the notion of extremal dynamics in a description using a _ continuous _ activity density seems problematic , since there is always one and only one extremal site . )
be that as it may , it seems plausible that if the field theory for dp @xcite were somehow modified to reflect extremal dynamics , the dominant nonlinearity would not change , so that @xmath281 would retain its value of four , as in dp .
the upper critical dimension @xmath282 for the bs model was established some time ago by boettcher and paczuski @xcite .
our argument suggests that extremal versions of other absorbing - state models have the same upper critical dimension as the corresponding nonextremal model .
we hope to test this prediction in future work .
studying the anisotropic abs model and its nonextremal counterpart , the cp2 model , we find that anisotropy is a relevant perturbation for extremal dp , while it is irrelevant for the corresponding nonextremal class .
we suspect that other perturbations , such as diffusion , may exhibit a similar pattern of relevance .
we are grateful to stefano zapperi and jafferson kamphorst leal da silva for informative discussions , and to paulo m. c. de oliveira for helpful comments on the manuscript .
we acknowledge cnpq and fapemig , brazil , for financial support .
p. grassberger , f. krause , and t. von der twer , j. phys .
a * 17 * , l105 ( 1984 ) ; p. grassberger , _ ibid_. * 22 * , l1103 ( 1989 ) ; h. takayasu and a. yu .
tretyakov , phys .
lett . * 68 * , 3060 ( 1992 ) ; i. jensen , phys .
e * 50 * , 3623 ( 1994 ) .
there are several ways of formulating a mean - field theory ( mft ) for extremal models .
first we consider an approach @xcite in which the probability of a site @xmath17 being chosen as the central site is proportional to @xmath284 ; extremal dynamics is recovered in the limit @xmath285 .
( in the present discussion the parameter @xmath105 bears no relation to the critical exponent denoted by the same symbol in the main text . )
applied to the bs model , this approach yields the distribution @xmath286 when @xmath285 @xcite . in the abs model
the distribution @xmath287 evolves via @xmath288 the first term represents a site with value @xmath65 being selected as the central site , which is only possible if @xmath67 . the second term
reflects updating three sites with new variables uniform on @xmath289 $ ] , with the integral representing the overall rate of events .
the final term represents updating of the two neighboring sites , and is obtained using the mean - field factorization of the nearest - neighbor joint probability density : @xmath290 .
note that in writing eq .
( [ mf1 ] ) we have associated a time increment @xmath291 , with @xmath13 the number of sites , with each event .
( [ mf1 ] ) admits an infinite set of stationary solutions for which @xmath292 on the interval @xmath293 .
these represent absorbing states . to seek an active stationary solution we let @xmath294 , and equate the time derivative to zero , yielding @xmath295 to determine @xmath296 we multiply eq .
( [ ss1 ] ) by @xmath297 and integrate from @xmath298 to @xmath249 , leading to @xmath299 $ ] , so that @xmath300 in the limit @xmath285 , we find , for @xmath301 , the singular density @xmath302 .
this is precisely the mf result for the original bs model .
[ when we take @xmath285 , the above expression reduces to @xmath303 for @xmath304 .
but this density is not normalized on @xmath289 $ ] and so must be rejected .
we are left with only absorbing stationary solutions for @xmath304 . ]
thus @xmath305 in the mft of the absorbing bak - sneppen model .
note that the parameter @xmath62 is irrelevant for @xmath306 , as was argued in sec .
a moment s reflection shows that the evolution of @xmath287 in the ( nonextremal ) cp3 model is given by eq .
( [ mf1 ] ) with @xmath307 , since all active sites are then equally likely to be chosen as the central site . taking the limit @xmath308 of the stationary solution , eq .
( [ mfsta ] ) , one finds , for @xmath309 , the stationary density @xmath310 eq .
( [ cp3st ] ) confirms that the stationary density of a ( nonextremal ) model exhibiting an absorbing state phase transition is characterized by a steplike singularity , as asserted in sec .
ii . for @xmath311 , eq . ( [ cp3st ] ) yields an unphysical , negative density , showing that @xmath312 for the cp3 , in the mean - field approximation .
the foregoing analysis is readily extended to the extremal - absorbing contact process ( cp@xmath116 ) defined in sec .
ii . the rate of events is again given by @xmath313 . at each event
, there is a probability of 1/2 that the central site ( which must have @xmath67 ) is replaced , while with probability 1/2 a neighbor is updated .
thus the loss terms in the equation for @xmath314 are : @xmath315p(x)$ ] .
the gain term corresponding to updating of the central site is simply @xmath316 , but for updating a neighbor it is @xmath317 $ ] , where @xmath318 is the probability that a given site @xmath17 has @xmath319 .
( the reason is that when updating an _ active _ neighbor , the new variable is chosen from the distribution uniform on @xmath71 $ ] . )
thus the mf equation of motion is to find the stationary solution @xmath82 we first note that for @xmath321 , setting @xmath322 to zero yields @xmath323 where @xmath324 . integrating eq .
( [ pxs ] ) from @xmath298 to @xmath249 , we find @xmath325 if we now multiply eq .
( [ pxs ] ) by @xmath297 and integrate over the same interval , we obtain @xmath326 leading to @xmath327 in the stationary state .
we see that @xmath328 as in the mft of the original contact process .
evaluating the integral in eq .
( [ int ] ) one finds @xmath329 , where @xmath330 . the stationary density is @xmath331 } { ( e^{-\beta \kappa } - e^{-\beta q } ) + e^{-\beta x}(1 - e^{-\beta \kappa } ) \theta(q - x ) } \label{pxscp}\ ] ] for @xmath248 , we have @xmath332 , and in the limit @xmath285 , @xmath333 which is normalized and exhibits step - function singularities at @xmath334 and @xmath94 . for @xmath335 on the other hand , @xmath336 and eq.([pxscp ] )
does not yield an acceptable probability density , and we conclude that the only stationary state is the absorbing one .
the critical point of the cp@xmath116 thus falls at @xmath328 in mf approximation . finally , for the extremal cp , the equation of motion is @xmath338 \label{mf3}\ ] ] with @xmath339 . to find the stationary solution we write @xmath340p(q)}{1 + i^{-1}e^{-\beta x } }
\label{cpxest}\ ] ] integrating from @xmath341 to @xmath62 and solving for @xmath342 we find @xmath343 where @xmath344 now multiply eq .
( [ cpxest ] ) by @xmath297 and integrate from 0 to 1 to obtain @xmath345 where @xmath346 $ ] and @xmath347 .
if @xmath348 , the first term on the r.h.s . of eq .
( [ one ] ) is nonzero and dominates as @xmath285 .
equating the first term to unity then leads to @xmath349 , and thence to @xmath327 which is positive for @xmath248 .
a simple calculation then yields the distribution of eq .
( [ pxscpext ] ) in the limit @xmath285 . if @xmath335 the above solution is not valid since it implies @xmath350 .
we therefore take @xmath351 , implying @xmath352 , and so @xmath353 .
( [ one ] ) now reads @xmath354 solving for @xmath296 and inserting the result in eq .
( [ cpxest ] ) , we find in this case @xmath355 .
these results have been verified via numerical integration .
another approach to formulating mft for the bs model considers extremal dynamics on an @xmath13-site complete graph or random - neighbor model ( two neighbors are selected at random each time a site is updated ) ; the stationary density @xmath82 becomes a step function in the infinite - size limit @xcite .
we now extend this approach to the abs model .
let @xmath356 = \int_0^x p(y ) dy$ ] be the distribution function and let @xmath357
. by definition @xmath358 is a non - increasing function with @xmath359 and @xmath360 , since @xmath361 outside the interval [ 0,1 ] .
activity in the abs model is predicated on the minimal site @xmath127 being smaller than @xmath62 ; the probability of this event , under the mf factorization , is @xmath362^n$ ] . given @xmath117 , updating the extremal site and two neighbors results , on the average , in the increment : @xmath363 - 2p(x ) + 3x \}$ ] , where @xmath364 , so that the first term represents loss of the minimal site , the second removal of two neighbors , and the third random replacement of the three site variables with numbers uniform on [ 0,1 ] .
if we adopt a time increment @xmath291 for each such event , the equation of motion for @xmath365 is @xmath366 + [ 1 - q(q , t)^n][3x - 2p(x , t ) ] \label{dpdt1}\ ] ] note that the evolution ceases if @xmath367 , i.e. , if there are no active sites .
( since @xmath358 is nonincreasing @xmath368 . ) for @xmath301 , the stationary solution to eq .
( [ dpdt1 ] ) corresponds to a density @xmath82 that approaches a step function , @xmath369 , as @xmath112 .
a simple calculation yields the dominant contribution for large @xmath13 : @xmath370 ( one should note however that the convergence is nonuniform in @xmath65 , being slower the closer @xmath65 is to the critical value of 1/3 . ) for @xmath304 we are unable to find an accepable stationary solution with @xmath371 ( i.e. , @xmath372 ) , for @xmath67 , and conclude that only absorbing solutions exist .
the analysis of the abs model on a complete graph confirms that in the infinite - size limit , the model enjoys the usual properties of the bs model for @xmath306 , and falls into the absorbing state for @xmath304 . the evolution of @xmath373 in the extremal cp follows , in mf approximation , the equation @xmath374 + \frac{x}{2 } - \frac{1}{2}p(x , t ) + \frac{1}{2 } [ x^ * p(q , t ) +
x q(p , t ) ] \label{dpdt2}\ ] ] where @xmath375 .
numerical integration shows that the solution converges , for large @xmath13 , to a stationary distribution consistent with the singular density found above in the limit @xmath285 .
1 . stationary activity density ( filled symbols ) @xmath140 versus system size @xmath160 in the one - dimensional abs model at the critical point .
open symbols : difference between @xmath140 and the fitting function , eq . ( [ fitrho ] ) , shifted vertically for visibility . fig
mean number of active sites @xmath167 in the abs model at the critical point , @xmath381 .
the solid curve represents the fitting function described in the text .
inset : a similar plot , for the critical _ anisotropic _ abs model . fig
position @xmath254 of the singularity in the stationary probability density of the extremal cp .
the upper line of singularities , @xmath249 , bifurcates from @xmath254 at the critical value @xmath78 . | extremal dynamics represents a path to self - organized criticality in which the order parameter is tuned to a value of zero .
the order parameter is associated with a phase transition to an absorbing state .
given a process that exhibits a phase transition to an absorbing state , we define an extremal absorbing " process , providing the link to the associated extremal ( nonabsorbing ) process .
stationary properties of the latter correspond to those at the absorbing - state phase transition in the former .
studying the absorbing version of an extremal dynamics model allows to determine certain critical exponents that are not otherwise accessible . in the case of the bak - sneppen ( bs ) model , the absorbing version is closely related to the @xmath0-avalanche " introduced by paczuski , maslov and bak [ phys .
rev .
e * 53 * , 414 ( 1996 ) ] , or , in spreading simulations to the bs branching process " also studied by these authors .
the corresponding nonextremal process belongs to the directed percolation universality class .
we revisit the absorbing bs model , obtaining refined estimates for the threshold and critical exponents in one dimension .
we also study an extremal version of the usual contact process , using mean - field theory and simulation .
the extremal condition slows the spread of activity and modifies the critical behavior radically , defining an
extremal directed percolation " universality class of absorbing - state phase transitions .
asymmetric updating is a relevant perturbation for this class , even though it is irrelevant for the corresponding nonextremal class .
@xmath1 electronic address : [email protected] + @xmath2 electronic address : [email protected] 0.5truecm |
the universality of random matrices is usually divided into bulk and edge universalities . edge universality concerns the distribution of the extreme eigenvalues . it is known that the extreme eigenvalues of a large class of wigner matrices exhibit universal limiting behavior .
the limiting distribution of the largest eigenvalue was first identified by tracy and widom @xcite for the gaussian ensembles .
edge universality for wigner matrices has first been proved by soshnikov @xcite ( see also @xcite ) for real symmetric and complex hermitian ensembles with symmetric distributions .
the symmetry assumption on the entries distribution was partially removed in @xcite .
edge universality without any symmetry assumption was proved in @xcite under the condition that the distribution of the matrix elements has subexponential decay and its first three moments match those of the gaussian distribution .
for wigner matrices with arbitrary symmetry class , edge universality was proven in @xcite under the assumption that the entries have @xmath12 moments .
recently , a necessary and sufficient condition for the edge universality of wigner matrices was given in @xcite .
the distribution of the largest eigenvalues of a random diagonal matrix @xmath4 whose entries @xmath13 are i.i.d .
real random variables is given by the order statistics of @xmath13 .
the fisher - tippett - gnedenko theorem ( see e.g. @xcite ) thus implies that the limiting distribution of the largest eigenvalue of @xmath4 belongs either to the gumbel , frchet or weibull family . in this paper
, we consider the interpolation between wigner matrices and real diagonal random matrices .
let @xmath2 be an @xmath14 real symmetric or complex hermitian wigner matrix whose centered entries have variance @xmath15 and subexponential decay .
let @xmath4 be an @xmath14 real diagonal random matrix whose entries are bounded i.i.d .
random variables .
for @xmath1 we set @xmath16 the matrices @xmath4 and @xmath2 are normalized in the sense that the eigenvalues of @xmath4 and @xmath2 are of order one .
if @xmath2 belongs to the gaussian unitary ensemble ( gue ) , the model is called the deformed gue .
it was shown in @xcite that the edge eigenvalues of the deformed gue are governed by the tracy - widom distribution for @xmath17 . at @xmath18
the fluctuations of the edge eigenvalues change from the tracy - widom to a gaussian distribution .
more precisely , johansson showed in @xcite that the limiting distribution of the edge eigenvalues for @xmath19 is given by the convolution of the tracy - widom and the centered gaussian distribution , with variance depending on @xmath20 .
these results have not been established for the gaussian orthogonal ensemble ( goe ) or for general wigner matrices . in the present paper ,
we consider the edge behavior of the deformed model in the regime @xmath21 with @xmath2 a real symmetric or complex hermitian matrix .
we show that there is , for certain @xmath4 , yet another transition for the limiting behavior of the largest eigenvalues of @xmath8 as @xmath22 varies . for simplicity , we assume that the distribution of the entries of @xmath4 is centered and given by the density @xmath23}(v)\,,\end{aligned}\ ] ] where @xmath24
, @xmath25 is a strictly positive @xmath26-function and @xmath27 is a normalization constant .
we primarily focus on the choices @xmath28 . from our first main result ,
theorem [ thm : main ] , it follows that there are @xmath3-independent constants @xmath29 and @xmath30 , such that , for @xmath31 and @xmath32 , the largest eigenvalue @xmath33 of @xmath8 satisfies @xmath34 where @xmath35 is a _ weibull distribution _ with parameter @xmath36 ; see .
