article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the central problem of nuclear structure theory is the solution of the many - body schrdinger equation ( mbse ) . for hamiltionans of interest in the nuclear case ,
an analytic solution is impossible , and one is compelled to use some approximation , either in the numerical solution of the equation or the specification of the hamiltonian , or both . approaching the problem with the aim of using as realistic a representation of the potential as possible usually means fitting a combination of a meson exchange and phenomenological interaction to low - energy nucleon - nucleon scattering data and properties of few - body systems . to get good agreement with experiment both two- and
three - body forces seem to be necessary .
recent examples of such potentials include the bonn @xcite , the argonne two - body @xcite with urbanna 3-body@xcite , nijmegen @xcite and moscow @xcite potentials , the last of which also incorporates quark degrees of freedom .
these forces share the property of having a hard repulsive core which is a natural consequence of meson - exchange .
it is this hard core which presents the difficulty in solving the mbse .
for instance , hartree - fock ( hf ) mean - field calculations with such interactions result in unbound nuclei .
treating corrections beyond the hf approximation order - by - order in perturbation theory is also unsuccessful since the interactions used are non - perturbative .
one has to solve the full mbse numerically in as exact a way as possible using techniques such as variational monte - carlo@xcite , green s function monte carlo@xcite , the coupled - cluster method@xcite , and the fermion hyper - netted chain model@xcite . using effective interactions derived from realistic potentials , no - core shell - model calculations have been made in light nuclei@xcite and heavier nuclei close to closed shells have been treated@xcite .
the computational difficulty of performing numerically exact solutions of the mbse has limited the techniques to light nuclei , for instance @xmath2 results have been published recently using the argonne @xmath3 and urbanna ix potentials in the gfmc framework@xcite . in this work
, it is seen that although the lightest nuclei are reproduced very well , the quantitative comparison of theory to data gets worse as @xmath4 increases .
this may be due to the necessarily phenomenological nature of the three body potential , a problem which may be overcome with re - fitting .
on the other hand , it is not obvious that higher - body forces will not prove necessary or that the concept of a bare interaction between nucleons is valid for small distances .
attempts were made in the late sixties primarily by the kerman group at mit to parameterize the nn interaction in such a way that it is weak in the sense of being perturbative .
such a weak interaction allows one to perform hartree - fock calculations to obtain a reasonable approximation to the full wavefunction and then to calculate corrections in perturbation theory .
while this technique seems very attractive , the result obtained were only moderately successful at reproducing experimental data@xcite , a fact which was presumed to be due to inadequacies in the potentials used .
the efficacy of developing a suitable interaction when similar , though more complicated , techniques were available for realistic interactions has been questioned@xcite and no better interaction was developed .
separable parameterizations , particularly the quadrupole - quadrupole interaction@xcite , have retained currency , as residual interactions @xcite . even when the interactions are too strong for regular perturbation theory , separable interactions requiring solution of brckner hartree - fock equations have proved fruitful@xcite because of their simplicity . on the other hand ,
interactions have been developed which are not intended for use in the full mbse , but rather to give good results with a hartree - fock calculation alone .
good quantitative success came with the zero - range density - dependent force of ehlers and moszkowski@xcite and skyrme s interaction@xcite , used in hf calculations by vautherin and brink @xcite and subsequently by many others , and also gogny s finite - range interaction@xcite .
skyrme s interaction has been particularly successful , in part due to its simple form , that of a delta function , which leads to easy calculation , even of the exchange part of the force .
this computational simplicity has allowed extensive study of the properties of nuclei to be made with the skyrme interaction across the entire range of nuclei in the periodic table @xcite . related somewhat to the skyrme - hartree - fock model
is the relativistic mean field ( rmf ) approach@xcite , which also gives single - particle motion in a mean field , but as a solution to the dirac equation as opposed to the schrdinger equation .
the rmf approach has some nice features such as the natural occurrence of the spin - orbit splitting without recourse to an assumed spin - orbit interaction .
these mean - field models are inherently single - particle in nature .
the forces used are not intended to be used in the mbse , nor are explicit corrections beyond the mean - field part of the framework , although skyrme s interaction can be considered as a phenomenological g - matrix equivalent@xcite , and in that sense includes a subset of possible correlation effects in the mean - field .
although appropriate for use in mean - field calculations , skyrme s interaction would actually diverge in perturbation theory because of the zero range .
this compels one to use a different interaction to obtain correlation beyond the mean field than was used to create it .
extra residual forces are used , such as pairing@xcite , or shell model interactions@xcite to allow for more general wavefunctions and obtain more accurate reproduction of the physics , or certain approximations are used such as rpa@xcite , which can describe certain observables , particularly those of giant resonance states and other forms of collective motion , but not others .
it is thought that correlations should be particularly important in nuclei at the limits of stability , where the nucleons nearest the fermi level couple strongly to the continuum .
we revisit the idea that it is possible to parameterize a nuclear interaction in such a way that it is weak enough with which to perform perturbation theory , thereby allowing correlated wavefunctions and observables to be calculated across the entire range of nuclei .
using the separable ansatz of previous `` weak '' interactions we have developed a density dependent interaction which we hope will provide some insight into the correlation structure of nuclear wavefunctions while retaining the quantitative power of contemporary effective interactions used in hartree - fock .
in contrast to previous work , the interaction is designed to be an effective interaction with parameters fitted to the properties of finite nuclei within the calculation framework for which it is intended .
the paper is organized as follows . in section 2
we briefly review basics of many - body perturbation theory .
the separable interaction , used in the present work is given and discussed in section 3 .
results of the calculation for doubly magic nuclei are summarized in section 4 .
derivation of the hf energy and potential is outlined in appendix a.
for standard perturbation theory to be successful , the hamiltonian must be separated into two parts , one which is solvable ( @xmath5 ) , and another which is `` small '' ( @xmath6 ) .
the hamiltonian for a many - fermion problem in which the particles interact via one- and two - body interactions may be written schematically as @xmath7 typically , the one - body part of the hamiltonian is just the kinetic energy .
this is a unsuitable choice for @xmath5 in the case of nuclei since the eigenstates of this operator , namely plane wave states , are not close enough to the exact eigenstates of the full hamiltonian and @xmath6 is thus not small . by adding and subtracting a one - body term , @xmath8 , of ones choice , the @xmath5 part of the hamiltonian may be solvable to give a wavefunction close to the solution of the full problem , thereby making the corrections from the @xmath6 part small enough for perturbation theory to succeed .
schematically the hamiltonian is now split up into two terms ; @xmath9 one practical choice of @xmath8 is a simple , analytically - solvable external potential , typically in the nuclear case of harmonic oscillator or woods - saxon form .
in such a case , all of the physics of the nuclear interaction is in the residual part , @xmath6 .
unfortunately , for sensible choices of two - body interaction , the residual part of the hamiltonian is non - perturbative in the basis obtained from the solution of the one - body part of the hamiltonian , and one has to perform more exact and exacting interaction shell - model calculations to arrive at a meaningful result .
traditionally , this has involved the diagonalisation of large matrices , or more recently , auxiliary field monte - carlo calculations in the smmc@xcite .
another popular choice for @xmath8 is that of the hartree - fock mean - field .
the hf mean - field potential is usually derived from the full hamiltonian by a variational principle .
viewed in this way , it is the one - body potential whose occupied eigenstates form the lowest energy slater - determinant many - body wavefunction possible for the full many - body hamiltonian .
using such a one - body potential , some physics of the two - body interaction is included in the single - particle problem . with a judicious choice of interaction , one ought to be able to produce very good results , since the approximation of the nucleus as a system of non - interacting particles in a mean - field is known to be a good one . using the hf potential for @xmath8 has the added attraction that it leads to vanishing first order corrections to the energy in perturbation theory . in this case , using wick s theorem@xcite , one can re - write ( [ eq : mbham ] ) as @xmath10 where @xmath5 is the hf mean - field hamiltonian and @xmath6 is the perturbing hamiltonian .
note that @xmath6 is just the full two - body interaction matrix elements with a time - ordered product of creation and annihilation operators .
the perturbation series for the energy is ordered by the number of matrix elements of the potential .
usually a diagrammatic representation is used@xcite .
one can write down the number of diagrams for any particular order of perturbation theory .
we have evaluated these for the vacuum amplitude up to order seven .
the results are presented in table [ tab : numdiags ] .
one sees that , for the method of direct evaluation of diagrams by order to be effective , the series must be sufficiently converged by fourth , or perhaps fifth , order . in this work ,
the second and third order diagrams for the vacuum amplitude , as shown in figs.[fig : seclab ] and [ fig : thilab ] , are calculated .
their algebraic form is given here as @xmath11 in which the tildes over the potential indicate the matrix element is antisymmetrized .
the state vectors label hf single - particle states , whose energies are given by the subscripted @xmath12 .
it is important to note that our interaction is not intended to fit scattering data , having , as it does , density dependence . on the level of the perturbation theory it is necessary to treat the density functions as just the spatial form of the interaction , rather than a representation of a many - body force
this is to be considered a part of the present model .
to do otherwise would be to surrender the simplifications our weak , separable potential affords . in the present work ,
the hartree - fock problem is solved in a basis of spherical harmonic oscillator states .
this yields , along with the hole states , a large number of particle states , with which to directly evaluate the sums of the perturbation series . a sufficient number of states is used so that the particle states are oscillatory over the size of the nucleus and that both the hf solution and the perturbation corrections are reasonably converged .
we have developed an interaction written in the form of a sum of separable terms , which is to say it is in the form @xmath13 .
the functions @xmath14 carry no angular momentum ( @xmath15 ) , and the force is dubbed a monopole - monopole interaction . for future applications
, it is in intended to include higher multipole forces , with @xmath16 , within our framework as these will presumably be necessary for calculation of excited states and deformed nuclei .
although higher multipole forces will contribute to spherical nuclei from the exchange term in hartree fock order and via correlations in perturbation theory , they are not included in the present calculation since it seems unwise to attempt to fit the parameters of such forces to spherical nuclei alone . in coordinate space ,
the monopole interaction is written as @xmath17 where the function @xmath18 is defined as @xmath19^{-1 } , \label{3:eq : fdef}\ ] ] for subscripts @xmath20 and @xmath21 . throughout this work ,
the three terms in ( [ eq : hamil ] ) are referred to , in the order they appear in the above expression , as the attractive , repulsive and derivative terms .
in addition , the spin - orbit force is taken to be @xmath22 which is similar to that used in the modified delta interaction @xcite .
the parameters @xmath23 , @xmath24 , @xmath25 , @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 , @xmath33 and @xmath34 are to be fitted to experimental data .
one notices that the two - body interaction consists of a sum of terms , each of which is separable in form and that the expressions for the attractive and repulsive terms in ( [ eq : hamil ] ) differ only by the values of their parameters the energy , @xmath35 , due to the interaction ( [ eq : hamil ] ) in the hartree - fock approximation is derived in appendix [ chap : hfenergy ] ( [ eq : eatt ] ) , ( [ eq : nonisoen ] ) , ( [ eq : ederv ] ) and is presented here ; @xmath36\big\}+\frac{1}{2}kn_d^2 + cn_w \label{eq : hfenergy}\end{aligned}\ ] ] where @xmath37 is the kinetic energy , @xmath38 is the direct coulomb energy plus exchange in the slater approximation .
the following quantities have been defined ; @xmath39 \nonumber \\
\delta n_\xi & = & \int { \rm d}^3\vec{r}\,\rho^{\beta_\xi}(\vec{r})\left [ \rho_p(\vec{r})-\rho_n(\vec{r})\right ] \nonumber \\ m_\xi^{(\tau\bar{\tau } ) } & = & \int\!\int { \rm d}^3\vec{r}_1{\rm d}^3 \vec{r}_2
\,\left[\rho_p(\vec{r}_1,\vec{r}_2)\rho^{\beta_\xi}(\vec{r}_1 ) \rho^{\beta_\xi}(\vec{r}_2 ) \rho_n(\vec{r}_1,\vec{r}_2 ) + \rho_n(\vec{r}_1,\vec{r}_2)\rho^{\beta_\xi}(\vec{r}_1 ) \rho^{\beta_\xi}(\vec{r}_2 ) \rho_p(\vec{r}_1,\vec{r}_2)\right ] \nonumber \\
n_d & = & \int { \rm d}^3\vec{r}\,\rho(\vec{r})\nabla^2\rho(\vec{r } ) \nonumber \\
% % m_d & = & \int\!\int { \rm d}^3\vec{r}_1{\rm d}^3\vec{r}_2 % % \rho(\vec{r}_1,\vec{r}_2 ) \nabla_1 ^
2\rho(\vec{r}_1 ) % % \nabla_2 ^ 2\rho(\vec{r}_2 ) \rho(\vec{r}_2,\vec{r}_2)\nonumber \\
n_w & = & \int { \rm d}^3\vec{r}\,\frac{1}{r}\frac{\partial \rho } { \partial r } \rho_w(\vec{r})\end{aligned}\ ] ] and the following densities are used : @xmath40 the variation of the total energy is carried out in appendix [ chap : hfenergy ] ( see [ eq : alphavarterm ] , [ eq : betavarterm ] , [ termpropg ] , [ eq : dervpot ] ) .
the resulting local hartree - fock potential is @xmath41\rho^{\beta_\xi}(x ) \nonumber \\ & - & w_\xi(\alpha_\xi/2)f_\xi^2\left[n_\xi^2+b_\xi(\delta n_\xi)^2-(1+b_\xi)m_\xi - a_\xi m_\xi^{(\tau\bar{\tau})}\right]\rho^{\alpha_\xi-1}(x)\nonumber \\ & -&w_\xi f_\xi\beta_\xi\left[(1+b_\xi)g_\xi(x)+a_\xi g_\xi^{(\tau \bar{\tau})}(x)\right ] \rho^{\beta_x-1}(x ) \nonumber \\ & + & \left[w_\xi b_\xi\beta_\xi
f_\xi\delta n_\xi\right ] \rho^{\beta_x-1}(x)\delta\rho(x)\big\ } \nonumber \\ & + & 2kn_d\nabla^2\rho(x)\label{eq : localpot}\end{aligned}\ ] ] which differs for protons ( @xmath42 ) and neutrons ( @xmath43 ) through the function @xmath44 the other newly - introduced functions in ( [ eq : localpot ] ) are @xmath45 \nonumber \\
g^{(\tau\bar{\tau})}_\xi(x ) & = & g_\xi^{(pn)}(x ) + g_\xi^{(np)}(x ) = \int \ ! { \rm d}^3 \vec{r } \left[\rho_p(\vec{r},\vec{x})\rho^{\beta_\xi}(r)\rho_n(\vec{x},\vec{r } ) + \rho_n(\vec{r},\vec{x})\rho^{\beta_\xi}(\vec{r})\rho_p(\vec{x},\vec{r})\right].\label{eq : geez}\end{aligned}\ ] ] in addition ,
the non - local component to the mean field is ( see [ eq : nonlocterm ] ) @xmath46 and there is a state - dependent potential from the spin - orbit interaction of the form @xmath47 where @xmath48 is the spin - orbit weight factor .
note that the one - body spin - orbit term could be taken as either a one - body force , or as a one - body potential deriving from a two - body force .
since the latter approach would render the perturbation calculation problematic due to the absence of a suitable of form of the two - body force , we choose the former approach .
hence , since the force is density - dependent , we have also included the rearrangement contribution to the hf potential .
only the non - rearrangement term actually gives rise to the spin - orbit splittings , but the rearrangement terms , coming as they do from a variational principle , result in a lowering of the hf energy . combining the potentials ( [ eq : localpot ] ) , ( [ eq : nonlocpot ] ) and ( [ eq : spinopot ] ) gives us the hf equation @xmath49 in this potential , as well as in the expression for the total energy ( [ eq : hfenergy ] ) , the exchange contribution from the derivative term
is omitted .
while it would , in principle , be desirable to include this term , the calculational complexity involved in doing so has forced the omission in the present case . however , for the main attractive and repulsive terms , the exchange part is much smaller than the direct in all nuclei , and the direct derivative term gives a rather small contribution to the mean - field and the binding energy in comparison to the other direct terms , so it is not considered an unwarranted approximation to neglect the effects of this term .
it might be objected that the form of the interaction is too unrealistic .
for instance , since it is separable it can not satisfy galilean invariance .
furthermore , the unusual form of the isospin operators gives rise to a different effective interaction for protons and neutrons .
as for the second point , since this is a density - dependent interaction , the effect of the difference between neutrons and protons comes from the densities as well as the isospin operators .
this being the case , it seems reasonable to allow for the `` stretching '' of the isospin operator as we have done , to allow greater freedom in fitting the interaction to data . as for the first issue
, the unusual form has been motivated by the desire to keep the force `` weak '' and is justified by the quality of the results produced . a possible method for restoring galilean invariance within our framework
is discussed in section [ sec : exper ] .
the choice of omitting a spin - spin ( @xmath50 ) type force yet having an isospin - isospin type force is motivated by the nuclei under study .
all the closed - shell nuclei are spin - saturated and would contribute only through the exchange term in the hf order . for this separable interaction ,
the space - exchange terms are rather small and a spin - spin force would add little to the results .
in addition , even if the effects in closed - shell nuclei are important , it does not seem reasonable to fit this term to closed - shell nuclei alone .
it remains an open question whether such a force will prove necessary or useful in open - shell nuclei .
it is interesting to compare the leading terms in the hf mean field to that of other models .
the first line of equation ( [ eq : localpot ] ) gives us this as @xmath51 where @xmath52 is a combination of constants .
the product @xmath53 is @xmath54 if @xmath55 then the product @xmath56 is constant and the leading mean field terms go like @xmath57 which , for the special case @xmath58 , are the same as the terms in the skyrme and gogny mean - field proportional to the parameters @xmath59 and @xmath60 , which give the bulk of the binding energy and saturation properties . in this work ,
we do not strictly keep @xmath61 , thus allowing for some a - dependence of the coefficients in the mean - field potential .
it has been found , however , that one can not go too far away from the equality and still obtain reasonable results .
that one can get similar results in a mean - field calculation from two very different interactions is reflected in the different constitution of the resulting perturbative part of the hamiltonian , @xmath6 .
in order to find the best set of parameters for the interaction ( [ eq : hamil ] ) , calculations have been made of 14 doubly closed - shell nuclei across the periodic table .
they are @xmath0o , @xmath62si , @xmath63ca , @xmath64ni , @xmath65zr , @xmath66sn , @xmath67gd and @xmath1pb .
the nuclei represent a selection of doubly - closed ( sub-)shell nuclei both close to and far from stability .
there is limited experimental information about @xmath68ni @xcite and @xmath69sn @xcite .
@xmath70ni has yet to be discovered .
the ability to reproduce the properties of such exotic nuclei will be important for applications of our technique and discrepancies will help direct refinements .
a hartree - fock code assuming spherical symmetry and representing wavefunctions in a basis of spherical harmonic oscillator states was used to calculate uncorrelated wavefunctions .
perturbation corrections to the binding energy were directly evaluated using the results of the hf calculation .
the results presented here were obtained in a basis of 12 expansion coefficients per single - particle wavefunction and iterated until the hf energy had converged to within 1kev .
the parameters of the force were fitted to binding energies to second order and charge radii , charge density distributions , single - particle energies and spin - orbit splittings to hf order of the nuclei listed above where experimental data were available , and are presented in table [ tab : params ] .
the results of the the calculated energies , in hf order and in each order of perturbation theory are presented in table [ tab : bind ] .
the differences between the hf energy @xmath71 and the experimental ground state energy @xmath72 and between hf plus second order perturbation correction ( [ eq:2 ] ) @xmath73 and experiment are shown in figure [ fig : percbind ] .
the experimental energies are taken from the mass table of audi and wapstra @xcite with two exceptions .
an estimate of the mass of the recently - discovered nuclei @xmath68ni @xcite and the measured mass of @xmath69sn @xcite .
the energy for @xmath70ni was taken from @xcite in which extrapolated values are given , which are thought to be in error by less than @xmath74 .
one sees from figure [ fig : percbind ] that most of the nuclei fit the binding energy to within @xmath752% .
the most obvious exception is @xmath0o which is quite under - bound .
this may be due to the omission of a center - of - mass correction which would undoubtedly go a long way to close the discrepancy in the energy@xcite .
it was not calculated in this case since a rigorous microscopic correction would destroy the mean - field which provides the essential basis for the perturbation calculation .
a phenomenological correction could have been calculated , but perhaps the most suitable method in our framework would be to include with our multipole forces an isoscalar dipole force which could be fitted to restore the translational invariance of the many - body hamiltonian , and evaluated exactly in perturbation theory . a general trend can be seen in which lighter nuclei are somewhat over - bound and the heaviest are under - bound .
it is the exceptions which conspire to stop the fitting algorithm from doing better , but the somewhat systematic nature of this discrepancy suggests that a better mass or isospin dependence may improve matters .
it is unclear as yet the extent to which multipole correlations or a spin - spin force would improve the fit to spherical nuclei .
that question awaits the study of deformed nuclei and excited states . in table
[ tab : skcomp ] a comparison is made of the quality of the fit to the binding energy to properties of the same nuclei calculated with a selection of skyrme parameterizations .
the parameterizations used are siii@xcite , skp@xcite , sly4@xcite and ski4@xcite . in this comparison
, it is seen that the energies from the different skyrme parameterizations are of a similar quality , all reproducing the binding energies of closed - shell nuclei very well , with only a few binding energies being reproduced no better than 1% including @xmath68ni whose experimental value is in any case not well known .
it is clear that the results from the separable force are somewhat worse . particularly problematic
is @xmath0o , whose large under - binding was mentioned above , and also @xmath68ni which is , as with the skyrme parameterizations , over - bound , although more so with the separable interaction . results for one - body properties are also presented .
comparison of the charge density results to experiment @xcite and to the selection of skyrme interactions is made in table [ tab : radii ] .
one - body observables are generally reproduced better in the hf calculation alone than the binding energies .
the comparison of the calculated charge radii with experiment is generally more favourable than the energy data .
the radius of oxygen is too large by about 5% which is consistent with its under - binding .
it can be seen that the agreement with experiment is of the same level as the skyrme interactions .
perturbative corrections to the one - body observables , such as the densities , and hence radii , will be calculated in future work .
figures [ fig : o16ff]-[fig : pb208ff ] show the electron scattering form - factors of a selection of the nuclei compared to experiment @xcite .
the proton density was corrected for the finite proton size by folding with a gaussian to give the charge density , from which the radii and form - factors were calculated .
the form - factors agree with experiment quite well , which is expected given the generally correct radii .
some typical single - particle energies for light and heavy nuclei are shown in figures [ fig : ca40spe ] and [ fig : pb208spe ] for @xmath76ca and @xmath1pb respectively .
the single - particle energies of a density - dependent hartree - fock calculation do not directly correspond to an experimental observable , so caution should be used in comparing values .
it can be seen that the level spacings and shell closures are better reproduced in @xmath1pb , which is true of heavy nuclei in general .
the comparatively poorer results in light nuclei seems to be common to mean - field approaches @xcite . in the case of @xmath76ca
, the gap at the fermi level may be widened by the inclusion of multipole forces which will link the occupied levels with the @xmath77 states , which is otherwise `` inert '' .
table [ tab : split ] shows spin - orbit splittings for some cases where the experimental values are known .
the `` experimental '' data presented represents that used in previous work for fitting effective interactions to data @xcite .
clearly the splittings are all systematically small .
this could be remedied by an increase in the spin - orbit coefficient , @xmath34 . in a previous work @xcite , a value 10% higher than ours was used for the same spin - orbit interaction , and
hence the spin - orbit splittings were more realistic .
the lower value used in our work is the result of a compromise between the reproduction of the spin - orbit splittings and the total binding energies .
this is a further indication that a more suitable spin - orbit potential needs to be sought .
the perturbation correction to the energy are seen to be rather small in all nuclei considered .
this is consistent with our goal that the mean field solution should be close to the exact solution of the mbse .
the size of the second order correlation is roughly constant across the periodic table .
it is characterized by a dimensionless strength parameter , @xmath78 , defined as @xmath79 it is related to the `` wound integral''@xcite and is proportional to the number of 2p2h states excited due to the second order perturbation .
figure [ fig : calcorr ] shows the correlation structure from the second order correction in the @xmath80 nucleus @xmath76ca , and the nuclei @xmath68ca and @xmath1pb .
the contribution to the second - order energy is defined as a function of one of the particle states @xmath81 ; @xmath82 the plot shows the contribution to the total second order energy correction as a function of the single particle energy @xmath83 , in 5 mev wide bins . in all three cases particles
are dominantly excited to low - lying states above the fermi level .
this results in a ground state with occupation probabilities similar to those which result from pairing forces .
it is also a further indication that perturbation theory makes sense for our interaction since it does not predict excitation of particles in the ground state to extremely high energies .
since the calculations were made only using a monopole force , the correlation structure is not expected to be complete .
only corrections involving simultaneous @xmath15 scattering of two particles is included .
an indication of this is seen in the difference between the results for @xmath80 and @xmath84 nuclei .
the second order correction in @xmath68ca is much larger than that in @xmath76ca , due to the possibility of an @xmath77 neutron exciting to the @xmath77 proton state while another proton excites to a neutron state .
this extra excitation is the labelled peak in fig .
[ fig : calcorr ] .
as well as having large wavefunction overlaps , the energy denominator in this case is much smaller than in any other possible monopole excitation which must excite any particle across major shells to keep all angular and isospin quantum numbers the same .
when general excitations are permitted by higher multipole forces @xmath85 , this difference between correlations in @xmath80 and @xmath84 nuclei will be smoothed out . for this reason , too , the correlation energies should not be considered too quantitatively at this stage , but rather as an indication of the perturbative properties of the interaction . comparing the form of the interaction to that of skyrme suggests other possible sources of improvement to the model .
one such may come from a better parameterization of the spin - orbit interaction .
a two - body form which fits the philosophy of the separable effective interaction has not been found , but may be necessary to give the correct contribution to the binding energy . in any case , the simple form used in the present work which depends on a radial derivative will need modification if it is to be applied to deformed nuclei .
it may also prove fruitful to explore a more general term dependent upon the derivatives of the density than the single term with parameter @xmath33 , such as is found in the skyrme interaction with two terms proportional to @xmath86 and @xmath87 , which often carry further exchange parameters @xmath88 and @xmath89 .
we have presented a new effective nuclear interaction which is designed for use in calculations which go beyond the mean - field . the technique of using perturbation theory to build correlations on top of the hartree - fock result is applicable to our interaction and results in small corrections to the single - particle behavior .
a monopole - monopole force alone gives reasonable results for the ground state properties of spherical doubly - magic nuclei .
it is expected that the addition of multipole forces will improve these results , particularly through the completion of the correlation structure .
such multipole forces will also presumably be important in giving the correct shapes of deformed nuclei , which are the subject of a forthcoming study and in the correct description of excited states .
the authors would like to acknowledge useful discussions with d. vautherin , p .-
reinhard , d. m. brink and d. j. dean .
this research was sponsored by the division of nuclear physics , u.s .
dept . of energy under contract
de - ac05 - 00or 22725 managed by ut
battelle , llc , and by research the uk epsrc , and by us doe grant no .
de - fg02 - 94er40834 .
the interaction is given in equation ( [ eq : hamil ] ) .
its expectation value is the contribution it makes to the total energy and is @xmath90 if we consider just the _ attractive _ term - that is , the term whose parameters have the subscript `` a '' - then we will obtain the contribution from the repulsive term by simply substituting the subscript `` a '' for `` r '' .
@xmath91 taking the first line , the matrix element is represented in space ( and spin and isospin ) coordinates ; @xmath92 where quantities defined in section [ sec : interaction ] are used .
note that integrals include sums over spinors and isospinors where appropriate , and the coordinates include spin and isospin coordinates where appropriate .
where densities are used , the summing over isospin states has already been done and where densities do not carry isospin labels , the isoscalar density is assumed .
see equations ( [ eq : den ] ) and ( [ eq : denmat ] ) for definitions . the second line of ( [ eq : eatt ] )
contains `` isospin - flipping '' operators whose action is to turn an isospin state where particle one is a proton and particle two a neutron , @xmath93 , into @xmath94 and vice - versa .
the direct contribution , in which the labels in the bra and the ket are in the same order is zero since all proton states are orthogonal to all neutron states .
the exchange term is similar to that in ( [ eq : nonisoen ] ) but with different isospin combination of the density matrices ; @xmath95 the third line of [ eq : eatt ] contains isospin - projection operators which have a value @xmath96 when @xmath97 and @xmath98 are like particles and @xmath99 when they are unlike . in the direct term
this gives an energy of @xmath100 ^ 2+\frac{1}{2}w_af_ab_a\left[\int{\rm d}^3\vec{r}\,\rho_n(\vec{r } ) \rho^{\beta_a}(\vec{r})\right]^2 \nonumber \\ & & -w_af_ab_a\int{\rm d}^3\vec{r}_1\,\rho_n(\vec{r}_1 ) \rho^{\beta_a}(\vec{r}_1 ) \int{\rm d}^3\vec{r}_2\,\rho_p(\vec{r}_2 ) \rho^{\beta_a}(\vec{r}_2)\nonumber \\ & = & \frac{1}{2}w_af_ab_a\left[\int{\rm d}^3\vec{r } ( \rho_p(\vec{r } ) - \rho_n(\vec{r}))\rho^{\beta_a}(\vec{r } ) \right]^2 = \frac{1}{2 } w_a f_a b_a ( \delta n_a)^2.\end{aligned}\ ] ] the exchange term gives a contribution only when @xmath97 and @xmath98 have the same isospin quantum number , which is just like the case for having no isospin operator there , so the contribution is like that in ( [ eq : nonisoen ] ) @xmath101 the hf mean - field is obtained by varying the total energy with respect to the single particle states .
this gives , for the case of the attractive term , without the explicit isospin dependence : @xmath102 the variation of the function @xmath103 is given by @xmath104^{-1 } \nonumber \\ & = & -f_a^2\int\,{\rm d}^3\vec{r}\,\frac{\rho^{\delta \alpha_a}(\vec{r } ) } { \delta\phi^*_b(\vec{x } ) } \nonumber \\ & = & -f_a^2\alpha_a\int\,{\rm d}^3\vec{r}\,\rho^{\alpha_a-1}(\vec{r } ) \frac{\delta \rho(\vec{r})}{\delta\phi_b^*(\vec{x } ) } \nonumber \\ & = & -f_a^2\alpha_a\rho^{\alpha_a-1}(\vec{x})\phi_b(\vec{x})\end{aligned}\ ] ] so that the contributions of the two terms in ( [ eq : varnoniso ] ) involving the variation of @xmath103 give a contribution to the hf mean field of @xmath105 the function @xmath106 is similar in form to @xmath103 and the functional variation proceeds in a similar manner ; @xmath107 and the contribution from the second term in ( [ eq : varnoniso ] ) to the mean - field is @xmath108 finally , the functional variation of the exchange matrix element @xmath109 is @xmath110 where use is made of the symmetry of the integral to combine the four terms into two . the last term in ( [ eq : varm ] )
gives rise to a local term in the mean - field of @xmath111 where @xmath112 has been defined as in ( [ eq : geez ] ) .
this completes the non - isospin - dependent part of the attractive force , and so also the repulsive by change of subscript .
the isospin - dependent terms are obtained in an analogous way , except that when the variation applies to the density of a single nucleon species , so the contribution to the mean - field applies only to that species . for the final term in equation ( [ eq : hamil ] ) , the so - called _ derivative _ term ,
only the direct part of the energy is at present considered .
it is @xmath114 the functional variation proceeds as @xmath115 the first term gives a contribution to the mean field of @xmath116 by integrating the second term by parts twice , one in fact gets exactly the same contribution to the mean - field again , so that the total contribution to the mean - field from the direct term of the derivative interaction is @xmath117 the exchange part of this term is not calculated .
m. leuschner , j. r. calarco , f. w. hersman , e. jans , g. j. kramer , l. lapiks , g. van der steenhoven , p. k. a. de witt huberts , h. p. blok , n. kalantar - nayestanaki , j. friedrich , _ phys .
_ * c 49 * , 955 ( 1994 ) . | we present calculations of ground state properties of spherical , doubly closed - shell nuclei from @xmath0o to @xmath1pb employing the techniques of many - body perturbation theory using a separable density dependent monopole interaction .
the model gives results in hartree - fock order which are of similar quality to other effective density - dependent interactions .
in addition , second and third order perturbation corrections to the binding energy are calculated and are found to contribute small , but non - negligible corrections beyond the mean - field result .
the perturbation series converges quickly , suggesting that this method may be used to calculate fully correlated wavefunctions with only second or third order perturbation theory .
we discuss the quality of the results and suggest possible methods of improvement . |
a weak perturbation of a hamiltonian system causes the onset of chaotic layers around separatrices of the unperturbed system and/or separatrices surrounding nonlinear resonances generated by the perturbation @xcite .
the system may be transported along the layer in a random - like fashion and this chaotic transport plays an important role in many physical phenomena @xcite . if the perturbation is sufficiently weak , then the layers are thin and the chaos is called _ local _ @xcite
. as the perturbation magnitude increases , the width of the layer grows and the layers corresponding to adjacent separatrices reconnect at some , typically non - small , critical value of the perturbation .
this conventionally marks the onset of _ global _ chaos @xcite i.e. chaos in a large region of the phase space , with chaotic transport throughout the whole relevant energy range .
the reconnection of the layers around separatrices of the resonances often correlates with the overlap in energy between neighbouring resonances calculated independently in the resonant approximation .
the latter constitutes the heuristic chirikov resonance - overlap criterion @xcite .
but the chirikov criterion may fail if the system is of the _ zero - dispersion _ ( zd )
type @xcite i.e. if the frequency of eigenoscillations possesses a local maximum or minimum as a function of its energy ( cf . also studies of related maps @xcite which are called _ nontwist _ , _ twistless _ or _ nonmonotonic twist _ maps ) . in such systems , there are typically two resonances of one and the same order @xcite , and their overlap in energy does not result in the onset of global chaos @xcite . even their overlap in phase space @xcite results typically only in the reconnection of the thin chaotic layers associated with the resonances .
as the amplitude of the time - periodic perturbation grows further , the layers may _ separate _
again @xcite .
an example of the evolution of resonances in the plane of energy and slow angle is given in fig .
1 ( the typical evolution of a real poincar section is shown e.g. in @xcite ) .
as it is known @xcite , any hamiltonian system with two or more separatrices belongs to the zd type : the eigenfrequency as a function of energy possesses a local maximum between each pair of adjacent separatrices . for the purpose of global chaos onset
, our letter @xcite has addressed the possibility to combine the overlap of resonances with each other ( typical of zd systems ) and their overlap with the chaotic layers associated with the separatrices . via numerical simulations , @xcite demonstrated that this is possible , leading to a scenario for global chaos onset which requires much smaller perturbation amplitudes than in the conventional case .
the letter @xcite suggested also a heuristic theory for this effect ( more details were presented in @xcite ) .
the _ present _ work develops the method for the quantitative description of chaotic layers in _ phase space _
, for the resonance frequency range .
we uncover the physical mechanism of their overlap with the resonances , and on this basis develop a detailed _ self - contained _ theory of the facilitated onset of global chaos .
we also discuss generalizations and applications .
the method for the description of the chaotic layers is of _ general importance _ : in conventional systems with a single - separatrix layer , it predicts a much larger maximum width in energy than what was assumed before @xcite .
the paper is organized as follows . sec .
ii introduces some relevant model example and presents the major results of the simulations : studying numerically the frequency dependence of the minimal amplitude of the ac drive for which global chaos occurs , @xmath0 , we show that @xmath0 possesses deep spikes at certain frequencies . sec .
iii gives the self - consistent asymptotic theory for the minima of the spikes , after assessing the boundaries of the relevant chaotic layers .
iv gives the theory for the spikes wings .
discussion of a few generalizations and applications is carried out in sec .
v. conclusions are drawn in sec .
the appendix describes in details the new method for the analysis of separatrix chaos .
as an example of a one - dimensional hamiltonian system possessing two or more separatrices , we use a spatially periodic potential system with two different - height barriers per period ( fig .
2(a ) ) : @xmath1 this model may relate e.g. to a pendulum spinning about its vertical axis @xcite or to a classical 2d electron gas in a magnetic field spatially periodic in one of the in - plane dimensions @xcite .
the interest to the latter system arose in the 90th due to technological advances allowing to manufacture magnetic superlattices of high - quality @xcite leading to a variety of interesting behaviours of the charge carriers in semiconductors @xcite .
2(b ) and 2(c ) show respectively the separatrices of the hamiltonian system ( 1 ) in the @xmath2 plane and the dependence of the frequency @xmath3 of its oscillation , often called _ eigenfrequency _ , on its energy @xmath4 .
the separatrices correspond to energies equal to the value of the potential barrier tops @xmath5 and @xmath6 ( fig .
the function @xmath7 is close to the extreme eigenfrequency @xmath8 for most of the range @xmath9 $ ] while sharply decreasing to zero as @xmath10 approaches either @xmath11 or @xmath12 .
0.2 cm 0.2 cm 0.2 cm add now a time - periodic perturbation : as an example , we use an ac drive , which corresponds to a dipole @xcite perturbation of the hamiltonian : @xmath13 the _ conventional _ scenario of global chaos onset between the separatrices of the system ( 2)-(1 ) is illustrated by fig . 3 .
the figure presents the evolution of the stroboscopic poincar section as @xmath14 grows while @xmath15 is fixed at an arbitrarily chosen value _ away _ from @xmath16 and its harmonics . at small @xmath14 , there are two thin chaotic layers around the inner and outer separatrices of the unperturbed system .
unbounded chaotic transport takes place only in the outer chaotic layer i.e. in a _ narrow _ energy range .
as @xmath14 grows , so do also the layers . at some critical value @xmath17 , the layers merge .
this may be considered as the onset of global chaos : the whole range of energies between the barrier levels is involved , with unbounded chaotic transport .
the states @xmath18 and @xmath19 ( where @xmath20 is any integer ) in the poincar section are associated respectively with the inner and outer saddles of the unperturbed system , and necessarily belong to the inner and outer chaotic layers , respectively .
thus , the necessary and sufficient condition for global chaos onset may be formulated as the possibility for the system placed initially in the state @xmath21 to pass beyond the neighbouring of the outer
states , @xmath22 or @xmath23 , i.e. the coordinate @xmath24 becomes @xmath25 or @xmath26 at sufficiently large times @xmath27 . a diagram in the @xmath28 plane , based on the above criterion , is shown in fig .
the lower boundary of the shaded area represents the function @xmath29 .
it has deep _ spikes _ i.e. cusp - like local minima .
the most pronounced spikes are situated at frequencies @xmath30 that are slightly less than the odd multiples of @xmath16 , @xmath31 the deepest minimum occurs at @xmath32 : the value of @xmath33 in the minimum , @xmath34 , is approximately 40 times smaller than the value in the neighbouring pronounced local maximum of @xmath29 at @xmath35 . as @xmath36 increases
, the @xmath36th minimum becomes less deep .
the function @xmath29 is very sensitive to @xmath15 in the vicinity of the minima : for example , a shift of @xmath15 down from @xmath37 by only 1% causes an increase of @xmath33 by @xmath38 .
the origin of the spikes becomes more clear looking at the evolution of the poincar section for @xmath39 as @xmath14 grows ( fig .
5 ) : it drastically differs from the conventional evolution shown in fig .
3 . for @xmath40 ( fig . 5(a ) ) , one can see four chaotic trajectories .
two of them are associated with the inner and outer separatrices of the unperturbed system , similarly to the conventional case ( cf .
they are marked by green and blue respectively .
these trajectories fill the corresponding chaotic layers , which will be referred below as the inner `` and outer '' separatrix layers respectively .
the other two chaotic trajectories marked by red and cyan are associated with the two nonlinear resonances of the 1st order .
examples of non - chaotic trajectories separating the chaotic ones are shown in brown .
as the perturbation amplitude @xmath14 increases , the outer separatrix layer sequentially absorbs other chaotic trajectories while large stability islands ( associated with the resonances ) arise in the layer . at @xmath41 ,
it has absorbed the red trajectory : the resulting chaotic layer is shown in blue in fig .
5(b ) . at @xmath42 ,
this chaotic layer has absorbed the cyan chaotic trajectory : the resulting chaotic layer is shown in blue in fig .
5(c ) @xcite .
finally , at @xmath43 the latter blue layer has merged with the inner separatrix layer @xcite ( see fig .
5(d ) ) , i.e. the onset of global chaos as defined above has occurred . even prior to the theoretical analysis
, one can draw a few conclusions from the evolution .
namely , if @xmath15 is close to the minimum of the spike of @xmath29 , then * the onset of global chaos occurs due to the _ combination _ of the overlap of chaotic layers associated with nonlinear resonances with each other and the overlap of the latter layers with the inner and outer separatrix layers ; * the width of the nonlinear resonances are large already at quite _ small _ amplitudes of the perturbation , so that the overlap with the chaotic layers around the original separatrices occurs at unusually small perturbation amplitudes ; * the onset of the overlap of at least one of the nonlinear resonances with the outer separatrix layer occurs at values of @xmath14 which are a few times smaller than those required for the onset of the overlap with the inner separatrix layer .
the above conclusions are also illustrated by fig .
6 which presents the evolution of the phase space of slow variables @xcite , action @xmath44 and slow angle @xmath45 , calculated in resonance approximation for the 1st - order spike ( see eq . ( 4 ) below ) . similarly , for the spikes of higher order , higher - order resonances are relevant
the eigenfrequency @xmath7 is close to its local maximum @xmath16 for most of the relevant range @xmath9 $ ] ( fig .
as shown below , @xmath7 approaches a _ rectangular _ form in the asymptotic limit @xmath46 .
hence , if the perturbation frequency @xmath47 is close to @xmath16 or its odd multiples , @xmath48 , then the energy width of nonlinear resonances becomes comparable to the width of the whole range between barriers ( i.e. @xmath49 ) at a rather small perturbation magnitude @xmath50 .
note that @xmath51 determines the characteristic magnitude of the perturbation required for the conventional overlap of the separatrix chaotic layers , when @xmath15 is not close to any odd multiple of @xmath52 ( fig .
3 ( c ) ) . thus ,
if @xmath53 , the nonlinear resonances should play a crucial role in the onset of global chaos ( cf .
fig . 5 ) .
we note that it is not entirely obvious _ a priori _ whether it is indeed possible to calculate @xmath54 within the resonance approximation : in fact , it is essential for the separatrices of nonlinear resonances to nearly touch the barriers levels , but the resonance approximation is obviously invalid in the close vicinity of the barriers ; furthermore , numerical calculations of resonances show that , if @xmath55 , the perturbation amplitude @xmath14 at which the resonance separatrix touches a given energy level in the close vicinity of the barriers is very sensitive to @xmath15 , apparently making the calculation of @xmath56 within the resonance approximation even less feasible .
nevertheless , we show below in a self - consistent manner that , in the asymptotic limit @xmath57 , the relevant boundaries of the chaotic layers lie in the range of energies @xmath10 where @xmath58 .
therefore , the resonant approximation is valid and it allows to obtain _ explicit _ asymptotic expressions both for @xmath59 and @xmath56 , and for the wings of the spikes in the vicinities of @xmath60 .
the _ asymptotic _
limit @xmath57 is the most interesting one from a theoretical point of view since this limit leads to the strongest facilitation of the global chaos onset and it is most accurately described by the self - contained theory .
most of the theory presented below assumes this limit and concentrates therefore on the results to the _ lowest _ order in the small parameter . on the applications side ,
the range of _ moderately small _
@xmath61 is more interesting , since the chaos facilitation is still pronounced ( and still described by the asymptotic theory ) while the area of chaos between the separatrices is not too small ( comparable with the area inside the inner separatrix ) : cf .
figs . 2 , 3 and 5 . to increase the accuracy of the theoretical description in this range
, we estimate the next - order corrections and develop an efficient numerical procedure allowing for the further corrections .
let @xmath15 be close to the @xmath36th odd @xcite harmonic of @xmath16 , @xmath62 .
over most of the range @xmath9 $ ] , except in the close vicinities of @xmath11 and @xmath12 , the @xmath36th harmonic of eigenoscillation is nearly resonant with the perturbation . due to this ,
the ( slow ) dynamics of the action @xmath63 and the angle @xmath64 @xcite can be shown to be described by the following auxiliary hamiltonian ( cf .
@xcite ) : @xmath65 where @xmath66 is the minimal energy ( over all @xmath67 ) @xmath68 ; @xmath69 and @xmath70 are , respectively , the frequency and the minimal coordinate of the conservative motion with a given value of energy @xmath10 ; @xmath20 is the number of right turning points in the trajectory @xmath71 $ ] of the conservative motion with energy @xmath10 and given initial state @xmath72 .
let us derive the explicit expressions for various quantities in ( 4 ) .
in the unperturbed case ( @xmath73 ) , the equations of motion ( 2 ) with @xmath74 ( 1 ) can be integrated @xcite ( see also eq . ( 60 ) below ) , so that we can find @xmath7 : @xmath75 } , \\ & & k=\frac{1}{2}\sqrt{\frac{(\sqrt{2e}+1)^2-\phi^2}{\sqrt{2e } } } \ , , \nonumber\end{aligned}\ ] ] where @xmath76=\int_0^{\frac{\pi}{2}}\frac{{\rm d}\phi}{\sqrt{1-k^2\sin^2(\phi)}},\ ] ] is the full elliptic integral of the first order @xcite . using its asymptotic expression , @xmath77 \simeq \frac{1}{2 } \ln\left ( { 16 \over 1-k^2 } \right),\ ] ] we derive @xmath7 in the asymptotic limit @xmath46 : @xmath78 the function @xmath7 ( 7 ) is close to its maximum @xmath79}\{\omega(e)\ } \simeq \frac{\pi}{2\ln ( 8/\phi ) } \,\ ] ] for most of the interbarrier @xcite range of energies @xmath80 $ ] ; on the other hand , in the close vicinity of the barriers , where either @xmath81 or @xmath82 become comparable with , or larger than , @xmath83 , @xmath7 sharply decreases to zero as @xmath84 . the range where this takes place
is @xmath85 , and its ratio to the whole interbarrier range , @xmath86 , is @xmath87 i.e. it goes to zero in the asymptotic limit @xmath57 : in other words , @xmath7 approaches a _ rectangular _ form . as it will be clear from the following , _ * it is this almost rectangular form of @xmath7 which determines many of the characteristic features of the global chaos onset in systems with two or more separatrices*_. one more quantity which strongly affects @xmath88 is the fourier harmonic @xmath89 .
the system stays most of the time very close to one of the barriers .
consider the motion within one of the periods of the potential @xmath90 , between neighboring upper barriers @xmath91 $ ] where @xmath92 .
if the energy @xmath93 lies in the relevant range @xmath9 $ ] , then the system will stay close to the lower barrier @xmath94 for a time @xcite @xmath95 during each period of eigenoscillation , while it will stay close to one of the upper barriers @xmath96 for most of the remaining of the eigenoscillation , @xmath97 hence , the function @xmath98 may be approximated by the following piecewise even periodic function : @xmath99\frac{\pi}{2}\frac{t_u}{t_l+t_u},\pi-\frac{\pi}{2}\frac{t_u}{t_l+t_u } \left[\right.,}^{\pi\quad { \rm at } \quad \psi\in \left[0,\frac{\pi}{2}\frac{t_u}{t_l+t_u } \right]\cup \left[\pi-\frac{\pi}{2}\frac{t_u}{t_l+t_u},\pi \right ] , } \right .
\nonumber \\ & & q(e,-\psi)-q_{lb}=q(e,\psi)-q_{lb } , \nonumber \\ & & q(e,\psi\pm 2\pi i)=q(e,\psi ) , \nonumber\\ & & i=1,2,3 , ... \nonumber\end{aligned}\ ] ] substituting the above approximation for @xmath100 into the definition of @xmath101 ( 4 ) , one can obtain : @xmath102 at barrier energies , @xmath103 takes the values @xmath104 as @xmath10 varies in between the barrier values , @xmath103 varies monotonously if @xmath105 and non - monotonously otherwise ( cf . fig .
but in any case , the significant variations occur mostly in the close vicinity of the barrier energies @xmath11 and @xmath12 while , for most of the range @xmath9 $ ] , the argument of the sine in eq . ( 11 ) is close to @xmath106 and @xmath107 is then almost constant : @xmath108}\frac{\sqrt{2}}{2j-1 } , \quad \
j= 1 , 2 , 3 , \ , \ldots , \\ & & \left| \ln \left ( \frac{1+\delta e/\phi}{1-\delta e/\phi } \right ) \right| \ll 2\ln \left ( \frac{1}{\phi }
\right ) , \nonumber\end{aligned}\ ] ] where @xmath109 $ ] means the integer part . in the asymptotic
limit @xmath57 , the range of @xmath110 where the approximate equality ( 12 ) for @xmath103 is valid approaches the whole range @xmath111-\phi,\phi[$ ] .
we emphasize that @xmath112 determines the strength of the nonlinear resonances : therefore , apart from the nearly rectangular form of @xmath7 , * the non - smallness of @xmath112 is one more factor giving rise to the strong facilitation of the global chaos onset*. we shall need also the asymptotic expression of the action @xmath113 . substituting @xmath7 ( 7 ) into the definition of @xmath114 ( 4 ) and carrying out the integration , we obtain @xmath115 we now turn to the analysis of the _ phase space _ of the resonance hamiltonian ( 4 ) . the evolution of the poincar section ( see fig . 5 and the related analysis in sec .
ii ) suggests that we need to find such _ separatrix _ of ( 4 ) which undergoes the following evolution as @xmath14 grows : for sufficiently small @xmath14 , the separatrix does not overlap chaotic layers associated with the barriers while , for @xmath116 , it does overlap them .
the relevance of such a condition will be justified further .
for @xmath117 with a given odd @xmath36 , the equations of motion of the system ( 4 ) read as follows @xmath118 any separatrix necessarily includes one or more unstable stationary points .
the system ( 14 ) may have several stationary points per @xmath119 interval of @xmath120 .
let us first exclude those points which are irrelevant to the separatrix undergoing the evolution described above .
given that @xmath121 , there are two unstable stationary points with @xmath113 corresponding to @xmath122 and @xmath123 .
they are irrelevant since , even for an infinitely small @xmath14 , each of them necessarily lies inside the corresponding barrier chaotic layer . if @xmath124 , then @xmath125 , so @xmath126 only if @xmath120 is equal either to 0 or to @xmath127 .
substituting these values into the second equation of ( 14 ) and putting @xmath128 , we obtain the equations for the corresponding actions : @xmath129 where the signs - `` and + '' correspond to @xmath130 and @xmath131 respectively .
a typical example of the graphic solution of equations ( 15 ) for @xmath132 is shown in fig .
two of the roots corresponding to @xmath131 are very close to the barrier values of @xmath113 ( we remind that the relevant values of @xmath14 are small ) .
these roots arise due to the divergence of @xmath133 as @xmath113 approaches any of the barrier values .
the lower / upper root corresponds to a stable / unstable point .
however , for any @xmath36 , both these points and the separatrix generated by the unstable point necessarily lie in the ranges covered by the barrier chaotic layers .
therefore , they are also irrelevant @xcite . for @xmath134 , the number of the roots of ( 15 ) in the vicinity of the barriers may be larger ( due to the oscillations of the modulus and sign of @xmath135 in the vicinity of the barriers ) but they all are irrelevant for the same reason , at least to leading - order terms in the expressions for the spikes minima .
consider the stationary points corresponding to the remaining four roots of equations ( 15 ) . as follows from the analysis of equations ( 14 ) linearized near the stationary points ( cf .
@xcite ) , two of them are stable elliptic points @xcite , often called _ nonlinear resonances _ , while two others are unstable hyperbolic points , often called _
saddles_. these saddles are of main interest in the context of our work .
they belong to the _ separatrices _ separating the regions of oscillations around the resonances from the regions of motion with a running slow angle @xmath120 .
we shall distinguish the relevant saddles as the saddles with the _ lower _ action / energy ( using the subscript @xmath136 ) and _ upper _ action / energy ( using the subscript @xmath137 ) .
the positions of the saddles in the @xmath138 plane are defined by the following equations ( cf . figs . 6 and 7 ) : @xmath139 } , \\ & & \tilde{\psi}_{sl}=\pi(1+g)/2 , \quad\quad \tilde{\psi}_{su}=\pi(1-g)/2,\nonumber \\ & & x_g(i_{sl})=x_{-g}(i_{su})=0 , \nonumber \\ & & \frac { { \rm d}x_g(i_{sl})}{{\rm d}i_{sl}}>0 , \quad\quad \frac { { \rm d}x_{-g}(i_{su})}{{\rm d}i_{su}}<0 , \nonumber\end{aligned}\ ] ] where @xmath140 are defined in eq .
( 15 ) while @xmath141 and @xmath142 are closer to @xmath143 than any other solution of ( 16 ) ( if any ) from below and from above , respectively .
given that the values of @xmath14 relevant to the minima of the spikes are small in the asymptotic limit @xmath57 , one may neglect the last term in the definition of @xmath144 in eq .
( 15 ) in the lowest - order approximation , so that the equations @xmath145 reduce to the simple resonance condition @xmath146 substituting here eq .
( 7 ) for @xmath3 , we obtain the explicit expressions for the energies in the saddles : @xmath147 the corresponding actions @xmath148 are expressed via @xmath149 by means of eq .
( 13 ) . for @xmath150 ,
the values of @xmath149 ( 18 ) lie in the range where the expression ( 12 ) for @xmath101 does hold true .
this will be explicitly confirmed by the results of the calculations based on this assumption .
using ( 16 ) for the angles and ( 18 ) for the energies , and the asymptotic expressions ( 7 ) , ( 12 ) and ( 13 ) for @xmath7 , @xmath151 and @xmath114 respectively , and allowing for the resonance condition ( 17 ) , we obtain explicit expressions for the values of the hamiltonian ( 4 ) in the saddles : @xmath152 + h\sqrt{2}.\ ] ] as the analysis of simulations suggests ( see the item 1 in the end of sec .
ii ) and as it is rigorously shown in the next subsection , one of the main conditions which should be satisfied in the spikes is the overlap in phase space between the separatrices of the nonlinear resonances , called _ separatrix reconnection _ @xcite . given that the hamiltonian @xmath153 is constant along any trajectory of the system ( 4 ) , the values of @xmath153 in the lower and upper saddles of the _ reconnected _ separatrices are equal to each other : @xmath154 that may be considered as the necessary and sufficient @xcite condition for the reconnection .
taking into account that @xmath155 ( see ( 19 ) ) , it follows from ( 20 ) that @xmath156 explicitly , the relations in ( 21 ) reduce to @xmath157 , \nonumber \\ & & \delta e^{(1)}\equiv \sqrt{\phi^2 - 64\exp(-\frac{n\pi}{\omega_f } ) } , \\ & & 0<\omega_m-\omega_f / n\ll \omega_m\equiv \frac{\pi}{2\ln(8/\phi ) } , \nonumber \\ & & n=1,3,5 , ... \nonumber\end{aligned}\ ] ] the function @xmath158 ( 22 ) monotonously decreases to zero as @xmath15 grows from @xmath159 to @xmath160 , where the line abruptly stops .
10 shows the portions of the lines ( 22 ) relevant to the left wings of the 1st and 2nd spikes ( for @xmath161 ) . the next step is to find a minimal value of @xmath14 for which the resonance separatrix overlaps the chaotic layer related to a potential barrier . with this aim ,
we study how the relevant outer boundary of the chaotic layer behaves as @xmath14 and @xmath15 vary .
assume that the relevant @xmath15 is close to @xmath160 while the relevant @xmath14 is sufficiently large for @xmath7 to be close to @xmath16 at all points of the outer boundary of the layer ( the results will confirm these assumptions ) .
then the motion along the regular trajectory infinitesimally close to the layer boundary may be described within the resonance approximation ( 4 ) .
hence the boundary may also be described as a trajectory of the resonant hamiltonian ( 4 ) .
this is explicitly proved in the appendix , using the separatrix map analysis that allows for the validity of the relation @xmath58 for all @xmath10 relevant to the boundary of the chaotic layer .
the main results are presented below . for the sake of clarity
, we present them for each layer separately , although they are similar in practice .
let @xmath15 be close to any of the spikes minima .
one of the key roles in the formation of the upper boundary of the layer is played by the angle - dependent quantity @xmath162 which we call the _ generalized separatrix split _ ( gss ) for the lower layer , alluding to the conventional _ separatrix split _ @xcite for the lower layer @xmath163 with @xmath164 given by eq .
( a11 ) @xcite .
accordingly , we use the term lower gss curve for the following curve in the @xmath138 plane : @xmath165 0.2 cm 2.5 cm _
a. relatively small @xmath14 _ 0.2 cm if @xmath166 , where the critical value @xmath167 is determined by eq .
( 41 ) ( its origin will be explained further ) , then there are differences in the boundary formation for the frequency ranges of _ odd _ and _ even _ spikes .
we describe these ranges separately . 0.2 cm 2.9 cm _ a.1 . odd spikes _ 0.2 cm in this case , the boundary is formed by the trajectory of the hamiltonian ( 4 ) _ tangent _ to the gss curve ( see fig .
16(a ) ; cf . also figs .
8(a ) , 9(b ) , 9(c ) ) .
there are two tangencies in the angle range @xmath111-\pi,\pi[$ ] : they occur at the angles @xmath168 where @xmath169 is determined by eq .
( a21 ) . in the ranges of @xmath14 and @xmath15 relevant to the spike minimum ,
the asymptotic expressions for @xmath170 and @xmath169 are : @xmath171 } \sqrt{\frac{n\pi}{8\ln\left(1/\phi\right ) } } + \pi\frac{1-(-1)^{\left[\frac{n}{4}\right]}}{2}.\end{aligned}\ ] ] hence , the asymptotic value for the deviation of the tangency energy @xmath172 from the lower barrier reduces to : @xmath173 the minimal energy on the boundary , @xmath174 , corresponds to @xmath175 @xmath159 or @xmath127 for even or odd values of @xmath176 $ ] respectively .
thus , it can be found from the equality @xmath177})/2\right)= \nonumber \\ & & = \tilde{h}\left(i_t^{(l)}\equiv i(e_t^{(l)}),\tilde{\psi}_t^{(l)}\right).\end{aligned}\ ] ] at @xmath57 , eq .
( [ 27 ] ) yields the following expression for the minimal deviation of energy on the boundary from the barrier : @xmath178 in the context of global chaos onset , the most important property of the boundary is that the _ maximal _ deviation of its energy from the barrier , @xmath179 , greatly exceeds both @xmath180 and @xmath181 . as @xmath182 , the maximum of the boundary approaches the saddle .
0.2 cm 2.9 cm _ a.2 . even spikes _ 0.2 cm in this case , the hamiltonian ( 4 ) possesses saddles in the close vicinity to the lower barrier ( see fig .
their angles differ by @xmath127 from those of :
@xmath183}}{2}+2\pi m , \\ & & m=0,\pm 1 , \pm 2 , \ldots , \nonumber\end{aligned}\ ] ] while the deviation of their energies from the barrier still lies in the relevant ( resonant ) range and reads , in the lowest - order approximation , @xmath184 the lower whiskers of the separatrix generated by these saddles intersect the gss curve while the upper whiskers in the asymptotic limit do not intersect it ( fig .
thus , it is the upper whiskers of the separatrix which form the boundary of the chaotic layer in the asymptotic limit .
the energy on the boundary takes the minimal value right on the saddle , so that @xmath185 similar to the case of the odd spikes , the _ maximal _ ( along the boundary ) deviation of the energy from the barrier greatly exceeds both @xmath180 and @xmath181 . as @xmath182 , the maximum of the boundary approaches the saddle . 0.2 cm 2.9 cm _
b. relatively large @xmath14 _ 0.2 cm if @xmath186 , the previously described trajectory ( the tangent one or the separatrix , for the odd or even spike ranges respectively ) is encompassed by the separatrix of the lower nonlinear resonance and typically forms the boundary of the major stability island inside the lower layer ( reproduced periodically in @xmath120 with the period @xmath119 ) .
the upper _ outer _ boundary of the layer is formed by the upper part of the _ resonance separatrix_. this may be interpreted as the absorption of the lower resonance by the lower chaotic layer .
0.75 cm 0.75 cm let @xmath15 be close to any of the spikes minima .
one of the key roles in the formation of the lower boundary of the layer is played by the angle - dependent quantity @xmath187 which we call the _ generalized separatrix split _ ( gss ) for the upper layer ; @xmath188 is the separatrix split for the upper layer : @xmath189 with @xmath190 given by eq .
accordingly , we use the term upper gss curve for the following curve in the @xmath138 plane : @xmath191 0.2 cm 2.5 cm _
a. relatively small @xmath14 _ 0.2 cm if @xmath192 , where the critical value @xmath193 is determined by eq .
( 42 ) ( its origin will be explained further ) , then there are some differences in the boundary formation in the frequency ranges of _ odd _ and _ even _ spikes : for odd spikes , the formation is similar to the one for even spikes in the lower - layer case and vice versa .
0.2 cm 2.9 cm _ a.1 .
odd spikes _ 0.2 cm in this case , the hamiltonian ( 4 ) possesses saddles in the close vicinity to the upper barrier , analogous to the saddles near the lower barrier in the case of even spikes .
their angles are shifted by @xmath127 from those of :
@xmath194}}{2}+2\pi m , \\ & & m=0,\pm 1 , \pm 2 , \ldots \nonumber\end{aligned}\ ] ] the deviation of their energies from the upper barrier coincides , in the lowest - order approximation , with @xmath195 : @xmath196 the upper whiskers of the separatrix generated by these saddles intersect the upper gss curve while the lower whiskers in the asymptotic limit do not intersect it .
thus , it is the lower whiskers of the separatrix which form the boundary of the chaotic layer in the asymptotic limit .
the deviation of energy from the upper barrier takes its minimal ( along the boundary ) value right on the saddle
, @xmath197 the _ maximal _ ( along the boundary ) deviation of the energy from the barrier greatly exceeds both @xmath198 and @xmath199 . as @xmath200 , the maximum of the boundary approaches the saddle
. 0.2 cm 2.9 cm _ a.2 . even spikes _ 0.2 cm the boundary is formed by the trajectory of the hamiltonian ( 4 ) _ tangent _ to the gss curve .
there are two tangencies in the angle range @xmath111-\pi,\pi[$ ] : they occur at the angles @xmath201 where @xmath202 is determined by eq .
( a41 ) . in the ranges of @xmath14 and @xmath15 relevant to the spike minimum , the expressions for @xmath188 and @xmath202 in the asymptotic limit @xmath57 are similar to the analogous quantities in the lower - layer case : @xmath203 } \sqrt{\frac{n\pi}{8\ln\left(\frac{1}{\phi}\right ) } } + \pi\frac{1+(-1)^{\left[\frac{n}{4}\right]}}{2 } .\end{aligned}\ ] ] hence , the asymptotic value for the deviation of the tangency energy @xmath204 from the upper barrier reduces to : @xmath205}}{2}-\tilde{\psi}_t^{(u)}\right|= \nonumber \\ & = & \frac{\pi^{3/2 } } { 2}\frac{h}{\sqrt{\ln\left(1/\phi\right)/n}}.\end{aligned}\ ] ] the maximal energy on the boundary , @xmath206 , corresponds to @xmath207})/2 $ ] .
thus , it can be found from the equality @xmath208})/2)= \nonumber \\ & & = \tilde{h}(i_t^{(u)}\equiv i(e_t^{(u)}),\tilde{\psi}_t^{(u)}).\end{aligned}\ ] ] at @xmath57 , eq .
( 39 ) yields the following expression for the minimal ( along the boundary ) deviation of energy from the barrier : @xmath209 0.2 cm 2.9 cm
_ b. relatively large @xmath14 _ 0.2 cm if @xmath210 ( cf .
8(a ) ) , the previously described trajectory ( tangent one or the separatrix , for the even and odd spikes ranges respectively ) is encompassed by the separatrix of the upper nonlinear resonance and typically forms the boundary of the major stability island inside the upper layer ( reproduced periodically in @xmath120 with the period @xmath119 ) .
the lower _
outer _ boundary of the layer is formed in this case by the lower part of the _ resonance separatrix_. this may be interpreted as the absorption of the upper resonance by the upper chaotic layer .
0.5 cm the description of chaotic layers given above and , in more details , in the appendix is the * first main result of this paper*. it provides a _ rigorous base _ for our intuitive assumption that the minimal value of @xmath14 at which the layers overlap corresponds to the reconnection of the nonlinear resonances with each other while the reconnected resonances touch one of the layers and touch / overlap another layer .
it is remarkable also that we have managed to obtain the _ quantitative _ theoretical description of the chaotic layers boundaries in the _ phase space _ , including even the major stability islands , that well fits the results of simulations ( see fig .
8(b ) ) .
the condition for the merger of the lower resonance and the lower chaotic layer may be written as @xmath211 the condition for the merger of the upper resonance and the upper chaotic layer may be written as @xmath212 for the global chaos onset related to the spike minimum , either of eqs . ( 41 ) and ( 42 ) should be combined with the condition of the separatrix reconnection ( 20 ) .
let us seek first only the leading terms of @xmath213 and @xmath214 .
then ( 20 ) may be replaced by its lowest - order approximation ( 21 ) or , equivalently , ( 22 ) .
using also the lowest - order approximation for the barriers ( @xmath215 ) , we reduce eqs .
( 41 ) , ( 42 ) respectively to @xmath216 where @xmath217 is given by ( 28 ) or ( 31 ) for the odd or even spikes respectively while @xmath218 is given by ( 35 ) or ( 40 ) for the odd or even spikes respectively .
the solution @xmath219 of the system of equations ( 22),(43 ) and the solution @xmath220 of the system of equations ( 22),(44 ) turn out _ identical _ to the _ leading _ order . for the sake of brevity
, we derive below just @xmath219 , denoting the latter , in short , as @xmath221 @xcite .
the system of algebraic equations ( 22 ) and ( 43 ) is still too complicated to find its exact solution .
however , we need only the _ lowest - order _ solution - and this simplifies the problem . still , even this simplified problem is not trivial , both because the function @xmath222 turns out to be non - analytic and because @xmath223 in ( 22 ) is very sensitive to @xmath15 in the relevant range . despite these difficulties
, we have found the solution in a _ self - consistent _ way , as briefly described below .
assume that the asymptotic dependence @xmath224 is : @xmath225 where the constant @xmath226 may be found from the asymptotic solution of ( 22 ) , ( 43 ) , ( 45 ) .
substituting the energies @xmath227 and @xmath228 in ( 7 ) and taking into account ( 28 ) , ( 31 ) , ( 35 ) , ( 40 ) and ( 45 ) , we find that , both for the odd and even spikes , the inequality @xmath229 holds in the whole relevant range of energies , i.e. for @xmath230.\ ] ] thus , the resonant approximation is valid in the whole range ( 47 ) .
( 12 ) for @xmath151 is valid in the whole relevant range of energies too .
consider eq .
( 43 ) in a more explicit form .
namely , we express @xmath15 from ( 43 ) , using eqs .
( 4 ) , ( 12 ) , and ( 13 ) , and using also ( 28)/(31 ) for odd / even spikes , and ( 45 ) : @xmath231 we emphasize that the value of @xmath180 enters explicitly only the term @xmath232 while , as it is clear from the consideration below , this term does not affect the leading terms in @xmath221 .
thus , @xmath180 does not affect the leading term of @xmath233 at all , while it affects the leading term of @xmath234 only _ implicitly _ :
@xmath180 lies in the range of energies where @xmath235 .
this latter quantity is present in the second term in the curly brackets in ( 48 ) and , as it is clear from the further consideration , @xmath234 is proportional to it . substituting ( 48 ) into the expression for @xmath223 in ( 22 ) , using ( 45 ) and keeping only the leading terms , we obtain @xmath236 substituting @xmath223 from ( 49 ) into eq . (
22 ) for @xmath158 and allowing for ( 45 ) once again , we arrive at the transcendental equation for @xmath237 : @xmath238 the approximate numerical solution of eq .
( 50 ) is : @xmath239 thus , the final leading - order asymptotic formulas for @xmath15 and @xmath14 in the minima of the spikes are the following : @xmath240 where the constant @xmath241 is the solution of eq . (
the rigorous derivation of the explicit asymptotic formulas for the minima of @xmath0 is the * second main result of this paper*. these formulas allow one to immediately predict the parameters for the weakest perturbation which may lead to global chaos . for @xmath161 ,
the numerical simulations give the following values for the frequencies in the minima of the first two spikes ( see fig .
4 ) : @xmath242 the values by the lowest - order formula ( 52 ) are : @xmath243 in rather good agreement with the simulations . the next - order correction for @xmath233 can be immediately found from eq .
( 48 ) for @xmath15 and eq .
( 52 ) for @xmath244 , so that @xmath245 the formula ( 55 ) agrees with the simulations even better than the lowest - order approximation : @xmath246 for @xmath14 in the spikes minima , the simulations give the following @xcite values ( see fig .
4 ) : @xmath247 the values by the lowest - order formula ( 52 ) are : @xmath248 the theoretical value @xmath249 gives a reasonable estimate for the simulation value @xmath250 .
the theoretical value @xmath251 gives the correct order of magnitude for the simulation value @xmath252 .
thus , the accuracy of the lowest - order formula ( 52 ) for @xmath234 is much lower than that for @xmath253 : this is due to the steepness of @xmath0 in the ranges of spikes ( the steepness , in turn , is due to the flatness of the function @xmath7 near its maximum ) .
moreover , as the number of the spike @xmath254 increases , the accuracy of the lowest - order value @xmath255 significantly decreases .
the latter can be explained as follows . for the next - order correction to @xmath255 ,
the dependence on @xmath61 reads as : @xmath256 at least some of the terms of this correction are positive and proportional to @xmath255 ( e.g. due to the difference between the exact equation ( 15 ) and its approximate version ( 17 ) ) while @xmath255 is proportional to @xmath257 .
thus , for @xmath161 , the relative correction for the 1st spike is @xmath258 while the correction for the 2nd spike is a few times larger i.e. @xmath259 .
but the latter means that , for @xmath161 , the asymptotic theory for the 2nd spike can not pretend to be a quantitative description of @xmath260 , but only provide the correct order of magnitude . besides , if @xmath134 while @xmath61 exceeds some critical value , then the search of the minimum involves eq .
( 66 ) rather than eq .
( 20 ) , as explained below in sec .
10(b ) and 11 ) .
altogether , this explains why @xmath250 is larger than @xmath249 only by 50@xmath261 while @xmath252 is larger than @xmath251 by 200@xmath261 .
the consistent explicit derivation of the correction to @xmath255 is complicated .
a reasonable alternative may be a proper _ numerical _ solution of the algebraic system of eqs .
( 20 ) @xcite and ( 41 ) for the odd spikes or ( 42 ) for the even spikes @xcite . to this end , in eqs .
( 20 ) @xcite and ( 41)/(42 ) we use : ( i ) the exact values of the saddle energies obtained from the exact relations ( 16 ) instead of the approximate relations ( 17 ) ; ( ii ) the exact formulas ( 5 ) and ( 6 ) for @xmath7 instead of the asymptotic expression ( 7 ) ; ( iii ) the exact functions @xmath151 instead of the asymptotic formula ( 12 ) ; ( iv ) the relation ( 27 ) and the calculation of the tangent
state @xmath262 by eqs .
( a11 ) , ( a22 ) for the odd spikes , or relation ( 39 ) and the calculation of the tangent
state @xmath263 by eqs .
( a41)-(a43 ) for the even spikes .
note that , to find the exact function @xmath151 , we substitute into the definition of @xmath151 in ( 4 ) the explicit @xcite solution for @xmath100 : @xmath264 , \nonumber \\ & & q(e,\psi)=\pi - q(e,\pi-\psi ) \quad\quad\quad\quad { \rm for } \quad \psi\in\left[\frac{\pi}{2},\pi\right ] , \nonumber \\ & & q(e,\psi)=q(e,2\pi-\psi ) \quad\quad\quad\quad\quad\quad { \rm for } \quad \psi\in\left[\pi,2\pi\right ] , \nonumber \\ & & \eta\equiv \frac{1}{2}(\sqrt{2e}-\phi+1){\rm sn}^2 \left(\frac{2k}{\pi}\psi \right ) \ , , \end{aligned}\ ] ] where @xmath265 is the elliptic sine @xcite with the same modulus @xmath266 as the full elliptic integral @xmath267 defined in ( 5),(6 ) .
the numerical solution described above gives : @xmath268 the agreement with the simulation results is : ( i ) excellent for @xmath269 for the both spikes and for @xmath234 for the 1st spike , ( ii ) reasonable for @xmath234 for the 2nd spike .
thus , if @xmath61 is _ moderately _ small , a much more accurate prediction for @xmath234 than that by the lowest - order formula is provided by the numerical procedure described above .
the goal of this section is to find mechanisms responsible for the formation of the spikes _ wings _
( i.e. the function @xmath0 in the ranges of @xmath15 slightly deviating from @xmath60 ) and to provide for their theoretical description . before developing the theory , we briefly analyze the simulation data ( fig .
4 ) , concentrating on the 1st spike .
the left wing of the spike is smooth and nearly straight .
the initial part of the right wing is also nearly straight @xcite though less steep .
but , at some small distance from @xmath270 , its slope changes jump - wise by a few times : compare the derivative @xcite @xmath271 at @xmath15 slightly exceeding @xmath37 ( see the left inset in fig .
4 ) and @xmath272 at @xmath273 ( see the main part of fig .
, even prior to the theoretical analysis , one may assume that there are a few different important mechanisms responsible for the formation of the wings .
consider the arbitrary @xmath254th spike .
we have shown in the previous section that , in the asymptotic limit @xmath57 , the minimum of the spike corresponds to the intersection between the lines ( 20 ) and ( 41)/(42 ) for odd / even spikes .
we recall that : ( i ) eq . ( 20 ) corresponds to the overlap in phase space between nonlinear resonances of the same order @xmath257 ; ( ii ) eq . ( 41)/(42 ) corresponds to the onset of the overlap between the resonance separatrix associated with the lower / upper saddle and the chaotic layer associated with the lower / upper potential barrier ; ( iii ) for @xmath30 , the condition ( 41)/(42 ) guarantees also the overlap between the upper / lower resonance separatrix and the chaotic layer associated with the upper / lower barrier @xcite .
0.75 cm 0.75 cm 0.75 cm if @xmath15 becomes slightly smaller than @xmath60 the resonances shift closer to the barriers while moving apart from each other .
hence , as @xmath14 increases , the overlap of the resonances with the chaotic layers associated with the barriers occurs earlier than with each other .
therefore , at @xmath274 , the function @xmath0 should approximately correspond to the reconnection of resonances of the order @xmath275 ( fig .
9(a ) ) . fig .
10(a ) demonstrates that even the asymptotic formula ( 22 ) for the separatrix reconnection line fits the left wing of the 1st spike quite well while the numerically calculated line ( 20 ) agrees with the simulations perfectly . if @xmath15 becomes slightly larger than @xmath60 then , on the contrary , the resonances shift closer to each other and more far from the barriers .
therefore , the overlap of resonances with each other occurs at smaller @xmath14 than the overlap between any of them and the chaotic layer associated with the lower / upper barrier ( cf . figs .
5(c ) and 5(d ) as well as 6(c ) and 6(d ) ) .
hence , it is the latter overlap which determines the function @xmath0 in the relevant range of @xmath15 ( fig .
9(b ) ) . fig .
10 shows that @xmath0 is indeed well approximated in the close vicinity to the right from @xmath60 by the numerical solution of eq .
( 41)/(42 ) for an odd / even spike and , for the 1st spike and the given @xmath61 , even by its asymptotic form , @xmath276 } { 2\sqrt{2 } } , \\ & & \delta e^{(1)}\equiv \sqrt{\phi^2 - 64\exp(-\frac{n\pi}{\omega_f } ) } , \nonumber \\ & & n\equiv 2j-1,\quad |\omega_f-\omega_s^{(j)}|\ll \omega_m \ , , \nonumber\end{aligned}\ ] ] 0.75 cm the mechanism described above determines @xmath0 only in the close vicinity of @xmath60 .
if @xmath277 becomes too close to @xmath16 or exceeds it , then the resonances are not of immediate relevance : they may even disappear or , if they still exist , their closed loops shrink , so that they can not anymore provide for the connection of the chaotic layers in the relevant range of @xmath14 .
at the same time , the closeness of the frequency to @xmath16 still may give rise to a large variation of action along the trajectory of the hamiltonian system ( 4 ) . for the odd / even spikes , the boundaries of the chaotic layers in the asymptotic limit @xmath57 are formed in this case by the trajectory of ( 4 ) which is tangent to the lower / upper gss curves ( for the lower / upper layer ) or by the lower / upper part of the separatrix of ( 4 ) generated by the saddle @xmath278/@xmath279
( for the upper / lower layer ) .
obviously , the overlap of the layers occurs when these trajectories coincide with each other , that may be formulated as the equality of @xmath153 in the corresponding tangency and saddle : @xmath280 note however that , for _ moderately _ small @xmath61 , the tangencies may be relevant both to the lower layer and to the upper one ( see the appendix ) .
indeed , such a case occurs for our example with @xmath161 : see fig .
therefore , the overlap of the layers corresponds to the equality of @xmath153 in the tangencies : @xmath281 to the lowest order , eq .
( 63 ) and eq .
( 64 ) read as : @xmath282 both the line ( 64 ) and the asymptotic line ( 65 ) well agree with the part of the right wing of the 1st spike situated beyond the immediate vicinity of the minimum from the right side , namely , to the right from the fold at @xmath283 ( fig .
the fold corresponds to the change of the mechanisms of the chaotic layers overlap
. if @xmath61 is moderately small while @xmath134 , the description of the far wings by the numerical lines ( 20 ) and ( 64 ) may be still quite good but the asymptotic lines ( 22 ) and ( 65 ) can not pretend to describe the wings quantitatively anymore ( fig .
10(b ) ) . as for
the very minimum of the spike and the wings in the close vicinity to it , one more mechanism may become relevant for their formation in this case ( figs .
10(b ) and 11 ) .
this mechanism may be explained as follows . if @xmath134 , then @xmath151 becomes zero in the close vicinity ( @xmath284 ) of the relevant barrier ( the upper / lower barrier , in the case of even / odd spikes : cf .
as follows from the equations of motion ( 14 ) , no trajectory can cross the line @xmath285 . in the asymptotic limit @xmath57 ,
provided @xmath14 is from the relevant range , almost the whole gss curve is farer from the barrier than the line @xmath285 , and the latter becomes irrelevant .
but , for a moderately small @xmath61 , the line may separate the whole gss curve from the rest of the phase space .
then the resonance separatrix can not connect to the gss curve even if there is a state on the latter curve with the same value of @xmath153 as on the resonance separatrix .
for a given @xmath15 , the connection requires then a higher value of @xmath14 : for such a value , the gss curve itself crosses the line @xmath285 . in the relevant range of @xmath14 ,
the resonance separatrix passes very close to this line , so that the connection is well approximated by the condition that the gss curve _ touches _ this line ( see the inset in fig . 11 ) : @xmath286
this mechanism is relevant for the formation of the minimum of the 2nd spike at @xmath161 , and in the close vicinity of the spike , on the left ( fig .
10(b ) ) . finally , let us explicitly find the _
universal asymptotic shape _ of the spike in the vicinity of its minimum .
first , we note that the lowest - order expression ( 62 ) for the spike between the minimum and the fold can be written as the _ half - sum _ of the expressions ( 22 ) and ( 65 ) ( which represent the lowest - order approximations for the spike to the left of the minimum , and to the right of the fold respectively ) .
thus , all three lines ( 22 ) , ( 62 ) and ( 65 ) intersect in one point .
the latter means that , in the asymptotic limit @xmath57 , the fold merges with the minimum : @xmath15 and @xmath14 in the fold asymptotically approach @xmath233 and @xmath234 respectively .
thus , though the fold is a generic feature of the spikes , it is not of main significance : the spike is formed basically by two straight lines .
the ratio between their slopes is universal .
so , introducing a proper scaling , we reduce the spike shape to a universal function ( fig .
12 ) : @xmath287 where @xmath288 and @xmath244 are the lowest - order expressions ( 52 ) respectively for the frequency and amplitude in the spike minimum , @xmath289 is the expression ( 55 ) for the frequency in the spike minimum , including the first - order correction , and @xmath237 is the constant ( 51 ) . beside the left and right wings of the universal shape ( the solid lines in fig .
12 ) , we also present in ( 67 ) the function @xmath290 ( the dashed line in fig .
12 ) : it purposes to show that , on one hand , the fold asymptotically merges with the minimum but , on the other hand , the fold is generic and the slope of the spike between the minimum and the fold has a universal ratio to any of the slopes of the major wings . even for a moderately small @xmath61 , like in our example ,
the ratios between the three slopes related to the 1st spike in the simulations are reasonably well reproduced by those in eq .
( 67 ) : cf .
10(a ) and 12 .
it follows from ( 67 ) that the asymptotic scaled shape is universal i.e. independent of @xmath61 , @xmath36 or any other parameter .
the description of the wings of the spikes near their minima , in particular the derivation of the spike universal shape , is the * third main result of this paper*.
the _ new approach _ for the treatment of separatrix chaos opens a broad variety of important generalizations and applications , some of which are discussed below . 1 .
it may be applied to _ any _ separatrix layer , including in particular the conventional _ single - separatrix _ case .
this is possible due to the characteristic dependence of the frequency of eigenoscillation on energy in the vicinity of any separatrix : the frequency keeps nearly a _ constant _ value even if the deviation of the energy from the separatrix _ strongly varies _ within a given scale of the deviation .
to the best of our knowledge , the latter idea was not exploited before , which is why probably it was thought to be impossible to analyze the dynamics of the separatrix map in an explicit form .
there were various estimates of the layer width in energy ( see @xcite and references therein ) but the quantitative analysis of the layer boundaries in the _ phase space _ was never done in the non - adiabatic case .
in contrast , our approach does analyze the dynamics of the separatrix map , and this allows us in particular to quantitatively describe the layer boundaries in case when the perturbation is resonant with the eigenoscillation in the relevant range of energy .
it follows from such a description that , for a given small amplitude @xmath14 of the perturbation , the maximal layer width in energy is much larger than it is assumed by former theories ( cf .
@xcite where the maximal width is assumed to be @xmath291 ) .
thus , we can quantitatively describe large _ jumps _ and _ peaks _ of the layer width as a function of the perturbation amplitude and frequency respectively @xcite .
moreover , a rough estimate on the basis of our approach indicates that , for many classes of systems , the relative range of such jumps / peaks ( i.e. the ratio between the upper and lower levels of the jump / peak ) _ diverges _ in the asymptotic limit @xmath292 .
2 . apart from the description of the boundaries , the approach allows us to describe the _
chaotic transport _ within the layer .
in particular , it may allow us to calculate a positive lyapunov exponent and to describe diffusion .
our approach may be generalized for a _ non - resonant _ perturbation .
the resonance approximation is not valid then but there still remains the property of a nearly constancy of the frequency of eigenoscillation within an arbitrary given scale of the deviation of energy from the separatrix .
this property may allow us to explicitly describe the dynamics of the separatrix map for any frequency of perturbation which is less than or of the order of the resonance frequency .
4 . apart from hamiltonian systems of the one and a half degrees of freedom and corresponding zaslavsky separatrix maps
, our approach may be useful in the treatment of other chaotic systems and separatrix maps ( see @xcite for the most recent major review on various types of separatrix maps and related continuous chaotic systems ) .
as concerns the _ facilitation of the global chaos onset _ between adjacent separatrices , we mention the following generalizations : * 1 .
* the spikes in @xmath0 may occur for an _ arbitrary hamiltonian _ system with two or more separatrices .
the asymptotic theory can be generalized accordingly .
the absence of pronounced spikes at _ even _ harmonics @xmath293 is explained by the symmetry of the potential ( 1 ) : the even fourier harmonics of the coordinate , @xmath294 , are equal to zero . for time - periodic perturbation of dipole type , like in eq .
( 2 ) , there are no resonances of even order due to this symmetry @xcite . if either the potential is _ non - symmetric _ or the additive perturbation of the hamiltonian is not an _ odd _ function of the coordinate , then even - order resonances do exist , resulting in the presence of the spikes in @xmath0 at @xmath295 . * 3 .
* there may be an additional facilitation of global chaos onset which is reasonable to call a secondary facilitation .
let the frequency @xmath15 be close to the frequency @xmath233 of the spike minimum while the amplitude @xmath14 be @xmath296 but still lower than @xmath0 .
then there are two resonance separatrices in the @xmath138 plane which are not connected by the chaotic transport ( cf . fig .
6(b ) and 5(b ) ) .
this system possesses the zero - dispersion property .
the trajectories of the resonant hamiltonian ( 4 ) which start in between the separatrices oscillate in @xmath113 ( as well as in @xmath297 ) .
the frequency @xmath298 of such oscillations along a given trajectory depends on the corresponding value of @xmath153 analogously to as @xmath3 depends on @xmath10 for the original hamiltonian @xmath74 : @xmath299 is equal to zero for the values of @xmath153 corresponding to the separatrices ( being equal in turn to @xmath300 and @xmath301 : see eq .
( 19 ) ) while possessing a nearly rectangular shape in between , provided the quantity @xmath302 is much smaller than the variation of @xmath153 within each of the resonances , @xmath303 and @xmath304 are the values of @xmath153 in the elliptic point of the lower and upper resonance respectively .
the maximum of @xmath299 in between @xmath300 and @xmath301 is asymptotically described by the following formula : @xmath305 if we additionally perturb the system in such a way that an additional time - periodic term of the frequency @xmath306 arises in the resonance hamiltonian , then the chaotic layers associated with the resonance separatrices may be connected by chaotic transport even for a rather small amplitude of the additional perturbation , due to a scenario similar to the one described in this paper
. there may be various types of such additional perturbation @xcite .
for example , one may _ add _ to @xmath307 ( 2 ) one more dipole time - periodic perturbation of _ mixed _ frequency ( i.e. @xmath308 ) .
alternatively , one may directly perturb the _ angle _ of the original perturbation by a _ low - frequency _ perturbation , i.e. the time - periodic term in @xmath307 ( 2 ) is replaced by the term @xmath309 recent physical problems where a similar situation is relevant are : chaotic mixing and transport in a meandering jet flow @xcite and reflection of light rays in a corrugated waveguide @xcite .
if the time - periodic perturbation is _ multiplicative _ rather than additive , the resonances become _ parametric _ ( cf.@xcite ) .
parametric resonance is more complicated and much less studied than nonlinear resonance .
nevertheless , the main mechanism for the onset of global chaos remains the same , namely the combination of the reconnection between resonances of the same order and of their overlap in energy with the chaotic layers associated with the barriers . at the same time , the frequencies of the main spikes in @xmath0 may change ( though still being related to @xmath16 ) .
we consider below the example when the periodically driven parameter is the parameter @xmath61 in ( 1 ) @xcite .
the hamiltonian reads as @xmath310 the term @xmath311 in @xmath307 ( 71 ) may be rewritten as @xmath312 .
the last term in the latter expression does not affect the equations of motion .
thus , we end up with an additive perturbation @xmath313 . in the asymptotic limit @xmath314 , the @xmath36th - order fourier component of the function @xmath315
can be shown to differ from zero only for the orders @xmath316 therefore one may expect the main spikes in @xmath0 to be at frequencies twice larger than those for the dipole perturbation ( 2 ) : @xmath317 this well agrees with the results of simulations ( fig .
13 ) . moreover , the asymptotic theory for the dipole perturbation may be immediately generalized to the present case : it is necessary only to replace the fourier component of the coordinate @xmath24 by the fourier component of the function @xmath315 : @xmath318 ( cf .
( 12 ) for @xmath101 ) .
we obtain : @xmath319 where @xmath237 is given in eqs . ( 50 ) and ( 51 ) . for @xmath320 ,
( 74 ) gives , for the 1st spike , values differing from the simulation data by about @xmath321 for the frequency and by about @xmath322 for the amplitude .
thus , the lowest - order formulas accurately describe the 1st spike even for a moderately small @xmath61 .
one more generalization relates to _ multi - dimensional _ hamiltonian systems with two or more saddles with different energies : the perturbation may not necessarily be time - periodic , in this case . the detailed analysis will be done elsewhere . finally , we point out some analogy between the facilitation of global chaos onset described in our work and the so called stochastic percolation in 2d hamiltonian systems described in @xcite : the merging of internal and external chaotic zones is also relevant there , like in our case . however , both the models and the underlying mechanisms are very different .
namely , the problem studied in @xcite is not of the zero - dispersion type ; and the two - dimensionality is inherently important .
let us turn now to a few detailed examples of applications of the facilitation of global chaos onset .
the first application relates to a classical electron gas in a magnetic superlattice @xcite , where the electrons may be considered as non - interacting quasi - particles moving on a plane perpendicular to the magnetic field spatially periodic in one of the in - plane directions ( we denote it as @xmath323-direction ) .
then the electron motion in the x - direction is described by the hamiltonian ( 1 ) in which @xmath24 and @xmath324 are the scaled electron coordinate @xmath323 and generalized momentum @xmath325 respectively , while the parameter @xmath61 is proportional to the reciprocal amplitude of the external magnetic field , @xmath326 , and to the generalized momentum @xmath327 in the second ( perpendicular to @xmath323 ) in - plane direction : see for details @xcite .
note that @xmath327 remains constant during the motion @xcite .
if an ac electric field is applied in the @xmath323-direction , then the model ( 2)-(1 ) becomes relevant .
the _ dc - conductivity _ in the @xmath323-direction is proportional to the fraction of electrons that can take part to the unbounded motion in the @xmath323-direction .
this fraction , in turn , significantly grows as the range of energies involved in the unbounded chaotic transport increases @xcite .
if electrons move in vacuum @xcite , then it may be possible to inject a beam of electrons which possess the same velocity . in this case
, the parameter @xmath61 in the model ( 1 ) has some certain value , so that the results obtained in the previous sections are directly applicable .
the spikes in @xmath0 mean a drastic increase of the dc - conductivity occurring at a very weak amplitude of the ac electric field .
the frequency of the ac field should be close to one of the spike frequencies .
the effect is especially pronounced for the 1st spike i.e. when @xmath328 $ ] .
if the electron motion takes place in a semiconductor @xcite , then the velocity in the @xmath329-direction is necessarily _ statistically distributed_. the same concerns the parameter @xmath61 then .
this might seem to smear the effect : cf .
@xcite where the _ conventional _ scenario of the onset of global chaos was exploited
. however , in the case of the zero - dispersion scenario suggested in our paper , it typically should not be so .
indeed , a statistical distribution of the velocity typically decreases _ exponentially sharply _ as the velocity exceeds some characteristic value @xmath330 : for high temperature @xmath331 , the boltzmann distribution of energy is relevant and , therefore , @xmath332 ; for low temperatures , the fermi distribution is relevant and , therefore , @xmath333 where @xmath334 is the fermi energy . on the other hand , in the range of small @xmath61 ,
the function @xmath335 does not significantly change even if @xmath61 changes by a few times .
hence , if @xmath336 ( where @xmath337 denotes the @xmath61 value corresponding to @xmath330 ) , then the frequency @xmath270 of the partial ( i.e. for a given value of @xmath338 ) 1st - order spike is nearly the same for most of the velocities in the relevant range @xmath339 , and it is approximately equal to @xmath340 $ ] .
similarly , @xmath341 .
thus , as a function of @xmath15 for fixed @xmath342 , the dc - conductivity should have a sharp maximum for @xmath343 .
if the parameter @xmath61 is time - dependent ( e.g. if the external magnetic field has an ac component or / and there is an ac electric force perpendicular to the @xmath323-direction ) , then the applications may be similar , with the only difference that the values of @xmath15 and @xmath14 in the minima of the main spikes differ from those for the additive perturbation , see fig .
13 and the related discussion .
the results of the present paper may also be of direct relevance for a pendulum spinning about its vertical axis @xcite , provided the friction is small . the periodic driving may be easily introduced mechanically , or electrically if the pendulum is electrically charged , or magnetically if the pendulum includes a ferromagnetic material . finally , we mention in this subsection that potentials similar to @xmath90 ( 1 ) , i.e. periodic potentials with two different - height barriers per period , may readily be generated for cold atoms by means of optical lattices @xcite . the dissipation may be suppressed by means of the detuning from the atomic resonance @xcite .
then the results of our paper are also of direct relevance to such systems . consider the noise - induced escape over a potential barrier in the presence of a non - adiabatic periodic driving .
for a moderately weak damping , such a driving decreases the activation barrier due to the resonant pushing the system in the range of resonant energies @xcite .
if the damping is even smaller , the decrease of the activation barrier becomes larger , due , however , to a different mechanism , typically related to the chaotic layer associated with the separatrix of the unperturbed system @xcite .
the lower energy boundary of the layer is smaller than the potential barrier energy , so it is sufficient that the noise pulls the system right to this boundary ( rather than to the very top of the potential barrier ) , after which the system may escape over the barrier purely dynamically .
if the eigenfrequency as a function of the energy possesses a local maximum , then the effect may be even more pronounced @xcite : the decrease of the activation barrier may become comparable to the potential barrier at unusually small amplitudes of the driving , provided the driving frequency is close to the extremal eigenfrequency .
one of the main mechanisms of the latter effect is closely related to phenomena discussed in the present paper . in the case of escape over
_ two _ barriers of different heights , the effect should become even more pronounced due to the mechanism responsible for the spikes of @xmath0 studied in the present paper . if the potential is periodic , e.g. like in optical lattices @xcite , the effect may lead to a drastic acceleration of the spatial diffusion . our results may be applied to the stochastic web formation @xcite .
if a harmonic oscillator is perturbed by a plane wave whose frequency is equal to the oscillator eigenfrequency or its multiple , then the perturbation plays two roles @xcite . on one hand , due to the resonance with the oscillator
, it transforms the structure of the phase space of the oscillator , leading to an infinite number of cells divided by a unique separatrix .
it has the form of a web of an infinite radius . on the other hand , the perturbation dresses " this separatrix by an exponentially narrow chaotic layer ( it is sometimes called stochastic layer ) .
such a web - like layer is called _
stochastic web_. it may lead to chaotic transport of the system for rather long distances both in coordinate and in energy . in case when either the resonance is not exact or / and the unperturbed oscillator possesses some nonlinearity , the perturbation generates _ many _ separatrices embedded into each other @xcite rather than one single infinite web - like separatrix .
then a significant chaotic transport in energy may arise only if the magnitude of the perturbation exceeds some critical value corresponding to the overlap of chaotic layers associated , at least , with two neighbouring separatrices . and
, still , the transport in energy remains limited since the width of the chaotic layer around each separatrix sharply decreases as the energy increases .
it can be shown @xcite that some types of additional time - periodic perturbation lead to a low - frequency dipole perturbation of the resonance system ( cf .
the paragraph preceding eq .
the structure of separatrices in the reduced system possesses properties similar to that of the system considered in the present paper .
indeed , in the region between the separatrices , the resonance system performs regular oscillations , and the frequency of such oscillations , as function of the value of the resonance hamiltonian , is equal to zero at each of the separatrices .
thus , it necessarily possesses a local maximum between energies corresponding to any two neighbouring separatrices , like in the case considered in the present paper .
if the additional perturbation has an optimal frequency related to one of these local maxima , then the overlap of chaotic layers associated with neighbouring separatrices is greatly facilitated , similarly to the case considered in the present paper .
moreover , the local maximum of the eigenfrequency changes from pair to pair of separatrices _ weakly _ , so that if the magnitude of the auxiliary perturbation exceeds the critical value even slightly the simultaneous overlap between _ many _
chaotic layers may occur .
then , the distance of the chaotic transport in energy greatly increases .
similar applications are relevant for the so called homogeneous ( sometimes called periodic ) stochastic webs @xcite and many other web - like stochastic structures @xcite . beside classical systems
, stochastic webs may arise in quantum systems too .
it was recently demonstrated , both theoretically @xcite and experimentally @xcite , that the stochastic web may play a crucial role in quantum electron transport in semiconductor superlattices subjected to stationary electric and magnetic fields . due to the spatial periodicity with a period of about a few nanometers ,
the system possesses narrow minibands in the electron spectrum .
it turns out that the description of the electron transport in the lowest miniband may be approximated by the model of a classical harmonic oscillator driven by a plane wave .
the role of the harmonic oscillator is played by the cyclotron motion while the wave arises due to the interplay between the cyclotron motion and bloch oscillations .
if the cyclotron and bloch frequencies are commensurate , then the phase space of such a system is threaded by a stochastic web .
this gives rise to the delocalization of electron orbits , that leads in turn to a strong increase of the conductivity @xcite .
however , this effect occurs only when the ratio between the electric and magnetic fields lies in the _ exponentially narrow _ regions corresponding to nearly integer ratios between the bloch and cyclotron frequencies .
the results of the present work suggest a method for a significant increase of the width of the relevant regions .
if the cyclotron and bloch frequencies are not exactly commensurate , then the stochastic web does not arise : rather a set of embedded separatrices arises provided the effective wave amplitude is sufficiently large . as discussed in the previous paragraph ,
even a rather weak time - periodic driving @xcite of the optimal frequency may significantly increase the area of the phase space involved in the chaotic transport .
this may provide for an effective control of the electron transport in such a system and may be used for developing electronic devices that exploit the intrinsic sensitivity of chaos ( cf .
a similar effect may be used also to control transmission through other periodic structures , e.g. ultra - cold atoms in optical lattices @xcite and photonic crystals @xcite .
we have developed a new _ general approach _ for the treatment of separatrix chaos .
this has allowed us to create a self - contained theory for the drastic facilitation of the global chaos onset between adjacent separatrices of a i d hamiltonian system subject to a time - periodic perturbation .
both the new approach and the theory of the facilitation are closely interwoven in our paper but , at the same time , each of these two items is relevant even on its own .
that is why we summarize them separately .
* i. * _ the new approach for separatrix chaos_. the approach is based on the separatrix map analysis which uses the characteristic property of the dependence of the frequency of eigenoscillation on energy in the vicinity of a separatrix : the frequency keeps nearly a _ constant _ value even if the deviation of the energy from the separatrix _ strongly varies _ within a given scale of the deviation . due to this
, the separatrix map evolves along the major part of the chaotic trajectory in _ regular - like _ way .
the deviation of the chaotic trajectory from the separatrix may vary along the regular - like parts of the trajectory in a much wider range than along the irregular parts . in the case of resonant perturbation
, we match the separatrix map analysis and the resonant hamiltonian approximation .
this allows us in particular to find the boundaries of the chaotic layers in the _ phase space _ , which well agrees with computer simulations ( fig .
the latter theory has been successfully applied by us to the problem of the global chaos onset in the double - separatrix case , which is summarized in the item ii below .
applications _ and _ generalizations _ of the approach include in particular the following .
* it may be applied to the conventional _ single - separatrix _ case .
in particular , our theory predicts that the _ maximal width _ of the separatrix chaotic layer in energy is typically @xmath344 , in contrast with former theories @xcite which assume that the maximal width is @xmath291 . * it allows to analyze the _ transport _ within the layer . *
it may be generalized for a _ non - resonant _ perturbation and for a _ higher _ dimension .
* ii . * _ the facilitation of the global chaos onset .
_ we have considered in details the characteristic example of a hamiltonian system possessing _ two or more separatrices _ , subject to a time - periodic perturbation .
the frequency @xmath3 of oscillation of the unperturbed motion necessarily possesses a _
local maximum _ @xmath16 as a function of energy @xmath10 in the range between the separatrices .
it is smaller than the frequency @xmath345 of eigenoscillation in the stable state of the hamiltonian system by a factor @xmath346 where @xmath347 is the difference of the separatrices energies , while @xmath348 is the difference between the upper separatrix energy and the stable state energy .
if @xmath349 , the function @xmath7 is close to @xmath16 for most of the energy range between the separatrices : in the asymptotic limit @xmath57 , @xmath7 approaches a _ rectangular _ form . besides , the amplitude @xmath101 of the @xmath36th fourier harmonic of the oscillation asymptotically approaches a _
non - small _ constant value in the whole energy range between separatrices .
these two properties are responsible for most of the characteristic features of the global chaos onset in between the separatrices .
the most striking one is a _ drastic facilitation _ of the global chaos onset when the perturbation frequency @xmath15 approaches @xmath16 or its multiples : the perturbation amplitude @xmath33 required for global chaos possesses , as a function of the perturbation frequency @xmath15 , deep spikes close to @xmath16 or its multiples .
on the basis of the theory for the boundaries of the chaotic layers , we have developed a _ self - consistent _ asymptotic theory for the _ spikes _ in the vicinity of the minima . in particular , the _ explicit _ asymptotic expressions for the very minima are given in eqs .
( 52 ) and ( 55 ) .
the minimal amplitude @xmath33 is smaller than the typical @xmath33 for @xmath15 beyond the close vicinity of @xmath16 by a factor @xmath350 .
the theory well agrees with the simulation results
. we have also found the mechanisms responsible for the spike wings ( figs .
the theory well fits the simulations ( fig .
the asymptotic shape of the spike is _ universal _ : it is described by eq .
( 67 ) ( fig .
the facilitation of the global chaos onset may have the following _ applications _ in particular : * drastic increase of the _ dc - conductivity _ of a 2d electron gas in a 1d magnetic superlattice ; * significant decrease of the _ activation barrier _ for noise - induced escape over double / multi - barrier structures , that may lead to a drastic acceleration of the _ diffusion _ in periodic structures ; * strong facilitation of the _ stochastic web _ formation .
the work was partly supported by intas grants 97 - 574 and 00 - 00867 , by the convention between the institute of semiconductor physics and pisa university for 2005 - 2007 , and by the ictp program for short - term visits .
we are grateful to d. escande , j. meiss and g. zaslavsky for discussions .
sms and rm acknowledge the hospitality of pisa university and institute of semiconductor physics respectively .
the chaotic layers of the system ( 2 ) associated with the separatrices of the unperturbed system ( 1 ) are described here by means of the separatrix map . to derive the map ,
we follow the method described in @xcite , but the analysis of the map significantly differs from existing ones @xcite . it constitutes the method allowing to calculate chaotic layer boundaries in the _ phase space _ ( rather than just in energy ) .
it also allows to quantitatively describe the _ transport _ within the layer in a manner different from existing ones ( cf . @xcite and references therein ) .
the typical function @xmath351 for the trajectory close to the inner separatrix ( the separatrix corresponding to the lower potential barrier ) is shown in fig .
one can resolve pulses in @xmath351 .
each of them consists of two approximately antisymmetric spikes @xcite .
the pulses are separated by intervals during which @xmath352 is relatively small . generally speaking ,
successive intervals differ of each other .
let us introduce the pair of variables @xmath353 and @xmath354 : the energy @xmath10 changes only during the pulses of @xmath351 and remains nearly unchanged during the intervals between the pulses , when @xmath357 is small @xcite .
we assign numbers @xmath358 to the pulses and introduce the sequences of @xmath359 corresponding to initial instants of pulses @xmath360 .
in such a way , we obtain the following map ( cf .
@xcite ) : \1 . let us denote with @xmath365 the instant within the @xmath363th pulse when @xmath366 is equal to zero ( fig .
the function @xmath367 is an odd function within the @xmath363th pulse and it is convenient to transform the cosine in the integrand in @xmath364 ( a2 ) as \2 .
each pulse of @xmath366 contains one positive and one negative spike .
the first spike can be either positive or negative . if @xmath10 changes during the given @xmath36th pulse so that its value at the end of the pulse is _ smaller _ than @xmath11 , then the first spikes of the @xmath363th and @xmath371st pulses have the _ same _ signs .
on the contrary , if @xmath10 at the end of the @xmath363th pulse is _ larger _ than @xmath11 , then the first spikes of the @xmath363th and @xmath371st pulses have _ opposite _ signs .
note that fig .
14 corresponds to the case when the energy remains above @xmath11 during the whole interval shown in the figure .
this obviously affects the sign of @xmath364 , and it may be explicitly accounted for in the map if we introduce a new discrete variable @xmath372 which characterizes the sign of @xmath366 at the beginning of a given @xmath363th pulse , the map similar to ( a5 ) was introduced for the first time in @xcite , and it is often called as the zaslavsky separatrix map . its mathematically rigorous derivation may be found e.g. in the recent major mathematical review @xcite .
the latter review describes also generalizations of the zaslavsky map as well as other types of separatrix maps .
the analysis presented below relates immediately to the zaslavsky map but it is hoped to be possible to generalize it for other types of the separatrix maps too .
the variable @xmath164 introduced in ( a5 ) will be convenient for the further calculations since it does not depend on @xmath363 in the lowest - order approximation .
a quantity like @xmath376 is sometimes called the _ separatrix split _ @xcite since it is conventionally assumed that the maximal deviation of energy on the chaotic trajectory from the separatrix energy is of the order of @xmath170 @xcite .
though we shall also use this term , we emphasize that the maximal deviation may be much larger .
dynamical chaos appears in the separatrix map ( a5 ) because @xmath377 .
various heuristic criteria were suggested for the estimate of the chaotic layer width in energy @xcite .
frequencies relevant to our problem are much smaller than the reciprocal width of the spikes of @xmath351 . for such frequencies ,
all these criteria @xcite give : the estimate ( a6 ) was used in our earlier theory @xcite .
but we found later that , for the case of small @xmath15 , the aforementioned criteria were _ insufficient _ , so that the estimate ( a6 ) was incorrect @xcite ( cf . also ) .
moreover , to search a uniform width of the layer is incorrect in cases like ours , where the width strongly depends on the angle . at the same time ,
the _ lowest - order _
formulas for the spike minimum @xmath221 are not affected by this , so that the results of @xcite ( with only the lowest - order formulas ) are correct .
still , the higher - order corrections ( quite significant for @xmath56 if @xmath61 is moderately small ) would be incorrect if they were calculated on the basis of the estimate ( a6 ) .
besides , the paper @xcite did not address the intriguing question : why does even a small excess of @xmath14 over @xmath0 result in the onset of chaos in a large part of the phase space between the separatrices , despite the fact that the width of the chaotic layers associated with the nonlinear resonances is _ exponentially small _ for @xmath379 ?
the analysis of the separatrix map presented below resolves these important problems . in the adiabatic limit @xmath380
, the excess of the upper boundary @xmath381 of the lower layer over the lower barrier @xmath11 does not depend on angle and equals @xmath382 @xcite ( cf . also ) . but @xmath15 relevant for the spike of @xmath0 can not be considered as an adiabatic frequency , despite its smallness , because it is close to @xmath16 or to its multiple while all energies at the boundary lie in the range where the eigenfrequency is also close to @xmath16 : let us explicitly evaluate @xmath164 .
given that the energy is close to @xmath11 , the velocity @xmath367 in @xmath164 ( a5 ) may be replaced by the corresponding velocity along the separatrix associated with the lower barrier , @xmath384 , while the upper limit in the integral may be replaced by infinity . besides , in the asymptotic limit @xmath57 , the interval between spikes within the pulse becomes infinitely long @xcite and , therefore , only short ( @xmath385 ) intervals corresponding to the spikes contribute to the integral in @xmath164 ( a5 ) . in the scale @xmath386
, they may be considered just as two instants : in the derivation of the first equality in ( a9 ) , we have also taken into account that the function @xmath389 is odd . in the derivation of the second equality in ( a9 ) , we have taken into account that the right turning point of the relevant separatrix is the top of the lower barrier and the distance between this point and the left turning point of the separatrix approaches @xmath127 in the limit @xmath57 . where the instant @xmath393 corresponds to the turning point of the separatrix to the left from the lower barrier , i.e. @xmath394 while @xmath395 for all @xmath396 .
the dependence @xmath397 by eq .
( a11 ) is shown for @xmath161 in fig .
for small frequencies , the asymptotic formula ( a9 ) well fits the formula ( a11 ) .
consider the _ dynamics _ of the map ( a5 ) , when @xmath15 is close to the spikes minima : @xmath398 where @xmath257 while @xmath399 .
let the energy at the step @xmath400 be equal to @xmath11 .
the trajectory passing through the state with this energy is chaotic since @xmath401 diverges as @xmath402 and , therefore , the angle @xmath403 is not correlated with the angle on the previous step @xmath404 .
the quantity @xmath405 is not correlated with @xmath406 either .
thus , @xmath407 may take any value in the range @xmath408 $ ] and @xmath405 may equally take the values 1 or -1 .
therefore , the energy may change on the next step by an arbitrary value in the interval @xmath409 $ ] .
thus , @xmath410 may have a positive value @xmath411 @xcite .
then , the approximate equality @xmath412 holds , provided the value of @xmath14 is from the relevant range .
allowing for this and recalling that we are interested only in those realizations of the map such that @xmath413 , the relevant realization of the map @xmath414 may be written as : one may expect that the further evolution of the map will approximately follow , for some time , the trajectory of the system ( 4 ) with the initial energy @xmath416 ( a12 ) , and with an arbitrary @xmath403 and initial slow angle @xmath120 somehow related to @xmath417 .
let us prove this explicitly .
consider two subsequent iterations of the map ( a5 ) : @xmath418 and @xmath419 with an arbitrary @xmath420 . while doing this , we shall assume the validity of ( a7 ) ( it will be clarified below when this is true ) from which it follows that : ( i ) @xmath421 , ( ii ) @xmath422 .
it will follow from the results that the neglected corrections are small in comparison with the characteristic scales of the variation of @xmath10 and @xmath423 ( cf .
the conventional treatment of the nonlinear resonance dynamics @xcite ) . besides
, it follows from ( a5 ) that , while the energy remains above the barrier energy , @xmath424 oscillates , so that @xmath425 and @xmath426 .
then , let us ( i ) use for @xmath164 the asymptotic formula ( a10 ) , ( ii ) take into account that the increase of @xmath363 by 1 corresponds to the increase of time by @xmath436 , and ( iii ) transform from the variables @xmath437 to the variables @xmath438 .
equations ( a15 ) reduce then to : @xmath439}\sin(\tilde{\psi } ) , \nonumber \\ & & \frac{{\rm d}\tilde{\psi}}{{\rm d}t}= n\omega-\omega_f , \\ & & \tilde{\psi}\equiv n\pi\frac{1-\sigma_0}{2}-\tilde{\varphi } , \quad\quad n\equiv 2j-1 .
\nonumber\end{aligned}\ ] ] equations ( a16 ) are identical to the equations of motion of the system ( 4 ) in the lowest - order approximation , i.e. to the equations ( 14 ) where @xmath101 is replaced by its asymptotic value ( 12 ) and the last term in the right - hand part of the second equation is neglected , being of higher order in comparison with the term @xmath440 .
apart from the formal identity of eqs .
( a16 ) and ( 14 ) , @xmath120 in ( a16 ) and @xmath120 in ( 14 ) are identical to each other .
necessarily @xmath365 corresponds to a turning point ( see fig .
14 ) while the corresponding @xmath64 is equal to @xmath441 or @xmath442 for the right and left turning point respectively ( see ( 4 ) ) i.e. @xmath443 , so that @xmath444 .
we name the quantity @xmath449 the _ generalized separatrix split _ ( gss ) for the lower layer . unlike the conventional separatrix split @xmath170 @xcite ,
it is _ angle - dependent_. the curve @xmath450 may be called then the gss curve for the lower barrier and denoted as @xmath451 .
finally , let us investigate an important issue : whether the transformation from the discrete separatrix map ( i.e. ( a13 ) and ( a14 ) ) to the differential equations ( a15 ) is valid for the very first step and , if it is so , for how long it is valid after that . the transformation is valid as long as @xmath452 i.e. as long as @xmath453 is not too close to the barrier energy @xmath11 . at the step @xmath454
, the system stays at the gss curve , with a given ( random ) angle @xmath455 from the range ( a18 ) .
thus , at this stage , the relation ( a7 ) is certainly valid ( for the relevant range of @xmath14 and for an angle a little away from the narrow vicinity of the multiples of @xmath127 ) .
hence , ( a7 ) is valid at the step @xmath457 too .
similarly , one can show that @xmath458 , etc .
thus , the transformation ( a13,a14)@xmath459(a15 ) is valid at this initial stage indeed , and the evolution of @xmath437 does reduce to the resonant trajectory ( 14 ) with an initial angle from the range ( a18 ) and the initial action ( a19 ) .
this lasts until the resonant trajectory meets the gss curve in the adjacent @xmath127 range of @xmath120 i.e. at @xmath460 such that the state @xmath461 satisfies the conditions : at this instant , the absolute value of the change of energy @xmath453 in the separatrix map ( a13 ) is equal to @xmath463 ( just because the state belongs to the gss curve ) but the sign of this change is negative since the sign of @xmath464 is opposite to the sign of @xmath465 .
therefore , at the step @xmath466 , the system gets to the very separatrix , and the regular - like evolution stops : at the next step of the map , the system may either again get to the gss curve with a new ( random ) angle from the range ( a18 ) and start a new regular - like evolution as described above ; or it may get to the similar gss curve _ below _ the barrier and start an analogous regular - like evolution in the energy range below the barrier , until it stops in the same manner as described above , etc .
this approach makes it possible to describe all features of the transport within the chaotic layer .
their detailed description will be done elsewhere while , in the present context , it is most important to describe the _ upper outer boundary _ of the layer .
it follows from the analysis carried out in the previous subsection that _ any _ state ( in the @xmath467 plane ) lying beyond the gss curve but belonging to any trajectory following the equations ( 14 ) which possesses common points with the gss curve belongs to the chaotic layer : the system starting from such a state will get , sooner or later , to the separatrix , where the chaotization will necessarily occur .
therefore , the _ upper boundary _ of the chaotic layer coincides with the trajectory following the equations ( 14 ) with the initial action ( a19 ) and an initial angle @xmath468 from the range ( a18 ) such that the trajectory deviates from the barrier energy more than a trajectory ( 14)-(a18)-(a19 ) with any other initial angle does
. there may be only two topologically different options for such a trajectory : either it is the trajectory _
tangent _ to the gss curve , or it is the separatrix trajectory which _ intersects _ the gss curve ( some schematic examples are shown in figs .
16(a ) and 16(b ) respectively ; some real calculations are shown in figs .
8 and 9 ) .
consider first values of @xmath14 which are large enough for the condition ( a7 ) to be satisfied ( the explicit criterion will be given in ( a31 ) ) but which are smaller than the value @xmath469 determined by eq .
( 41 ) ( its meaning is explained below ) .
the further analysis within this range of @xmath14 differs for the ranges of @xmath15 relevant to _ odd _ and _ even _ spikes , and we consider them separately .
let us seek the state @xmath471 ( with @xmath169 within the range @xmath1110,\pi[$ ] ) where the resonant trajectory curve is _ tangent _ to the gss curve . with this aim , we equal both the actions and the derivatives of both curves . the equality of actions immediately yields @xmath472 via @xmath169 : @xmath473 .
the derivative along the gss curve is obtained by differentiation of @xmath474 .
the derivative along a resonant trajectory can be found dividing the first dynamic equation in ( 14 ) by the second one . substituting the expression of @xmath472 via @xmath169 into the equality of the derivatives ,
we obtain a closed equation for @xmath169 , and its solution immediately gives us the relevant @xmath455 : @xmath475_{e = e_t^{(l)}}=0 , \\ & & e_t^{(l)}\equiv e_b^{(1)}+h|\epsilon^{(low)}|\sin(\tilde{\psi}_{t}^{(l ) } ) , \nonumber\\ & & \tilde{\psi}_{t}^{(l)}\in [ 0,\pi ] , \quad\quad n\equiv 2j-1 , \quad\quad j=1,3,5,\ldots , \nonumber \\ & & \tilde{\psi}(0)=\tilde{\psi}_{t}^{(l)}. \nonumber\end{aligned}\ ] ] a careful analysis of the phase space structure shows that , in the present case ( i.e. when @xmath476 while @xmath254 is odd ) , there is no separatrix of the resonant hamiltonian ( 4 ) which would both intersect the gss curve and possess points above the tangent trajectory @xcite .
thus , for this range of @xmath14 , the outer boundary of the chaotic layer is formed by the trajectory following the dynamical equations ( 14 ) with the initial angle by ( a22 ) and initial action ( a19 ) ( fig .
16(a ) ) . let us find the lowest - order solution of eq .
we neglect the term @xmath477 ( the result will confirm the validity of this ) and use the lowest - order expression for the relevant quantities : namely , eqs . ( a10 ) and
( 12 ) for @xmath164 and @xmath101 respectively , and the lowest - order expression for @xmath478 which can be derived from eq . ( 11 ) : the next step is to find the _
minimal _ value of the energy on the boundary of the layer , @xmath174 .
it follows from the analysis of the dynamical equations ( 14 ) that the corresponding angle @xmath484 is equal to @xmath159 if @xmath485 ( i.e. @xmath486 ) or to @xmath127 if @xmath487 ( i.e. @xmath488 ) : cf .
given that the hamiltonian ( 4 ) is constant along any trajectory ( 14 ) while the boundary coincides with one of such trajectories , the values of the hamiltonian ( 4 ) in the states @xmath489 and @xmath490 should be equal to each other . in the explicit form , this equality may be written as let us find the lowest - order solution of eq .
assume that @xmath174 still belongs to the range of @xmath10 where @xmath58 ( the result will confirm this assumption ) .
then the integrand in ( a27 ) goes to zero in the asymptotic limit @xmath57 and , hence , the integral may be neglected ( the result will confirm the validity of this ) .
the remaining terms in eq .
( a27 ) should be treated very carefully . in particular , it is insufficient to use the lowest - order value ( 12 ) for @xmath101 since it is the difference between @xmath492 and @xmath493 that matters .
moreover , the approximate equality @xmath494 does not apply here either since , as it follows from eq .
( a23 ) , the derivative @xmath495 may strongly vary in the range @xmath496 $ ] if @xmath497 ( the result will show that it is the case ) . that is why it is necessary to use for @xmath101 the more accurate expression ( 11 ) . allowing for the asymptotic expression ( a25 ) of @xmath169 and keeping only the lowest - order terms
, one can finally reduce eq .
( a27 ) to the relation it is necessary and sufficient that the condition @xmath500 is satisfied at the _ minimal _ and _ maximal _ energies of the boundary to ensure that the second equality in ( a7 ) holds true , i.e. that @xmath7 is close to @xmath16 for _ all _ points of the boundary . in this case ,
@xmath151 and @xmath495 have different signs for all @xmath10 within the relevant range ( i.e. where @xmath506 , @xmath507 ) : cf .
( 12 ) and ( a23 ) .
then , in the asymptotic limit @xmath57 , eq . ( a22 ) for the tangency does not have any solution for @xmath169 in the relevant range @xcite . there may be only solutions very close to some of @xmath127 integers , and the corresponding energies @xmath172 are very close to @xmath11 i.e. @xmath508 : therefore they are irrelevant . while the energy ( which may be found as the appropriate solution of eq .
( 15 ) ) lies in the relevant vicinity of the lower barrier ( fig .
16(b ) ) . in the lowest - order approximation
, this saddle energy reads : this saddle ( denoted in fig .
16(b ) as @xmath279 ) generates a separatrix .
its upper whiskers go to the similar adjacent saddles ( shifted in @xmath120 by @xmath119 ) . in the asymptotic limit @xmath57 , the upper whiskers are much steeper than the gss curve and hence they do not intersect it @xcite . as concerns the lower whiskers , they do intersect the gss curve and , moreover , two intersections lie in the relevant energy range ( fig .
16(b ) ) . let us show this explicitly .
let us write the expression for the hamiltonian ( 4 ) in the relevant vicinity of the barrier energy ( i.e. where @xmath511 , keeping , in the expression , both the lowest - order terms and the terms of next order ( in particular , we use eq .
( 11 ) for @xmath151 and take into account that @xmath512 for the relevant range of @xmath10 ) : @xmath513}h\sqrt{2}\left(1+\frac{n\pi\ln \left(\frac{2\phi}{\delta}\right)}{8\ln \left(\frac{4{\rm e}}{\phi}\right)}\right)\cos(\tilde{\psi}),\\ & & \omega_m-\omega(e+\delta)\ll\omega_m .
\nonumber\end{aligned}\ ] ] the hamiltonian @xmath153 should possess equal values at the saddle @xmath279 and at the intersections of the separatrix and the gss curve . let us denote the angle of the intersection in the range @xmath1110,\pi[$ ] as @xmath514 , and let us denote the deviation of its energy @xmath515 from @xmath11 as @xmath516 .
assuming that @xmath517 ( the result will confirm this ) so that @xmath518}(1-(\tilde{\psi}_i^{(l)}-\tilde{\psi}_s^{(l)})^2/2)\approx ( -1)^{[n/4]}(1-(\delta_i^{(l)}/\delta_l)^2/2)\approx ( -1)^{[n/4]}(1-(\delta_i^{(l)}/h)^2/4)$ ] , the equality of the values of @xmath153 reads as : let us assume that , in the asymptotic limit @xmath57 , @xmath520 ( the result will confirm this ) .
then the left - hand part is asymptotically smaller than the first term in the right - hand part .
so , eq . ( a36 ) implies , in the asymptotic limit , that the right - hand side equals zero . expressing @xmath14 via @xmath521 from eq .
( a34 ) , we finally obtain a closed transcendental equation for @xmath522 : thus , we have proved the following asymptotic properties of the separatrix generated by the saddle @xmath279 : 1 ) the lower whiskers of the separatrix do intersect the gss curve in the relevant range of @xmath10 ( i.e. where the resonant approximation is valid ) , 2 ) the upper whiskers of the separatrix do _ not _ intersect the gss curve ( there is no solution of eq .
( a36 ) in the range @xmath527 ) .
the former property confirms the self - consistence of the asymptotic theory for even spikes ; the latter property means that the _ upper outer boundary _ of the lower chaotic layer is formed by the _
upper whiskers of the separatrix generated by the saddle @xmath279_. as @xmath14 grows , the boundary of the layer raises up while the lower part of the resonance separatrix , on the contrary , goes down .
they reconnect at the critical value of @xmath14 , @xmath529 , determined by eq .
( 41 ) , that may be considered as the absorption of the resonance by the chaotic layer .
if @xmath14 grows further , then the gss curve and the resonance separatrix intersect . as a result , the trajectory starting from the state with the angle ( a22 ) and action ( a19 ) , for odd spikes , or from the saddle @xmath279 , for even spikes , is _ encompassed _ by the resonance separatrix .
so , it does not form the outer boundary of the layer anymore .
rather it forms the inner boundary i.e. the boundary of the main island of the stability inside the layer , repeated periodically in @xmath120 with a period @xmath119 ( cf .
analogous islands in the upper layer in fig . 8) . unless the lower chaotic layer reconnects with the upper one , the _ outer _ boundary of the lower layer is formed by the upper part of the _ resonance separatrix_. the relevant initial angle @xmath455 on the gss curve corresponds to the intersection of the gss curve with the resonance separatrix ( cf .
the analogous situation for the upper layer in fig . 8) .
similarly to the lower - layer case , one may consider the ranges of relatively small @xmath14 ( namely , smaller than @xmath530 determined by eq .
( 42 ) ) and relatively large @xmath14 ( i.e. @xmath531 ) . in the former range ,
the formation of the boundary occurs in a manner which is , in a sense , opposite to that for the lower - layer case . for even spikes ,
the lower outer boundary is formed by _ tangency _ while , for odd spikes , it is formed by the lower part of the _ separatrix _ generated by the saddle @xmath278 , analogous to the saddle @xmath279 in the lower - layer case @xcite .
@xmath532_{e = e_t^{(u)}}=0 , \\ & & e_t^{(u)}\equiv e_b^{(2)}-h|\epsilon^{(up)}|\sin(\tilde{\psi}_{t}^{(u)})\nonumber \\ & & \tilde{\psi}_{t}^{(u)}\in\left[0,\pi \right ] , \quad\quad n\equiv 2j-1 , \quad\quad j=2,4,6,\ldots , \nonumber \\ & & \tilde{\psi}(0)=\tilde{\psi}_{t}^{(u)},\nonumber\end{aligned}\ ] ] where @xmath535 is the time dependence of the velocity along the separatrix associated with the upper barrier and the instant @xmath393 is chosen so that @xmath536 is equal to the coordinate of the lower barrier while @xmath537 for @xmath538 . the dependence @xmath539 in eq .
( a43 ) is shown for @xmath161 in fig .
15(b ) .
the lowest - order solution of ( a41 ) is given in eq .
( 37 ) , so that @xmath543 is approximated by eq .
. the maximal energy on the lower boundary of the layer corresponds to @xmath544 if @xmath545 or @xmath159 if @xmath546 and is determined by eq .
the asymptotic value of the minimal deviation from the upper barrier of the energy at the boundary , @xmath198 , is given in eq .
( 40 ) . for odd spikes ,
the boundary is formed by the lower part of the separatrix generated by the saddle @xmath278 .
the angle of the saddle is given in eq .
( 33 ) while the deviation of its energy from the barrier is approximated , to the lowest - order approximation , by eq .
( 34 ) .
as @xmath14 grows , the boundary of the layer goes down while the upper part of the upper resonance separatrix goes up .
they reconnect at @xmath547 , as determined by eq .
( 42 ) , that may be considered as the absorption of the resonance by the chaotic layer . for larger @xmath14
, the boundary of the layer is formed by the lower part of the upper resonance separatrix ( fig . 8) , unless the latter intersects the lower gss curve ( in the latter case , @xmath548 marks the global chaos onset ) .
howard and s.m .
hohs , phys .
rev . * a 29 * , 418 ( 1984 ) ; d. del - castillo - negrete , j.m .
greene , p.j .
morrison , physica * d 61 * , 1 ( 1996 ) ; h. r. dullin , j. d. meiss , d. sterling , nonlinearity * 13 * , 203 ( 2000 ) ; a.d .
morozov , chaos * 12 * , 539 ( 2002 ) .
soskin , o.m .
yevtushenko , and r. mannella , in _ noise in complex systems and stochastic dynamics ii _ , eds . z. gingl , j.m .
sancho , l. schimansky - geier , j. kertesz , proceedings of spie vol .
5471 ( spie , bellingham , wa , 2004 ) p. 355 .
this value of @xmath14 is so close to the critical value @xmath33 ( the latter is about 0.0049 , as seen from fig .
4 ) that the blue and green layers nearly touch each other .
that is why we did not draw a regular trajectory separating the layers .
we still use different colours for the trajectories starting from @xmath549 and @xmath550 in order to demonstrate that their mixing occurs very slowly , which indicates that @xmath14 is only slightly above the critical @xmath33 . for sufficiently small @xmath61 and @xmath14 , the separatrix generated by the unstable point forms the boundary of the upper chaotic layer , but this affects only the higher - order terms in the expressions for the spikes minima ( see below ) .
( 20 ) is the _ sufficient _ ( rather than only necessary ) condition for the separatrix reconnection since there is no any other separatrix which would lie in between the separatrices generated by the saddles and . with account for the next - order corrections
, the spike minimum @xmath221 coincides with @xmath219 in case of an odd spike or with @xmath220 in case of an even spike .
this occurs because , in case of odd spikes , @xmath553 increases / decreases as @xmath10 approaches the relevant vicinity of the upper / lower barrier , while it is vice versa in case of even spikes . and
the larger @xmath112 the farer the resonance separatrix extends : in other words , the reconnection of the barrier chaotic layer with the resonance separatrix requires a smaller value of @xmath14 at the barrier where @xmath112 , in the relevant vicinity of the barrier , is larger . for a given @xmath55 ,
the numerical error on @xmath33 is rather large since the corresponding simulation involves a very long integration time .
we estimate the relative numerical error of @xmath554 to be @xmath555 .
note however that this error practically does not affect the measured values of the frequencies @xmath60 in the minima of the spikes : the error in the measurement of @xmath33 is systematic while the spikes are very sharp .
we estimate the error of @xmath556 to be @xmath557 .
still , this procedure does not pretend to be exaustive since it takes into account only corrections within the description by means of the resonant hamiltonian ( 4 ) , while ignoring a correction to the resonant hamiltonian ( 4 ) itself .
a derivation of the latter correction requires the use of the second - order averaging method @xcite , that goes beyond the scope of the present paper . on the other hand
, even our incomplete procedure gives a good agreement with the simulations .
it means that the corrections coming from the second - order averaging are probably insignificant in comparison with the corrections of other origins .
obviously , this motion is necessarily three - dimensional .
however , if the direction of the magnetic field is perpendicular to the direction of its periodicity , then the motion in the plane perpendicular to the magnetic filed is _ separated _ from the motion along the field .
thus , the in - plane dynamics is two - dimensional while being reduced in the relevant case to the one - dimensional dynamics ( 2)-(1 ) ( cf .
@xcite ) .
spikes correspond to the motion over any of the minima of the potential , first in one and then ( after the reflection from one of the upper barriers ) in the opposite direction .
if @xmath61 is small , the spikes within the pulse are separated by a long interval since the reflection point is situated close to the top of the upper barrier , where the motion is slow .
the latter is valid for any @xmath403 except in the close vicinity of multiples of @xmath127 while the state @xmath558 ( a5 ) in the latter range of @xmath403 turns out irrelevant to the boundary , as shown further down .
for odd numbers @xmath559 , there are separatrices which lie in the range of @xmath10 where @xmath560 i.e. much closer to the barrier than the tangent trajectory : these separatrices are therefore irrelevant . in case of a _ moderately _ small @xmath61 , they may intersect the gss curve . and then , the tangent trajectory lying above the separatrix necessarily exists , so that the boundary of the layer is formed by this tangent trajectory . for any
ac - driven spatially periodic hamiltonian system , the _ upper _ energy boundary of the layer associated with the unbounded separatrix diverges in the adiabatic limit @xmath380 @xcite .
however , this divergence is not relevant for the present problem for the following reasons .
the lower chaotic layer relates to the _ bounded _ separatrix while , for the upper ( unbounded ) layer , it is the _ lower _ boundary of the layer which is relevant for the onset of global chaos in between the separatrices .
moreover , even for the upper boundary of the upper layer , the divergence is not yet manifested for the driving parameters @xmath561 in the vicinity of the spikes minima ( cf . | we have developed the _ general method _ for the description of _ separatrix chaos _ , basing on the analysis of the separatrix map dynamics . matching it with the resonant hamiltonian analysis ,
we show that , for a given amplitude of perturbation , the maximum width of the chaotic layer in energy may be much larger than it was assumed before .
we apply the above theory to explain the drastic facilitation of global chaos onset in time - periodically perturbed hamiltonian systems possessing two or more separatrices , previously discovered ( prl 90 , 174101 ( 2003 ) ) .
the theory well agrees with simulations .
we also discuss generalizations and applications .
examples of applications of the facilitation include : the increase of the dc conductivity in spatially periodic structures , the reduction of activation barriers for noise - induced transitions and the related acceleration of spatial diffusion , the facilitation of the stochastic web formation in a wave - driven or kicked oscillator . |
one of the defining characteristics of active galactic nuclei ( agn ) is that their x - ray emission is variable .
x - ray light curves from seyfert 1 galaxies show unpredictable and seemingly aperiodic variability ( lawrence et al . 1987 ; m@xmath2hardy 1989 ) .
such random variability is often referred to as _ noise _ , meaning that it is the result of a stochastic , as opposed to deterministic , process . in this context
the ` noise ' is intrinsic to the source and not a result of measurement errors ( such as poisson noise ) , i.e. the signal itself is the output of a noise process .
one of the most common tools for examining agn variability ( and noise processes in general ) is the fluctuation power spectral density ( psd ) which represents the amount of variability power ( mean of the squared amplitude ) as a function of temporal frequency ( timescale@xmath3 ) .
the high frequency psds of seyferts are usually well - represented by power - laws over a broad range of frequencies ( @xmath4 , where @xmath5 is the power at frequency @xmath6 ) with slopes @xmath7 ( green , m@xmath2hardy & lehto , 1993 ; lawrence & papadakis , 1993 ; edelson & nandra , 1999 ; uttley , m@xmath2hardy & papadakis , 2002 ; vaughan , fabian & nandra 2003 ; markowitz et al .
such a spectrum , with a slope @xmath8 is usually called ` red noise ' ( for an introduction to red noise see press 1978 ) .
if seyfert 1 light curves are viewed as the product of a stochastic ( in this case red noise ) process then the specific details of each individual light curve provide little physical insight .
each light curve is only one _ realisation _ of the underlying stochastic process , i.e. it is one of the _ ensemble _ of random light curves that might be generated by the process .
each new realisation will look different and these changes are simply statistical fluctuations inherent in any stochastic process ( as opposed to changes in the nature of the process itself ) .
therefore one should expect two light curves to have different characteristics ( such as mean and variance ) even if they are realisations of the same process . on the other hand , data from deterministic processes , for example the energy spectrum of a non - varying source or the light curve of a strictly periodic source ( such as a pulsar ) , should be repeatable within the limits set by the measurement errors .
it is the average properties of the variability ( such as the psd ) that often provide most insight into the driving process .
for example , the psd of any real red noise process can not continue as a steep power - law indefinitely to longer timescales or the integrated variability power would diverge .
therefore the psds of agn variability must break to a flatter index at low frequencies ; the position of such a break would represent a characteristic variability timescale and may yield information about the underlying driving process .
recent timing studies have indeed found evidence that the steep power - law psds of seyfert 1s show a flattening , or turnover , at low frequencies ( edelson & nandra 1999 ; uttley et al .
2002 ; markowitz et al .
2003 ) . in many cases however the data are not adequate for psd analysis . in these situations
the variability is usually described in terms of the statistical moments ( e.g. the sample mean and variance , etc . ) . however , due to the stochastic nature of red noise variability there is a large degree of randomness associated with these quantities . in practice
this means that it is difficult to assign meaningful errors to the variance .
this in turn makes it difficult to quantitatively compare variances , say from repeated observations of the same source ( and thereby test whether the variability is _ stationary _ ) .
such an analysis might be desirable ; it could in principle reveal changes in the ` state ' of the source if its variability properties were found to evolve with time .
this paper discusses this and related problems that are encountered when examining the variability properties of agn .
particular emphasis is placed on the mathematical properties and implications of the inherent randomness in the variability .
the mathematical details are well understood from the general theory of stochastic processes ( e.g. priestley 1981 for spectral analysis ) but some of the practical consequences for agn observations have not been discussed in detail . on the basis of simulated data some recipes are developed that may serve as a useful guide for observers wishing make quantitative use of their variability analysis without the recourse to e.g. extensive monte carlo simulations .
the paper is organised as follows .
section [ sect : stats ] defines the estimators to be discussed ( namely the periodogram and the variance ) .
simulated data are used to illustrate various aspects of these estimators ; section [ sect : simulations ] describes methods for producing artificial red noise time series .
section [ sect : stationarity ] discusses the stationarity of time series .
sections [ sect : errors_intrinsic ] and [ sect : errors_poisson ] discuss two sources of uncertainty associated with measuring variability amplitudes , the first due to the stochastic nature of the variability and the second due to flux measurement errors .
section [ sect : mrk766 ] gives an example using a real _ xmm - newton_observation of .
finally , a brief discussion of these results is given in section [ sect : disco ] and the conclusions are summarised in section [ sect : conc ] .
the psd defines the amount of variability ` power ' as a function of temporal frequency .
it is estimated by calculating the _ periodogram _ , which is an estimator of the continuous psd @xmath5 .
the periodogram is therefore specific to each realisation of the process , whereas the psd is representative of the true , underlying process . ]
( priestley 1981 ; bloomfield 2000 ) . for an evenly sampled light curve ( with a sampling period @xmath9 )
the periodogram is the modulus - squared of the discrete fourier transform ( dft ) of the data ( press et al .
1996 ) . for a light curve comprising a series of fluxes @xmath10 measured at discrete times @xmath11 : @xmath12 at @xmath13 evenly spaced frequencies @xmath14 ( where @xmath15 ) , @xmath16 is the nyquist frequency , @xmath17 .
note that it is customary to subtract the mean flux from the light curve before calculating the dft .
this eliminates the zero - frequency power .
the periodogram , @xmath18 , is then calculated by choosing an appropriate normalisation @xmath19 ( see appendix [ sect : pds_norm ] for more on periodogram normalisations ) .
for example @xmath20 if the time series is a photon counting signal such as normally encountered in x - ray astronomy , and is binned into intervals of @xmath9 , the effect of poisson noise is to add an approximately constant amount of power to the periodogram at all frequencies . with the above normalisation this constant poisson noise level is @xmath21 ( assuming the light curve is not background subtracted ) .
the periodogram of a noise process , if measured from a single time series , shows a great deal of scatter around the underlying psd .
in particular , the periodogram at a given frequency [ @xmath22 is scattered around the psd [ @xmath5 ] following a @xmath24 distribution with two degrees of freedom ( van der klis 1989 ) : @xmath25 where @xmath26 is a random variable distributed as @xmath24 with two degrees of freedom , i.e. an exponential distribution with a mean and variance of two and four , respectively .
the periodogram is distributed in this way because the real and imaginary parts of the dft are normally distributed for a stochastic process is even so the periodogram at this frequency is distributed as @xmath27 , i.e. with one degree of freedom . ]
( section 6.2 of priestley , 1981 ; jenkins & watts 1968 ) .
the expectation value of the periodogram is equal to the psd but its standard deviation is 100 per cent , leading to the larger scatter in the periodogram ( see fig .
[ fig : two_curves ] ) . see leahy et al .
( 1983 ) , van der klis ( 1989 ) , papadakis & lawrence ( 1993 ) , timmer & knig ( 1995 ) and stella et al .
( 1997 ) for further discussion of this point .
when applied to real data the periodogram is an _
inconsistent _ estimator of the psd , meaning that the scatter in the periodogram does not decrease as the number of data points in the light curve increases ( jenkins & watts 1968 ) .
in order to reduce this scatter the periodogram must be smoothed ( averaged ) in some fashion . as the number of data points per bin increases ( either by binning over frequencies or averaging over many data segments ) the scatter in the binned periodogram decreases , i.e. the averaged periodogram is a _ consistent _ estimator of the psd ( see papadakis & lawrence 1993 and van der klis 1997 for more on binned periodogram estimates ) .
a further point is that periodograms measured from finite data tend to be biased by windowing effects which further complicate their interpretation ( van der klis 1989 ; papadakis & lawrence 1993 ; uttley et al . 2002 and see below ) .
the integral of the psd between two frequencies ( @xmath28 and @xmath29 ) yields the contribution to the expectation value of the ( ` true ' ) variance due to variations between the corresponding timescales ( @xmath30 and @xmath31 ) .
this result follows from parseval s theorem ( see e.g. van der klis 1989 ) @xmath32 correspondingly , for a discrete time series the integrated periodogram yields the observed variance for that particular realisation @xmath33 where @xmath34 is the frequency resolution of the dft ( @xmath35 ) . the total variance of a real light curve is equal to its periodogram integrated over the frequency range @xmath36 to @xmath37 .
the sample variance ( which will differ from observation to observation ) is given by : @xmath38 where @xmath39 is the arithmetic mean of @xmath40 .
in the limit of large @xmath41 these two variance estimates are identical .
the normalised variance is simply @xmath42 .
in order to elucidate the properties of the variance of red noise data , random light curves were generated from power - law psds similar to those of agn .
[ fig : two_curves ] shows two artificial time series and their periodograms .
it is worth reiterating that the large scatter in the periodograms is an intrinsic property of stochastic processes it does not depend on the number of data points and is not related to poisson noise in the data .
these artificial time series were produced using the algorithm of timmer & knig ( 1995 ) .
this generates random time series with arbitrary broad - band psd , correctly accounting for the intrinsic scatter in the powers ( i.e. equation [ eqn : pds_scatter ] ) .
other methods of generating random light curves include the related ` summing of sines ' method ( done et al .
note that it is not correct to randomise only the phases of the component sine functions , their amplitudes must also be randomised .
otherwise this method does not account for this intrinsic scatter in the powers .
shot - noise models can produce red noise time series with certain psd shapes ( see lehto 1989 ) .
there also exist various mathematical tricks for producing data with specific power - law psd slopes .
data with a @xmath43 psd ( often called ` flicker noise ' ) can be generated using the half - integral method outlined in press ( 1978 ) , while @xmath44 ( ` random walk ' ) data can be generated using a first - order autoregressive process ( ar[1 ] ) , essentially a running sum of gaussian deviates ( see deeming 1970 and scargle 1981 for more on such methods ) .
the method of timmer & knig ( 1995 ) is used below as this can generate time series from an arbitrary psd and is computationally efficient . some caution should be applied when using these routines to produce artificial time series . as mentioned briefly in the previous section ,
periodograms measured from real data tend to be biased by windowing effects . for uninterrupted but finite observations data of red noise processes the most important of these effects is ` red noise leak ' the transfer of power from low to high frequencies by the lobes of the window function ( see e.g. deeter & boynton 1982 ; van der klis 1997 ) .
if there is significant power at frequencies below the lowest frequency probed by the periodogram ( i.e. on timescales longer than the length of the observation ) this can give rise to slow rising or falling trends across the light curve .
these trends contribute to the variance of the light curve .
thus variability power ` leaks ' into the data from frequencies below the observed frequency band - pass .
the degree to which this occurs , and the resultant bias on the measured periodogram , depend on the shape of the underlying psd and the length of the observation ( papadakis & lawrence 1995 ; uttley et al .
for flat psd slopes ( @xmath45 ) the amount of leakage from low frequencies is usually negligible .
since agn light curves usually contain significant power on timescales longer than those probed ( see section [ sect : weak_non - stationarity ] ) the effects of red noise leak must be included in simulations of agn light curves .
this can be achieved by using the timmer & knig ( 1995 ) algorithm to generate a long light curve from a psd that extends to very low frequencies and then using a light curve segment of the required length .
data simulated in such a fashion will include power on timescales much longer than covered in the short segment . the effects of measurement errors ( e.g. poisson noise ) can be included in the simulations using standard techniques ( e.g. press et al .
a stationary process is one for which the statistical properties ( such as mean , variance , etc . )
do not depend on time .
[ fig : curve1 ] shows an artificial red noise time series together with its mean and variance measured every 20 data points .
the simulation was produced from a single , well - defined psd ( which did not vary ) .
the process is therefore stationary .
it would have been reasonable to expect the resulting time series ( a realisation of the process ) to appear stationary .
this is not the case however ; both the mean and variance change with time ( panels 2 and 3 ) .
this has nothing whatsoever to do with measurement errors - the simulation has zero errors .
this simulation demonstrates that , when dealing with red noise , fluctuations in variance are not sufficient to claim the variability process is non - stationary .
as the purpose of time series analysis is to gain insight into the process , not the details of any specific realisation , a more robust approach is needed to determine whether these data were produced by a time - stationary process or a non - stationary process .
it would be more insightful to consider whether the _ expectation values _ of the characteristics ( such as the variance ) are time - variable .
the expectation values should be representative of the properties of the underlying process , not just any one realisation .
see section 1.3 of bendat & piersol ( 1986 ) for a discussion of this point . for
a process with a steep red noise psd ( @xmath46 ) , the integrated periodogram will diverge as @xmath47 .
this means that ( following equation [ eqn : correspondence1 ] ) the variance of a red noise time series with a steep psd will diverge with time . in this case
there is no well - defined mean ; press & rybicki ( 1997 ) describe this form of variability as ` weakly non - stationary . ' for time series with power spectra flatter than this the variance converges as @xmath47 , thus for a white noise process with a flat psd ( @xmath48 ) , the variance will converge as the observation length increases , and there will be a well - defined mean on long timescales . of course
, for any real process the psd must eventually flatten such that the power does not diverge ( i.e. @xmath49 on sufficiently long timescales ) . thus , weak non - stationarity is entirely due to observations sampling only the steep part of the psd of a source . but in agn this flattening occurs on timescales much longer than those probed by typical observations . for instance , _ xmm - newton _ observations of agn typically last for @xmath50 s whereas in many objects the psd is steep until @xmath51 s and in some cases probably much longer ( edelson & nandra 1999 ; uttley et al . 2002 ; markowitz et al . 2003 ) .
therefore on the timescales relevant for most x - ray observations , agn light curves should be considered weakly non - stationary .
( top ) , @xmath52 ( middle ) and @xmath53 ( bottom ) .
each distribution is derived from 50,000 realisations .
as the psd gets steeper the distribution of variances becomes less gaussian and more like a @xmath24 distribution with a low effective degrees of freedom .
[ fig : fvar_dist1 ] , title="fig:",width=5 ] + ( top ) , @xmath52 ( middle ) and @xmath53 ( bottom ) .
each distribution is derived from 50,000 realisations .
as the psd gets steeper the distribution of variances becomes less gaussian and more like a @xmath24 distribution with a low effective degrees of freedom .
[ fig : fvar_dist1 ] , title="fig:",width=5 ] + ( top ) , @xmath52 ( middle ) and @xmath53 ( bottom ) .
each distribution is derived from 50,000 realisations .
as the psd gets steeper the distribution of variances becomes less gaussian and more like a @xmath24 distribution with a low effective degrees of freedom .
[ fig : fvar_dist1 ] , title="fig:",width=5 ] fluctuations in the statistical moments ( such as mean and variance ) of a light curve are intrinsic to red noise processes .
therefore , even in the absence of measurement errors ( e.g. no poisson noise ) the means and variances of two light curves produced by exactly the same process can be significantly different .
this can be seen in fig .