however , if @xmath11 , then there are @xmath3-independent constants @xmath37 and @xmath38 , such that @xmath39 where @xmath40 denotes the cumulative distribution function of the centered _ gaussian distribution _ with variance @xmath41 ; see appendix c. we remark that neither nor depend on the symmetry type of the wigner matrix @xmath2 . the appearance of the weibull distribution in the model is indeed expected when @xmath22 grows sufficiently fast with @xmath3 , since in this case the diagonal matrix dominates the spectral properties of @xmath8 .
however , it is quite surprising that the weibull distribution already appears for @xmath22 order one , since the local behavior of the eigenvalues in the bulk of the deformed model mainly stems from the wigner part , and the contribution from the random diagonal part is limited to mesoscopic fluctuations of the eigenvalues ; see @xcite .
having identified two possible limiting distributions of the largest eigenvalues , it is natural to ask about the behavior of the associated eigenvectors . before considering the deformed model ,
we recall that the eigenvectors of wigner matrices with subexponential decay are completely delocalized , as was proved by erds , schlein and yau @xcite . for deformed wigner matrices
we show that the eigenvectors of the largest eigenvalues are _ partially localized _ in the regime where the edge behavior holds .
more precisely , we prove that one component of the ( @xmath42-normalized ) eigenvectors associated with eigenvalues at the extreme edge carries a weight of order one , while the other components each carry a weight of order @xmath43 ; see theorem [ thm : local ] .
if , however , the edge behavior holds , all eigenvectors are completely delocalized .
although we do not prove it explicitly , we claim that the bulk eigenvectors of the model with for the choice of @xmath44 , are completely delocalized ( for any choice of @xmath21 ) .
this can be proved with the very same methods as in @xcite . to understand the transition from partial localization to delocalization
, further efforts are required .
the phenomenology described above is reminiscent of the one for heavy - tailed wigner matrices .
for instance , consider real symmetric wigner matrices whose entries distribution function decays as a power law , i.e. , @xmath45 for some slowly varying function @xmath46 .
it was proved by soshnikov @xcite that the linear statistics of the largest eigenvalues are poissonian for @xmath47 ; in particular , the largest eigenvalue has a frchet limit distribution .
later , auffinger , ben arous and pch @xcite showed that the same conclusions hold for @xmath48 as well .
recently , it was proved by bordenave and guionnet @xcite that the eigenvectors of models satisfying are weakly delocalized for @xmath49 . for @xmath50 , it is conjectured @xcite that there is a sharp `` metal - insulator '' transition . in @xcite it
is proved that the eigenvectors of sufficiently large eigenvalues are weakly localized for @xmath51 . to clarify the terminology `` partial localization ''
we remark that it is quite different from the usual notion of localization for random schrdinger operators .
the telltale signature of localization for random schrdinger operators is an exponential decay of off - diagonal green function entries , which implies the absence of diffusion , spectral localization etc .
for the anderson model in dimensions @xmath52 such an exponential decay was first obtained by frhlich and spencer @xcite using a multiscale analysis .
later , a similar bound was presented by aizenman and molchanov @xcite using fractional moments . due to the mean - field nature of the wigner matrix @xmath2 ,
there is no notion of distance for the deformed model . instead , our localization result states that most of the mass of the eigenvectors is concentrated on a few sites , whose locations are independent and uniformly distributed .
this result agrees with the predictions of formal perturbation theory . yet
, there are some similarities with the anderson model in @xmath52 : in the anderson model localization occurs where the density of states is ( exponentially ) small @xcite ; this is known to happen close to the spectral edges or for large disorder .
further it is strongly believed that the anderson model admits extended states , i.e. , the generalized eigenvectors in the bulk are expected to be delocalized .
moreover , it was proven by minami @xcite that the local eigenvalue statistics of the anderson model can be described by a poisson point process in the strong localization regime .
it is also conjectured that the local eigenvalue statistics in the bulk are given by the goe statistics , respectively gue statistics if time - reversal symmetry is broken .
we remark that there are some partial results on bulk universality for the deformed model ; see @xcite .
we also mention that the localization result we prove in this paper also differs from that for random band matrices , where all the eigenvectors are localized , even in the bulk .
we refer to @xcite for more details on the localization / delocalization for random band matrices .
next , we outline the proofs of our main results .
it was first shown by pastur @xcite that the empirical eigenvalue distribution of the deformed model converges to a deterministic distribution in the limit @xmath10 under some weak assumption on @xmath6 . however , this limiting eigenvalue distribution , referred to as the _ deformed semicircle law _ in the following , is in general different from wigner s semicircle law and depends on the limiting distribution of @xmath6 .
the deformed semicircle law can be defined in terms of a functional equation for the stieltjes transforms of the limiting eigenvalue distributions of @xmath6 and @xmath2 @xcite . restricting the discussion to the special case
when the entries of @xmath4 follow the centered jacobi distribution in with @xmath31 , we showed in @xcite that the deformed semicircle law , henceforth denoted by @xmath53 , is supported on a single interval and shows either of the following behavior close to the upper edge : @xmath54 for @xmath55 , @xmath56 , where @xmath57 denotes the distance from @xmath58 to the upper endpoint of the support of @xmath53 ; see lemma [ general case - large lambda ] below . in case
the square root behavior prevails , we are going to show that the largest eigenvalue of @xmath8 satisfies , whereas in case we have a `` convex decay '' with @xmath31 , is satisfied . in a first step
, we derive a _ local law _ for the empirical eigenvalue density : under some moment conditions the convergence of the empirical eigenvalue distribution to the semicircle law also holds on very small scales . denoting by @xmath59 , @xmath60 , the green function or resolvent of @xmath8 , convergence of the empirical eigenvalue distribution on scale @xmath61 around an energy @xmath62 is equivalent to the convergence of the averaged green function @xmath63 , @xmath64 . in a series of papers
@xcite erds , schlein and yau showed that the semicircle law for wigner matrices also holds down to the optimal scale @xmath65 , up to logarithmic corrections . in
@xcite a `` fluctuation average lemma '' was introduced that yielded optimal bounds on the convergence of @xmath66 for wigner matrices in the bulk @xcite and up to the edge @xcite on scales @xmath67 . below this scale
the eigenvalue density remains fluctuating even for large @xmath3 . in @xcite the green function @xmath68 and its average @xmath66
have been used to prove edge universality for generalized wigner matrices . in @xcite we derived a _ local deformed semicircle law _ for the deformed ensemble under the assumption that @xmath53 has a square root behavior at the endpoints .
in the present paper , we derive a local law at the _ extreme edge _ in case @xmath53 shows a convex decay at the edge .
we propose , however , a slightly different path than the one taken in @xcite : we condition on the random variables @xmath13 and show that @xmath66 converges , for `` typical '' realizations of @xmath13 , on scale @xmath69 ; see proposition [ prop : step 2_4 ] .
in particular , we show that the typical eigenvalue spacing at the extreme edge is of order @xmath70 ; as is suggested by the convex decay in .
similar to the wigner case , see e.g. , @xcite , this is accomplished by deriving a self - consistent equation for @xmath66 .
however , the analysis of the self - consistent equation is quite different from the wigner case , due to the absence of the usual stability bound ; see @xcite . in a second step
, we can use the self - averaging property of the wigner matrix @xmath2 , to show that the _ imaginary part _ of @xmath71 can be controlled on scales much smaller than @xmath72 : the technical input here is the `` fluctuation averaging lemma '' @xcite .
our proof relies on the basic strategy of @xcite . however , in our setup the diagonal entries of @xmath73 are not uniformly bounded , which requires several changes to previous arguments . to complete the proof of our first main result , theorem [ thm : local ]
, we note that the imaginary part of @xmath71 can be written as @xmath74 where @xmath75 are the eigenvalues of @xmath8 .
thus , having control on the left side for @xmath76 allows tracking the individual eigenvalues at the extreme edge , where their typical spacing much bigger than @xmath72 . finally , we point out the main steps in the proof of the partial localization of eigenvectors ; see theorem [ thm : local ] for precise results .
it is well - known that information on the averaged green function @xmath66 can be translated via the helffer - sjstrand formula to information on the density of states ; see , e.g. , @xcite .
since the typical eigenvalue spacing at the edge is , for the case at hand , much larger than @xmath72 , the helffer - sjstrand formula also allows to translate information on the diagonal green function entries @xmath77 into information on the eigenvectors at the edge . relying on estimates on the green function
, we can then prove `` partial localization '' of the eigenvectors at the edge .
the paper is organized as follows : in section [ definition and results ] , we introduce the precise definition of the model and state the main results of the paper . in section [ prelim ]
, we collect basic notations and identities for the resolvent of @xmath8 . in section [ mfc and hat mfc ] , we prove the first main result of the paper , theorem [ thm : main ] , using estimates on the stieltjes transform of the deformed semicircle measure .
( see also corollary [ cor : main ] . ) in sections [ location ] and [ sec : zlemma ] , we prove important lemmas on the location of the extreme eigenvalues , including the local law , which have crucial roles in the proof of theorem [ thm : main ] . in section
[ sec : local ] , we prove the second main result of the paper , theorem [ thm : local ] , on the partial localization of the eigenvectors at the edge .
proofs of some technical lemmas are collected in the appendices a , b and c. _ acknowledgements : _ we thank horng - tzer yau for numerous helpful discussions and remarks .
we are also grateful to paul bourgade , lszl erds and antti knowles for discussions and comments .
ji oon lee is partially supported by the basic science research program of the national research foundation of korea , grant 2011 - 0013474 .
the stay of kevin schnelli at ias is supported by the fund for math .
in this section , we define our model and state our main results . for a ( probability ) measure , @xmath78 , on @xmath79 , we define its stieltjes transform by @xmath80 note that @xmath81 is an analytic function in the upper half plane , satisfying @xmath82 , @xmath83 . as first shown in @xcite , the stieltjes transform of the limiting spectral distribution of the interpolating model satisfies the equation @xmath84 where @xmath44 is the distribution of the i.i.d .
random variables @xmath13 .
equation is often called the pastur relation .
it is shown in @xcite that has a unique solution .
moreover , it is easy to check that @xmath85 , thus @xmath86 determines an absolutely continuous probability measure on @xmath79 , whose density , @xmath53 , is given by @xmath87 the measure @xmath53 has been studied in details in @xcite ; for example , it was shown that @xmath53 is an analytic function inside its support .
setting @xmath88 , reduces to @xmath89 and one immediately checks that in this case @xmath53 is given , as expected , by the standard semicircular measure , @xmath90 , which is characterized by the density @xmath91 .
the measure @xmath53 is often called the _ additive free convolution _ of the semicircular law and the measure @xmath44 ( up to the scaling by @xmath22 ) . more generally , the additive free convolution of two ( probability ) measures @xmath92 and @xmath93 , usually denoted by @xmath94 , is defined as the distribution of the sum of two freely independent non - commutative random variables , having distributions @xmath92 , @xmath93 respectively ; we refer to @xcite . similarly to , the free convolution measure @xmath94 can be described in terms of a set of functional equations for the stieltjes transforms ; see @xcite . for a discussion of regularity properties of @xmath94
we refer to @xcite .
free probability theory turned out to be a natural setting for studying global laws for such ensembles ; see , e.g. , @xcite . for more recent treatments , including local laws
, we refer to @xcite .
[ assumption wigner ]
let @xmath2 be an @xmath14 random matrix , whose entries , @xmath95 , are independent , up to the symmetry constraint @xmath96 , centered , real ( complex ) random variables with variance @xmath15 and with subexponential decay , i.e. , @xmath97 for some positive constants @xmath98 and @xmath99 .
in particular , if @xmath95 are complex random variables , @xmath100 if @xmath95 are real random variables , @xmath101 let @xmath4 be an @xmath14 diagonal random matrix , whose entries @xmath13 are real , centered , i.i.d .
random variables , independent of @xmath102 , with law @xmath44 .
more assumptions on @xmath44 will be stated below . without loss of generality , we assume that the entries of @xmath4 are ordered , @xmath103 for @xmath104 , we consider the random matrix @xmath105 we choose for simplicity @xmath44 as a jacobi measure , i.e. , @xmath44 is described in terms of its density @xmath106}(v)\,,\end{aligned}\ ] ] where @xmath107 , @xmath108)$ ] such that @xmath109 , @xmath110 $ ] , and @xmath27 is an appropriately chosen normalization constant . we assume , for simplicity of the arguments , that @xmath44 is centered , but this condition can easily be relaxed .
we remark that the measure @xmath44 has support @xmath111 $ ] , but we observe that varying @xmath22 is equivalent to changing the support of @xmath44 . since @xmath44 is absolutely continuous , we may assume that holds with strict inequalities .
properties of @xmath53 with the special choice for @xmath44 and with @xmath21 , have been studied in @xcite ; see also @xcite .
for example , the support of @xmath53 consists of a single interval . for @xmath62 ,
we denote by @xmath57 the distance to the endpoints support of @xmath53 , i.e. , @xmath112\,,\quad\qquad ( e\in{\mathbb{r}})\,.\end{aligned}\ ] ] in the following we will often abbreviate @xmath113 . in the present paper , we are mainly interested in the limiting behavior of the largest , respectively smallest , eigenvalues of the interpolating matrix , for @xmath21 . for concreteness , we focus on the upper edge and comment on the lower edge in remark [ remark about lower edge ] . the following lemma is taken from @xcite ; see also @xcite for statement ( 1 ) .
[ general case - large lambda ] let @xmath44 be a centered jacobi measure defined in with @xmath31 . define @xmath114 then , there exist @xmath115 such that the support of @xmath53 is @xmath116 $ ] . moreover , * if @xmath11 , then for @xmath117 , @xmath118 for some @xmath119 ; * if @xmath9 , then @xmath120 and , for @xmath117 , @xmath121 for some @xmath119 .
moreover , @xmath122 satisfies @xmath123 .
since @xmath124 we find that @xmath125 .
similarly , we also have that @xmath126 .
[ remark about lower edge ] for @xmath127 , the analogue statements to lemma [ general case - large lambda ] hold for the lower endpoint @xmath128 of the support of @xmath53 , with @xmath129 and @xmath130 replaced by @xmath131 when @xmath132 , there exist , for any @xmath1 , @xmath133 , such that @xmath134 $ ] .
moreover , for any @xmath1 , there exists @xmath135 such that @xmath136\,.\end{aligned}\ ] ] if @xmath137 or @xmath138 , the analogous statement to holds only at the lower edge or at the upper edge , respectively .
these results can be proved using the methods of @xcite ; see @xcite for more details . in @xcite , spectral properties of the interpolating matrix
have been analyzed in detail under the assumption that holds , i.e. , it was assumed that either @xmath22 is sufficiently small or @xmath139 .
denote by @xmath140 the ordered eigenvalues of the matrix @xmath141 , @xmath142 in the following , we fix some @xmath143 , independent of @xmath3 , and consider the largest eigenvalues @xmath144 of @xmath8 .
all our results also apply mutatis mutandis to the smallest eigenvalues @xmath145 of @xmath8 as can readily be checked .
the first main result of the paper shows that the locations of the extreme eigenvalues are determined by the order statistics of the diagonal elements @xmath13 .
recall that we denote by @xmath44 the distribution of the ( unordered ) centered random variables @xmath13 .
[ thm : main ] let @xmath2 be a real symmetric or complex hermitian wigner matrix , satisfying the assumptions in definition [ assumption wigner ] .
assume that the distribution @xmath44 is given by with @xmath146 and fix some @xmath9 ; see .
let @xmath147 be a fixed constant independent of @xmath3 , denote by @xmath148 the @xmath149-th largest eigenvalue of @xmath141 and let @xmath150 .
then the joint distribution function of the @xmath151 largest rescaled eigenvalues , @xmath152 converges to the joint distribution function of the @xmath151 largest rescaled order statistics of @xmath13 , @xmath153 as @xmath154 , where @xmath155 . in particular
, the cumulative distribution function of the rescaled largest eigenvalue @xmath156 converges to the cumulative distribution function of the weibull distribution , @xmath157 where @xmath158 in section [ location ] we obtain estimates on the speed of convergence of ; see corollary [ cor : main ] .
[ rem : gaussian ] for @xmath9 , the typical size of the fluctuations of the largest eigenvalues is of order @xmath159 ( with @xmath31 ) as we can see from theorem [ thm : main ] . for @xmath160 , on the other hand , the fluctuations for the largest eigenvalue become , in the limit @xmath10 , gaussian with standard deviation of order @xmath72 .
( see appendix c for more detail . )
theorem [ thm : main ] shows that the extreme eigenvalues of @xmath8 become , for @xmath9 , uncorrelated in the limit @xmath10 .
in fact , extending the methods presented in this paper ( by choosing @xmath161 ) , one can show that the point process defined by the ( unordered ) rescaled extreme eigenvalues of @xmath8 converges in distribution to an inhomogeneous poisson point process on @xmath162 with intensity function determined by @xmath22 and @xmath44 .
our second main result asserts that the eigenvectors associated with the largest eigenvalues are `` partially localized '' for @xmath9 .
we denote by @xmath163 the components of the eigenvector @xmath164 associated to the eigenvalue @xmath165 .
all eigenvectors are normalized as @xmath166 .
[ thm : local ] let @xmath2 be a real symmetric or complex hermitian wigner matrix satisfying the assumptions in definition [ assumption wigner ] .
assume that the distribution @xmath44 is given by with @xmath146 and fix some @xmath9 ; see .
let @xmath147 be a fixed constant independent of @xmath3 . then there exist constants @xmath167 , depending only on @xmath168 , @xmath22 and @xmath44 , such that @xmath169 and @xmath170 in section [ sec : local ] we obtain explicit expressions for the constants @xmath167 .
[ rem : bulk ] in the preceding paper @xcite , we proved that all eigenvectors are completely delocalized when @xmath160 .
this shows the existence of a sharp transition from the partial localization to the complete delocalization regime .
we say that an eigenvalue @xmath148 is in the bulk of the spectrum of @xmath8 if @xmath171 $ ] , for any ( small ) @xmath172 and sufficiently large @xmath3 . following the proof in @xcite ,
can prove that the eigenvectors associated to eigenvalues in the bulk are completely delocalized if @xmath173 .
assuming that @xmath44 is given by with @xmath174 ( and @xmath175 ) , we showed in @xcite that all eigenvectors of @xmath8 are completely delocalized up to the edge .
theorems [ thm : main ] and [ thm : local ] remain valid for deterministic potentials @xmath4 , provided the entires @xmath13 satisfy some suitable assumptions ; see definition [ v assumptions ] in section [ mfc and hat mfc ] for details .
[ rem : perturbation ] from we find @xmath176 which is in accordance with the fact that holds and that , typically , @xmath177 where we used . in the remaining sections , we prove theorems [ thm : main ] and [ thm : local ] .
we state our proofs for complex hermitian matrices .
the real symmetric case can be dealt with in the same way .
in this section , we collect basic notations and identities . for high probability estimates we use two parameters @xmath178 and @xmath179 :
we let @xmath180 for some fixed constant @xmath119 .
we say an event @xmath181 has @xmath182-high probability , if @xmath183 for @xmath3 sufficiently large .
similarly , for a given event @xmath184 we say an event @xmath181 holds with @xmath182-high probability on @xmath184 , if @xmath185 for @xmath3 sufficiently large . for brevity
, we occasionally say an event holds with high probability , when we mean with @xmath182-high probability .
we do not keep track of the explicit value of @xmath186 in the following , allowing @xmath186 to decrease from line to line such that @xmath187 . from our proof
it becomes apparent that such reductions occur only finitely many times .
we define the resolvent , or _
green function _ , @xmath68 , and the averaged green function , @xmath66 , of @xmath8 by @xmath188 frequently , we abbreviate @xmath189 , @xmath190 , etc .
we refer to @xmath191 as spectral parameter and often write @xmath64 , @xmath62 , @xmath192 .
we will use double brackets to denote the index set , i.e. , for @xmath193 , @xmath194 \cap { \mathbb{z}}\,.\ ] ] we use the symbols @xmath195 and @xmath196 for the standard big - o and little - o notation .
the notations @xmath197 , @xmath198 , @xmath199 , @xmath200 , refer to the limit @xmath154 unless otherwise stated , where the notation @xmath201 means @xmath202 .
we use @xmath41 and @xmath203 to denote positive constants that do not depend on @xmath3 .
their value may change from line to line . finally , we write @xmath204 , if there is @xmath119 such that @xmath205 , and , occasionally , we write for @xmath3-dependent quantities @xmath206 , if there exist constants @xmath207 such that @xmath208
. let @xmath209 .
then we define @xmath210 as the @xmath211 minor of @xmath8 obtained by removing all columns and rows of @xmath8 indexed by @xmath212 .
note that we do not change the names of the indices of @xmath8 when defining @xmath210 . more specifically
, we define an operation @xmath213 , @xmath214 , on the probability space by @xmath215 then , for @xmath209 , we set @xmath216 and define @xmath217 the green functions @xmath218 , are defined in an obvious way using @xmath210 .
moreover , we use the shorthand notation @xmath219 abbreviate @xmath220 , @xmath221 and use the convention @xmath222 , if @xmath212 , @xmath223 , else . in green function entries
@xmath224 we refer to @xmath225 as lower indices and to @xmath226 as upper indices .
finally , we set @xmath227 here , we use the normalization @xmath15 , instead @xmath228 , since it is more convenient for our computations .
the next lemma collects the main identities between resolvent matrix elements of @xmath8 and @xmath210 .
let @xmath229 be an @xmath14 matrix .
consider the green function @xmath230 , @xmath231 .
then , for @xmath232 , the following identities hold : * _ schur complement / feshbach formula:_[feshbach ] for any @xmath149 , @xmath233 * for @xmath234 , @xmath235 * for @xmath234 , @xmath236 * for @xmath237 , @xmath238 * _ ward identity : _ for any @xmath149 , @xmath239 where @xmath240 . for a proof
we refer to , e.g. , @xcite .
[ cauchy interlacing ] there is a constant @xmath203 such that , for any @xmath241 , @xmath242 , we have @xmath243 the lemma follows from cauchy s interlacing property of eigenvalues of @xmath8 and its minor @xmath244 . for a detailed proof
we refer to @xcite . for @xmath245 , with , say , @xmath246
, we obtain @xmath247 .
we collect here some useful large deviation estimates for random variables with slowly decaying moments .
[ lemma.lde ] let @xmath248 and @xmath249 be centered and independent complex random variables with variance @xmath250 and having subexponential decay @xmath251 for some positive constants @xmath98 and @xmath99 . for @xmath252 ,
let @xmath253 and @xmath254 .
then there exists a constant @xmath255 , depending only on @xmath256 and @xmath98 , such that for @xmath257 and @xmath258 the following estimates hold . * @xmath259 \mathbb{p}\left(\left|\sum_{i=1}^n\overline{a}_i b_{ii}a_i-\sum_{i=1}^n \sigma^2 b_{ii}\right|\ge ( \varphi_n)^{\xi}\sigma^2\left(\sum_{i=1}^n|b_{ii}|^2\right)^{1/2}\right)&\le \mathrm{e}^{-(\log n)^{\xi}}\,,\label{lde2}\\[1 mm ] \mathbb{p}\left(\left|\sum_{i\not = j}^n\overline{a}_i b_{ij}a_j\right|\ge(\varphi_n)^{2\xi}\sigma^2\left(\sum_{i\not = j}|b_{ij}|^2 \right)^{1/2}\right)&\le\mathrm{e}^{-(\log n)^{\xi}}\label{lde3}\,,\end{aligned}\ ] ] + for @xmath3 sufficiently large ; * @xmath260 + for @xmath3 sufficiently large . for a proof
we refer to @xcite .
we now choose @xmath203 in such that @xmath261 .
finally , we point out the difference between the random variables @xmath95 and @xmath13 : from , we obtain @xmath262 with @xmath182-high probability , whereas @xmath263 $ ] , almost surely .
in this section , we outline the proof of theorem [ thm : main ] . we first fix the diagonal random entries @xmath13 and consider @xmath264 , the deformed semicircle measure with fixed @xmath13 .
the main tools we use in the proof are lemma [ mfc estimate ] , where we obtain a linear approximation of @xmath265 , and lemma [ hat bound ] , which estimates the difference between @xmath265 and @xmath266 , the latter being the stieltjes transform of @xmath264 . using proposition [ prop : mu_k ] that estimates the eigenvalue locations in terms of @xmath266 , we prove theorem [ thm : main ] .
in this subsection we define an event @xmath267 , on which the random variables @xmath13 exhibit `` typical '' behavior . for this purpose
we need some more notation : denote by @xmath268 the constant @xmath269 which only depends on @xmath168 .
fix some small @xmath270 satisfying @xmath271 and define the domain , @xmath272 , of the spectral parameter @xmath191 by @xmath273 using spectral perturbation theory , we find that the following a priori bound latexmath:[\ ] ] notice that , by the central limit theorem , @xmath1210 converges to a gaussian random variable with mean @xmath1195 and variance @xmath1212 . since , @xmath1213 with high probability , the desired lemma follows .
when @xmath13 are fixed , we may follow the proof of theorem 2.21 in @xcite to get @xmath1214 , we find that the leading fluctuation of the largest eigenvalue comes from the gaussian fluctuation obtained in lemma [ lem : gaussian ] .
this also shows that there is a sharp transition in the distribution of the largest eigenvalue from a gaussian law to a weibull distribution as @xmath22 crosses @xmath129 .
capitaine , m. , donati - martin , c. , fral , d. , fvrier , m. : _ free convolution with a semi - circular distribution and eigenvalues of spiked deformations of wigner matrices _ , electron .
j. probab .
* 16 * , 1750 - 1792 ( 2011 ) .
voiculescu , d. , dykema , k. j. , nica , a. : _ free random variables : a noncommutative probability approach to free products with applications to random matrices , operator algebras and harmonic analysis on free groups _ , american mathematical society ( 1992 ) . | we consider random matrices of the form @xmath0 , @xmath1 , where @xmath2 is a real symmetric or complex hermitian wigner matrix of size @xmath3 and @xmath4 is a real bounded diagonal random matrix of size @xmath3 with i.i.d .
entries that are independent of @xmath2 .
we assume subexponential decay of the distribution of the matrix entries of @xmath2 and we choose @xmath5 , so that the eigenvalues of @xmath2 and @xmath6 are typically of the same order .
further , we assume that the density of the entries of @xmath4 is supported on a single interval and is convex near the edges of its support . in this paper
we prove that there is @xmath7 such that the largest eigenvalues of @xmath8 are in the limit of large @xmath3 determined by the order statistics of @xmath4 for @xmath9 .
in particular , the largest eigenvalue of @xmath8 has a weibull distribution in the limit @xmath10 if @xmath9 .
moreover , for @xmath3 sufficiently large , we show that the eigenvectors associated to the largest eigenvalues are partially localized for @xmath9 , while they are completely delocalized for @xmath11 .
similar results hold for the lowest eigenvalues . * extremal eigenvalues and eigenvectors of deformed wigner matrices * ji oon lee + department of mathematical sciences + _ korea advanced institute of science and technology
_ + [email protected]
kevin schnelli + school of mathematics + _ institute for advanced study _
+ [email protected] _ ams subject classification ( 2010 ) _ : 15b52 , 60b20 , 82b44 _ keywords _ : random matrix , local semicircle law , delocalization , localization _ _ |
in spite of remarkable progress in study of @xmath6-violation phenomena in both k and b sectors it remains an interesting issue .
the standard model ( sm ) successfully describes existing experimental data by a single phase of the ckm matrix , although it is hard to believe that this phase is the only source of @xmath6-violation .
for example , the baryon asymmetry of the universe can not be explained by the ckm phase only and at least one additional source of the @xmath6-violation is required .
a good place to look for new @xmath6-violating phases is a measurement of processes where the sm @xmath6-violation is vanishing or very suppressed while additional or alternative sources of @xmath6-violation can produce a sizable effect .
such interesting observables are the electric dipole moment of neutron which is extremely small in the sm , and the transverse lepton polarization in three - body decays of kaons and b - mesons @xcite . in this paper , we study the @xmath5-odd muon polarization in the decay @xmath7 ( @xmath8 ) .
namely , we investigate the possibility to measure the vector and axial - vector kaon form factors , @xmath3 and @xmath4 , from the @xmath5-odd muon polarization emerging due to the electromagnetic final state interactions ( fsi ) .
also we analyze potential contributions to @xmath0 from various extensions of the sm . in this study
we focus on the dalitz plot region where @xmath8 events have a large angle between photon and muon momenta .
this is the region , where @xmath5-odd muon polarization exhibits the best sensitivity to @xmath3 and @xmath4 .
moreover , we found that in this region the ratio of the possible contribution to @xmath0 from new physics and fsi contribution to @xmath0 becomes the largest .
hence , the analysis of the @xmath8 events at large @xmath9 will provide the best accuracy in measurement of the relevant parameters of the new physics or the strongest constraints on them .
it is found that @xmath0 dependence on the kaon form factors gives a possibility to unambiguously determine in @xmath8 decay the signs of the sum of the kaon form factors without fixing their difference .
combined with measurement of the normal muon polarization , which is very sensitive to @xmath3 and @xmath4 @xcite , this allows the values of the kaon form factors to be unambiguously extracted with 1% accuracy from this experiment .
investigating the prospects of new experiments in searching for new physics we show that , generally , they are limited mostly by the uncertainty in fsi predictions rather than by the anticipated statistical error .
the outline of the paper is as follows . in section [ secii ]
we introduce the parameters describing muon transverse polarization @xmath0 in the @xmath8 decay , recall the relevant formulae and estimate the expected sensitivity to @xmath0 in forthcoming experiments . in sec .