[ fig : curve1 ] ( panels 2 and 3 ) , where each 20 point segment of the light curve shows a different mean and variance .
these random fluctuations in variance are however governed by the normal statistical rules of noise processes and can thus be understood in a statistical sense .
any given series is only one realisation of the processes and its periodogram will show the scatter predicted by equation [ eqn : pds_scatter ] .
the integrated periodogram ( which gives the variance ; equation [ eqn : correspondence ] ) will therefore be randomly scattered around the true value for the psd of the process .
the variance in a specific time series is given by @xmath54 i.e. the variance of a given realisation is a sum of @xmath26 distributions weighted by the psd for even @xmath41 .
however , for large @xmath41 this will make a negligible difference to the sum ( cf .
equation 2.9 of van der klis , 1989 ) . ] .
( this assumes biases such as red noise leak are not significant .
if this is not true then these biases will further distort the distribution of variances . )
it is the expectation value of the variance that is representative of the integrated power in the psd , and thus the average amplitude of the variability process ( eqn .
[ eqn : correspondence1 ] ) .
thus , while the expectation value of the variance is equal to the integrated psd , each realisation ( time series ) of the same process will show a different variance even if the parent variability process is stationary .
this is particularly important for steep psd time series ( i.e. weakly non - stationary data ) since the variance does not converge as more data are collected .
only if the time series spans timescales on which the integrated power converges at low frequencies ( i.e. @xmath55 ) will the variance converge as the length of time series increases .
these points are illustrated by fig .
[ fig : fvar_dist1 ] , which shows the distributions of variances in random time series with three different psd slopes ( @xmath56 ) .
this plot was produced by generating 50,000 random time series ( each 100 points long ) for each psd slope and measuring the variance of each one .
the scatter in the variance is entirely due to random fluctuations between different realisations because the psd normalisation was kept fixed and no instrumental noise was added .
the shape of the distribution of variances can be seen to depend on the psd slope .
consider a white noise process ( @xmath48 , i.e. @xmath57 ) .
the periodogram of each realisation is randomly scattered around its flat psd . following equation [ eqn : sum_var ]
the variance is simply the sum of the @xmath13 @xmath26-distributed powers in the periodogram , and these are evenly weighted ( psd is constant ) .
the sum of @xmath13 @xmath26 distributions follows a @xmath58 distribution .
this tends to a normal distribution as @xmath41 increases .
thus the variance of a white noise process is approximately normally distributed ( as can be seen in fig .
[ fig : fvar_dist1 ] ) and converges as @xmath41 increases .
the fractional standard deviation of @xmath58 is given by @xmath59 , so for the 100 point light curves used the ( 1@xmath60 ) fractional width of the variance distribution is @xmath61 per cent , in agreement with the simulations ( fig .
[ fig : fvar_dist1 ] , top panel ) . for time series with steeper psds
the lower frequency periodogram points contribute more strongly to the sum than the higher frequency points .
the variance of such a time series is therefore dominated by a few low frequency powers and thus resembles a @xmath62 distribution with low ` effective degrees of freedom . '
the distribution of variances in red noise data is dependent on the underlying psd and is , in general , non - gaussian ( fig .
[ fig : fvar_dist1 ] ) .
the fractional standard deviation of @xmath62 is @xmath63 , which tends to unity as the psd gets steeper ( i.e. as effective @xmath64 ) .
thus the largest fluctuations in variance ( up to a limit of @xmath65 per cent rms ) are expected to result from very steep psd slopes .
as discussed above , when examining agn light curves one should expect random changes in the mean and variance with time ( between segments of a long observation or between observations taken at different epochs ) . this is true even if the measurement errors are zero and does not depend on the number of data points used ( due to the weak non - stationarity ) . however , it is also possible that the underlying process responsible for the variability itself changes with time ( e.g. the psd changes ) , in which case the variability is non - stationary in a more meaningful sense ` strongly non - stationary . '
such changes in the variability process could provide insight into the changing physical conditions in the nuclear regions . on the other hand the random changes expected for a red noise process yield no such physical insight .
the question thus arises : how does one tell , from a set of time series of the same source , whether they were produced by a strongly non - stationary process ?
in other words , is it possible to differentiate between differences in variance caused by real changes in the variability process ( physical changes in the system ) , and random fluctuations expected from red noise ( random nature of the process ) ?
if the process responsible for the variability observed in a given source is stationary then its psd is constant in time
. the expectation value of the absolute ( un - normalised ) variance will therefore be the same from epoch to epoch , but the individual variance estimates will fluctuate as discussed in section [ sect : stochasticity ] .
this makes it difficult to judge , from just the variances of two light curves taken at two epochs , whether they were produced by a stationary process .
given sufficient data it is , however , possible to test whether the expectation values of the variance ( estimated from an ensemble of light curves ) at two epochs are consistent with a stationary process . the methods most frequently employed involve comparing the psds ( estimated from the binned periodogram ) at different epochs .
if the psds show significant differences ( at a given confidence level ) the variability process can be said to be strongly non - stationary . as an example of this , the psds of x - ray binaries evolve with time , and the way in which the variability properties evolve provides a great deal of information on the detailed workings of these systems ( see e.g. belloni & hasinger 1990 ; uttley & m@xmath66hardy 2001 ; belloni , psaltis & van der klis 2002 ; pottschmidt et al .
2002 ) .
papadakis & lawrence ( 1995 ) suggested a method suitable for testing whether large agn datasets display evidence for strongly non - stationary variability . again
this method works by comparing the psds from different time intervals , in this case by determining whether the differences between two periodograms are consistent with the scatter expected based on equation [ eqn : pds_scatter ] .
in particular , they define a statistic @xmath67 based on the ratio of two normalised periodograms . if @xmath67 deviates significantly from its expected value for stationary data ( if @xmath68 ) then the hypothesis that the data are stationary can be rejected ( at some confidence level ) .
a different approach is compare variances @xmath69 derived from @xmath70 observations of the same source ( either segments of one long observation or separate short observations ) . in order to test whether the @xmath69 differ significantly ( i.e. more than expected for a red noise process ) a measure of the expected scatter is required .
this error could be obtained directly from the data ( by measuring the standard deviation of multiple estimates ) or through simulations ( based on an assumed psd shape ) .
an empirical estimate of the mean and standard deviation of the variance can be made given @xmath70 non - overlapping data segments .
the @xmath70 segments each yield an estimate of the variance data points in order to yield a meaningful variance .
] , @xmath69 .
each of these is an independent variable of ( in general ) unknown but identical distribution ( unless the process is strongly non - stationary ) .
the central limit theorem dictates that the sum of these will become normally distributed as @xmath70 increases .
therefore by averaging the @xmath70 variance estimates it is possible to produce an averaged variance ( @xmath71 ) and assign an error bar in the usual fashion ( e.g. equation 4.14 of bevington & robinson 1992 ) .
this gives a robust estimate of the variance and the standard deviation of the @xmath70 variances around the mean gives an estimate of the uncertainty on the mean variance .
if several sets of data segments are acquired it is therefore possible to compare the mean variance of each set statistically ( since each has an associated uncertainty ) .
for example , with two long _ xmm - newton _ observations of the same source , taken a year apart , one could measure the variance for each observation ( by breaking each into short segments and taking the mean variance of the segments ) .
thus it would be possible to test whether the variability was stationary between the two observations .
this method of estimating the mean and standard deviation of the variance requires a large amount of data . of order
@xmath72 data points are needed to produce a single well - determined estimate of the mean variance and its error .
a typical _ xmm - newton _ observation of a bright seyfert 1 galaxy ( @xmath73 ks duration ) is only likely to yield enough data for one estimate of the mean variance .
thus this method is suitable for testing whether the mean variance has changed from observation to observation .
[ fig : curve1 ] ( panel 4 ) demonstrates this empirically derived mean variance and its error bar on a long , simulated time series . these data were produced by calculating the variances @xmath69 in bins of @xmath74 data points ( panel 3 ) and then averaging @xmath75 variances to produce a mean variance with error bar ( panel 4 ) .
these averaged variances are consistent with constant , as expected ; fitting these data with a constant gave @xmath76 .
figure [ fig : curve2 ] shows the rms amplitude is constant with flux .
these tests indicate that the integrated psd is consistent with being constant with time ; the variance does not change significantly from epoch to epoch ( or as function of flux ) , as expected for a stationary process .
the advantage of the above method is that it requires no assumption about the shape of the psd , the drawback is that it requires a substantial amount of data to produce a single , robust variance estimate .
an alternative approach is to estimate the standard deviation of the variances @xmath69 based on simulations . given an assumed shape for the psd it is possible to calculate the distribution of variances expected for a stationary process ( see section [ sect : stochasticity ] ) .
some example distributions are shown in fig .
[ fig : fvar_dist1 ] , which clearly demonstrates how the distribution depends on the slope of the psd .
the distribution becomes more normal at flatter slopes and more asymmetric at steep slopes . for a given psd shape these distributions
are well defined ( by eqn .
[ eqn : sum_var ] ) and can be computed through monte carlo simulations .
this makes it possible to estimate limits within which one would expect the variance to be distributed if the process is stationary .
the two primary factors that affect the distribution of variance are the psd shape and the sampling of the light curve ( the length of the data segments in the case of contiguously binned light curves ) .
table [ table ] gives the expected confidence limits for four different psd shapes and five different lengths for the data segments .
these values were computed by simulating one very long light curve with the assumed psd shape and breaking it into @xmath77 separate segments ( of specified length ) .
the variance within each segment was measured and the distribution of the @xmath77 variances was calculated .
the @xmath78 per cent confidence interval was calculated by finding the @xmath79th and @xmath80th percentiles of the variance distribution ( in general these upper and lower bounds will differ because the distribution is asymmetric ) .
the numbers given in the table are the boundaries of the @xmath78 and @xmath81 per cent confidence regions estimated by averaging the results from @xmath82 runs .
the limits are given in terms of @xmath83 because they are multiplicative .
that is , from a particular realisation the variance is expected to be scattered within some factor of the true variance ( for which the absolute normalisation is irrelevant ) .
the factors are tabulated in terms of their logarithms ( since multiplicative factors in linear - space become additive offsets in log - space ) .
the psd used for the simulations was chosen to match that expected for agn , i.e. a steep power - law at high frequencies ( with a slope of @xmath84 ) breaking to a flatter slope ( @xmath85 ) at low frequencies .
the frequency of the break was fixed to be @xmath86 , in other words the break timescale was @xmath77 times the bin size .
( the absolute size of the time bins is arbitrary in the simulations . when comparing the simulated results to real data sampled with e.g. @xmath87 s time resolution , the break timescale in the simulated psd is thus @xmath87 ks . )
[ table ] . expected scatter in variance estimates .
the @xmath78 and @xmath81 per cent intervals are presented in terms of @xmath83 .
( the @xmath81 per cent interval is given in * bold*. ) the boundaries were calculated from monte carlo simulations of light curves .
the psd was chosen to be a broken power - law with a slope of @xmath88 below the break ( at a frequency @xmath86 ) and a slope above the break of @xmath84 .
the simulated data segments were chosen to be @xmath89 points long ( the timescale of the break in the psd being at @xmath77 , in arbitrary units ) . [ cols= " <
, > , > , > , > , > " , ] the numbers given in the table provide an approximate prescription for the expected scatter in the variance of a stationary process with a red noise psd similar to that of agn .
the simulated light curve shown in fig .
[ fig : curve1 ] was used to demonstrate the use of this table . in this case
the psd is know to have a slope @xmath90 , and the variances ( shown in panel 3 of fig .
[ fig : curve1 ] ) were calculated every @xmath91 points .
therefore , the @xmath78 interval for the expected variance is given by @xmath92 . taking the mean variance as the expectation value for @xmath93 ,
this translates to @xmath94 .
the interval boundaries were calculated by converting the logarithmic value into a linear factor and multiplying by the sample mean ( assumed to represent the true variance ) .
this interval is shown on fig .
[ fig : sim_result ] by the dotted lines .
the corresponding @xmath81 per cent confidence interval is also marked .
as expected the individual variances fall within the expected region .
however , the @xmath78 per cent region spans an order of magnitude in variance .
thus even order of magnitude differences in variance between short sections of a light curve are to be expected and do not necessarily indicate that the underlying process is not stationary .
subtle changes in the psd will thus be difficult to detect by examining the raw variances as the intrinsic scatter is so large .
such changes could be revealed by comparing averaged variances or comparing the psds as described above .
the datasets considered thus far have been ideal , in the sense that they are free from flux uncertainties . in real life , however , a light curve @xmath10 will have finite uncertainties @xmath96 due to measurement errors ( such as poisson noise in the case of an x - ray photon counting signal ) .
these uncertainties on the individual flux measurements will contribute an additional variance .
this leads to the use of the ` excess variance ' ( nandra et al .
1997 ; edelson et al . 2002 ) as an estimator of the _ intrinsic _ source variance .
this is the variance after subtracting the contribution expected from measurement errors @xmath97 where @xmath98 is the mean square error @xmath99 the normalised excess variance is given by @xmath100 and the fractional root mean square ( rms ) variability amplitude ( @xmath0 ; edelson , pike & krolik 1990 ; rodriguez - pascual et al .
1997 ) is the square root of this , i.e. @xmath101 the statistic @xmath95 is often chosen in preference to @xmath102 although the two convey exactly the same information .
@xmath95 is a linear statistics and can therefore give the rms variability amplitude in per centage terms . the choice of whether to quote @xmath95 or @xmath103 is usually purely one of presentation .
it is worth noting that the monte carlo results given in section [ sect : sim_table ] , to estimate the expected scatter on the variance , can also be applied to its square root .
the expected boundaries of the confidence region of the logarithm of the rms is approximately half those of the variance .
specifically , @xmath104 and similarly @xmath105 .
an x - ray light curve of an agn can be split into different energy bands .
the light curves in each band will be strictly simultaneous and can be used to test whether the x - ray variability is a function of energy .
for example , one might examine the ratio of a soft band light curve to a hard band light curve .
the statistical significance of any variations in the ratio can be quantified by propagating the measurement errors and applying an appropriate test , such as the @xmath24 test of the constant ratio hypothesis .
if the ratio shows variations greater than those expected from the errors then the two light curves are intrinsically different and the source does indeed show spectral variability .
such changes in the energy spectrum with time can in principle provide valuable clues to the nature of the x - ray source
. this test does not provide any quantitative description of the spectral variability .
another tool for investigating spectral variability is the rms spectrum , i.e. the rms variability amplitude ( or @xmath95 ) as a function of energy .
see e.g. inoue & matsumoto ( 2001 ) , edelson et al .
( 2002 ) , fabian et al .
( 2002 ) and schurch & warwick ( 2002 ) for some examples of rms spectra from agn .
however , when examining rms spectra it is often not clear whether changes in the amplitude with energy reflect real energy - dependence of the intrinsic variability amplitude or are caused by random errors in the light curves .
the finite measurement errors on the individual fluxes ( e.g. due to poisson noise ) will introduce some uncertainty in the estimated rms amplitudes .
an estimate of this uncertainty would help answer the question posed above , namely whether features in rms spectra are the result of random errors in the data or represent spectral variations intrinsic to the source .
the problem of how to assess the uncertainty on the excess variance ( or @xmath95 ) is a long - standing one ( e.g. nandra et al . 1997 ; turner et al . 1999 ; edelson et al .
the standard error formulae presented in the literature ( e.g. turner et al . 1999 ; edelson et al .
2002 ) are formally valid in the case of un - correlated gaussian processes . typically agn light curves at different x - ray energies are strongly correlated and are not gaussian . however , when searching for subtle differences in amplitude between simultaneous and correlated light curves it may be more useful to have an indication of the uncertainty resulting from the finite flux errors .
a monte carlo approach was used to develop a prescription of the effect of measurement errors on estimates of @xmath95 ( and @xmath106 ) .
a short red noise light curve was generated .
poisson noise was added ( i.e. the individual flux measurements were randomised following the poisson distribution ) and the excess variance was recorded .
the fluxes of the original light curve were randomised again and the excess variance recorded , this was repeated many times .
the distribution of excess variances was then used to determine the uncertainty in the variance estimate caused by poisson noise .
full details of the procedure are given in appendix [ sect : mc ] for these simulations it was found that the error on @xmath106 decreases as the s / n in the light curve is increased according to : @xmath107 see appendix b for details of this equation and its equivalent in terms of @xmath0 . as this only accounts for the effect of flux measurement errors ( such as poisson noise ) in a given light curve
it can be used to test whether two simultaneously observed light curves of the same source , but in different bands , show consistent amplitudes .
a demonstration of this using real data is given in the following section .
this uncertainty does not account for the random scatter intrinsic to the red noise process , therefore the absolute value of the rms spectrum will change between realisations ( i.e. from epoch to epoch ) .
but if a source shows achromatic variability then the values of @xmath0 calculated in each energy band ( at a given epoch ) should match to within the limits set by the poisson noise ( i.e. the fractional rms spectrum should be constant to within the uncertainties given by the above equation ) .
differences in @xmath0 significantly larger than these would indicate that the source variability amplitude is a function of energy .
this would then mean the psd amplitude / shape is different in different energy bands , or there are multiple spectral components that vary independently .
the above uncertainty estimates can be used to test the hypothesis that the source variability is achromatic .
if significant differences between energy bands are detected ( as in the case of presented below ) then these errors should not be used to fit the rms spectrum .
the assumption that the differences are due only to measurement errors is no longer the case . in such situations
the light curves in adjacent energy bands are likely to be partially correlated and so @xmath24-fitting of the rms spectrum is not appropriate .
the differences in excess variance will be a combination of intrinsic differences and measurement errors .
their uncertainty will therefore be more difficult to quantify .
panel 2 and 3 : mean count rate and excess variance measured from segments of 20 points .
panel 4 : averaged normalised excess variance measured by binning the individual variance estimates into groups of 20 .
this average variance is inconsistent with constant .
panel 5 : fractional rms amplitude measured from segments of 20 points .
panel 6 : averaged fractional rms amplitude measured by binning the individual amplitudes into groups of 20 .
this average fractional amplitude is consistent with constant .
this contrasts with the situation shown in fig .
[ fig : curve1 ] [ fig : mrk766_1 ] , width=8 ] in this section a long ( @xmath108 s ) _ xmm - newton _ observation of the bright , variable seyfert 1 galaxy markarian 766 is used to illustrate the points discussed above .
the data were obtained from the _ xmm - newton_data archive .
details of the observation are discussed in mason et al .
( 2003 ) and an analysis of the psd is presented in vaughan & fabian ( 2003 ) . was observed by _ xmm - newton _ ( jansen et al .
2001 ) over the period 2001 may 20 2001 may 21 ( rev .
the present analysis is restricted to the data from the pn european photon imaging camera ( epic ) , which was operated in small window mode . extraction of science products from the observation data files ( odfs ) followed standard procedures using the _ xmm - newton _ science analysis system ( sas ) v5.3.3 .
source data were extracted from a circular region of radius 35 arcsec from the processed image and only events corresponding to patterns 04 ( single and double pixel events ) were used .
background events were extracted from regions in the small window least effected by source photons , these showed that the background rate increased dramatically during the final @xmath109 s of the observation .
this section of the data was excluded , leaving @xmath110 s of uninterrupted data .
the light curves were corrected for telemetry drop outs ( less than 1 per cent of the total time ) and background subtracted .
the errors on the light curves were calculated by propagating the poisson noise . the broad band ( 0.210 kev ) light curve extracted from the pn is shown in fig . [ fig : mrk766_1 ] ( panel 1 ) .
as was the case for the simulated data shown in fig [ fig : curve1 ] , the mean and variance ( calculated every 20 data points ) s. ] show changes during the length of the observation ( panels 2 and 3 ) .
the expected range for the excess variance , calculated using the results of section [ sect : sim_table ] ( and assuming a psd slope of @xmath111 ) , is marked in figure [ fig:766_result ] .
[ fig:766_result2 ] shows the same data in terms of normalised excess variances .
neither of these show fluctuations larger than expected for a stationary process . but given the large expected scatter this is a rather insensitive test . in the case of the light curve however , there are sufficient data to examine variations of the average variance with time , allowing a more sensitive test for non - stationarity . by averaging the excess variance estimates ( in time bins containing 20 excess variance estimates ) significant changes in the variance with time
are revealed ( panel 4 ) .
this contrasts with the simulated data shown in fig .
[ fig : curve1 ] .
the binned excess variance is inconsistent with a constant hypothesis : fitting with a constant gave @xmath112 for @xmath113 degrees of freedom ( @xmath114 ) , rejected at @xmath81 per cent confidence .
the average variance is therefore changing with time , indicating the variability is strongly non - stationary . ) as a function of flux for the light curve ( compare with fig .
[ fig : curve2 ] ) .
bottom panel : the average it fractional rms amplitude ( @xmath115 ) as a function of flux . clearly the absolute rms amplitude is a function of flux , but this dependence is removed in the fractional rms . [
fig : mrk766_rms ] , title="fig:",width=6 ] + ) as a function of flux for the light curve ( compare with fig . [
fig : curve2 ] ) .
bottom panel : the average it fractional rms amplitude ( @xmath115 ) as a function of flux . clearly the absolute rms amplitude is a function of flux , but this dependence is removed in the fractional rms .
[ fig : mrk766_rms ] , title="fig:",width=6 ] a careful inspection of fig .
[ fig : mrk766_1 ] ( panels 3 and 4 ) shows the individual variance estimates have a tendency to track the source count rate .
this is difficult to discern from the individual variances ( panel 3 ) , due to the larger intrinsic scatter , but much clearer in the averaged variances ( panel 4 ) .
this can be seen clearly in fig .
[ fig : mrk766_rms ] ( top panel ) where the rms amplitude ( @xmath116 ) is shown as a function of count rate . to produce this plot the individual rms estimates ( fig .
[ fig : mrk766_1 ] , panel 3 ) were sorted by count rate and binned by flux ( such that there were @xmath91 estimates per bin ) .
the error on the mean rms was calculated in the standard fashion ( see above ) .
this indicates that the source does show a form of genuine non - stationarity : the absolute rms amplitude of the variations increases , on average , as the source flux increases .
this effect has been noted in other seyferts ( uttley & m@xmath2hardy 2001 ; edelson et al .
2002 ; vaughan et al .
2003 ) and is due to a linear correlation between rms and flux ( see uttley et al . in prep . for further discussion of this effect
non - stationarity of this form can be ` factored out ' by using the normalised amplitude ( @xmath95 or @xmath102 ) instead of the absolute values . normalising each variance ( or rms ) estimate by its local flux
removes this trend .
the bottom panel of fig .
[ fig : mrk766_rms ] shows that @xmath95 is indeed constant with flux ( fitting a constant gave @xmath117 for @xmath118 ) .
[ fig : mrk766_1 ] ( panels 5 and 6 ) shows @xmath95 and its average as a function of time ; the average is consistent with staying constant ( @xmath119 for @xmath118 ) .
the variability of does show genuine ( strong ) non - stationarity , in the sense that the absolute rms increases linearly with flux , but this trend can be removed by using normalised units
@xmath95 ( and therefore the normalised excess variance ) is consistent with being constant ( with time and flux ) .
the above analysis suggests that , after accounting for the effect of the rms flux correlation , there is no other evidence for strong non - stationarity in the rapid variability of .
this was confirmed using the @xmath67-test of papadakis & lawrence ( 1995 ; see their appendix a ) .
a periodogram was calculated for three consecutive segments of @xmath120 s duration , and normalised to fractional units ( see appendix [ sect : pds_norm ] ) .
the @xmath67 value was computed by comparing periodograms at frequencies below @xmath121 hz ( above which the poisson noise power becomes comparable to the source variability ) .
for each pair of periodograms the value of @xmath67 was within the range expected for stationary data ( specifically @xmath122 , within one standard deviation of the expected value ) .
the variability amplitude as a function of energy was calculated by measuring @xmath95 from light curves extracted in various energy ranges .
the results are shown in fig .
[ fig : mrk766_2 ] and the errors were calculated using equation [ eqn : fvar_err1 ] to account for the effect of poisson noise . the variability amplitude is clearly a function of energy , i.e. shows significant spectral variability .
this was confirmed by a fourier analysis of the light curves in different energy bands ( vaughan & fabian 2003 ) which revealed complex energy - dependent variability .
the rms spectrum was re - calculated for light curves containing only single pixel ( pattern 0 ) events and again for double pixel ( patterns 14 ) events .
these two sets of data were extracted from the same detector and using identical extraction regions etc . after accounting for the difference in count rate between single and double pixel events the two sets of light curves should be identical except for the effects of poisson noise .
the two rms spectra should be the same except for poisson errors . comparing the ratio of the two rms spectra using a @xmath123-test ( against the hypothesis of unity ratio ) gave @xmath124 degrees of freedom .
comparing the difference of the two rms spectra to the hypothesis of zero difference gave identical results and shows the two rms spectra are indeed fairly consistent .
this test indicates that for real data the error formula given above does provide a reasonable description of the uncertainty induced by photon noise .
, width=6 ]
the analysis of stochastic processes , such as x - ray variability of agn , is conceptually different from the analysis of deterministic data such as time - averaged spectra ( see discussions in e.g. jenkins & watts 1968 ; priestley 1981 ; bendat & piersol 1986 ; bloomfield 2000 ) .
for example , when observing the spectrum of a constant source one expects repeatability of the data to within the limits set by measurement errors , i.e. each new realisation of the spectrum should be consistent within the errors . in agn variability analysis
it is the signal itself that is randomly variable ; one does not expected repeatability of quantities such as the mean or variance .
these statistical moments will change ( randomly ) with each new light curve even if there are no measurement errors .
the stochastic nature of red noise processes means that it is usually only their _ average _ properties that can provide physical insight .
non - deterministic data should be handled statistically .
for example , it is customary to examine the timing properties of x - ray binaries using psds estimated from the average periodogram of an ensemble of light curves ( e.g. van der klis 1995 ) .
averaging over many independent realisations reduces the random fluctuations inherent in the noise process . in most agn timing studies
however there are rarely enough data to construct averages in this way ( but see papadakis & lawrence 1993 and uttley et al .
2002 for more on psd estimation for agn ) . as a result of this relative lack of data , agn timing studies often emphasise the properties of a single light curve . but
emphasis on the detailed properties of any single realisation of a stochastic process can be misleading .
for example , agn light curves show large fluctuations in variance .
these changes provide little insight as they are expected even when the underlying physical process responsible for the variability is constant .
rather , they may simply be statistical fluctuations intrinsic to the stochastic process .
all red noise processes show random fluctuations in both mean and variance and the variance will be distributed in a non - gaussian fashion with a large scatter ( see sections [ sect : stochasticity ] and [ sect : errors_intrinsic ] ) .
previous claims of non - stationary variability based on changes in variance ( e.g. nandra et al . 1997 ; dewangan et al . 2002 ; gliozzi , sambruna & eracleous 2003 ) should therefore be treated with caution since they did not account for this intrinsic scatter ( see also section 3.3.1 of leighly 1999 for a discussion of this point ) .
real changes in the psd would indicate genuine non - stationarity and reflect real changes in the physical conditions of the variability process .
such changes can be measured from the average properties of the light curve , such as the averaged periodogram or the averaged variance ( see section [ sect : errors_intrinsic ] ) .
a different issue is that differences between the variance of simultaneous light curves obtained in different energy bands can be examined using the excess variance ( or @xmath95 ) statistic .
it is possible to estimate the uncertainty in the excess variance due to errors in the flux measurements .
this uncertainty , accounting only for measurement ( e.g. poisson ) errors , can be used when testing for spectral variability , as demonstrated in section [ sect : mrk766 ] .
estimators such as the excess variance provide a useful , if crude , means of quantifying the variability of agn .
even though the stochastic nature of agn light curves makes it difficult to estimate variability amplitudes robustly from short observations , the excess variance can provide useful information .
for example an analysis of the excess variances measured from short observations of seyfert 1 galaxies demonstrated that the variability amplitude ( at a given range of timescales ) is inversely correlated with the luminosity of the source ( nandra et al .
1997 ; leighly 1999 ; markowitz & edelson 2001 ) .
although random fluctuations in variance are expected for agn light curves the range of variances observed is far larger than could be accounted for by this effect alone .
another example is given in section [ sect : mrk766_stationary ] when it is demonstrated that the average variance of is a function of the flux of the source .
a similar effect has been observed in x - ray binaries ( uttley & m@xmath66hardy 2001 ) .
a discussion of the implications of this result will be given in uttley et al . in prep .
this paper discusses some aspects of quantifying the variability of agn using simple statistics such as the variance .
various possible uses of these are presented and some possible problems with their significance and interpretation are brought to light .
the primary issues are as follows : 1 . in order to search for non - stationary variability in an ensemble of short light curves ( or short light curve segments ) one can test whether the individual variances are consistent with their mean .
two practical methods are presented ( sections [ sect : bin_var ] and [ sect : sim_table ] ) .
2 . in the first method
the mean variance and its error are calculated at various epochs by binning the individual variance estimates .
this is most useful when searching for subtle changes in variability amplitude but requires large datasets ( in order that the variance can be sufficiently averaged ) .
3 . in the second method
the individual variance estimates are compared with the expected scatter around the mean .
the expected scatter is calculated using monte carlo simulations of stationary processes .
the table gives some examples of the scatter expected for various psd shapes typical of agn .
this table can therefore be used to provide a ` quick look ' at whether the observed fluctuations in the variance are larger than expected .
one drawback is that , because the intrinsic scatter in the variance is rather large for red noise data , this method is only sensitive to very large changes in the variability amplitude .
another drawback is that one has to assume a shape for the psd .
4 . the excess variance can also be used to quantify how the variance changes as a function of energy ( section [ sect : spec_var ] ) .
an approximate formula is presented ( based on the results of monte carlo simulations ) that gives the expected error in the excess variance resulting from only observation uncertainties ( flux errors such as poisson noise ) .
this can be used to test for significant differences in variance between energy bands .
if the normalised excess variances ( or @xmath125s ) are found to differ significantly between energy bands this implies the psd is energy dependent and/or there are independently varying spectral components .
possibly the most robust yet practical approach to variability analysis from agn data is to test the validity of hypotheses using monte carlo simulations .
this approach has yielded reliable psd estimates for seyfert galaxies ( green et al .
1999 ; uttley et al . 2002
; vaughan et al . 2003 ; markowitz et al . 2003 ) and has been used to test the reliability of cross - correlation results ( e.g. welsh 1999 ) amongst other things .
section [ sect : simulations ] discusses some methods for simulating red noise data .
we are very grateful to the referee , andy lawrence , for a thoughtful report that prompted significant improvements to the manuscript .
sv and pu acknowledge support from pparc .
this paper made use of observations obtained with _ xmm - newton _ , an esa science mission with instruments and contributions directly funded by esa member states and the usa ( nasa ) .
99 belloni t. , hasinger g. , 1990 , a&a , 227 , l33 belloni t. , psaltis d. , van der klis m. , 2002 , apj , 572 , 392 bendat j. s. , piersol a. g. , 1986 , _ random data : analysis and measurement procedures _ , wiley ( new york ) bevington p. r. , robinson d. k , .
1992 , _ data reduction and error analysis for the physical sciences _ , mcgraw - hill ( new york ) bloomfield p , 2000 , _ fourier analysis of time series _ , wiley ( new york ) deeming t. j. , 1970 , aj , 75 , 1027 deeming t. j. , 1975 , ap&ss , 36 , 137 deeter j. e. , boynton p. e. , 1982 , apj , 261 , 337 dewangan g. c. , boller th . , singh k. p. , leighly k. m. , 2002 , a&a , 390 , 65 done c. , madejski g. m. , mushotzky r. f. , turner t. j. , koyama k. , kunieda h. , 1992 , apj , 400 , 138 edelson r. , pike g. f. , krolik j. h. , 1990 , apj , 359 , 86 edelson r. , nandra k. 1999 , apj , 514 , 682 edelson r. et al .
2002 , apj , 568 , 610 fabian a. c. et al .
2002 , mnras , 335 , l1 green a. r. , m@xmath2hardy i. m. , lehto h. j. 1993 , mnras , 265 , 664 green a. r. , m@xmath2hardy i. m. , done c. , 1999 , mnras , 305 , 309 gliozzi m. , sambruna r. m. , eracleous m. , 2003 , apj , 584 , 176 inoue , h. , matsumoto , c. , 2001 , adspr , 28 , 445 jenkins g. m. , watts , d. g. , 1968 , _ spectral analysis and its applications _ , holden - day ( san fancisco ) lawrence a. , watson m. g. , pounds k. a. , elvis m. , 1987 , nature , 325 , 694 lawrence a. , papadakis i. , 1993 , apj , 414 , l85 leahy d. a. , darbro w. , elsner r. f. , weisskopf m. c. , kahn s. , sutherland p. g. , grindlay j. e. , 1983 , apj , 266 , 160 lehto h. j. , 1989 , in j. hunt , b. battrick , eds , _ two topics in x ray astronomy _ , ( esa sp-296 ; noordwijk : esa ) , p499 leighly k. , 1999 , apjs , 125 , 297 markowitz a. , edelson r. , 2001 , apj , 547 , 684 markowitz a. et al .
2003 , apj , in press ( astro - ph/0303273 ) mason k. o. , et al .
2003 , apj , 582 , 95 m@xmath2hardy i. m. , 1989 , in j. hunt , b. battrick , eds , _ two topics in x ray astronomy _ , ( esa sp-296 ; noordwijk : esa ) , p1111 nandra k. , george i. m. , mushotzky r. f. , turner t. j. , yaqoob t. , 1997 , apj , 476 , 70 papadakis i. e. , lawrence a. , 1993 , mnras , 261 , 612 papadakis i. e. , lawrence a. , 1995 , mnras , 272 , 161 pottschmidt k. , 2002 , a&a , submitted ( astro - ph/0202258 ) press w. h. 1978 , comments on astrophysics , 7 , 103 press w. h. , teukolsky s. a. , vetterling w. t. , flannery b. p. , 1996
, _ numerical recipes
_ , cambridge univ . press .
( cambridge ) press w. h. , rybicki g. b. , 1997 , in _ astronomical time series _ , d. maoz , a. sternberg , e. m. leibowitz ( eds . ) , kluwer ( dordrecht ) , p61 priestley m. b. , 1981 , _ spectral analysis and time series _
, academic press ( london ) rodriguez - pascual p. m. et al . , 1997 , apjs , 110 , 9 scargle j. d. , 1981 , apjs , 45 , 1 schurch n. j. , warwick r. s. , 2002 , mnras , 334 , 811 stella l. , arlandi e. , tagliaferri g. , israel g. l. , 1997 , in _ ` applications of time series analysis in astronomy and meteorology ' _ , t. subba rao , m. b. priestley , o. lessi ( eds . ) , chapman & hall ( london ) ( astro - ph/9411050 ) timmer j. , knig m. 1995 , a&a , 300 , 707 turner t. j. , george i. m. , nandra k. , turcan d. , 1999 , apj , 524 , 667 uttley p. , m@xmath2hardy i. m. , 2001 , mnras , 323 , 26 uttley p. , m@xmath2hardy i. m. , papadakis , i. 2002 , mnras , 332 , 231 van der klis m. , 1989 , in _ timing neutron stars _ , h. ogelman , e. p. j. van den heuvel ( eds . ) , kluwer ( dordrecht ) , nato asi series c 262 , p27 van der klis m. , 1995 , in _ x - ray binaries _ , w. h. g. lewin , j. van paradijs , e. p. j. van den heuvel ( eds . ) , cambridge univ . press ( cambridge ) , p252 van der klis m. , 1997 , in _
statistical challenges in modern astronomy ii _ , ed .
babu , e.d .
feigelson , springer - verlag ( new york ) , p321 vaughan s. , fabian a. c. , nandra k. 2003 , 339 , 1237 vaughan s. , fabian a. c. , 2003 , mnras , 341 , 496 welsh w. f. , 1999 , pasp , 111 , 1347
the periodogram is calculated by normalising the modulus - squared of the dft ( see equation [ eqn : ft ] ) : @xmath126 there are a variety of options for the normalisation @xmath19 used in the literature , each has desirable properties . in the mathematical literature on time
series analysis a normalisation of the form @xmath127 is standard ( e.g. priestley 1981 ; bloomfield 2000 ) .
however , this normalisation is generally not used for time series analysis in astronomy because the periodogram then depends on flux of the source and the binning of the time series . below
are listed three of the most commonly used normalisations , which only differ by factors of @xmath39 , the mean count rate in cts / s ( @xmath128 ) .
the factor of two is present in all these normalisations to make the periodogram ` one sided , ' meaning that integrating over positive frequencies only yields the correct variance .
1 . @xmath129 defined by van der klis ( 1997 ) ( see also miyamoto et al . 1991 ) .
this is the normalisation most often used in analysis of agn and x - ray binaries because the integrated periodogram yields the fractional variance of the data .
the units for the periodogram ordinate are ( rms / mean)@xmath130 hz@xmath3 ( where rms / mean is the dimensionless quantity @xmath0 ) , or simply hz@xmath3 .
+ if a light curve consists of a binned photon counting signal ( and in the absence of other effects such as detector dead - time ) the expected poisson noise ` background ' level in its periodogram is given by @xmath131 where @xmath39 is the mean source count rate , @xmath132 is the mean background count rate , @xmath133 is the sampling interval and @xmath134 is the time bin width .
the factor of @xmath135 accounts for aliasing of the poisson noise level if the original photon counting signal contained gaps .
if the light curve is a series of contiguous time bins ( i.e. @xmath136 ) and has zero background ( which is approximately true for many _ xmm - newton _ light curves of agn ) then this reduces to @xmath137 .
+ for a light curve with gaussian errors @xmath138 the noise level in the periodogram is @xmath139 2 .
@xmath140 originally due to leahy et al .
this has the property that the expected poisson noise level is simply 2 ( for continuous , binned photon counting data ) .
if the light curve consists only of poisson fluctuations then the periodogram should be distributed exactly as @xmath26 .
it is this property that makes this normalisation the standard for searching for periodic signals in the presence of poisson noise ( see leahy et al . 1983 ) .
if the input light curve is in units of ct s@xmath3 then the periodogram ordinate is in units of ct s@xmath3 hz@xmath3 .
3 . @xmath141 this is the normalisation used in equation [ eqn : pds ] .
this gives the periodogram in absolute units [ e.g. ( ct s@xmath3)@xmath130 hz@xmath3 ] and so the integrated periodogram gives the total variance in absolute units [ e.g. ( ct s@xmath3)@xmath130 ] for a contiguously binned light curve with poisson errors the noise level is @xmath144 , and for gaussian errors the noise level is @xmath145 .
to estimate the effect on @xmath106 due only to poisson noise the basic strategy was as follows . 1 .
generate a random red noise light curve .
this acts as the ` true ' light curve of the source .
2 . add poisson noise , i.e. draw fluxes from the light curve according to the poisson distribution .
this simulates ` observing ' the true light curve .
error bars were assigned based on the ` observed ' counts in each bin ( @xmath146 ) .
3 . measure the normalised excess variance @xmath106 of the observed light curve .
this will be different from the variance of the true light curve because of the poisson noise .
steps 2 and 3 were repeated , using the same true light curve , to obtain the distribution of @xmath106 .
[ fig : fvar_dist2 ] shows some results . in this example
the ` true ' light curve was generated with a @xmath53 psd and normalised to a pre - defined mean and variance , e.g. @xmath147 ( @xmath148% ) .
this light curve was then observed ( i.e. steps 2 and 3 were repeated ) @xmath149 times . ] .
the three panels correspond to different mean count rates for the true light curve ( i.e. different s / n of the observation ) .
the ( @xmath150 ) widths of the @xmath106 distributions are monte - carlo estimates of the size of the error bars on @xmath106 due to poisson noise . as is clear from fig .
[ fig : fvar_dist2 ] the distribution of @xmath106 becomes narrower , i.e. the error on @xmath106 gets smaller , as the s / n of the data increases .
obviously in the limit of very high s / n data the measured value of @xmath106 will tend to the ` true ' value ( in this case @xmath151 ) , i.e. @xmath152 as @xmath153 .
it should also be noted that the distributions are quite symmetrically centred on the correct value , indicating that @xmath106 is an unbiased estimator of the intrinsic variance in the light curve , even in relatively low s / n data . in order to assess how the error on @xmath106 changes with s / n
, the width of its distribution was measured from simulated data at various different settings of s / n and intrinsic variance ( i.e. @xmath42 ) .
width of the distribution at each setting was calculated from only @xmath154 ` observations ' of each light curve . in order
that no particular realisation adversely affect the outcome , and to increase the statistics , this was repeated for @xmath91 different random light curves ( of the same fractional variance ) and the width of the @xmath106 distributions were averaged ( i.e. the whole cycle of steps 13 was repeated @xmath91 times ) .
thus for each specified value of s / n and fractional variance , the error on @xmath106 is estimated from @xmath149 simulated ` observations . '
these monte carlo estimated errors on the normalised excess variance are shown in fig .
[ fig : signal / noise ] . from 10,000 ` observations ' of the same light curve . in each case the ` true ' @xmath106 is 0.04 ( dotted line ) .
the top panel used the lowest s / n data , the bottom panel used the highest s / n data .
the mean number of counts per bin in the simulated light curves was 15 ( top ) , 30 ( middle ) and 100 ( bottom ) .
as the s / n increases ( count rate increases ) the distribution of @xmath106 becomes narrower .
( note : this is different from fig . [
fig : fvar_dist1 ] , which shows how the variance changes between different realisations of the same stochastic process . )
[ fig : fvar_dist2 ] , title="fig:",width=5 ] + from 10,000 ` observations ' of the same light curve . in each case
the ` true ' @xmath106 is 0.04 ( dotted line ) . the top panel used the lowest s / n data , the bottom panel used the highest s / n data .
the mean number of counts per bin in the simulated light curves was 15 ( top ) , 30 ( middle ) and 100 ( bottom ) .
as the s / n increases ( count rate increases ) the distribution of @xmath106 becomes narrower .
( note : this is different from fig . [
fig : fvar_dist1 ] , which shows how the variance changes between different realisations of the same stochastic process . )
[ fig : fvar_dist2 ] , title="fig:",width=5 ] + from 10,000 ` observations ' of the same light curve . in each case
the ` true ' @xmath106 is 0.04 ( dotted line ) . the top panel used the lowest s / n data , the bottom panel used the highest s / n data .
the mean number of counts per bin in the simulated light curves was 15 ( top ) , 30 ( middle ) and 100 ( bottom ) .
as the s / n increases ( count rate increases ) the distribution of @xmath106 becomes narrower .
( note : this is different from fig .
[ fig : fvar_dist1 ] , which shows how the variance changes between different realisations of the same stochastic process . )
[ fig : fvar_dist2 ] , title="fig:",width=5 ] ( resulting from poisson noise ) as a function of the number of counts per bin .
compare with fig .
[ fig : fvar_dist2 ] .
the solid curve shows the function described in the text ( equation [ eqn : xs_error ] ) .
[ fig : signal / noise ] , width=6 ] the solid lines show the functions defined by equation [ eqn : xs_error ] ( which was obtained by fitting various trial functions to the monte carlo results ) .
clearly this equation gives a very good match to the monte carlo results .
if the variability is not well detected , either because the s / n is low or the intrinsic amplitude is weak , then @xmath155 .
it is the first term on the right hand side of equation [ eqn : xs_error ] that dominates .
if the variability is well detected , i.e. @xmath156 , then it is the second term that dominates . in the former case
the deviations from the mean are dominated by the errors and the fluxes are approximately normally distributed . in this regime
the error equation becomes the same as that given in equation a9 of edelson et al .
( 2002 ) . in the latter case
the deviations in the light curve are enhanced by the intrinsic variance .
the second term is similar to the first except multiplied by a factor @xmath158 to account for this .
equation [ eqn : xs_error ] can be used to give the uncertainty on @xmath0 thusly @xmath159 and this is the equation used to derive the errors shown in fig .
[ fig : mrk766_2 ] . in the two regimes
this becomes : @xmath160 in the first instance , when the variability is not well detected , @xmath161 should be preferred over @xmath0 as negative values of @xmath161 are possible .
additional monte carlo simulations confirmed the above equations are valid for both gaussian and poisson distributed flux errors .
it is worth reiterating that this error accounts only for measurement errors on the fluxes .
it does not account for the intrinsic scatter in the fluxes inherent in any red noise process . | we review some practical aspects of measuring the amplitude of variability in ` red noise ' light curves typical of those from active galactic nuclei ( agn ) .
the quantities commonly used to estimate the variability amplitude in agn light curves , such as the fractional rms variability amplitude , @xmath0 , and excess variance , @xmath1 , are examined .
their statistical properties , relationship to the power spectrum , and uses for investigating the nature of the variability processes are discussed .
we demonstrate that @xmath1 ( or similarly @xmath0 ) shows large changes from one part of the light curve to the next , even when the variability is produced by a stationary process .
this limits the usefulness of these estimators for quantifying differences in variability amplitude between different sources or from epoch to epoch in one source .
some examples of the expected scatter in the variance are tabulated for various typical power spectral shapes , based on monte carlo simulations .
the excess variance can be useful for comparing the variability amplitudes of light curves in different energy bands from the same observation .
monte carlo simulations are used to derive a description of the uncertainty in the amplitude expected between different energy bands ( due to measurement errors ) .
finally , these estimators are used to demonstrate some variability properties of the bright seyfert 1 galaxy markarian 766 .
the source is found to show a strong , linear correlation between rms amplitude and flux , and to show significant spectral variability .
[ firstpage ] galaxies : active galaxies : seyfert galaxies : individual ( ) x - rays : galaxies methods : data analysis |
kohn - sham density functional theory ( ksdft ) is the most widely used framework for electronic - structure calculations , and plays an important role in the analysis of electronic , structural and optical properties of molecules , solids and nano - structures .
the efficiency of ksdft depends largely on the computational cost associated with the evaluation of the electron charge density for a given potential within a self - consistent field ( scf ) iteration .
the most straightforward way to perform such an evaluation is to partially diagonalize the kohn - sham hamiltonian by computing a set of eigenvectors corresponding to the algebraically smallest eigenvalues of the hamiltonian .
the complexity of this approach is @xmath1 , where @xmath2 is the number of electrons in the atomistic system of interest . as the number of atoms or electrons in the system increases , the cost of diagonalization becomes prohibitively expensive .
although linear scaling algorithms @xcite are attractive alternatives for improving the efficiency of ksdft , they rely on using the nearsightedness principle @xcite , which asserts that the density perturbation induced by a local change in the external potential decays away from where the perturbation is applied .
one can then truncate elements of the density matrix away from the diagonal .
such truncation can be in practice applied only to insulating systems whose density matrix elements decay exponentially away from the diagonal , but not to metallic systems at low temperature , for which the density matrix elements decay only algebraically away from the diagonal . the recently developed pole expansion and selected inversion ( pexsi ) method @xcite provides an alternative way for solving the kohn - sham problem without using a diagonalization procedure , and without invoking the nearsightedness principle to truncate density matrix elements . compared to existing techniques ,
the pexsi method has a few salient features : 1 ) pexsi expresses physical quantities such as electron density , free energy , atomic forces , density of states and local density of states in terms of a spectral projector which is evaluated without computing any eigenvalues or eigenvectors .
2 ) the computational cost of the pexsi technique scales at most as @xmath3 ) .
the actual computational cost depends on the dimensionality of the system : the cost for quasi-1d systems such as nanotubes is @xmath4 _ i.e. _ linear scaling ; for quasi-2d systems such as graphene and surfaces ( slabs ) is @xmath5 ; for general 3d bulk systems is @xmath6 .
this favorable scaling hinges on the sparse character of the hamiltonian and overlap matrices , but not on any fundamental assumption about the localization properties of the single particle density matrix .
3 ) the pexsi technique can be accurately applied to general materials system including small gapped systems and metallic systems , and remains accurate at low temperatures .
4 ) the pexsi method has a two - level parallelism structure and is by design highly scalable .
the recently developed massively parallel pexsi technique can make efficient usage of @xmath7 processors on high performance machines .
5 ) pexsi can be controlled with a few input parameters , and can act nearly as a black - box substitution of the diagonalization procedure commonly used in electronic structure calculations . in order to benefit from the pexsi method
, the hamiltonian and overlap matrices must be sparse .
this requirement is satisfied if localized discretization is used for representing the kohn - sham hamiltonian by a finite sized matrix .
examples of localized discretization include numerical atomic orbitals @xcite , gaussian type orbitals @xcite , the finite difference @xcite and finite element @xcite methods , adaptive curvilinear coordinates @xcite , optimized nonorthogonal orbitals @xcite and adaptive local basis functions @xcite .
in contrast , the plane - wave basis set is not localized and therefore can not directly benefit from the pexsi method . even though they are formally localized , the number of degrees of freedom per atom associated with methods such as the finite difference and the finite element is usually much larger than that associated with other methods such as numerical atomic orbitals , leading to an increase of the preconstant factor in the computational cost .
therefore the finite difference and the finite element methods may not benefit as much from the pexsi technique as those based on numerical atomic orbitals .
we note also that the use of hybrid functionals with an orbital - based exact exchange term @xcite may significantly impact the sparsity pattern of the hamiltonian matrix and increase the computational cost . in previous work @xcite , the applicability of the pexsi method was demonstrated for accelerating atomic orbital based electronic structure calculations . with the sequential implementation of the pexsi method
, it was possible to perform electronic structure calculations accurately for a nanotube containing 10,000 atoms discretized by a single-@xmath8 ( sz , minimal ) basis , and to perform geometry optimization for a nanotube that contains more than @xmath9 atoms with a double-@xmath8 plus polarization ( dzp ) basis .
however , the sequential implementation does not benefit from the inherent parallelism in the pexsi method , and therefore leads to limited or no improvement for general electronic structure calculations .
the contribution of this paper is twofold : 1 ) we present the siesta - pexsi method , which combines the siesta method @xcite based on numerical atomic orbitals and the recently developed massively parallel pexsi method @xcite .
the siesta - pexsi method can be efficiently scalable to more than @xmath0 processors .
we provide performance data for a range of systems , including strong and weak scaling characteristics , and illustrate the crossover points beyond which the new approach is more efficient than diagonalization .
the accuracy of the result obtained from the siesta - pexsi method is nearly indistinguishable from the result obtained from the diagonalization method .
2 ) we develop a hybrid scheme of density of states estimation and newton s method to obtain the chemical potential .
we demonstrate that the scheme is highly efficient and robust with respect to the initial guess , with or without the presence of gap states .
the siesta - pexsi approach has been implemented as a new solver in siesta , with built - in heuristics that balance efficiency and accuracy , but at the same time offering full control by the user .
this paper is organized as follows . in section
[ sec : theory ] , we describe the massively parallel pexsi technique , and how to integrate it with the siesta method
. we also present a new method to update the chemical potential . in section [ sec : numerical ] , we report the performance of the siesta - pexsi method on several problems . throughout the paper ,
we use @xmath10 to denote the imaginary part of a complex matrix @xmath11 .
we use @xmath12 to denote the discretized hamiltonian matrix and the corresponding overlap matrix obtained from a basis set @xmath13 such as numerical atomic orbitals .
similarly @xmath14 denotes the single particle density matrix operator , and the corresponding electron density is denoted by @xmath15 .
the matrix @xmath16 denotes the single particle density matrix represented in the @xmath13 basis . in pexsi ,
@xmath16 and related matrices are approximated by a finite @xmath17-term pole expansion , denoted by @xmath18 respectively .
however , to simplify notation , we will drop the subscript @xmath17 and simply use @xmath19 to denote the approximated matrices unless otherwise noted .
the ground - state electron charge density @xmath15 of an atomistic system can be obtained from the self - consistent solution to the kohn - sham equations @xmath20 \psi_i(x ) = \psi_i(x ) \varepsilon_i , \label{kseqs}\ ] ] where @xmath21 is the kohn - sham hamiltonian that depends on @xmath15 , @xmath22 are the kohn - sham orbitals which in turn determine the charge density by @xmath23 with occupation numbers @xmath24 that can be chosen according to the fermi - dirac distribution function @xmath25 where @xmath26 is the chemical potential chosen to ensure that @xmath27 and @xmath28 is the inverse of the temperature , i.e. , @xmath29 with @xmath30 being the boltzmann constant . the most straightforward method to solve
the kohn - sham problem is to expand the orbitals @xmath31 as a linear combination of a finite number of basis functions @xmath32 , and thus recast as a ( generalized ) eigenvalue problem within an iterative procedure to achieve self - consistency in the charge density .
the computational complexity of this approach is @xmath33 , where @xmath34 is the number of basis functions and is generally proportional to the number of electrons @xmath2 or atoms in the system to be studied .
this approach becomes prohibitively expensive when the size of the system increases .
formally , the electronic - structure problem can be recast in terms of the one - particle density matrix defined by @xmath35 with @xmath26 chosen so that @xmath36 = n_e$ ] .
@xmath37 can thus be evaluated without the need for diagonalization , if the fermi function is approximated by a linear combination of a number of simpler functions .
this is the fermi operator expansion ( foe ) method @xcite . while most of the foe schemes require as many as @xmath38 or @xmath39 terms of simple functions ( with @xmath40 the spectrum width ) , the recently developed pole expansion @xcite is particularly promising since it requires only @xmath41 terms of simple rational functions .
the pole expansion has the analytic expression @xmath42 we refer readers to ref . for more details .
the complex shifts @xmath43 and weights @xmath44 are determined only by @xmath45 and the number of poles @xmath17 .
all quantities in the pole expansion are known explicitly and their calculation takes negligible amount of time .
following the derivation in ref . , we can use ( [ eqn : polerho ] ) to approximate the single particle density matrix @xmath37 by its @xmath17-term pole expansion , denoted by @xmath46 as @xmath47 where @xmath48 $ ] is a collective vector of basis functions , @xmath49 , @xmath50 , and @xmath16 is an @xmath51 matrix represented in terms of the @xmath13 basis . to simplify our notation ,
we will drop the subscript @xmath17 originating from the @xmath17-term pole expansion approximation unless otherwise noted .
it would seem that the need to carry out @xmath17 matrix inversions in ( [ eqn : gammapole ] ) would mean that the computational complexity of this approach is still close to the @xmath33 scaling of diagonalization .
however , what is really needed in practice is just the electron density in real space , that is @xmath52 when the basis functions @xmath53 are compactly supported in real space , the product of two functions @xmath54 and @xmath55 will be zero when they do not overlap .
these @xmath56 pairs can be excluded from the summation in eq . .
consequently , we _ only _ need @xmath57 such that @xmath58 in eq . .
this set of @xmath57 s is a subset of @xmath59 . to obtain these _ selected elements _ , we need to compute the corresponding elements of @xmath60 for all @xmath61 .
we emphasize that we compute the selected elements of the density matrix because only these elements are needed to compute physical quantities such as charge density , energy and forces , due to the localized character of the basis set .
the computed selected elements of the density matrix are accurate , and should be regarded as if we performed a conventional @xmath33 calculation first , and then only kept the corresponding selected elements of the density matrix . in principle
we could retrieve any matrix element of the density matrix , simply by enlarging the set of `` selected elements '' .
this process is fundamentally different from the usage of `` near - sightedness '' approximation , which throws away the information of the density matrix beyond the truncation region .
the recently developed selected inversion method @xcite provides an efficient way of computing the selected elements of an inverse matrix . for a ( complex ) symmetric matrix of the form @xmath62 , the selected inversion algorithm first constructs an @xmath63 factorization of @xmath11 , where @xmath64 is a block lower triangular matrix called the cholesky factor , and @xmath65 is a block diagonal matrix .
the computational scaling of the selected inversion algorithm is only proportional to the number of nonzero elements in the cholesky factor @xmath64 , which is @xmath66 for quasi-1d systems , @xmath67 for quasi-2d systems , and @xmath68 for 3d bulk systems , thus achieving universal asymptotic improvement over the diagonalization method for systems of all dimensions .
it should be noted that the selected inversion algorithm is an _ exact _
method for computing selected elements of @xmath69 if exact arithmetic is employed , and in practice the only source of error originates from the roundoff error .
in particular , the selected inversion algorithm does not rely on any localization property of @xmath69 .
in addition to computing the charge density at a reduced computational complexity in each scf iteration , we can also use this pole - expansion and selected - inversion ( pexsi ) technique to compute the free energy and the atomic forces efficiently without diagonalizing the kohn - sham hamiltonian . following the derivation in ref .
, the relevant expressions are : @xmath70
+ \mu n_e - \frac12 \iint \frac{{\rho}(x ) { \rho}(y)}{{\left\lvertx - y\right\rvert } } { \,\mathrm{d}}x { \,\mathrm{d}}y\\ & + e_{{{\mathrm{xc}}}}[{\rho } ] - \int v_{{{\mathrm{xc}}}}[{\rho } ] { \rho}(x ) { \,\mathrm{d}}x , \end{split } \label{eqn : helmholtzpole}\ ] ] @xmath71 + { \mathrm{tr}}\left [ \gamma^e \frac{\partial s}{\partial r_i } \right ] .
\label{eqn : forcepole}\ ] ] where the energy and free - energy density matrices @xmath72 and @xmath73 are given by pole expansions with _ the same poles _ as those used for computing the charge density : @xmath74 since the trace terms in and require only the @xmath75th entries of @xmath76 for @xmath75 satisfying @xmath77 or @xmath78 , the needed elements of the energy - density matrices can be computed without additional complexity .
in addition to its favorable asymptotic complexity , the pexsi method is also inherently more scalable than the standard approach based on matrix diagonalization when it is implemented on a parallel computer .
the parallelism in pexsi exists at two levels .
first , the selected inversions associated with different poles ( usually on the order of @xmath79 ) are completely independent .
second , each selected inversion itself can be parallelized by using the parallel selected inversion method called pselinv @xcite .
a parallel selected inversion consists of the following steps : 1 . the rows and columns of the matrices @xmath80 and @xmath81 are reordered to reduce the number of nonzeros in the triangular factor of the @xmath63 decomposition of @xmath82 .
2 . a parallel symbolic factorization of @xmath82
is performed to identify the location of the nonzero matrix elements in @xmath64 .
the numerical @xmath63 decomposition ( or equivalent @xmath83 decomposition ) of @xmath84 is performed .
the desired selected elements of @xmath85 are computed from @xmath64 and @xmath65 .
step 1 can be performed in parallel by using the parmetis @xcite or the pt - scotch @xcite software packages .
its cost is much smaller compared to the numerical factorization , and only needs to be done once per scf cycle .
although for symmetric matrices only @xmath86 factorization is needed , the pexsi package currently use the superlu_dist software package @xcite to perform steps 2 and 3 in parallel with @xmath83 factorization .
the @xmath83 factorization contains equivalent information as in @xmath86 factorization but can be twice as expensive in the worst case due to the lack of usage of the symmetry property of the matrix .
pexsi and pselinv has a independent data structure , and allows to be interfaced with other sparse direct solvers .
the cost of symbolic factorization in step 2 is usually much lower than the numerical factorization .
the numerical factorization procedure can be described in terms of the traversal of a tree called the _
elimination tree_. each node of the tree corresponds to a block of continuous columns of @xmath87 .
a node @xmath88 is the parent of a node @xmath89 if and only if @xmath90 in superlu_dist , each node is distributed among a subset of processors block cyclically .
the traversal of the elimination tree proceeds from the leaves towards the root .
the update of @xmath91 is performed in parallel on a subset of processors assigned to @xmath91 and its children nodes , and the main operations involved in the update are a number of dense matrix - matrix multiplications .
in addition to parallelism within the update of a supernode , additional concurrency can be exploited in the traversal of different branches of the elimination tree for updating different supernodes .
all of these techniques can be used in step 4 to compute selected elements of @xmath92 . in this step , the elimination tree is traversed from the root down towards the leaves .
as each node is traversed , selected elements of @xmath92 within the columns that are mapped to that node are computed through a number of dense matrix - matrix multiplications .
communication is needed among processors that are mapped to the node and its ancestors .
multiple nodes belonging to different branches of the elimination tree can be traversed simultaneously if the update of these nodes do not involve communications with the same ancestor . for sufficiently large problems
, superlu_dist can achieve substantial speedup in the numerical factorization when hundreds to thousands of processors are used .
similar or better speedup factors can be observed when selected elements of @xmath92 are computed from the distributed @xmath64 and @xmath65 factors .
it is shown that for large matrices a single selected inversion can scale to @xmath93 cores , and if @xmath94 poles are used in the pole expansion of the fermi - dirac function , then total number of computational cores that can be efficiently utilized by the parallel pexsi method is @xmath95 .
we will give some more concrete examples in section [ sec : numerical ] to demonstrate the performance of our implementation of the parallel pexsi algorithm .
compared to pexsi , it is generally difficult to efficiently use that many computational cores in a dense matrix calculation algorithm such as those implemented in scalapack .
the main bottleneck of the computation using scalapack is the reduction of a dense matrix to a tridiagonal matrix and the back transformation of the eigenvectors of a tridiagonal matrix , which are inherently sequential .
moreover , the cost of the diagonalization method scales cubically with respect to the matrix dimension .
this limits the size of the matrix that can be handled by scalapack , as well as the parallel scalability on massively parallel computers .
although some recent progress has been made to make these transformations more efficient @xcite , making diagonalization scalable on more than tens of thousands of cores remains a challenging task .
the chemical potential @xmath26 required in the pole expansion is not known a priori .
its value must in general be determined so that the total number of electrons is appropriate : @xmath96 = { \mathrm{tr}}[\gamma \phi^t \phi ] = { \mathrm{tr}}[\gamma s ] .
\label{eqn : nmu}\ ] ] as the right - hand - side is a non - decreasing function of @xmath26 , the chemical potential can be efficiently obtained by newton s method , maybe combined with extra safeguards such as bisection .
the required derivative @xmath97 can be computed with very little extra cost using the pole expansion of the derivative of the fermi - dirac distribution @xmath98 , which can be constructed , for the reasons illustrated in ref . , by using the same shifts @xmath61 as those in : @xmath99 , \label{}\ ] ] @xmath100 when newton s method is used , the convergence of @xmath26 is rapid near the correct chemical potential .
however , the standard newton s method may not be robust enough when the initial guess is far away from the correct chemical potential .
it may give , for example , too large a correction when @xmath101 is close to zero , as when @xmath26 is near the edge or in the middle of a band gap .
one way to overcome the above difficulty is to use an approximation to the function @xmath102 to narrow down the region in which the correct @xmath26 must lie .
this function can be seen effectively as a ( temperature smeared ) cumulative density of states , counting the number of eigenvalues in the interval @xmath103 .
we can evaluate @xmath104 , its zero - temperature limit , without computing any eigenvalues of @xmath105 .
instead , we perform a matrix decomposition of the shifted matrix @xmath106 , where @xmath64 is unit lower triangular and @xmath65 is diagonal .
it follows from sylvester s law of inertia @xcite , which states that the inertia ( the number of negative , zero and positive eigenvalues ) of a real symmetric matrix does not change under a congruent transform , that @xmath65 has the same inertia as that of @xmath107 .
hence , we can obtain @xmath104 by simply counting the number of negative entries in @xmath65 .
note that the matrix decomposition @xmath108 can be computed efficiently by using a sparse @xmath86 or @xmath83 factorization in real arithmetic .
it requires fewer floating point operations than the complex arithmetic direct sparse factorization used in pexsi . to estimate @xmath109 for a finite @xmath28
, we use the identity @xmath110 and perform an integration by parts to obtain @xmath111 the integral in can be evaluated numerically by sampling @xmath112 and @xmath104 at a number of quadrature points @xmath113 and performing a weighted sum of @xmath114 for @xmath115 .
the @xmath116 evaluations of @xmath117 can be performed simultaneously using the @xmath63 factorization - based inertia counting procedure described above , with @xmath116 groups of processors . since the derivative of the fermi - dirac function is sharply peaked , we can approximate @xmath109 in a given interval by sampling @xmath104 in a slightly wider interval . fig .
[ fig : ne ] shows the number of electrons at zero temperature @xmath104 obtained from inertia counting procedure , and the interpolated finite temperature profile @xmath118 at @xmath119 k for a dna system ( with finite gap ) and a sih system ( with zero gap ) near the fermi energy . while the finite temperature smearing effect is negligible for insulators at @xmath119 k , it is more pronounced for metals and leads to a more smooth @xmath118 which is suitable for applying newton s method to find the chemical potential . on the other hand ,
the inertia counting procedure obtains the global profile of the cumulative density of states , and does not suffer from the problem of being trapped in intermediate band gaps .
this approximation to the function @xmath118 is then available for use in determining the approximate placement of the true @xmath26 by simple root finding : @xmath120 . in practice , we need to start from an interval which is large enough to contain the chemical potential . for systems with a gap
the interval should also contain the gap edges . in this case ,
two roots are sought : @xmath121 and @xmath122 , where @xmath123 is small , say 0.1 , and so @xmath124 and @xmath125 will be estimates of the band edges from which @xmath26 can be determined as @xmath126 .
the fidelity of @xmath118 , and thus the quality of the estimation of @xmath26 , depends on the density with which the interval can be sampled , which with a fixed number @xmath116 of sampling points for @xmath104 will increase as the interval is narrowed . in practice
, the above procedure is repeated with progressively smaller intervals until either the @xmath26 estimate stabilizes or the size of the interval is small enough .
unless the original interval is very large , 2 or 3 inertia counts are enough to provide an adequate @xmath26 from which to start the pexsi process , and the number of inertia counts can be one or even zero ( i.e. the inertia counting procedure is turned off ) in subsequent scf iterations .
newton s method in the solver then takes over for the final refining of @xmath26 .
the hybrid procedure of inertia counting and newton s method is efficient and robust for both insulating and metallic systems .
this will be further demonstrated in section [ sec : numerical ] .
the density of states @xmath127 is defined as @xmath128 the factor of two comes from spin degeneracy . as advanced in the previous section ,
the cumulative density of states ( cdos ) @xmath129 is exactly the function @xmath104 discussed there . to evaluate @xmath127
, we sample @xmath104 at a set of @xmath130 using the inertia counting procedure , and use an appropriate interpolation scheme to approximate @xmath104 for other values of @xmath131 . the approximation to the dos
is then obtained by numerical differentiation .
we remark that inertia counting is not the only diagonalization - free method for computing the dos .
other methods that make use of matrix vector multiplications only are also possible ( see _ e.g._a recent review @xcite ) .
another physical quantity that can be easily approximated via selected inversion is the localized density of states ( ldos ) , defined as @xmath132 and representing in fact the contribution to the charge density of the states with eigenvalues in the vicinity of @xmath131 , as filtered by the @xmath133 function .
when a finite basis set @xmath13 is used , the ldos can be represented as @xmath134 where @xmath135 is a matrix of the coefficients of the expansion in @xmath13 of the orbitals @xmath136 , and @xmath137 is a diagonal matrix of the eigenvalues .
it follows from the sokhotski - plemelj formula @xcite @xmath138 that @xmath139 where symbol @xmath140 in stands for the cauchy principal value . combining eq . and
, we obtain the following alternative expression for the ldos : @xmath141^{-1 } \phi^{t}(x ) .
\end{split } \label{eqn : ldosgreen}\ ] ] note that eq . allows us to compute the ldos without using any eigenvalue or eigenvector . in practice ,
we take @xmath142 to be a small positive number , and the local dos can be approximated by @xmath143^{-1}_{ij } , \label{eqn : ldosgreen}\ ] ] which is similar to the computation of the electron density .
again only the selected elements of the matrix @xmath144^{-1}$ ] are needed for the computation of the ldos for each @xmath131 , and the selected elements can be obtained efficiently by the pselinv procedure .
siesta is a density functional theory code which uses finite - range atomic orbitals to discretize the kohn - sham problem @xcite , and thus handles internally sparse h , s , and single - particle density matrices .
siesta is therefore well suited to implement the pexsi method along the lines explained above .
the pexsi method can be directly integrated into siesta as a new kind of electronic - structure solver .
conceptually , the interface between siesta and pexsi is straightforward .
the existing siesta framework takes care of setting up the basis set and of constructing the sparse @xmath80 and @xmath81 matrices at each iteration of the self - consistent - field ( scf ) cycle . @xmath80 and @xmath81 are passed to the pexsi module , which returns the density matrix @xmath16 and , optionally , the energy - density matrix @xmath145 ( needed for the calculation of forces ) and the @xmath73 matrix that can be used to estimate the electronic entropy .
siesta then computes the charge density to generate a new hamiltonian to continue the cycle , until convergence is achieved .
energies and forces are computed as needed .
the details of the interface are controlled by a number of parameters which provide flexibility to the user , especially in regard to the bracketing and tolerances involved in the determination of the chemical potential , and in the context of parallel computation .
we list some of the more relevant parameters in table [ tab : param ] , but we should note that even finer control is possible by other , more specialized parameters .
.main siesta - pexsi parameters [ cols="^,^",options="header " , ] another way to study the weak scaling of the siesta - pexsi approach is to examine how the total computational cost , which is the product of the wallclock time and the number of processors ( i.e. , the first two columns for each method in table [ tab : exampleweakscaling ] ) , change with respect to the problem size . for perfect weak
scaling , the change in cost should match that predicted by the computational complexity of the sequential algorithm .
we plot the cost involved in solving each instance of the dna , c - bn and h@xmath146o problems in fig .
[ fig : efficiencyweakscaling ] , which shows that the data for each series falls approximately on a line .
the ( fitted ) slope of the cost for dna , which includes both factorization and selected inversion , is observed to be approximately 1.3 , higher than the @xmath147 ideal scaling for quasi-1d systems . for c
- bn the slope of the fitting line is approximately 1.7 , which is a bit larger than the @xmath148 ideal scaling for quasi-2d systems .
the slope of the line for the h@xmath146o series is approximately 2.18 , again slightly larger than the expected @xmath149 scaling for 3d systems .
the observed degradation in parallel scaling ( relatively larger for dna ) comes about because neither selected inversion nor factorization ( which has the same asymptotic complexity ) can scale perfectly due to the complicated data communication and task dependency involved .
it should be noted also that the ideal weak - scaling data in table [ tab : exampleweakscaling ] depends on the size and processor count chosen for the smallest instance of each series .
we have already mentioned that all systems considered in our timings have been run with complete parallelization over poles .
this limits the gains from increasing processor counts to the actual efficiency of selected - inversion and factorization within a pole , as a function of @xmath150 .
roughly , the computational load for a given instance will depend on the number of non - zeros in the cholesky factor @xmath64 , which is plotted as a function of matrix size in fig . [ fig : cholesky_nnz ] .
it can be seen that c - bn has the highest such non - zero density , closely followed by h@xmath146o .
this is because c - bn has a more densely packed structure than h@xmath151o , and thus a larger prefactor , but there is a clear trend for a faster increase in the number of non - zeros in the cholesky factor for h@xmath151o , in agreement with the asymptotic complexity of 3d and quasi-2d systems , respectively .
the dna system has significantly fewer non - zero values in l for a given size of h , and so it can use a relatively smaller number of processors per pole efficiently . for the systems shown on fig .
[ fig : efficiencyweakscaling ] , as a function of the dimension of @xmath80 . ]
we can also see from fig .
[ fig : efficiencyweakscaling ] that the crossover point at which siesta - pexsi becomes more efficient ( in terms of cost ) than the standard diagonalization procedure in siesta is around matrix dimension @xmath34=7000 ( @xmath152700 atoms ) for the dna problem , @xmath34=50,000 ( @xmath1524,000 atoms ) for the c - bn problem and @xmath34=40,000 ( @xmath1525,000 atoms ) for the h@xmath146o problem .
the actual orderings of the crossover points depend on the dimensionality ( through the power of n in the scaling ) , and also on a prefactor which is system dependent .
since more processors can be used efficiently , the advantage of siesta - pexsi in time - to - solution performance is even more pronounced than the benefit from the reduction of total work load , and is already evident for smaller systems , as can be seen in table [ tab : exampleweakscaling ] .
the conclusions regarding the performance comparison between pexsi and diagonalization , being based on the asymptotic scaling of the algorithm , remain largely valid even taking into account the efficiency improvements that can been obtained by refactoring some of the internals of scalapack @xcite .
number of processors ) for the first scf step for the h@xmath146o , c - bn , and dna examples .
the points correspond to the various problem sizes as listed in table [ tab : examplefeatures ] .
the times for diagonalization show some variations due to differences in the construction of @xmath80 and the building of the @xmath16 matrix , and are represented by the gray stripe .
the configurations and results can be found in more detail in table [ tab : exampleweakscaling ] . ] as we discussed in section [ subsec : chemicalpotential ] , the current implementation of siesta - pexsi uses a combination of an inertia counting bracketing technique and newton s method to determine the chemical potential @xmath26 that satisfies the nonlinear equation @xmath153 , where @xmath154 is defined in . in this section
, we demonstrate the effectiveness of this approach for both metallic and insulating systems .
the metallic example is the sih system with @xmath155 electrons . since our calculation is spin - restricted , the odd number of electrons is solely controlled through the fractional occupation of the h - derived state in the gap .
any small perturbation of the chemical potential can change the number of electrons , and the perturbation is temperature dependent ( we use 300 k ) . despite the relatively small system size ,
this can be considered as the most difficult case for finding the chemical potential without having access to the eigenvalues .