[ seciii ] , an approach to determine the signs of the sum of the kaon form factors from the fsi polarization is discussed and the accuracy of this method is estimated . in sec .
[ seciv ] we study the discovery potential of the @xmath0 measurements in search for new physics .
specifically , we estimate the upper bounds on the @xmath0 in multi - higgs doublet models , supersymmetric extensions of the standard model , left - right symmetric model and leptoquark models .
[ secv ] contains conclusions and final remarks .
introducing three unit vectors @xmath10 with @xmath11 and @xmath12 being the four - momenta of @xmath13 and @xmath14 , respectively , one can define longitudinal ( @xmath15 ) , normal ( @xmath16 ) and transverse ( @xmath0 ) components of the muon polarization as the corresponding contributions to the squared matrix element of the @xmath8 decay , @xmath17\;,\ ] ] with @xmath18 being a unit vector along muon spin and @xmath19 is @xmath20 where the internal bremsstrahlung ( @xmath21 ) , structure dependent ( @xmath22 ) and interference contributions ( @xmath23 ) are given as follows @xcite @xmath24\ !
\label{fsd } f_\mathrm{sd}&= m^4_k x^2\bigg[|f_v+f_a|^2\frac{\lambda^2 } { 1-\lambda } \left(1-x-\frac{r_\mu } { \lambda}\right ) \nonumber \\ & \qquad+|f_v - f_a|^2(y-\lambda)\bigg]\ ; , \\ \label{fibsd } f_\mathrm{ibsd}&= - 4m_k m^2_\mu\bigg[{\mathop\mathrm{re}\nolimits}[f_k(f_v+f_a)^ * ] \left(1-x-\frac{r_\mu}{\lambda}\right ) \nonumber\\ & \qquad -{\mathop\mathrm{re}\nolimits}[f_k(f_v - f_a)^*]\frac{1-y+\lambda}{\lambda}\bigg ] \;.\end{aligned}\ ] ] here we used the standard notations @xmath25 , @xmath26 , and @xmath27 , @xmath28 with @xmath29 , @xmath30 being the photon and muon energies in the kaon rest frame , respectively ; @xmath31 is the fermi constant , @xmath32 is the corresponding element of the cabibbo
maskawa ( ckm ) matrix and @xmath33 mev is the kaon decay constant . in terms of these variables
the differential decay width reads @xmath34 the transverse muon polarization @xmath0 is determined using the partial decay width @xmath35 this is a @xmath5-odd observable ( both @xmath15 and @xmath16 are @xmath5-even ) , hence in the @xmath5-invariant theory its value equals zero at tree level .
moreover , @xmath0 does not have tree - level contributions from the @xmath6-violating phase in the ckm matrix .
these two features make @xmath0 a very promising observable for new @xmath6-violating physics searches . at the same time , as will be seen in the sec .
[ seciii ] , analysis of loop contributions from the standard model ( fsi ) is also of a special physical interest . for the following analysis
let us estimate the level of precision which could be achieved experimentally .
the running e246 experiment at kek @xcite dedicated for a measurement of @xmath0 in the decay @xmath36 has only a limited sensitivity of about @xmath37 to @xmath0 in @xmath8 @xcite .
there is a proposal for a new experiment in which a statistical sensitivity @xmath38 in @xmath8 can be reached @xcite .
the main features of this experiment include a high resolution measurement of neutral particles from @xmath39 and @xmath8 decays , usage of an active muon polarimeter which provides information about stopped muons ( stopping point , momentum ) , positron direction , and also detects photons , and a highly efficient photon veto system covering nearly @xmath40 solid angle .
this approach allows to accumulate @xmath8 events for all @xmath9 angles between photon and muon momenta due to efficient photon veto detector .
this system eliminates @xmath41 decays which are the main background source at large @xmath9 .
using parameters of the proposed detector @xcite and the kaon beam intensity of about @xmath42 per second expected at the jhf @xcite one can estimate the sensitivity to @xmath0 which could be achieved in this experiment . with the analyzing power of the detector @xmath43 and the kinematical attenuation factor @xmath44
, the expected sensitivity to @xmath0 in some region @xmath45 of the dalitz plot can be expressed as @xmath46 where @xmath47 is the number of @xmath8 events in the region @xmath45 . in this experiment ,
the most effective suppression of background events are anticipated in the kinematic region with @xmath48 mev , @xmath49 mev . there
the number of @xmath8 events accumulated for one year running period is estimated to be @xmath50 . in what follows we adopt these cuts as well as the number of events .
the branching ratio of this decay for @xmath51 mev and @xmath52 mev integrated over @xmath53 wide bins in @xmath9 is presented in fig .
[ fig:2 ] .
decay for @xmath53 wide bins in @xmath9 ; @xmath51 mev , @xmath52 mev .
dependence of this branching ratio on the values of the @xmath3 , @xmath4 form factors is quite weak . ] from this plot one can estimate the sensitivity to @xmath0 in various @xmath9 regions .
due to lack of understanding of the qcd low - energy structure , there is no definite prediction for the values of the @xmath3 and @xmath4 form factors : the calculation of them is a model dependent procedure .
so , the measurement of these form factors would give a possibility to make a choice between various candidates for the correct description of the qcd low - energy limit .
the distribution of the @xmath8 decay width over dalitz plot allows only the absolute values of the sum and difference of the kaon form factors to be measured , since the term @xmath54 ( see eq . ) is small . the terms with linear and quadratic dependence on @xmath3 and @xmath4
give the comparable contribution in some regions of the dalitz plot that could , in principle , make it possible to measure the signs as well as the magnitudes of the form factors .
unfortunately , in the region where the linear terms grow , the dominant contribution to @xmath8 ( internal bremsstrahlung , which depends only on @xmath55 ) also increases , that significantly reduces the sensitivity of dalitz plot measurements to the kaon form factors . in practice ,
the situation is even worse , since only the absolute value of the sum of the kaon form factors has been measured with good accuracy in these measurements , while their difference still has only lower and upper bounds @xcite : @xmath56 recently , both vector and axial - vector form factors have been measured in @xmath57 and @xmath58 decays @xcite .
these decays are generalizations of @xmath59 for the case of a virtual photon in the final state , so the kaon form factors @xmath3 and @xmath4 are believed to be the same in all these processes . the combined fit for both four - body decay experiments results in @xmath60 these values are in a good agreement with @xmath61 predictions @xcite of the chiral perturbation theory ( chpt ) @xmath62 it is worth to note that actually @xmath3 and @xmath4 are not constants , but some functions of the momentum of the lepton pair , @xmath64 , with @xmath65 being kaon four - momentum . in chpt @xmath63-dependence emerges due to higher order corrections , which have not been calculated yet .
their magnitude can be estimated as @xmath66 where @xmath67 and @xmath68 are masses of the first strange hadronic vector ( @xmath69 ) and axial - vector ( @xmath70 ) resonances , respectively .
this estimate implies corrections as large as 25% , hence the experimental data on @xmath59 should be fitted with at least two additional parameters , that generally decreases the chances to determine @xmath3 and @xmath4 in @xmath59 experiments . in @xmath71
both vector and axial - vector form factors get additional dependence on the non - zero @xmath72 that also increases the bias in the corresponding fitting procedure . in ref .
@xcite experimental data was fitted assuming constant form factors and form factors depending on @xmath63 and @xmath72 @xmath73 with @xmath74 being mass of @xmath75 meson .
this dependence appears in the approximation of dominance of the contribution of lowest lying resonances @xcite .
the experimental data favors the dependence ( [ fqq ] ) over the constant @xmath76 , though the accuracy achieved in this experiment did not allow definite confirmation or rejection of @xmath63- , @xmath72-dependence .
the determination of the @xmath63-dependence of the form factors may be done in future experiments .
in particular , analysis of the @xmath16 muon polarization in @xmath8 decays can improve knowledge about the @xmath76 behavior for sure @xcite .
however , at present , the unknown @xmath63-dependence imports some uncertainty into predictions for muon asymmetry to be measured in future experiments .
this concerns fsi contribution as well as possible contribution from new physics , since both are functions of the kaon form factors . to avoid all these difficulties one can try to find a physical observable which strongly depends on @xmath3 and @xmath4 and
can be measured with an accuracy sufficient to distinguish the signs of kaon form factors in the @xmath8 decay .
this implies that : _
( i ) _ the observable we are interested in should exhibit linear dependence on @xmath3 and @xmath4 in some region of the dalitz plot ; _ ( ii ) _ in this region the differential partial width should be sufficiently large ; _ ( iii ) _ the @xmath63-corrections in this region should be small enough to unambiguously determine the sign of the kaon form factors .
below we consider the transverse muon polarization calculated in the framework of the standard model as a physical observable , which pins down the signs of kaon form factors in @xmath8 decay . in the framework of the standard model ,
the contribution to @xmath0 emerges only at loop levels due to the fsi .
this contribution has been recently calculated @xcite at one - loop level .
the expression for @xmath77 can be read out from ref .
@xcite after changing the definitions of the form factors : @xmath78 and @xmath79 .
the asymmetry is positive at any relevant @xmath80 and its absolute value ranges from zero to @xmath81 .
the dalitz plot distribution of @xmath0 for several values of the form factors satisfying , is presented in fig .
[ fig : pt_dalitz ] .
the transverse muon polarization emerging from the fsi exhibits the behavior we are looking for to pin down the signs of the kaon form factors .
indeed , the muon transverse polarization is sensitive to the signs of @xmath3 and @xmath4 , especially at large angles @xmath9 between photon and muon momenta . to illustrate this point we present the @xmath0 values integrated over @xmath53intervals as a function of @xmath9 in fig .
[ fig:1 ] .
( fsi ) asymmetry for @xmath53 wide bins in @xmath9 .
upper region corresponds to negative @xmath82 and lower one to positive . ]
the value of @xmath0 associated with a bin @xmath83 is determined as @xmath84 where the additional relevant cuts are @xmath85 , @xmath86 .
we evaluate @xmath0 for four different sets of the form factors corresponding to the boundaries of the allowed intervals , .
one can see that the predictions for @xmath0 differ significantly , and in the region of large angles , @xmath87 , _ the sign of @xmath82 can be unambiguously determined from the @xmath0 analysis only_. suppression of @xmath41 decays becomes less efficient at very large @xmath9 angles because of low efficiency of the detection of low energy photons from asymmetrical @xmath88 decays , so we adopt the region @xmath89 as the realistic region with the best sensitivity to the kaon form factors .
the difference between the @xmath0 values increases with @xmath9 and reaches its maximum value at @xmath90 , as seen from fig .
[ fig:1 ] .
although the differential @xmath8 branching ratio decreases at large @xmath9 , its value is still reasonable at large @xmath9 , see figure [ fig:2 ] .
the fraction of the @xmath8 events within @xmath91 is about 7% . by making use of eq .
( [ eq : pt ] ) , the statistical sensitivity to @xmath0 in this dalitz plot region is estimated to be @xmath92 to unambiguously determine the sign of @xmath82 without any additional data on the difference of the form factors one has to be able to distinguish the @xmath0 values with @xmath93 , @xmath94 and @xmath95 , @xmath96 ( the two closest lines in fig .
[ fig:1 ] ) .
for these two sets of the form factors , the difference between @xmath0 values in the region @xmath97 is @xmath98 , that allows determination of the sign of the sum of the form factors at the level of 1.6 @xmath99 .
note , there is no usual degeneracy related to the signs of the sum of the form factors , that provides a possibility to pin down the sign of the sum without fixing the difference . if the difference between the form factors is measured with better precision than in eq .
one has to compare the situations with equal @xmath100 and different signs of the sum of the form factors , that gives the difference in @xmath0 values in the same region @xmath101 of @xmath102 depending on the precise value of @xmath100 .
then the statistical sensitivity ( [ statsens ] ) allows to determine the signs at the level of @xmath103 @xmath99 .
so , measurement of @xmath0 allows signs of the form factors to be distinguished .
it is worth to note in passing , that at large @xmath9 the main contribution to the @xmath0 ( fsi ) comes from the dalitz plot region @xmath104 , where the uncertainty associated with the momentum of the lepton pair @xmath63 is @xmath10515% .
this uncertainty is small enough to be neglected , because it is significantly less than the difference in @xmath0 for the form factors of different signs , see fig .
[ fig:1 ] .
recently we found @xcite that the normal muon polarization @xmath16 is extremely sensitive to the values of the form factors . since measurement of @xmath0 pins down the sign of the form factors
, it will provide a possibility of an independent cross - check of the results from measurement of @xmath16 in the future experiment @xcite , which will determine @xmath3 and @xmath4 with accuracy as high as 1% @xcite basing only on its own experimental data .
for a general investigation of possible contribution of some new @xmath6-violating sources to the transverse muon polarization in @xmath8 we first introduce the most general four - fermion interaction @xmath106 where @xmath107 , @xmath108 , @xmath109 , and @xmath110 , arising from new physics , denote scalar , pseudoscalar , vector , and axial - vector coupling constants , respectively .
their contribution to @xmath8 decay amplitude may be taken into account by the redefinition of the usual kaon form factors as follows , @xmath111 with @xmath112 the constant @xmath113 is related to quark condensate as @xmath114 and may be evaluated from the masses of kaon and quarks , @xmath115 gev .
the scalar type interaction @xmath107 does not contribute to @xmath8 decay because of parity , and will not be considered below .
note , however , that it contributes to the transverse muon polarization in the @xmath116 ( @xmath39 ) decays ( while the pseudoscalar interaction @xmath108 does not ) .
the imaginary parts of the new coupling constants are responsible for @xmath6-violation and could give rise to the tree - level contributions to the muon polarization @xcite , @xmath117\frac{\lambda}{1-\lambda } \times\left(1-x-\frac{r_{\mu}}{\lambda}\right)\\ & \hphantom{\times\biggl\{\,}+ { \mathop\mathrm{im}\nolimits}[f_k(f_v - f_a)^*]\biggr\}\ , .
\end{split}\ ] ] it is convenient to rewrite @xmath118 as @xmath119 with @xmath120{\mathop\mathrm{im}\nolimits}(\delta_p)\ , , \label{pt - va}\end{aligned}\ ] ] where @xmath121\;,\end{gathered}\ ] ] @xmath122\;.\end{gathered}\ ] ] to illustrate the sensitivity of the transverse muon polarization to new physics we adopt the experimental values ( [ measured - values ] ) for the kaon form factors and plot the integrated over @xmath53 @xmath9-intervals @xmath0 values as a function of @xmath9 ( see the definition ( [ theta - definition ] ) ) in fig .
[ fig : fsi - fvfa - exp ] .
the distribution of @xmath123 and @xmath124 over the dalitz plot are completely determined by the functions @xmath125 and @xmath126 $ ] , presented in fig .
[ sv_sv - sa ] . for @xmath53 wide bins in @xmath9 ;
the kaon form factors are equal to the measured values ( [ measured - values ] ) . ] indeed , the values of new @xmath6-violating coupling constants provide only the normalization factors , see eqs .
( [ pt - va ] ) . to understand the general behavior we present @xmath123 and @xmath124 values for @xmath53 wide bins in @xmath9 at @xmath127 in fig . [
fig : new - phys ] . for @xmath53 wide bins in @xmath9 at @xmath127 ;
the kaon form factors are equal to the measured values ( [ measured - values ] ) .
[ fig : new - phys ] ] comparing figs .
[ fig : fsi - fvfa - exp ] and [ fig : new - phys ] one can realize that while both fsi and new - physics contributions are peaked at large @xmath9 , the slope is steeper in the former case : @xmath128 this shows that the sensitivity to new physics increases the region of dalitz plot with large @xmath9 angles . indeed , the best way to distinguish fsi and new - physics contributions is to compare the @xmath9-dependence and @xmath0 averaged over the large angles , @xmath129 . for this range
we obtain @xmath130 taking into account the expected experimental sensitivity one may hope to detect effects of new physics leading to @xmath131 or @xmath132 as small as @xmath133 at 1 @xmath99 level . in the estimation of the sensitivity presented above we neglected the contribution to @xmath0 from the fsi .
this is possible if one reduces the uncertainty in the prediction of @xmath134 to values smaller than @xmath135 . to this end
careful determination of the @xmath4 and @xmath3 form factors from other experiments is needed .
for example , from @xmath136 , see eq . , or from the measurement of the normal muon polarization @xmath16 in @xmath8 decay , which is also a very sensitive to the form factors observable @xcite .
simultaneous measurement of both @xmath16 and @xmath0 in one experiment with the same statistical sensitivity provides good opportunity for accurate determination of @xmath4 and @xmath3 from @xmath16 and , therefore , precise calculation of the fsi contribution to measured @xmath0 and clear separation the non - sm @xmath0 from physics background .
the values of @xmath137 , @xmath138 and @xmath139 are completely determined by the physics beyond the standard model .
below we consider several extensions of the sm and present the corresponding bounds on @xmath137 , @xmath138 and @xmath139 from existing experimental data . in a general two - higgs doublet ( 2hd )
model @xcite up- and down - type quarks , as well as leptons , couple to both two - higgs doublets .
the up- and down - type quark mass matrices can not be diagonalized simultaneously , that is the reason for flavor changing neutral current to be induced at tree level .
the relevant terms read @xcite @xmath140 u \\\label{2hd - lagr } \times\bar{\nu}{\left(}1+\gamma_5{\right)}\mu \end{gathered}\ ] ] where @xmath141 is the mass of charged higgs boson , @xmath142 are coupling constants of charged higgs boson and corresponding leptons and @xmath143 are mixing matrices parameterizing up- and down - type quark interactions with the lightest neutral higgs boson .
this interaction results in the muon transverse polarization @xmath144\frac{b_0z}{2\sqrt{2}g_fm_h^2|v_{us}| } \sum_i{\mathop\mathrm{im}\nolimits}[v_{is}^*\tilde{\xi}_{i1}^u+v_{ui}^*\tilde{\xi}_{i2}^{d*}]\;.\ ] ] the measurement of the @xmath145 gives @xmath146 gev @xcite , perturbativity bound ( see , e.g. , @xcite ) implies @xmath147 gev@xmath148 and , if the only relevant terms in eq .
( [ pt2hd ] ) are @xmath149 and @xmath150 , neutral kaon system constrains them to be @xmath151 @xcite . with account of eq
( [ pta ] ) this favors the limit @xmath152}{10^{-2}}{\right)}\;. \end{split}\ ] ] it is more than 300 times larger than the expected experimental sensitivity to @xmath0 @xcite .
note , that without special fine - tuning between @xmath153 $ ] and @xmath154 $ ] , the strongest limit on @xmath0 comes from the search for transverse muon polarization in @xmath155 decay @xcite , since the interaction ( [ 2hd - lagr ] ) generally provides both @xmath108 and @xmath107 of the same order @xcite ( see sec .
[ st - param ] ) .
this yields @xmath156 at 95% cl . on the other hand , depending on the structure of the mixing matrices
@xmath158 , the lower bound on @xmath0 in 2hd model can be as large as 0.1 .
in these models three different higgs doublets ( @xmath159 , @xmath160 , @xmath161 ) couple to up- , down - type quarks and leptons , respectively .
the relevant interaction between sm fermions and charged higgs bosons for @xmath8 decay reads @xmath162{2}}\sum_{i=1}^2\biggl\{h_i^+ [ \alpha_i\bar{u}vm_d{\left(}1+\gamma_5{\right)}d \\ + & \beta_i\bar{u}m_uv{\left(}1-\gamma_5{\right)}d+\gamma_i\bar{\nu}m_l{\left(}1+\gamma_5{\right)}e]\biggr\}+h.c.\ ; , \end{split}\ ] ] where @xmath163 is the ckm matrix , @xmath164 , @xmath165 and @xmath166 are diagonal down - type quark , up - type quark and lepton mass matrices , respectively , and @xmath167 , @xmath168 and @xmath169 ( @xmath170 ) are complex mixing parameters in higgs sector .
this interaction results in @xcite @xmath171{\left(}\frac{m_k^2}{m_1 ^ 2}-\frac{m_k^2}{m_2 ^ 2}{\right)}\\&\times & { \left(}{\mathop\mathrm{im}\nolimits}{\gamma_1\alpha_1^*}-\frac{m_u}{m_s}{\mathop\mathrm{im}\nolimits}{\gamma_1\beta_1^*}{\right)}\label{pt3hd}\end{aligned}\ ] ] where @xmath172 are the masses of charged higgs bosons .
the current experimental constraints on the parameters of the theory ( see , e.g. , ref .
@xcite for a collection of relevant formulae ) give the same order bound on @xmath0 as from @xcite for @xmath0 in @xmath155 decay ( as far as @xmath108 and @xmath107 emerging from lagrangian ( [ thdml ] ) are generally of the same order ) , @xmath173 this upper limit is about 30 times larger than the expected sensitivity of the new experiment
. in the minimal supersymmetric extension of the standard model ( mssm ) the relevant four - fermion interactions emerge due to w - boson and charged higgs boson exchanges with couplings being enhanced by squark - gluino loops .
right - handed current interaction is generated by the diagram with gluino , stop and sbottom particles running in the loop with left - right mass insertions both for stop and sbottom .
this gives rise to the effective interaction @xcite @xmath174}{36\pi } v_{31}^{u_{{{\scriptscriptstyle \rm r}}}}v_{32}^{d_{{{\scriptscriptstyle \rm r}}}*}v_{33}^{sckm*}\;,\end{aligned}\ ] ] where @xmath175 and @xmath176 are the soft supersymmetry breaking trilinear terms for stops and sbottoms , @xmath177 is higgs superfield mixing parameter , @xmath178 is the ratio of the two higgs vacuum expectation values , @xmath179 is gluino mass , @xmath180 and @xmath181 are the rotations in the generation space between up - type right - handed squarks and down - type right - handed squarks and corresponding quark partners , respectively ; @xmath182 is the super ckm matrix associated with w - squark - squark couplings , and @xmath183 is given by the integral @xmath184 for w - boson exchange , the interaction ( [ efsusy ] ) provides the polarization @xmath185}{18\pi } \frac{{\mathop\mathrm{im}\nolimits}[{v_{31}^{u_{{{\scriptscriptstyle \rm r}}}}v_{32}^{d_{{{\scriptscriptstyle \rm r}}}*}v_{33}^{sckm*}}]}{|v_{us}|}\;. \label{ptsusywr1}\end{aligned}\ ] ] current experimental constraints on the parameters of the theory @xcite leads to @xmath186 where we set @xmath187 , @xmath188 , @xmath189 , and @xmath190\simeq0.5 $ ] .
the charged higgs boson exchange enhanced by gluon - stop - sbottom loops gives rise to the effective interaction @xcite @xmath191}{3\pi}i_1\frac{m_tm_\mu\tan{\beta}}{m_{h^+}^2 } \frac{a_t\cot{\beta}\!+\!\mu}{m_{\tilde{g } } } v_{31}^{u_{{{\scriptscriptstyle \rm r}}}}v_{32}^{d_{{{\scriptscriptstyle \rm l}}}*}v_{33}^{h*}\;,\nonumber\\ \!\!c_2\!\!=\!\!\frac{\alpha_s[m_{{{\scriptscriptstyle \rm susy}}}]}{3\pi}i_1\frac{m_bm_\mu\tan{\beta}}{m_{h^+}^2 } \frac{a_b\tan{\beta}\!+\!\mu}{m_{\tilde{g } } } v_{31}^{u_{{{\scriptscriptstyle \rm l}}}}v_{32}^{d_{{{\scriptscriptstyle \rm r}}}*}v_{33}^{h*}\ ; , \nonumber\\\nonumber i_1\!\!=\!\!\int_0 ^ 1dz_1\int_0^{1-z_1}\!\!\!\!\!\!\!\!\!dz_2\frac{2 } { 1+{\left(}\frac{m^2_{\tilde{t}}}{m^2_{\tilde{g}}}-1{\right)}z_1+{\left(}\frac{m^2_{\tilde{b}}}{m^2_{\tilde{g}}}-1{\right)}z_2}\;,\end{gathered}\ ] ] where @xmath192 is the mixing in the coupling between charged higgs and up - type right - handed and down - type left - handed squarks , @xmath193 and @xmath194 are the rotations in the generation space between up - type left - handed squarks and down - type left - handed squarks and corresponding quark partners , respectively .
this results in @xmath195}{3\pi}i_1 \frac{b_0m_t\tan\beta}{m_{h^+}^2}\frac{a_t\cot{\beta}+\mu}{m_{\tilde{g } } } \\&\times&\frac{{\mathop\mathrm{im}\nolimits}[{v_{31}^{u_{{{\scriptscriptstyle \rm r}}}}v_{32}^{d_{{{\scriptscriptstyle \rm l}}}*}v_{33}^{h*}}]}{|v_{us}|}\nonumber\\\nonumber & -&{\left(}\sigma_v-\sigma_a{\right)}\frac{\alpha_s[m_{{{\scriptscriptstyle \rm susy}}}]}{3\pi}i_1 \frac{b_0m_b\tan\beta}{m_{h^+}^2}\frac{a_b\tan{\beta}+\mu}{m_{\tilde{g } } } \\&\times&\frac{{\mathop\mathrm{im}\nolimits}[{v_{31}^{u_{{{\scriptscriptstyle \rm l}}}}v_{32}^{d_{{{\scriptscriptstyle \rm r}}}*}v_{33}^{h*}}]}{|v_{us}|}\;. \label{ptsusywr2}\end{aligned}\ ] ] with the estimate @xmath196\simeq0.5 $ ] and @xmath197\simeq0.5 $ ] and the same settings as listed below eq .
( [ ptwestimate ] ) both terms in eq .
( [ ptsusywr2 ] ) are of the order @xmath37 .
so , @xmath198 and @xmath199 are 10 and 100 times larger , respectively , than the expected experimental sensitivity .
note in passing , that without a special cancellation associated with light superpartners , charge higgs boson gives large contribution to @xmath145 .
this yields the bound @xmath146 gev @xcite , which decreases the upper limit on @xmath199 by an order of magnitude .
in this section we consider the supersymmetric extensions of the standard model with the violation of the r - parity and the lepton number .
the relevant superpotential is given by @xmath200 with @xmath201 and @xmath202 being the chiral superfields of lepton doublets and singlets , and @xmath203 , @xmath204 denote chiral superfields of quark doublets and down - type singlets , respectively . in this model
the contribution to @xmath8 decay arises due to the interaction @xmath205 where @xmath206 are masses of the left - handed sleptons .
the resulting transverse muon polarization reads @xcite @xmath207\frac{b_0}{m_\mu|v_{us}| } \frac{1}{2\sqrt{2}g_fm^2_{\tilde{e}_{{{\scriptscriptstyle \rm l_i}}}}}\;.\ ] ] the strongest relevant experimental limit on the parameters of the model comes from the measurement of @xmath208 decay rate @xcite , and one can obtain from eqs .
( [ pta ] ) , ( [ ptsusy - w / or ] ) @xmath209}{3.8\times10^{-7}}{\right)}{\left(}\frac{m_{\tilde{\nu}_1}}{m_{\tilde{e}_{{{\scriptscriptstyle \rm l_1}}}}}{\right)}^2\nonumber\\&+ & { \left(}\frac{{\mathop\mathrm{im}\nolimits}[\lambda_{232}^*\lambda_{312}']}{3.8\times10^{-7}}{\right)}{\left(}\frac{m_{\tilde{\nu}_3}}{m_{\tilde{e}_{{{\scriptscriptstyle \rm l_3}}}}}{\right)}^2 \biggr)\nonumber\end{aligned}\ ] ] with @xmath210 being sneutrino masses .
this contribution ( [ ptsusy - w / or - numbers ] ) is obviously too small to be detected .
note in passing that in models with relevant hierarchy in slepton sector or in models with some cancellation of the sparticle contributions to @xmath208 decay , one could expect @xmath0 in @xmath8 at the level of @xmath211 . in models with left - right gauge symmetries
@xmath212 @xcite the four - fermion interaction which contributes to @xmath8 decay is given by @xmath213 where @xmath214 is left - right complex mixing parameter , @xmath215 is right - handed ckm matrix and @xmath216 are coupling constants of @xmath217 and @xmath218 , respectively .
this interaction results @xcite in @xmath219 with the current experimental constraints on the parameters of the theory @xcite one can obtain from eqs .