[ fig : convsih ] illustrates the number of inertia counting cycles and pexsi solver steps needed to obtain a chemical potential that satisfies @xmath156 for the converged sih system when using 60 poles .
the calculation starts from a wide initial @xmath26 search interval of @xmath157 ev , and uses an adaptive electron tolerance which starts at a coarse @xmath158=0.1 and progressively tightens towards the target value of @xmath159 as the deviation from self - consistency in the h matrix elements moves towards its tolerance target of @xmath160 ry .
the number of inertia counting steps needed to narrow down the interval that contains the true chemical potential is roughly @xmath161 in the first few scf steps . after the @xmath162th iteration
, the inertia counting procedure is turned off .
the number of pexsi solver steps needed is at most @xmath163 in the first few scf steps , and it becomes 2 after the @xmath164th scf step .
[ fig : convsih ] also shows the absolute error @xmath165 at each scf step , which is always below the set tolerance .
our test for insulating systems is the dna-1 system with @xmath166 electrons , at @xmath119 k and using 50 poles .
[ fig : convdna1 ] illustrates the number of inertia counting and pexsi solver steps needed to reduce @xmath167 to a level below a final tolerance of @xmath159 , starting at a coarser tolerance of @xmath168 .
the calculation also starts from a wide initial interval of @xmath157 ev .
usually two inertia counting steps are needed in the first few scf iterations .
the second iteration requires one more due to a large jump in h. after the @xmath162th iteration , the inertia counting procedure is completely turned off . throughout the scf cycle , which ends when the deviation from self - consistency in h is below @xmath160 ry
, just one pexsi solver call is needed to provide a solution with a number of electrons within the tolerance .
this is because @xmath26 is kept within the gap by the bracketing procedure . in section [ subsec : dos ] , we described how to use siesta - pexsi to obtain spectral information of the atomistic system such as dos and ldos without computing eigenvalues and eigenvectors of @xmath105 . in fig .
[ fig : dos ] ( a ) we plot the dos obtained from the standard diagonalization method and the inertia counting procedure implemented in siesta - pexsi for the c - bn@xmath169 system near the fermi level ( `` e@xmath170 '' ) .
we observe nearly perfect agreement between the dos curves obtained from these two approaches .
[ fig : dos ] ( b ) shows the dos near the fermi level for the dna-25 system using both the inertia counting method and diagonalization .
the usage of pexsi for evaluating dos without obtaining eigenvalues is also significant for systems at large size . for the dna-25 system ,
using 64 @xmath150 and 3200 processors to evaluate @xmath171 points in the dos near the fermi energy ( each group of 64 processors evaluate @xmath172 points of cumulative dos with inertia counting ) only takes @xmath173 sec , while diagonalization using the same number of processors takes @xmath174 sec .
it should be noted that the performance of pexsi can further be improved by simply using more cores : after the scf converges , one can restart the calculation to compute the dos using the converged density matrix , and use @xmath175 processors to compute @xmath171 points in the dos in parallel . in such case , the wall clock can be reduced to @xmath176 sec .
[ fig : ldos ] presents the ldos for the sih system for an energy interval of width 0.4 ev around the fermi level , showing the state due to the added h atom in the bulk si system .
we have combined the pole expansion and selected inversion ( pexsi ) technique with the siesta method for kohn - sham density functional theory ( kdsft ) calculation .
the resulting siesta - pexsi method can efficiently use more than @xmath0 processors , and is particularly suitable for performing large scale _ ab initio _
materials simulation on high performance parallel computers .
the siesta - pexsi method does not compute eigenvalues or eigenvectors of the kohn - sham hamiltonian , and its accuracy is fully comparable to that obtained from the standard matrix diagonalization based siesta calculation for general systems , including insulating , semi - metallic , and metallic systems at low temperature .
the current implementation of siesta - pexsi does not yet support spin - polarized systems , but this can be achieved with minor changes to the code .
furthermore , the code does not support * k*-point sampling , as in principle it is not needed for large - enough systems .
however , in some cases , as in graphene and similar semi - metallic systems , an appropriate computation of the spectral properties such as the dos does need * k*-point sampling even when very large supercells are used .
we will address this problem in future work , which requires modifying pexsi to handle complex hermitian @xmath80 matrices .
in such case the shifted matrix @xmath177 is no longer symmetric but is only structurally symmetric .
this further development will also be useful for _ ab initio _ study of non - collinear magnetism and spin - orbit coupling effects .
this work was partially supported by the laboratory directed research and development program of lawrence berkeley national laboratory under the u.s .
department of energy contract number de - ac02 - 05ch11231 , by scientific discovery through advanced computing ( scidac ) program funded by u.s .
department of energy , office of science , advanced scientific computing research and basic energy sciences , by the center for applied mathematics for energy research applications ( camera ) , which is a partnership between basic energy sciences and advanced scientific computing research at the u.s department of energy ( l. l. and c. y. ) , by the european community s seventh framework programme [ fp7/2007 - 2013 ] under the prace project grant agreement number 283493 ( g. h. ) , and by the spanish mineco through grants fis2009 - 12721-c04 - 03 , fis2012 - 37549-c05 - 05 and csd2007 - 00050 ( a. g. ) .
38ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1016/j.parco.2011.05.002 [ * * , ( ) ] @noop * * , ( ) @noop ( ) @noop _ _ ( , , ) | we describe a scheme for efficient large - scale electronic - structure calculations based on the combination of the pole expansion and selected inversion ( pexsi ) technique with the siesta method , which uses numerical atomic orbitals within the kohn - sham density functional theory ( ksdft ) framework .
the pexsi technique can efficiently utilize the sparsity pattern of the hamiltonian and overlap matrices generated in siesta , and for large systems has a much lower computational complexity than that associated with the matrix diagonalization procedure .
the pexsi technique can be used to evaluate the electron density , free energy , atomic forces , density of states and local density of states without computing any eigenvalue or eigenvector of the kohn - sham hamiltonian .
it can achieve accuracy fully comparable to that obtained from a matrix diagonalization procedure for general systems , including metallic systems at low temperature .
the pexsi method is also highly scalable . with the recently developed massively parallel pexsi technique , we can make efficient use of more than @xmath0 processors on high performance machines .
we demonstrate the performance and accuracy of the siesta - pexsi method using several examples of large scale electronic structure calculations , including 1d , 2d and bulk problems with insulating , semi - metallic , and metallic character . |
even after a cursory reading of the papers published in the seminal vatican conference on stellar populations it becomes clear the key role that rr lyrae played in the development of what we now call stellar populations and in the shaping of galactic structure .
their relevance is also supported by the first sentence of one the papers given by w. baade at that conference : _ variable stars , in particular the cepheids and the cluster - type variables , have becoming increasingly important in the exploration of our own and other galaxies . _ the relevance of rr lyrae stars in stellar astrophysics is further supported by the fact that rr lyrae are the most popular old , low - mass distance indicators . in early times
it was assumed that their mean luminosity was constant , but soon after the spectroscopic investigations showed that they do obey a visual magnitude metallicity relation .
therefore , the knowledge of their metallicity was fundamental for accurate estimates of their individual distances . however , high - resolution spectroscopy in single slit mode was prohibitive for large numbers of rr lyrae stars .
the @xmath3s method invented by george @xcite opened a new path in the extensive use of field rr lyrae as stellar tracers of the galactic halo @xcite and of the bulge @xcite .
rr lyrae are also fundamental laboratories to constrain the physical mechanisms that drive their pulsation instability . after the pioneering investigations by @xcite based on nonlinear radiative models ,
the modelling of rr lyrae pulsation behavior was at the cross - roads of several theoretical and observational investigations .
once again the papers by @xcite and @xcite have been a benchmark for those interested in understanding the impact that the different physical mechanisms ( convection ) and input physics ( opacity , equation of state ) have on light and radial velocity curves .
the structure of this paper is the following . in
2 we briefly discuss pros and cons of cluster and field rr lyrae . the pulsation framework developed by our group during the last few years
is outlined in 3 , together with the added values and the drawbacks of the bailey diagram ( amplitude versus period ) . in 5
we discuss the current observational scenario of the k - band period - luminosity relation of cluster rr lyrae . in 6 we discuss the diagnostics currently adopted to estimate the primordial helium content , in particular we focus our attention on rr lyrae stars . in the last section we briefly mention the possible paths that the next generation of observing facilities will open during the next few years .
field and cluster rr lyrae bring forward several distinctive advantages worth discussing in detail : _ i)_distance cluster rr lyrae provide accurate distance estimates of galactic globular clusters ( ggcs ) , and in turn more accurate estimates of their absolute age .
unfortunately , they are not ubiquitous , i.e. their occurrence in ggcs does depend on the morphology of the horizontal branch .
they are not present in ggcs that have either a very red or a very blue hb morphology .
this means that they are affected by the so - called second parameter problem ( e.g. , kunder et al .
2011 ) . _ ii)_ensemble properties cluster and field rr
lyrae stars do show different period distributions , i.e. they are affected by the so - called oosterhoff dichotomy @xcite .
however , rr lyrae in mc globulars are oosterhoff intermediate @xcite .
the same outcome applies to rr lyrae in different dwarf spheroidal galaxies in the local group ( e.g. carina , @xcite ) .
the fraction of first overtone ( fo , rrc ) and fundamental ( f , rrab ) variables is also affected by the hb morphology and changes when moving from ggcs to dwarf galaxies to the galactic field @xcite . this indicates that pulsation properties of rr lyrae do depend on the chemical and dynamical evolution of their stellar environment . _
iii)_evolutionary status empirical evidence and theoretical predictions indicate that the range in visual magnitude between the zero - age - horizontal - branch ( zahb ) and the end of central helium burning is correlated with the metallicity @xcite .
together with this intrinsic evolutionary effect , we are also facing the problem that we still lack a firm photometric diagnostic to constrain the evolutionary status of individual rr lyrae stars .
this drawback applies to both cluster and field rr lyrae and causes a series of problems not only in the distance estimate , but also in the estimate of the observed zahb luminosity level , and in turn on the parameters correlated with this evolutionary feature , namely the r parameter ( @xcite ; troisi et al .
2011 , in preparation ) , the @xmath3@xmath4 @xcite and the relative ages of ggcs @xcite . a possible solution to overcome
this problem is to estimate the individual distances using the nir pl relations , since they are minimally affected by evolutionary effects and by a possible spread in mass inside the instability strip @xcite . on the basis of the true distance modulus and of their individual reddenings ,
the same objects can be located in the absolute cmd ( @xmath5 vs @xmath6 ) to constrain their evolutionary status ( bono et al .
2011 , in preparation ) . _
iv)_evolutionary / pulsation connection cluster variables are also relevant to constrain the input physics adopted in pulsation and evolutionary codes .
observables predicted by pulsation models ( periods , modal stability , pulsation amplitudes ) and by evolutionary models ( lifetimes , mass - luminosity relation , effective temperature ) do depend on the same intrinsic parameters ( stellar mass , chemical composition ) .
the comparison between theory and observations using the same objects ( rr lyrae ) , seen as variables and as hb stars , provide a fundamental sanity check for both the micro ( opacity , equation of state , cross sections ) and the macro ( gravitational settling , mixing , mass - loss ) physics .
this is a fundamental stepping stone to improve the plausibility of the physical assumptions adopted in pulsation and evolutionary models , since the former rely on the mass - luminosity relation predicted by evolutionary models .
this is the path that has been followed to address several open problems ( oosterhoff dichotomy , sandage period effect , hysteresis mechanism , topology of the instability strip ) that are at the cross - roads between theory and observations .
_ iv)_statistics and progenitors the number of ggcs that host several tens of rr lyrae stars is quite limited @xcite .
the new photometric surveys are disclosing several thousands of field rr lyrae .
this means that robust comparisons can only be based on field stars .
however , for cluster variables we have accurate knowledge of the absolute age and chemical composition of the progenitors . for field
rr lyrae we can either estimate ( photometric indices ) or measure ( spectroscopy ) the chemical composition , but we can barely constrain their absolute age , and in turn the mass of the progenitor . _
v)_helium abundance more than 30 years ago it was suggested that the mass - luminosity ratio of rr lyrae , the so - called a - parameter @xcite can be adopted , via the pulsation relation @xcite , to estimate the helium content of individual variables @xcite .
the application of this diagnostic to cluster rr lyrae is very powerful , since the precision is only limited by the size of the sample .
the two main drawbacks of this approach is that the he estimates are affected by the precision of the color - temperature relation and by the evolutionary status of individual objects .
a breath of fresh air on this delicate topic arrived with the discovery by @xcite of both he i and he ii emission and absorption lines in field rr lyrae stars .
the lines show up during the rising branch of rr lyrae stars and appear to be the consequence of the shock propagation @xcite .
the content of 2 indicates that rr lyrae stars are the cross - roads of several open astrophysical problems .
several of these problems call for new theoretical and observational insights . in particular ,
the theoretical approach does require a theoretical framework dealing with both evolutionary and pulsation properties of radial variables .
our group undertook this project almost twenty years ago @xcite . in the following , we briefly mention recent findings concerning the predicted bailey diagram and the k - band period - luminosity relation .
we constructed new sets of pulsation models by using the hydrodynamical code developed by @xcite and updated by @xcite [ see also @xcite for a similar approach ] , @xcite . the physical assumptions adopted to compute these models will be described in marconi et al .
( 2011 , in preparation ) .
we adopted the opal radiative opacities released in 2005 by @xcite and the molecular opacities by @xcite . to properly constrain the pulsation properties of stars , we typically cover a wide range in metal abundances ( scaled - solar , @xmath7-enhanced ) .
moreover , to constrain the dependence of the pulsation properties of stars on the he abundance we also adopt different values of this crucial parameter . for each fixed chemical composition
the stellar mass of stars was fixed by using the evolutionary prescription for @xmath7-enhanced structures provided by @xcite and available on the basti database .
note that for each fixed metal content the mass of the he - enhanced models was estimated assuming the same cluster age ( 13 gyr ) . together with the luminosity predicted by evolutionary models we often adopted a brighter luminosity level to account for the possible occurrence of evolved stars .
the reader interested in a more detailed discussion concerning evolutionary and pulsation ingredients is referred to ( * ? ? ?
* and references therein ) .
the period distribution and the bailey diagram ( luminosity amplitude vs pulsation period ) are very robust observables , since they are independent of distance and reddening corrections .
after the seminal discovery of this diagram @xcite , @xcite found that objects with different metal abundances display different trends .
the use of the pulsation amplitude as a proxy of the metal content is still lively debated both from the observational ( @xcite ) and the theoretical @xcite point of view .
the reader interested in a detailed discussion concerning the development and the use of the bailey diagram is referred to the detailed paper by smith et al .
( these proceedings ) . in the following ,
we focus our attention on the use of the bailey diagram to constrain the helium content of rr lyrae stars .
the bolometric light curves predicted by nonlinear , convective , pulsation models are typically transformed into the observational plane using the bolometric corrections and the color - temperature transformations provided by castelli et al .
( 1997a , b ) .
figure 1 shows predicted b - band amplitudes for fundamental ( top ) and first overtone ( bottom ) rr lyrae models computed at fixed metal content ( z=0.001 ) and primordial helium contents ranging from y=0.24 to y=0.38 ( see labeled values ) . to constrain the dependence of the luminosity amplitude on evolutionary effects the models for each fixed chemical composition
were constructed using two different luminosity levels .
data plotted in this figure indicate that the he content marginally affects the pulsation behavior of rr lyrae .
the top panel shows that an increase of 30% in he content ( 0.24 vs 0.30 ) causes a systematic shift in the pulsation period .
thus , suggesting that the difference in the helium enhanced models is mainly due to the increase in the luminosity level .
the different sets of models disclose that structures with a difference in helium content of @xmath150% might have , at fixed period , similar pulsation amplitudes .
the amplitudes of the he enhanced models y=0.30 , y=0.38 display minimal changes , since the latter group was constructed assuming the same luminosity levels and slightly smaller stellar masses ( 0.60 vs 0.65 @xmath8 ) .
the first the overtone pulsators ( bottom panel ) show similar trends concerning the he dependence .
this circumstantial evidence indicates that the bailey diagram is not a good diagnostic to constrain the helium content of cluster and field rr lyrae .
moreover , in dealing with luminosity amplitudes we need to keep in mind two relevant limits .
_ i)_ the theoretical bailey diagram is affected by uncertainties on the mixing length parameter ( see e.g. * ? ? ?
a decrease in the convective efficiency causes larger amplitudes , but the pulsation periods are minimally affected .
this means that the mixing length affects the slopes of the predicted amplitude
period relations , but the systematic drift as a function of the he content is not affected ( marconi et al .
2011 , in preparation ) .
_ ii)_ recent space ( corot , @xcite ) and ground - based @xcite observations indicate that the fraction of stars affected by the blazhko phenomenon is higher than previously estimated ( @xmath950% , @xcite .
rr lyrae variables are relatively bright and have been detected in several local group galaxies ( e.g. @xcite ) and can be easily identified from their characteristic light curves .
the most popular methods to estimate their distances is to use either the visual magnitude
metallicity relation or the near - infrared ( nir ) period - luminosity ( pl ) relation ( e.g. @xcite ) or parallaxes and proper motions @xcite .
the reader interested in independent approaches based on rr lyrae to estimate stellar distances is referred to the thorough investigations by , e.g. , @xcite . the visual magnitude
metallicity relation appears to be hampered by several uncertainties affecting both the zero - point and the slope @xcite . on the other hand , the nir pl relation seems very promising , since it shows several relevant advantages .
@xcite demonstrated , on an empirical basis , that rr lyrae do obey to a well defined @xmath10-band pl relation .
the reason why the pl relation shows up in the nir bands is due to the fact that the bc in the nir bands , in contrast with the optical bands , steadily decreases when moving from the hot ( blue ) to the cool ( red ) edge of the rr lyrae instability strip .
this means that they become brighter as they become redder .
the pulsation periods at fixed stellar mass and luminosity become longer , since redder rr lyrae have larger radii .
the consequence of this intrinsic property is that periods and magnitudes are strongly correlated when moving toward longer wavelengths .
theoretical and empirical evidence indicates that the nir pl relations of rr lyrae are robust methods to determine stellar distances . + _
i)_ the nir pl relations are minimally affected by evolutionary effects inside the rr lyrae instability strip .
the same outcome applies for the typical spread in mass inside the rr lyrae instability strip @xcite .
therefore , rr lyrae distances based on the nir pl relations are minimally affected by systematics introduced by their evolutionary status .
_ ii)_ theory and observations indicate that fundamental ( f ) and first overtone ( fo ) rr lyrae do obey independent nir pl relations .
_ iii)_ theory and observations indicate that the nir pl relations are linear over the entire period range covered by f and fo pulsators .
the nir pl relations also have three observational advantages . + _
i)_ the nir magnitudes are minimally affected by uncertainties on the reddening . _
ii)_ the nir amplitudes are at least a factor of 2 - 3 smaller than in the optical bands . therefore , accurate estimates of the mean nir magnitudes can be obtained with a modest number of observations .
moreover , empirical k - band light curve templates @xcite can be adopted to further improve the accuracy of the mean magnitudes .
_ iii)_ thanks to 2mass , accurate samples of local nir standard stars are available across the sky .
this means that both relative and absolute nir photometric calibrations can be easily accomplished .
the use of the nir pl relations is also affected by two limits . + _
i)_ empirical estimates of the slope of nir pl relations show a significant scatter from cluster to cluster .
they range from @xmath11 ( ic@xmath12 , sollima et al .
2006 ) to @xmath13 ( m@xmath14 , sollima et al .
2006 ) and it is not clear whether the difference is intrinsic or caused by possible observational biases . _
ii)_ current predictions indicate that the intrinsic spread of the nir pl relations decreases by taking into account either the metallicity or the hb - type of the horizontal branch ( hb ) @xcite .
therefore accurate distance estimates of field rr lyrae do require an estimate of the metallicity .
however , no general consensus has been reached yet concerning the value of the coefficient of the metallicity term in the nir period - luminosity - metallicity ( plz ) relations .
the current estimates for the @xmath10-band range from @xmath15 @xcite to @xmath16 @xcite mag / dex by using cluster and field rr lyrae , respectively .
= @xmath17 + 0.127 ) .
the red line shows the linear fit of the plk relation and the slope is labeled .
, width=302 ] to address these problems our group undertook a long - term project aimed at providing homogeneous and accurate nir photometry for several gcs hosting a good sample of rr lyrae and covering a wide range of metal abundances . in the following ,
we briefly discuss the slope of the k - band pl relation when moving from metal - poor to metal - intermediate gcs .
* m92)* this is the most metal - poor cluster in our sample @xcite . according to the metallicity scale by @xcite based on feii lines , the iron content is [ fe / h][email protected]@xmath190.07 .
we observed eight fundamental and three first overtones .
the latter were fundamentalized and we found a slope of @[email protected] , where the error only accounts for the uncertainty on the linear fit to the data . by using the k - band pl relation provided by @xcite we found a true distance modulus of @[email protected] mag , that agrees quite well with similar estimates available in the literature ( * ? ? ?
* and references therein ) .
* reticulum)* this is an lmc metal - poor cluster @xcite .
according to the metallicity scale by @xcite based on calcium triplet lines , the iron content is [ fe / h][email protected]@xmath190.1 .
we observed 21 fundamental and five first overtones .
the latter were fundamentalized and we found a slope of @[email protected] , where the error only accounts for the uncertainty on the linear fit to the data .
by using the k - band pl relation provided by @xcite we found a true distance modulus to reticulum of @[email protected] mag , that agrees quite well with similar lmc distances available in the literature .
* m5)* this is the metal - intermediate ( [ fe / h][email protected]@xmath190.06 ) cluster in our sample with the richest sample of first overtone ( 52 ) and fundamental ( 24 ) pulsators @xcite . by using the entire sample we found a slope of @[email protected] .
we also found a true distance modulus to m5 of @[email protected] mag , that agrees quite well with distances based on different distance indicators .
the good agreement also applies to the kinematic distance .
this seems an important finding , since this geometrical methods provides distance that are systematically larger than the distances based on different distance indicators @xcite .
centauri based on data collected by @xcite .
, width=302 ] * @xmath0 cen)* finally , we also estimated the distance to @xmath0 cen and observed 93 first overtones and 104 fundamentals @xcite .
the nature of this massive stellar system is not well established yet @xcite . however , there is general consensus that the stellar content of this system shows a metallicity distribution with multiple peaks @xcite .
the same outcome applies to the heavy element abundances ( @xcite , and references therein ) . to test the dependence of the slope on the iron abundance we split the sample into a metal - poor and a less metal - poor subsample .
we found that the slopes , within the errors , are very similar . therefore , we decided to perform a linear fit over the entire sample and we found a slope of @[email protected] . by using the calibration of the k - band pl relation provided by @xcite we found a true distance modulus to @xmath0 cen of @[email protected] mag , that agrees quite well with distance estimates available in the literature @xcite .
the iron content of the ggcs we have already investigated , neglecting @xmath0 cen , ranges from @xmath182.38 to @xmath181.26 dex .
the slope of the k - band pl relation should change from @[email protected] @xcite to @[email protected] @xcite .
the former slope is based on models that include the metallicity term , while the latter include the hb type .
current evidence indicates that the slope of the k - band pl relation is marginally affected by iron abundance , thus supporting the results by by theory and by @xcite .
however , more accurate data both in the metal - poor and in the metal - rich regime are required before we can reach firm conclusions concerning the metallicity dependence of both the slopes and the zero - points of the nir pl relations .
precise abundances of primordial helium ( @xmath21 ) together with the abundances of a few other light elements can provide robust constraints on the primordial nucleosynthesis , and in particular on the baryonic density of the universe .
unfortunately , stellar spectra display strong photospheric helium absorption lines in the visual spectral range only at high effective temperatures ( @xmath22 10,000 k ) .
high - mass stars are useless in constraining @xmath21 , since their material was already polluted by previous stellar generations .
low - mass stars only attain hot effective temperatures during central helium - burning phases .
the typical evolutionary lifetime of these phases called hot and extreme horizontal - branch phases is of the order of 100 myr .
however , these stellar structures can not be adopted to constrain @xmath21 , since their surface abundance are affected by gravitational settling and/or by radiative levitation @xcite . to overcome current observational limits , accurate estimates of @xmath21
have been provided using he emission lines in hii region of blue compact galaxies .
current estimates agree , within the errors , with the @xmath21 abundance provided by wmap @xcite . however , the agreement is not solid because @xmath21 is used as a prior in the cosmological solutions .
moreover , the spectroscopic abundances based on nebular lines might be affected by systematic errors @xcite .
the observational scenario concerning the helium abundance has been recently enriched by the possible occurrence of a variation in the helium abundance of globular cluster ( gc ) stars .
the presence of he - enriched stars in gcs has been suggested to explain not only the presence of multiple unevolved sequences , but also the presence of extended blue hb tails @xcite .
this working hypothesis relies on the well established anti - correlations between the molecular band - strengths of cn and ch @xcite and between o na and mg
al measured in evolved ( rg , horizontal branch [ hb ] ) , and in unevolved ( main sequence [ ms ] ) stars of the gcs investigated so far with high - resolution spectra @xcite .
more recently , deep hubble space telescope photometry disclosed the presence of multiple stellar populations in several massive gcs .
together with @xmath0 centauri @xcite multiple stellar sequences have been detected in gcs covering a broad range of metal contents : ngc 2808 @xcite , m54 @xcite and ngc 1851 @xcite .
some of these multiple sequences ( @xmath0 cen , ngc 2808 , ngc 1851 ) might be explained either with a he - enhanced @xcite , or with a cno - enhanced @xcite sub - population .
however , we still lack an empirical validation of the occurrence of he - enhanced sub - population(s ) in gcs . to overcome the spectroscopic problems
, @xcite suggested to use the r - parameter the number ratio between hb and rgb stars brighter than the luminosity level of the hb at the rr lyrae instability strip ( is) to estimate the initial he abundance of cluster stars .
+ two independent approaches to estimate the helium content in gcs were also suggested by @xcite .
the @xmath3-parameter , the difference in magnitude between the ms at ( b - v)@xmath23=0.7 and the luminosity level of the hb at the rr lyrae is , and the a - parameter the mass - to - luminosity ratio of rr lyrae stars .
the pros and cons of the quoted parameters were discussed in a thorough investigation by @xcite .
theoretical and empirical limits affecting the precision of the r - parameter have been also discussed by @xcite , @xcite and @xcite .
the key feature of the quoted parameters is that they are directly or indirectly connected with the hb luminosity level . in spite of the improvements in the photometric precision , in the sample of known cluster hb stars ,
we still lack firm empirical methods to estimate the hb luminosity level in gcs .
this problem is partially due to substantial changes in the hb morphology when moving from metal - poor ( blue hb ) to metal - rich ( red hb ) gcs .
moreover , we still lack a robust diagnostic to constrain the actual off - zahb evolution of hb stars ( @xcite . to overcome several of the above drawbacks , @xcite suggested a new method based on the difference in luminosity between the rgb bump and the main sequence benchmark at the same color of the bump .
the new method shows several indisputable advantages , but also a strong dependence on the metal content .
a new approach has been suggested by @xcite who detected , in a few rg stars in @xmath0 cen , the chromospheric he i line at 10830 .
the he line was detected in more metal - rich stars and it seems to be correlated with al and na , but no clear correlation was found with fe abundance .
however , a more detailed non - lte analysis of the absolute abundance of he is required before firm conclusions can be drawn concerning the occurrence of a spread in helium content .
interestingly enough , @xcite discovered both he i and he ii emission and absorption lines in field rr lyrae stars .
the lines show up during the rising branch of rr lyrae stars and appear to be the consequence of the shock propagation soon after the phases of minimum radius @xcite .
the first detection of helium lines in low - mass variables dates back to @xcite , who detected helium lines in type ii cepheids .
the key advantage of rr lyrae stars , when compared with detections of helium lines in hot and extreme hb stars ( behr 2003 ) , is that they have an extended convective envelope .
therefore , they are minimally affected by gravitational settling and/or radiative levitation @xcite .
the drawback is that the measurement of the helium abundance requires hydrodynamical atmosphere models accounting for time - dependent , convective transport , radiative transfer equations , together with the formation and the propagation of sonic shocks .
empirical evidence indicates that he absorption and emission lines in rr lyrae stars take place along the rising branch of the light curve .
plain physical arguments suggest that they are triggered by the formation and the development of strong shocks across the phases of maximum compression .
to further constrain the physical mechanism(s ) driving the occurrence of these interesting phenomena , new high - resolution , high signal - to - noise spectra are required to constrain the dependence of he lines on evolutionary and pulsation properties of rr lyrae stars and eventually to estimate the helium abundance .
recent theoretical and observational perspectives indicate that rr lyrae stars are heading for a new golden age .
this is due to the huge amount of field rr lyrae stars that have already been collected by extended photometric surveys .
a sample of @xmath11,200 field rr lyrae were collected by the northern sky variability survey ( nsvs , @xcite ) and analyzed by @xcite .
they cover a distance of @xmath17 - 9 kpc in the solar neighborhood and reach a limit magnitude of v=15 mag .
an even more ambitious project was realized by the all sky automated survey ( asas , @xcite ) , since they covered the southern sky up to @xmath24=+28 ( @xmath1 75% of their sky ) .
the survey has a limiting magnitude of v@xmath114 and they identified more than 1,450 rr lyrae within 4 kpc in the solar neighborhood .
deep multiband photometric ( ugriz ) and spectroscopic data for almost 500 rr lyrae have been collected by the sloan digital sky survey ii ( sdss ii , @xcite , sesar et al . these proceedings ) .
they have been able to investigate the spatial distribution of halo rr lyrae stars up to galactocentric distances of 5 - 100 kpc .
a sizable sample of field rr lyrae ( more than 2,000 ) was found by @xcite using v and r - band archival data collected by the southern edgeworth - kuiper belt object ( sekbo ) survey .
this survey covers 1675 square degrees along the ecliptic to a mean depth of v=19.5 , i.e. up heliocentric distances of @xmath150 kpc .
however , the real quantum jump concerning the photometric precision , the depth and the sampling of current rr lyrae surveys , is given by the optical gravitational lensing experiment iii survey ( ogle iii , @xcite ) .
they collected v and i - band data over a time interval of almost ten years , toward the galactic bulge and identified the richest sample ever collected of field rr lyrae , i.e. more than 16,800 stars .
the near future appears even more promising not only concerning the new optical ( pan starrs ) and nir ( vista ) photometric surveys , but also for the spectroscopic surveys ( m2fs@magellan , flames@vlt ) .
this means that together with pulsation properties , also the metallicity of a relevant fraction of field rr lyrae will become available .
it is noteworthy that to attack the open problems mentioned above requires a multiwavelength , a spectroscopic , and a theoretical approach @xcite : a challenge that we have to deal with to further improve our knowledge of the detailed structure of the galactic spheroid and to improve the accuracy of the rr lyrae distance scale .
the goal is not trivial : not only to further constrain the structure of the galactic bulge and the interaction with the galactic bar , but also to trace the possible transition from thick to thin disc stars ( kinemuchi et al . 2006 ) .
note that up to now , we still lack firm empirical evidence concerning the presence of old , low - mass stars like rr lyrae stars in the thin disc .
there are a few field stars that are good candidates ( tv lib ) , but their intrinsic properties might be significantly different than typical rr lyrae @xcite
. the rr lyrae can be even more relevant as beacons to trace the stellar streams in the galactic halo ( marconi et al .
2006 ) and to constrain the outermost radial extent of the galactic halo . | we present some recent findings concerning the use of rr lyrae as distance indicators and stellar tracers .
we outline pros and cons of field and cluster rr lyrae stars and discuss recent theoretical findings concerning the use of the bailey ( amplitude vs pulsation period ) diagram to constrain the possible occurrence of helium enhanced rr lyrae stars .
nonlinear , convective rr lyrae models indicate that the pulsation properties of rr lyrae stars are minimally affected by the helium content .
the main difference between canonical and he enhanced models is due to the increase in luminosity predicted by evolutionary models .
moreover , we focus our attention on the near - infrared period - luminosity ( pl ) relation of rr lyrae and summarize observational evidence concerning the slope of the k - band pl relation in a few globulars ( m92 , reticulum , m5 , @xmath0 cen ) covering a range in metallicity of @xmath11 dex .
current findings suggest that the slope has a mild dependence on the metal content when moving from the metal - poor to the metal - intermediate regime .
finally , we also discuss the use of rr lyrae stars either to estimate ( helium indicator : a - parameter ) or to measure ( absorption and emission lines ) the helium content .
[ 1996/06/01 ] 2@xmath2 2c ii 4c iv 2fe ii 3fe iii 1 mg i 2 mg ii 2si ii 4si iv 2al ii 3al iii 1o i 1n
i 1h i = |
requirements engineering ( re ) is a discipline that constitutes a holistic key to successful development projects as the elicitation , specification and validation of precise and stakeholder - appropriate requirements are critical determinants of quality @xcite . at the same time , re is characterised by the involvement of interdisciplinary stakeholders and uncertainty as many things are not clear from the beginning of a project .
hence , re is highly volatile and inherently complex by nature .
although the importance of a high quality re and its continuos improvement has been recognised for many years , we can still observe industry struggling in defining and applying a high quality re @xcite .
the diversity of how re is performed in various industrial environments , each having its particularities in the domains of application or the software process models used , dooms the discipline to be not only a process area difficult to improve , but also difficult to investigate for common practices and shortcomings . from a researcher s perspective
, experimental research in re thereby becomes a crucial and challenging task .
it is crucial , as experimentation of any kind in re , ranging from classical action research through observational studies to broad exploratory surveys , are fundamentally necessary to understand the practical needs and improvement goals in re , to steer problem - driven research and to investigate the value of new re methods via validation research @xcite .
it is challenging , because we still need a solid empirical basis that allows for generalisations taking into account the human factors that influence the anyway hardly standardisable discipline like no other in software engineering . in consequence ,
qualitative research methods are gaining much attention @xcite , and survey research has become an indispensable means to investigate re .
although we are confident about the value of survey research to understand practical needs and to distill improvement goals in re , we still lack a solid empirical survey basis .
the reason seems to lie in an ironically paradoxical circumstance : the appropriate design of a survey in re and the descriptive interpretation of the results going beyond purely observational , qualitative analyses and reasoning is very challenging , because we still lack empirically grounded theories in re @xcite . in turn
, we still lack such theories in re as we still have no empirically sound survey basis .
available surveys in re either investigate isolated techniques in application , or they focus on a small data population ( single countries or companies ) so that the findings of the surveys are hardly generalisable they can not be viewed as representative . yet missing is a series of empirical investigations of practical problems and needs in re that allows for empirical generalisations to steer future research in a problem - driven manner . as a long - term goal
, we want to establish an open and generalisable set of empirical findings about practical problems and needs in re that allows us to steer future research in a problem - driven manner .
to this end , we aim at conducting a continuously and independently replicated , globally distributed survey on re that investigates the state of the practice and trends including industrial expectations , status quo , experienced problems and how those problems manifest themselves in the process .
we contribute the design of a globally distributed family of surveys on re and first results from its start conducted in germany .
the survey will be replicated from 2013 on in a series of countries to manifest a clear understanding on the practical needs in re and the inference of practically relevant improvement goals .
the remainder of the paper is as follows . in sect .
[ sec : relatedwork ] , we introduce available contributions in the context of our research , which gaps are left open , and how we intent to close those gaps . in sect .
[ sec : design ] , we introduce the design of the family of surveys , including the research questions , the used methodology and instrumentation , an initial theory ( to be extended during the replications ) , and the data analysis and validity procedures .
the first results of the survey currently ongoing in germany are presented in sect .
[ sec : results ] , before giving a concluding discussion in sect .
[ sec : conclusion ] .
we directly focus on empirical investigations in our area and delimit from any philosophical discussions , opinion papers or solution proposals ( see @xcite ) of which several valuable ones exist
. we can classify related empirical investigations into two major areas : investigations of techniques and methods , and investigations of general practices and contemporary phenomena in industrial process environments . in both areas , we find survey research as well as technical action research among case and field studies . contributions that investigate techniques and methods analyse , for example , selected requirements phases and which techniques are suitable to support typical tasks in those phases .
zowghi et al .
@xcite , for example , conducted a survey about which techniques support the elicitation phase .
a broader investigation of all phases is performed by cox et al .
@xcite who analysed the perceived value of the re practices proposed by sommerville and sawyer .
an exemplary survey on the choice of elicitation techniques is carried out by carrizo et al .
studies like those reveal the effects of given techniques when applying them in practical contexts .
another type of studies on techniques and methods is often driven by the objective of investigating the improvement of specific variables when applying different techniques in same or similar contexts .
for instance , we investigated the effects two different process models in re had on the quality of the resulting artefacts ( e.g. specification documents ) by performing technical action research @xcite .
abraho et al .
@xcite raise the level of abstraction in this research area and set those kind of studies into the context of a framework that supports the validation of methods based on user perception while testing the framework with a family of experiments .
in general , those studies give the opportunity to test the sensitivity of existing re approaches in an industrial context , but they rely on a problem domain explored in advance and focus on pre - defined improvement goals . to reveal such industrial improvement goals and explore the problem domain to steer research activities , we mostly rely on field studies and surveys .
one of the most commonly known survey is the chaos report of the standish group , examining , inter alia , root causes for project failures of which most ones are to be seen in re , such as missing user involvement .
whereas the report is known to have serious flaws in its design negatively affecting the validity of the results @xcite , other studies , such as the success study , conduct a similar investigation of german companies including a detailed and reproducible study design .
still , both surveys exclusively investigate failed projects and general causes at the level of overall processes . a similar focus , but exclusively set in the area of re , had the study of kamata et al .
they analysed the criticality of the single parts of the ieee software requirements specification std .
830 - 1998 on project success . the focus of those studies , however , does not support investigation of contemporary phenomena and problems in industrial re environments .
such investigations have , for example , been indirectly conducted by damian et al .
they analysed process improvements in re and the relation to payoffs regarding , for example , productivity and the final product quality .
nikula et al .
@xcite present a survey on re at organisational level of small and medium size companies in finland . based on their findings , they inferred improvement goals , e.g. on optimising knowledge transfer .
a study investigating the industrial reluctance on software process improvement was performed by staples et al .
they discovered different reasons why organisations do not adopt normative improvement solutions dictated by cmmi , such as no clear benefit and the relation to the company size .
a field study with a broader data population of 60 cases in one company has been performed by enam et al .
they could infer recommendations to practitioners , such as the involvement of users in the elicitation process .
a more curiosity - driven study to analyse typical project situations in companies was presented by us in @xcite .
we could discover 31 project characteristics that directly influence re .
a survey that directly focused on discovering problems in practical settings was performed by hall et al .
they empirically underpin the problems discussed by hsia et al .
@xcite and investigated a set of critical organisational and project - specific problems , such as communication problems , inappropriate skills or vague requirements , while those problems matched to a large extent with project characteristics we could discover .
the previous non - exhaustive list of contributions reveals valuable observations when applying methods and techniques in sensitive , industrial contexts .
another introduced type of studies being directly related to our contribution comprehends surveys that focus on the industrial status quo and problems in re .
although giving valuable insights into industrial environments , those studies do by now not allow for generalisation as they focus on single aspects in re , such as problems in re processes or re improvements , or they focus on small subject populations ( e.g. focus groups in single companies ) and , thus , these studies by now remain not representative .
to close this gap in literature , we designed a family of surveys in joint collaboration with different researchers .
the initial theory for the surveys relies on available study results as the ones introduced in the previous section and we expect it to change ( along with the variables ) over the years during replications due to an expected learning curve at us researchers .
in addition , we present the first results from the survey conducted in germany and illustrate a replication outline beginning from 2013 in different countries . by bringing together different interdisciplinary communities , the survey shall build an empirical basis for empirical generalisations and problem - driven research in re .
in the following , we introduce the design of the family of surveys . our overall long - term objective is to lay the empirical foundation to be continuously replicated in different countries .
those surveys aim at a generalisable investigation of the state of the practice and trends in re including practitioners expectations , the status quo in re and its improvement , and contemporary problems experienced in companies .
to support the dissemination of the results and the collaboration among the research communities , we provide a shared survey infrastructure relying on the same questionnaire , and we disclose the anonymised data to the promise repository .
in the following , we formulate four research questions that build the frame for our surveys .
afterwards , we design the overall methodology and introduce the instrument , i.e. the ( typed ) questions and the categories , in sect . [
sec : instrument ] . in sect .
[ sec : theory ] , we introduce an initial theory and our expectations we have gathered so far on basis of available literature , before concluding with the data analysis and validity procedures .
we formulate four research questions to steer the overall design of the surveys shown in tab .
[ tab : rqs ] .
.research questions[tab : rqs ] [ cols= " < , < " , ] [ tab : problems - general ] there , we received a more mixed picture of re in practice .
on the one hand , six of the problems we offered were moderately agreed with by most of the respondents : communication flaws within the team and with the customer , incomplete and/or hidden requirements , inconsistent requirements , time boxing / not enough time and underspecified requirements that are too abstract and allow for various interpretations .
on the other hand , two problems were disagreed with : weak knowledge of the customer s application domain and weak relationship to customer .
the rest was considered mostly neutral or was moderately disagreed with .
the deviations were mostly small ( 1 ) . in four problems
, we have a deviation of 1.5 , which suggests a slightly higher diversity in the answers .
accordingly , we could not find any large or significant correlation between the problems stated by the respondents and the answers selected in rq 1 and 2 ( e.g. the relevance given to the artefacts and problems experienced w.r.t . ,
inconsistency ) . finally , we asked the respondents to rank the problems they have experienced according to their criticality .
the most often mentioned problem in this ranking are incomplete and/or hidden requirements .
also mentioned often are time boxing / not enough time and inconsistent requirements .
the most frequent answer for how those selected problems manifest themselves in the process was change requests and additional effort ( e.g. for meetings ) . when asked which of the selected problems they saw as a major reason for experienced project failure , the highest answer was given to incomplete requirements ( 8 times ) , followed by communication flaws with the customer ( 6 ) and and moving targets as well as time boxing each being selected 5 times .
the results on project failures , however , do not yet allow for a clear interpretation as not all respondents selected problems . the major problem with standards seems to be that they do not guide the requirements engineers enough how to create specification documents .
also the integration with risk management is sometimes a problem but the deviations are higher and , hence , it seems to be different in different companies .
the unclear terminology also receives some minor complaints .
the picture is more mixed for general re problems , but there are also no overwhelmingly large problems .
the classic re problems , communication , incomplete requirements , inconsistent requirements and not enough time seem to dominate .
the application domain and the relationship to customers are in most companies not problematic .
in this paper , we contributed the design of a global family of surveys to overcome the problem of by now isolated investigations in re that are not yet representative . with the family of surveys , we aim at establishing an empirically sound basis for understanding practical trends and problems in re and for inferring representative improvement goals . hence , the family of re surveys will build a continuous and generalisable empirical basis for problem - driven research .
the family of surveys relies on an initial theory induced from literature , and it is designed in joint collaboration with different researchers from different countries .
an additional pilot phase rounded out the validity procedures we could perform in advance .
we presented the first results from the survey conducted in germany where 30 respondents of different companies participated ( with an answer rate of 29% ) and showed initial trends in re as well as problems the respondents experience in their practical settings .
the first replication of the survey is currently performed in the netherlands and the synthesis of both surveys will be further disseminated to bring together the various empirical and re - specific research communities .
further replications are planned from 2013 on by the researchers that already committed themselves to our undertaking .
each replication uses the same infrastructure and is based on the same questionnaire while the interpretation of the results will be performed independently .
to guarantee a reproducible design of the survey and of the results , we commit ourselves to disclose the anonymised data of each replication to the promise repository .
the overall objective is to establish a regularly performed survey replication to continuously adapt and , finally , manifest a theory on the practical status quo and problems in re .
based on our interpretations of the results w.r.t .
the theory presented in sect .
[ sec : theory ] , we could confirm selected empirical findings available in literature .
for example , our results indicate that artefacts still remain underrepresented in available company standards for re while the practitioners rate the definition of artefacts structures and contents in those standards to be important .
furthermore , we could already confirm the general reluctance of practitioners against available normative re improvement standards , such as ones based on cmmi , which put the focus on assessing re standards against pre - defined processes and methods . regarding the problems observed in the respondents project environments , we could rank the following as the most problematic ones : incomplete and inconsistent requirements , communication flaws within teams and with customers , and time boxing . while the first four ones relate to existing evidence
, the last one is , to us , remarkable .
we can directly derive two implications from our contributions .
first , the results confirm a first coarse theory we draw on the basis of different isolated studies .
this already allows researchers to steer their problem - driven research , i.e. they can define improvement goals on basis of a survey that already goes beyond isolated investigations and validation research ; for instance , in the area of qualitative re improvement methods and/or in the area of artefact orientation .
second , the family of surveys is and will remain open .
this allows not only researchers to reproduce the results and their interpretation , but also practitioners to evaluate their own re situation against overall industrial trends .
although we have first results from the survey conducted in germany and can be confident about the next replications , we are aware that the design has still limitations .
most importantly and as already discussed , the initial theory is based on available contributions that investigate different aspects in re in an isolated manner .
hence , the theory still needs to evolve and mature along with the variables in the instrument over the years during replications due to expected learning curves at us researchers , before we can finally establish a reliable and empirically grounded theory .
we plan the further coordination of the replications , their synthesis and dissemination as future work .
the dissemination comprehends both empirical and re research communities to support a variety of conceptual work in re on basis of empirical sound findings . to this end
, we cordially invite further researchers to join in for further replications over the next years to establish a generalisable empirical basis on the state of the practice in requirements engineering .
we want to thank all participants of isern and eessmod for fruitful discussions .
in particular , we want to thank silvia abraho , barry boehm , giovanni cantone , michael chaudron , maya daneva , marcela genero , lars pareto , and roel wieringa for participating on the thematic workshops and giving us valuable feedback on the basic design of the family of surveys .
furthermore , we want to thank the internal reviewers listed in tab . [
tab : review ] for their effort spent , and the external reviewers for their commitment to the whole undertaking . finally , we are grateful to michael mlynarski for supporting us in the pilot phase and to our project partners for their participation .
m. broy . .
in a. levi , e. savas , h. yenigun , s. balcisoy , and y. saygin , editors , _ proceedings of the 21th international symposium on computer and information sciences ( iscis 2006 ) _ , volume 4263 , pages 2434 .
springer - verlag berlin , 2006 .
d. mendez fernandez , k. lochmann , b. penzenstadler , and s. wagner . .
in _ proceedings of the 15th international conference on evaluation and assessment in software engineering ( ease 2011 ) _ , pages 104113 . institution of engineering and technology ( iet ) , 2011 . | * context : * for many years , we have observed industry struggling in defining a high quality requirements engineering ( re ) and researchers trying to understand industrial expectations and problems .
although we are investigating the discipline with a plethora of empirical studies , those studies either concentrate on validating specific methods or on single companies or countries .
therefore , they allow only for limited empirical generalisations .
* objective : * to lay an empirical and generalisable foundation about the state of the practice in re , we aim at a series of open and reproducible surveys that allow us to steer future research in a problem - driven manner .
* method : * we designed a globally distributed family of surveys in joint collaborations with different researchers from different countries .
the instrument is based on an initial theory inferred from available studies . as a long - term goal ,
the survey will be regularly replicated to manifest a clear understanding on the status quo and practical needs in re . in this paper , we present the design of the family of surveys and first results of its start in germany .
* results : * our first results contain responses from 30 german companies .
the results are not yet generalisable , but already indicate several trends and problems . for instance , a commonly stated problem respondents see in their company standards are artefacts being underrepresented , and important problems they experience in their projects are incomplete and inconsistent requirements .
* conclusion : * the results suggest that the survey design and instrument are well - suited to be replicated and , thereby , to create a generalisable empirical basis of re in practice .
1000000 1000000 |
it has long been established that microscopic deterministic chaos provides an efficient mechanism for the mixing of orbits in the phase
space of dynamical systems , leading to the decay of statistical correlations as time evolves .
thus , chaotic dynamics can magnify small scale fluctuations and justify the existence of macroscopic variables like entropy and temperature , which are of central importance in an analysis based on non equilibrium statistical mechanics @xcite .
the main purpose of the work presented here is to study the `` melting transition '' in a microplasma model , using standard methods , as well as certain recently developed techniques for chaos detection , such as the _ smaller alignment index _ ( sali ) @xcite and its extension to the so called _ generalized alignment index _ ( gali ) @xcite .
this latter approach is based on geometrical aspects of the microscopic dynamics and has significant computational advantages over more classical indicators based either on local dynamics , such as the lyapunov exponents , or statistical properties , such as the mean temperature or the kolmogorov
sinai entropy .
in fact , the use of these novel indices often constitutes an improvement of several orders of magnitude in cpu and dynamical time for the identification of the chaotic or ordered nature of single orbits @xcite .
the utility of sali ( or gali ) methods in detecting dynamical regime changes in few particle hamiltonian systems @xcite is due to their sensitivity in tracing out the geometrical properties of the tangent dynamics of the flow .
thus , they provide accurate information about regime changes when important parameters of the system are varied .
for example , the existence of `` sticky '' regions and the occurrence of slow diffusion in weakly chaotic domains is detectable and the distinction between order and strongly chaotic motion is easily made by these methods @xcite . in the case of
fully developed chaotic motion , sali is particularly efficient , since it decays exponentially as @xmath0 where @xmath1 are the two largest lyapunov exponents @xcite , while for ordered orbits sali@xmath8const.@xmath9 .
the system we consider here consists of @xmath10 interacting particles described by a hamiltonian function of the form @xmath11 where the kinetic energy part @xmath12 is quadratic in the generalized momenta @xmath13 and the potential energy @xmath14 is a function of its generalized position coordinates @xmath15 . in the case of confined systems , one takes @xmath16 as the sum of two terms : @xmath17 , where @xmath18 represents the potential of the trap ( being a smooth positive function ) , while @xmath19 accounts for the interactions amongst the @xmath10 particles . the fact that @xmath10 is a finite number classifies the system as `` small '' , in contrast to a `` large '' thermodynamic system ( where one lets @xmath10 and the volume @xmath16 tend to infinity in such a way that @xmath20 is a finite real constant ) @xcite
small , finite hamiltonian systems in contrast to large , infinite , systems do not exhibit phase transitions in the standard sense .
the characteristic discontinuities or singularities in thermodynamic functions or their derivatives , which is the signature of a phase transition in large systems would now appear as steep but continuous changes in their thermodynamic functions or their derivatives @xcite .
phase transitions in small systems , have also been associated , quite recently , with certain topology changes in configuration space @xcite .
these developments are stemming from earlier work on hamiltonian systems exhibiting chaotic instabilities associated with singular behaviour of their configuration space curvature fluctuations at their phase transition point @xcite .
furthermore , investigations of the temperature dependence of the largest lyapunov exponent and other observables related to the `` topological hypothesis '' and the issue of phase transitions in many degrees of freedom systems is thoroughly presented in @xcite and references therein . in this paper ,
our model hamiltonian system describes a microplasma characterized by long range ( non shielded ) coulomb interactions described by the potential @xmath19 .
the microplasma is confined in a penning trap given by the potential @xmath18 and our aim is to study the motion of its ions as the energy increases .
the system evolves , in general , in 3 spatial dimensions and contains a relatively small number of ions for which the tracking of individual trajectories is numerically as well as experimentally feasible @xcite .
it is important to note that lyapunov exponents have already been used to investigate spatially extended plasmas @xcite as well as 1dimensional wave particle plasma models @xcite .
moreover , as gaspard @xcite has demonstrated , for a realistic hamiltonian model in 3 spatial dimensions containing a relatively small number of ions , the long range nature of the coulomb interaction @xmath19 makes the maximum lyapunov exponents behave differently in microplasmas than in many
particle systems with short range interactions , such as a hard ball fluid or systems with shielded ionic interactions ( yukawa potentials ) . a detailed analysis of the differences between shielded ( yukawa like ) and unshielded coulomb potentials can be found in the book by baus and tejero @xcite . indeed , in microplasmas composed of more than a few dozen of ions , the behavior of the maximum lyapunov exponent , especially in the `` gas phase '' , was explained following purely statistical mechanics arguments @xcite .
the plan of the paper is the following : the hamiltonian representation of our system and a summary of previous work are presented in sec .
2 . some background material related to the sali method and the spectrum of lyapunov exponents are given in sec .
the calculation of lyapunov exponents spectra for our system is presented in sec .
4 , while the detection of the `` melting transition '' as a passage from weak to strong chaos is demonstrated in sec .
finally , our conclusions are presented in sec .
let us consider a microplasma of @xmath10 ions of equal mass @xmath21 and electric charge @xmath22 in a penning trap with electrostatic potential @xmath23 and constant magnetic field along the @xmath24 direction with a vector potential of the form @xmath25 then , the hamiltonian of the full system is given by @xmath26 where @xmath27 is the position of the @xmath28th ion , @xmath29 is the euclidean distance between the @xmath28th and @xmath30th ions and @xmath31 is the vacuum permittivity . in the penning trap ,
the ions are subjected to a harmonic confinement in the @xmath24 direction with frequency @xmath32 while in the perpendicular direction ( due to the cyclotron motion ) they rotate with frequency @xmath33 .
thus , in a frame rotating around the @xmath24 axis at the larmor frequency @xmath34 , the ions feel a harmonic confinement of frequency @xmath35 in the direction perpendicular to the magnetic field . in the rescaled time
@xmath36 , position @xmath37 and energy @xmath38 with @xmath39 , the hamiltonian describing the motion takes the form @xmath40+\sum_{i=1}^{n}\bigl[\bigl(\frac{1}{8}-\frac{\gamma^{2}}{4}\bigr)(x_{i}^{2}+y_{i}^{2})+\frac{\gamma^{2}}{2}z_{i}^{2}\bigr]+\sum_{i
< j}\frac{1}{r_{ij}}=e\label{mic_plas_ham}\ ] ] where @xmath41 is the total energy of the system , @xmath42 and @xmath43 are the positions and canonically conjugate momenta respectively , @xmath44 is the euclidean distance between different ions @xmath45 given by @xmath46 and @xmath47 .
the ions are trapped in bounded motion under the condition that @xmath48 the trap is called prolate if @xmath49 , isotropic if @xmath50 and oblate if @xmath51 .
so , the motion is quasi 1dimensional in the limit @xmath52 and quasi 2dimensional in the limit @xmath53 .
the @xmath54 direction is a symmetry axis and hence the @xmath54 component of the angular momentum @xmath55 is conserved , being thus , a second integral of the motion .
we suppose from now on that the angular momentum is equal to zero ( i.e. @xmath56 ) and that the motion is studied in the larmor rotating frame . in @xcite , it was shown that the motion of the ions governed by hamiltonian is generally very sensitive to initial conditions and has at least one positive maximum lyapunov exponent , while a first study of its dependence on the energy , number of ions and trap geometry was also presented . at low kinetic energies , where the microplasma forms an ion crystal ,
chaos is considerably reduced , since the motion is quasi
harmonic around stable equilibrium configurations . on the other hand , at high temperatures ( or kinetic energies ) , the maximum lyapunov exponent decreases , as coulomb interactions become negligible and the microplasma forms a thermal cloud of nearly independent ions moving throughout the full extent of the harmonic potential of the trap .
finally , for intermediate values of the energy , there is a regime of chaotic behavior which becomes wider as the number of ions increases .
it is in this regime that the maximum lyapunov exponent attains a peak , whose value increases as a function of the number of ions .
a crucial aspect related to our work is that of the geometry of the trap .
as has been demonstrated in @xcite for prolate traps the maximum lyapunov exponent ( and therefore the k - s entropy of the system ) show a distinctively smoother increase in comparison to the oblate traps in the energy values interval from @xmath57 ( its minimum ) to @xmath58 ( its maximum ) .
the transition of interest for our analysis is taking place within a small interval of energies at the beginning , i.e. just after @xmath57 .
therefore , the slower the increase of the function @xmath59 the higher the resolution required and the more accurate ( and time consuming ) the numerical computation needed .
of course , if one argues in analogy to the standard phase transition theory , the dimensionality of the problem makes it easier to detect transitions in two or three spatial dimensions . but
systems with long range interactions may still exhibit a transition even in one dimension .
this justifies the choice of small @xmath60 adopted in the present paper as being both physically challenging and numerically tractable . at zero temperature ,
the system freezes at a crystalline state ( see @xcite and in particular refs . [
20 ] [ 25 ] therein ) reminiscent of wigner crystals @xcite . in that case
, the ion crystal is composed of several concentric rings for the case of an oblate ( quasi 2dimensional ) trap or a single line of ions in the case of a prolate ( quasi 1dimensional ) trap .
as temperature rises slightly above zero , the ions execute quasi
periodic motion about stable periodic orbits moving around their equilibrium position .
this is the regime where stable normal modes of vibration play an important role , as expected .
at slightly higher temperatures , the system `` bifurcates '' from the regime of quasi harmonic motion to one where it executes quasiperiodic motion in the form of a collective soft - mode .
we use the term `` soft - mode '' here since the dynamics is similar to what would be expected under a collective force on the particles , which diminishes as the energy increases .
the frequencies of the normal vibrational modes become smaller and smaller and tend to zero at the transition . as the energy continues to increase ,
these soft modes overcome the energy barriers and set out to explore broader domains of phase space . at even higher temperatures ,
their motion becomes erratic and the ion crystal `` melts '' , entering a regime of strongly chaotic motion in phase space .
as the temperature increases further , the ions form a thermal cloud in which the mean coulomb potential energy starts to become negligible with respect to the mean kinetic energy and mean harmonic potential energy .
it has been shown that the maximum lyapunov exponent @xmath61 reaches its peak value at energies well above the `` melting phase '' , while this transition is not manifested in the behavior of the lyapunov exponents as a function of the number of particles @xmath10 or the energy @xmath41 .
finally , for energies beyond the peak value of the maximum lyapunov exponent , the motion is dominated by the kinetic energy part and the system becomes amenable to a statistical mechanical treatment with considerable accuracy , in spite of the small number of particles present .
indeed , at such high temperatures @xmath62 , the spatial disorder of the microplasma can be described in terms of its thermodynamic entropy and its maximum lyapunov exponent @xmath61 expressed by the theoretical estimate @xcite @xmath63 which may be used to characterize the crossover to the regime of thermal cloud motion , albeit at a high computational cost for the determination of @xmath61 . and @xmath64 ( prolate quasi 1dimensional geometry ) : ( a ) at @xmath65 the motion is dominated by its stable normal modes and `` crystalline '' behavior is observed .
( b ) at @xmath66 a transition to chaotic motion occurs , where `` melting '' is expected to take place .
( c ) at @xmath67 the onset of `` thermal cloud '' behavior is evident.[fig:4 ] , scaledwidth=100.0%,scaledwidth=50.0% ] furthermore , in @xcite , the maximum value of @xmath61 is numerically observed to obey a power law with respect to the energy as well as the number of ions .
thus , one reaches the conclusion that , for a prolate trap geometry ( quasi 1dimensional ) , the `` melting transition '' and its associated microscopic dynamical behavior should be detectable at low energies and few number of ions ( e.g. for @xmath68 and @xmath64 ) .
the different characteristic kinds of motion in such a setting are depicted in fig .
[ fig:4 ] .
panel ( a ) presents a typical behavior of the ions at energy @xmath65 , where the dynamics is dominated by stable normal modes that correspond to `` crystalline '' behavior . at the slightly increased energy @xmath66 of panel ( b ) we observe the occurrence of a transition from stable collective ( quasi periodic ) motion to chaotic behavior .
finally , in panel ( c ) , the onset of a fully `` thermal '' cloud behavior is apparent at the energy @xmath67 .
so , for the rest of the paper we consider the case of @xmath68 ions in the prolate trap geometry with @xmath64 .
( as in eq .
, dimensionless scale ) on the energy @xmath41 for the microplasma system of @xmath68 ions in a prolate penning trap @xmath69 .
the error bars represent one standard deviation from the time averaged temperature @xmath62 .
clearly , the crossover from ordered to weakly chaotic motion around @xmath70 does not occur in a manner justifying the use of statistical mechanics considerations .
in the inset , the ratio of the standard deviation of the time averaged @xmath62 to the time averaged temperature @xmath62 shows that they are comparable.[temper],scaledwidth=100.0%,scaledwidth=50.0% ] the temperature , in dimensionless units , @xmath71 ( where @xmath72 denotes the euclidean norm of the @xmath28th ions velocity @xmath73 ) averaged over the time @xmath74 which is a proportional quantity of the mean kinetic energy of the ions is plotted in fig .
[ temper ] , as a function of the energy @xmath41 for the microplasma with @xmath68 ions in the prolate penning trap with @xmath64 .
evidently , the melting of the crystal here is not associated with a sharp increase of @xmath62 at some critical energy , as would be expected from a first order phase transition .
note also in fig .
[ temper ] that the fluctuations of @xmath62 ( measured by its standard deviation ) can not provide a clear indication of the crossover from a regime of stable oscillations , through `` melting '' , to strongly developed chaotic motion .
so , the use of statistical mechanics considerations must focus on the importance of large fluctuations .
indeed , the inset of fig .
[ temper ] shows that the fluctuations of the time averaged temperature @xmath62 , in the crossover regime , are comparable with the values of the time averaged temperature itself .
one can , therefore , say that there is no `` macroscopic '' methodology for detecting dynamical regime changes in system that might be useful to theorists or experimentalists .
it , therefore , becomes especially important to adopt a different approach and attempt to study dynamical regime changes in our microplasma system , by performing a detailed study of its _ microscopic _ dynamics . to this end
, we shall focus hereafter on the lower energy range of hamiltonian , where the transition from `` crystalline like '' to `` liquid like '' collective behavior takes place . in the next section , we review the methods used for the detection of these dynamical regimes , i.e. the smaller alignment index ( sali ) @xcite and the spectrum of lyapunov exponents @xcite and compare the information provided by these approaches .
as we shall see , in the crossover regime , the sali exhibits a rather intricate `` stair like '' power law decay to zero , with different lyapunov exponents , reminiscent of the `` stickiness '' behavior typically observed in the neighborhood of resonance islands of hamiltonian systems @xcite . in our case
, this type of weakly chaotic dynamics will turn out to characterize the energy range over which the passage from ordered to irregular dynamics occurs , through what we call the `` melting transition '' .
the sali method was initially introduced in @xcite and has been applied successfully to distinguish between ordered and chaotic dynamics in maps of various dimensions @xcite , in hamiltonian systems @xcite , as well as in problems of celestial mechanics @xcite , galactic dynamics @xcite , field theory @xcite and nonlinear 1dimensional lattices @xcite .
following @xcite , one considers the @xmath75dimensional phase space of an arbitrary autonomous hamiltonian system @xmath76 where @xmath77 is the canonical position vector , @xmath78 is the corresponding canonical conjugate momentum vector and @xmath41 is the total constant energy .
the time evolution of an orbit of hamiltonian associated with an initial condition @xmath79 at time @xmath80 is defined as the solution of hamilton s equations of motion @xmath81 and is called the orbit @xmath82 of eqs . passing through @xmath83 . to define the smaller alignment index ( sali ) , one uses the _ variational equations _ , which represent the linearization of hamilton s equations of motion about a reference orbit @xmath82 of the system and
are defined as @xmath84 where @xmath85 is the jacobian of the right hand side of eqs .
calculated about the orbit @xmath82 .
the vectors @xmath86 are known as deviation vectors of the @xmath82 .
if we follow two of them , say @xmath87 and @xmath88 , we can define the _ smaller alignment index _ ( sali ) as @xmath89 where @xmath90 denotes the usual euclidean norm defined in @xmath91 . if @xmath82 is chaotic then
@xmath92 since both deviation vectors tend to align with the direction of the maximum lyapunov exponent as @xmath74 increases @xcite .
furthermore , it has been shown that the time evolution of the sali for chaotic orbits tends to zero exponentially at a rate related to the difference of the two largest lyapunov exponents @xmath61 and @xmath93 as @xcite @xmath94 on the other hand , if the orbit is ordered , sali exhibits small oscillations about a constant @xmath95 $ ] .
this is so , because both deviation vectors tend to become tangential to the torus on which the orbit is evolving while remaining mutually linearly independent in time due the existence of local or global integrals of motion @xcite .
it follows that this sharply different behavior of the sali for ordered and chaotic orbits makes it a reliable and computationally fast tool for distinguishing between order and chaos in hamiltonian systems of any number of degrees of freedom .
the choice of the initial deviation vectors is arbitrary and in general does not affect the method apart from some very special cases demonstrated theoretically and verified numerically in a number of previous works @xcite .
one of the most standard and well established means for extracting information about the nature of a given orbit of a dynamical system is to calculate its maximum _ lyapunov characteristic exponent _ ( lce ) @xmath61 .
if @xmath96 the orbit is characterized as chaotic .
the theory of lyapunov exponents was first applied to characterize chaotic orbits by oseledec @xcite , while the connection between lyapunov exponents and exponential divergence of nearby orbits was given in @xcite .
benettin et al .
@xcite studied the problem of the computation of all lces theoretically and proposed in @xcite an algorithm for their efficient numerical evaluation .
in particular , @xmath61 is computed as the limit for @xmath97 of the quantity @xmath98 where @xmath99 , @xmath100 are deviation vectors from the orbit we want to characterize , at times @xmath101 and @xmath102 respectively .
similarly , all other lces , @xmath93,@xmath103 are computed as limits for @xmath97 of analogous quantities , @xmath104,@xmath105 . in the present paper
, we shall compute the values of all lyapunov exponents ( called the lyapunov spectrum ) using the algorithm proposed by benettin et al .
@xcite .
let us also recall that the lyapunov spectrum of an @xmath10 degree of freedom hamiltonian system , i.e. @xmath106 , exhibits a basic symmetry with @xmath107 of its members having the opposite sign of the other @xmath107 and two exponents , i.e. @xmath108 , @xmath109 , being equal to 0 .
so , the discussion from here on concerns only the positive half of the lyapunov spectrum , @xmath110 where @xmath111 .
the calculation of the lyapunov spectrum is computationally demanding , as its convergence often requires the integration of trajectories over very long time intervals .
this convergence depends on the inverse of the largest lyapunov exponent or _
lyapunov time_. so , the closer @xmath61 is to zero the longer the integration is needed to obtain reliable estimates for the full spectrum . on the other hand , of particular importance to statistical mechanics
is the connection of the sum of all positive lyapunov exponents ( an index of chaos based on microscopic quantities ) to the kolmogorov
sinai entropy @xmath112 ( a measure of disorder of a macroscopic nature ) , given by pesin s celebrated theorem @xcite @xmath113 in the case of zero escape rates .
this connection has opened new directions in the applications of thermodynamics , using the underlying microscopic dynamics of chaotic orbits ( for a review see @xcite ) .
for the purposes of simplifying our analysis , let us consider the case of few ions , e.g. @xmath68 and @xmath64 ( i.e. a small system in a prolate trap ) . as mentioned above , in this setting , it is easier to detect dynamical regime changes , such as seen in fig .
[ fig:1 ] , where we have plotted the kolmogorov
sinai entropy @xmath112 ( see eq . ) as a function of the energy @xmath41 of the microplasma hamiltonian .
both @xmath112 and lyapunov exponents grow less steeply than in the oblate case ( quasi 2dimensional trap ) for the region of small energies , as already pointed out in @xcite .
this slow increase of @xmath112 and @xmath114 s exhibits an inflection point at low energies as is evident in the inset of fig .
[ fig:1 ] .
we remark that this inflection point occurs just above the energy threshold that permits ions to move around their fixed points , while beyond this energy we observe the transition from weak to strong chaotic behavior .
note that @xmath112 as well as the lyapunov exponents exhibit a maximum around @xmath115 .
, for @xmath64 and @xmath68 , computed by the eq .
( dashed highest line ) and the lyapunov exponents ( black solid lines ) are plotted from the maximum down to the lowest one ( 15th ) as a function of the energy .
note that , as shown in the inset , their maximum occurs at energies close to @xmath115 , which is much higher than the regime where the `` melting transition '' takes place ( see text).,scaledwidth=100.0%,scaledwidth=50.0% ] as a first attempt to derive useful information from the spectrum of lyapunov exponents ( see fig .
[ fig:1 ] ) , in the energy range where the `` melting transition '' is expected to occur , we calculate the logarithms of the ratios of successive pairwise differences of lyapunov exponents @xmath116 as a function of the energy ( see fig .
[ fig:2 ] ) .
the calculation of all @xmath114 s was carried out for each trajectory up to a final integration time @xmath117 , ensuring that a relative convergence to 4 or 5 significant decimal digits has been achieved .
note in fig .
[ fig:2](a ) that the ratios ( [ les_ratios ] ) fluctuate wildly initially and only for @xmath118 do they start to settle down to small oscillations about their ultimate values , which , at least for the first 2 or 3 ratios appear to be quite distinct .
this suggests that it is within an interval @xmath119 where @xmath120 that we should expect to find the `` melting transition '' , where the positive lyapunov exponents are very small , far from the values they attain in the regime of strong chaos ( see inset of fig .
[ fig:1 ] for @xmath118 ) .
( see eq . ) , for @xmath64 and @xmath68 , as a function of the energy @xmath41 .
note the characteristic gaps in the spectrum appearing for large values of @xmath41 .
( b ) : same as in panel ( a ) for @xmath121 and small values of @xmath41 , with the characteristic gaps in the spectrum becoming erratic in the regime of weak chaos , settling down to more clearly defined values around @xmath118.,title="fig:",scaledwidth=50.0%,scaledwidth=50.0% ] ( see eq . ) , for @xmath64 and @xmath68 , as a function of the energy @xmath41 .
note the characteristic gaps in the spectrum appearing for large values of @xmath41 .
( b ) : same as in panel ( a ) for @xmath121 and small values of @xmath41 , with the characteristic gaps in the spectrum becoming erratic in the regime of weak chaos , settling down to more clearly defined values around @xmath118.,title="fig:",scaledwidth=50.0%,scaledwidth=50.0% ] undoubtedly , although the information concerning the existence of two distinct dynamical regimes at small energies is somehow contained in the plots of fig .
[ fig:2 ] , expressions still fail to provide a precise picture of this change .
this is perhaps due to inaccuracies of the inherent averaging process used in the calculation of the @xmath122 at low energies , where the lyapunov exponents are very small .
we , therefore , proceed to study more carefully the underlying local microscopic dynamics by employing the sali method we introduced previously .
in this section , we apply the procedure outlined in sec . 3.1 to analyze microscopically trajectories of the few particle ( i.e. @xmath68 ions ) microplasma system described by the hamiltonian in the prolate trap with @xmath64 .
we are thus able to identify at low energies a regime of weak chaos , through which the system passes from a `` crystalline like '' type of ordered motion to the strongly chaotic behavior of collective soft modes .
this is achieved by detecting qualitative changes in the sali behavior for trajectories within the energy interval where the transition occurs .
[ fig:3 ] shows the behavior of the sali as a function of time @xmath74 at selected number of energies .
these plots represent three distinct classes of long time behaviors for typical trajectories within the interval @xmath123 .
we have also computed , for the same sample of energies , the corresponding lyapunov spectra for sufficiently long integration times ( typically @xmath117 ) to make sure that the lyapunov exponents have converged to their limiting values up to a desired accuracy ( typically to 4 or 5 significant decimal digits ) .
[ fig:3](a ) represents the sali evolution for a representative trajectory of system at @xmath124 .
we see that it fluctuates around non
zero positive values up to an integration time of order @xmath117 , indicating that the motion ( at least up to that time ) is ordered and quasi - periodic .
by contrast , in fig . [ fig:3](b )
, we follow an orbit with energy @xmath125 and find that sali decays to zero exponentially fast , driven by the first gap in the lyapunov spectrum of fig .
[ fig:2 ] , according to the formula @xmath126 @xcite . in this case , the first two lyapunov exponents of the trajectory , @xmath127 and @xmath128 , are large enough and distinct , implying strongly chaotic motion .
finally , let us turn our attention to the intermediate and more interesting case where sali exhibits a stair like decay to zero as a function of time , at energies @xmath129 ( see fig . [ fig:3](c ) ) .
choosing , for example , a trajectory with @xmath130 we observe that sali differs significantly compared to what is shown in fig .
[ fig:3](a ) or [ fig:3](b ) . at first
, one might think that , as in fig .
[ fig:3](a ) , the orbit is quasi - periodic , since sali `` sticks '' to non zero values for a fairly long time interval ( @xmath131 ) . then , its behavior changes qualitatively , showing a near exponential decay to zero @xmath132 .
this suggests that , even though the lyapunov exponents ( and their differences ) are very small , eq . still holds , as in the strongly chaotic domains of phase space @xcite . for @xmath68 , @xmath64 and 5 typical energies :
( a ) @xmath124 where the motion is ordered .