( [ ptv ] ) , ( [ ptlr ] ) @xmath220\;,\end{gathered}\ ] ] that is 20 times larger than the expected statistical uncertainty of @xmath0 ( [ statsens ] ) .
there exist two leptoquark models contributing to @xmath0 in @xmath8 decay .
the quantum numbers of the leptoquarks under the standard model group are @xmath221 and the general couplings involving these leptoquarks are given by @xmath222 the relevant terms read @xmath223 where @xmath224 are leptoquark masses , and they provide @xcite @xmath225 \frac{b_0}{m_\mu|v_{us}|}\frac{\sqrt{2}}{8g_fm_{\phi_1}^2}\ ; , \label{ptlq1}\\ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!p_{t_{ii}}\!\!&=&\!\!-{\left(}\sigma_v-\sigma_a{\right)}{\mathop\mathrm{im}\nolimits}[\lambda_2^{2i}\lambda_2'^{12 * } ] \frac{b_0}{m_\mu|v_{us}|}\frac{\sqrt{2}}{8g_fm_{\phi_2}^2}\;. \label{ptlq2}\end{aligned}\ ] ] with the current experimental constraints on the parameters of the models with leptoquarks , the strongest limit on @xmath0 comes from the measurement of the @xmath0 in @xmath155 decay @xcite ( supposing that @xmath108 and @xmath107 constants are generally of the same order in this theory ) , that yields @xmath226 at 95% cl .
this bound is larger than the expected experimental sensitivity by a factor of 30 .
the analysis presented in this paper concerns the information which could be drawn from the experiment on the measurement of the transverse muon polarization @xmath0 in @xmath227 decay .
it was found that the fsi @xmath0 distribution over the dalitz plot is sensitive to the values of the kaon form factors .
the best sensitivity is exhibited in the region of large angles @xmath9 between muon and photon .
the @xmath0 calculated in this region allows to pin down the signs of the kaon form factors in spite of the uncertainties associated with the unknown dependence of the form factors on the momentum of the lepton pair . to this
end the statistics of @xmath228 @xmath8 events with large @xmath229 is required .
it was recently demonstrated @xcite that the normal muon polarization @xmath16 is very sensitive to the signs and the values of the kaon form factors .
since @xmath16 emerges at tree level , the statistics expected in the new experiment proposed at jhf allows to determine for sure the kaon form factors with a few percent accuracy .
the unknown dependence of the form factors on the momentum of the lepton pair should be determined by fitting the experimental data on the @xmath16 dalitz plot distribution .
finally , it was shown that possible new - physics contributions and the fsi contribution to @xmath0 show different behavior over the dalitz plot .
in particular , at large @xmath9 the new - physics contributions to @xmath0 grow by one and a half orders of magnitude , that suggests to consider this region as the most interesting to probe the new physics responsible for @xmath5-violating effects . in most cases
the contribution from new physics can exceed the measurable level of @xmath0 by one or two orders of magnitude .
moreover , extraction of form factor values from @xmath16 allows the background contribution from fsi to the non - sm t - violating muon polarization to be determined with small uncertainty .
it is worth to note , that the signs of pion form factors @xmath230 and @xmath231 have not been measured yet @xcite .
the experimental situation there is similar to kaons , though @xmath230 has a definite cvc prediction and the pion form factors are almost constants over the whole dalitz plot .
the fsi contributions to @xmath0 is of the order of @xmath232 ( @xmath211 ) for @xmath233 ( @xmath234 ) @xcite . from the analysis given above
, we can suggest that the measurement of the transverse lepton polarization in @xmath235 would allow to pin down the signs of the pion form factors , if possible contributions to @xmath0 from physics beyond the standard model are negligible .
this issue will be considered elsewhere .
the work is supported in part by rfbr grant 02 - 02 - 17398 and by the program scopes of the swiss national science foundation , project no .
7supj062239 . the work of f.b .
is supported in part by crdf grant rp1 - 2364-mo-02 .
the work of d.g . is supported in part by the intas ysf 2001/2 - 142 .
99 t. d. lee , phys .
d * 8 * , 1226 ( 1973 ) . s. weinberg , phys .
d * 42 * , 860 ( 1990 ) ; r. garisto and g. l. kane , phys .
d * 44 * , 2038 ( 1991 ) ; g. belanger and c. q. geng , phys . rev .
d * 44 * , 2789 ( 1991 ) ; g. h. wu and j. n. ng , phys .
b * 392 * , 93 ( 1997 ) [ arxiv : hep - ph/9609314 ] .
e. gabrielli , phys .
b * 301 * ( 1993 ) 409 .
j. bijnens , g. ecker and j. gasser , nucl .
b * 396 * , 81 ( 1993 ) [ arxiv : hep - ph/9209261 ] .
c. h. chen , c. q. geng and c. c. lih , phys .
d * 56 * , 6856 ( 1997 ) [ arxiv : hep - ph/9709447 ] .
y. g. kudenko , phys .
nucl . * 65 * , 244 ( 2002 ) [ yad .
fiz . * 65 * , 269 ( 2002 ) ] [ arxiv : hep - ex/0103007 ] .
y. g. kudenko and a. n. khotyantsev , phys .
nucl . * 63 * , 820 ( 2000 ) [ yad .
fiz . * 63 * , 890 ( 2000 ) ] ; yu .
kudenko , measurement of t - violation in k@xmath236 decays at jhf , proceedings of the international workshop on jhf science , march 4 - 7,1998 , tsukuba , japan , v.ii , p.39 .
m. furusaka _ et al_. , the joint project for high - intensity proron accelerators , kek report 99 - 4 , 1999 .
s. c. adler _ et al .
_ [ e787 collaboration ] , phys . rev .
* 85 * , 2256 ( 2000 ) [ arxiv : hep - ex/0003019 ]
. k. hagiwara _ et al .
_ [ particle data group collaboration ] , phys .
d * 66 * , 010001 ( 2002 ) .
a. a. poblaguev _ et al .
_ , phys . rev
. lett . *
89 * , 061803 ( 2002 ) [ arxiv : hep - ex/0204006 ] .
f. l. bezrukov , d. s. gorbunov and y. g. kudenko , arxiv : hep - ph/0302106 .
d. y. bardin and e. a. ivanov , sov .
j. part .
nucl . * 7 * , 286 ( 1976 ) [ fiz .
yadra * 7 * , 726 ( 1976 ) ] . v. v. braguta , a. e. chalov and a. a. likhoded , phys .
d * 66 * , 034012 ( 2002 ) [ arxiv : hep - ph/0205203 ] .
p. gambino and m. misiak , nucl .
b * 611 * , 338 ( 2001 ) [ arxiv : hep - ph/0104034 ] .
v. d. barger , j. l. hewett and r. j. phillips , phys .
d * 41 * , 3421 ( 1990 ) .
m. kobayashi , t. t. lin and y. okada , prog .
. phys . * 95 * , 361 ( 1996 ) [ arxiv : hep - ph/9507225 ] .
y. grossman , nucl .
b * 426 * , 355 ( 1994 ) [ arxiv : hep - ph/9401311 ] .
g. h. wu and j. n. ng , phys .
d * 55 * , 2806 ( 1997 ) [ arxiv : hep - ph/9610533 ] .
m. abe _ et al .
_ , [ arxiv : hep - ex/0211049 ] .
d. choudhury and p. roy , phys .
b * 378 * , 153 ( 1996 ) [ arxiv : hep - ph/9603363 ] .
j. c. pati and a. salam , phys .
d * 10 * , 275 ( 1974 ) ; r. n. mohapatra and j. c. pati , phys .
d * 11 * , 566 ( 1975 ) ; r. n. mohapatra and j. c. pati , phys .
d * 11 * , 2558 ( 1975 ) ; g. senjanovic and r. n. mohapatra , phys .
d * 12 * , 1502 ( 1975 ) .
g. barenboim , j. bernabeu , j. prades and m. raidal , phys .
d * 55 * , 4213 ( 1997 ) [ arxiv : hep - ph/9611347 ] .
f. l. bezrukov and d. s. gorbunov , phys .
d * 66 * , 054012 ( 2002 ) [ arxiv : hep - ph/0205158 ] . | we study the potential of the measurement of the transverse muon polarization @xmath0 in the @xmath1 decay with the sensitivity of @xmath2 .
it is shown that the forthcoming experiment can measure the contribution of the electromagnetic final state interactions to @xmath0 that gives a possibility to unambiguously determine the signs of the sum of the kaon form factors @xmath3 and @xmath4 even without fixing their difference .
we also estimate the sensitivity of this experiment to the new physics , which could give rise to @xmath5-violation : multi - higgs doublet models , supersymmetric extensions of the standard model , left - right symmetric model and leptoquark models . |
the amplitude analysis of non - leptonic three body b decays became an important tool to determine the ckm phases @xcite , and also to observe cp - asymmetry . using this method ,
the belle and babar collaborations have recently @xcite extracted a fraction asymmetry for the channel @xmath2 . with this kind of analysis
one could also explore the asymmetry associated with the interference between two neighbour resonances decaying into the same three body final state .
the dalitz plot analysis needs some _ a priori _ model , with all possible dynamical components and a correct functional form to be used to fit the dalitz plot distribution .
arbitrary distribution functions can be used to get a good fit , but they have no physical meaning . the non - resonant component
which , in general , is spread all over the phase space , can mimic other dynamical components , through the interferences with the resonances present in the same phase space .
this shadowing phenomenon was observed in charm three body decays in the e791 experiment @xcite , where the overestimated contribution of the non - resonant amplitude replaced , in a wrong way , the contribution ( and the very existence ) of the scalar mesons @xmath3 and @xmath4 .
the function above could fit the data with an acceptable confidence level around 1% .
certainly this distribution was more useful for the belle analysis than the usual constant non - resonant distribution .
however , it has to be employed with caution .
the use of an empirical parametrization , with a form without dynamical content , may hide the physical meaning of the dalitz plot and , even worse , may yield inadequate parameters for the contribution of the resonance amplitudes and for the cp asymmetries .
this is so because in the dalitz distribution the parameters are highly correlated among themselves and also with the non - resonant amplitude .
the parametrization ( [ anr ] ) was proposed to describe non - uniformities in the non - resonant three - body decays .
data suggest the formation of a @xmath13 with momentum predominantly smaller than the momentum of the @xmath14 and @xmath15 in the process @xmath16 .
how can we understand this ? in principle
there are two different cathegories of dynamical processes contributing to the non - resonant structures in the dalitz plot .
one is a genuine three - body decay that is a consequence of the partonic structure of the weak decay @xcite and therefore will be called parton interaction , or pi .
the other comes from final state hadron - hadron interactions @xcite and we shall call it fsi . in both approaches one is forced to make non - trivial calculation assumptions and the parameters and form factors are not well known .
however as we are dealing with dalitz distributions , the comparison of these different contributions with the experimental distributions might discriminate which dynamical mechanism is more realistic to describe data . in fig .
1 we show the diagrams relevant for the partonic description of this process . figs .
1a and 1b show the direct and penguin contribution to the `` current induced '' processes . figs .
1c ) , 1d ) and 1e ) show the direct ( 1c ) ) and and penguin ( 1d ) and 1e ) ) contributions to the `` transition '' processes and and figs .
1f ) and 1 g ) show the direct and penguin `` annihilation '' processes .
looking at diagrams 1a ) to 1e ) we can see that in the beginning both @xmath17 and @xmath18 quarks can carry a large momentum . when they emit a @xmath19 or a gluon
, the bosons can also carry a large momentum and therefore , with the exception of the @xmath20 and @xmath21 quarks ( which come always from the vacuum and are soft ) any of the quarks in the final state may carry a large momentum . this large momentum
is then transferred to the final mesons .
the final momentum distribution of the three mesons will be eventually non - uniform but there is no reason for producing a softer @xmath13 .
the `` entanglement '' of the @xmath17 , @xmath18 and bosonic lines will distribute more or less democratically the initial high momenta .
in sharp contrast to this situation are diagrams 1f ) and 1 g ) , where after the boson emission and absorption the high momenta are _ only _ with the @xmath18 and @xmath22 quarks . in the middle of the diagrams we see
the @xmath13 made of @xmath18 and @xmath20 quarks pairs taken from the vacuum .
this @xmath13 will be comparatively softer than the negative pions of diagrams 1a ) - 1e ) .
when the final three mesons are produced through final state interactions we can also understand the softer @xmath13 in a very simple way : first the @xmath23 decays into @xmath14 and @xmath24 and then these interact creating @xmath14 , @xmath15 and @xmath13 .
the first two mesons carry the valence quarks of the more energetic intermediate @xmath14 and @xmath24 and therefore will have higher momenta .
the @xmath13 is produced with quarks taken from the vacuum and therefore is softer .
this is illustrated in fig.[fsi1 ] .
thus the appearance of a comparatively soft @xmath13 is nothing but the manifestation of the `` leading particle effect '' , so frequently observed in other processes in hadron physics . to the best of our knowledge
, this kind of process has been first observed in a dalitz plot by the aachen - berlin - cern collaboration in the reaction @xmath25 @xcite and also in @xmath26 @xcite . shortly after the experimental observation it was theoretically understood in @xcite . in that paper
it was explained in terms of double regge graphs .
this mechanism implies the production of a soft pion in the central region , responsible for a bump at the corner of the dalitz plot , hence the name _ cornering effect_. the analysis performed in @xcite , suggests that in the chiral limit approximation , i.e. when @xmath27 , diagram 1f ) vanishes .
it also suggests that the most important process is the one shown in fig .
1 g ) , which has a striking similarity with the diagram of fig .
both have the same final quark flow and final meson formation . from the above qualitative discussion and from figures 1 g ) and [ fsi1 ]
we conclude that in the decay @xmath28 we expect to see a softer @xmath13 which will yield a non - uniform dalitz distribution .
this effect comes both from the partonic weak decay and from final state hadronic interactions and has , in both cases , the same physical origin : valence quarks form hard mesons in the final state while softer mesons are produced from the vacuum . neglecting the flavor change ( @xmath29 ) in fig
. 1 g ) , we can say that in both cases , pi and fsi , we have a leading particle effect . in
what follows we will implement these ideas in a more quantitative way .
in view of the uncertainties in the partonic description , specially the use of the factorization hypothesis , we will develop here only the hadronic final state interaction approach , for which the relevant lagrangian densities and form factors are better known from a large body of phenomenological studies at low and intermediate energies .
therefore we shall suppose that the non - resonant three - body decay @xmath30 proceeds through a hadronic scattering between two intermediate mesons @xmath31 and @xmath24 in the two - body decay @xmath32 .
hadronic final state interactions have a long story and follow two different approaches : regge models and meson exchange models .
the former have been considered in many works @xcite whereas the latter were discussed in @xcite .
regge theory is formulated in the high energy limit and in it all the amplitudes respect unitarity constraints .
the interactions are represented by reggeon exchanges , the pomeron being usually the most important one .
as one moves to the low energy region , secondary reggeons become important , additional assumptions have to be made and the theory looses predictive power .
it is not clear when the energy starts to be `` high '' and regge theory has been applied to energies of the order of a few gev .
meson exchange models describe hadronic interactions fairly well at intermediate energies ( a few hundreds mev ) .
they contain uncertainties associated with higher order diagrams , multiple exchange terms , coupling constants and the spatial extension of hadrons .
these uncertainties are translated into form factors , which , in turn , can be either calculated ( e.g. with qcd sum rules @xcite ) or simply parametrized and fitted to data .
a positive aspect of meson exchange models is that one can use effective lagrangians for them and so enforce chiral symmetry , which is known to be very important at intermediate energies . on the other hand , in this kind of model ,
the amplitudes are not unitary .
the lack of unitarity is believed to become a serious problem at increasing energies .
these considerations suggest that with inceasing energies we should change the dynamical description , going from meson exchange to reggeon exchange , but is not clear _ at which energy _
this should be done . in this work
we shall study final state interactions with meson dynamics . in principle
the @xmath1 meson can decay into inumerous hadronic two - body intermediate states which will subsequently decay into @xmath33 .
these states include : @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath42 , @xmath43 , @xmath44 , @xmath45 ... these mesons can undergo a two - to - three reaction , exchanging virtual mesons and creating the @xmath13 .
this is illustrated in fig .
[ fsi1 ] for the @xmath35 intermediate state , which we will use as working example .
a more careful analysis reveals that some of the above mentioned two - body states can be excluded because their interactions yielding a @xmath46 final state would violate @xmath47 parity ( this is the case of the @xmath34 and @xmath42 ) .
some other states are much less abundant as two body decays .
this is the case of the @xmath40 and @xmath41 states , which branching fraction is one order of magnitude smaller than the @xmath35 one .
other diagrams are suppressed by dual topological unitarity ( dtu ) @xcite .
finally , some other states , inspite of being different , are expected to yield amplitudes which are qualitatively very similar .
this is , for example , the case of the states @xmath35 and @xmath36 . replacing @xmath24 by @xmath48 in fig .
[ fsi1 ] will lead to a similar diagram with the vertex @xmath49 instead of @xmath50 . this will change the corresponding coupling constant and the form factor .
the former is unable to distort the dalitz plot and irrelevant for our discussion . in both vertices
the form factor will suppress highly virtual internal @xmath51 mesons .
we therefore expect these two diagrams to give similar results .
even after a careful scrutiny and the elimination of suppressed diagrams and of `` double counting '' , there will remain a few diagrams to be considered . here
we shall work out in detail only the process depicted in fig .
this will show schematically how the final state meson dynamics will lead to inhomogeneities in the dalitz plot . before starting the evaluation of the feynman diagram given in fig .
[ fsi1 ] , one last remark is in order . in other fsi calculations , such as @xcite the starting point is the @xmath1 meson , which decays into two off - shell mesons , which , in turn exchange a third virtual meson .
this process is described by a loop diagram , from which the absorptive part is considered . in our fig .
[ fsi1 ] this procedure would be equivalent to close the lower part of the diagram forming a loop with a virtual @xmath14 and a virtual @xmath24 . here , for simplicity , we take them to be real . in this way we will give emphasis to the creation of the soft @xmath13 and consequent dalitz plot distortion .
the loop integral will introduce some smearing in our result and presumably change the normalization .
since our purpose in this work is to discuss this process qualitatively , we postpone to the future a more general calculation , including effects of the kaon and pion off - shellness .
the effective lagrangians relevant to the calculation can be constructed from a chiral lagrangian @xcite .
they are : @xmath52 with the above interaction lagrangians , we can write the amplitude for the diagram in fig . [ fsi1 ] as : = 4ig_kkg_g_^p_5p_2p_1p_3((p_1-p_3)^2-m_^2 ) ( ( p_5-p_2)^2-m_^2 ) , [ amp ] where @xmath53 and @xmath54 denote the momenta of the @xmath14 and @xmath24 in the initial state ; @xmath55 and @xmath56 are those of the @xmath57 and @xmath15 in the final state of the diagram in fig .
it is important to notice that @xmath58 , where @xmath59 is the momentum of the @xmath1 meson , since we are considering the final state interactions in the two - body decay @xmath60 . in order to take into account the effects of hadron internal structure we follow ref .
@xcite and introduce , in the amplitude , the form factor : f(q_,q_)=(^4 ^ 4+(q_^2-m_^2)^2 ) ( ^4 ^ 4+(q_^2-m_^2)^2 ) , where @xmath61 and @xmath62 are the four momenta of the intermediate off - shell vector mesons , _
i.e. _ , @xmath63 and @xmath64 and @xmath65 is a cut - off parameter taken to be @xmath66 gev .
parametrizing the amplitude of the weak decay @xmath60 through the weak decay coupling @xmath67 , we can write the amplitude for the three - body decay , @xmath68 , in terms of the @xmath69 ( @xmath70 ) and @xmath71 ( @xmath72 ) momenta as : & & a_k= -32i^8g_bkg_kkg_g_(q_1 ^ 2 - 4m_^2)(^4+(q_1 ^ 2/4-m_^2)^2 ) + & & ^()_p_bp _ ( ) _ ( q_2 ^ 2 - 4m_^2)(^4+(q_2 ^ 2/4-m_^2)^2 ) , [ ampf ] where @xmath73 , @xmath74 and @xmath75 . in terms of these momenta
, the three - body decay rate is given by : d=|a_k|^22 ^ 4 ^ 5m_b ( ( + -p_b)^2-m_^2)((p_b .
+ -.)^2-m_^2)((p_b-)^2-m_k^2 ) d^4 d^4
. [ rate ] evaluating eq .
( [ rate ] ) and using the delta functions to perform some of the integrals , we finally write the decay rate in the standard form for the dalitz plot : = d(_k ) d _ , [ frate ] where , as in eq .
( [ anr ] ) , @xmath76 and @xmath77 . in the rest frame of the @xmath1 meson and due to the delta functions , in evaluating
@xmath78 one has to use : p^2=2m_^2 + 2m_k^2-m_b^2 , with e_p = m_k^2-m_^2m_b || = m_b , where @xmath79 . for the other momenta and scalar products we have : q_1 ^ 2=2s_23 - 2m_b^2 + 2m_k^2 + 4m_^2 + 4p .
, q_2 ^ 2=2s_13 - 2m_b^2 + 2m_^2 + 4m_k^2 - 4p . ,
p.=e_pe_k-|||_k| , p.&=&e_pe_-|||_| ( & + & | ) , & = & ( m_k^2-m_^2-m_b^2+s_13+s_23 + + & + & .(s_13-m_^2)(s_23-m_k^2)m_b^2 ) 14|_k||_| , e_k = m_b^2+s_13-m_^22m_b , e_=m_b^2+s_23-m_k^22m_b , |_|=2m_b , |_k|=2m_b , .=m_b^2-m_k^22 . for a given value of @xmath80 ( @xmath81 ) , the range of @xmath82 is determined by imposing @xmath83 and is given by @xcite : ( s_23)_min&=&(e^*_1+e^*_2)^2-(+)^2 + ( s_23)_max&=&(e^*_1+e^*_2)^2-(-)^2 where e^*_1&=&s_13-m_k^2+m_^22 , + e^*_2&=&m_b^2-s_13-m_^22 . before presenting our numerical results we would like to emphasize that _ we do not wish to reproduce the absolute normalization of the decay rate_. the only purpose of our calculation is to show that the meson exchange mechanism is able to produce distortions in the dalitz plot of the @xmath68 decay . in fig .
[ dali ] we show the dalitz plot of a monte carlo simulation of eq .
( [ frate ] ) .
the distribution shown in the figure is consistent with the parametrization in eq.([anr ] ) but it goes to zero near the threshold .
this result gives support to the expectation presented in section ii : the @xmath14 and @xmath15 carry the valence quarks from the intermediate @xmath14 and @xmath24 and are therefore hard , while the @xmath13 is created from the vacuum and is therefore softer .
the dalitz distribution of fig .
[ dali ] can be parametrized as where f_i ( x ) = [ fermi ] with @xmath84 gev@xmath85 , @xmath86 gev@xmath87 , @xmath88 gev@xmath89 , @xmath90 gev@xmath87 and @xmath91 gev@xmath89 .
the meson exchange mechanism considered above could play the same role in the @xmath92 three - body decay , where one does not see any inhomogeneity in the non - resonant amplitude . in the case of the @xmath93
, the three - body decay could proceed through final state interactions between the @xmath94 and @xmath15 mesons in the two - body decay @xmath95 . in fig .
[ diagra2 ] we show the feynman diagram for the @xmath92 decay through the @xmath96 interaction .
we see that there are no mesons that could be exchanged satisfying all the conservation laws required by strong interactions .
this means that the diagram does not exist and the three - body decay @xmath92 does not proceed through final state interactions .
a similar situation occurs in the thre - body decay @xmath97 , since there is no possible meson exchange that would lead the two mesons @xmath98 and @xmath15 in the two - body decay @xmath99 into the final three mesons : @xmath100 . on the other hand
, the three - body decay @xmath101 can proceed through the two - body decay @xmath102 , followed by final state interactions between @xmath14 and @xmath13 , as can be seen by the feynman diagrams in fig .
[ diagra3 ] .
therefore , in the case of the three - body decay @xmath101 , one expects some inhomogeneity in the non - resonant amplitude , consistent with a hard @xmath14 , a hard @xmath103 and a softer @xmath104 . this effect was not included in the data analysis performed by the babar colaboration in ref .
@xcite , where a homogeneous parametrization for the non - resonant component was used . to close this section
we would like to comment the belle results on baryonic three body decays .
in this context the fsi approach was first used in @xcite to study rare baryonic @xmath1 decays .
the study of the decays @xmath105 , @xmath106 @xcite revealed a low mass enhancement of the system @xmath107 .
however this enhancement in the low mass ( close to the threshold ) baryon - antibaryon system was not seen in the decays @xmath108 and @xmath109 @xcite .
in terms of fsi we can very easily understand these results .
the mass of the baryon - antibaryon system is low because the @xmath110 is soft . in the case of light meson production
we have the following sequence of decays : @xmath111 .
the intermediate pion splits into a proton and an antiproton and the latter interacts with the kaon exchanging a @xmath51 or an @xmath112 .
the emerging @xmath110 has all the antiquarks coming from the vacuum and will therefore carry low momentum . in the case of @xmath113 production
the @xmath1 goes first to an intermediate state with @xmath113 , such as , for example , @xmath114 .
the pion could then split into @xmath115 and @xmath110 but now the subsequent elastic interaction between the @xmath113 and the @xmath110 is ozi suppressed , since the two mesons have no quarks in common .
therefore , in this case there is no soft @xmath110 from fsi and no low mass enhancement in the @xmath116 system . this same problem has been addressed in @xcite where a similar conclusion was obtained .
we hope to have made clear that , for experimental purposes , it is highly desirable to have a parametrization for the amplitude of non - resonant three body decays .
this parametrization should have a physical basis and should be derived from theory .
the already existing parametrization given in eq .
( [ anr ] ) is not satisfactory from this point of view . in order to improve it
we have considered one possible mechanism responsible for inhomogeneities in the dalitz plot , which we have called final state interactions ( fsi ) .
we have sketched calculations of the dalitz distributions with this mechanism .
these calculations can certainly be improved , but already at this stage they can teach us how to obtain distortions in the non - resonant three body decays .
fsi trace back the observed dalitz distributions to a manifestation of the leading particle effect .
this work has been partly supported by fapesp and cnpq .
a. deandrea , m. ladisa , v. laporta , g. nardulli and p. santorelli , int .
j. mod .
phys . a 21 ( 2006 ) 4425 ; h. y. cheng , int .
j. mod .
phys . a 21 ( 2006 ) 650 .
j. f. donoghue , e. golowich , a. a. petrov and j. m. soares , phys .
lett . 77 ( 1996 ) 2178 .
n. isgur , k. maltman , i. weinstein and t. barnes , phys .
64 ( 1990 ) 161 ; h. lipkin and b.s .
zou , phys .
d 53 ( 1996 ) 6693 ; p. geiger and n. isgur , phys .
67 ( 1991)1066 ; x.q .
bugg and b.s .
zou , phys .
d 55 ( 1997)1421 . | we evaluate the non - resonant decay amplitude of the process @xmath0 using an approach based on final state hadronic interactions described in terms of meson exchanges . we conclude that this mechanism generates inhomogeneities in the dalitz plot of the @xmath1 decay . |
the _ swift _ satellite @xcite certainly revolutionised the grb field , but it has also shaken some of our pre_swift _ strong beliefs .
one of those was the reassuring division between the prompt and the afterglow phases understood in terms of internal shocks between relativistic shells followed by the external shock of the fireball with the circumburst medium .
this scenario could well explain the erratic and violent prompt emission lasting up to a few tens of seconds , and most of the ( then existing ) observations of the afterglow , typically starting a few hours after the trigger .
the afterglow phase was thought to be somewhat better understood , since it was constructed upon the very solid basis of the conservation of energy and momentum , shock acceleration physics and the synchrotron process as the main radiation mechanism ( that was even confirmed by the detection of a small , but significant , linear polarisation ; see @xcite for a recent review ) . although some problem remained to be solved for the prompt phase ( as the inherently low efficiency of internal shocks and the hardness of the observed spectra if the emission process is synchrotron by cooling electrons , @xcite ) ,
the overall scenario was satisfying , especially after the discovery that long grbs were indeed associated with powerful supernovae , confirming that the central engine had to be an hyper accreting stellar size black hole .
the _ swift _ capability of a very fast slew allowed to explore the very early phases , soon discovering the unexpected steep flat steep " behaviour of the light
curve , especially in the x - rays @xcite .
although interpreted in several ways ( see e.g. @xcite for a recent review ) , none seems conclusive . the spectral slope does not change across the temporal break from the shallow to the normal decay phase , ruling out a changing spectral break as a viable explanation .
an hydrodynamical or geometrical nature of the break is instead preferred .
specific models often focused only on the x ray behaviour , but there is another crucial information that can not be overlooked : _ the x ray and optical light
curves often do not track one another _
( e.g. @xcite ) , in a way that can not be accommodated by the standard external shock / fireball scenario .
it is this remarkable characteristic that suggests that there must be two components contributing to the observed flux .