( b ) @xmath125 ( strong chaotic behavior ) where we also plot the theoretical prediction sali@xmath133 for comparison .
( c ) @xmath134 and @xmath135 , plotting also the corresponding theoretical predictions .
note that all three vertical axes are logarithmic .
[ fig:3],title="fig:",scaledwidth=50.0%,scaledwidth=40.0% ] for @xmath68 , @xmath64 and 5 typical energies : ( a ) @xmath124 where the motion is ordered .
( b ) @xmath125 ( strong chaotic behavior ) where we also plot the theoretical prediction sali@xmath133 for comparison .
( c ) @xmath134 and @xmath135 , plotting also the corresponding theoretical predictions .
note that all three vertical axes are logarithmic .
[ fig:3],title="fig:",scaledwidth=50.0%,scaledwidth=40.0% ] for @xmath68 , @xmath64 and 5 typical energies : ( a ) @xmath124 where the motion is ordered .
( b ) @xmath125 ( strong chaotic behavior ) where we also plot the theoretical prediction sali@xmath133 for comparison .
( c ) @xmath134 and @xmath135 , plotting also the corresponding theoretical predictions .
note that all three vertical axes are logarithmic .
[ fig:3],scaledwidth=50.0%,scaledwidth=40.0% ] if we slightly increase the energy , however , to @xmath136 , sali displays an intermediate and rather intricate behavior , following a stair like decay to zero , shown by the middle curve in fig . [
fig:3](c ) . here
, the exponential law does not explain the sali decay to zero .
it appears that the motion lies within a weakly chaotic domain and `` sticks '' temporarily to islands of regular motion , executing a multi stage diffusion process @xcite . finally , by increasing the energy to @xmath135
, we observe that sali decays to zero exponentially fast , following closely the theoretical prediction @xmath126 , with @xmath137 and @xmath138 .
these results imply that the corresponding trajectory is fully chaotic and the system has `` melted '' , passing to a regime with strongly mixing properties .
we now examine more closely this step wise decay of the sali , observed in fig .
[ fig:3](c ) , over a range of energies spanning the melting transition . to this end
, we start with our microplasma in the form of a wigner crystal , at @xmath58 and proceed to the onset of thermal cloud formation , increasing the energy by steps of @xmath139 up to @xmath140 .
, for @xmath64 and @xmath68 , as a function of the energy @xmath141 $ ] where @xmath142 is the difference of the two largest lyapunov exponents and @xmath143 is the slope calculated by linear regression performed on the @xmath144 vs. @xmath74 curves .
we also plot the line @xmath145 to indicate the energy range where @xmath146 and eq .
are satisfied .
[ fig:5 ] , scaledwidth=100.0%,scaledwidth=50.0% ] one way to quantify the departure of the sali from the exponential decay law ( see fig .
[ fig:3](c ) ) is to calculate for each energy in the interval @xmath147 $ ] the quantity @xmath148 defined as the ratio of the difference between the two largest lyapunov exponents @xmath2 and @xmath143 , the linear regression estimate of this difference , obtained from semi logarithmic plots of sali , like those depicted in fig . [
fig:3](b ) and ( c ) .
the result is shown in fig .
[ fig:5 ] .
as expected , for the strongly chaotic cases where @xmath149 , we obtain values of @xmath150 near 1 with clearly negative @xmath151(sali ) slopes . the interesting part of fig .
[ fig:5 ] , however , is observed in the energy interval @xmath152 where @xmath143 becomes one order of magnitude smaller than @xmath142 .
it is in this regime that we observe the `` melting transition '' for our system .
the divergence of sali from its theoretical estimate of eq . signifies the presence of other collective modes of motion .
this motion is organized around islands of stability , suggesting a certain type of `` stickiness '' around different tori at successive time intervals . during these time intervals sali decays in a step like manner by different power laws .
in fact , we can provide more evidence to support the above interpretation of the dynamics in the `` melting phase '' , using a recent generalization of the sali , called the _
generalized alignment indices _ ( gali ) @xcite .
these indices represent the volumes of a parallelepiped , called gali@xmath153 , formed by @xmath154 unit deviation vectors , which ( a ) for ordered motion , oscillate about positive constants as long as @xmath155 ( @xmath156 being the dimension of the torus ) and decay following power laws for @xmath157 , while ( b ) for chaotic orbits , they all vanish exponentially with exponents that become more negative as @xmath158 increases and depend on more lyapunov exponents @xcite . in practise , since the computation of the gali requires the calculation of a fairly big number of large determinants at every time step , we have adopted an alternative way of evaluating it which is significantly faster in cpu time , the so called _ linear dependence index _
( ldi ) @xcite . thus , in fig .
[ fig:6 ] , we use this approach to compute the gali@xmath153 , @xmath159 and find that they demonstrate the existence of an ( at least 6dimensional ) torus for @xmath124 @xcite , while they show a step wise decrease for @xmath65 , which continues to hold up to @xmath160 , as predicted by the sali ( @xmath161 ) calculations ( compare fig .
[ fig:3](a ) , ( c ) with fig .
[ fig:6](a ) and ( b ) ) .
of course , as the energy increases , the ( weakly ) chaotic nature of the orbits becomes evident at earlier times ( e.g. for @xmath65 we have @xmath162 and for @xmath135 we get @xmath163 for the ldi ( or gali ) to become @xmath164 ) as we see in fig . [ fig:6](b ) and ( c ) .
( = gali@xmath153 ) , @xmath165 as a function of time @xmath74 for the microplasma system of @xmath68 ions in a prolate trap ( @xmath64 ) for 3 typical energies . ( a ) ldi@xmath166ldi@xmath167 for the energy @xmath124 .
( b ) same as in panel ( a ) for the slightly bigger energy @xmath65 .
( c ) same as in panel ( a ) for the bigger energy @xmath135 .
note that all axes are logarithmic.[fig:6],title="fig:",scaledwidth=50.0%,scaledwidth=50.0% ] ( = gali@xmath153 ) , @xmath165 as a function of time @xmath74 for the microplasma system of @xmath68 ions in a prolate trap ( @xmath64 ) for 3 typical energies .
( a ) ldi@xmath166ldi@xmath167 for the energy @xmath124 .
( b ) same as in panel ( a ) for the slightly bigger energy @xmath65 .
( c ) same as in panel ( a ) for the bigger energy @xmath135 .
note that all axes are logarithmic.[fig:6],title="fig:",scaledwidth=50.0%,scaledwidth=50.0% ] ( = gali@xmath153 ) , @xmath165 as a function of time @xmath74 for the microplasma system of @xmath68 ions in a prolate trap ( @xmath64 ) for 3 typical energies .
( a ) ldi@xmath166ldi@xmath167 for the energy @xmath124 .
( b ) same as in panel ( a ) for the slightly bigger energy @xmath65 .
( c ) same as in panel ( a ) for the bigger energy @xmath135 .
note that all axes are logarithmic.[fig:6],title="fig:",scaledwidth=50.0%,scaledwidth=50.0% ] thus , the above results suggest that the behavior of the sali during dynamical regime changes is especially useful , as it alerts us to examine quantities like @xmath150 ( see eq . ) , related to variations in the differences of lyapunov exponents .
such quantities indeed furnish important information , not readily available by other more standard microscopic chaotic indicators or by related macroscopic entropic quantities .
in this paper , we have reported results demonstrating the occurrence of dynamical regime changes in a hamiltonian system describing a microplasma confined in a prolate quasi 1dimensional configuration and characterized by long range coulomb interactions .
more specifically , in the lower energy regime , we macroscopically detected the transition from `` crystalline like '' to `` liquid like '' behavior , through what we call the `` melting phase '' .
as expected , well beyond this phase , the microplasma exhibits strong chaotic behavior that may be described by the macroscopic variables of statistical mechanics .
the question , therefore , is how can one determine the dynamical nature and identify the energy range of this `` melting transition '' from ordered to strongly chaotic behavior . to this end
, we first showed that the `` melting '' of our quasi 1dimensional crystal ( composed of few ions confined in a prolate trap ) is not associated with a sharp increase of the temperature at some critical energy , as might have been expected .
furthermore , the positive lyapunov exponents ( and the kolmogorov
sinai entropy expressed by their sum ) attain their highest values at energies much higher than the regime where the microplasma `` melts '' .
thus , it appears that there is no clear `` macroscopic '' methodology for identifying and studying this `` melting '' process in detail .
next , we argued that even though information about dynamical regime changes must be contained in the lyapunov exponents , lyapunov spectra by themselves still fail to provide , a more precise picture of this change . indeed , while the ratios of the larger of them show strong fluctuations at low energies , these require further analysis before they can reveal , with any precision , the energies over which the `` melting transition '' occurs . for these reasons , we found it useful to employ a more accurate tool , called the sali method , to study in more detail the local microscopic dynamics of the microplasma system . in this way
, we discovered an energy range of _ weak chaos _
, where the positive lyapunov exponents are very small and sali exhibits a stair like decay to zero with varying decay rates . as @xmath168 in chaotic domains ,
this inspired us to look more closely at the statistical fluctuations of the differences of the two largest lyapunov exponents @xmath2 in that regime .
we , thus , observed that these differences exhibit their largest fluctuations over a definite energy interval , where `` melting '' occurs in a way reminiscent of `` sticky '' orbits , executing a multi stage diffusion process near the boundaries of resonance islands in the phase space of hamiltonian systems .
the above results were also supported by the use of an extended set of indices which generalize sali to the case of more than 2 deviation vectors .
finally , we remark that it is also the rapid convergence of these indices , which turns them into efficient diagnostic tools that may be used to replace demanding molecular simulations in identifying weakly chaotic regimes in multi particle systems .
as we intend to show in a future publication , these indices can provide a viable alternative to the notoriously time consuming simulations required to study the presence of weak chaos and slow diffusive effects in the dynamics of metastable states .
the authors would like to thank professors g. nicolis , p. gaspard , g
forti and ch .
skokos as well as dr .
p. de buyl for fruitful discussions , encouragement and support during the preparation of the paper .
we would also like to thank the referees for their constructive remarks and suggestions . the work of ch .
a. was financially supported by the pai 20072011 `` nosy nonlinear systems , stochastic processes and statistical mechanics '' ( fd9024cu1341 ) contract and the work of v. b. by the prodex program of the european space agency under contract no .
esa ao2004070 ( c90241 ) .
the computer programs of this work were ran in the high performance multiprocessor systems `` pythia '' of ulb and `` turing '' of uop .
g. benettin , l. galgani , a. giorgilli , and j - m .
lyapunov characteristic exponents for smooth dynamical systems and for hamiltonian systems : a method for computing all of them .
part 1 : theory .
, 15:9 20 , march 1980 .
g. benettin , l. galgani , a. giorgilli , and j - m .
lyapunov characteristic exponents for smooth dynamical systems and for hamiltonian systems : a method for computing all of them .
part 2 : numerical application .
, 15:21 30 , march 1980 .
skokos , ch .
antonopoulos , and t. c. bountis . detecting chaos , determining the dimensions of tori and predicting slow diffusion in fermi pasta
ulam lattices by the generalized alignment index method . , 165:5 14 , 2008 . c. yannouleas and u. landman .
symmetry breaking and quantum correlations in finite systems : studies of quantum dots and ultracold bose gases and related nuclear and chemical methods .
, 70(12):2067 2148 , 2007 . | we present results demonstrating the occurrence of changes in the collective dynamics of a hamiltonian system which describes a confined microplasma characterized by long range coulomb interactions . in its lower energy regime , we first detect macroscopically , the transition from a `` crystalline like '' to a `` liquid like '' behavior , which we call the `` melting transition '' .
we then proceed to study this transition using a microscopic chaos indicator called the _ smaller alignment index _
( sali ) , which utilizes two deviation vectors in the tangent dynamics of the flow and is nearly constant for ordered ( quasi periodic ) orbits , while it decays exponentially to zero for chaotic orbits as @xmath0 , where @xmath1 are the two largest lyapunov exponents . during the `` melting phase '' , sali exhibits a peculiar , stair like decay to zero , reminiscent of `` sticky '' orbits of hamiltonian systems near the boundaries of resonance islands
this alerts us to the importance of the @xmath2 variations in that regime and helps us identify the energy range over which `` melting '' occurs as a multi stage diffusion process through weakly chaotic layers in the phase space of the microplasma .
additional evidence supporting further the above findings is given by examining the @xmath3 indices , which generalize sali ( = @xmath4 ) to the case of @xmath5 deviation vectors and depend on the complete spectrum of lyapunov exponents of the tangent flow about the reference orbit .
@xmath6interdisciplinary center for nonlinear phenomena and complex systems ( cenoli ) , service de physique des systmes complexes et mcanique statistique , + universit libre de bruxelles , 1050 , brussels , belgium + @xmath7center for research and applications of nonlinear systems ( crans ) , + department of mathematics , + university of patras , 26500 , patras , greece |
delayed dynamical systems ( ddss ) serve to model diverse phenomena in physics @xcite ( prominently in optics ) , but also in other fields @xcite such as engineering , biology , climatology or ecology .
more than one decade ago @xcite it was found that delayed systems with one constant delay can be studied like extended systems , and they present not only analogies but equivalent phenomena such as pattern instabilities @xcite and high - dimensional chaos @xcite .
chaos in ddss is considered as a particular type of spatio - temporal chaos for which the delay plays the role of the system size .
there are a number of studies concerning the tangent dynamics of systems with time delay .
previous works mainly focused on the lyapunov exponents ( les ) and properties obtained from them ( dimension , entropy , ... ) @xcite .
less is known about the associated tangent space directions , in particular , there are no studies analyzing the structure of lyapunov vectors ( lvs ) apart from the main one ( pointing along the most unstable direction )
. correlations of the main lv were only recently addressed @xcite taking advantage of the generic mapping of a dds into an extended system , but with contradicting results .
when considering lvs other than the main one , one must distinguish among different vector types .
the so - called characteristic lvs @xcite constitute the only intrinsic set of vectors that is univocally defined and is covariant with the dynamics .
however , their numerical computation is difficult in high - dimensional systems .
we adapt to ddss one of the methods to compute characteristic lvs @xcite .
in this paper we report on the generic properties that lvs exhibit in ddss in the large delay limit .
our numerical and theoretical results indicate that the lvs ( the main one and the others ) behave in qualitative and quantitative terms like in one - dimensional dissipative systems with spatiotemporal chaos .
ddss may exhibit chaos even in the simplest situation in which the main variable is a scalar and its evolution is determined by a delay - differential equation with one constant delay : @xmath0 where @xmath1 . in a lyapunov analysis we are interested in the evolution of infinitesimal perturbations @xmath2 : @xmath3 which is also a delayed equation with @xmath4 , and where @xmath5 and @xmath6 are functions : @xmath7 , @xmath8 . if the dds is chaotic an initial random perturbation becomes after some time aligned with the main lv ( reaching a stationary state in a statistical sense ) , and the average exponential growth rate is the largest le of the system @xmath9 .
numerically , this is simply achieved by integrating eqs .
( [ dds ] ) and ( [ tangent ] ) for long enough times . for non - leading lvs , more involved algorithms are needed ( see below ) .
we focus our study on systems in which delayed and non - delayed terms are separated : @xmath10 .
our results are expected to be generic for this class of models . in our study , we have assumed the non - delayed part is linear that did not revealed relevant differences . ] , @xmath11 , and all nonlinearities appear in the retarded component @xmath12 .
many important time - delayed systems , including the mackey - glass ( mg ) @xcite and ikeda models @xcite , and optical delayed feedback systems @xcite , can be expressed in this mathematical form . in particular , we have studied numerically different nonlinear functions : the mackey - glass ( mg ) , @xmath13 , @xcite ; and the nonlinear function @xmath14 that appears in some optical cryptosystems with delayed feedback @xcite .
almost identical results in qualitative and quantitative terms are obtained for these two systems .
therefore , for the sake of brevity , we choose to present the results only for the mg model , also used in @xcite . in our simulations
we have integrated numerically eq .
( [ dds ] ) using a third - order adams - bashforth - moulton predictor - corrector method @xcite , while linear equations , _ e.g. _ ( [ tangent ] ) , have been integrated using an euler method with the non - delayed part integrated semi - implicitly .
the parameters we used for the mg model are @xmath15 , @xmath16 and the time step was @xmath17 . in a numerical integration of a time - delayed system @xcite , the temporal discretization makes the system finite - dimensional ( with @xmath18 degrees of freedom ) but this has no effect in the results whatsoever since an increase of time resolution ( _ i.e. _ decreasing @xmath19 ) does not modify the largest les ( and their associated vectors ) . in several previous works
@xcite it was found useful to map the dds into an equivalent spatially extended dynamical system of `` size '' @xmath20 that evolves at discrete times @xmath21 as follows .
we express the continuous time as @xmath22 where @xmath23 is the `` spatial '' position and @xmath24 is the discrete time .
this spatial representation can be applied to both , the state of the system , @xmath25 , and the perturbations , @xmath26 .
the benefit of this spatial representation is that one can analyze the dds with the tools available for spatially extended dynamical systems .
we shall use the spatial representation of delayed systems here to analyze lvs and analyze the existence of long - range correlations as well as some other dynamical properties .
there is some degree of confusion in the literature regarding the definition and computation of lvs .
many authors think of lvs as the orthonormal frame of vectors that results as a byproduct of computing the les with the algorithm of benettin et al .
these vectors are usually called gram - schmidt or _ backward _ lvs @xcite and they span the subspaces in tangent space that , at present time @xmath27 , have grown at exponential rates @xmath28 since the remote past .
backward lvs @xmath29 depend on the particular scalar product adopted for the gram - schmidt orthogonalization in benettin s algorithm ( though the subspaces they span are genuine ) .
this and other undesired properties render the backward lvs unsuited to analyze problems like extensivity or hyperbolicity questions .
ruelle and eckmann @xcite noticed time ago that one can define a different set of lvs , the so - called _ characteristic _ @xcite lvs @xmath30 , that are intrinsic to the dynamics and are univocally defined ( _ i.e. _ independent of how the scalar product is defined ) .
each of these vectors is covariant with the dynamics , and the corresponding @xmath31-th le is indeed recovered by either a forward or backward integration of an infinitesimal perturbation initially aligned with the @xmath31-th characteristic lv .
actually , the perturbation will be aligned with the lv at all times ( hence the covariance ) .
however for backward lvs ( other than the main one , because @xmath32 ) the corresponding les are only recovered integrating backward . until very recently , no efficient numerical algorithms were available to compute characteristic lvs in large extended systems . in this paper
we calculate the set of characteristic lvs by adapting to ddss the method recently introduced by wolfe and samelson @xcite . in order to find the @xmath31th characteristic lv one has to compute the first @xmath33 forward lvs in addition to the first @xmath31 backward lvs .
forward lvs @xmath34 are obtained like backward lvs but integrating the perturbations backward in time ( from the remote future to the present time @xmath27 ) .
they have to be calculated using the transpose jacobian matrix so that les are obtained with the usual ordering @xcite : @xmath35 . for our time - delayed system
this procedure encompasses to integrate : @xmath36 the diagonal coefficient @xmath37 is identical to that in eq .
( [ tangent ] ) .
the off - diagonal term has a curious structure stemming from transposing a jacobian matrix with nonzero elements @xmath20 temporal units below the main diagonal .
thus , in eq .
( [ tangentb ] ) , we have @xmath38 .
notice that , as the integration runs backward in time ( @xmath39 ) , the minus sign in the left hand side is canceled . as occurs with backward lvs , periodic gram - schmidt orthonormalizations are needed to avoid the collapse of all perturbations along the most unstable direction .
note also that , due to the delay , eq .
( [ dds ] ) can not be integrated backward , and hence it is necessary to store the trajectory @xmath40 at every time step in the computer to be used in the computation of ( [ tangentb ] ) .
once both backward lvs and forward lvs have been computed the characteristic lvs are easily calculated following the prescriptions in ref .
one delicate point in delayed systems is that one must make sure that both forward and backward lv sets correspond exactly to the same time interval @xmath41 .
in this section we study the form of both backward ( gram - schmidt ) and characteristic lvs corresponding to the largest les , their localization properties , and the existence of long - range correlations .
lvs are strong localized objects , therefore , like in refs .
@xcite , it is very convenient to work with the associated fields obtained after a logarithmic transformation : @xmath42 ( and likewise for backward lvs )
. we will refer to @xmath43 as the @xmath31th _ surface _ due to the similarities of its dynamics with that of the kinetic roughening of fractal surfaces in growth models ( see details below ) . also recall that the norm is irrelevant because a lv only indicates a direction in tangent space , and this implies that the mean height of @xmath43 is irrelevant ( only its profile fluctuations matter ) ; this should be beard in mind for the theoretical analysis below .
( @xmath44 t.u . ) .
( b ) difference of 2nd and 3rd lv - surfaces with respect to the 1st lv - surface .
the existence of plateaus evidences the piecewise copy structure of lv surfaces with respect to the 1st one.,width=283 ] figure [ vectors](a ) shows a snapshot of the characteristic lv surfaces @xmath45 , @xmath46 , and @xmath47 corresponding to the three largest les t.u .
there are 168 positive les .
the @xmath48-th le vanishes , with @xmath49 . ] .
we find that non - leading lv surfaces are approximately piecewise copies of the leading lv surface @xmath45 , see fig .
[ vectors](b ) .
this ` replication property ' is a highly nontrivial phenomenon that was originally discovered to occur in chaotic extended dissipative systems @xcite .
we also emphasize that characteristic lvs exhibit a tendency to clusterize , contrary to backward lvs whose localization sites are scattered due to the imposed orthogonality . indeed , as exemplified by the snapshot shown in fig .
[ vectors](a ) , the first three characteristic lvs localize ( reach their largest magnitude ) at the same point ( completely overlapping on the leftmost region in this particular snapshot ) .
this does not occur all the time but in an intermittent manner .
interestingly , these structural features namely , replication and clustering have recently been reported to occur generically for lvs in chaotic spatially extended systems @xcite .
this deepens in the analogy between ddss and systems with extensive chaos in one dimension , which happens to hold even at the level of non - leading lvs .
regarding the existence of long - range correlations , we find that lv surfaces have a self - affine spatial structure at long scales , which translates into power - law correlations .
a detailed analysis of the spatial structure can be best achieved by computing the fourier transform of the surfaces at discrete times @xmath21 given by @xmath50 , with wavenumbers @xmath51 $ ] , where @xmath20 is the system size in this representation .
for a given lv surface @xmath43 , the fourier transform of the two - point correlator @xmath52 , is the so - called structure factor : @xmath53 where the brackets denote average over time @xmath21 and realizations .
therefore , the structure factor @xmath54 directly informs about the @xmath31th lv surface correlations at scale @xmath55 .
the curves from top to bottom correspond to @xmath56 .
the insets show the dependence of the local slopes of different curves . for @xmath57
the asymptotic slope is @xmath58 whereas for @xmath59 the asymptotic slopes are @xmath60 and @xmath61 for backward and characteristic lvs , respectively.,width=283 ] in fig .
[ sf ] it can be seen that the structure factor for the first lv - surface decays asymptotically as @xmath62 , with the so - called roughness exponent @xmath63 ( in agreement with the result in @xcite for the ikeda model ) .
figure [ sf ] also reveals the existence of a particular crossover wavenumber @xmath64 for the @xmath31th lv .
the exponents observed for @xmath65 are @xmath66 and @xmath67 for characteristic and backward lvs , respectively .
these numbers coincide with those previously reported in dissipative systems in one dimension @xcite .
our interpretation of these results is that for backward lvs the @xmath68 dependence seems to be the consequence of residual correlations with a geometric origin in the orthogonality of the basis .
the @xmath66 exponent of characteristic lvs indicates they convey more information among distant parts of the system than backward lvs ; which can be understood as a consequence of the covariance of characteristic lvs as they are consistent with both past and future evolutions [ characteristic lvs with @xmath59 are `` saddle solutions '' of ( [ tangent ] ) forced by ( [ dds ] ) , see @xcite ] .
as we have seen in the previous section , the surface @xmath45 associated with the main lv exhibits scale - invariant correlations in the @xmath69 limit .
this strongly suggests that this lv surface should belong to one of the universality classes of surface growth .
the fact that the lv surfaces obey scaling laws and that systems with spatiotemporal chaos could be divided into a few universality classes , according to the scaling of the associated surfaces , has implications in our understanding of chaos in extended systems from a statistical physics point of view .
moreover , the generic replication and clustering of characteristic lvs along the main vector direction in extended systems @xcite ( also observed here for time - delay systems , see previous section ) makes it even more interesting to determine the universality class of the main lv , since this is expected to provide a great deal of information about the structure of space and time correlations of the lv corresponding to leading as well as sub - leading unstable directions . for a wide class of one - dimensional spatio - temporal chaotic systems ( see @xcite )
the statistical features of the main lv surface are well captured by the 1@xmath70-kardar - parisi - zhang ( kpz ) @xcite stochastic surface growth equation : @xmath71 where @xmath72 is a white noise .
the kpz equation defines itself an important universality class of surface growth .
regarding chaotic systems , in many of them the main lv surface belongs to the kpz class .
this includes coupled - map lattices , coupled symplectic maps , the complex ginzburg - landau model , lorenz 96 model , among others @xcite .
only lvs of anharmonic hamiltonian lattices are known @xcite to exhibit correlations that are clearly inconsistent with kpz exponents ( @xmath73 ) .
statistical mechanics teaches us that , given the fact that kpz represents a dynamical universality class , one would expect that a large collection of different systems could belong to the kpz universality despite their apparent differences . only symmetries and conservation laws would determine the universality class .
that may explain why hamiltonian ( energy conserving ) systems , in contrast to dissipative systems , do not generally belong to the kpz universality . on this basis ,
ddss were also proposed @xcite to belong to a different universality class , namely the zhang model , arguing that these systems break the @xmath74 symmetry in the spatial representation .
however , this conclusion was in contradiction with an earlier work @xcite where kpz was postulated .
our aim in this section is to clarify this question by means of a theoretical analysis and extensive numerical simulations .
our starting point is the evolution equation in tangent space ( [ tangent ] ) , of which the main lv is the asymptotic solution . in the spatial representation ( see sec . [ model ] ) we perform the change of variables @xmath75 , so that ( [ tangent ] ) is now written as : @xmath76 where @xmath23 is the ` spatial ' position and @xmath24 is the discrete time .
we have seen in the previous section that a description in terms of surfaces instead of vectors themselves is more appropriate due to the strong localization of the latter . moreover , to relate eq .
( [ 2 ] ) with one of the stochastic partial differential equations modeling surface growth , we wish to approximate the discrete differences in @xmath21 by a partial derivative . in sum
, we have to take two steps : 1 .
transform to a surface : @xmath77 .
approximate @xmath21 by a continuous variable : @xmath78 .
note that these two steps do not commute .
the treatment of eq .
( [ 2 ] ) by snchez _ et .
@xcite proceeded with step ( ii ) before step ( i ) . in more detail , after step ( ii ) , eq .
( [ 2 ] ) becomes : @xmath79 and now transforming to the surface picture [ step ( i ) ] one has : @xmath80 so that we get an equation for the field @xmath81 .
this equation was already analyzed in ref .
@xcite and we summarize some of its properties in the following . on the one hand , the random drift term @xmath82 gives rise to an effective diffusion term @xmath83 . on the other hand ,
the presence of @xmath6 in the denominator induces large fluctuations of those terms proportional to @xmath84 every time that @xmath6 takes values close to zero .
usually , the probability density function is algebraic at the lowest order : @xmath85 with @xmath86 to be normalizable ( and @xmath87 in general ) . in turn
, @xmath88 is heavy tailed with large fluctuations @xmath89 ) .
these reasoning led the authors of ref .
@xcite to conclude that the main lv surface in ddss behaves following ( [ h_sanchez ] ) and generically falls into the universality class of the linear zhang model @xcite : @xmath90 that describes surface growth driven by a heavy - tailed noise @xmath91 with a probability distribution @xmath92 for @xmath93 and the index @xmath94 ( @xmath95 if @xmath96 ) .
in this work we propose an alternative analysis of eq .
( [ 2 ] ) , where step ( i ) is carried out before step ( ii ) .
this would be more suited because one applies the nonlinear transformation ( i ) prior to the approximation ( ii ) .
thus after defining the corresponding lv surface , @xmath97 , eq .
( [ 2 ] ) becomes @xmath98 } , \label{interm}\ ] ] where the choice of sign @xmath99 comes from the absolute value and is irrelevant for the arguments that follow .
we approximate the difference @xmath100 by the partial derivative @xmath101 . with no further approximations small enough so that the exponential in eq .
( [ interm ] ) can be expanded and only the lowest order may be retained yields eq .
( [ h_sanchez ] ) .
we propose here a different derivation that avoids this assumption and , remarkably , leads to a different result . ] and taking logarithms in eq .
( [ interm ] ) we get : @xmath102 where the term @xmath103 is again a fluctuating noise - like source due to the chaotic character of the trajectories . in the neighborhood of @xmath104 the probability distribution of @xmath6
is ( [ pv ] ) , and thus large values of @xmath105 are exponentially rare : @xmath106 , _ i.e. _ the noise is not heavy tailed .
we can obtain a more intuitive equation by making use of the small gradient approximation , @xmath107 . expanding ( [ logeq ] ) in the form : @xmath108 , \label{logeq2}\ ] ] where the prototypical quadratic term @xmath109 of kpz appears .
if @xmath5 is fluctuating , the drift term @xmath110 again yields ( at large scales , after a spatial averaging ) an effective diffusion term . in this case , eq .
( [ logeq2 ] ) would lead to the kpz eq .
( [ kpz_eq ] ) .
in contrast , if @xmath5 is a constant , like for instance in the case of the mg model , a diffusion term would also eventually appear far from the small gradient limit as a result of the absolute value in @xmath111 in ( [ logeq ] ) .
finally , notice that the noise - like term @xmath112 , which is different from that in ( [ zhang_eq ] ) , will not lead to rare events and its distribution will be exponentially decaying in general . at variance with snchez _
et al . _
derivation @xcite ( _ cf .
_ sec .
[ zhang_sec ] ) we have arrived here at the lv surface equation ( [ logeq2 ] ) by expanding on the local slope @xmath113 , which is generally expected to be a reasonable approach when the roughness exponent @xmath114 since the spatial average over a window of extent @xmath115 should scale as @xmath116 . .
the lower panel shows the local derivative . for the largest value of the delay , @xmath117 t.u . , the local derivative exhibits a plateau about @xmath118 , close to the theoretical value for kpz ( @xmath119 ) .
the curves are averages over at least 1000 realizations.,width=283 ] in the field of growing interfaces it is customary to quantify the temporal features by looking at the growth of the surface width at a given time when started from a flat initial perturbation [ @xmath120 . to do so we let random initial perturbations @xmath121 to evolve and measure the average surface width at time @xmath21 as @xmath122 ^ 2 } \right\rangle,\ ] ] where @xmath123 is the mean surface position and brackets denote averaging over different initial realizations .
the growth exponent @xmath124 is defined as the exponent of the transient power - law growth : @xmath125 , before the surface fluctuations saturate ( for @xmath126 ) ; while @xmath127 , when saturation is reached ( for @xmath128 ) and the perturbation is virtually aligned with the lv : @xmath129 .
the key point now is that @xmath124 equals @xmath130 for kpz , while it is @xmath131 for the zhang model with @xmath132 @xcite .
this makes the time exponent @xmath124 an excellent index to distinguish between kpz and zhang behavior ( note that the spatial exponent @xmath133 is the same in both models : @xmath63 ) .
figure [ beta ] shows that @xmath134 grows with an exponent that , for large enough delays , progressively approaches the kpz growth exponent @xmath135 , which strongly supports the kpz asymptotic ( @xmath136 ) scaling for chaotic ddss . in the paper by snchez et al . @xcite the largest time delay used @xmath137 t.u
. resulted in an estimation @xmath138 . in the light of our simulations with much larger delays
we conclude that the value of @xmath20 used in ref .
@xcite was insufficient to detect the true asymptotic scaling exponent .
in addition to the growth exponent , ref .
@xcite also invoked the existence of multiscaling of the main lv surface as an argument supporting its identification with the universality class of the linear zhang model ( [ zhang_eq ] ) .
this model , contrary to the kpz equation , leads to surfaces that exhibit multiscaling , _
i.e. _ strongly non - gaussian tails , induced by extreme events .
if we compute the @xmath139th - height - height correlation function for points separated a distance @xmath140 , @xmath141 ( where the overline denotes the average over @xmath142 and the brackets denote a realizations average ) one finds that @xmath143 , and multiscaling exists if the roughness exponents @xmath144 depend on the index @xmath139 . , @xmath145 .
multiscaling is observed at small @xmath140 . above a certain characteristic size @xmath146 no multiscaling is observed .
note that @xmath147 is insensitive to @xmath20 , and the curves overlap almost perfectly ( except at large @xmath140 due to finite - size effects).,width=283 ] we have computed the @xmath139th order height - height correlation function @xmath148 , given by eq .
( [ g_q ] ) , for several values of @xmath20 ( _ i.e. _ system sizes ) ; see fig .
[ multiscaling ] .
the distribution of the surface fluctuations shows clear signs of multiscaling with local roughness exponents @xmath144 that depend on @xmath139 for length scales below a typical scale of about @xmath149 .
these non - gaussian features disappear at larger scales ( @xmath150 ) , where the @xmath151 for all @xmath139 . for truly extreme
event dominated fluctuations , this length scale is expected to diverge ( however slowly ) with the system size , so that the fluctuation distribution is truly non - gaussian in the thermodynamic limit .
this slow divergence was indeed measured for models in the zhang universality class @xcite . in the case of time - delay systems
, this would correspond to having @xmath152 increasing with @xmath20 .
in contrast , we find that the characteristic length @xmath146 remains constant even after an increase of the delay of 16 times . again , the simulations by snchez _ et .
@xcite were carried out in systems with delays that were too short to obtain conclusive evidence on the dependence of @xmath147 with @xmath20 .
certainly , we can not rule out for sure an extremely weak dependence of @xmath147 on @xmath20 , but if it exists it must be sub - logarithmic and well beyond the precision that we can reach in our simulations .
we conclude that multiscaling of the lv surface fluctuations in ddss seems to be a short scale phenomenon that has no effect in the thermodynamic limit where the universality class is defined ; in this case described by the kpz equation .
in this work we have implemented ( to our knowledge for the first time ) characteristic lvs in ddss .
adaptation of the method proposed in @xcite to this kind of systems together with the computer capabilities nowadays available allowed us to reach systems with fairly large delays , which serves to investigate the `` thermodinamic limit '' of these systems .
our results for the lvs coincide quantitatively with those obtained in extended dissipative systems @xcite .
in addition we have revisited the question of which universality class the main lv belongs to .
after simulations with very large delays we may conclude that the main lv surface falls into the universality class of the kardar - parisi - zhang equation .
our theoretical arguments support this conclusion as well . in sum ,
dds are equivalent to extended dynamical systems in the sense that infinitesimal perturbations exhibit the same exponents characterizing spatiotemporal correlations .
dds have been traditionally considered to be different because of the lack of extensivity of the lyapunov spectrum : the positive exponents approach zero as @xmath153 @xcite and the ( kolmogorov - sinai ) entropy saturates with @xmath20 .
however the identification of @xmath20 with a size implies that comparisons should be done in temporal units of @xmath154 , and extensivity is then recovered . as @xmath155 ,
the redefined les @xmath156 do not decay to zero with @xmath20 , and the lyapunov spectrum converges to a density in the thermodynamic limit .
d.p . acknowledges support by csic under the junta de ampliacin de estudios programme ( jae - doc ) .
financial support from the ministerio de ciencia e innovacin ( spain ) under project no .
fis2009 - 12964-c05 - 05 is acknowledged . | we compute lyapunov vectors ( lvs ) corresponding to the largest lyapunov exponents in delay - differential equations with large time delay .
we find that characteristic lvs , and backward ( gram - schmidt ) lvs , exhibit long - range correlations , identical to those already observed in dissipative extended systems .
in addition we give numerical and theoretical support to the hypothesis that the main lv belongs , under a suitable transformation , to the universality class of the kardar - parisi - zhang equation .
these facts indicate that in the large delay limit ( an important class of ) delayed equations behave exactly as dissipative systems with spatiotemporal chaos . |
the hadron resonance gas model has been extremely successful in reproducing lattice quantum chromodynamics results in the hadron gas phase @xcite .
the underlying assumption is that it is possible to describe an interacting gas of hadrons in the ground states through a non - interacting gas of hadrons and resonances .
the expression for thermodynamic quantities in the hrg model is then simply a sum over the contributions from each resonance . in this sense ,
the only `` variable '' in the model is the hadronic spectrum one feeds into it .
the original idea by rolf hagedorn was that a limiting temperature @xmath1 for matter exists , called hagedorn temperature , which can not be exceeded , and therefore a system of hadrons responds to an arbitrary increase in the energy density by creating more and more resonances rather than by increasing the temperature @xcite .
following these arguments , he showed that these resonances would need to populate a spectrum which is exponentially increasing in mass , when counting all states with the corresponding degeneracy : @xmath2 summing over the resonances with degeneracy @xmath3 and mass @xmath4 .
it was later shown @xcite that @xmath1 can be interpreted as the limiting temperature for hadronic matter , if a phase transition occurs to a deconfined state of quarks and gluons .
presently , we know from lattice qcd calculations that the deconfinement transition is a smooth crossover for vanishing baryonic chemical potential @xcite . in this sense , the hagedorn temperature can be analogous to the critical temperature of the deconfinement transition , and , therefore , it is reasonable to expect the effects related to the increased states population to appear and become relevant close to such transition .
the list of experimentally measured resonances is continuously updated by the particle data group @xcite , and confirms the original idea from hagedorn @xcite .
moreover , there exist quark models @xcite that predict the presence of further , not yet discovered , hadronic states .
the latter states also show an exponential behavior in the mass spectrum , as shown in fig.[hag_spec ] . including further states in an exponentially increasing mass spectrum can have far reaching effects in the field of heavy - ions .
for instance , they play a role in the transport coefficients close to the critical temperature region @xcite , influence elliptical flow @xcite , can improve thermal fits @xcite , and affect the chemical equilibration time in the hadron gas phase @xcite .
thus , it is important to determine the precise number of resonances that exist otherwise a systematic bias affects numerous comparisons between theory and experiment .
the hadron resonance gas model has been employed to extract information about chemical freeze - out in heavy ion collisions through comparison of particle yields @xcite , particle ratios @xcite and ratios of fluctuations of conserved charges ( susceptibilities ) @xcite , with experimental data . in the past few years
, such comparisons of yields and particle ratios with data from both rhic and the lhc seemed to indicate the existence of different freeze - out temperatures for light and strange matter @xcite . when comparing to the experiment , effects due to resonance decays , acceptance cuts and isospin randomization need to be taken into account @xcite , which can also be affected by the inclusion of more states in the hadronic population @xcite . for strangeness , the agreement of the hadron resonance gas model with lattice results is not always satisfactory , so that it was suggested @xcite that the inclusion of quark model states is necessary .
it was also pointed out that this inclusion would reduce the difference in temperature between light and strange freeze - out . however , in @xcite the effects of resonance decays were not taken into account .
thus , we used the known branching ratios from the pdg list @xcite to extrapolate the branching ratios for the quark model states , carefully enforcing quantum numbers and mass conservation . by comparing net - charge and net - proton fluctuations ( @xmath5 ) with star data @xcite from the beam energy scan , it was possible to extract the values of @xmath6 and @xmath7 at chemical freeze - out .
the addition of quark model states only has a small decrease of the freeze - out temperature , as seen in fig.[tmub ] . + from lattice qcd ( wuppertal budapest ) to preliminary star data for @xmath8200 gev . ] from lattice qcd ( wuppertal budapest ) to preliminary star data for @xmath8200 gev . ] to extract freeze - out parameters from first principles , one can use ratios of susceptibilities as discussed in @xcite . if lattice qcd showed a separation between light and strange freeze - out @xcite , this would have an heavy impact for example on the hadronization schemes used in hydrodynamical simulations . in principle , one could compare lattice qcd to experiment directly from susceptibilities of strangeness .
however , lattice only has global contributions whereas the only strangeness - related fluctuations presently available from experiment are those of charged kaons . in order
to isolate charged kaons on the lattice , partial pressures for charged strange mesons can be used @xcite , where ratios of susceptibilities are shown to be of the form @xmath9 where @xmath10 and @xmath11 are supplied by the wuppertal budapest collaboration .
+ in fig.[lat_star ] @xmath12 is compared from lattice qcd with preliminary star data . due to large uncertainties in the experimental data ,
only a lower bound at @xmath13 can be extracted at the moment .
in this proceedings two different approaches for studying light and strange freeze - out temperatures are shown .
adding quark model states with their decays to the hadronic spectrum , lowers the light freeze - out temperature by a small amount .
second , a new method for extracting strangeness freeze - out parameters from first principles using strange charged sucpetibilities is presented to resolve the tension between the light and strange sector . in both cases ,
smaller error bars are needed to better check the strange sector .
this material is based upon work supported by the national science foundation under grant no .
phy-1513864 and by the u.s .
department of energy , office of science , office of nuclear physics , within the framework of the beam energy scan theory ( best ) topical collaboration .
this work contains lattice qcd data provided by the wuppertal - budapest collaboration .
an award of computer time was provided by the incite program .
this research used resources of the argonne leadership computing facility , which is a doe office of science user facility supported under contract de - ac02 - 06ch11357 .
the work of j. g. and a. p. was supported by the dfg grant sfb / tr55 .
the authors gratefully acknowledge the gauss centre for supercomputing ( gcs ) for providing computing time for a gcs large - scale project on the gcs share of the supercomputer juqueen [ 30 ] at jlich supercomputing centre ( jsc ) .
9 c. ratti _ et al . _
[ wuppertal - budapest collaboration ] , nucl .
a * 855 * , 253 ( 2011 ) doi:10.1016/j.nuclphysa.2011.02.052 m. bluhm , p. alba , w. alberico , a. beraudo and c. ratti , nucl .
a * 929 * , 157 ( 2014 ) doi:10.1016/j.nuclphysa.2014.06.013 r. hagedorn , nuovo cim .
suppl . * 3 * , 147 ( 1965 ) .
n. cabibbo and g. parisi , phys .
b * 59 * , 67 ( 1975 ) .
doi:10.1016/0370 - 2693(75)90158 - 6 w. broniowski , w. florkowski and l. y. glozman , phys . rev .
d * 70 * , 117503 ( 2004 ) doi:10.1103/physrevd.70.117503 p. m. lo , m. marczenko , k. redlich and c. sasaki , phys .
c * 92 * , no . 5 , 055206 ( 2015 ) doi:10.1103/physrevc.92.055206 s. capstick and n. isgur , phys .
d * 34 * , 2809 ( 1986 ) [ aip conf .
* 132 * , 267 ( 1985 ) ] .
doi:10.1103/physrevd.34.2809 , 10.1063/1.35361 s. godfrey and n. isgur , phys .
d * 32 * , 189 ( 1985 ) .
doi:10.1103/physrevd.32.189 j. noronha - hostler , j. noronha and c. greiner , phys .
* 103 * , 172302 ( 2009 ) doi:10.1103/physrevlett.103.172302 j. noronha - hostler , j. noronha and c. greiner , phys .
c * 86 * , 024913 ( 2012 ) doi:10.1103/physrevc.86.024913 g. p.
kadam and h. mishra , nucl .
a * 934 * , 133 ( 2014 ) doi:10.1016/j.nuclphysa.2014.12.004 s. pal , phys . lett .
b * 684 * , 211 ( 2010 ) doi:10.1016/j.physletb.2010.01.017 j. noronha - hostler , j. noronha , g. s. denicol , r. p. g. andrade , f. grassi and c. greiner , phys .
c * 89 * , no . 5 , 054904 ( 2014 ) doi:10.1103/physrevc.89.054904 j. f. paquet , c. shen , g. s. denicol , m. luzum , b. schenke , s. jeon and c. gale , phys .
c * 93 * , no . 4 , 044906 ( 2016 ) doi:10.1103/physrevc.93.044906 j. noronha - hostler , c. greiner and i. a. shovkovy , phys .
lett . * 100 * , 252301 ( 2008 ) doi:10.1103/physrevlett.100.252301 j. noronha - hostler , m. beitel , c. greiner and i. shovkovy , phys .
c * 81 * , 054909 ( 2010 ) doi:10.1103/physrevc.81.054909 m. beitel , k. gallmeister and c. greiner , phys .
c * 90 * , no . 4 , 045203 ( 2014 ) doi:10.1103/physrevc.90.045203 m. beitel , c. greiner and h. stoecker , phys .
c * 94 * , no . 2 , 021902 ( 2016 )
doi:10.1103/physrevc.94.021902 j. stachel , a. andronic , p. braun - munzinger and k. redlich , j. phys .
ser . * 509 * , 012019 ( 2014 ) doi:10.1088/1742 - 6596/509/1/012019 a. andronic , p. braun - munzinger and j. stachel , phys .
b * 673 * , 142 ( 2009 ) erratum : [ phys .
b * 678 * , 516 ( 2009 ) ] doi:10.1016/j.physletb.2009.02.014 , 10.1016/j.physletb.2009.06.021 s. borsanyi , z. fodor , s. d. katz , s. krieg , c. ratti and k. k. szabo , phys .
* 113 * , 052301 ( 2014 ) doi:10.1103/physrevlett.113.052301 r. bellwied , s. borsanyi , z. fodor , s. d. katz and c. ratti , phys .
lett . * 111 * , 202302 ( 2013 ) doi:10.1103/physrevlett.111.202302 | heavier resonances are continually being added to the hadronic spectrum from the particle data group that follow an exponentially increasing mass spectrum .
however , it has been suggested that even further states predicted from quark models are needed in the hadronic spectrum in order to improve the agreement between the hadron resonance gas model predictions and lattice qcd data .
we find that the inclusion of such states with extrapolated branching ratios slightly decreases the freezeout temperature . to eliminate ambiguities , we introduce a first principle method to extract the freeze - out temperature for charged kaons from experimental data , which yields a lower bound of @xmath0145 mev for the highest collision energy at rhic . |
cosmic strings @xcite are linear vortex defects predicted to be formed at a cosmological phase transition during which the vacuum manifold would not be simply connected .
the first interest in studying them comes from the fact that since a typical grand unified theory ( gut ) predicts a few phase transitions ( whose order , as far as the strings are concerned , does actually not matter @xcite ) , and because the vacuum structure needed to form strings happens to be generically realized , one can , following vilenkin @xcite , reasonnably assume that cosmic strings have an existence probability of at least @xmath1 . although they are not the only possible topological defects that could be formed in such phase transitions , they have the advantage , as compared for instance with domain walls and monopoles which must be somehow inflated away @xcite , to be at present compatible ( as are as well the textures @xcite ) with all existent cosmological data @xcite , while being also possibly responsible for the formation of large scale structure @xcite and the observed anisotropies in the cosmic microwave background ( cmbr ) @xcite .
most models based on these strings assume that they are generated at the gut phase transition , so that the dimensionless parameter @xmath2 , with @xmath3 the newton constant and @xmath4 the energy per unit string length , giving the expected relative order of magnitude of any gravitational effect due to these strings ( e.g. , light deflection @xcite or cmbr temperature fluctuations @xcite ) , was assumed to be @xmath5 .
another kind of strings was proposed by witten @xcite in 1985 who pointed out the possibility that bosonic or fermionic superconducting currents could be trapped in the strings , thereby inducing many electromagnetic effects , such as , for instance , a new scenario for structure formation @xcite .
shortly thereafter , it was shown by davis and shellard @xcite and independently by carter @xcite that , although the regular strings can not be potentially responsible for a cosmological catastrophe ( i.e. , the remnant mass density would not exceed the critical density ) because of gravitational radiation and the absence of any stabilising mechanism , the situation was completely different for current - carrying string loops since in the latter case , there exists centrifugally supported equilibrium configurations ( called vortons @xcite or rings @xcite ) which would overfill the universe @xcite by many orders of magnitude if they were stable ( a point which still demands further clarification and is presently under investigation @xcite ) , this stabilizing mechanism being enhanced when electromagnetic corrections are taken into account @xcite , and should not be confused with the much less efficient `` spring '' , or magnetostatic support mechanism @xcite .
the witten mechanism to produce currents in cosmic strings has been studied by many authors , interested in particular by their internal microscopic structure @xcite , and who exhibited clearly the characteristic features of what should be expected in these objects , such as the existence of a maximum ( spacelike ) current ( or the current quenching phenomenon ) and a phase frequency threshold ( for timelike currents ) .
independently , a `` macroscopic '' formalism @xcite was derived that allows one in principle to evaluate the dynamics of any current - carrying string configuration once its equation of state , relating the energy per unit length @xmath4 to the tension @xmath6 , is given .
this equation of state , for the witten simple model describing strings , and whose properties are believed to be qualitatively ( if not quantitatively ) similar to more complicated ( and realistic ) strings models , was indeed obtained ( albeit unfortunately only numerically ) , so that it has now become possible to study realistically the cosmological importance of superconducting cosmic strings , and their astrophysical , gravitationally induced , signature , since the structure of the spacetime surrounding a string of any kind is also known @xcite . the purpose of this article is to show that there exists yet another mechanism by which a string can become not only superconducting , but also current - carrying without invoking any extra external field ( e.g. , electromagnetic ) , hence the name spontaneous current generation for this mechanism , comparable in many respects to a similar mechanism existing in @xmath7he vortex lines @xcite .
this phenomenon occurs for particular values of the underlying microscopic parameters , when the current - carrier field is neither a scalar nor a fermionic field ( these cases being essentially equivalent due to the two dimensional description of the vortex ) , but a charged - coupled vector field such as the intermediate @xmath8 in the electroweak theory . to illustrate this mechanism in the most realistic possible way
, we shall consider a simple string - forming extension @xcite of the standard electroweak theory @xcite ( this latter model being string - free since its vacuum manifold , isomorphic to the 3sphere @xmath9 , is simply connected ) .
other motivations @xcite such as supersymmetry or superstrings inspired models also lead , generally as a low energy limit , to the model we wish to consider , namely that in which an extra@xmath10 is gauged , this new symmetry being spontaneously broken .
it is interesting to notice that because of the large number of experimentally testable phenomenological consequences of such a model , the energy scale of the symmetry breaking involved @xmath11 say , and thus the energy per unit length of the corresponding strings , is in fact constrained to exceed @xcite 300 500 gev ( depending on the couplings ) , which is actually close to the upper limit provided by the vorton mechanism @xcite , namely @xmath12 tev .
we shall first introduce our string forming model , as well as the vortex solutions themselves , in the simple case to begin with where no current is flowing in the strings , i.e. , the so - called nielsen
olesen @xcite solutions , or kibble type vortices .
then we go on the spontaneous current generation itself which is shown to be be due to an electromagnetic instability of the vacuum for massless @xmath0 field at zero temperature , and finally discuss how the phenomenon is responsible for a spontaneous breaking of the lorentz boost symmetry along the initially kibble - like string .
the electroweak theory @xcite is based on the spontaneous breaking of the @xmath13 symmetry by means of an @xmath14 doublet higgs field @xmath15 down to the electromagnetic @xmath10 symmetry .
this means that the vacuum has the topology of the quotient group @xmath16 , which is isomorphic to @xmath14 , i.e. , it has the topology of the 3sphere and is therefore simply connected . as a result ,
topologically stable cosmic strings are not present in this model ( and in fact , due to the experimental bound on the higgs mass @xmath17 gev @xcite , even string - like solutions in this model are dynamically unstable @xcite ) . in order to investigate the structure of cosmic strings in a realistic model that would take the electroweak theory into account
, it is thus necessary to modify this theory first .
there are basically two different approaches that can be followed to extend this model .
the first one consists in assuming the higgs realisation of the symmetry breaking not to be fundamental , and to consider instead dynamical symmetry breaking such as in the chiral approach @xcite involving the @xmath18 symmetry .
this leads to the existence of semi - topological defects [ because only one direction of @xmath19 is actually gauged ] , which may be shown @xcite to be dynamically stable and moreover superconducting .
the second approach , the one we shall follow , is to regard the higgs representation , and thus the higgs field @xmath15 itself , as fundamental , and to extend the gauge group .
it turns out that the most simple such extension one can think of , consisting in an extra @xmath10 , also generates topologically stable ( and superconducting ) cosmic strings @xcite .
the string - forming model we shall now examine is the following ( this section is essentially useful to fix the notation used throughout ) : initially , the symmetry @xmath20 ( with @xmath21 the extra hypercharge ) is broken down to @xmath13 by means of a higgs field @xmath22 , and this is followed by the usual electroweak phase transition .
the model is minimal in the sense that we assign a vanishing @xmath21 hypercharge for the @xmath15 field , and symmetrically assume @xmath22 to be an @xmath14 singlet .
we shall altogether neglect the fermionic sector of the model , but it may be remarked that the hypercharge @xmath21 , with the previous assignement made on the higgs fields , coincides with @xmath23 ( up to a normalisation factor absorbable in the fermionic fields ) in this sector , and that even though @xmath24 is broken , the baryonic and leptonic numbers @xmath25 and @xmath26 are conserved . also the model will be anomaly - free provided one includes a right - handed neutrino .
we therefore start with the lagrangian density ( again , without the fermions ) @xmath27 where the ( classical ) potential between the higgs fields is ( we assume that both phase transitions are second order so we neglect the logarithmic corrections @xcite in this zero temperature effective theory , see , however , ref .
@xcite on that point ) @xmath28 and we have set the covariant derivative @xmath29 with @xmath30 the generators of @xmath31 in the representation of the particle upon which the derivative acts , @xmath32 , @xmath33 and @xmath34 the gauge coupling constants of @xmath31 , @xmath35 and @xmath24 respectively , and the kinetic terms of the gauge vectors are expressed through @xmath36 @xmath37 the higgs doublet is understood as @xmath38 , and its vacuum expectation value ( vev ) is experimentally known to be @xmath39 gev .
we are now interested in a vortex solution of this model , of the kind proposed by nielsen and olesen @xcite , for the @xmath22 field .
since we are concerned by classical solutions , it is necessary that we fix the gauges . for the string - forming fields ,
there is a particularly convenient gauge choice : if the vortex solution is taken to be aligned along the @xmath40 axis ( which is always possible since the curvature effects can be locally neglected ) , we can choose a cylindrical coordinate system , and in this system , the phase of the @xmath22 field is identified with the angular coordinate @xmath41 .
the nielsen olesen vortex solution then takes the simple form @xmath42 with @xmath43 the winding number @xcite .
let us now turn to the electroweak fields .
because of the disjoint structure of the initial invariance , we have not lost any freedom in going to the vortex gauge .
we can thus choose the most convenient gauge with regard to the subsequent interpretation , namely the unitary gauge , in which only the neutral component of @xmath15 is considered : @xmath44 $ ] . before going any further in the resolution of the euler
lagrange equations for this system , we wish to examine in more details what occurs in the strings core .
the string solution is defined as the set of points in space where @xmath45 .
moreover , the vacuum ( or the false vacuum in the case of the strings core ) should represent a minimum of the potential ( [ pot ] ) .
varying this potential for @xmath46 and @xmath47 , we see that the extremization yields two differents possibilities , namely , far from the strings core , i.e. , in the usual vacuum , @xmath48 whereas in the strings core with @xmath49 , then @xmath46 should satisfy ( not taking the kinetic terms into account for the moment ) @xmath50 from which we can conclude that two cases may occur in principle .
the first case , already studied elsewhere @xcite , is for @xmath51 , with @xmath52 which corresponds to a shift in the @xmath14 doublet higgs vev at @xmath53 . the second case , to which we now turn definitely , is for @xmath54 .
if the underlying parameters are such that this inequality is satisfied , then there is no real solution to eq .
( [ s - vac ] ) .
thus , one finds that the real minimum of the potential is now at @xmath55 as long as @xmath56 , where @xmath57 [ given by inserting a nonzero value for @xmath58 into eq .
( [ s - vac ] ) and solving for @xmath55 ] .
1 illustrates the internal string structure which is obtained when the kinetic terms are included .
this figure represents a solution of the field equations derived from the lagrangian ( [ lag ] ) under the gauge assumptions and with the vortex solution ( [ no ] ) , with zero vector fields @xmath60 and @xmath61.this solution was obtained by means of a successive over relaxation method @xcite , and the distances are in units of the inverse @xmath22 mass @xmath62 . more details concerning the numerical procedure itself and the stability of the solution can be found in ref .
@xcite , but here , and in particular in the next section , we shall be mainly interested in what occurs close to the strings core , namely the symmetry restoration .
for the time being , let us just remark that since the higgs field @xmath46 is real , there is no associated current with it , so the fact that it be trapped in the string , its vev varying from @xmath53 to @xmath63 , merely changes the actual value of the string s energy per unit length , but otherwise does not break the lorentz boost invariance along the string .
therefore , setting the stress - energy tensor in the form @xmath64 with @xmath65 and @xmath66 two unit timelike and spacelike vector respectively , tangent to the strings worldsheet , @xmath4 being the energy per unit length and @xmath6 the tension , the lorentz invariance requires , whether there is a higgs condensate or not , that the equation of state be that of goto nambu , i.e. , @xmath67cte @xcite .
this is important because once the current - generation mechanism which we will investigate in the next section has been at work , this degeneracy in the stress - energy tensor eigenvalues is spontaneously raised , so the lorentz invariance is spoiled .
the electroweak vacuum surrounding a cosmic string of the kind we just investigated is in fact not stable .
this can be seen as follow : in the standard vacuum , the @xmath8 particles are charged and massive because of the higgs field @xmath15 vev .
now , close to the string core , as we have just seen , this vev actually vanishes so the @xmath8 particles become charged and massless . as a result , they can be created by pair through any fluctuation of the electromagnetic field , but since they are charged , they can actually be considered themselves as the sources for these electromagnetic fluctuations . more precisely , as will be shown in this section , fluctuations in the @xmath0 field yield a corresponding nonvanishing @xmath68 and @xmath69 , with nonzero gradients .
this implies nonzero electric and magnetic fields which are used as negative masses for the @xmath0 particles .
the vacuum surrounding a cosmic string is thus unstable and there is a spontaneous current generation in the form of @xmath0 flowing along the strings . to see how this phenomenon actually occurs ,
let us concentrate on the stress energy tensor @xmath70 given by @xmath71 and in particular the energy density @xmath72 .
setting as usual @xcite @xmath73 , @xmath74 , @xmath75 , @xmath76 , @xmath77 , @xmath78 , @xmath79 , and using the vortex ansatz ( [ no ] ) gives , if one considers only a configuration where radial electric and orthoradial magnetic fields are present ( i.e. , with @xmath80 , @xmath81 , @xmath82 and @xmath83 the only nonvanishing components of the photonic and the @xmath69 fields ) @xmath84 + w^-_{\mu t } \left [ w^{+\mu } ( sa_t + c z_t ) - w^+_t ( sa^\mu + c z^\mu ) \right]\right\ } \nonumber \\ & + & { 1\over 2}ig\left\ { w^+_{\mu\nu}\left [ w^{-\nu } ( sa^\mu + c z^\mu ) - w^{-\mu } ( sa^\nu + c z^\nu ) \right ] + w^-_{\mu\nu } \left [ w^{+\mu } ( sa^\nu + c z^\nu ) - w^{+\mu } ( sa^\nu + c z^\nu ) \right ] \right\ } \nonumber\\ & + & { 1\over 4}g^2h^2(|w_r|^2 + |w_z|^2 + { 1\over r^2 } |w_\theta |^2 + |w_t|^2 ) \nonumber\\ & + & ig[(sa'_z + cz'_z)(w_r^- w^+_z - w_z^-w^+_r ) + ( sa'_t + cz'_t)(w_r^- w^+_t - w_t^-w^+_r ) ] \nonumber\\ & - & g^2 \big\ { \big| ( sa_z + cz_z ) w^+_z + ( sa_t+cz_t)w^+_t \big| ^2 \nonumber\\ & + & { 1\over 2}(w^+_r w^-_t - w^-_r w^+_t ) ^2 + { 1\over 2}(w^+_z w^-_t - w^-_z w^+_t ) ^2 + { 1\over 2}(w^+_r w^-_z - w^-_r w^+_z ) ^2 \nonumber\\ & + & { 1\over 2 r^2 } \left [ ( w^+_r w^-_\theta - w^-_r w^+_\theta ) ^2 + ( w^+_z w^-_\theta - w^-_z w^+_\theta ) ^2 + ( w^+_t w^-_\theta - w^-_t w^+_\theta ) ^2 \right ] \nonumber\\ & - & ( |w_r|^2 + |w_z|^2 + { 1\over r^2 } |w_\theta |^2 + |w_t|^2 ) \left [ ( sa_z + cz_z)^2 + ( sa_t+cz_t)^2 \right ] \big\ } , \label{bigu}\end{aligned}\ ] ] where a prime means differentiation with respect to the radial coordinate @xmath85 .
let us now consider the quadratic terms in @xmath8 that are present in eq .
( [ bigu ] ) , i.e. , the effective mass matrix @xmath86 , defined by @xmath87 because of the coupling between the @xmath0 fields and the gradients of the photon and the @xmath69 fields , this mass matrix is in fact nondiagonal , and one can easily derive the eigenvalues as @xmath88 @xmath89 @xmath90 where , to first order in @xmath32 , the eigenvectors are @xmath91 @xmath92,\ ] ] @xmath93,\label{w3}\ ] ] and we have set @xmath94 , @xmath95 the radial ( respectively orthoradial ) component of the electric ( resp .
magnetic ) field .
it can be remarked on eqs .
( [ m1 ] ) , ( [ m2 ] ) and ( [ m3 ] ) that @xmath96 has a nonpositive definite mass , which is not the case for all other components of this gauge field .
therefore , since we are seeking a minimum energy configuration , it seems safe to assume @xmath97 .
setting , for simplicity , @xmath98 , one has @xmath99
@xmath100 @xmath101 where now @xmath102 and @xmath103 are arbitrary .
we recover the previous results @xcite @xmath104 or @xmath105 for the purely magnetic or electric cases respectively ( i.e. , when only one component of @xmath68 and @xmath69 is explicitely considered ) . under the assumptions given by eqs .
( [ wr ] ) , ( [ wz ] ) and ( [ wt ] ) , it is now simple to derive the effective potential for the @xmath0 field , namely : @xmath106 so this field effectively behaves as a higgs `` scalar '' in the region of parameter space where its squared mass is negative .
this is indeed possible close to the strings core since there , as we have seen , still in the case where @xmath107 , the higgs field @xmath46 vanishes , so to first order in @xmath32 , eq .
( [ m3 ] ) implies that @xmath108 is actually negative .
thus , any fluctuation in @xmath0 , the latter being coupled with the photon , will generate a small fluctuation in @xmath109 , @xmath110 , @xmath111 and @xmath112 , which , if the initial perturbation was axisymmetric , will produce an electric or a magnetic field .
since @xmath0 is effectively massless in the core of the string , the energy in the electromagnetic perturbation is already sufficient to create a pair @xmath113 , which can in turn be seen as the source for the electromagnetic fields . since these electromagnetic fields are necessary to support the @xmath0 condensate in the strings core , one is led to the conclusion that a current has been spontaneously generated .
we shall now investigate in more details this mechanism .
to exhibit the instability of the electroweak vacuum surrounding a nielsen - olesen string ( [ no ] ) , we turn to the euler - lagrange equations derivable from the lagrangian ( [ lag ] ) , which we expand to first order in the coupling constant @xmath32 and to lowest order in the various fields involved to consider the case of a perturbation in @xmath96 [ given by eq .
( [ w3 ] ) ] . for the photon
, we have @xmath114 \right\ } = rigs \left ( w^{-\alpha\beta } w^+_\beta - w^{+\alpha\beta } w^-_\beta \right),\label{apert}\ ] ] a similar equation applying for the @xmath69 field with @xmath115 replaced by @xmath116 , so also similar conclusions can actually be drawn for both fields , and @xmath117 \right\ } = rigs \left [ w^{-\beta\alpha } ( sa_\beta + cz_\beta ) + ( sa^{\beta\alpha } + cz^{\beta\alpha})w^-_\beta \right ] , \label{wpert}\ ] ] close to the strings core , in the symmetry restoration region where @xmath55 . in fact , because there , @xmath55 , neglecting a possible backreaction due to outer region couplings , we see that the background nielsen - olesen string is essentially unaffected by the inclusion of the @xmath0 , @xmath68 and @xmath69 fields .
we shall now examine a perturbation in @xmath118 in the form @xmath119 . inserting this form into eq .
( [ apert ] ) yields @xmath120 @xmath121 @xmath122 equations in which we shall assume for now on that @xmath123 is ignorable , and we will denote the differentiation with respect to the radial coordinate @xmath85 by a prime , whereas a dot will mean a derivative with respect to time .
we shall also work in the lorentz gauge @xmath124 since in this gauge the equations for the various components of @xmath68 are decoupled . actually , the gauge condition together with the field equation ( [ ar ] ) with @xmath125 ignorable yields , upon differentiation with respect to @xmath85 @xmath126 and inserting this into eq .
( [ ar ] ) yields @xmath127 so that setting @xmath128 gives @xmath129 with @xmath130 a constant and @xmath131 the two dimensionnal laplacian in the transverse plane .
so @xmath132 , and the radial dependent part obeys a schrdinger equation @xmath133 with a positive definite potential @xmath134 .
therefore , the eigenvalues @xmath135 of eq .
( [ schro ] ) are all positive and the string state is stable against @xmath136 perturbations . thus
, neglecting @xmath136 in the resulting stationnary configurations is justified and we shall consistently set @xmath137 in what follows . inserting eq .
( [ gauge ] ) into eq .
( [ at1 ] ) is the last step toward the definite equations describing the dynamics of @xmath109 and @xmath110 when a perturbation in @xmath96 is applied , and we find @xmath138 @xmath139 out of which the spontaneous current mechanism can be clearly exhibited .
the linearized field equations ( [ az ] ) and ( [ at ] ) have a first immediate consequence , namely that @xmath140 .
it turns out that this relation is still valid when the whole set of classical field equations are used , so the overall field configuration is in fact determined by the value of the ratio @xmath141 .
we will return to that point later .
moreover , eqs .
( [ az ] ) and ( [ at ] ) , being inhomogeneous because of the source term due to the @xmath0 field [ a direct consequence of the nonabelian nature of @xmath142 , the configuration @xmath143 and @xmath144 is not solution of the field equations , and more generally there is also no solution with vanishing gradient .
but this is precisely the condition for the squared mass @xmath108 to be nonpositive definite .
thus , we know that there exist unstable modes in eq .
( [ wpert ] ) .
these modes will grow exponentially , as we shall now show , as will also @xmath110 and @xmath109 , until they reach an equilibrium configuration where the quadratic terms become large enough to stop the instability . to examine the @xmath0 instability , we note first that eq .
( [ wpert ] ) is not well adapted since it was written for the usual components of @xmath145 .
instead , we write an equivalent linearized action [ obtained by retaining only the lowest order tems in the lagrangian ( [ lag ] ) ] for the field @xmath146 alone , namely @xmath147 \nonumber\\ & + & g(sa_t+cz_t)\big [ { \beta_t \over \sqrt{\beta_z^2 + \beta_t^2 } } ( w'^\star w + w^\star w ' ) + i{2\beta_z^2 + \beta_t^2\over\beta_z^2 + \beta_t^2 } ( \dot w w^\star - \dot w^\star
w ) \big ] \nonumber\\ & + & { 2g\over\sqrt{\beta_z^2 + \beta_t^2 } } \big [ \beta_z ( sa'_z+cz'_z ) - \beta_t ( sa'_t+cz'_t ) \big ] |w|^2 \big\ } \end{aligned}\ ] ] whose variations yield the following euler - lagrange equation : @xmath148 where we have set @xmath149 , @xmath150 , @xmath151 @xmath152,\ ] ] @xmath153 and @xmath154 .\ ] ] although it is impossible to solve eq .
( [ wfin ] ) , some information regarding the transition from the nonconducting state to the current - carrying state can be obtained from it if one makes an `` adiabatic '' hypothesis , namely assumes that the transition is slow enough that the time derivative in the gauge fields can be neglected compared to their spatial gradients . in this hypothesis , setting @xmath155 into eq .
( [ wfin ] ) yields @xmath156 where the function @xmath157 depends on the radial coordinate @xmath85 only .
thus , the oscillatory modes @xmath158 satisfy a dispersion relation @xmath159 whose solutions have an imaginary part for @xmath160 , i.e. , in the low frequency limit ( thereby justifying the `` adiabatic '' hypothesis ) . in the limit of zero frequency ( @xmath161 ) , we can estimate roughly , in order of magnitude , the expected value of the timescale necessary for the string to become current - carrying ( namely @xmath162 ) : assuming the current - carrier field to have an amplitude @xcite @xmath163 , and taking @xmath164 to be the typical distance over which the fields vary , then eqs . ( [ az ] ) and ( [ at ] ) give @xmath165 , and we find @xmath166 i.e. , a time independent of the string thickness , an expected result since the background string and the electroweak fields are decoupled in the core ( again , neglecting backreaction ) .
it should be mentionned that for coupling values @xmath51 , the same mechanism actually applies , but that in this case , the initial configuration and only metastable and although the current will definitely be spontaneously generated , it will be through tunnelling . as a result , the life time of the noncurrent - carrying configuration is in fact increased by a factor depending on the strings thickness and the value of @xmath46 at @xmath53 , these giving the order of magnitude of the expected potential barrier and its width : the standard wkb approximation then gives an extra exponential factor in eq .
( [ time ] ) .
the spontaneous current generation mechanism we have just discussed has in fact many interesting consequences , including , we believe , cosmological ( notably in the framework of the vorton problem @xcite which becomes even more unavoidable in this context ) , and in this section , we wish to emphasize a particular effect , namely that generating a current this way breaks the lorentz boost invariance along the string spontaneously .
the basic reason that this occurs is that the particle that gets trapped in the string is in fact a @xmath167 , i.e. , a vectorial particle , and a nonvanishing vev for a vector picks a privileged direction in space time . also we shall exhibit the internal microscopic structure of the string and compare it with what is obtained in the simple witten @xcite bosonic toy model . as was already said , the current generation phenomenon will spontaneously break the lorentz boost symmetry along the string : before the @xmath0 condenses in the strings core , the energy per unit length and the tension are both equal so a boost in the @xmath40 direction does not change the physics of the system .
now , when a perturbation in @xmath0 is applied , as we have just seen , the @xmath0 and @xmath68 fields vev increase exponentially , so the degeneracy of the stress energy tensor is raised exponentially as times passes by , until again the system reaches a stationnary configuration .
it is therefore an actual spontaneous mechanism because it does not require any external field and it needs not even exist in the expanding universe ( as is the case for the usual symmetry breaking higgs mechanism ) .
a stationnary configuration obtained this way consists in a @xmath0 field together with , as argued before , any value for the ratio @xmath141 .
in fact , it turns out that the only thing to know to determine ( nearly ) entirely the configuration is whether this ratio is less or greater than unity , for once this is known , it suffices to apply a boost along the string to remove one of the field @xmath109 or @xmath110 . as a result , the only interesting cases are the magnetic case for which one can always set @xmath168 and @xmath169 , the electric case having @xmath170 and @xmath171 , and the null or lightlike case with @xmath172 , @xmath173 and @xmath174 , as can be seen on eqs .