[ f1 ] shows two examples of optical and x ray afterglows , to illustrate the different behaviour of the optical and x ray light curves in grb 061126 , while they track one another in grb 060614 .
we ( @xcite , see also @xcite ) then used a simple phenomenological 2component model ( briefly described in the following section ) to fit _ both _ the x ray and optical light curves of _ swift _ grbs with redshift and a well sampled optical light
curve , allowing to estimate the optical extinction due to dust in the host galaxies . as of march 2008
, they amount to 33 grbs .
the logic we follow is to see if with a minimum number of parameters we can fit the observed light curves of these bursts .
if so , we can find some hints for a more physical interpretation .
we assume that at all times the flux is the sum of two components : a standard afterglow plus ( for the time being ) a completely phenomenological second component . * component 1 : * this is the synchrotron radiation produced by the standard forward shock caused by the fireball running into the circum burst material . for this component
we follow the prescription of @xcite , that requires 6 free parameters , plus the assumption of an homogeneous or a wind like profile of the circumburst density .
these are : i ) the isotropic equivalent kinetic energy of the fireball @xmath1 ; ii ) its initial bulk lorentz factor @xmath2 ; iii ) the value of the circum burst medium density @xmath3 ;
iv)v ) the `` equipartition '' parameters @xmath4 and @xmath5 ; vi ) the slope of the relativistic electron energy distribution @xmath6 . *
component 2 : * it is treated phenomenologically , since its form / origin is not currently known , though it can be possibly ascribed to the extension in time of the early prompt emission ( as discussed below ) .
we parametrise this component with the only criterion of minimising the number of free parameters . for simplicity
we assume that its spectral shape is constant in time and is described by a broken power
law , with spectral indices @xmath7 and @xmath8 below and above the break frequency @xmath9 .
the temporal parameters are described by the flat and steep decay indices , @xmath10 and @xmath11 respectively , and the time @xmath12 at which the two behaviours join .
the time @xmath12 is the time at which the shallow phase ( of the steep flat
steep behaviour ) ends .
then we have 3 spectral and 3 temporal parameters , and we must add one normalisation .
component 2 is thus determined by 7 free parameters , that can be rather well constrained by observations when this component dominates the emission . in this case @xmath10 , @xmath11 and @xmath12 can be directly determined , as well as one spectral index ( usually @xmath13 , since the late prompt emission is usually dominating in the x ray range ) .
, as a function of the isotropic energy of the prompt emission , @xmath14 .
the dashed line corresponds to the least square fit , @xmath15 \propto e_{\rm \gamma , iso}^{0.86}$ ] ( chance probability @xmath16 , excluding the outlier grb 070125 ) .
right panel : the kinetic energy @xmath1 after the prompt emission as a function of @xmath17 .
the dashed line is a least square fit , yielding @xmath18 ( chance probability @xmath19 , excluding grb 070125 ) .
adapted from @xcite .
, title="fig:",height=321 ] , as a function of the isotropic energy of the prompt emission , @xmath14 .
the dashed line corresponds to the least square fit , @xmath15 \propto e_{\rm \gamma , iso}^{0.86}$ ] ( chance probability @xmath16 , excluding the outlier grb 070125 ) .
right panel : the kinetic energy @xmath1 after the prompt emission as a function of @xmath17 .
the dashed line is a least square fit , yielding @xmath18 ( chance probability @xmath19 , excluding grb 070125 ) .
adapted from @xcite .
, title="fig:",height=321 ]
fig . [ f1 ] shows two examples of our modelling ( in @xcite one can find all the 33 grbs we have analysed ) . in these two illustrative examples both grbs
are dominated by component 2 at late times in both bands , while the standard afterglow emission ( i.e. the synchrotron emission from the external shock ) is important only at very early times .
the complex shape of the optical light curve of grb 061126 is due to the emerging of component 2 at @xmath201000 s. the break at @xmath21 s has nothing to do with a jet break , but is the end of the shallow component ( i.e. it is @xmath12 , the time break of component 2 , see @xcite for more discussion ) .
if this burst ( of known @xmath22 ) had to follow the ghirlanda " relation , then it should have a jet break a few days after trigger ( rest frame time ) as indicated by the vertical line .
but at this time the emission is dominated by component 2 , _ thus the jet break is hidden , and can not be observed_. see also @xcite for more examples and discussion of this point . for grb 060614 the entire observed light curves are dominated by component 2 : also in this burst the very prominent and _ achromatic _ break at @xmath2040,000 s ( rest frame ) has again nothing to do with a jet break , being associated to @xmath12 .
note that in our modelling we do not treat the steep phase immediately following ( in most , but not all grbs ) the prompt phase , nor the x ray or optical flares .
consider also that the standard afterglow code we use assume isotropic emission , and so it can not produce any jet break ( i.e. for simplicity we do not use the opening angle of the jet as a free parameter ) . and
@xmath23 , as labelled .
the x ray luminosity profile is flatter than @xmath23 and closer to a @xmath24 decay .
however , this behaviour is due to the contribution in some grbs of the afterglow emission at late times , flattening the overall light curve .
in fact the right panel shows the modelled x ray light curves . on the left
we show component 2 ( i.e. the late prompt " ) , in the middle the afterglow and on the right their sum .
the dashed lines correspond to @xmath25 or @xmath26 , as labelled . adapted from @xcite .
, title="fig:",height=321 ] and @xmath23 , as labelled .
the x ray luminosity profile is flatter than @xmath23 and closer to a @xmath24 decay .
however , this behaviour is due to the contribution in some grbs of the afterglow emission at late times , flattening the overall light curve .
in fact the right panel shows the modelled x ray light curves . on the left
we show component 2 ( i.e. the late prompt " ) , in the middle the afterglow and on the right their sum .
the dashed lines correspond to @xmath25 or @xmath26 , as labelled . adapted from @xcite .
, title="fig:",height=321 ] considering now the entire sample of 33 grbs , we can see if there is some relation between the ( kinetic ) energy of the fireball @xmath1 and the energy emitted during the prompt phase ( i.e. @xmath14 ) .
similarly , we can see how the energy emitted by component 2 , approximated by @xmath27 , relates with @xmath14 .
we are aware of the danger of finding signs of a correlation when considering two energies ( or luminosities ) , since both quantities are function of redshift , but we can compare the two plots . fig .
[ f2 ] shows that @xmath27 correlates with @xmath14 better than @xmath1 .
this suggests that the early prompt and component 2 are related .
the most important results of our analysis is the found distribution of @xmath11 , the decay index of component 2 after @xmath12 . fig .
[ f3 ] shows a clustering around @xmath28 .
this value is equal to the time profile of the accretion rate of fallback material in supernovae ( @xcite , @xcite , @xcite ) , and to the average decay of the x ray flare luminosity , as analysed by @xcite .
this is _ not _ the average observed decay slope : the x ray light
curves are flatter than @xmath29 ( see fig . [ f4 ] ) , but this is due to the contribution , especially at early and late times , of the emission of the standard afterglow . the light
curves of component 2 are indeed steeper , on average , than the total emission that reproduces the data .
the fact that @xmath30 strongly suggests that component 2 can be interpreted as due to the late time accretion of fallback mass , namely material that failed to reach the escape velocity from the exploding progenitor star , and falls back .
this can continue for weeks , enough to sustain late prompt emission even at very late times .
furthermore , the fact that also x ray flares follow a similar behaviour suggests that both x ray flares and component 2 have a common origin , related to the accretion of the fallback material .
the puzzling features of the early afterglow disclosed by _ swift _ have revealed an unforeseen complexity , difficult to explain in terms of the standard internal / external shock scenario .
many alternatives have been proposed ( see the introduction of @xcite for a brief list of models and @xcite for a review ) , but only a few are able to explain a different behaviour of the optical and x ray light curves .
uhm & beloborodov @xcite and genet , daigne & mochkovitch @xcite suggested that the x ray plateau emission is not due to the forward , but to the reverse shock running into ejecta of relatively small ( and decreasing ) lorentz factors .
the optical can instead be due to the standard emission of the forward shock .
this scenario requires an appropriate @xmath31distribution of the ejecta , and also the suppression of the x ray flux produced by the forward shock .
we ( @xcite ) instead suggested that the plateau phase of the x ray emission ( and sometimes even of the optical ) is due to a prolonged activity of the central engine ( see also @xcite ) , responsible for a `` late
prompt '' phase : after the early standard " prompt the central engine continues to produce for a long time ( i.e. days ) shells of progressively lower power and bulk lorentz factor .
the dissipation process during this and the early phases occur at similar radii ( namely close to the transparency radius ) .
the reason for the shallow decay phase , and for the break ending it , is that the @xmath31factors of the late shells are monotonically decreasing , allowing to see an increasing portion of the emitting surface , until all of it is visible .
then the break occurs when @xmath32 , at @xmath12 .
the shallow phase is then the result of a balance : the total emitted luminosity is decreasing ( with a decay slope @xmath11 ) , but for some time ( before @xmath12 ) the surface visible to us is increasing ( because of the decreasing @xmath31 ) .
the combination of these two effects flattens the observed flux decay , that appears to be characterised by @xmath10 .
after @xmath12 the entire emitting surface is visible , and we see the real " decay slope @xmath11 . with respect to other models ,
late prompt " scenario is very economic , since the total extra energy needed is a relatively small fraction of what used during the real prompt phase .
finally , we would like to collect the different pieces of evidence found in these studies , to try to construct an heuristic simple scenario .
let us assume that , following the death of a massive star , a rapidly spinning black hole is formed .
the fast rotation of the equatorial material prevented it to immediately fall into the black hole .
this material forms a very dense torus .
viscosity and angular momentum conservation makes this torus to spread , and accretion takes place .
this accretion phase corresponds to the real prompt emission , possibly mediated by the strong magnetic field formed in the vicinity of the black hole , making the blandford & znajek @xcite mechanism at work .
the energy stored in a maximally spinning black hole of 2 solar masses , and that can be extracted , amounts to @xmath33 erg .
we need only a few per cent of it .
when the bulk of the dense torus has been accreted , there must be a discontinuity in the accretion rate , to explain the steep behaviour of the light
curve following the real prompt emission .
this discontinuity could be associated to the transition from the end of the accretion of the material of the dense torus and the beginning of the accretion of the fallback material .
after this transition phase accretion proceeds at a reduced rate : this phase is associated to fallback , with its typical time profile .
a reduced accretion most likely corresponds to a reduced magnetic field in the vicinity of the black hole , and so to a reduced capacity to extract the spin energy of the black hole .
occasionally , however , fragmentation of the accreting material leads to temporary enhanced accretion , that we can associate to the x ray flares .
9 n. gerhels , g. chincarini , p. giommi , et al . , _ apj _ , * 611 * , 1005 ( 2004 ) s. covino , proceedings of the conference `` the coming of age of x - ray polarimetry '' , rome , italy , 27 - 30 april 2009 ( astro ph.0906.5440 ) g. ghisellini g. , a. celotti , d. lazzati d. , _ mnras _ , * 313 * , l1 ( 2000 ) g. tagliaferri , m. goad , g. chincarini , et al . , 2005 , _ nature _ , * 436 * , 985 j.a .
nousek , c. kouveliotou , d. grupe , et al . , 2005 , _
apj _ , * 642 * , 389 b. zhang , 2007 , _ advances in space research _ , * 40 * , issue 8 , p. 1186
( astro ph/0611774 ) a. panaitescu , p. meszaros , d. burrows , n. nousek et al .
2006 , _ mnras _ , * 369 * , 2059 a. panaitescu , _ il nuovo cimento _ , * 121 * , 1099 ( 2006 ) g. ghisellini , m. nardini , g. ghirlanda , a. celotti , _ mnras _ , * 393 * , 253 ( 2209 ) m. nardini , g. ghisellini , g. ghirlanda , a. celotti , this volume ( 2009 ) l. nava , g. ghisellini , g. ghirlanda , et al .
, _ mnras _ , * 377 * , 1464 ( 2007 ) a. panaitescu , p. kumar , _ apj _ , * 543 * , 66 ( 2000 ) g. ghirlanda , g. ghisellini & d , lazzati , 2004 , _ apj _ , * 616 * , 331 g. ghirlanda , l. nava , g. ghisellini , c. firmani , _ a&a _ ,
* 466 * , 127 ( 2007 ) d. lazzati , r. perna , m.c .
begelman , _ mnras _ , * 388 * , l15 ( 2008 ) r.a .
chevalier , _ apj _ , * 346 * , 847 ( 1989 ) a.i .
macfadyen , s.e .
woosley , a. heger , _ apj _ , * 550 * , 410 ( 2001 ) w. zhang , s.e .
woosley , a. heger , 2007 , _ apj _ , * 679 * , 639 ( 2007 ) l.z .
uhm & a.m. beloborodov , _ apj _ , * 665 * , l93 ( 2007 ) f. genet , f. daigne , & r. mochkovitch , _ mnras _ , * 381 * , 732 ( 2007 ) g. ghisellini , g. ghirlanda , l. nava & c. firmani , _ apj _ , * 658 * , l75 ( 2007 ) d. lazzati & r. perna , _ mnras _ , * 375 * , l46 ( 2007 ) r.d .
blandford , r.l .
znajek , _ mnras _ , * 179 * , 433 ( 1977 ) | we call prompt " emission of gamma ray bursts ( grbs ) the erratic and violent phase of hard x - ray and soft @xmath0ray emission , usually lasting for tens of seconds in long grbs
. however , the central engine of grbs may live much longer .
evidence of it comes from the strange behaviour of the early afterglow " , seen especially in the x ray band , characterised by a
steep flat steep " light curve ,
very often not paralleled by a similar behaviour in the optical band .
this difference makes it hard to explain _ both _ the optical and the x ray emission with a unique component .
two different mechanisms seem to be required .
one can well be the standard emission from the forward shock of the fireball running through the interstellar medium .
the second one is more elusive , and to characterise its properties we have studied those grbs with well sampled data and known redshift , fitting _ at the same time _ the optical and the x ray light
curves by means of a composite model , i.e. the sum of the standard forward shock emission and a phenomenological additional component with a minimum of free parameters .
some interesting findings have emerged , pointing to a long activity of the central engine powered by fallback material .
address = inaf osservatorio astronomico di brera , via e. bianchi , 46 i23807 , merate , italy address = sissa / isas , via beirut 24 , i34151 , trieste , italy , address = inaf osservatorio astronomico di brera , via e. bianchi , 46 i23807 , merate , italy |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.