( [ az ] ) and ( [ at ] ) .
it can then be shown @xcite that the most general configuration will be @xmath175 where , without lack of generality @xcite , the phase function @xmath176 could be chosen as @xmath177 .
the `` energy per unit length '' is then @xmath178 \upsilon ^2 + { \beta_z\partial_z\psi + \beta_t\partial_t \psi \over \sqrt{\beta^2_z+\beta^2_t } } \upsilon \upsilon ' + { 1\over 2 } ( a'^2_z + a'^2_t + z'^2_z + z'^2_t ) \nonumber\\ & + & { g^2\over 8c^2 } h^2 ( z_z^2 + z_t^2 ) + { 1\over 4 } g^2 h^2 \upsilon ^2 + { 1\over 2 } g^2 \upsilon ^4 + g^2 \upsilon ^2 [ ( sa_z+cz_z)^2 + ( sa_t+cz_t)^2]\nonumber\\ & + & g(sa_z+cz_z)\big\ { { \beta_z\upsilon\upsilon'\over \sqrt{\beta^2_z+\beta^2_t } } + \big [ \partial_z\psi + { \beta_t \over \beta^2_z+\beta^2_t } ( \beta_t \partial_z\psi + \beta_z\partial_t\psi ) \big ] \upsilon^2 \big\ } \nonumber\\ & + & g(sa_t+cz_t)\big\ { { \beta_t\upsilon\upsilon'\over \sqrt{\beta^2_z+\beta^2_t } } + \big [ \partial_t\psi - { \beta_z \over \beta^2_z+\beta^2_t } ( \beta_t \partial_z\psi + \beta_z\partial_t\psi ) \big ] \upsilon^2 \big\ } \nonumber\\ & -&{g^2\upsilon^2\over 2(\beta^2_z+\beta^2_t ) } \big [ ( sa_z+cz_z)\beta_z + ( sa_t+cz_t ) \beta_t \big]^2 - { g\upsilon^2\over \sqrt{\beta^2_z+\beta^2_t } } \big [ ( sa'_z+cz'_z)\beta_z + ( sa'_t+cz'_t ) \beta_t \big ] \big\ } , \\ \label{utilde}\end{aligned}\ ] ] and the `` tension '' @xmath179 \upsilon ^2 - { \beta_z\partial_z\psi + \beta_t\partial_t \psi \over \sqrt{\beta^2_z+\beta^2_t } } \upsilon \upsilon ' -{1\over 2 } ( a'^2_z + a'^2_t + z'^2_z + z'^2_t)\nonumber\\ & -&{g^2\over 8c^2 } h^2 ( z_z^2 + z_t^2 ) - { 1\over 2 } g^2 \upsilon ^4 - g(sa_z+cz_z)\big\ { { \beta_z\upsilon\upsilon'\over \sqrt{\beta^2_z+\beta^2_t } } + \big [ \partial_z\psi - { \beta_t \over \beta^2_z+\beta^2_t } ( \beta_t \partial_z\psi + \beta_z\partial_t\psi ) \big ] \upsilon^2 \big\ } \nonumber\\ & - & g(sa_t+cz_t)\big\ { { \beta_t\upsilon\upsilon'\over \sqrt{\beta^2_z+\beta^2_t } } + \big [ \partial_t\psi + { \beta_z \over \beta^2_z+\beta^2_t } ( \beta_t \partial_z\psi + \beta_z\partial_t\psi ) \big ] \upsilon^2 \big\}\nonumber\\ & -&{g^2\upsilon^2\over 2(\beta^2_z+\beta^2_t ) } \big [ ( sa_z+cz_z)\beta_z - ( sa_t+cz_t ) \beta_t \big]^2
+ { g\upsilon^2\over \sqrt{\beta^2_z+\beta^2_t } } \big [ ( sa'_z+cz'_z)\beta_z + ( sa'_t+cz'_t ) \beta_t \big ] \big\ } , \\ \label{ttilde}\end{aligned}\ ] ] where the quotes [ and the subsequent tildes in eqs .
( [ utilde ] ) and ( [ ttilde ] ) ] in the previous denominations for the energy per unit length and tension come from the fact that the stress tensor also has a nondiagonal part @xmath180 \nonumber\\ & + & g^2\big [ ( w^+_r)^2 w^-_t w^-_z + ( w^-_r)^2 w^+_t w^+_z - |w_r|^2 ( w^+_z w^-_t + w^+_t w^-_z ) - 2 |w_r|^2 ( sa_z+cz_z)(sa_t+cz_t)\big ] \big\ } , \end{aligned}\ ] ] and this part can be removed by choosing the reference frame in which either @xmath103 or @xmath102 vanishes , which is always possible provided they are not equal , i.e. , provided the current is not lightlike . in this case
however , the eigenvectors @xmath181 and @xmath182 of the stress energy tensor ( [ tmn ] ) are both null and the definition of the eigenvalues as energy per unit length and tension becomes unclear .
investigation of such lightlike currents in strings is posponed for future work @xcite . therefore , in general , as was already discussed in ref .
@xcite , the knowledge of the energy per unit length and the tension in the particular cases @xmath183 , @xmath184 and @xmath185 , @xmath186 will be sufficient for macroscopic applications .
for these two cases , it turns out that @xmath4 and @xmath6 ( without the tilde , i.e. , after diagonalization of the stress tensor , this operation corresponding to choosing [ @xmath170 , @xmath187 or [ @xmath168 , @xmath188 ) actually depend on the gauge fields @xmath189 and @xmath190 only through the new field functions @xmath191 and the orthogonal field @xmath192 where @xmath193 denotes either @xmath40 or @xmath194 depending on to whether one considers a magnetic or an electric state respectively , and where the parameter @xmath195 is not in fact arbitrary since it is given by @xmath196 . as a result ,
the realistic string model given by the lagrangian ( [ lag ] ) is nearly as simple as the witten @xcite bosonic toy model in the sense that out of the initial 18 field functions ( @xmath47 , @xmath46 , @xmath197 , @xmath198 , @xmath189 , and @xmath199 ) , the internal microscopic structure is fully determined by the knowledge of 6 field functions [ @xmath200 , @xmath201 , @xmath202 , @xmath203 , @xmath204 and @xmath205 and two free parameters , namely the winding number @xmath43 and the state parameter @xmath206 .
this is to be compared to the original witten bosonic model whose structure needs the knowledge of already 4 field functions and the same two free parameters , and with very similar field equations . thus we believe that , apart from the spontaneous current generation mechanism discussed in the present article , most qualitative conclusions regarding this simple model should apply as well to this more realistic model .
in examining the internal structure of cosmic strings arising in the most simple string - forming extension of the standard electroweak model , we have found that , because of the nonabelian nature of @xmath207 , the field @xmath0 can condense spontaneously in a strings core if the coupling constant between the string - forming higgs field and the usual @xmath14 doublet higgs is less than a critical value .
this phenomenon can be understood in the following way : for certain values of the coupling constants between the string forming higgs field @xmath22 and the @xmath14 doublet higgs field @xmath15 , the latter has a vanishing vev close to the strings core , so the initial @xmath207 symmetry is restored .
therefore , the intermediate vector bosons , just like the photon , remains massless in this region .
any exitation of the photon field will thus have enough energy to generate a pair @xmath113 through vacuum fluctuations . in turn , the effectively massless @xmath0 particle , being charged , is responsible for the existence of a nonvanishing electromagnetic field .
this turns out to be in fact an unstable fluctuation mode , and nonzero vev for @xmath0 and the photon therefore build up spontaneously . by using the field equations for the electroweak fields in the symmetry restored region , we have been able to exhibit explicitely this instability , and to estimate what we believe to be a lower bound on the time necessary for the current to be generated
this timescale is , as expected , independent of the underlying string parameters provided the latter are such that the symmetry restoration mechanism actually occurs .
it should be remarked that because the phenomenon here described is essentially electromagnetic and involves only the @xmath0 particle , the timescale found could have been deduced on dimensionnal ground , namely @xmath208 .
although this is a huge time compared to the characteristic length of the string if the underlying string forming theory is at gut scale , it is still sufficiently short to be irrelevant for cosmological considerations .
thus , current - carrying strings seem quite generic in string - forming gut models since the potential ( [ pot ] ) is in fact very general even as a low energy limit and , as we have seen , the current formation mechanism is independent of the background string structure .
considering the results of ref .
@xcite and the present calculation , it can be concluded that for any value value of the coupling between the string forming theory and the electroweak fields , the resulting strings are superconducting in the sense of witten , whether the current builds up through tunneling ( high frequencies metastability @xcite ) or instability .
thus , if cosmic string exist , and if they are not arbitrarily decoupled from the low energy physics ( a requirement of `` naturalness '' ) , then , they are superconducting . since the present knowledge in high energy physics tells us that approximately half of the plausible gut theories contains cosmic strings , it means that we can estimate the existence probability of superconducting cosmic string to be also of the order 1/2 .
a final remark seems appropriate at this point : the current generation we have exhibited here relies in fact entirely on the nonabelian nature of @xmath14 .
this means that for a string forming gut , it will exist as well since gut models usually involve large unifying groups with nonabelian couplings , and various higgses
. therefore , gauge bosons having masses of the order of the gut scale should spontaneously condense in the string core , within a timescale given this time by @xmath209 , where @xmath32 is the gut group coupling constant .
the cosmological relevance of such effects is then obvious since many of these gauge bosons are responsible for baryon number violation , so that in particular , these strings would enhance the primordial baryon number asymmetry .
i wish to thank b. carter , a .- c .
davis , p. fayet , j. garriga , m. hindmarsh , e. p. s. shellard and a. vilenkin for many interesting and improving discussions .
this work was supported by serc grant # 15091-aoz - l9 .
r. davis , , 3705 ( 1987 ) ; n. turok , , 2225 ( 1989 ) ; d. spergel , n. turok , w. press , b. ryden , , 1038 ( 1991 ) ; j. borrill , e. copeland , a. liddle , phys .
b * 258 * , 310 ( 1991 ) ; l. perivolaropoulos , , 1858 ( 1992 ) d. bennet , f. r. bouchet , , 257 ( 1988 ) ; , 2408 ( 1990 ) ; a. albrecht , n. turok , _ ibid _ * 40 * , 973 ( 1989 ) ; b. allen , e. p. s. shellard , _ ibid _ * 64 * , 119 ( 1990 ) ; a. stebbins , s. veeraraghavan , r. brandenberger , j. silk , n. turok , , 1 ( 1987 ) ; r. h. brandenberger , preprint brown - het-906 b. carter , in _ the early universe and cosmic structures _ , proceedings of the xth moriond astrophysics meeting , les arcs , france , 1990 , edited by a. blanchard and j. tran thanh vn ( editions frontires , gif - sur - yvette , 1990 ) e. copeland , m. hindmarsh , n. turok , phys .
lett . * 58 * , 1910 ( 1987 ) ; r. l. davis , e. p. s. shellard , phys .
b * 209 * , 485 ( 1988 ) ; _ ibid _ * 207 * , 404 ( 1988 ) ; c. t. hill , h. m. hodges , m. s. turner , , 263 ( 1988 ) ; a. babul , t. piran , d. n. spergel , phys .
b * 202 * , 307 ( 1988 ) ; e. copeland , d. haws , m. hindmarsh , n. turok , nucl .
b * 306 * , 908 ( 1988 ) ; d. haws , m. hindmarsh , n. turok , phys . lett .
b * 209 * , 255 ( 1988 ) p. peter , , 1091 ( 1992 ) ; _ ibid _ * 46 * , 3335 ( 1992 ) b. carter , in _ formation and evolution of cosmic strings _ , edited by g. w. gibbons , s. w. hawking and t. vachaspati ( cambridge university press , cambridge , england , 1990 ) ; phys .
b * 228 * , 466 ( 1989 ) t. m. helliwell , d. a. konkowski , , 1918 ( 1986 ) ; i. moss , s. poletti , phys .
b * 199 * , 34 ( 1987 ) ; a. babul , t. piran , d. n. spergel , _ ibid _ * 209 * , 477 ( 1988 ) ; p. amsterdamski , p. laguna - castillo , , 877 ( 1988 ) ; b. linet , class .
quantum grav .
* 6 * , 435 ( 1989 ) ; p. peter , d. puy , , ( 1993 ) t.w.b .
kibble , g. lazarides , q. shafi , phys . lett .
b * 113 * , 237 ( 1982 ) ; p. fayet , s. ferrara , phys .
rep . * 32 * , 249 ( 1977 ) ; p. fayet , nucl .
b * 347 * , 743 ( 1990 ) r. w. robinett , j. rosner , , 3036 ( 1982 ) ; c. n. leung , j. rosner , , 2132 ( 1984 ) f. del aguila , m. quirs , f. zwirner , nucl .
b * 287 * , 419 ( 1987 ) ; t. m. bibilashvili , g. r. dvali , phys .
b * 263 * , 359 ( 1991 ) ; _ ibid _ * 248 * , 259 ( 1990 ) l. s. durkin , p. langacker , phys .
b * 166 * , 436 ( 1986 ) ; particle data group , , part 2 ( 1992 ) ; phys . lett .
b * 276 * , 247 ( 1992 ) ; k. hagiwara , r. najima , m. sakuda , n. terunuma , , 815 ( 1990 ) ; m. c. gonzalez garcia , j. f. w. valle , , 360 ( 1990 ) ; u. amaldi _ et al .
_ , , 1385 ( 1987 ) ; f. del aguila , j. m. moreno , m. quirs , phys .
b * 254 * , 497 ( 1991 ) ; g. altarelli _ et al .
b * 263 * , 459 ( 1991 ) ; p. langacker , m. luo , , 278 ( 1992 ) | it is shown that in models including the standard electroweak theory and for some particular values of the underlying parameters , electric currents can be spontaneously generated in cosmic strings , without the need of any external field ( e.g. , electric or magnetic ) as is required in most models .
this mechanism is then shown to break spontaneously the lorentz invariance along the initially goto - nambu string . the characteristic time needed for the current to build up
is estimated and found to lowest order to depend only on the mass of the intermediate @xmath0 vector boson and the fine structure constant . |
the properties of neutrinos attract a strong attention since neutrinos alone are the elementary particles showing the evidence of new physics beyond the standard model . in particular
, recent experiments of the superkamiokande ( sk ) detector @xcite and the sudbury neutrino observatory ( sno ) @xcite have shown that neutrino have non - zero mass and mixing angles by solar @xcite and atmospheric @xcite neutrino data .
these neutrino oscillation experiments enable us to constrain the squared mass difference between mass eigenstates and mixing angles .
the solar neutrino experiments strongly support the matter - induced large mixing angle ( lma ) solution ( @xmath4 @xcite ) and the atmospheric neutrino experiments have proved that @xmath5 and @xmath6 are maximally mixing ( @xmath7 @xcite ) .
however , there are several parameters which are not sufficiently constrained .
one such parameter is @xmath8 , for which only upper bound is known from reactor experiment @xcite ( @xmath9 ) .
in addition , the nature of neutrino mass hierarchy ( normal or inverted ) is also matter of controversy .
there is another astrophysical source of neutrinos , or supernova .
neutrino burst from the next galactic core - collapse supernova will be captured by the present underground detectors , such as sk and sno .
it is widely believed that the detected events will provide rich information not only on the mechanism of supernovae but also on the physical properties of neutrinos themselves .
in fact , neutrino signals from sn1987a which were detected by the kamiokande @xcite and imb @xcite detectors permitted many discussions about the fundamental properties of neutrinos @xcite , although the detected event numbers were quite small ( 11 for kamiokande , 8 for imb ) .
what we can learn from the next galactic supernova is considered in many articles ( for a review , see ref .
@xcite ) . for example
, we can constrain the properties of neutrino oscillation such as @xmath8 and the mass hierarchy @xcite .
in addition to neutrino mass , non - zero magnetic moment is another nature of neutrinos beyond the standard model .
the most stringent upper bound of the neutrino magnetic moment is obtained from cooling argument : any new physics that would enhance the neutrino diffusion , such that the cooling time drops below the observed duration , must be excluded .
the recently revised upper bound is @xmath10 @xcite , where @xmath11 is the bohr magneton .
the above value is several orders of magnitude more stringent than the existing limit from reactor @xcite and solar neutrino experiments @xcite , and comparable to the bound from globular cluster red giant cooling arguments @xcite , which give @xmath12 . on the other hand , from the minimally - extended standard model calculation
, we obtain @xmath13 , which is many orders of magnitude smaller than the current experimental upper bounds .
however , some particle physics models have been proposed to induce a larger magnetic moment of order @xmath14 @xcite , which is comparable to the current upper bounds .
if neutrinos have the non - zero magnetic moment , it leads to precession between left- and right - handed neutrinos in sufficiently strong magnetic fields expected in the inner region of the core - collapse supernova @xcite . in general ,
non - diagonal elements of the magnetic moment matrix are possible and neutrinos can be changed into different flavors and chiralities @xcite . furthermore , with the additional effect of the coherent forward scattering by matter , neutrinos can be resonantly converted into those with different chiralities @xcite by a mechanism similar to the well known mikheyev - smirnov - wolfenstein ( msw ) effect @xcite .
this resonant spin - flavor ( rsf ) conversion induced by the neutrino magnetic moment in the supernova has been studied by many authors @xcite .
almost all these earlier publications considered the rsf conversions between two - flavor neutrinos ( @xmath15 ) .
however , since all three - flavor neutrinos and anti - neutrinos are emitted from the supernova core , we should calculate conversion probabilities using three - flavor formulation to obtain the spectrum of the neutrinos in the detectors .
fortunately , our knowledge of the neutrino mixing parameters permits the three - flavor analyses , which have not been considered because of many unknown parameters in those days . in this paper
, we study the rsf effect in detail , including three flavor neutrinos and anti - neutrinos .
particularly , a new crossing diagram , which deals with both msw and rsf effects , is given . from the diagram
, we find that the combination of the msw and the rsf effects makes crossing schemes very interesting to investigate , and the expected flux would be different from that obtained in the case of pure msw or pure rsf effect .
in addition , we calculate numerically the conversion probabilities of neutrinos and anti - neutrinos in the supernova using three - flavor formulation . in the calculations ,
we adopt the latest neutrino mixing parameters , which are constrained by the solar and the atmospheric neutrino experiments , and assume the normal mass hierarchy .
however , there are two unknown parameters of neutrinos , @xmath16 and @xmath8 . , and magnetic field .
we give in section [ sec : models ] the adopted properties in our calculation .
] for @xmath16 , we assume the value @xmath17 , ( see section [ sec : formulation ] ) , the results would also apply to any other value of @xmath16 provided that the supernova magnetic field strength is rescaled accordingly . ] and for @xmath8 , two large and small values ( @xmath18 ) . using the calculated conversion probabilities ,
we give the expected energy spectra at sk and sno detectors , which show various behaviors as we change the value of @xmath19 .
finally , we discuss the model uncertainties and how to obtain the implication for the value of neutrino magnetic moment from the obtained energy spectra .
this paper is organized as follows . in section [ sec : formulation ] , we give the formulation used in our calculation , which includes all three flavor neutrinos and anti - neutrinos . in section [ sec : models ] , the neutrino oscillation models and supernova models ( original neutrino spectrum , density , @xmath20 , and magnetic field profiles ) considered in this paper are illustrated and their validity is also discussed . in section [ sec :
level crossing schemes and qualitative illustrations ] , we give a new crossing diagram , with which qualitative discussions concerning neutrino conversions in supernova matter become possible . the numerically calculated conversion probabilities and energy spectra at sk and sno are shown in section [ sec : results ] .
finally , detailed discussions on model uncertainties and how to obtain the implication for neutrino magnetic moment are presented in section [ sec : discussion ] .
the interaction of the magnetic moment of neutrinos and magnetic fields is described by @xmath21 where @xmath22 is the component of the neutrino magnetic moment matrix , @xmath23 the magnetic field transverse to the direction of propagation , @xmath24 and @xmath25 the right- and left - handed neutrinos , respectively , and @xmath26 and @xmath27 denote the flavor eigenstate of neutrinos , i.e. , @xmath28 and @xmath29
. if neutrinos are the dirac particles , right - handed neutrinos and left - handed anti - neutrinos are undetectable ( sterile neutrinos ) , since they do not interact with matter .
the spin - flavor conversion into these sterile neutrinos suffers strong constraints from observation of neutrinos from sn1987a @xcite . on the other hand ,
if neutrinos are the majorana particles , @xmath30 s are identical to anti - particles of @xmath31 s and interact with matter , and the constraint above becomes considerably weak . in this paper , we assume that neutrinos are majorana particles .
the diagonal magnetic moments is forbidden for the majorana neutrinos , and therefore , only the conversion between different flavors is possible , e.g. , @xmath32 .
the coherent forward scattering with matter induces the effective potential for neutrinos , which is calculated using weak interaction theory . the effective potential due to scattering with electrons is given as @xmath33 where @xmath34 is electron number density , @xmath35 the fermi coupling constant , and @xmath36 the weinberg angle .
the @xmath37 sign in front refers to @xmath38 and @xmath39 and that in the parentheses to @xmath40 and @xmath41 .
the difference between @xmath42 and @xmath43 neutrinos comes from the existence of charged - current interaction .
the subscript @xmath44 of @xmath45 refers to the first and the second @xmath37 sign .
the ordinary msw effect between @xmath46 and @xmath47 is caused by the potential difference , @xmath48 .
to include the rsf effect , which causes the conversion between neutrinos and anti - neutrinos , we should take into account the neutral - current scattering by nucleons : @xmath49 where @xmath50 are proton and neutron number density , respectively .
for neutrinos we add @xmath51 to the potential and for anti - neutrinos @xmath52 .
therefore , the rsf conversion between @xmath53 and @xmath47 obeys the potential difference @xmath54 where @xmath55 is the density , @xmath56 the nucleon mass , and @xmath57 is the number of electrons per baryon .
( when we obtain eq .
( [ eq : delta v ] ) , we assumed charge neutrality @xmath58 . ) in this subsection , we give the simplest formulation between @xmath53 and @xmath5 , and consider the properties of the rsf conversion .
the time evolution of the mixed state of @xmath53 and @xmath5 is described by the schrdinger equation @xmath60 where @xmath61 is the radius from the center of the star , @xmath62 the transition magnetic moment , and @xmath63 is defined by @xmath64 the resonance occurs when @xmath65 , thus , the necessary condition is @xmath66 .
( if @xmath67 , the @xmath68 conversion occurs conversely , at almost the same place . )
several earlier authors @xcite pointed out that the rsf conversion occurs quite efficiently in the region above the iron core and below the hydrogen envelope of collapsing stars , namely , in the `` isotopically neutral region '' ( o+si , o+ne+mg , o+c , and he layers ) , in which the value of @xmath20 is slightly less than 0.5 [ typically , @xmath69 .
actually , the pre - supernova model used in our calculation have the similar profile , as is shown in section [ sub : supernova progenitor model ] . when the resonance is adiabatic ( or when the density is slowly changing at that point and the magnetic field is strong enough ) , the significant conversion occurs .
the adiabaticity of the rsf conversion can be derived by the same argument used in the case of the msw conversion @xcite .
it is given by @xmath70 where the subscript `` res '' means that the value is evaluated at the resonance point .
therefore , if the magnetic field is sufficiently strong at the resonance point , the @xmath15 conversion occurs completely . at the end of this subsection
, we discuss the possibility that @xmath71 conversion occurs in solar magnetic field .
there are several works examining the rsf conversion as a solution to the solar neutrino problem @xcite . in this paper , however , we postulate that the solution to the solar neutrino problem is given by the ordinary msw effect ( lma solution ) .
in fact , if we take the mixing parameters of lma ( @xmath72 ) in calculating the rsf resonance condition @xmath73 , the required density is found to be larger than the central density of the sun .
( in this calculation , the sign of @xmath74 is changed in eq .
( [ eq : delta h ] ) , since @xmath67 in the sun .
this corresponds to @xmath75 conversion . ) here , we present the three - flavor ( six - component ) formulation of neutrino mixing by @xmath76 where @xmath77 @xmath78 @xmath79 and @xmath80 .
( we assume cp phase @xmath81 in eq .
( [ eq : u ] ) for simplicity . )
the resonant flavor conversion occurs when the two diagonal elements in matrix in eq .
( [ eq : three - flavor ] ) have the same value .
there are five such resonance points , which are for @xmath82 ( msw - l ) , @xmath83 ( msw - h ) , @xmath15 ( rsf - l ) , @xmath84 ( rsf - h ) , and @xmath85 conversions .
suffixes `` -l '' and `` -h '' attached to `` msw '' and `` rsf '' indicate whether the density at the resonance points are lower or higher .
hereafter , we neglect @xmath86 conversion , since it is always nonadiabatic as well as including this makes discussion further complicated .
we adopt the neutrino mixing parameters introduced to explain the observations of the solar and the atmospheric neutrino experiments @xcite , and assume the normal mass hierarchy . from the atmospheric neutrino experiments , we use the values @xcite @xmath87 and from the solar neutrino experiments @xcite ,
@xmath88 in our calculations . for @xmath8 ,
which is not sufficiently constrained , we adopt two large and small values , @xmath89 where each value causes the adiabatic and nonadiabatic msw - h conversions , respectively . as for neutrino magnetic moment , there are three transition moments for majorana neutrino ( eq .
( [ eq : magnetic moment ] ) ) .
we assume that all of them are near the present upper bound ( i.e. , @xmath90 ) .
however , as we will show in section [ sub : supernova magnetic field ] , @xmath91 alone governs results .
we use a realistic model of a collapse - driven supernova by the lawrence livermore group @xcite to calculate the neutrino energy spectrum at production .
we show in fig .
[ fig : fluence ] the time - integrated energy spectra of neutrinos ( see ref .
@xcite for detail ) , where the distance to the supernova is set to 10 kpc .
( for discussions from here , we adopt this assumption , @xmath92 . )
both the flux integrated over entire time during neutrino burst and that integrated until 0.5 seconds after core bounce are shown in the same figure .
although the neutrino burst persists @xmath93 10 seconds , about half of neutrinos are emitted by 0.5 seconds , according to numerical simulation by lawrence livermore group @xcite .
the reason for using time integrated flux until 0.5 seconds is given in the next subsection .
the mean energies are different between flavors : @xmath94 for fully time - integrated spectrum , while for that integrated until 0.5 seconds , @xmath95 this hierarchy of mean energies is explained as follows .
since @xmath96 s interact with matter only through the neutral - current reactions in supernova , they are weakly coupled with matter compared to @xmath46 s and @xmath53 s .
thus the neutrino sphere of @xmath96 s is deeper in the core than that of @xmath46 s and @xmath53 s , which leads to higher temperatures for @xmath96 s .
the difference between @xmath46 s and @xmath53 s comes from that the core is neutron rich and @xmath46 s couples with matter more strongly , through @xmath97 reactions .
we use the precollapse model of massive stars of woosley and weaver @xcite .
the progenitor mass was set to be @xmath98 , and the metalicity the same as that of the sun .
the upper panel of fig .
[ fig : profiles ] shows @xmath99 and @xmath100 profiles of this progenitor model . in the same figure
, we also show @xmath101 and @xmath102 . at intersections between @xmath103 and @xmath104 ,
the rsf and msw conversions occur ( see eq .
( [ eq : delta h ] ) ) .
however , in fact , since the density profile changes drastically during neutrino burst ( @xmath105 sec ) owing to shock propagation , we can not use the static progenitor model when calculating the conversion probability in supernova @xcite .
we show in fig .
[ fig : shock ] the _ time - dependent _ @xmath106 and @xmath99 profile calculated by lawrence livermore group @xcite , compared with the static model adopted in our calculation @xcite . from the figure , the shock wave changes the @xmath106 profile in the resonance region , which is responsible for rsf conversions , at @xmath107 seconds after bounce , while that at @xmath108 seconds is almost the same as the static progenitor model .
thus , from here , we confine our discussion until 0.5 seconds after core bounce , since in that case , using the static progenitor model is considered to be a good approximation . calculating for full time - scale during neutrino burst with the time - dependent density and @xmath20 profiles makes discussion very complicated , and it is out of scope of this study , while it will be our future work .
we assume that the global structure of the magnetic field is a dipole moment and the field strength is normalized at the surface of the iron core with the value of @xmath109 gauss .
the reason for this normalization is as follows . as pointed out by totani and sato @xcite , the global magnetic field normalized at the surface of the newly born neutron star , such as used in ref .
@xcite , is not a good assumption .
it is because that the magnetic fields of a nascent neutron star are hard to affect the far outer region , such as the isotopically neutral region , within the short time scale of neutrino burst .
the magnetic fields , therefore , should be normalized by the fields which are static and existent before the core collapse . in addition , as discussed in the previous subsection , since the shock wave does not affect the resonance region at @xmath108 seconds after bounce , it is also expected that the magnetic field strength at the resonance points is not seriously changed at that time .
the strength of such magnetic fields above the surface of the iron core may be inferred from the observation of the surface of white dwarfs , since both are sustained against the gravitational collapse by the degenerate pressure of electrons .
the observations of the magnetic fields in white dwarfs show that the strength spreads in a wide range of @xmath110 gauss @xcite . taking account of the possibility of the decay of magnetic fields in white dwarfs , it is not unnatural to consider the magnetic fields up to @xmath111 gauss at the surface of the iron core . then , in eq .
( [ eq : three - flavor ] ) , @xmath112 , where @xmath2 is the strength of the magnetic field at the equator on the iron core surface , @xmath113 the radius of the iron core , and @xmath114 the angle between the pole of the magnetic dipole and the direction of neutrino propagation .
hereafter , we assume @xmath115 . from the magnetic field , density , and
@xmath20 profiles of supernova we can calculate the adiabaticity parameter of each resonance points with eq .
( [ eq : adiabaticity ] ) .
the lower panel of fig .
[ fig : profiles ] shows adiabaticity parameters of rsf conversions as a function of radius .
those for both @xmath116 and @xmath117 gauss , when @xmath118 is assumed , are shown . from the figure
, we see that the rsf - h conversion can be adiabatic when the @xmath119 , while the rsf - l conversion is always nonadiabatic under any conditions considered here
. therefore , the values of transition magnetic moment @xmath120 do not matter , while that of @xmath91 alone governs results .
from the formulation which includes three - flavor neutrinos and anti - neutrinos ( eq .
( [ eq : three - flavor ] ) ) , we give a new level crossing diagram , which is an expanded version of the ordinary diagram including msw resonance alone ( see fig . 5 in ref .
@xcite ) . we show it in fig .
[ fig : crossing ] , in which the adiabatic conversion means that the neutrinos trace the solid curve at each resonance point ( i.e. , the mass eigenstate does not flip ) , while the nonadiabatic conversion the dotted line .
the figure clearly includes not only ordinary msw resonances but also the rsf effects , and it is expected that the combination effect of msw and rsf makes this scheme very interesting to investigate . from the earlier publications on the msw effect , we know
that the msw - l is adiabatic for lma solution , and msw - h is adiabatic for large @xmath8 and nonadiabatic for small @xmath8 @xcite .
the adiabaticity of the rsf - l and rsf - h depends on the magnetic field strength and the density profile at the resonance point ( see eq .
( [ eq : adiabaticity ] ) ) , which are given in section [ sec : models ] and fig .
[ fig : profiles ] . here
, we qualitatively consider the flavor transition of neutrinos under some simple conditions . at the supernova core , electron neutrinos and anti - neutrinos
are produced as the mass eigenstate in matter , owing to large matter potential .
the other flavor ( @xmath43 ) neutrinos and anti - neutrinos at production are not mass eigenstates , however , we can make them by linear combination of these flavor eigenstates .
( these states corresponds to @xmath121 in fig .
[ fig : crossing ] . ) then , neutrinos and anti - neutrinos of all flavors propagate in supernova matter along the curves in fig . [
fig : crossing ] , being affected by four resonances , and appear from the stellar surface as pure vacuum mass eigenstates or their mixing states .
when both the rsf - l and rsf - h are nonadiabatic , the situation is the same as the pure msw case , i.e. , the conversions occurs as @xmath122 from these considerations , we can estimate the spectral deformation of @xmath46 and @xmath53 , which are the easiest to detect at sk and sno . when the msw - h is adiabatic , we obtain @xmath123 where @xmath124 is the flux of the @xmath26-type neutrinos at detection , @xmath125 the flux at production which is shown in fig .
[ fig : fluence ] , and subscript @xmath126 , since the flux of these flavors are considered to be the same . with the same manner , for nonadiabatic msw - h , we obtain @xmath127 when the rsf - h is adiabatic and rsf - l nonadiabatic , which is actually the case for strong magnetic field ( see fig . [
fig : profiles ] ) , the same arguments lead to for adiabatic msw - h : @xmath128 @xmath129 and for nonadiabatic msw - h : @xmath130 @xmath131 the adiabatic rsf - h conversion , thus , deforms the energy spectra , and the difference from the case of the pure msw effect is clear .
we calculated eq .
( [ eq : three - flavor ] ) numerically with adopted models given in section [ sec : models ] , and obtained the conversion probabilities for each flavor .
figure [ fig : prob_l ] shows those for @xmath53 ( upper panel ( a ) ) and @xmath5 ( lower panel ( b ) ) , under the condition that @xmath132 , and @xmath133 . since @xmath6 s ( @xmath134 s ) are maximally mixing with @xmath5 s ( @xmath135 s ) , we did not show @xmath136 and @xmath137 , which are almost the same as that for @xmath138 .
we can see three clear conversions , one rsf and two msws .
the rsf - h conversion occurs at @xmath139 between @xmath53 and @xmath6 .
( the apparent @xmath15 conversion occurs , since @xmath5 s are maximally mixing with @xmath6 s . ) after that , neutrinos propagate in matter inducing the ordinary msw effects , which are studied in detail by earlier authors @xcite ( msw - h at @xmath140 , msw - l at @xmath141 ) .
the strange behavior of @xmath142 at @xmath143 is considered due to the nonadiabatic rsf - l conversion , while the probability is not changed as a whole .
figure [ fig : prob_s ] is the same one as fig .
[ fig : prob_l ] , but it is for @xmath144 instead of @xmath145 . in this case , we do not see the msw - h conversion , since it is nonadiabatic .
we show in fig .
[ fig : prob_em ] , the dependence of the conversion probability @xmath142 on the energy and the magnetic field strength at the iron core , assuming @xmath146 . from fig .
[ fig : prob_em ] ( a ) , we see that the resonance point goes to outer region and the conversion becomes more and more nonadiabatic , as the neutrino energy becomes larger . in the same manner ,
[ fig : prob_em ] ( b ) indicates that the conversion becomes more adiabatic for larger magnetic field strength .
these behaviors can be understood qualitatively using fig .
[ fig : profiles ] and eq .
( [ eq : adiabaticity ] ) . at the end of this subsection
, we compare the calculated conversion probabilities with ones which are evaluated with time - dependent density profile .
figure [ fig : conv_prob ] shows @xmath147 and @xmath148 as a function of neutrino energy . since how the shock propagation changes
the structure of magnetic field at resonance points is not known well , we assume the same static dipole profile as for magnetic field , and the figure is for @xmath117 g. as discussed in section [ sub : supernova magnetic field ] , this assumption is considered to be a good one for @xmath108 seconds after bounce , however , probably incorrect for @xmath149 seconds . anyway , from the figure , we can confirm that the probabilities are in good agreement between the model 0.5 sec and the static progenitor model .
sk is a water herenkov detector with 32,000 tons of pure water based at kamioka in japan .
the relevant interactions of neutrinos with water are as follows : @xmath150 where c.c .
stand for charged- and neutral - current interactions , respectively .
sk is not in operation now because of the unfortunate accident , which occurred on 12th november 2001 .
however , it is expected to be repaired using remaining photomultiplier tubes ( pmts ) , and will restart the detection by the end of this year ( 2002 ) with lower performance .
the effect of the accident on its performance is expected not to be serious for supernova neutrinos , because the fiducial volume will not change , and the threshold energy change ( from 5 mev to about 7 - 8 mev ) influences the event number very little .
although the energy resolution will become about @xmath151 times worse , it does not matter for our consideration . in the calculations , we used the energy threshold and the energy resolution after the accident . for the reaction cross sections , we refer to ref .
the @xmath152 is the largest reaction and the energy of recoil positrons is related to that of injected neutrinos by simple formula @xmath153 , where @xmath154 are neutron and proton mass , respectively .
the energy spectrum of electrons ( positrons ) , therefore , reflects that of @xmath53 , which is easily estimated in some simple cases using e.g. , eqs .
( [ eq : f_ebar_a_na ] ) , ( [ eq : f_ebar_na_na ] ) , ( [ eq : f_ebar_a_a ] ) , and ( [ eq : f_ebar_na_a ] ) . figure [ fig : spectra_sk ] shows the energy spectra of electrons ( positrons ) for first 0.5 seconds evaluated with @xmath155 and several values of @xmath2 , which are calculated with conversion probabilities , original neutrino spectra , reaction cross sections , and detector performance .
when we increase the value of @xmath2 , the energy spectrum shifts to higher energy region and the shift saturates at @xmath156 .
this is because at @xmath157 the rsf - h conversion starts to become adiabatic and it achieves completely adiabatic transition at @xmath158 . for @xmath144
, we obtained almost the same figure as fig .
[ fig : spectra_sk ] , and we do not show it here .
sno is a herenkov detector filled with 1,000 tons of heavy water .
the interactions of neutrinos with heavy water are @xmath159 and reactions represented by eqs .
( [ eq : nue+e ] ) to ( [ eq : nuebar+o ] ) .
furthermore , the produced neutrons are detected through the interaction with surrounding nuclei as @xmath160 where @xmath161 and @xmath162 represent the @xmath163-rays whose energies are 6.3 mev and 8.6 mev , respectively , and nacl is added to efficiently capture neutrons . owing to these neutron capture reactions , the neutrino detection through the neutral - current reaction eq .
( [ eq : nux+d ] ) is possible .
furthermore from the charged - current reaction of @xmath53 s ( eq .
( [ eq : nuebar+d ] ) ) , we receive more than two signals , i.e. , one from the preceding recoil positron , and the others from the delayed @xmath163-rays by neutron captures .
thus , using this criterion , we can discern the @xmath164 events from the other detection channels . for the cross sections of reactions with deuterons
, we referred to ref .
figure [ fig : spectra_sno_l ] shows the energy spectra for first 0.5 seconds evaluated with @xmath165 and several values of @xmath2 . )
does not contribute to energy spectrum , we neglect it from here on . ]
the upper panel ( a ) shows the total energy spectra and the lower panel ( b ) the spectra of ( total@xmath166 ) events which can be obtained using the delayed coincidence signals , where the line types are the same as those used in fig .
[ fig : spectra_sk ] . as @xmath2 inceases ,
the energy spectrum of the total events becomes harder , while that of the ( total@xmath167 ) events softer .
these properties can be explained as follows .
since almost all the events come from @xmath168 and @xmath164 reactions , the ( total@xmath166 ) events are roughly coincide with @xmath168 events .
thus , from the qualitative discussion given in section [ sec : level crossing schemes and qualitative illustrations ] , we can obtain the properties in fig .
[ fig : spectra_sno_l ] ( b ) ( see eqs .
( [ eq : f_e_a_na ] ) and ( [ eq : f_e_a_a ] ) ) .
since the energy spectrum due to @xmath164 becomes harder for larger @xmath2 ( see eqs .
( [ eq : f_ebar_a_na ] ) and ( [ eq : f_ebar_a_a ] ) ) , the total spectrum also becomes harder but with the suppressed degree of the shift , because that of the @xmath168 events relaxes .
figure [ fig : spectra_sno_s ] is the same one as fig .
[ fig : spectra_sno_l ] , but for the different value of @xmath8 ( @xmath144 ) . for larger @xmath2 ,
the spectrum of the total events becomes harder , while that of ( total@xmath167 ) events changes little ( see eqs .
( [ eq : f_e_na_na ] ) , ( [ eq : f_ebar_na_na ] ) , ( [ eq : f_e_na_a ] ) , and ( [ eq : f_ebar_na_a ] ) ) .
in this study , we used the original neutrino spectrum calculated by lawrence livermore group @xcite . unfortunately
, their study as well as the other published full numerical supernova collapse simulations have not yet included the nucleon bremsstrahlung process or nucleon recoils , even though it is no longer controversial that these effects are important .
recent studies ( e.g. @xcite ) including all these processes have shown that average @xmath96 energy exceeds the average @xmath53 energy by only a small amount , 10% being a typical number .
if this is the case , the spectral deformation , particularly at sk detector , will be little and it is difficult to obtain some information on the flavor conversion mechanism in supernova .
however , it is premature to conclude that their results are correct , since it is based on the neutrino transport study on the background of an assumed neutron star atmosphere , and this approach lacks hydro - dynamical self - consistency .
further it is also because the mean energies and their ratios change significantly between the supernova bounce , accretion phase , and the later neutron star cooling phase .
thus , at present , even if the galactic supernova neutrino burst were detected by sk and sno , it would be very difficult to obtain useful information on whether the rsf effect actually occurs in supernova and to obtain implication for neutrino magnetic moment , owing to large uncertainties concerning original neutrino spectrum .
in the future , however , it might be plausible that these uncertainties will be removed by development of numerical simulations , and the signal of supernova neutrino burst would provide valuable implications for the flavor conversion mechanisms . in discussions from here ,
we present the method to probe conversion mechanisms from the detected neutrino signals , _ assuming _ that the uncertainties concerning the supernova models are considerably reduced by the future development of numerical approaches . as we have shown in section [ sec : results ] , the deformation of spectral shape occurs when @xmath157 .
therefore the ratio of high energy events to low energy ones is a good measure to investigate the magnetic field strength in supernova or neutrino magnetic moment .
we adopted as that two quantities below : @xmath169 figure [ fig : ratio_sk ] shows @xmath170 as a function of @xmath171 for @xmath165 , where error bars include only statistical errors .
( here , we re - introduce the angular factor @xmath172 since we also discuss the dependence on magnetic field orientation below . )
we obtained almost the same figure for @xmath144 .
therefore , sk data will provide statistically sufficient information on the magnetic field strength in the supernova and neutrino magnetic moment , and these consequences are independent of @xmath8 .
for example , the value of @xmath170 which is found to be larger than @xmath173 indicates that the rsf conversion occurred in the supernova , and we can constrain the value of @xmath174 assuming the supernova and presupernova models . the same figure for sno is fig .
[ fig : ratio_sno ] , where the upper panel ( a ) is for @xmath165 and the lower panel ( b ) for @xmath144 .
if we obtain @xmath175 which is larger than @xmath176 , it suggests that the supernova magnetic field is quite strong , @xmath177 , and neutrinos have non - zero magnetic moment .
however , since the statistical errors of @xmath178 are rather large because of smallness of data , we can use the sno data to confirm the sk result .
in addition , if we know the value of the angular factor @xmath172 from the optical observation such as the hubble space telescope ( hst ) , we can constrain the magnetic field strength @xmath2 ( not the combination @xmath171 ) from the neutrino observation .
actually , in the case of sn1987a , which occurred in the large magellanic cloud ( lmc ) , @xmath114 is constrained with the procedure as follows .
the hst images of sn1987a show a bright elliptical ring surrounding the optical remnant . this inner ring has been interpreted as a ring of high - density material in the equatorial plane of the progenitor , caused by the impact of the high velocity stellar wind of the blue supergiant progenitor with its earlier , slower red giant wind , and ionized by the uv flash from the supernova explosion @xcite .
the inner ring is known to be inclined to the line of sight by @xmath179 . thus , assuming that the magnetic dipole moment of the star orient to the rotation axis , we can constrain @xmath172 to be @xmath180 from the hst observation . however
, if the next supernova occurred in the galactic center , it would be quite difficult to estimate @xmath114 since the galactic center is optically thick from the earth .
on the other hand , if it occurred in the optically thin environment such as lmc , we would be able to constrain @xmath114 from optical observations and @xmath2 from the neutrino signals .
although lmc is located at @xmath181 kpc away from the earth , sk data would permit statistically sufficient discussions on the supernova magnetic field .
for example , in that case , the @xmath182 statistical errors of @xmath170 are as large as @xmath183 errors in the case of the galactic supernova ( @xmath184 ) , which are also shown in fig .
[ fig : ratio_sk ] , and it will be sufficient to distinguish whether @xmath185 or @xmath186 .
in addition to next galactic supernova , the observation of supernova relic neutrino ( srn ) background would be available to obtain information on flavor conversion mechanism in supernova .
the flux of srn is calculated by @xmath187 where @xmath188 , @xmath189 is supernova rate per comoving volume at redshift @xmath190 , @xmath191 energy spectrum of emitted neutrinos , and @xmath192 the redshift when the gravitational collapses began ( see ref .
@xcite for detailed discussion ) .
most recently , sk collaboration set an upper bound of @xmath193 for the srn flux in the energy region @xmath194 mev @xcite .
that limit is the same order as some typical theoretical predictions such as ref .
for example , ando et al .
@xcite predicted that the total srn flux integrated over entire energy was @xmath195 , while the corresponding sk limit calculated with the predicted spectral shape is @xmath196 .
since the theoretical calculations contain many ambiguities such as the supernova rate in the universe and neutrino spectrum from each supernova , this severe observational srn limit can provide a number of valuable information on the various fields of astrophysics and cosmology .
further , in the near future , it is expected that the upper limit will be much lower ( about factor 3 ) when the current sk data of 1,496 days are reanalyzed using some technique to reduce background against detection @xcite . in that case , the further severer constraint not only on the star formation history but also on the nature of neutrinos might be obtained . since the rsf mechanism further enhances the average energy of @xmath53 compared to the ordinary msw effect , the current and the future expected upper limit on the srn flux might provide very useful information on that mechanism .
however , in order to give precise srn prediction including the rsf mechanism , we must calculate conversion probability in supernova for full time - scale during the neutrino burst .
therefore , the prediction is difficult at present stage , and it is also one of our future works .
we investigated three - flavor neutrino oscillation with non - zero neutrino magnetic moment in supernova magnetic fields .
we gave a new crossing diagram which includes not only ordinary msw resonance but also magnetically - induced rsf effect . from the diagram , it was found that there occurred four resonances , two were induced by the rsf effect and two by the pure msw . in addition , the qualitative behavior of neutrino conversions and expected neutrino spectrum at the detector were found to be well illustrated .
we also numerically calculated conversion probability in the supernova to obtain the expected energy spectrum at the detector . in that calculation ,
realistic neutrino mixing parameters inferred by recent neutrino oscillation experiments and realistic density and @xmath20 profiles of presupernova stars were adopted . as for magnetic field in the supernova , we assumed the dipole structure , and that the field strength was normalized at the surface of the iron core , which was inferred from the observations of white dwarfs .
although the shock wave propagation changes density profile in the resonance region drastically and is also expected to change magnetic filed profile at the same region , we showed that if we confine our discussion until 0.5 seconds after core bounce , using static progenitor model and static magnetic field structure is expected to be a good approximation .
as a result of numerical calculations , it was found that if the magnetic field at the surface of the iron core is sufficiently strong ( @xmath197 , where we assumed @xmath198 ) , the rsf - h conversion becomes adiabatic and complete @xmath84 conversion occurs .
hence , non - zero neutrino magnetic moment with the strong magnetic field enhances the energies of @xmath53 and hardens the energy spectrum at sk and sno .
thus , in the future , if the uncertainty concerning original neutrino spectrum is removed by development of numerical simulations for supernova explosions , we might be able to obtain information on neutrino magnetic moment as well as the magnetic field strength in supernova using the energy spectrum at detectors .
in addition , it could be also plausible using the current and the future supernova relic neutrino observations .
we would like to thank superkamiokande collaboration including y. suzuki , m. nakahata , y. fukuda , and a. suzuki for useful discussion on supernova relic neutrino background .
this work was supported in part by grants - in - aid for scientific research provided by the ministry of education , science and culture of japan through research grant no .
s14102004 .
y. totsuka , rep .
prog . phys . *
55 * , 377 ( 1992 ) ; k. nakamura , t. kajita , and a. suzuki , in _ physics and astrophysics of neutrinos _ , edited by m. fukugita and a. suzuki ( springer - verlag , tokyo , 1994 ) , p. 249 .
( solid curve ) , @xmath53 ( dashed curve ) , and @xmath96 ( dotted curve ) , where @xmath96 represents @xmath47 and @xmath199 . for each line type ,
the larger one represents fully time - integrated spectrum , while smaller one spectrum integrated until 0.5 seconds after core bounce .
[ fig : fluence],width=566 ] combination which is responsible for the rsf conversions @xmath200 , and the dashed curve for the msw conversions @xmath201 .
two horizontal bands are for @xmath202 and @xmath203 .
( the band width comes from the energy range 5 - 70 mev . ) at intersections between the solid , dashed curves and the horizontal bands , the rsf and msw conversions occur ( see eq .
( [ eq : delta h ] ) ) .
lower panel : the adiabaticity parameter of rsf conversions ( eq .
( [ eq : adiabaticity ] ) ) as a function of radius , where the indicated values are magnetic field strength at the iron core ( @xmath2 ) when @xmath204 is assumed .
[ fig : profiles],width=566 ] profiles of supernova calculated by lawrence livermore group @xcite .
upper panel : the @xmath106 profile of supernova , which is responsible for rsf conversions . the profile at 0.5 , 0.7 , 2 , and 10 seconds after bounce , are shown .
the profile calculated with static progenitor model @xcite is also shown , for comparison .
the resonance points , which are represented by the intersections between the @xmath106 curves and horizontal bands ( @xmath205 ) , are affected by the shock propagation at @xmath107 seconds .
lower panel : the same figure , but for @xmath99 profile , which is responsible for msw conversions .
[ fig : shock],width=566 ] and @xmath206 represent the mass eigenstates of neutrinos and anti - neutrinos in matter , respectively , and @xmath207 and @xmath208 the mass eigenstates at production , which are superpositions of @xmath5 and @xmath6 or @xmath135 and @xmath134 .
there are four resonance points , msw - l , msw - h , rsf - l , and rsf - h .
the adiabatic conversion means that the neutrinos trace the solid curve at each resonance point ( i.e. , the mass eigenstate does not flip ) , while the nonadiabatic conversion the dotted line .
[ fig : crossing],width=566 ] under the condition that @xmath209 , and @xmath210 .
the solid curve represents @xmath211 , the dotted curve @xmath212 , the short - dashed curve @xmath213 , and the long - dashed curve @xmath214 .
( b ) the same figure as ( a ) , but for those of @xmath5 .
the solid curve represents @xmath215 , the dotted curve @xmath216 , the short - dashed curve @xmath142 , and the long - dashed curve @xmath217 .
[ fig : prob_l],width=566 ] and @xmath148 as a function of neutrino energy , which are evaluated with time - dependent density and @xmath20 profiles at 0.5 ( solid line ) , 0.6 ( short - dashed line ) , 0.7 ( dotted line ) , and 2 ( long - dashed line ) seconds after bounce .
those obtained with static progenitor model is also shown as dot - dashed line .
the magnetic field profile is assumed to be the same one as the static dipole model ( @xmath221 g ) .
[ fig : conv_prob],width=566 ] and several values of @xmath2 .
the energy spectra for @xmath222 are almost the same .
the mean energy increases as @xmath2 becomes larger , and saturates around @xmath223 ( completely adiabatic rsf - h conversion ) .
[ fig : spectra_sk],width=566 ] and several values of @xmath2 .
the line types are the same as those used in fig .
[ fig : spectra_sk ] .
( a ) total numbers of events at sno . as @xmath2 increases ,
the spectrum becomes harder .
( b ) numbers of events at sno , where the @xmath164 events ( eq .
( [ eq : nuebar+d ] ) ) are subtracted from the total events .
the spectrum moves conversely compared to the total events ( a ) , as the values of @xmath2 .
[ fig : spectra_sno_l],width=566 ] , as a function of @xmath171 for @xmath224 , where error bars include only statistical errors .
the solid error bars are at @xmath182 level and the dashed ones at @xmath225 level .
the thick and thin bands are the cubic spline interpolations of @xmath182 and @xmath183 error bars , respectively .
[ fig : ratio_sk],width=566 ] , where the upper panel ( a ) is for @xmath224 and the lower panel ( b ) for @xmath226 .
the error bars include only statistical errors , and are at @xmath182 level .
the thick and thin bands are the cubic spline interpolations of these error bars .
[ fig : ratio_sno],width=566 ] | we investigate resonant spin - flavor ( rsf ) conversions of supernova neutrinos which are induced by the interaction of neutrino magnetic moment and supernova magnetic fields . from the formulation which includes all three - flavor neutrinos and anti - neutrinos , we give a new crossing diagram that includes not only ordinary msw resonance but
also magnetically - induced rsf effect . with the diagram , it is found that four conversions occur in supernova , two are induced by the rsf effect and two by the pure msw .
we also numerically calculate neutrino conversions in supernova matter , using neutrino mixing parameters inferred from recent experimental results and a realistic supernova progenitor model .
the results indicate that until 0.5 seconds after core bounce , the rsf - induced @xmath0 transition occurs efficiently ( adiabatic resonance ) , when @xmath1 , where @xmath2 is the strength of the magnetic field at the surface of iron core .
we also evaluate the energy spectrum as a function of @xmath3 at the superkamiokande detector and the sudbury neutrino observatory using the calculated conversion probabilities , and find that the spectral deformation might have possibility to provide useful information on neutrino magnetic moment as well as magnetic field strength in supernovae . |
the sources of ultra - high energy cosmic rays have remained elusive in spite of the enormous progress reached on the experimental side , with present day detectors reaching apertures @xmath6km@xmath7 sr yr .
the differential energy spectrum , the chemical composition and the distribution of arrival directions on the sky are as many clues to the nature of the source .
there is now a consensus on the existence of the gzk cut - off ( greisen 1966 ; zatsepin & kuzmin 1966 ) , which has been observed by two different experiments ( abbasi et al .
2008 , abraham et al .
the existence of this gzk cut - off puts on solid ground the models which attribute the origin of ultra - high energy cosmic rays to powerful astrophysical objects distributed on cosmological scales , such as powerful radio - galaxies ( rachen & biermann 1993 ) , gamma - ray bursts ( milgrom & usov 1995 , vietri 1995 , waxman 1995 ) , or magnetars ( arons 2003 ) .
experimental results on the chemical composition and anisotropies are however far more confusing . while the most recent analysis of hires data points to a proton composition above the ankle ( abbasi et al . 2005 , belz 2009 ) , the fluorescence data of the pierre auger observatory ( pao ) rather indicates an increasingly heavier composition above this energy , with the last data points at @xmath8eev close to expectations for iron ( unger et al .
2007 , wahlberg et al . 2009 ) . regarding the distribution of arrival directions on the sky ,
there exist various contradictory claims , see for instance the correlation with the super - galactic plane reported by stanev et al .
( 1995 ) , contradicted by the results of the agasa experiment ( takeda et al .
1999 ) ; the claim for a correlation with bl lac objects in tinyakov & tkachev ( 2001 , 2002 ) , gorbunov et al . ( 2002 , 2004 ) , which has been debated ( evans , ferrer & sarkar 2003 , 2004 ; tinyakov & tkachev 2004 ) ; as well as the possible detections of multiplets in various datasets , e.g. takeda et al .
( 1999 ) , uchihori et al .
( 2000 ) , farrar , berlind & hogg ( 2006 ) , the significance of which is questioned in abbasi et al .
( 2004 ) and finley & westerhoff ( 2004 ) ; or , finally , the recent data of the pao , which reveal a correlation with nearby active galactic nuclei ( agn ) , the maximal signal being obtained for a search radius of @xmath9 around agn located closer than 75 mpc and for energies above @xmath10eev ( abraham et al .
2007 , 2008b ) .
in general , these issues of anisotropies and chemical composition are discussed separately . however , as we argue in the present paper , one can construct a powerful test of the chemical composition by using the anisotropy signal at various energies .
this becomes particularly interesting when one notes that measurements of chemical composition are relatively sensitive to the details of the shower reconstruction , in particular to the extrapolation of hadronic models at energies beyond those currently tested in accelerators ( see wibig 2008 , ulrich et al .
2009a , b and wibig 2009 for recent discussions of this issue ) .
moreover , measurements of chemical composition rely on the use of fluorescence detectors whose duty cycle is rather low and can not be made event by event , so that they are limited by statistics .
the essence of the test proposed in this paper is to exploit the fact that a source in the sky emitting heavy nuclei of charge @xmath0 at an energy @xmath11 is expected to produce a similar anisotropy pattern at energies @xmath12 via the proton component which is expected to be associated with the same source .
our expectation for the existence of a proton component relies on the theoretical expectation , that if protons are present in the plasma in which particle acceleration takes place , then they should be accelerated along side with the heavy nuclei .
as discussed in some detail at the end of [ sec : test - chem ] , a source accelerating heavy nuclei to energy @xmath11 is quite generally expected to accelerate protons to energy @xmath12 .
although the isotropic background `` noise '' increases at lower energies , we show in [ sec : test - chem ] that the signal - to - noise ratio for the detection of the anisotropy also increases .
this prediction does not depend at all on the modelling of astrophysical magnetic fields as it only relies on the property that protons of energy @xmath12 follow the same path in the intervening magnetic fields and produce the same angular image as heavy nuclei of charge @xmath0 and energy @xmath11 .
the proposed test is discussed in more detail in section [ sec : test - chem ] . when applied to the most recent data of the pao ( section [ sec : pao ] )
, it shows that the signal that is responsible for the apparent anisotropy pattern at energies @xmath13eev must not be heavy but light nuclei , provided this anisotropy is confirmed by future data , and provided the pao does not see evidence for anisotropy at lower energies , as seems to be the case ( abraham et al .
2007 ) .
we discuss the implications of these results in the last two sections . in section
[ sec : agn ] , we argue that local agn ( including fr i radio - galaxies ) do not possess the power required to accelerate protons to ultra - high energies .
when combined with the previous conclusions , this leads to the conclusion that the current pao data do not support agn as sources of the highest energy cosmic rays .
finally , section [ sec : disc ] summarizes the findings and concludes that the current data on composition and anisotropy suggest one of the following : ( i ) the shower modelling or , what would be more interesting , the hadronic theoretical models of shower reconstruction are in error at high energy ; ( ii ) the composition switches abruptly from heavy to light above @xmath14eev ; ( iii ) the source injects primarily heavy nuclei ( which seems unlikely ) ; ( iv ) the anisotropy seen by the pao is a statistical artefact .
as argued below , one can use the results of searches for anisotropy at various energies in order to constrain the chemical composition of ultra - high energy cosmic rays .
the basic claim is the following : if one detects anisotropy above some energy @xmath1 , but not below , then the chemical composition is most likely light above @xmath15 , because one would have otherwise observed a similar anisotropy pattern at an energy @xmath16 , with @xmath0 the assumed average charge of the cosmic rays at energies @xmath17 .
consider a region of the sky , of angular size @xmath18 and solid angle @xmath19 , showing an excess of particles above isotropic expectations above some energy threshold @xmath15 .
the most natural interpretation is to infer the existence of one or more sources in this direction , the image of which is spread over @xmath20 , because of smearing by @xmath18 in the intervening magnetic fields , or because @xmath20 subtends the source distribution . for the sake of the argument
, we assume that the spectrum produced by the source consists of protons and of heavy nuclei of charge @xmath0 . in
all known source models , one expects that the ratio of the elemental spectra @xmath21 and more generally @xmath22 at a given energy @xmath23 , where @xmath24 denotes the maximal energy at the source for protons .
for instance , the composition ratio of protons to iron peak elements ( @xmath25 ) in the galactic cosmic ray spectrum , which is roughly consistent with the solar abundance when compared at a given energy per nucleon , implies a source composition at a given energy @xmath26 ( as taken from the recent atic-2 data , panov et al .
the only energy regime in which one may obtain @xmath27 is @xmath28 , with @xmath29 the maximal energy for the nuclei of charge @xmath0 ( see also further below ) .
we also assume here that the maximal energy scales as @xmath0 ; as argued further below , this is a conservative choice .
now , let us ask what would be seen if the composition above @xmath1 were heavy . consider the quantity @xmath30 which characterizes the signal - to - noise ratio of the anisotropy signal : @xmath31 represents the excess number of events over isotropic expectations , i.e. the difference between the total number of events observed above @xmath1 within @xmath20 and @xmath32 , the number of events expected from the isotropic background . in the above model , @xmath33 corresponds on average to the number of events produced by the sources inside @xmath20 .
a detection of anisotropy thus corresponds to @xmath34 .
one can now compute a similar quantity , but corresponding to the proton component at energies @xmath2 , denoted @xmath35 .
these protons have a same magnetic rigidity than the heavy component above @xmath1 , therefore they follow the same path in the galactic and extra - galactic magnetic fields , and consequently they are to produce a similar anisotropy pattern in the sky , up to the composition ratio of the elemental spectra and up to the level of background noise associated with the isotropic component .
assuming that the elemental spectra at injection have the same power - law form @xmath36 at @xmath37 , and that the total number of events observed scales as @xmath38 , with @xmath39 the slope of the observed all - sky differential spectrum , one finds @xmath40 the fudge factor @xmath41 incorporates the effect of energy losses ; it is discussed in more detail in the following . at energies above the ankle , one has @xmath42 ( abbasi et al .
2008 , abraham et al .
therefore , if the source emits a power - law with @xmath43 , one finds that the signal - to - noise ratio for detecting the anisotropy is larger at @xmath16 than at @xmath1 by a factor @xmath44 , generally expected to be significantly larger than unity !
clearly , the softer the source spectrum , the larger the gain in signal - to - noise ratio with decreasing @xmath0 .
the factor @xmath45 accounts for the difference between the propagated and injected spectra .
given that the energy loss distance for protons of energy @xmath16 is much larger than the energy loss distance for heavy nuclei of energy @xmath1 , if @xmath46ev , one expects @xmath47 .
indeed , regarding the ( primary ) proton component , energy losses can be safely neglected at @xmath16 energies , since the attenuation length is of order of a gpc or more at eev energies , which is much larger than the depth up to which which anisotropies can be produced , this latter being of order of the inhomogeneity scale of the large scale structure @xmath48mpc or of the distance to the closest source , whichever is larger .
therefore , @xmath45 approximately corresponds to the reciprocal of the attenuation factor for heavy nuclei of charge @xmath0 and energy @xmath17 propagating over the distance scale @xmath49 to the sources inside @xmath20 .
the attenuation factor is understood as the ratio of the flux above @xmath17 after propagation on a distance @xmath49 to the injected flux above this energy .
for the case of iron primaries , this attenuation factor is @xmath50 above @xmath51eev at a distance of 10 mpc , becoming @xmath52 at a distance of 100 mpc ( counting in the propagated flux all secondary nuclei with @xmath53 ) .
for oxygen , the attenuation factor is @xmath54 at 10 mpc , becoming @xmath55 at 100 mpc ( counting in all c , n and o secondary nuclei ) ; see bertone et al .
( 2002 ) for an illustration of the relative fragility of ultra - high energy cno nuclei vs iron peak elements .
actually , this enhancement factor @xmath45 should be further increased by the number of secondary protons with energy @xmath2 produced by the photo - disintegration of the primary heavy nuclei with energy @xmath17 .
this number is not easy to quantify but it can be significant . for instance , recent simulations by aloisio , berezinsky & gazizov ( 2008 ) indicate that the signal should be dominated by protons at energies @xmath56ev even if the source composition were strongly enriched in iron nuclei [ assuming of course that @xmath57ev ] .
allard et al .
( 2008 ) also conclude that heavy nuclei could be found in abundance at the detector only if the composition were essentially dominated by fe group nuclei at energies beyond @xmath58ev . as discussed in this latter reference , and above , the generic case that gives rise to a heavy composition at ultra - high energies
is when the proton maximal energy at the source is smaller than the gzk cut - off . in the following , we adopt a conservative point of view and neglect the effect of energy losses , i.e. we set @xmath59 .
let us now discuss the scaling of @xmath60 with charge .
the maximal energy is given by the comparison of the acceleration timescale ( which depends on rigidity @xmath12 ) with the minimum of the escape timescale ( which also depends on @xmath12 ) , the age of the source or dynamical timescale ( which do not depend on @xmath11 , @xmath0 ) and the energy losses timescale ( which depends on @xmath11 , @xmath0 and mass number @xmath61 ) .
if the latter timescale does not provide the dominant limitation to the acceleration mechanism , then the maximal energy scales as the rigidity , as assumed above .
if energy losses are dominant however , one expects @xmath62 , as explained in the following .
then the above argument would remain valid , as it only requires that protons exist at energies @xmath2 , i.e. that @xmath63 .
note that in the more extreme case in which protons and heavy nuclei had similar maximal energies , the composition at ultra - high energies @xmath17 would be dominated by protons , and therefore the anisotropy would be due to those protons since they have a higher rigidity than the heavy nuclei of a same energy .
for photo - hadronic interactions , the energy loss timescales for protons and iron - peak nuclei are known to be comparable , while that of intermediate elements is generally smaller ; in that case , one would expect @xmath64 .
regarding synchrotron losses , the energy loss timescale scales as @xmath65 , therefore @xmath66 if the acceleration timescale @xmath67 and synchrotron losses dominate at the highest energies .
in that case , the scaling @xmath68 remains correct to within a factor of order unity .
the above conclusions agree with the recent simulations of allard & protheroe ( 2009 ) . finally , accounting for a more complex chemical composition with more than two species would not modify significantly the results shown in eq .
( [ eq : sigma ] ) as long as the anisotropy above @xmath1 is indeed due to the species of charge @xmath0 . in practice
, it will be useful to test the composition above @xmath1 by checking the anisotropies at various @xmath69 , with @xmath70 representative of the most abundant elements .
in the following , we provide specific examples of the above test applied to the most recent data of the pao .
we will show in particular that the above argument remains valid when one considers the all - sky anisotropy pattern due to an ensemble of close - by sources .
abraham et al . ( 2007 , 2008b ) have reported a correlation of 20/27 arrival directions of the highest energy events ( @xmath71ev ) with nearby ( @xmath72mpc ) active galactic nuclei
. however , the typical model of ultra - high energy cosmic ray origin in agn refers to strongly beamed fanaroff - riley ii
( fr ii ) sources with giant radio lobes ( e.g. rachen & biermann 1993 ) , while 19 out of these 20 correlating agn in the pierre auger dataset belong to the seyfert or liner class , only one being a fanaroff - riley i ( fr i ) radio - galaxy ( cen a ) .
these correlating agns are thus more likely to be tracers of the local large scale structure in which the actual sources of ultra - high energy cosmic rays camouflage and indeed , the arrival directions are compatible with a distribution of sources following the large scale structure ( kashti & waxman 2008 , ghisellini et al .
2008 , zaw , farrar & greene 2009 ) .
the extended dataset up to 2009 of the pao reveals a significantly weaker correlation , with only 26 out of 58 events above @xmath51eev lying within @xmath9 of an agn located within 75mpc ( hague et al .
this level of correlation is compatible with that expected for sources distributed as the large scale structure ( see discussion in kotera & lemoine 2008 and aublin et al .
2009 for a detailed discussion of the correlation with cosmological catalogs ) .
it is important to note that the departure from isotropy lies at the @xmath73 level . therefore one should be careful not to overinterpret this excess . in the following
, we use these pao results as a concrete example for the test developed in the previous section .
interestingly , the pao data up to 2007 has also shown evidence for clustering of events in the region around cen a , with as many as @xmath74 events within @xmath75 ( gorbunov et al .
2008 ) , as well as correlation of @xmath76 out of 27 events with a sample of radio - galaxies ( moskalenko et al . 2008 ; nagar & matulich 2008 ) .
this has triggered a surge of interest in models of ultra - high energy cosmic ray origin in cen a as well as forecast studies of neutrino and gamma - ray expected signals from cen a ( e.g. cuoco & hannestad 2008 , gupta 2008 , halzen & omurchadha 2008 , holder et al .
2008 , kachelriess et al .
however , such a correlation does not necessarily imply that cen a is the source of these cosmic rays .
first of all , and as noted in abraham et al .
( 2008b ) , hague et al .
( 2009 ) , the evidence for clustering is based on an a posteriori analysis , so that it is difficult to quantify the level of significance .
even if future data confirm this clustering , one must keep in mind that cen a lies in front of one of the largest concentrations of matter in the local universe , at @xmath77mpc from us , so that the correlation could be accidental . finally , as we argue in section [ sec : cena ] , the cen a agn / jet / lobe system is too weak to accelerate protons up to the observed energies @xmath78eev .
even more intriguing is the fact that the chemical composition measured by the pao becomes increasingly heavier above the ankle at @xmath79eev , with the last data points at @xmath14eev lying close to expectations for iron ( unger et al .
2007 , wahlberg et al . 2009 ) .
such a measurement is in obvious conflict with the most recent analysis of hires data , which indicate a pure proton composition up to @xmath80eev ( belz 2009 ) .
as we argue in the following , it is also at odds with the observed anisotropy pattern above @xmath51eev , unless the composition abruptly changes from heavy to light above @xmath14eev .
let us first discuss the anisotropy toward cen a , using the most recent pao dataset , in particular , @xmath81 events located within @xmath82 out of a total of @xmath83 events above @xmath51eev ( hague et al .
2009 ) , assuming that the clustering is not a statistical accident . in a first approximation , one can model the anisotropy pattern as in the previous section , with one ( or several clustered ) source(s ) contributing a fraction @xmath84 of the total flux , the remaining @xmath85 being accounted for by an isotropic background . we do not need to assume that cen a itself is a source , but we center the anisotropic signal on the location of cen a in order to match the observed pattern .
given that @xmath86 events are expected on average from this region of the sky if the sources are distributed isotropically , we set @xmath87 in the following .
we assume that the sources inject protons and iron nuclei with powerlaw spectra @xmath88\right\}$ ] , producing an angular image of size @xmath89 .
as before , this angle could represent smearing due to angular deflection , in which case @xmath90 , or the angular size of the ensemble of sources .
although we assume that all heavy nuclei are iron , the following could be easily generalized to any kind of mixed composition without changing the basic result .
we also neglect the energy losses for the spectrum of the anisotropic component , as before .
this is motivated by the proximity of matter in this direction of the sky and , as discussed before , incorporating energy losses would not diminish the signal we are to extract out of the data , while it would introduce other free parameters .
the remaining @xmath85 fraction of the flux is modelled with an isotropic background , e.g. as produced by far away sources .
let us first assume that the composition is dominated by iron at the highest energies .
we set @xmath91eev , and @xmath92 above @xmath1 .
we also set @xmath93eev , so that essentially no proton contribute to the signal above @xmath1 .
in the top panel of fig . [
fig : ptsa ] we plot the histogram of the number of events as a function of the angular distance to the source ( solid line ) .
the colored region indicates the expectations for a purely isotropic signal .
the departure from isotropy is clear , although the uncertainty on the signal in each bin is substantial given the small number of events involved . here
the number of events within @xmath82 of cen a is @xmath94 for the model ( slightly lower than the observed 12 events , yet within the uncertainty ) and @xmath86 for the isotropic expectations so that the inferred @xmath95 . in the middle panel of fig .
[ fig : ptsa ] , we plot the same histogram , but at an energy @xmath96eev . in order to simulate the anisotropy signal ,
we have proceeded as follows .
we have assumed that the total number of events increases ( with decreasing energy ) according to the observed data ( in particular the pao spectrum published in yamamoto et al .
2007 ) , and we have calculated the contribution to this flux above @xmath97 for all the protons injected by the source above this energy . we have assumed that the ratio of protons to iron is 1:0.06 at a given energy well below @xmath60 , corresponding to the composition ratio of proton to iron peak elements ( @xmath25 ) in the galactic cosmic ray spectrum reconstructed at the source . the iron nuclei injected by the source at energies @xmath98eev add up to the isotropic background and do not contribute to the anisotropy signal .
as made obvious in fig .
[ fig : ptsa ] , the gain in signal - to - noise ratio is very large : the number of events observed above @xmath99eev within @xmath82 is now 5150 , for 2330 expected , so that @xmath100 .
this gain by a factor @xmath101 can be recovered from eq .
( [ eq : sigmap ] ) . finally , the lower panel of fig . [ fig : ptsa ] shows the ratio of the differential histograms @xmath102 for the source proton contribution plus isotropic background ( filled circles ) to the isotropic background @xmath103 normalized to a same total number of events , including the error bars evaluated through the monte carlo simulation .
the empty circles show the same quantity , but for a model in which the source injects only protons , even up to the highest energies , assuming that @xmath89 represents magnetic deflection , i.e. @xmath90 : we have set @xmath92 above @xmath51eev in this model , so that the anisotropy signal above @xmath51eev is similar to that of the previous iron+proton model shown in the top panel of fig .
[ fig : ptsa ] , but the low energy signal is essentially isotropic , see the bottom panel .
the conclusions to be drawn are clear .
if the clustering seen toward cen a is real , the events responsible for this anisotropy pattern toward cen a above @xmath51eev are unlikely to be heavy nuclei , otherwise pao would have observed a much stronger anisotropy at energies @xmath104eev ( abraham et al .
2007 , 2008b ) . in the present section ,
we investigate how the anisotropy signal with respect to nearby agn , as reported in hague et al .
( 2009 ) can be used to discriminate the chemical composition at the highest energies .
we proceed in a similar way as before , but we now assume that the spectrum is composed of two components : one that follows the local large scale structure , contributing a fraction @xmath84 to the flux above @xmath51eev , and one isotropic component that makes up for the remainder , @xmath85 . as a proxy for the local large scale structure , we use the pscz catalog of galaxies ( saunders et al .
1991 ) , as in previous studies ( kashti & waxman 2008 , kotera & lemoine 2008 ) . out of simplicity , we assume that the sources follow the pscz distribution without bias . however , we allow for the possibility of a deflection angle @xmath89 in intervening magnetic fields and we explore two possible values @xmath105 and @xmath92 above @xmath51eev .
this choice is motivated by the resolution of the pscz catalog , approximately @xmath106 , and by the possibility that this signal is composed of heavy nuclei . in practice
, we proceed through monte carlo simulations as follows .
we construct many mock catalogs of angular sky distributions , each mock catalog being a sample of @xmath83 events above @xmath51eev ( or the corresponding number at any other energy ) .
the arrival direction of each event in the sample is drawn at random , according to a probability law reproducing the pscz distribution and pao coverage if it belongs to the @xmath84 anisotropic fraction , or reproducing isotropic arrival directions and pao coverage if the event belongs to the remaining @xmath85 fraction of isotropic background . in order to draw an event from the pscz catalog
, we include the appropriate weight factors for the flux dilution with inverse distance squared , for the pscz exposure function and for the pao aperture .
then we randomly displace this arrival direction by @xmath89 in the plane of the sky .
the energy losses are accounted for by the depth of the catalog , which is limited to @xmath107mpc .
we have set @xmath108 , which allows to produce an anisotropy signal close to the level observed : the pao indeed records 24 events out of 58 events above @xmath51eev located within @xmath109 of an agn closer than @xmath110mpc , while the above model produces @xmath111 events within @xmath109 if @xmath105 and @xmath112 events if @xmath92 .
the model with 90% of the flux above @xmath51eev coming from sources distributed as the large scale structure with negligible deflection thus gives a number of correlating agn within @xmath109 that is comparable to that seen by pao , to within the uncertainty .
a deflection angle @xmath92 provides a signal that is marginally too low , but given the simplifying assumptions made above , this is not crucial .
it certainly indicates that larger deflection angles will probably result in too small anisotropies .
it is interesting to note that the anisotropy signal peaks at a separation which is substantially smaller than @xmath89 , i.e. @xmath113 for @xmath92 .
this of course results from the pao procedure , which selects the closest agn to the arrival direction and not to the source direction , i.e. after the direction at the source has been modified by @xmath89 .
the differential histograms corresponding to these models and isotropic expectations are shown in fig .
[ fig : lssa ] as a function of the angular separation to the closest agn drawn from the vron - cetty & vron ( 2006 ) catalog , with an arbitrary cut at @xmath110mpc , following the pao analysis .
these histograms are shown for the two possible values of @xmath89 , as indicated . here , we find @xmath114 and @xmath115 for @xmath105 and @xmath92 respectively .
the signal is thus marginal in terms of departure from isotropy .
we now plot the ratios of the differential histograms expected in these two models at low energy @xmath116ev to the histogram expected for purely isotropic arrival directions , as before . we have assumed a conservative composition ratio 1:0.06 between proton and iron as in the previous section , and we have neglected energy losses on the propagated spectra .
the results are shown in fig .
[ fig : lssb ] , which clearly reveals a stronger anisotropy signal at these energies . in terms of signal - to - noise
, we find @xmath117 and @xmath118 respectively for @xmath105 and @xmath92 ( @xmath119 and @xmath120 respectively , @xmath121 ) , which significantly exceed the signal - to - noise ratios obtained above @xmath51eev .
this generalizes the argument presented in section [ sec : test - chem ] for the case of a more complex distribution of point sources on the sky , superimposed on an isotropic background .
it also shows that the present angular data of the pao disfavors a heavy composition at ultra - high energies . strictly speaking
, this test demonstrates that the _ anisotropic _ part of the signal can not be due to heavy nuclei _ but _ one can not exclude that the isotropic background is made up of heavy nuclei .
however , our above model also suggests that a significant fraction ( 90% in fig .
[ fig : lssa],[fig : lssb ] ) of the flux at @xmath13eev is produced by the anisotropic component if one takes at face value the anisotropy results of the pao .
this suggests that the bulk of ultra - high energy cosmic rays above @xmath51eev are protons .
on general grounds , one can construct a lower bound to the magnetic luminosity that a source must exhibit in order to be able to accelerate particles up to @xmath58ev ( norman et al .
1995 , waxman 1995 , lyutikov & ouyed 2005 , waxman 2005 ) .
let us first briefly describe the general , model independent argument given in waxman ( 1995 ) . consider an astrophysical source driving a flow of magnetized plasma , with characteristic magnetic field strength @xmath122 and velocity @xmath123 .
imagine now a conducting wire encircling the source at radius @xmath124 .
the potential generated by the moving plasma is given by the time derivative of the magnetic flux , and is therefore given by @xmath125 where @xmath126 .
a proton which is allowed to be accelerated by this potential drop would reach energy @xmath127 .
the situation is somewhat more complicated in the case of a relativistic outflow , where @xmath128 . in this case
, the proton is allowed to be accelerated only over a fraction of the radius @xmath124 , comparable to @xmath129 . to see this , one must realize that as the plasma expands , its magnetic field decreases , so the time available for acceleration corresponds , say , to the time of expansion from @xmath124 to @xmath130 . in the observer frame
this time is @xmath131 , while in the plasma rest frame it is @xmath132 .
thus , a proton moving with the magnetized plasma can be accelerated over a transverse distance @xmath133 .
this sets a lower limit to the product of the magnetic field and source size , which is required to allow acceleration to @xmath134 , @xmath135 .
this also sets a lower limit to the rate @xmath136 at which energy is carried by the out flowing plasma , and which must be provided by the source ( waxman 1995 , 2005 ) , @xmath137 where we have generalized the last equation to particles of charge @xmath0 .
let us consider next the application of the general argument given above to the commonly assumed scenario , where particle acceleration within the outflowing plasma is achieved via diffusive acceleration . in this case
, one writes the acceleration timescale @xmath138 as a multiple @xmath139 of the larmor time , and one compares it to the dynamical timescale available for acceleration .
let us discuss this bound in some detail .
for the sake of definiteness , we assume an outflow of lorentz factor @xmath140 and opening angle @xmath141 ( the isotropic case is recovered in the limit @xmath142 ) . the quantity @xmath139 depends on the diffusion coefficient of the particle , hence on its energy and the general properties of the magnetic field of the accelerating region . in acceleration mechanisms which involve scattering against magnetic inhomogeneities ( such as fermi i or fermi ii ) , @xmath143 , where @xmath144 is the scattering timescale and @xmath145 the diffusion coefficient .
the parallel diffusion coefficient can be written as @xmath146 ( see for instance casse , lemoine & pelletier 2002 ) , with @xmath147 the fraction of turbulent magnetic energy density ( in units of total magnetic energy density ) and @xmath148 the coherence length .
the exponent @xmath149 depends on the spectral index @xmath150 of the ( one dimensional ) turbulence power spectrum : @xmath151 for @xmath152 , @xmath153 for @xmath154 . from this , one easily verifies that @xmath155 is general , meaning that the so - called bohm regime @xmath156 is a limiting regime . in non - relativistic fermi shock acceleration ( @xmath157 or @xmath158 ) , @xmath159 , with @xmath160
the shock velocity in the comoving frame and @xmath145 refers to quantities measured in the upstream frame .
hence @xmath161 .
one recovers a similar result for non - relativistic fermi ii acceleration , with @xmath162 , @xmath163 denoting the alfvn speed . if acceleration takes place in a ( non - relativistic ) shear flow with velocity gradient @xmath164 , the timescale @xmath165 ( rieger , bosch - ramon & duffy 2007 ) .
the deconfinement limit corresponds to @xmath166 , in which case the limiting acceleration timescale becomes comparable to @xmath167 , as above .
note that the acceleration timescale is larger ( hence acceleration is less efficient ) when @xmath168 . in the ultra - relativistic shock limit @xmath169 ,
one expects @xmath170 ( up to a factor of order unity , see pelletier , lemoine & marcowith 2009 ) , with the condition @xmath171 .
since @xmath139 increases with @xmath172 , this process is in general not efficient .
finally , in the mildly relativistic regime @xmath173 , one expects @xmath174 , hence @xmath175 , whose minimum value is of order one provided the bohm regime diffusion applies and the magnetic field is nearly fully turbulent . to summarize , @xmath176 in general , and @xmath177 corresponds to a maximally efficient acceleration process .
the constraint @xmath178 , with @xmath179 the dynamical timescale at radius @xmath124 ( in the lab frame ) , now gives : @xmath180 with @xmath181 the observed energy in units of @xmath58ev .
the presence of @xmath182 does not mean that in the limit @xmath183 , this lower bound vanishes , since @xmath184 more than compensates for this term .
similarly , the above bound does not vanish in the limit @xmath185 , since lateral escape losses become prominent when @xmath186 .
the corresponding timescale @xmath187 , hence @xmath188 . in the case of agn jets ,
one finds in general @xmath189 , hence both limits are comparable . according to the above discussion , the most optimistic values for the parameters entering these equations are @xmath190 , @xmath191 and @xmath192 .
therefore @xmath193erg / s can be considered as a firm lower bound on the source magnetic luminosity . since most of the agn seen to correlate within @xmath109 of the pao events are seyfert galaxies , with bolometric luminosities well below @xmath194ergs
/ s , the agn that they harbor can not be the source of light particles at @xmath58ev .
we also note that zaw , farrar & greene ( 2009 ) have shown that one third of this sample is actually made of star forming galaxies with little or no agn activity .
hence the correlation with stricto - sensu agn objects is actually weaker than reported . actually even fr i radio - galaxies , tev blazars and bl lac objects do not possess the required power to accelerate @xmath195 particles up to @xmath58ev , since their magnetic luminosities are of order @xmath196ergs / s ( celotti & ghisellini 2008 ) .
such objects might possibly accelerate heavy nuclei up to energies around the gzk cut - off but , as discussed above , the anisotropy pattern at high and especially low energies would then appear inconsistent with the pao results .
further anisotropy studies in both energy ranges would certainly further constrain their possible contribution .
the compilation of celotti & ghisellini ( 2008 ) has been done in the frame of leptonic synchrotron and inverse compton emission . in hadronic blazar models ,
the magnetic field in the blazar zone is typically one order of magnitude larger than in leptonic models , so that these objects might fulfill the magnetic luminosity bound to accelerate protons up to @xmath58ev .
however , in this case , acceleration would occur in the blazar zone hence the emission should be beamed forward . in order to escape further expansion losses in the magnetized jets ,
the accelerated protons would have to be converted into neutrons , which would decay back to protons on a distance scale @xmath197mpc , i.e. outside the jet .
one should therefore observe a correlation of the arrival directions with blazars , not with radio - galaxies seen offside ( rachen 2008 ) .
however the pierre auger observatory reports no correlation with blazars ( harari et al .
2007 ) , so that this possibility is equally at odds with present observational results .
keeping in line with the modelling of celotti & ghisellini ( 2008 ) , only flat spectrum radio quasars ( the likely equivalent to the most powerful fr ii sources ) seem capable of producing jets with @xmath198ergs / s .
but , considering the highest energy pao event ( @xmath199ev ) , one finds that the smallest angular separation of this event to the fr ii sources located within 130 mpc , as compiled by massaglia ( 2007 ) is already @xmath200 ( for ngc 4261 ) .
the next objects are 3c 296 ( @xmath201 away ) and pks1343 - 60 ( @xmath202 away ) .
the closest blazar ( classified as bl lac in the vron & vron - cetty catalog ) located closer than @xmath203mpc lies @xmath204 away from this highest energy event ( tex 0554 + 534 ) .
recently , it has been proposed that transients in active galactic nuclei could power up the engine up to the luminosities required , thereby evading the above constraints ( see for instance farrar & gruzinov 2009 , dermer et al .
however , as demonstrated by waxman & loeb ( 2009 ) , such transients should produce counterpart flares in x - rays through the concomittant acceleration of electrons , which should have been picked up by existing surveys .
their non - detection strongly argues against such flaring scenarios .
given its proximity and the apparent clustering of pao events in this direction of the sky , cen a has received a lot of attention . in particular , several authors have suggested that this agn could host a site of acceleration to ultra - high energies ( cavallo 1978 ; romero et al .
1996 ; farrar & piran 2000 ; gorbunov et al . 2008 ; hardcastle et al .
2009 , osullivan , reville & taylor 2009 ) .
let us therefore discuss it in some more detail .
centaurus a is classified as a fr i / misaligned blazar ( see israel 1998 for a review ) .
its bolometric luminosity @xmath205erg / s and its jet kinetic power is inferred to be @xmath206erg / s .
therefore , this source should not be able to accelerate light particles to @xmath58ev .
the detailed modelling of the spectral energy distribution of its nucleus by chiaberge , capetti & celotti ( 2001 ) provides further information on the emission zone : @xmath207 cm , @xmath208 g and @xmath209 .
this corresponds to a magnetic luminosity @xmath210ergs / s , a value that seems quite reasonable in view of the jet kinetic power and of the bolometric luminosity , and which agrees relatively well with estimates of the magnetic field in the inner jet at larger distances from the core , when assuming @xmath211constant .
the papers which have discussed the issue of acceleration to ultra - high energy in some detail are romero et al .
( 1996 ) , which has argued that acceleration can take place in the x - ray knots of the inner jet through diffuse shock acceleration , hardcastle et al .
( 2009 ) and osullivan , reville & taylor ( 2009 ) , which have discussed the stochastic acceleration of protons in the giant lobes of cen a and gorbunov et al .
( 2008 ) relying on the model of neronov et al .
( 2007 ) of particle acceleration in a black hole magnetosphere .
romero et al . ( 1996 ) uses the inferred values of @xmath212 g ( upstream field ) , @xmath213kpc , @xmath214 and @xmath215 .
however , since the larmor radius @xmath216 ( with @xmath217ev , @xmath218 g ) , even the hillas criterion ( hillas 1984 ) is not satisfied . balancing the acceleration timescale with the escape timescale @xmath131 , one derives a maximal energy @xmath219ev ( @xmath220 , @xmath221 ) .
as discussed above , this maximal energy lies in the correct ballpark for accelerating nuclei to ultra - high energies .
however the anisotropy signal would then appear incompatible with the pao data ; in particular , a strong anisotropy signal at eev energies should have been detected .
hardcastle et al .
( 2009 ) have argued that protons could be accelerated up to @xmath222ev through stochastic acceleration in the magnetized turbulence of the giant lobes of cen a. however , one can show that their estimate of the acceleration timescale is far too optimistic because , when applied to the electrons , it produces a maximal energy well in excess of the inferred @xmath223ev . in detail , using the estimate of hardcastle et al .
( 2009 ) for the acceleration timescale , @xmath224 , and their bohm scaling @xmath225 , one infers the maximal energy for electrons by comparing @xmath226 to the synchrotron and inverse compton energy loss timescales , leading to @xmath227ev .
the above maximal electron energy is five orders of magnitude larger than the inferred maximal energy , hence the above acceleration timescale is far too optimistic and , consequently , the maximal proton energy must be much lower than @xmath58ev .
this conclusion agrees with those of osullivan , reville & taylor ( 2009 ) ; see also casse , lemoine & pelletier ( 2002 ) for a similar discussion regarding the acceleration in the lobes of powerful radio - galaxies .
finally , neronov et al . ( 2007 ) have argued that in a magnetic field of strength @xmath228
g threading a maximally rotating black hole of mass @xmath229 , heavy nuclei could be accelerated to energies of order @xmath58ev by exploiting the potential drop @xmath230v .
however , this maximal energy is quite optimistic because it ignores radiative energy losses , which are highly efficient in the nuclei of powerful agn ( norman et al .
1995 , henri et al .
1999 ) . in any case
, the poynting flux that is required , @xmath231erg / s far exceeds any other luminosity measured elsewhere in the galaxy , and in particular the estimate obtained from multi - band spectral modelling of the nucleus emission ( chiaberge , capetti & celotti 2001 ) .
let us start by summarizing the results obtained so far .
we have discussed how one can can test the chemical composition of ultra - high rays on the sky by comparing the anisotropy signals at various energies .
our main result is to show that if anisotropies are observed above some energy @xmath1 and the composition is assumed to be heavy at that energy ( nuclei of charge @xmath0 ) , one should observe at energies @xmath2 a substantially stronger anisotropy signal , which is associated with the proton component emitted by the sources that are responsible for the anisotropy pattern .
these conclusions do not depend at all on the modelling of the intervening magnetic fields and they are robust with respect to the parameters characterizing the sources . we have then applied this test to the most recent data of the pao .
we find that , if these data are taken at face value ( and if the pao arrival directions around the ankle appear isotropic ) , the events that are responsible for the anisotropy signal reported toward cen a and toward the nearby agn should not be heavy nuclei .
interestingly , this result appears to be at odds with the current pao results on the chemical composition at the highest energies .
one can not exclude at the present time that the observed anisotropies are statistical accidents ; only the acquisition of a larger set of data will tell .
one can not exclude either that the composition switches abruptly from heavy to light at @xmath232eev .
finally , one can not exclude that a systematic bias affects the composition measurements of the pao . in any case
, the present discussion indicates that , if both the anisotropy and the heavy composition are confirmed by future data , some possibly important information is to be extracted out of this apparent contradiction . the theoretical and experimental uncertainties in the extrapolation of the @xmath233 cross - section to center - of - mass energies @xmath234tev are one of the possible sources of biases in shower reconstruction .
recently , it has been argued that , if this cross - section were underestimated by some @xmath235% at these energies , one might reconcile the existing @xmath236 measurements with a pure proton composition at energies above the ankle ( ulrich et al .
2009a , 2009b ; see however wibig 2009 ) . in this respect , and in the light of the above discussion , the apparent anisotropy towards cen a offers a very valuable opportunity to test this possibility .
it suffices , in principle , to test the anisotropy pattern against a future ( independent ) data set at various energies by comparing the number of events detected within a predetermined solid angle in a predetermined direction to the expected number for isotropic arrival directions ; at the same time , one needs to perform a dedicated composition measurement for cosmic rays from this particular region of the sky , in order to avoid as much as possible the contamination due to the isotropic contribution to the all - sky flux .
then , one should apply the test discussed above .
the natural question that follows is what can be inferred about the sources of ultra - high energy cosmic rays if , following the anisotropy results of pao and the above discussion , one assumes that most ultra - high energy cosmic rays are protons .
let us now explore this line of reasoning .
first of all , one concludes that the existing data of the pao disfavor the agn model of ultra - high energy cosmic ray origin .
indeed , we have discussed in section [ sec : agn ] a robust lower bound to the magnetic luminosity that a source must possess in order to be able to accelerate particles of charge @xmath0 up to @xmath58ev , @xmath237erg / s .
this bound allows to rule out the local radio quiet agn as sources of @xmath58ev protons .
we have also argued that , while fr i radio - galaxies do not seem to possess the required luminosity to accelerate protons up to @xmath58ev , fr ii radio - galaxies and blazars do not correlate with the highest energy events .
although radio - loud galaxies could possibly accelerate heavy nuclei up to energies close to the gzk cut - off , as we have discussed they would produce an anisotropy pattern predominantly at @xmath238eev energies rather than at ultra - high energies , in conflict with the pao results .
we have discussed the particular case of cen a in some detail , as this source has recently received a lot of attention due to the observed clustering of events around it .
we have provided a critical discussion of acceleration in this object and argued that its agn / jet / lobe system can not produce protons at the energies required , in good agreement with the general expectations from the above luminosity bound .
that being said , the excess clustering in this particular direction , if real , remains to be interpreted .
first , if sources are distributed according to the large scale structure , one expects a certain number of events to coincide accidentally with the direction of cen a , because of the location of the centaurus and shapley superclusters in this direction . using the pscz catalog of galaxies as a proxy for the distribution of the sources , with a depth of @xmath239mpc , and accounting for the pao exposure , one expects @xmath240 events above @xmath51eev within @xmath241 of cen a nucleus ( and @xmath242 for isotropic arrival directions ) for @xmath243 observed , or @xmath244 events within @xmath82 ( and @xmath86 for isotropic arrival directions ) for @xmath81 observed .
note that the large scale structure is poorly sampled in the proximity of cen a due to its low galactic latitude .
nevertheless , the above numbers already indicate that the contamination from the surrounding large scale structure in the direction to cen a is substantial . in particular , at the present level of statistics , one can not claim observing a significant departure from a model in which the sources are distributed as the large scale structure .
it should also be recalled here that the pao analysis of clustering around cen a is an a posteriori one , hence one may not therefore assign a reliable significance to the detection of the anisotropy . moreover ,
if sources are gamma - ray bursts or magnetars , one should include in the flux prediction the contribution of such sources located in the cen a host galaxy itself .
it is well known that the probability of detecting ultra - high energy cosmic rays from nearby gamma - ray bursts is extremely small unless the arrival times of these ultra - high energy cosmic rays are sufficiently dispersed ( waxman 1995 , waxman & miralda - escud 1996 ) .
consider for instance gamma - ray bursts with rate of occurence @xmath245 in cen a ; these can be seen if either of their jets points into a solid angle @xmath246 where @xmath89 is the typical deflection acquired by the cosmic ray en route to the detector , giving a detection probability @xmath247 , with @xmath248 ; @xmath249 represents the spread of arrival times and @xmath250 the angular deflection for particles of energy @xmath251eev traveling from cen a to the detector through a magnetic field of strength @xmath252 g and coherence length @xmath253kpc ( waxman & miralda - escud 1996 ) .
the above probability suggests that one should not detect events from gamma - ray bursts emitted directly toward the observed .
however , if the jet of the gamma - ray burst hits one of the lobes of cen a , the particles can be redistributed in all directions , in particular towards the detector .
the magnetic field inside the lobe , @xmath254 g is indeed sufficient to impart a deflection of order unity on the lobe distance scale @xmath48kpc to protons of @xmath255eev . in this case , the expected time delay and time dispersion are of the order of the travel time to the lobes , @xmath256yr , so that the expected number of gamma - ray bursts that one can see at any time , through rescattering of the particles on the lobes of cen a is : @xmath257 for an apparent solid angle of the lobes @xmath258 as viewed from the host galaxy of cen a , which is close to @xmath259 .
the last gamma - ray burst(s ) of cen a may then contribute to the flux above the gzk cut - off up to : @xmath260 with @xmath261yrs , @xmath262 for an injection spectrum with @xmath263 , and @xmath264ergs the energy injected by one gamma - ray burst in ultra - high energy cosmic rays .
this flux should be spread over the angular scale of the lobes , @xmath265 . at @xmath255eev ,
pao records a diffuse flux @xmath266 .
after correcting for the solid angle of the cen a image and the pao aperture towards cen a , one finds that the last gamma - ray bursts in cen a should contribute up to @xmath267% of the observed pao flux , well within the range of the observed anisotropy .
we also note that this contribution from sources inside cen a would improve the overall correlation of all the pao data above @xmath51eev with sources distributed as the local large scale structure .
finally , we remark that the above effect of rescattering on the lobes provides a clear example of the possible biases discussed in kotera & lemoine ( 2008 ) , namely that the pao could be preferentially observing the last scattering centers ( in the present case , the lobes of cen a ) rather than the source itself ( the gamma - ray bursts located in the core of cen a ) .
abbasi , r. u. , et al .
( the hires collaboration ) , 2004 , , 610 , l73 abbasi , r. u. , et al .
( the hires collaboration ) , 2005 , , 623 , 164 abbasi , r. u. , et al .
( the hires collaboration ) , 2008 , physical review letters , 100 , 101101 abraham , j. et al .
( the pierre auger collaboration ) , 2007 , science , 318 , 938 abraham , j. , et al .
( the pierre auger collaboration ) , 2008a , physical review letters , 101 , 061101 abraham , j. et al .
( the pierre auger collaboration ) , 2008b , astroparticle physics , 29 , 188 allard , d. , busca , n. g. , decerprit , g. , olinto , a. v. , & parizot , e. 2008 , journal of cosmology and astro - particle physics , 10 , 33 allard , d. , & protheroe , r. j. 2009 , arxiv:0902.4538 aloisio , r. , berezinsky , v. , & gazizov , a. 2008 , arxiv:0803.2494 arons , j. 2003 , , 589 , 871 belz , j. ( the hires collaboration ) , _ proc .
( malfa , italy , sept .
2008 ) aublin , j. et al .
( the pierre auger collaboration ) , 2009 , arxiv:0906.2347 , icrc 2009 bertone , g. , isola , c. , lemoine , m. , & sigl , g. 2002 , , 66 , 103003 casse , f. , lemoine , m. , & pelletier , g. 2002 , , 65 , 023002 cavallo , g. 1978 , , 65 , 415 celotti , a. , ghisellini , g. , 2008 , mnras , 385 , 283 chiaberge , m. , capetti , a. , & celotti , a. 2001 , , 324 , l33 cuoco , a. , & hannestad , s. 2008 , , 78 , 023007 dermer , c. d. , razzaque , s. , finke , j. d. , & atoyan , a. , 2008 , arxiv:0811.1160 evans , n. w. , ferrer , f. , sarkar , s. , 2003 , prd 69 , 3501 .
evans , n. w. , ferrer , f. , sarkar , s. , 2004 , prd 69 , 8302 .
farrar , g. r. , & piran , t. 2000 , arxiv : astro - ph/0010370 farrar , g. r. , berlind , a. a. , & hogg , d. w. 2006 , , 642 , l89 farrar , g. r. , & gruzinov , a. , 2009 , , 693 , 329 finley , c. b. , & westerhoff , s. 2004 , astroparticle physics , 21 , 359 ghisellini , g. , ghirlanda , g. , tavecchio , f. , fraternali , f. , & pareschi , g. 2008 , arxiv:0806.2393 gorbunov , d. s. , tinyakov , p. g. , tkachev , i. i. , troitsky , s. v. , 2002 , apj 577 , l93 .
gorbunov , d. s. , tinyakov , p. g. , tkachev , i. i. , troitsky , s. v. , 2004 , astro - ph/0406654 .
gorbunov , d. s. , tinyakov , p. g. , tkachev , i. i. , & troitsky , s. v. 2008 , arxiv:0804.1088 greisen , k. , 1966 , prl 16 , 748 .
gupta , n. 2008 , journal of cosmology and astro - particle physics , 6 , 22 hague , j. d. et al .
( the pierre auger collaboration ) , arxiv:0906.2347 , icrc 2009 halzen , f. , & omurchadha , a. 2008 , arxiv:0802.0887 harari , d. , et al .
( the pierre auger collaboration ) , 2007 , arxiv:0706.1715 hardcastle , m. j. , cheung , c. c. , feain , i. j. , & stawarz , l. 2009 , mnras , 393 , 1041 henri , g. , pelletier , g. , petrucci , p. o. , & renaud , n. 1999 , astroparticle physics , 11 , 347 hillas , a. m. 1984 , , 22 , 425 holder , j. , & for the veritas collaboration 2008 , arxiv:0810.0471 israel , f. p. 1998
, , 8 , 237 kachelriess , m. , ostapchenko , s. , & tomas , r. 2008 , arxiv:0805.2608 kashti , t. , & waxman , e. 2008 , journal of cosmology and astro - particle physics , 5 , 6 kotera , k. , & lemoine , m. 2008 , , 77 , 123003 lyutikov , m. , & ouyed , r. 2005 , arxiv : astro - ph/0507620 massaglia , s. , 2007 , nuc .
168 , 302 milgrom , m. , & usov , v. 1995 , , 449 , l37 moskalenko , i. v. , stawarz , l. , porter , t. a. , & cheung , c. c. 2008 , arxiv:0805.1260 nagar , n. m. , & matulich , j. 2008 , arxiv:0806.3220 neronov , a. , semikoz , d. , & tkachev , i. 2007 , arxiv:0712.1737 norman , c. a. , melrose , d. b. , & achterberg , a. 1995 , , 454 , 60 osullivan , s. , reville , b. , & taylor , a. m. , 2009 , arxiv:0903.1259 pelletier , g. , lemoine , m. , & marcowith , a. , 2008 , arxiv:0707.3459 panov , a. d. , et al .
( atic-2 collaboration ) 2006 , arxiv : astro - ph/0612377 rachen , j. p. , & biermann , p. l. 1993 , , 272 , 161 rachen , j. , 2008 , in proceedings of `` rencontres de blois 2008 - challenges in particle astrophysics '' , blois ( france ) , may 18 - 23 2008 , ed
. j. dumarchez rieger , f. m. , bosch - ramon , v. , & duffy , p. 2007
, , 309 , 119 romero , g. e. , combi , j. a. , perez bergliaffa , s. e. , & anchordoqui , l. a. 1996 , astroparticle physics , 5 , 279 saunders , w. , frenk , c. , rowan - robinson , m. , lawrence , a. , & efstathiou , g. 1991 , , 349 , 32 stanev , t. , biermann , p. l. , lloyd - evans , j. , rachen , j. p. , watson , a. a. , 1995 , prl 75 , 3056 .
takeda , m. , et al . 1999 , , 522 , 225 tinyakov , p. g. , tkachev , i. i. , 2001 , pisma zh .
74 , 3 [ jetp lett .
tinyakov , p. g. , tkachev , i. i. , 2002 , astropart .
tinyakov , p. g. , tkachev , i. i. , 2004 , prd 69 , 8301 .
uchihori , y. , nagano , m. , takeda , m. , teshima , m. , lloyd - evans , j. , watson , a. a. , 2000 , astropart .
ulrich , r. , engel , r. , mller , s. , pierog , t. , schssler , f. , & unger , m. 2009a , arxiv:0906.0418 ulrich , r. , engel , r. , mller , s. , schssler , f. , & unger , m. 2009b , arxiv:0906.3075 unger , m. et al .
( the pierre auger collaboration ) , icrc-2007 , arxiv:0706.1495 vron - cetty , m .-
, & vron , p. 2006
, , 455 , 773 vietri , m. 1995 , , 453 , 883 wahlberg , h. et al . ( the pierre auger collaboration ) , 2009 , arxiv:0906.2319 , icrc 2009 waxman , e. 1995 , phys .
lett . , 75 , 386 .
waxman , e. , & miralda - escud , j. 1996 , , 472 , l89 waxman , e. , 2005 , phys .
scripta , t121 , 147 waxman , e. , & loeb , a. , 2008 , arxiv:0809.3788 wibig , t. 2008 , arxiv:0810.5281 wibig , t. 2009 , , 79 , 094008 yamamoto , t. et al .
( the pierre auger collaboration ) , icrc 2007 zatsepin , g. t. , kuzmin , v. a. , 1966 , pisma zh .
fiz . 4 , 114 [ jetp .
lett . 4 , 78 ] .
zaw , i. , farrar , g. r. , & greene , j. e. 2009 , , 696 , 1218 | this paper proposes and discusses a test of the chemical composition of ultra - high energy cosmic rays that relies on the anisotropy patterns measured as a function of energy . in particular
, we show that if one records an anisotropy signal produced by heavy nuclei of charge @xmath0 above an energy @xmath1 , one should record an even stronger ( possibly much stronger ) anisotropy at energies @xmath2 due to the proton component that is expected to be associated with the sources of the heavy nuclei .
this conclusion remains robust with respect to the parameters characterizing the sources and it does not depend at all on the modelling of astrophysical magnetic fields . as a concrete example , we apply this test to the most recent data of the pierre auger observatory . assuming that the anisotropy reported above 55 eev is not a statistical accident , and that no significant anisotropy has been observed at energies @xmath3eev , we show that the apparent clustering toward cen a can not be attributed to heavy nuclei .
similar conclusions are drawn regarding the apparent excess correlation with nearby active galactic nuclei .
we then discuss a robust lower bound to the magnetic luminosity that a source must possess in order to be able to accelerate particles of charge @xmath0 up to @xmath4eev , @xmath5erg / s . using this bound in conjunction with the above conclusions
, we argue that the current pao data does not support the model of cosmic ray origin in active radio - quiet or even radio - loud galaxies .
finally , we demonstrate that the apparent clustering in the direction of cen a can be explained by the contribution of the last few gamma - ray bursts or magnetars in the host galaxy thanks to the scattering of the cosmic rays on the magnetized lobes . |
the abundance of massive clusters of galaxies provides sensitive constraints on the cosmological parameters that govern structure growth in the universe .
however , a prerequisite for this is reliable mass measurements for large samples of clusters with well - understood selection criteria .
cluster mass measurements used have traditionally come from virial analysis of the velocity dispersion measurements of cluster galaxies ( e.g. , frenk et al .
1990 ; carlberg et al . 1997 ; borgani et al . 1999 ) , or x - ray temperature measurements of the hot intra - cluster gas under the assumption that the gas is in hydrostatic equilibrium ( for a review see * ? ? ?
satellite observatories such as rosat , asca , xmm - newton and _ chandra _ have made increasingly accurate x - ray temperature measurements of clusters , and have produced well - defined cluster samples of sufficient size to accurately measure the x - ray temperature and luminosity functions ( e.g. * ? ? ?
* ; * ? ? ?
* ) . however , the relation between cluster mass and x - ray temperature and luminosity , respectively , must be determined to convert these into a reliable cluster mass function .
x - ray luminosities are available for large samples of clusters , but the x - ray luminosity is highly sensitive to the complex physics of cluster cores , making it challenging to relate to cluster mass . measuring x - ray temperatures is observationally much more demanding , but the x - ray temperature is mainly determined by gravitational processes and is hence more directly related to cluster mass than x - ray luminosity . both from simulations and observations
the intrinsic scatter in mass around the mass temperature relation is thus found to be much smaller ( @xmath7 , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) than the scatter in mass around the mass luminosity relation ( @xmath8 , * ? ? ?
two different routes have been followed for determining the mass temperature relation .
most studies have used a small sample ( up to about a dozen ) of supposedly well understood clusters for which the assumptions underlying the mass determination should be fulfilled to a high degree .
the main concern for this approach is that the selected clusters may not be representative of the whole cluster population , and therefore the derived mass temperature relation may only apply to a subset of clusters . alternatively , the mass temperature relation may be determined from a large sample of more objectively selected clusters .
this is more fruitful when comparing such a locally determined mass temperature relation to a sample of high - redshift clusters where the data quality does not allow a similar selection of the `` most suitable '' clusters .
also , mass temperature relations derived from simulations are usually based on a large range of simulated clusters with no pre - selection .
hence it is most appropriate to compare observationally obtained mass temperature relations determined from all available clusters to the relations from simulations . on the other hand , for some of the clusters in such a sample the hydrostatic assumption may be invalid , making x - ray based mass determinations unreliable for a subset of the clusters . a larger scatter ( which may not be symmetric ) around the mean mass temperature relation may be expected , when such clusters are included . there are still poorly understood systematic uncertainties associated with establishing the mass temperature relation .
the normalization of the mass temperature relation based on cluster masses determined from x - ray data @xcite , tend to differ significantly between studies , and from the expectations based on numerical simulations ( e.g. , evrard et al .
1996 ; eke et al .
1998 ; pen 1998 ; borgani et al . 2004 ) .
the determination of this normalization is currently the dominating source of discrepancies between the reported values for the power spectrum normalization on the scale of galaxy clusters , @xmath0 , derived from the observed cluster temperature function ( huterer & white 2002 ; seljak 2002 ; pierpaoli et al .
2003 ; henry 2004 ) . observations using x - ray based mass determinations have traditionally favored low normalizations ( and hence low values of @xmath0 ) , while simulations have favored somewhat higher normalizations .
gravitational lensing provides an opportunity to measure cluster masses without invoking the assumption of hydrostatic equilibrium in the hot intra - cluster gas implicit in the x - ray based mass determinations .
also , in this case the measurement of cluster mass is truly independent of the x - ray temperature measurement .
hjorth , oukbir & van kampen ( 1998 ) used weak gravitational mass measurements for eight clusters drawn from the literature to find a relation between mass over cluster - centric radius , and temperature .
they determined a normalization of this relation consistent with the value predicted by evrard et al .
( 1996 ) , but with a preference for somewhat higher cluster masses ( if the redshift scaling of equation [ eq : theorymt ] is assumed , see below ) . however , smith et al .
( 2005 ; hereafter s05 ) determined a mass temperature relation with a normalization significantly lower than indicated by the hjorth et al .
( 1998 ) study .
s05 based their results on a sample of 10 clusters with weak lensing masses and temperatures determined from _
chandra _ data . here
, we present a new weak gravitational lensing - based measurement of the normalization of the mass temperature relation . the main improvements with respect to the work of hjorth et al .
( 1998 ) and s05 is that we use a significantly larger cluster sample which represents a significant fraction of all the clusters in an even larger sample with well - defined objective selection criteria ( dahle et al . 2002 ; h. dahle 2006 , in preparation ) .
an additional improvement over the work of hjorth et al .
( 1998 ) is that the weak lensing analysis has been performed in a consistent way for all clusters , using the same shear estimator and making the same assumptions about e.g. , the typical redshift of the lensed galaxy population and the degree of contamination by cluster galaxies .
we note that the early data set of clusters with published weak lensing masses used by hjorth et al .
( 1998 ) is biased at some level towards systems that were observed because of `` extreme '' properties , such as being the hottest or most x - ray luminous system known at the time , or having a large number of strongly gravitationally lensed arcs .
furthermore , we note that our gravitational lensing measurements are made at larger radii than probed by s05 , requiring smaller extrapolations to estimate the mass within e.g. , the virial radii of the clusters . the data set used for the analysis
is described in
[ sec : data ] , our results for the mass temperature relation and @xmath0 are presented in [ sec : results ] , and our results are compared to other work and the implications discussed in [
sec : discussion ] . except
when specifically noted otherwise ( for easy comparison to previous results using different cosmologies ) , we assume a spatially flat cosmology with a cosmological constant ( @xmath5 , @xmath9 ) , and the hubble parameter is given by @xmath10 .
our weak lensing data set is a sample of 30 clusters ( see table [ tab : dataset ] ) , of which 28 were included in the weak lensing cluster sample of dahle et al .
data for two additional clusters come from a recent extension of this data set ( h. dahle 2006 , in preparation ) .
the clusters targeted for these weak lensing studies were generally selected to lie above an x - ray luminosity limit @xmath11 ergs s@xmath12 ( this luminosity limit is for our chosen cosmology with @xmath13 ) and within a redshift range @xmath14 .
the observed clusters were selected from the x - ray luminous cluster samples of briel & henry ( 1993 ) and ebeling et al .
( 1996;1998;2000 ) .
the cluster samples of the first two of these papers are based on correlating an optically selected cluster sample ( abell 1958 ; abell , corwin , & olowin 1989 ) with x - ray sources from the rosat all - sky survey ( rass ; trmper et al .
1993 ) , while the two last papers contain x - ray flux limited cluster catalogs , also based on rass . of the total sample of 30 clusters , three ( , , and )
are drawn from the briel & henry ( 1993 ) sample and two ( and ) are drawn from the xbac sample of ebeling et al .
( 1996 ) . of the remaining 25 clusters , 22
are included in the x - ray brightest cluster sample ( bcs ) of ebeling et al .
( 1998 ) , while three ( , , and ) come from its low - flux extension ( ebcs ; ebeling et al .
2000 ) . of the bcs and ebcs clusters in our sample ,
24 objects are included in a volume - limited sample of 35 clusters selected from the bcs and ebcs samples ( dahle 2006 ) .
hence , while our sample is not strictly physically well - defined ( in the sense that the availability of an x - ray temperature measurement is one of the defining selection criteria ) , it still has significant overlap with a well - defined cluster sample . in a recent paper ,
stanek et al . ( 2006 ) discuss how a significant scatter around the mean mass - luminosity relation may cause a significant malmquist bias in x - ray flux - limited cluster samples , causing high - mass , low flux clusters to drop out at high redshifts .
this would result in a bias in the mass - luminosity ( or mass - temperature ) relation derived based on such a sample .
we note , however , that although the rass - based samples from which our cluster sample is drawn are flux - limited , the cluster sample discussed here quite closely approximates a volume - limited sample , and we therefore expect any such bias to be negligible .
the observations were made with the @xmath15 uh8k mosaic ccd camera and the @xmath16 tek ccd camera at the 2.24 m university of hawaii telescope and with the @xmath16 alfosc ccd camera at the 2.56 m nordic optical telescope .
all clusters were imaged in both the @xmath17- and @xmath18-band , with typical total exposure times of 3.5h in each passband for the uh8k data and @xmath19h for the data obtained with the more sensitive @xmath16 detectors .
the seeing was in the range @xmath20 for all the imaging data used for the weak lensing analysis .
the median seeing was 082 in the @xmath17-band and 09 in the @xmath18-band .
this gave typically @xmath21 usable background galaxies per square arcminute , or a `` figure of merit '' value of @xmath22deg@xmath23 , as defined by kaiser ( 2000 ) .
as noted below , the background galaxies were selected based on signal to noise ratio rather than magnitude , with limits corresponding to @xmath24 and @xmath25 for point sources .
the observations and data reduction of the data set used for the weak lensing mass measurements are described in detail by dahle et al .
( 2002 ) .
major efforts are being made to improve the methods for the estimation of weak gravitational lensing , particularly in connection with ongoing and future studies of `` cosmic shear '' based on wide - field optical surveys .
the requirements for the precision of shear estimates in these surveys are substantially more stringent than for weak lensing observations of massive clusters , given the significantly weaker lensing effects measured in random fields . in this work ,
we have used the shear estimator of kaiser ( 2000 ) , which was `` blind - tested '' ( along with several other shear estimators ) by heymans et al .
( 2005 ) , using simulated lensing data .
the shear estimator of kaiser ( 2000 ) is more mathematically rigorous than the currently most widely used shear estimator ( kaiser , squires , & broadhurst 1995 ) , but it displays a significant non - linear response to shear , unlike most other shear estimators . if we correct our shear values using a second order polynomial based on the test results of heymans et al .
( 2005 ) , we find that most cluster masses stay within @xmath26% of the mass calculated based on uncorrected shear values .
furthermore , the change in average cluster mass is @xmath27% , i.e. , there is very little systematic shift in mass . in the end , we chose not to apply this correction , since it would , in a few cases , require extrapolations outside the range of shear values over which the shear estimator has been tested . for more details about the practical implementation of this shear estimator ,
see dahle et al .
( 2002 ) . to convert the measurements of weak gravitational shear into actual cluster masses , the distances to the background galaxies need to be known .
the background galaxy redshifts were estimated from spectroscopic and photometric redshifts in the hubble deep field ( for details , see dahle et al .
2002 ) . for our data set and chosen cosmological model , the average value of the ratio between the lens - source and observer - source angular diameter distances , @xmath28/@xmath29 , is well approximated by the relation @xmath30 within the redshift range of our cluster sample .
this then provides an effective critical surface density for lensing ( @xmath31 ; where @xmath32 is the angular diameter distance to the cluster ) , which is used for deriving cluster masses from the shear estimates .
the quoted value of @xmath33 corresponds to the value at large cluster radii ; at smaller radii a correction term has to be employed to account for contamination by cluster galaxies , as discussed below and illustrated in figure [ fig : cluscont ] .
the observable galaxy shape distortions caused by gravitational lensing provide a measurement of the reduced tangential shear , @xmath34 , where @xmath35 is the tangential component of the shear and @xmath36 is the convergence .
we fit an nfw - type mass density profile , @xmath37 ( navarro , frenk , & white 1997 ) , to the observed reduced shear profile @xmath38 of each cluster .
here , @xmath39 is the critical density of the universe at the redshift of the cluster , and @xmath40 we assumed a concentration parameter @xmath41 , corresponding to the median halo concentration predicted by bullock et al .
( 2001 ) for a @xmath42 cluster from simulations of dark matter halos in a @xmath4cdm universe . here , @xmath43 , and @xmath44 , where @xmath45 is defined as the radius within which the average mass density is 200 times the critical density @xmath46 , and
@xmath47 is the virial radius of the cluster .
the lensing properties of the nfw model have been calculated by bartelmann ( 1996 ) and wright & brainerd ( 2000 ) . from our fit
, we calculated @xmath48 , the mass enclosed by the radius @xmath49 .
the mass estimates are listed in table [ tab : dataset ] .
the shear measurements used for the fit were made at clustercentric radii @xmath50 for the clusters that were observed with @xmath16 ccd cameras and @xmath51 for the clusters that were observed with the uh8k camera . by comparison , we find @xmath49 values typically in the range @xmath52 for the clusters we study here . in many cases ,
we need to extrapolate the nfw profile out to @xmath49 ( in table [ tab : dataset ] we list the ratio of the outermost radii of our shear measurements , @xmath53 , to @xmath49 , and note that @xmath54 for our chosen nfw model ) . in this extrapolation , we assume the median nfw concentration parameter given above . hence , any intrinsic scatter in @xmath55 will introduce an extra uncertainty in the cluster mass estimates . if we assume a random scatter around the mean value of @xmath55 at the level ( a @xmath56 @xmath57 ) predicted by bullock et al .
( 2001 ) , we find a corresponding scatter in the mass estimates of 20% for our data set . this additional scatter is not included in the uncertainties of the listed mass measurements in table [ tab : dataset ] , but is considered further in section 3.2 .
the measured gravitational lensing signal is sensitive to the two - dimensional surface mass distribution , including mass associated with the cluster outside @xmath49 , and random structures seen in projection along the line of sight ( metzler , white & loken 2001 ; hoekstra 2001 ; clowe , de lucia & king 2004 ; de putter & white 2005 ) .
this will introduce additional uncertainty ( and potentially a net bias ) to any lensing - based estimates of the cluster mass contained within a 3d volume .
studies based on simulated clusters ( e.g. , clowe et al .
2004 ) indicate that the net bias is no more than a few percent when 3d cluster masses are estimated by fitting observations of @xmath38 to predictions from theoretical models of the mass distribution , such as the nfw model
. however , the scatter in the mass estimates from projection effects amount to a weak lensing mass dispersion of @xmath58 for massive galaxy clusters , which should be added to the observational uncertainties of the lensing mass estimates . in this paper , we have assumed a lensing mass dispersion of 0.26 resulting from projection effects , corresponding to the value estimated by metzler , white & loken ( 2001 ) from their n - body simulations . although these authors considered a somewhat different mass estimator , more recent estimates indicate a similar mass dispersion for the nfw profile fitting method that we have used .
this additional mass uncertainty has been added in quadrature to the uncertainties of @xmath48 values listed in table [ tab : dataset ] .
the absence of reliable information about the individual redshifts of the faint galaxies used for the weak lensing measurement will inevitably result in some degree of confusion between lensed background galaxies and unlensed cluster galaxies .
the magnitude of this effect will depend on the projected number density of cluster galaxies , and should thus have a strong dependency on cluster radius .
hence , a radially dependent correction factor was applied to the shear measurements to correct for contamination from cluster galaxies in the faint galaxy catalogs that were used to measure the gravitational shear ( these catalogs included all galaxies in the cluster fields that were detected at a signal - to - noise ratio @xmath59 , with no additional selection based on e.g. , galaxy color ) .
the magnitude of this correction was estimated from the radial dependence of the average faint galaxy density in two `` stacks '' of clusters observed with the uh8k camera , one at @xmath60 and the other at @xmath61 , assuming that the contamination is negligible at the edge of the uh8k fields , @xmath62 mpc from the cluster center .
the estimated degree of contamination is shown in figure [ fig : cluscont ] . given the difference in cluster redshift , the similarity of the two curves in figure [ fig : cluscont ] may be somewhat surprising , as one would naively expect the more distant clusters to display a significantly lower surface density of cluster galaxies .
however , there are several competing effects that affect the observed galaxy density at a fixed angular radius : firstly , if all cluster galaxies were detectable regardless of cluster redshift , the change in apparent image scale with redshift should increase the surface density by a factor given by the square of the ratio of the angular diameter distances . on the other hand , a fixed angular radius would correspond to a larger physical cluster - centric radius ( and hence lower galaxy density in physical units ) at the larger redshift , the difference depending on the slope of the radial galaxy density profile . at the radii probed in this study ,
both the radial surface mass density profile and the number density profile of bright cluster galaxies follow approximately the power law behavior of a singular isothermal sphere ( @xmath63 , with @xmath64 ) .
hence , the physical number density ( in galaxies / mpc@xmath65 ) at a fixed angular radius should decrease as the inverse of the ratio of angular diameter distances .
in addition , the faintest galaxies drop below the detection limit at higher redshift , the effect depending on the slope of the cluster luminosity function around @xmath66 . assuming a schechter ( 1976 ) luminosity function with a faint - end slope @xmath67 ( typical of rich clusters ) and @xmath68 , the luminosity function can be integrated down to the detection limit ( corresponding to @xmath69 and @xmath70 at @xmath71 and @xmath72 , respectively ) , to estimate the fraction of cluster galaxies that drop out at the higher redshift ( @xmath73% ) .
finally , a redshift - dependence given by @xmath74 was assumed to account for galaxy evolution in the clusters .
the combination of all these effects would predict a surface density of cluster galaxies which is 7% less at @xmath75 , compared to @xmath71 , for a fixed cluster richness .
even this small difference would be erased by a slight decrease in the assumed values of the slopes @xmath76 and @xmath77 .
a faint - end slope of the luminosity function of @xmath78 would be sufficient to remove the predicted difference in galaxy surface density at the two different redshifts .
based on _ hubble space telescope _ ( hst ) wfpc2 imaging of the galaxy cluster ( which is similar to the clusters studied here in terms of optical richness , lensing mass and x - ray properties ) , pracy et al .
( 2004 ) find that the cluster core shows a relative depletion of dwarf galaxies , leading to a radial profile of faint galaxies which is significantly shallower than the sis prediction .
for the `` intermediate '' dwarf population ( @xmath79 , similar to the range in absolute magnitude of cluster galaxies in our faint galaxy catalogs ) , these authors find a radial distribution with a slope @xmath80 . assuming a similar slope for
the radial distribution of the faint cluster galaxies in our catalogs would also remove the predicted difference between the two curves in figure [ fig : cluscont ] . based on this figure , and the above discussion
, we conclude that we are probably justified in ignoring the redshift - dependence in our cluster galaxy contamination correction .
the level of cluster galaxy contamination for individual clusters will generally differ from the mean level calculated above , as there will be significant cluster - to - cluster variations in the abundances of cluster dwarf galaxies .
based on the sample of clusters observed with the uh8k camera , the scatter in dwarf galaxy richness was estimated to be @xmath81 ( this estimate also includes variations in the field galaxy density caused by uncorrelated large - scale structures along the line of sight , and hence the true scatter in dwarf galaxy richness of the clusters is somewhat overestimated ) . by employing the mean contamination correction calculated above rather than an estimate appropriate for each cluster ,
we introduce an additional scatter of up to @xmath82 in our mass estimates .
this additional scatter is not included in the uncertainties of the tabulated mass measurements in table [ tab : dataset ] , but is considered further in section 3.2 .
eight of the clusters in our sample were also included in the combined strong and weak gravitational lensing study of s05 , based on observations of a sample of 10 x - ray luminous galaxy clusters at @xmath83 using hst wfpc2 .
these authors estimated the projected cluster mass within a clustercentric radius of @xmath84 kpc , assuming an einstein - de sitter ( @xmath85 , @xmath86 ) cosmology .
figure [ fig : s05comp ] shows a comparison of the mass values listed by s05 with our cluster mass estimates , using the best - fit nfw model to derive projected cluster masses , assuming the same cosmology as s05 .
these authors assumed a spatially constant contamination of 20% cluster galaxies in their background galaxy catalogs at radii @xmath87 , while we find an average contamination of 30% for our data at these radii .
hence , for the plot in figure [ fig : s05comp ] we have adjusted our radially dependent contamination correction such that the average contamination at small radii is consistent with that assumed by s05 .
we find that our cluster mass estimates are generally consistent with those of s05 , although with a tendency for higher masses ( by about 30% ) .
for clusters in the weak lensing data set , we compiled a list of corresponding x - ray temperatures from the literature . for many of these clusters , their global temperature , or even a temperature map , has been determined using data from _ chandra _ and/or xmm - newton .
however , these temperatures constitute a rather heterogeneous sample , for which the systematics are not well established .
consequently , x - ray temperatures were primarily drawn from the samples of @xcite , @xcite and @xcite , each providing a homogeneous measure of the global cluster temperature ( i.e. , temperature measured within a cluster - centric distance close to @xmath49 ) for a large fraction of the clusters in the weak lensing sample .
all these authors derived temperatures based on analysis of asca spectra . for two clusters not in either of the samples mentioned above
we extracted published temperatures from other sources .
specifically , from the works of @xcite and @xcite we extracted temperatures estimated by these authors by fitting an isothermal plasma model with the galactic absorbing column density as a free parameter .
the temperatures taken from @xcite were derived by fitting an isothermal plasma model with the nominal galactic absorbing column density fixed , but since only energies above 1 kev are used in the @xcite spectral fitting the fixed column density should not introduce systematic effects relative to the temperatures from the @xcite and the @xcite samples . for the remaining two clusters we extracted
published temperatures obtained in a similar way ( see table [ tab : dataset ] ) .
for the clusters in two or more samples , their derived temperatures agree within the uncertainties , and for any cluster the derived temperatures differ by less than 20% between samples . also , the mean of temperature differences between any two of the samples by @xcite , @xcite , and @xcite is less than 3% , indicating the low level of systematic temperature variance between different analyses .
although the isothermal plasma model has proven too simplistic for nearby clusters , the global cluster temperature is straightforward to derive from observations as well as simulations , enabling a rather direct comparison between observations and theory .
furthermore , for the majority of distant ( @xmath88 ) clusters only global isothermal temperatures can be obtained in the foreseeable future .
hence , we refrain from going into the detailed spatial and spectral modeling of the intra - cluster gas .
the effects of cluster dynamics , `` cooling cores '' , non - sphericity etc . generally affects the global temperatures only at the 10%-20% level ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
lcccccccc & 0.255 & @xmath89 & 0.30 & @xmath90 & & & + & 0.197 & @xmath91 & 0.50 & @xmath92 & & & + & 0.206 & @xmath93 & 1.62 & & & & @xmath94 + & 0.230 & @xmath95 & 1.32 & @xmath96 & & & + & 0.203 & @xmath97 & 1.20 & & @xmath98 & & + & 0.171 & @xmath99 & 0.21 & @xmath100 & @xmath101 & @xmath102 & + & 0.288 & @xmath103 & 0.59 & & & @xmath104 & + & 0.182 & @xmath105 & 0.36 & @xmath106 & @xmath107 & @xmath108 & + & 0.282 & @xmath109 & 0.39 & @xmath110 & & @xmath111 & + & 0.217 & @xmath112 & 0.30 & @xmath113 & @xmath114 & @xmath115 & + & 0.285 & @xmath116 & 1.45 & @xmath117 & & & + & 0.206 & @xmath118 & 1.51 & @xmath119 & @xmath120 & @xmath121 & + & 0.299 & @xmath122 & 1.58 & & & @xmath123 & + & 0.226 & @xmath124 & 0.55 & @xmath125 & & @xmath126 & + & 0.325 & @xmath127 & 2.58 & @xmath128 & & & + & 0.280 & @xmath129 & 0.33 & @xmath130 & & & + & 0.228 & @xmath131 & 0.45 & @xmath132 & & @xmath133 & + & 0.253 & @xmath134 & 0.39 & @xmath135 & @xmath136 & @xmath137 & + & 0.171 & @xmath138 & 0.44 & & & @xmath139 & + & 0.320 & @xmath140 & 1.23 & @xmath141 & & @xmath142 & + & 0.153 & @xmath143 & 0.23 & @xmath144 & & @xmath145 & + & 0.229 & @xmath146 & 0.47 & @xmath147 & & & + & 0.152 & @xmath148 & 0.27 & @xmath149 & @xmath150 & @xmath151 & + & 0.228 & @xmath152 & 0.45 & & @xmath153 & @xmath154 & + & 0.224 & @xmath155 & 0.32 & @xmath156 & @xmath157 & @xmath158 & + & 0.258 & @xmath159 & 0.58 & & @xmath160 & @xmath161 & + & 0.345 & @xmath162 & 0.44 & @xmath163 & & & + & 0.164 & @xmath164 & 0.39 & & & & @xmath165 + & 0.235 & @xmath166 & 0.38 & @xmath167 & & & + & 0.291 & @xmath168 & 0.47 & & @xmath169 & @xmath170 & +
from the virial relation @xmath171 between cluster mass , @xmath172 , inside radius @xmath173 , galaxy velocity , @xmath174 , and gas temperature , @xmath175 , combined with the definition of mass within an over - density of 500 times the critical density @xmath176 where the term @xmath177 describes the evolution of the over - density for a given cosmology , the mass temperature relation is obtained as @xmath178 in this study , we have measured @xmath48 , relating to @xmath179 through @xmath180 for the average cluster redshift with our chosen nfw model .
we take into account that the slope may deviate from the simple theoretical expectation @xmath181 and normalize the relation at 8 kev since our sample is dominated by massive clusters .
hence , mass temperature relations were obtained by fitting the data in table [ tab : dataset ] using the bces(@xmath182 ) estimator of @xcite to the following parameterization of the mass temperature relation @xmath183 the redshift - dependent factor , @xmath184 , contained in eq.[eq : fittedmtrelation ] must be calculated individually for each cluster , as this would otherwise produce an artificial 15% variation in mass over the redshift range spanned by our cluster sample . in effect , the normalization of the relation refers to the present epoch ( @xmath185 ) .
the fitting procedure of @xcite takes uncertainties in temperatures as well as in weak lensing masses into account , and makes no assumptions about the intrinsic scatter of both quantities .
results from fitting sub - samples as well as the full sample are presented in table [ tab : tstar ] and figure [ fig : mtrel ] . for the full data set ( for those clusters with temperature from more than one sample the temperature was taken in prioritized order from @xcite , @xcite , and @xcite ) we find the following normalization of the mass temperature relation at 8 kev @xmath186 and a slope of @xmath187 .
it is evident that the slope of the mass temperature relation is not well - determined since our data only span a modest range at the high mass / high temperature end of the cluster distribution . in fact , it is not obvious that there is a tight mass temperature relation at the high temperature end probed here . the concentration of clusters around @xmath188 kev enables a robust measurement of the normalization of the mass temperature relation at the high mass end . even though the slope of the mass temperature relation varies substantially between the three x - ray sub - samples @xcite the best fit normalizations agree within their statistical uncertainty .
we note that for all fits , the four different regressions of @xcite all result in normalizations within 20% .
the strongest constraints on the mass temperature relation normalization are obtained by taking advantage of previous studies of massive clusters @xcite , showing that the mass temperature relation slope is close to @xmath2 as expected from simple gravitational collapse models @xcite .
hence , in order to express the normalization of the mass temperature relation in terms of the characteristic temperature @xmath189 @xcite we assume @xmath181 ( with a representative uncertainty of 10% , e.g. * ? ? ?
as is custom for quoting @xmath190 values we adopt the redshift dependence factor @xmath191^{-3/2}$ ] @xcite where @xmath192 is the mean overdensity inside the virial radius in units of the critical density at the relevant redshift ( @xmath193 and @xmath184 differs by 7% at @xmath71 ) .
we find @xmath194 for our full sample and results from calculations of @xmath190 based on various subsamples are listed in table [ tab : tstar2 ] . from the @xmath0 - @xmath190 relation plotted by @xcite in their figure 2 , we find @xmath195 , based on our full sample .
we note that this relation is valid only for an intrinsic scatter in temperature of @xmath196 around the mean mass temperature relation .
a larger intrinsic scatter will imply a lower value of @xmath0 .
we provide our constraints on the intrinsic scatter below .
the squared scatter in lensing mass , @xmath48 , around the best fit , @xmath197 , is the sum of the squared measurement error , the squared intrinsic scatter , and the squared systematic errors @xmath198 .
the main systematic errors in the lensing mass ( see section 2.1 ) arise from extrapolating the assumed nfw mass profile out to @xmath49 ( due to cluster - to - cluster variations in the assumed concentration parameter @xmath199 ) and from the separation of cluster / background galaxies ( due to cluster - to - cluster richness variations ) .
each of these introduces a scatter of 20% in the lensing mass , hence @xmath200 . for the full sample
we find @xmath201 which is larger than expected from the mean lensing mass error , @xmath202 and the systematic errors , indicating either a sizable intrinsic scatter in mass or that the measurement / systematic errors are severely under - estimated .
accounting for errors in both mass and temperature , we find an intrinsic scatter in @xmath190 of @xmath203 .
there is a 70% probability that the scatter in temperature is larger than 10% , favoring somewhat lower values of @xmath0 than quoted above .
however , most of the scatter is caused by the low mass clusters .
we looked into whether relaxed clusters and non - relaxed clusters have the same normalization of the mass temperature relation .
our `` relaxed '' cluster sample consists of a586 , a963 , a1835 , a1995 , a2204 , a2261 , rxj1720 , and rxj1532 .
these are clusters with `` spherical '' optical and x - ray morphology , and no known cluster - scale dynamic disturbances . for
the relaxed clusters we find a normalization of the mass temperature relation of @xmath204 while the normalization for the non - relaxed clusters is @xmath205 ( see figure [ fig : relax ] ) .
the higher normalization of relaxed clusters is supported by the fact that the mean mass of relaxed clusters is a factor @xmath206 larger than the mean mass of `` non - relaxed '' clusters , although the relaxed and the non - relaxed clusters span roughly the same temperature range .
the scatter in mass for the relaxed sample ( @xmath207 ) is similar to the scatter for the non - relaxed sample ( @xmath208 ) .
the mean error for both samples is @xmath209 .
either relaxed clusters spread as much around their mass temperature relation as clusters in general , or we have used a poor definition of `` relaxed '' clusters . however , the fact that the normalization of the mass temperature relation for relaxed clusters is higher than for non - relaxed clusters indicates that there is a physical difference between the two sub - samples . from the present study
, it thus seems that relaxed clusters do not form a tighter mass temperature relation than clusters in general . for a given mass ,
non - relaxed clusters are found to be @xmath210% hotter than relaxed clusters .
since we consider global , isothermal temperatures , the presence of `` cooling cores '' in relaxed clusters will result in a lower global temperature than the virial temperature .
however , this effect is at the 10%-20% level ( e.g. , * ? ? ?
* ; * ? ? ?
* ) so this can not alone explain the temperature difference between relaxed and non - relaxed clusters .
based on the mass temperature relation from 10 clusters ( 3 of which are considered relaxed ) , s05 also find that non - relaxed clusters are hotter than relaxed clusters .
an objective classification of the degree of relaxation for a sizeable cluster sample is required for further quantifying the size of this effect .
ccr @xmath211 & @xmath212 & ota & mitsuda ( 2004 ) + @xmath213 & @xmath214 & allen ( 2000 ) + @xmath215 & @xmath216 & white ( 2000 ) + @xmath217 & @xmath218 & all + @xmath219 & @xmath220 & `` relaxed '' + @xmath221 & @xmath222 & `` non - relaxed '' + ccr @xmath223 & @xmath224 & ota & mitsuda ( 2004 ) + @xmath225 & @xmath224 & allen ( 2000 ) + @xmath226 & @xmath224 & white ( 2000 ) + @xmath227 & @xmath224 & all + @xmath228 & @xmath224 & `` relaxed '' + @xmath229 & @xmath224 & `` non - relaxed '' + lcccr x - rays & 0.09 & @xmath230 & fitted & @xcite + & 0.09 & @xmath231 & fitted & @xcite + & & & + lensing & 0.23 & @xmath232 & fixed & s05 + & 0.23 & @xmath233 & fitted & this study + & & & + simulations & 0.04 & @xmath234 & fixed & @xcite + & 0.00 & @xmath235 & fitted & @xcite +
based on the hitherto largest sample of x - ray luminous clusters with measured lensing masses , we derive a normalization of the mass temperature relation at the high mass end , @xmath236 . this value is higher than the lensing based mass temperature normalization of s05 , based on a smaller cluster sample , but is consistent with this within @xmath237 errors ; see table [ tab : mnorm ] .
temperature relations with masses determined from x - ray data tend to have a lower normalization than lensing based relations , and they are only marginally consistent with our normalization .
this is also the case for the two recent studies of @xcite and @xcite based on smaller samples of lower mass ( and hence cooler ) clusters . @xcite
measured cluster masses inside @xmath49 from x - ray observations of a sample of 13 low redshift clusters with a median temperature of 5.0 kev while @xcite determined the normalization from x - ray derived masses of 10 nearby clusters with a mean temperature of 4.8 kev .
the two studies agree on the same normalization , higher than previous x - ray mass based studies , but there still seems to be a @xmath238 discrepancy between x - ray and lensing derived mass temperature relations .
we note , however , that the lensing based and x - ray based normalizations are made at different redshifts , and that this discrepancy would vanish if the redshift - dependence predicted by the self - similar collapse model in equation [ eq : theorymt ] were neglected . given the heterogeneous nature of these data sets , any claim of significant departures from self - similarity would be premature , but this clearly provides an interesting avenue for future research , involving even larger cluster samples spanning a wider interval in redshift .
we confirm the result of smith et al .
( 2005 ) that non - relaxed clusters are on average significantly hotter than relaxed clusters .
this is qualitatively consistent with n - body / hydrodynamical cluster simulations which show that major mergers can temporarily boost the x - ray luminosities and temperatures well above their equilibrium values ( e.g. * ? ? ?
in contrast to several previous ( mainly x - ray mass based ) published mass temperature relations , the normalization derived in this study is in good agreement with the normalization derived from numerical simulations . however , the accuracy of the normalization is not good enough to discriminate between simulations including different physical processes .
our results show that x - ray based measurements of the cluster abundances , after reducing the major systematic uncertainties associated with the mass temperature normalization , give an amplitude of mass fluctuations on cluster scales that is consistent with other methods .
this lends additional support to the `` concordance model '' cosmology , and lends credence to the basic assumptions of gaussian density fluctuations .
our determination of @xmath239 is higher than most @xmath0 determinations from cluster data ( for a compilation of these , see e.g. , * ? ? ?
however , our finding is consistent with the value derived from weak gravitational lensing in the combined deep and wide cfht legacy survey ( @xmath240 ; semboloni et al .
2005 ) based on the halo model of density fluctuations ( smith et al .
it is also consistent with the cmb+2dfgrs+ly@xmath241 forest result ( @xmath242 ) of spergel et al .
( 2003 ) , with the joint cmb + weak lensing analysis of contaldi , hoekstra , & lewis ( 2003 ) , which gave @xmath243 , and with cmb analyses @xcite yielding @xmath244 .
however , the more recent 3-year wmap results ( spergel et al . 2006 ) give a significantly lower value of @xmath0 , and also a preference for a value of @xmath245 lower than 0.3 .
also , results from the recent 100 square degree weak lensing survey ( benjamin et al . 2007 ) favor a lower value of @xmath246 for @xmath5 .
we note that our quoted value of @xmath0 is based on the assumption that the intrinsic scatter about the mass temperature relation is @xmath196 , and that our @xmath0 estimate will be biased high if the true scatter significantly exceeds this value @xcite .
the limiting factor of our measurement of the normalization of the mass temperature relation is the magnitude of the measurement errors ( dominating the systematic errors , estimated to be @xmath247 ) . in order for the mass temperature relation to be a competitive route for constraining cosmological parameters and to discriminate between simulations with different input physics
, the normalization must be measured to better than @xmath248% accuracy
. however , there are good prospects for improving on these results in the near future .
firstly , the superior spectro - imaging capabilities of _ chandra _ and xmm - newton will allow the construction of large , homogeneous cluster temperature samples . a comparison to tailored simulations with realistic physics , analyzed in the same way as observations , will advance our understanding of systematics and the link between the mass temperature relation and structure formation ( c.b .
hededal et al . 2007 , in preparation ) .
secondly , more accurate weak lensing - based mass measurements of a larger sample of clusters are feasible as large mosaic ccd cameras that can probe intermediate - redshift clusters beyond their virial radii are now common , and the cluster sample could easily be doubled from a similar survey in the southern celestial hemisphere .
finally , we note that a more direct measurement of @xmath0 from weak lensing by clusters is possible , provided that weak lensing mass estimates are available for a large , well - defined , volume - limited cluster sample .
this makes it feasible to calculate the cluster mass function directly from the lensing masses , rather than indirectly via the x - ray temperature function ( dahle 2006 ) . since mass estimates based on baryonic tracers of the total cluster mass only enters indirectly as a selection criterion ( e.g. , clusters selected based on x - ray luminosity above a certain threshold ) , the method is less susceptible to systematic and random errors , as it does not require an accurate characterization of the scatter around the mean mass temperature relation .
we thank per b. lilje and jens hjorth for valuable comments on a draft version of this paper .
we also thank the anonymous referee for comments and suggestions that improved the presentation of our results .
kp acknowledges support from the danish national research council , the carlsberg foundation , and instrument center for danish astrophysics .
the dark cosmology centre is funded by the danish national research foundation .
hd acknowledges support from the research council of norway and travel support from nordita . | the main uncertainty in current determinations of the power spectrum normalization , @xmath0 , from abundances of x - ray luminous galaxy clusters arises from the calibration of the mass temperature relation .
we use our weak lensing mass determinations of 30 clusters from the hitherto largest sample of clusters with lensing masses , combined with x - ray temperature data from the literature , to calibrate the normalization of this relation at a temperature of 8 kev , @xmath1 .
this normalization is consistent with previous lensing - based results based on smaller cluster samples , and with some predictions from numerical simulations , but higher than most normalizations based on x - ray derived cluster masses .
assuming the theoretically expected slope @xmath2 of the mass temperature relation , we derive @xmath3 for a spatially - flat @xmath4cdm universe with @xmath5 .
the main systematic errors on the lensing masses result from extrapolating the cluster masses beyond the field - of - view used for the gravitational lensing measurements , and from the separation of cluster / background galaxies , contributing each with a scatter of 20% . taking this into account ,
there is still significant intrinsic scatter in the mass temperature relation indicating that this relation may not be very tight , at least at the high mass end .
furthermore , we find that dynamically relaxed clusters are @xmath6 hotter than non - relaxed clusters . |
first j102347.6 + 003841 ( hereafter j1023 ) was discovered by bond et al .
( 2002 ) in the radio band and initially classified as a magnetic cataclysmic variable .
optical studies revealed signs for the presence of an accretion disc in 2001 ( szkody et al . 2003 ; wang et al .
2009a ) , which led to identify this system as a neutron star low - mass x - ray binary ( lmxb ; thorstensen & armstrong 2005 ; homer et al .
no enhanced x - ray emission was reported .
the discovery of a 1.69-ms radio pulsar , psr j1023 + 0038 , in a 4.75-h binary system ( archibald et al .
2009 ) made j1023 the first system showing the potential to alternate its radio pulsar phase , powered by rotation , to an x - ray state , powered by accretion and typical of lmxbs .
the recycling scenario of neutron stars in lmxbs in fact does not necessarily proceed as a stable phase : a large majority of lmxbs are indeed transients and when the neutron star is already spinning at the ms level , accretion episodes ( outbursts ) can alternate to radio ejection phases in the same system ( campana et al .
1998 ; burderi et al .
this picture has been successfully confirmed by igr j182452452 in the globular cluster m28 , which presented itself as a millisecond radio pulsar , an accreting millisecond x - ray pulsar and again as a millisecond radio pulsar on a time - scale of a few weeks ( papitto et al .
a similar system is the lmxb xss j122704859 : radio , optical and x - ray observations suggested that this system has recently switched to a millisecond radio pulsar phase ( bassa et al .
2014 ; bogdanov et al .
the detection of 1.69-ms radio pulsations provided compelling evidence of this state change ( roy et al .
2014 ) . since its discovery in 2007 june
, j1023 has been regularly monitored with several radio facilities . a long - term radio monitoring campaign from mid-2008 to
mid-2012 revealed eclipses , interpreted as the result of magnetic activity of the companion ( archibald et al .
j1023 was detected with the westerbork synthesis radio telescope on 2013 june 15 , but it was not detected by the lovell telescope at jodrell bank eight days later ( stappers et al .
deeper observations at the same or higher frequencies confirmed this disappearance ( stappers et al .
2014 ) . during this period ( 2013 july
august ) , j1023 remained below the sensitivity threshold of the burst alert telescope hard x - ray monitoring instrument on board _
swift_. this results in a 1550 kev upper limit for the luminosity ( for a source distance of 1368 pc ; deller et al .
2012 ) of @xmath0 ( stappers et al . 2014 ) and rules out a major accretion outburst from this system .
j1023 remained undetected in the radio also during an observation in 2013 october ( patruno et al . 2014 ) .
as many other millisecond radio pulsars , j1023 is a known gamma - ray emitter ( tam et al .
the _ fermi _ satellite revealed an increase in the gev flux ( 100 mev300 gev ) by a factor of @xmath1 compared to the long - term average during the periods 2013 jun 21aug 17 and aug 17sep 15 , respectively ( stappers et al .
2014 ) . despite some variability at time - scales of a few days to a few weeks
, the gamma - ray flux remained at a high level for most of the time until , at least , 2013 november ( takata et al .
2014 ) . in the x - ray band
, j1023 has been observed several times in 2013 by _ swift_/xrt .
the system appears to be much brighter on october 18 than on june 10 and 12 , reaching a 0.310 kev luminosity of @xmath2 ( kong 2013 ; patruno et al .
2014 ) , a factor @xmath3 brighter than its quiescent level ( @xmath4 ; archibald et al . 2010 ; bogdanov et al . 2011 ) .
moreover , evidence for aperiodic variability is found on short time - scales , down to @xmath5 s. _ nustar _ observations also revealed an increase in the 379 kev luminosity of the system , from @xmath6 between june 1012 to @xmath7 and up to @xmath8 between october 1921 ( tendulkar et al .
2014 ) . in the optical , the system was @xmath9 mag brighter on october 16 - 23 than in june and somewhat bluer ( halpern et al .
this is confirmed by uv observations gathered with _
swift_/uvot , since an increase by @xmath10 mag is observed in the uvw1 filter ( patruno et al .
optical spectroscopy performed between 2013 october 22 and december 16 shows the presence of several broad emission lines such as h@xmath11 , h@xmath12 @xmath135876 , and h@xmath12 @xmath136678 ( halpern et al .
2013 ; linares et al . 2014 ;
takata et al . 2014 ) .
the h@xmath11 line is strong with an equivalent width of @xmath14 and double - peaked , indicating the presence of an accretion disc .
these values are similar to those observed during the 2001 optically active phase ( wang et al .
2009a ) . in this paper
, we confirm and improve this picture with new observational data .
we report on a much larger number of _ swift _ observations in the x - ray and uv / optical bands in section 2 .
we focus on new optical and infrared photometry of this system in section 3 .
we report the results of new optical spectroscopy in section 4 .
we derive a strictly contemporaneous spectral energy distribution ( sed ) in section 5 .
we put forward a theoretical interpretation based on a completely enshrouded radio pulsar in section 6 .
conclusions are drawn in section 7 .
the _ swift _ satellite ( gehrels et al . 2004 ) observed j1023 51 times between 2013 october 18 and 2014 may 12 .
the observations were carried out both in the soft x - ray ( 0.310 kev ) and in the uv / optical bands with the xrt and the uvot on board _ swift _ , respectively .
we report the main characteristics of the observations in table [ log ] .
the first 10 _ swift _ observations were already analysed by takata et al .
( 2014 ) .
[ cols="<,^,^,^,^ " , ] [ tab_ew ] radial velocities were measured from the four san pedro mrtir optical spectra through cross - correlation with a g5 v template star .
prior to cross - correlating , the spectra were barycentric corrected and rebinned to a uniform velocity scale .
cross - correlation was performed in the range 5130 - 6300 , after masking the h@xmath12 @xmath15 emission line from the accretion disc .
a fit with a constant plus sinusoid function provides a systemic velocity @xmath16 km s@xmath17 and a radial velocity @xmath18 km s@xmath17 ( see fig .
[ radvel ] ) .
although the large uncertainties ( mainly due to the low statistics ) , these values are fully consistent with the results obtained by thorstensen & armstrong ( 2005 ) during the system quiescence .
j1023 was observed strictly simultaneously by _ swift _ and rem for @xmath19 min on 2013 november 30 , while it was at orbital phase 0.9 ( near the neutron star superior conjunction ) .
we extracted the corresponding source and background x - ray spectra adopting as extraction region a circle centred on the source with a radius of 15 pixels in the former case and a circle positioned in a location free from known x - ray sources with a radius of 30 pixels in the latter .
we created the ancillary response file for the extracted spectrum with xrtmkarf , in order to correct the count rate for the presence of bad pixels , vignetting and hot columns .
finally , we assigned the latest version of the redistribution files .
fluxes in the optical and infrared bands were estimated through the aperture photometry technique .
the unabsorbed sed from the nir to the x - ray bands is shown in fig .
[ sed ] .
we attempted to account for the nir , optical , uv and x - ray sed with the simple model of an irradiated star plus the contribution of a shock front . for the star component we assumed an irradiated blackbody model ( for the details of the modelling see eqs .
[ 8]-[9 ] of chakrabarty 1998 ) which depends on the irradiating luminosity ( @xmath20 ) , the source distance ( @xmath21 ) , the radius of the companion star ( @xmath22 ) , the albedo of the star ( @xmath23 ) and the binary separation ( @xmath24 ) .
we fitted the data by using @xmath20 as a free parameter and fixing @xmath25 pc ( deller et al .
2012 ) , @xmath26 ( archibald et al . 2009 ) and @xmath27 .
the binary separation @xmath28^{1/3}$ ] was obtained by fixing @xmath29 and @xmath30 ( deller et al .
we modelled the shock front component with a simple power law , by leaving free to vary the normalization and the power law index .
the best fit is obtained for @xmath31 erg s@xmath17 , consistent with the estimated dipole spin - down luminosity of the ms radio pulsar ( @xmath32 \times 10^{34}{\rm \ erg \ ; s^{-1}}$ ] ; archibald et al .
however , such model provides a rather poor fit to the data , particularly in the uv region ( @xmath33 for 10 d.o.f . ; see fig .
[ sed ] , left - hand panel ) . to check for a more realistic solution , we tried to fit our data by adding to the model the contribution of an irradiated accretion disc by using @xmath20 and the internal disc radius ( @xmath34 ) as free parameters ( eqs .
[ 10]-[15 ] of chakrabarty 1998 ) .
we fixed the x - ray albedo of the disc to 0.95 ( chakrabarty 1998 ) and the mass transfer rate to 10@xmath35 yr@xmath17 , as predicted by verbunt ( 1993 ) for a short - period x - ray binary with a main - sequence companion star and where the mass transfer is kept going by loss of angular momentum from the system .
for each possible radius @xmath34 , we assumed an outer disc radius of @xmath36 ( where @xmath24 is the binary separation ) .
the fit improves ( @xmath37 with respect to the star plus shock model ; see fig . [
sed ] , right - hand panel ) and we obtain an acceptable solution for @xmath38 erg s@xmath17 and an inner radius for the emitting region @xmath39 cm .
more sophisticated models are beyond the scope of the paper .
millisecond radio pulsars in binary systems can interact with the companion star if the orbital separation is small enough . in recent years
an increasing number of the so - called spider radio pulsars have been discovered ( see roberts 2011 for a review ) , following the first discoveries of the redback psr j1740 - 5340 in the globular cluster ngc 6397 ( with a non - degenerate companion , damico et al .
2001 ) and the black widow psr b1957 + 20 ( with a brown dwarf companion , reynolds et al .
the interaction between the relativistic pulsar wind and matter outflowing from the companion can give rise to different geometries depending on the relative strengths of the two components .
in the case of a strong relativistic wind ( high pressure ) , matter outflowing from the companion is pushed in a narrow cometary tail around it , as in the case of psr b1957 + 20 . in the case of a weaker relativistic pulsar wind ( or strong matter pressure )
more extended patterns can be obtained , giving rise to extensive eclipses of the radio signal during the orbital period .
a further case has been envisaged by tavani ( 1991 ) , where a large amount of matter is outflowing from the companion , completely engulfing the radio pulsar . in this regime
the radio pulsar is still active , but its signal is completely undetectable in the radio band due to the high free free absorption .
arons & tavani ( 1993 ) developed the theory of high - energy emission by the relativistic shock produced by the pulsar wind in the nebula surrounding the binary and by the shock constraining the mass outflow from the companion star .
particularly compelling to j1023 is the case of psr b125963 .
this is a millisecond pulsar orbiting a high mass be companion in a very eccentric orbit .
far from the companion , the x - ray emission is dominated by shock - powered high - energy emission produced by the interaction between the relativistic wind from the pulsar and matter outflowing from the companion .
the spectrum is described by a hard power law with @xmath40 , extending up to 200 kev ( and more ) as detected with compton / osse ( grove et al .
1995 ) . during its orbital evolution
, the pulsar disappears close to periastron . during the passage
an increase in luminosity and a softening of the power law @xmath41 are observed ( tavani & arons 1997 ) . a change to a propeller regime has been excluded ( campana et al .
1995 ) .
three characteristic radii define the fate of matter falling on to a magnetized , fast spinning neutron star : the magnetospheric radius , @xmath42 ( where the incoming matter pressure balances the magnetic dipole pressure ) , the corotation radius , @xmath43 ( where matter in keplerian orbit corotates with the neutron star ) , and the light cylinder radius , @xmath44 ( where field lines attached to the neutron star rotate at the speed of light ) .
the last two radii depend only on the neutron star spin and in the case of j1023 their values are @xmath45 km ( for a neutron star mass of @xmath46 ) and @xmath47 km .
the magnetospheric radius depends on the neutron star magnetic field and on the mass inflow rate .
it can be expressed as @xmath48 where @xmath49 is the mass accretion rate on to the magnetosphere in units of @xmath50 g s@xmath17 , @xmath51 and @xmath52 are the neutron star mass and radius in units of @xmath46 and 10 km , respectively , and @xmath53 is the magnetic field normalized to @xmath54 g ( perna , bozzo & stella 2006 ) .
due to the inclination of the magnetic moment with respect to the rotation axis , the magnetospheric radius can be larger up to a factor 1.5 ( perna et al .
2006 ) .
the magnetic field of j1023 has been estimated to be @xmath55 g ( archibald et al . 2013 ) .
depending on the mass inflow rate , matter can reach the neutron star surface ( @xmath56 , accretion powered ) , be halted or strongly reduced at the neutron star magnetosphere by the propeller mechanism ( @xmath57 , still accretion powered ) or be ejected by the pulsar pressure ( @xmath58 , spin - down powered ) .
these regimes occur for lower and lower mass accretion rates .
when a radio pulsar reactivates it is more difficult to quench it .
this is simply due to the fact that the mass inflow pressure has overcome the radiation pulsar pressure throughout the binary separation ( campana et al .
1998 ; burderi et al .
this is why it is so difficult to quench a fast spinning millisecond radio pulsar when it reactivates .
it is not easy to account for the minimal luminosity needed to quench j1023 .
following burderi et al .
( 2001 ) , we can estimate a quenching luminosity of @xmath59 .
this is much larger than the observed x - ray luminosities and would suggest that the radio pulsar is still active , but unobservable .
we note that a propeller model has been proposed to explain the emission characteristics of xss j12270 - 4859 ( papitto , torres & li 2014 ) and such model may apply to j1023 in its current state as well . in this model
the power law emission is interpreted as resulting from synchrotron emission at the interface between the disc and the magnetosphere .
based on energetic grounds , this model is slightly disfavoured by the data , despite the appealing feature of easily explaining the presence of a ( truncated ) accretion disc .
the minimum luminosity expected in the propeller regime is obtained when the magnetospheric radius is close to the light cylinder radius . in the case of j1023 ,
this luminosity is @xmath60 erg s@xmath17 , whereas the maximum luminosity expected in the propeller regime is @xmath61 erg s@xmath17 .
the observed mean x - ray luminosity is @xmath62 erg s@xmath17 and during the flare it rises to @xmath63 erg s@xmath17 .
by contrast the full spin - down luminosity is@xmath64 erg s@xmath17 based on _ nustar _ data , tendulkar et al .
( 2014 ) estimated an average value for the x - ray luminosity which is close to the minimum luminosity expected in the propeller regime , whereas the peak luminosity should be in the propeller luminosity interval .
in this work we presented the results of 51 _ swift _ xrt and uvot observations of the binary millisecond radio pulsar psr j1023 + 0038 carried out in the 2013 october2014 may period and the results of optical and nir photometry and optical spectroscopy of this system .
the x - ray spectrum of j1023 is best modelled by an absorbed power law , with a column density of @xmath65 @xmath66 and a photon index @xmath67 .
the contribution from any possible soft component is negligible and no iron line features can be detected .
spectral changes can be accounted for by changes of the power law normalization alone .
these observational features are consistent with shock - powered emission produced by the pulsar / outflow interaction ( tavani & arons 1997 ) and with the enshrouding of j1023 by a large amount of matter , material which is however not enough to quench the radio pulsar .
the enshrouding is testified by the mild increase in the column density and by the softening of the power law photon index compared to the previous quiescent period ( archibald et al .
2010 ; bogdanov et al .
2011 ) : the former rose from * @xmath68 * to @xmath69 @xmath66 , the latter changed from @xmath70 to @xmath67 ( tavani & arons 1997 ) . at the end of 2014 february
, x - ray flares were detected , with a 0.310 kev luminosity of @xmath71 ( a factor @xmath72 higher than the average x - ray luminosity ) .
the peak luminosity of j1023 during flaring is large , encompassing a sizeable fraction of the spin - down luminosity and thus leaving open the possibility for emission in the propeller state ( papitto et al .
2014 ) .
the shock - powered emission scenario can be well reconciled with both the reported switch off in the radio band and the increase in the gamma - ray emission . according to this ,
the radio pulsar signal is completely shielded by scattering and/or free free absorption by a large amount of outflowing matter , whereas the increase in the gamma
ray emission can be naturally explained in terms of an enhancement of the shock emission ( archibald et al .
2013 ; patruno et al .
2014 ; stappers et al . 2014 ;
takata et al . 2014 ; tendulkar et al . 2014 ) .
a strong correlation is found between the x - ray and the uv count rates in the 2013 november 6 - 19 period both on time - scales of days and of a few tens of seconds , thus suggesting that the same emission mechanism is powering part of the x - ray and uv emission .
optical and infrared photometric observations show that the companion star is irradiated by the spin - down emission of the radio pulsar , but also suggest that other emission mechanisms must be at work .
in fact , double - peaked emission lines in the optical spectra indicate that an accretion disc exists around the system , as shown also from multi - wavelength observations perfomed since 2013 june ( patruno et al . 2014 ; stappers et al .
2014 ; takata et al . 2014 ; tendulkar et al . 2014 ) .
the peak separation of the horns of the emission lines is @xmath73 km s@xmath17 .
this value is consistent with those reported by halpern et al .
( 2013 ) , takata et al .
( 2014 ) and linares et al .
( 2014 ) , suggesting that the disc thickness has not changed significantly between 2013 october and december .
the sed from the nir to the x - ray band is well represented by a model consisting of an irradiated companion , an accretion disc and a shock emission .
the sed reveals a minimal contribution of the disc to the x - ray emission and may indicate that accretion on to the neutron star is not occurring ( see also the sed reported by takata et al .
in fact , this is what is expected in order not to completely quench the pulsar mechanism and thus the relativistic wind .
the sed also shows that more than half of the uv emission ( @xmath74 per cent fractional contribution to the total flux ) is emitted from the accretion disc , with the remaining part arising from the intra - binary shock ( @xmath75 per cent ) .
the modelling of the sed allowed us to estimate the inner radius of the emitting region ( @xmath76 @xmath77 ) , consistent with the value reported by takata et al .
( 2014 ) for the inner edge of the accretion disc , and the spin - down luminosity of the radio pulsar ( @xmath38 erg s@xmath17 ) , possibly involving some shielding .
if this picture is correct , the disc should not extend down to the magnetosphere , but it should be halted further outside .
the disc should then be partially supported or strongly evaporated by the neutron star spin - down luminosity , as it has been suggested for some white dwarfs ( meyer & meyer - hofmeister 1994 ) .
future multi - wavelength monitoring campaigns may shed further light on this and , more in general , on the phenomenology of the system emission .
we thank the referee for useful comments . we thank alessandro patruno and alessandro papitto for productive discussions on x - ray variability and on radio pulsar - dominated and accretion - dominated regimes .
sc thanks neil gehrels for approving _
swift _ observations of j1023.we thank san pedro mrtir staff for all their help during the observation run .
we thank silvia galleti for service mode observations at the loiano telescope , and antonio de blasi for night assistance .
tmd acknowledges funding via an eu marie curie intra - european fellowship under contract no .
2011 - 301355 .
ejb and hof acknowledge conacyt grant 129204 .
hof is funded by a postdoctoral unam grant .
fgrf is funded by a conacyt grant for postgraduate studies .
this work made use of data supplied by the uk _ swift _ science data centre at the university of leicester and was funded in part by european research council advanced grant 267697 ` 4@xmath78 sky ' : extreme astrophysics with revolutionary radio telescopes .
roberts m. s. e. , 2011 , in aip conf .
1357 , radio pulsars : an astrophysical key to unlock the secrets of the universe , ed .
m. burgay , n. damico , p. esposito , a. pellizzoni & a. possenti- am .
melville , ny , p. 127 | the binary millisecond radio pulsar psr j1023 + 0038 has been recently the subject of multi - wavelength monitoring campaigns which revealed that an accretion disc has formed around the neutron star ( since 2013 june ) .
we present here the results of x - ray and uv observations carried out by the _ swift _ satellite between 2013 october and 2014 may , and of optical and nir observations performed with the rem telescope , the liverpool telescope , the 2.1-m telescope at the san pedro mrtir observatory and the 1.52-m telescope at the loiano observing station .
the x - ray spectrum is well described by an absorbed power law , which is softer than the previous quiescent epoch ( up to 2013 june ) .
the strong correlation between the x - ray and the uv emissions indicates that the same mechanism should be responsible for part of the emission in these bands .
optical and infrared photometric observations show that the companion star is strongly irradiated .
double - peaked emission lines in the optical spectra provide compelling evidence for the presence of an outer accretion disc too . the spectral energy distribution from ir to x - rays is well modelled if the contributions from the companion , the disc and the intra - binary shock emission are all considered .
our extensive data set can be interpreted in terms of an engulfed radio pulsar : the radio pulsar is still active , but undetectable in the radio band due to a large amount of ionized material surrounding the compact object .
x - rays and gamma - rays are produced in an intra - binary shock front between the relativistic pulsar wind and matter from the companion and an outer accretion disc .
the intense spin - down power irradiates the disc and the companion star , accounting for the uv and optical emissions .
[ firstpage ] stars : individual : psr j1023 + 0038 stars : neutron pulsars : general x - rays : binaries . |
n - body problem is a classical problem in mathematics and celestial mechanics . since euler s collinear solutions and lagrange s equilateral triangle solutions , any more `` clearly '' ( including the existence and the shape of the orbit ) solutions
have not been found except figure - eight solution , a planer three - body orbits with equal masses , which was proved strictly by chenciner and montgomery in @xcite in 1999 .
this remarkable and interesting orbit arouse many authors curiosity in studying more choreography orbits or other interesting symmetric constraints acting on the orbits of n - body problems , see , etc .
furthermore , some investigations of a possible star system in real space have been performed in practice ( see @xcite ) . among those authors studies ,
the variational approach and the theory of critical points perform an important role . by minimizing the lagrangian action on a space of loops symmetry with respect to a well - chosen symmetry group , which also has some simple conditions on it ,
ferrario and terracini give a fairly general condition on symmetry groups of the loop space in @xcite .
but their results `` are not suitable to prove the existence of those orbits that are found as minimizers in classes of paths characterized by homotopy conditions or a mixture of symmetry and homotopy conditions '' .
a further study based on @xcite is carried by shibayama to distinguish the solutions , which have been obtained by shibayama in @xcite and some of which by chen in @xcite . in this paper
, we study the figure - eight constraint carefully in section [ se3 ] , and draw a conclusion that even bodies with equivariant shift on figure - eight will collide at least on the original point , and odd bodies will not collide in a dense subset of @xmath0 $ ] .
then in section [ se4 ] , we compose figure - eight from simple to double ( or more , that we do not describe in this paper ) by setting odd bodies on them reasonably . and in section [ se5 ] , it is practical to mix figure - eight with other constraints on @xmath1 in space . then along gordon and zelati s @xcite line , in section [ se6 ] ,
we have proved the existence of generalized solutions and non - linear solutions for these periodic orbits .
the numerical study in section [ se4 ] highly enlightens us that we can arrange many ( even or odd ) bodies on figure - eight by choosing an appropriate _
phase differences_. obviously , all the study is based on figure - eight constraint , hence this paper is a continued study on chenciner and montgomery s significant work .
the periodic solutions for n - body problem is described by the following second order ordinary differential equations in general : @xmath2 @xmath3 where @xmath4 is the mass of @xmath5th body , @xmath6 is the position of @xmath5th body in @xmath7 , @xmath8 is the number of bodies , @xmath9 is the period , and generally the singular potential function @xmath10 is @xmath11 with assumptions : @xmath12 and @xmath13 , @xmath14 @xmath15 where @xmath16 is usually described by @xmath17 when @xmath18 in ( [ uil1 ] ) , it is called _ newton _ n - body problem . the figures , as the numerical examples to illuminate the existence of the solutions in this paper , are all coming from newton n - body problem .
poincar has shown that it is impossible for @xmath19 to find an explicit expression for the general solution .
so even for the three - body problem , it is impossible to describe all the solutions .
then the numerical approach and/or endowing orbits with some symmetric constraints are quite reasonable directions for the study in n - body problem .
no matter which direction adopted by the authors it is , the lagrangian @xmath20 @xmath21 acts an important role in their study in finding the periodic solutions for n - body problems . in one direction , finding some symmetric orbits , chenciner and montgomery have introduced a new `` torsional '' symmetric constraint to the possible orbits for three - body problem with equal masses in 1999 .
the `` torsional '' symmetry is described by the following expressions : let @xmath22 , @xmath23 , the orbit of the @xmath5-th body is @xmath24 @xmath25,\ ] ] @xmath26 where @xmath27 , @xmath28 and @xmath29 are generators of the action of klein group @xmath30 on @xmath31 and on @xmath32 : @xmath33 @xmath34 by the constraints ( [ e1])-([e3 ] ) , the existence is proved in @xcite for three equal masses with the exclusion of collision .
it is called simple choreography orbits , figure - eight solution . after that , many choreography / choreographies solutions are found numerically by many authors , e.g. sim , arioli , barutello and terracini @xcite etc .
one of the facile development to simple choreography is considering more bodies on the same figure - eight .
we can develop ( [ e1 ] ) and ( [ e2 ] ) by @xmath35 @xmath36.\ ] ] hence , ( [ e4 ] ) , ( [ e5 ] ) and ( [ e3 ] ) describe @xmath8 bodies on the same figure - eight with equivalent shift @xmath37 . by ( [ e3 ] )
, one can deduce that @xmath38 @xmath39 when we set @xmath40 , @xmath41 it shows that @xmath42 is an odd function .
let @xmath43 , one can deduce the following formulas from ( [ q1 ] ) and ( [ q2 ] ) , @xmath44 @xmath45 that is @xmath46 this means the original point @xmath47 is one of the points of _ self - intersection_s of @xmath42 . from the expression in ( [ e4 ] ) , one has @xmath48
@xmath49 now , in the case of even bodies on figure - eight , i.e. @xmath50 , @xmath51 , the following is from ( [ q3 ] ) , @xmath52 for all @xmath53 , we consider @xmath54 we notice this in ( [ q1 ] ) @xmath55 together ( [ qi ] ) with ( [ qik ] ) , it shows that when @xmath56 , @xmath57 that is the @xmath5-th and @xmath58-th bodies will collide at the time @xmath56 . with the assumption of ( [ e2]),([e4 ] ) and ( [ e5 ] ) , even bodies on figure - eight orbit
will collide on original point . since the constraint is more free , the shape of the possible orbit is more complex .
it is reasonable to make a further strict assumption , a simple " orbits constraint , on figure - eight .
we set the further _ simple _ figure - eight assumption : @xmath59 it means that the choreography orbit of @xmath60 has one and only one _ self - intersection _ on figure - eight .
now we consider the case for the odd bodies on figure - eight .
let @xmath61 , @xmath62 and suppose that @xmath6 and @xmath63 collide at @xmath64 , then @xmath65 according to assumption ( [ e7 ] ) , one has @xmath66 without losing generality , set @xmath67 which means @xmath68 since @xmath69 , the expression above is contradictory
. then we have with the assumption ( [ e2]),([e4 ] ) , ( [ e5 ] ) and ( [ e7 ] ) , odd bodies will not collide on _ simple _ figure - eight orbits . it is obvious that assumption ( [ e7 ] ) can be generalized by @xmath70 where @xmath71 [ th1 ] there is a dense subset in @xmath0 $ ] , such that odd bodies on figure - eight without collision in it .
* proof : * one can easily verify that @xmath72 is dense in @xmath0 $ ] , i.e. , for @xmath73 $ ] and @xmath74 , there exists @xmath75 , such that @xmath76 .
then @xmath72 is at least one of the dense subset of @xmath0 $ ] we need .
this completes the proof .
if we could make a further generalization for assumption ( [ e71 ] ) by @xmath77,\quad i=1,2,\ ] ] we would have gotten a _
perfect _ result .
but we can not get the generalization now .
it means the figure - eight perhaps has some self - intersection points on time @xmath64 , which is not in @xmath72 , e.g. @xmath64 is an irrational point .
it is obvious , when we record @xmath78 , a loop space , we may have the same result by modifying @xmath79 * figure [ eight13 ] * in section [ se7 ] is one of our numerical examples for odd bodies on figure - eight .
we consider some two choreographies orbits , the double - eight orbits , which arises from figure - eight . with this idea , one can also figure out other more choreographies orbits .
we describe the following double - eight orbits as an example for n - body problem . based on the aforementioned results , we set @xmath80 and @xmath81 are all odd in @xmath82 , and the number of bodies is @xmath83
. then we make a two - choreographies assumption : one figure - eight reclines on @xmath84-axis , the other on @xmath85-axis .
that is , one figure - eight @xmath42 satisfies @xmath86 @xmath87 @xmath88 the other figure - eight @xmath89 satisfies @xmath90 @xmath91 @xmath92 where @xmath93 and @xmath94 can be regarded as the _ phase difference_s after equivariant shifts . in order to guarantee that any two bodies in two figure - eight separately will not collide on the origin point , specially , we set @xmath95 , @xmath96 , where @xmath97 is prime to @xmath98
. when we set @xmath99 not satisfy ( [ q3 ] ) but ( [ d3 ] ) , we will get the orbits with the constraint such that two figure - eights recline on the same @xmath84-axis ( or @xmath85-axis ) .
the * figures [ f3_3e01]*-*[l3_5n05 ] * in section [ se7 ] , are the numerical examples for double - eight orbits . to construct some mixed - symmetric constraints on simple choreography or two choreographies for @xmath101 , we set @xmath102 and @xmath103 satisfy ( [ d1])-([d3 ] ) on @xmath104 plane , and in @xmath105-coordinate , @xmath1 , satisfy one of the following constraints : * symmetry 1 ( simple choreography orbit ) @xmath106.\ ] ] * symmetry 2 ( simple choreography orbit ) @xmath107,\ ] ] and @xmath108.\ ] ] * symmetry 3 ( two choreographies orbits ) + let @xmath109 be a simple choreography and @xmath110 another one , we set 1 .
@xmath60 and @xmath99 locate on the same figure - eight on @xmath104 plane ; 2 .
@xmath1 satisfy ( [ con1 ] ) and ( [ con2 ] ) ; 3 .
@xmath111 .
see * figures [ r3x11e01]-[m8ccl08e02 ] * for numerical examples . from the discussion in section [ se3 ] and section [ se4 ]
, we can choose many @xmath112 , @xmath113 , @xmath114 , such that @xmath115 groups of odd bodies can be arranged on figure - eight without any collision in a dense subset of @xmath0 $ ] .
[ th3 ] for all @xmath19 , @xmath116 , @xmath8 bodies can be arranged on figure - eight without any collision in a dense subset of @xmath0 $ ] .
* proof : * if @xmath19 is odd , by theorem [ th1 ] , the proof is as follows .
if @xmath8 is even , it can be decomposed by two odds @xmath98 and @xmath117 , such that @xmath83 .
then we set @xmath118 @xmath119 @xmath120 @xmath121 @xmath122 @xmath123 where @xmath97 is prime to @xmath98 . through the discussion in the counterpart of section [ se4 ]
, one can easily verify that the theorem follows .
* figures [ even06e01]-[even12e01 ] * are numerical examples for theorem [ th3 ] .
in this section , we mainly consider the existence of generalized solutions and non - collision solutions for mixed - symmetry and double - eight . at
first , we assemble some well - known facts from @xcite , @xcite , @xcite and @xcite about sobolev space @xmath124 and the ordinary space @xmath125 , ( \mathbb{r}^{d})^{n})$ ] .
* let @xmath126 , ( \mathbb{r}^{d})^{n})$ ] , then the inner product can be defined by @xmath127 and the corresponding norm is @xmath128 .
* let @xmath129 , then the inner product can be defined by @xmath130 and the corresponding norm is @xmath131 .
* by sobolev imbedding theorems , weak convergence in @xmath132 implies uniform convergence , i.e. , the weak @xmath133 topology is stronger than the @xmath134 topology . *
let @xmath135 be a functional from @xmath133 to @xmath136 .
a sequence @xmath137 is called a _ minimizing sequence _ , if @xmath138 * a functional @xmath139 is _ weakly lower semi - continuous _ if @xmath140 * a weakly lower semi - continuous functional @xmath135 on @xmath133 has a minimum in @xmath133 , if @xmath135 has a bounded minimizing sequence .
since it is difficult to exclude the collision without any more additional constraints for the orbits , we introduce the following two definitions , both of which partially exclude the collision in some sense .
we denote @xmath141 \ | \ q_{i}(t)=q_{j}(t),\ \mbox{for some}\ i \neq j \right\}.\ ] ] we call @xmath142 , ( \mathbb{r}^{d})^{n})$ ] a _ non - collision solution _ of ( [ n ] ) and ( [ p ] ) , if @xmath143 .
[ defi2 ] we call @xmath144 a _ generalized solution _ of ( [ n ] ) and ( [ p ] ) , if \a ) @xmath145 ; \b ) @xmath60 satisfy ( [ n ] ) ; \c ) @xmath146 where @xmath147\backslash \delta$ ] and @xmath148 is a constant . in order to overcome the singularity of @xmath10
, the following condition is usually used as an efficient method in considering the existence of minimum of functional @xmath135 .
let some @xmath149 , @xmath150 ( @xmath151 ) and @xmath152 where @xmath153 is small .
then we say @xmath16 satisfies _ strong force ( sf ) condition_. for the individuality of the mixed - symmetric ( the same as double - eight ) solutions , we have to consider them respectively .
but in the following theorem , we describe our result at one time for all the symmetric constraints , which have been mentioned therein .
[ th2 ] there are infinitely many generalized solutions for ( [ n])-([p ] ) with anyone of those symmetric constraints . furthermore
, if @xmath10 satisfies ( sf ) condition , there will be infinitely many non - collision solutions .
the proof of theorem [ th2 ] is extremely along the line from gordon and zelati @xcite .
the key points locate on the estimation of @xmath154 and the construction of those nested intervals to cover @xmath0\backslash\delta$ ] . in the ongoing proof ,
the existence of periodic solutions with symmetry 1 is proved as a candidate of those mixed - symmetries and double - eight .
we set @xmath155 @xmath156 @xmath157 it s easy to verify that the critical points of @xmath135 on @xmath158 are non - collision solutions of ( [ n])-([p ] ) .
and one can easily check that the critical points of @xmath159 are actually those of @xmath135 on @xmath158 . * step1 : * if @xmath16 does not satisfy ( sf ) condition , we can modify it by the following formula @xmath160 where @xmath161 is small and @xmath162 then the modified @xmath163 satisfies ( sf ) condition , and the corresponding lagrangian functional is @xmath164 where @xmath165 * step2 : * in this step
, we will show the existence of a minimum of @xmath135 .
let the infimum of @xmath166 be @xmath167 consider a minimizing sequence @xmath168 such that @xmath169 when @xmath170 .
then , when @xmath171 is large enough , we have @xmath172 hence for one term of @xmath166 , we can deduce that @xmath173 where @xmath148 is a constance and @xmath174 .
we notice that from ( [ q1])-([q2 ] ) , one can deduce the period of @xmath103 is @xmath175 , and @xmath42 is symmetrical about @xmath103-axis , then one has @xmath176 hence we can estimate the norm of @xmath60 , @xmath177 then for @xmath178 , @xmath179 ( sufficiently large ) , the norm for one of the minimizing sequence is @xmath180 this implies the existence of @xmath181 with @xmath182 , for all @xmath183 such that @xmath184 and @xmath185 on condition ( sf ) , it follows that @xmath186 for every sequence @xmath187 such that @xmath188 weakly in @xmath133 and strongly in @xmath134 if @xmath189 .
this proves that @xmath190 is a minimum for @xmath166 on @xmath191 .
this minimum is a non - collision solution of ( [ n])-([p ] ) .
then we have proved that ( [ n])-([p ] ) has at least one non - collision solution if condition ( sf ) is satisfied .
* step3 : * this step is the same as the corresponding part of the proof for theorem1.1 in @xcite , so we only describe the sketch : we can find a generalized solution @xmath192 of ( [ n])-([p ] ) by constructing compact nested interval cover of @xmath0\backslash \delta$ ] to show @xmath193 and @xmath194 when @xmath195 .
this process shows that @xmath192 satisfies definition [ defi2 ] .
we can also prove @xmath196 is a @xmath9-periodic solution of ( [ n])-([p ] ) if @xmath60 is a @xmath9-periodic solution , where @xmath105 is a positive integer , then we can find infinitely many @xmath9-periodic solutions of ( [ n])-([p ] ) .
this completes the proof .
in this section , we along the line in theorem [ th2 ] , finding the critical points of @xmath135 , to list some of numerical examples for those theoretic consideration . in the following figures , @xmath197 is the masses , some of which are divided into two parts with a semicolon to distinguish the masses of @xmath98 bodies and @xmath117 bodies , @xmath93 and @xmath94 as defined above . in those figures with sub - figures ,
the top left sub - figure is the projection of the orbits on @xmath84-@xmath85 plane , the top right is the projection of the orbits on @xmath85-@xmath105 plane , the bottom left is the projection of the orbits on @xmath84-@xmath105 plane , and the bottom right is the orbits in space . during our numerical study
, it shows some interesting phenomena .
the followings are some of them for further consideration : * when we set some other constraints on @xmath1 at will , e.g. the following symmetries 4 - 6 , we have gotten some interesting figures numerically ( see figure [ r3x12e01]-[r3x11e03 ] ) . *
* symmetry 4 @xmath198 @xmath199.\ ] ] * * symmetry 5 @xmath200.\ ] ] * * symmetry 6 @xmath201,\ ] ] @xmath202 + see * figures [ r3x12e01]-[r3x11e03 ] * for numerical examples . * in figure - eight
, we can arrange more bodies on it , but we can not find the `` 8 '' including more self - intersections , because the dense subset @xmath72 is so dense that we can not separate those points from others by a modern computer . * in double - eight ,
the shape of those two `` 8 '' depends on the proportion of the masses .
the orbit with less masses looks more slim , and the bodies on it shuttle rapidly across the interspace of the bodies on the other orbit . * in double - eight , the shape of the orbits also depends on the phase difference and the initial values of functional @xmath135 . * in symmetry 1 , 4 , 5 and 6 , there are two of the projections of the orbits become slimmer and slimmer .
it seems that one of the constraints acting on three projections is `` stronger '' than the other two . * in symmetry 2 and 3 , theorem [ th3 ] guarantees that we can arrange many bodies on those orbits , no matter even or odd . ; @xmath203.,width=192 ] ; @xmath203.,width=192 ] ; @xmath203.,width=192 ] ; @xmath204.,width=192 ] ; @xmath204.,width=192 ] ; @xmath204.,width=192 ] ; @xmath205.,width=192 ] ; @xmath205.,width=192 ] ; @xmath205.,width=192 ] ; @xmath205.,width=192 ] ; @xmath205.,width=192 ] ; @xmath205.,width=192 ] ; @xmath206.,width=192 ] ; @xmath206.,width=192 ] ; @xmath206.,width=192 ] ; @xmath204.,width=192 ] ; @xmath204.,width=192 ] ; @xmath204.,width=192 ] ; @xmath206.,width=192 ] ; @xmath206.,width=192 ] ; @xmath206.,width=192 ] with different scale.,width=288 ] with different scale.,width=288 ] .$],width=288 ] .$],width=288 ] , \alpha_{1}=0 , \alpha_{2}=\frac{t}{8}$].,width=288 ] , \alpha_{1}=0 , \alpha_{2}=\frac{t}{8}$].,width=288 ] , \alpha_{1}=0 , \alpha_{2}=\frac{t}{12}$],width=288 ] , \alpha_{1}=0 , \alpha_{2}=\frac{t}{12}$],width=288 ] the following figures are numerical examples for theorem [ th3 ] : , \newline \alpha_{1}=0 , \alpha_{2}=\frac{t}{36}$].,width=192 ] , \newline \alpha_{1}=0 , \alpha_{2}=\frac{t}{36}$].,width=192 ] , \newline \alpha_{1}=0 , \alpha_{2}=\frac{t}{36}$].,width=192 ]
_ for double - eight _ : @xmath208 for @xmath89 in double - eight , we have the same estimation . :
since ( [ con2 ] ) is the same condition as in symmetry 1 , we can at least draw the same conclusion as in symmetry 1 , i.e. @xmath209 : we set @xmath210 , it is easy to verify that @xmath211 | in this paper , we first describe how we can arrange any bodies on figure - eight without collision in a dense subset of @xmath0 $ ] after showing that the self - intersections of figure - eight will not happen in this subset .
then it is reasonable for us to consider the existence of generalized solutions and non - collision solutions with mixed - symmetries or with double - eight constraints , arising from figure - eight , for n - body problem .
all of the orbits we found numerically in section [ se7 ] have not been obtained by other authors as far as we know . to prove the existence of these new periodic solutions , the variational approach and critical point theory
are applied to the classical n - body equations . and along the line used in this paper
, one can construct other symmetric constraints on n - body problems and prove the existence of periodic solutions for them .
msc - class : 70f10 , 37c80 , 70g75 |
the vela pulsar is one of the closest and brightest radio pulsars known .
polarization observations made shortly after its discovery showed that it is highly linearly polarized and that the position angle of the radiation followed an s - shape curve as a function of pulse longitude ( radhakrishnan & cooke 1969 ) .
this showed that the radio emission was related to the geometry of the magnetic field lines near the pole .
the rotating vector model ( rvm ) of radhakrishnan & cooke has been extensively used in a large number of pulsars to obtain the spin axis - magnetic axis and the viewing angle - magnetic axis geometries . in the vela pulsar , the angle between the spin and magnetic axes ,
@xmath0 is estimated to lie between 60 and 90 , and the impact angle , @xmath1 is @xmath2 ( krishnamohan & downs 1983 ( hereafter kd83 ) ; lyne & manchester 1988 ; rankin 1993 ) . in the only detailed single pulse study of vela to have appeared in the ( recent ) literature , kd83 observed 87000 pulses at 2.3 ghz with a time resolution including dispersion smearing of 750@xmath3s .
their main conclusion was that the pulse profile consists of four different components which were emitted at different heights in the magnetosphere .
they also found that the pulse shape depended on the intensity with the strongest pulses arriving early with respect to the average profile . in some pulsars ,
the s - shaped sweep of polarization is broken by one or more jumps of @xmath4 .
these are known as ` orthogonal ' jumps ( backer & rankin 1980 ) .
it is generally believed that emission from two orthogonally polarized modes are superposed ( mckinnon & stinebring 1998 ) . at a given pulse longitude the dominant mode then determines the position angle . in the vela pulsar
, kd83 found that only one mode was present at all pulse longitudes for all 87000 pulses .
sub - pulse structure is ubiquitous in pulsars on a number of different timescales .
so - called microstructure is often seen and can have both a typical width ( generally a few tens of @xmath3s ) and be quasi - periodic in a given pulsar ( e.g. lange et al .
surprisingly , for the vela pulsar no microstructure analysis has ever been published .
the phenomenon of giant pulses has only been detected in two pulsars , the crab pulsar ( e.g. lundgren et al .
1995 ) and psr b1937 + 21 ( cognard et al . 1996 ) .
the working definition of giant pulses is a flux density in a single pulse which is more than 10 times the mean flux density .
both the crab and psr b1937 + 21 show occasional pulses in excess of 1000 times the mean pulsed flux . in the vast majority of pulsars , very few if any single pulse fluxes exceed 10 times the mean flux density .
the crab is one of the youngest known pulsars whereas psr b1937 + 21 is a millisecond pulsar with very rapid rotation rate . as cognard et al .
( 1996 ) point out , the common feature these pulsars share is the highest ( estimated ) magnetic field at their light cylinder , although whether this is related to the physics of the giant pulses is unclear . in this parameter ,
vela ranks 22nd in the current catalogue of 965 pulsars .
as part of a project investigating single pulses from a large sample of pulsars , we observed the vela pulsar at two different observing frequencies between 2000 march 14 and 17 using the 64-m parkes radiotelescope .
the center frequencies of the observations were 660 and 1413 mhz ; at these frequencies the system equivalent flux density is 120 and 26 jy respectively .
each receiver consists of two orthogonal feeds sensitive to linear polarization .
the signals are down - converted and amplified before being passed into the backend .
the backend , cpsr , is an enhanced version of the caltech baseband recorder ( jenet et al .
it consists of an analogue dual - channel down - converter and digitizer card which yields 2-bit quadrature samples at 20 mhz .
the data stream is written to dlt for subsequent off - line processing allowing all 4 stokes parameters to be computed .
at both frequencies , 30 minutes of data or @xmath520000 pulses were recorded . before each observation , a 90-s observation of a pulsed signal , directly injected into the receiver at a @xmath6 angle to the feed , is made .
this enables instrumental polarization to be corrected .
observations of the flux calibrator hydra a were made at each of the observing frequencies ; this allows absolute fluxes to be obtained .
the data were processed off - line using a workstation cluster at the swinburne supercomputer centre .
data reduction involves coherent de - dispersion ( hankins & rickett 1975 ) and includes quantization error corrections as described by jenet & anderson ( 1998 ) .
the data are folded at the apparent topocentric period of the pulsar and the full stokes profiles for each pulse are written to disk .
flux calibration and instrumental calibration are then carried out using information contained in the observation of the pulsed ( calibration ) signal .
the data in each frequency channel is corrected for the rotation measure of the pulsar and all the frequency channels are then summed to produce the final profile . at 1413 mhz
there are 2048 time - bins per pulse period for an effective time resolution of 44 @xmath3s , comparable to the scatter broadening at this frequency . at the lower frequency ,
the time resolution is 88 @xmath3s , but the scatter broadening dominates and the effective time resolution is @xmath50.5 ms at 660 mhz .
as far as we are aware , the literature has not pointed out the existence of microstructure in the vela pulsar , generally because the time resolution of the observations has been too coarse .
figure 1 shows ` typical ' single pulses from vela at 1413 mhz with a time resolution of 44@xmath3s .
microstructure is clearly seen .
this microstructure is ubiquitous in virtually every pulse at 1413 mhz .
we conclude that the smooth pulse profiles shown by kd83 are an artifact of their coarse time resolution . at lower observing frequencies , pulse scatter broadening smooths over the microstructure features , although they are still visible in the 660 mhz data .
a complete investigation of the microstructure in the vela pulsar will be presented elsewhere ( kramer et al . 2001 ) .
figure 2 shows a histogram of the mean flux density for the 20085 pulses recorded at 1413 mhz .
more than 95% of the pulses are within a factor of 2 of the mean flux density , 99.5% are within a factor of 3 and there are no pulses greater than 10 times the mean flux density .
this distribution is typical of many , perhaps most pulsars .
there are no giant pulses in vela in the same sense as those in the crab pulsar and psr b1937 + 21 .
it is well known that vela does not show any nulls ( e.g. kd83 ; biggs 1986 ) , and indeed we also see no evidence of nulling in the @xmath540,000 pulses collected at the two observing frequencies .
the weakest single pulses have continuum flux densities of 1.3 and 0.3 jy at 660 and 1413 mhz respectively .
we computed the modulation index , @xmath7/@xmath8 where @xmath7 is the rms and @xmath8 the mean intensity in the @xmath9th bin , for each bin in the pulse profile .
as also shown by kd83 , the modulation index is low in the centre of the pulse profile and increases towards the wings . to show this effect in a more striking way , we computed the quantity @xmath10=(max@xmath11@xmath8)/@xmath7 where max@xmath11 is the maximum intensity in the @xmath9th bin .
@xmath10 as a function of pulse phase is shown in figure 3 .
it can be seen that the value of @xmath12 is consistent with gaussian statistics in the centre of the pulse but that both the leading and trailing edges of the pulse have individual pulses which are extremely high amplitude with respect to the mean intensity .
a very similar result is seen at 660 mhz .
the three main features in this plot will be discussed in more detail below .
the sweep of the position angle across the pulse in the vela pulsar shows the characteristic s shape without any of the ` orthogonal ' jumps seen in a large number of other pulsars .
the swing of the position angle is well fitted by the rvm and shows that the line - of - sight cuts very close to the magnetic axis , i.e. @xmath13 and @xmath14 . to investigate whether any individual pulse shows emission in the orthogonal mode we compared the position angle of each bin in each pulse to that of the position angle in the integrated profile at the same pulse phase .
of the 40,000 pulses recorded at 660 and 1413 mhz , _
none _ shows evidence for orthogonal mode emission in the region of the profile s peak , consistent with the results of kd83 at 2295 mhz .
however , investigating the feature centered near phase + 4 ms in figure 3 , we discovered that it is a result of only 21 pulses at 1413 mhz .
these pulses not only exhibit integrated flux densities more than 10 times that of the integrated profile at this longitude , but they also have a consistently small fractional polarization . similarly , at 660 mhz there are 23 pulses with the same characteristics . of these ,
all had low net polarization but , where polarization was present it was orthogonal to the integrated polarization at this pulse phase .
figure 4 shows the profile resulting from the summation of the 21 pulses taken at 1413 mhz .
the large intensity of this ` bump ' feature compared to the integrated profile and the orthogonal mode emission can clearly be seen .
we also note that the circular polarization is negative under the bump emission and this is of the opposite sign to that in the integrated profile , as is commonly seen in orthogonal mode emission . introducing a jump of @xmath4 in the position angle of the bump region results in the identical results for @xmath0 and @xmath1 when fitting the rvm .
the midpoint of the ` bump ' emission occurs 3.86 ms after the main pulse peak at 1413 mhz and 4.37 ms afterwards at 660 mhz , a significant difference .
thus the profile is wider at 660 mhz than at 1413 mhz .
the ` bump ' component is also stronger relative to the main pulse at 1413 mhz than at 660 mhz by nearly a factor of 2 .
this frequency dependence is in the direction expected in the radius - to - frequency mapping paradigm .
such a high intensity , relatively rarely - appearing component has not been detected in any other pulsar although the interpulse in psr b0950 + 08 may show similar properties .
as stated above , at 1413 mhz there are 21 occurrences of the bump component in the @xmath520000 pulses .
a look at the pulse numbers , however , reveals a far from random sequence .
there are 7 cases in which the differences in pulse number between appearances of the bump component is less than 300 . of these ,
5 pairs have intervals either in the range 130 - 140 pulses or 260 - 280 pulses and other related harmonics are also present .
we computed fluctuation spectra from the single pulses , producing results which are somewhat different to those of kd83 , probably due to our much better time resolution . including 15360 pulses in our analysis
, we indeed find periodicities of @xmath15 and @xmath16 cycles / period , i.e. every @xmath17 and @xmath18 pulses , in the ` bump ' region .
these periodicities are significant at the @xmath196@xmath20 level .
we also detect an even lower frequency feature in the main peak region , which could be associated with a ` drift ' in micropulses , which we sometimes observe .
a longer dataset is required to determine the significance of these features ; observations at 1400 and 2400 mhz will be carried out in the near future . at 1413 mhz
there are 14 pulses for which @xmath21 and which do not occur in the bump region .
the individual pulses are shown superimposed in figure 5 .
of these 14 , the two which occur earliest in pulse phase are both centered at phase @xmath22 ms ( relative to the peak of the integrated profile ) and are very narrow with a full - width at half - maximum of only @xmath550@xmath3s .
the other 12 pulses have peak phases between @xmath23 and @xmath24 ms and have typical half - widths of @xmath5400@xmath3s .
the brightest of these pulses has a peak flux density in excess of 2500 jy , more than 40 times the peak flux density in the integrated profile and more than 130 times the mean flux density at this pulse phase . at 660 mhz
the results are very similar .
one high amplitude pulse is seen at early phases and 19 at phases just prior to the main pulse .
the peak amplitudes are again in excess of 2000 jy , this is only a factor of @xmath510 higher than the peak of the main pulse at this frequency .
we call all these features giant micro - pulses : they are not true giant pulses because their mean flux density does not exceed 10 times the mean flux density , but their peak flux densities are very large in absolute terms and they appear far more often than would be expected from any ` normal ' flux probability distribution .
the polarization of these giant micro - pulses is extremely high with a linear fraction close to 100% .
furthermore , the position angle of the polarization is exactly that expected by extrapolating the main position angle swing back to these longitudes .
finally , 13 of the 14 pulses have small negative circular polarization and one has positive circular polarization .
although kd83 also noted that strong intensity pulses arrived at early phases , they make no explicit mention of single giant pulses .
however , they seem to have employed a time - averaging technique which may have had the effect of broadening the giant pulses ( because they do not all arrive at the same pulse phase ) and reducing their amplitude .
our secondary feature more than 2 ms earlier than the main pulse peak was not covered in their ` on - pulse ' bins .
if we assume that the line - of - sight crosses the magnetic axis at the peak of the profile , we can compute the emission height at 660 and 1413 mhz by measuring the delay of the steepest swing of the position angle relative to the profile peak ( blaskiewicz et al .
1991 ; hibschman & arons 2000 ) . using the result of our rvm fits ,
the delay is 1.3 ms at 660 mhz and 1.04 ms at 1413 mhz leading to emission heights of 98 and 78 km respectively , although the 660 mhz value has a significant error bar due to the scatter broadening .
these emission heights are typical in pulsars , and the frequency scaling is also within the range seen in other pulsars ( e.g. kramer et al .
1994 ) . in turn , by assuming dipolar field lines , the emission heights can be used to derive the half - opening angle of the cone , @xmath25 , respectively 13@xmath26 and 11.6@xmath26 at the two frequencies .
knowledge of @xmath27 and the viewing geometry can be used to derive the width of the pulse profile .
( normally this process is used in reverse , with the known pulse width used to derive @xmath27 ) . for @xmath28 and @xmath29 ,
the pulse full - width should be @xmath30 or @xmath31 ms .
even when considering that this width is to be measured at a very low intensity level , it seems much larger than observed in the integrated profile .
however , the emission from the ` bump ' component now extends the pulse profile towards the trailing edge , matching the expected width quite nicely .
luo & melrose ( 2001 ) have recently proposed that the trailing edges of wide cones have their emission suppressed due to cyclotron absorption within the pulsar magnetosphere .
such a mechanism may explain why these ` bump ' pulses are seen so rarely .
the presented calculation assumes that the magnetic axis coincides with the pulse peak , tempting the speculation that we are missing the leading edge of the cone . perhaps extending to a phase of @xmath32 ms ,
the cone may not emit at all or only very rarely .
we note that the giant micro - pulses do increase the pulse width on the leading edge but only by 1 ms or so .
one could also assume that , with the extra bump component , we are seeing the full pulse width .
in this case , the magnetic pole crossing would be close to phase @xmath33 ms .
however , this reasoning does not fit well with a number of the observations .
first , the steepest swing of the position angle sweep occurs _
prior _ to this point . also , the frequency dependence of the pulse width and the spectral index of the various features show that the ` main ' pulse is more likely to be the core component and the ` bump ' more likely to be conal ( e.g. kramer et al .
1994 ) . considering the apparent two groups of giant micro - pulses
, it is unclear whether we are dealing with two separate phenomenon or whether the small number statistics are preventing us from seeing these features at all pulse phases near @xmath34 ms .
however , the giant micro - pulses at the earliest phases are of much shorter duration than those occuring later , and the same pattern is seen at 660 mhz as at 1413 mhz . what is certain is that they all lead the main pulse ; there are no such features anywhere after the main pulse peak .
these giant micro - pulses seem to have more in common with the giant pulses in psr b1937 + 21 than with the crab giant pulses . in the former ,
the giant pulses all occur at the same pulse phase , whereas in the crab the giant pulses can arrive anywhere within the pulse window . in vela ,
the giant micro - pulses all arrive within a narrow phase range and lead the main pulse ( unlike in psr b1937 + 21 where the giant pulses lag the main pulse ) .
backer , d. c. & rankin , j. 1980 , apj supp .
series , 42 , 143 + blaskiewicz , m. , cordes , j. m. & wasserman , i. 1991 , apj , 370 , 643 + cognard , i. , shrauner , j. a. , taylor , j. h. & thorsett , s. e. 1996 , apj , l81 + foster , r. s. & hankins , t. 1995 , apj , 453 , 433 + hankins , t. h. & rickett , b. j. 1975 , meth .
, 14 , 55 + hibschman , j. a. & arons j. , 2000 , astro - ph/0008117 + jenet , f. a. , cook , w. r. , prince , t. a. & unwin , s. c. 1997 , pasp , 109 , 707 + jenet , f. a. & anderson , s. b. 1998 , pasp , 110 , 1467 + kramer , m. , johnston , s. & van straten , w. 2001 , in preparation + kramer , m. , wielebinski , r. , jessner , a. , gil , j. a. , seiradakis , j. h. , 1994 , a&as , 107 , 515 + krishnamohan , s. & downs , g. s. 1983 , apj , 265 , 372 ( kd83 ) + lange , c. , kramer , m. , wielebinski , r. & jessner , a. 1998 , a&a , 332 , 111 + lundgren , s. c. , cordes , j. m. , ulmer , m. , matz , s. m. , lomatch , s. , luo , q. & melrose , d. b. , 2001 , mnras , submitted .
+ lyne , a. g. & manchester , 1988 , mnras , 234 , 477 + mckinnon , m. m. & stinebring , d. r. 1998 , apj , 502 , 883 + radhakrishnan , v. & cooke , d. j. 1973 , astrophysical letters , 3 , 225 + rankin , j. m. , 1993 , apjs , 85 , 145 | we present high time resolution observations of single pulses from the vela pulsar ( psr b083345 ) made with a baseband recording system at observing frequencies of 660 and 1413 mhz .
we have discovered two startling features in the 1413 mhz single pulse data .
the first is the presence of giant micro - pulses which are confined to the leading edge of the pulse profile .
one of these pulses has a peak flux density in excess of 2500 jy , more than 40 times the integrated pulse peak .
the second new result is the presence of a large amplitude gaussian component on the trailing edge of the pulse profile .
this component can exceed the main pulse in intensity but is switched on only relatively rarely .
fluctutation spectra reveal a possible periodicity in this feature of 140 pulse periods .
unlike the rest of the profile , this component has low net polarization and emits predominantly in the orthogonal mode .
this feature appears to be unique to the vela pulsar .
we have also detected microstructure in the vela pulsar for the first time .
these same features are present in the 660 mhz data .
we suggest that the full width of the vela pulse profile might be as large as 10 ms but that the conal edges emit only rarely . |
star formation in galaxies generally occurs in star clusters instead of isolated stars , at least 20@xmath6 and possibly all stars form in clusters or associations ( fall 2004 ) .
young massive star clusters ( yscs , with masses often @xmath7 10@xmath3m@xmath4 ) , which are thought to be the products of violent star - forming episodes triggered by galaxy collisions , mergers , and close encounters ( de grijs 2003,2004 , and references therein ) , or generally form in the disks of isolated spirals with higher efficiency in environments of high star formation rate ( larsen 2004a , and references therein ) , are important for studying the ongoing star formation , stellar populations , and the evolutionary histories of their parent galaxies .
yscs are thought to formed in giant molecular clouds ( gmcs ) , and concentrated in star - forming clumps ( over - dense regions , or cores ; e.g. , elmegreen 2004 ) .
the majority of important studies of extragalactic star clusters in the last years have involved use of the _ hubble space telescope _ ( _ hst _ ) , with its unprecedented spatial resolution ( @xmath10.04@xmath2 ) and full uv / optical bands ( 0.1 - 1.0@xmath5 m ) access for studying stellar populations , in particular the blue coverage for age - dating young clusters ( e.g. , de grijs et al . 2003 ; larsen 2004b ) .
the ages and masses of star clusters can be estimated using color - magnitude and/or color - color diagrams and compared with stellar population synthesis models ( e.g. , johnson et al .
2003 ; harris et al .
previous studies of extragalactic star clusters paid more attention to their general properties based on optical images , such as their luminosity and mass functions ( e.g. , whitmore et al .
1999 ; elmegreen et al .
2001 ) , the size of the clusters ( e.g. , larsen 1999 ) , and the mechanisms of cluster formation and disruption ( e.g. , elmegreen 2004 ) . nevertheless , mid - infrared observations of extragalactic star clusters are also very important , for understanding the properties of heavily obscured clusters ( e.g. , bontemps et al . 2001 ) , and dust environments of young cluster - forming systems ( e.g. , zhang et al .
spitzer space telescope _
s ( werner et al .
2004 ) observations in mid - infrared with higher sensitivity and better angular resolution than previous observations ( e.g. , _ iso _ ) , provide a new opportunity to study both young and old stellar populations and star formation in normal ( e.g. , pahre et al 2004 ; wu et al .
2005 ; calzetti et al . 2005 ) , starburst ( cannon et al .
2005,2006ab ) , and interacting / merging ( e.g. , wang et al .
2004 ; smith et al . 2005 ; elmegreen et al . 2006 ) galaxies .
the four irac bands from 3.6 to 8.0 @xmath5 m probe both stellar continuum and warm dust emissions ( of the so - called polycyclic aromatic hydrocarbon , or pah , and dust continuum emissions ) , and mips 24@xmath5 m band probes the warm dust emissions from the very small grains ( vsgs ) .
although _ spitzer _ images are unable to resolve individual star clusters which can be well resolved by _ hst _ , they can be used to study the mid - infrared properties of the infrared bright knots / clumps which may be collections of ob associations and dense clusters of several hundred parsecs in size ( efremov 1995 ; elmegreen et al . 2006 ) .
arp 24 ( ngc 3445 ) forms a triplet with ngc 3440 and ngc 3458 , at a separation of 9.9@xmath8 and 14.0@xmath8 , respectively .
numerous h ii region candidates exist in its spiral arm and disk . at the end of its southern spiral arm is a shred that may be a separate galaxy or was one in the past .
arp 24 is at distance of 27.6 mpc ( 1@xmath2 corresponds to 134 parsecs ) and with a total infrared luminosity of l@xmath9 @xmath1 4.8@xmath1010@xmath11l@xmath12 ( bell 2003 ) .
van den bergh ( 1995 ) classified it as a transitional object that appear intermediate between spirals that have central bulges and objects having central regions that are resolved into stars and knots .
bker et al .
( 2002,2003 ) found it contains a nuclear star cluster in the central region instead of a spiral bulge .
arp 24 is a target of the `` spiral , bridges , and tails '' ( sb&t ) guest observer cycle 1 _ spitzer _ program ( pi : c. struck ) , which is for studying the distribution of star formation in a sample of colliding galaxies with a wide range of tidal and splash structures ( see smith et al . 2005,2006 ; hancock et al .
2006 for details ) . in this paper
, we present an analysis of data from _ hst _ , _ spitzer _ , sdss , _ galex _ , h@xmath0 image , and spectroscopic observations , for studying the properties and possible formation scenarios of the yscs in the interacting galaxy arp 24 .
the observations and relevant data reduction are presented in @xmath132 ; results on the multi - wavelength emissions from the yscs , the discovery of a ysc complex ( yscc ) , and other physical properties of the yscs are described in @xmath133 .
possible formation scenarios of the yscs , and the pah and warm dust emissions of the infrared knots in this system are discussed in @xmath134 .
the major results of this work are summarized in @xmath135 .
the u ( f300w ) and i - band ( f814w ) images of arp 24 were taken with the wfpc2 wide field camera and planetary camera on board _ hst _ , with exposure times of 600 and 640 seconds , respectively .
the images were obtained as b associations from the eso / st - ecf science archive @xmath14 . the photflam ( in ergs@xmath15@xmath16@xmath15 ) , exptime ( exposure duration ) keywords in the header of each image
were used for converting instrumental magnitudes to flux densities , and to vega magnitudes based on the zeropoints given in wfpc2 data handbook @xmath17 .
the broadband optical images ( @xmath18 ) were derived from the sdss data archive ( york et al .
2000 ; stoughton et al .
the background fitted by low - order legendre polynomial was subtracted from each band image , after masking out all bright sources ( zheng et al .
1999 ; wu et al .
then the counts were converted to flux densities and ab magnitudes @xmath19 .
figure 1 shows the three color image of arp 24 derived from the sdss data archive .
north is up , east is to the left as denoted by the cross hair .
the far - ultraviolet ( fuv ; 1516 ) and near - ultraviolet ( nuv ; 2267 ) images were derived from the ultraviolet atlas of nearby galaxies ( gil de paz et al .
2006 ) based on images obtained with nasa s satellite , the _ galaxy evolution explorer _ ( _ galex _ ; martin et al .
keywords of mean sky - background level ( sky ) and zero point in ab magnitudes scale ( zp ) in the headers of the fuv & nuv images were used for sky - subtractions and flux calibrations , respectively . the narrow - band interference filter image centered near the red - shifted h@xmath0 and the associated continuum filter ( r band ) image of arp 24
were taken on 23 february , 2006 , by the 2.16 m telescope at xinglong observatory of naoc @xmath20 , with the bao faint object spectrograph and camera ( bfosc ) , using a lick 2048 @xmath10 2048 ccd as detector .
the field of view of the ccd is close to 10@xmath8 @xmath10 10@xmath8 , with a pixel size of 0.305@xmath2 .
the `` h@xmath0 '' images actually include a component from the [ nii]@xmath21@xmath216548,6583 doublet in addition to the balmer line flux .
standard ccd reductions include overscan and bias subtractions , flatfield correction and cosmic - ray removal were applied to the images using iraf package .
then the astrometric calibrations were computed to place the images on the sdss frames , using several field stars as reference , and the accuracy of the calibration is better than 1@xmath2 .
sdss images were used as references to our r - band image for flux calibration , based on field stars and the transformations between sdss magnitudes and ubvrcic given by lupton ( 2005 ) @xmath22 .
r - band image was used for subtracting the stellar continuum from the h@xmath0 image .
we selected five unsaturated stars over the field of view and adjusted values from the scaled r - band subtracted h@xmath0 image that these stars were canceled in a statistical sense .
broadband mid - infrared images of arp 24 were acquired with the infrared array camera ( irac , fazio et al .
2004 ) and multiband imaging photometer for spitzer ( mips , rieke et al .
2004 ) on board the _ spitzer space telescope_. the basic calibrated data ( bcd ) were part of the lockman hole field in the _ spitzer _ wide - field infrared extragalactic ( swire ) survey ( lonsdale et al .
following preliminary data reduction by the _
science center pipeline , images of each of the four irac ( 3.6 , 4.5 , 5.8 and 8 @xmath5 m ) and mips 24@xmath5 m bands were mosaicked , after pointing refinement , distortion correction and cosmic - ray removal ( fazio et al .
2004 ; huang et al . 2004
; wu et al .
2005 ; surace et al . 2005 ) .
the mosaicked images have pixel sizes of 0.6@xmath2 and angular resolutions ( full width at half maximum , fwhm ) of 1.9@xmath2 , 2.0@xmath2 , 1.9@xmath2 and 2.2@xmath2 for irac four bands , and pixel size of 1.225@xmath2 and fwhm of 5.9@xmath2 for mips 24@xmath5 m band , respectively .
figure 2 shows _ spitzer _
irac 3.6 , 4.5 , 5.8 , 8@xmath5 m and mips 24@xmath5 m images of arp 24 .
the continuum - subtracted h@xmath0 image was also shown to emphasize the good correspondence of the peaks between h@xmath0 and 24@xmath5 m emissions .
aperture photometries were performed on _ hst _
wfpc2 f300w & f814w images using 0.2@xmath2 radius circular apertures .
photometries on sdss , _ galex _ , _ spitzer _ , and continuum - subtracted h@xmath0 images were performed using apertures of 12@xmath2 and 120@xmath2 diameters for individual regions and the entire galaxy , respectively .
all photometric values were derived without aperture corrections .
both the irac and mips 24@xmath5 m bands have calibration uncertainties at the 10@xmath6 level ( dale et al .
2005 ; wu et al .
the global values of the 70@xmath5 m and 160@xmath5 m fluxes of arp 24 were derived from the catalogues of swire data release 3 ( surace et al .
2005 ) .
the optical spectrum of the ysc complex ( see @xmath133.2 ) in arp 24 was taken on 07 april , 2006 .
the observation was carried out with 2.16 m telescope and the bfosc . a g4 grism ( from 4000 to 8000 ) and slit width of @xmath1 2@xmath2 were used in this observation , gives a dispersion of roughly about 200 / mm .
a relatively higher resolution spectrum of this complex was taken on 06 may , 2006 , with 2.16 m telescope and the omr spectrograph .
a 1200 l / mm grating ( from 6300 to 7100 ) , blazed at 6700 , and slit width of 2@xmath2 were used in this observation , gives a dispersion of about 50 / mm .
optical spectrum of the nuclear region in arp 24 was taken on 05 may , 2006 , with the omr spectrograph and a 300 l / mm grating ( from 3800 to 8000 ) with slit width of 2@xmath2 .
the unprocessed frames were reduced by standard ccd procedure using iraf package .
the ccd reductions include bias subtraction , flatfield correction and cosmic - ray removal .
the wavelength calibrations were carried out using fe / ar ( for bfosc ) and he / ne / ar ( for omr ) lamps .
kpno irs standard stars were observed each night for carrying out the flux calibrations .
from sdss ( fig . 1 ) and _ spitzer _ ( fig .
2 ) images , we found that the nucleus of arp 24 appears relatively redder in optical and bright at 3.6 and 4.5 @xmath5 m , but is absent in irac 8@xmath5 m and mips 24@xmath5 m bands . in the mips
24@xmath5 m image we identified four infrared bright knots ( labeled with k0 , k1 , k2 , k3 ) in arp 24 , which are bluer in optical and bright at 8@xmath5 m and 24@xmath5 m .
the brightest infrared knot ( k0 ) was found to be associated with a star cluster complex based on _ hst _ images ( see next section ) , so we re - labeled it with yscc. many gaseous structures and filamentaries were also shown up in irac 5.8 , 8@xmath5 m and mips 24@xmath5 m bands .
most of the emission at 8@xmath5 m ( and also at 5.8 @xmath5 m to a lesser extent ) arises from the so - called aromatic features ( usually attributed to pahs ) at 6.2 , 7.7 , and 8.6 @xmath5 m in the photo - dissociation regions ( pdr ) , which are often associated with the warm dust in or near star forming regions ( h ii regions ) , thus can be used as tracers of star formation ( peeters et al .
2004 ; wu et al . 2005 ) . mips 24@xmath5 m emissions , which is mainly due to hot dust emissions from the very small grains ( vsgs ) , were also thought to be good measures of the sfrs of galaxies ( wu et al .
2005 ; calzetti et al . 2005 ; prez - gonzlez et al .
thus , we conclude that the four infrared bright knots ( yscc , k1 , k2 , k3 ) are dominated by young stars and are sites of active star formation .
the h@xmath0 image of arp 24 shows strong h@xmath0 emissions in these regions , which also indicates young stellar population there ( see fig .
the ultraviolet to infrared spectral energy distributions ( seds ) were shown in figure 3 and table 1 . from comparison with
the empirical sed templates generated using the grasil code ( silva et al .
1998 ; see also more detailed descriptions in jarrett et al .
2006 ) , we find that the yscc , which has very strong 24@xmath5 m dust emission relative to pahs , is very similar to the prototype starburst galaxy m 82 in mid - infrared , but exhibits large excess of starlight at ultraviolet and short optical wavelength relative to the m 82 template ( see figure 3b ) .
this is similar to the tidal tail super massive clusters in tadpole , in which the significant blue excess is considered to be related to recent star formation ( jarrett et al .
2006 ) and the gas / dust was expected to be blown away by stellar winds and/or supernovae explosions .
the other three infrared bright knots ( k1 , k2 , k3 ) , which indicate strong pah and dust continuum emissions , are comparable to star formation regions in late - type sc galaxies ( e.g. , ngc 6946 ) and the infrared - bright regions of the tadpole disk ( jarrett et al .
2006 ) . the brightest infrared knot in arp 24 ( k0 ) was found to be associated with a complex of five star clusters ( yscc for short ) , based on u and i - band wfpc2 images from _ hst _ ( fig .
this complex is within a region of @xmath1 0.95@xmath2 radius ( 127pc ) in size , and apart from the nuclear star cluster in the galaxy center by about 20.5@xmath2 ( @xmath1 2.7kpc ) .
the measured fluxes and absolute magnitudes of the star clusters in the yscc , the nucleus , and other three infrared bright knots ( k1 , k2 , k3 ) were shown in table 2 @xmath23 .
the different m@xmath24-m@xmath25 colors between clusters in each region primarily reflect the mean age of the stellar population . the color - magnitude diagrams ( cmd ) of the yscc and three infrared bright knots in arp 24
are shown in figure 5 .
the ages and masses of the clusters were estimated using starburst99 instantaneous models ( leitherer et al .
1999 ) , with salpeter initial mass function ( imf , @xmath26 ) between 0.1 and 120 m@xmath12 and metallicities z=0.02 & 0.008 .
the models are also shown in the cmds , ages along the evolutionary lines which are drawn for cluster masses of 10@xmath3 ( solid ) and 4@xmath1010@xmath27 ( dotted ) m@xmath4 are indicated by diamonds ( z=0.02 ) and plus ( z=0.008 ) place in t=1 myr intervals and begin from 1 myr on the left .
the clusters in the yscc and k1 are within ages of around 3 - 5 myr and masses of about 10@xmath3m@xmath4 .
the clusters in other two infrared bright regions ( k2 & k3 ) are relatively less massive ( @xmath1 4@xmath1010@xmath27m@xmath4 ) but have similar ages ( @xmath1 4 - 6 myr ) to that in the yscc and k1 .
the masses estimated here are consistent with that derived from _ hst
_ f300w / f814w luminosities and the mass - to - light ratio ( m / l ) of @xmath1 0.01 - 0.02 in visual band for young clusters with a 10@xmath28 burst ( chandar et al .
the masses of the star clusters in the yscc and k1 are consistent with that of yscs or super star clusters ( de grijs 2003 ) observed in other interacting galaxies ( e.g. , ic 2163 & ngc 2207 , elmegreen et al .
2001 ; ngc4038/39 , whitmore et al . 1999 ) , which are thought to be the progenitors of luminous globular clusters ( e.g. , ma et al . 2006ab ; kravtsov 2006 ) . optical spectrum of the nucleus in arp 24 ( fig .
6 ) indicates a mixture of populations of different ages , with strong h@xmath29 absorption which indicates the existence of a large number of evolved a - type stars ( age @xmath1 10@xmath30yr ; see , e.g. , wang & wei 2006 ) , and strong h@xmath0 emission which traces ongoing star formation .
a template spectrum corresponds to an instantaneous - burst model with a young population of 100 myr and metallicity z=0.02 plus an old population of 11 gyr was also plotted for comparison .
this result is in agreement with previous studies that the nuclear clusters are massive and dense star clusters which form stars recurrently until the present day ( walcher et al . 2005,2006 ) .
unfortunately , its hst / stis spectra which can provide better separation of nuclear star cluster light from underlying galaxy light than our ground - based spectrum ca nt be used for stellar population analysis ( due to its low s / n ratio ; @xmath1 2.9 , see rossa et al .
. optical spectrum of the yscc in arp 24 ( fig .
7 ) shows strong hydrogen and oxygen emission lines which indicates young stellar populations and active star formation .
the derived redshift is 0.007 , matching the redshift ( 0.0069 ) of the galaxy derived from the updated zwicky catalog ( falco et al .
the emission lines were fitted with gaussian profiles and the line fluxes were listed in table 3 .
the line ratios are h@xmath0/h@xmath31=3.13 , [ oiii]/h@xmath31=3.57 , and [ nii]/h@xmath0=0.12,[sii]/h@xmath0=0.19 . using standard optical line ratio diagnostic diagrams and comparing with the mappings iii code ( kewley et al . 2001 ; kewley & dopita 2002 ) results , we computed photo - ionization models and estimated the ionization parameter q ( cm s@xmath15 ) of the yscc , which is about 4@xmath1010@xmath32 .
the dust extinction can be estimated from the balmer decrement ( calzetti 2001 ) : @xmath33 where the value of the intrinsic luminosity ratio h@xmath0@xmath34/h@xmath31@xmath34 is 2.87 for temperature
t = 10,000 k and case b recombination ( osterbrock 1989 ) , and the value of the differential extinction k(h@xmath31)-k(h@xmath0 ) between h@xmath0 and h@xmath31 is 1.163 ( calzetti 2001 ) .
then the extinction of the yscc can be calculated from a@xmath35=r@xmath35e(b - v ) ( r@xmath35 = 3.1 ) , and the value is about 0.25 . it is smaller than the attenuations in m 81 ( a@[email protected] ; see , e.g. , hill et al .
1995 ) , m 51 ( a@xmath35@xmath13 ; see calzetti et al . 2005 ) , and the archetype starburst galaxy m 82 ( a@[email protected] ; mayya et al .
2006 ) .
the metallicity was estimated using the line ratios , following vacca & conti ( 1992 ) : @xmath36 where @xmath37\lambda4959)+i([oiii]\lambda5007)}{i(h\beta)}\ ] ] the measured metallicity ( 12+log(o / h ) ) of the yscc is about 8.37 , or 0.51z@xmath4 if adopted the value of 8.66 for solar abundance ( asplund et al .
this relation ( r@xmath38@xmath1log(o / h ) ) has been shown to be affected by the ionizing photon hardness .
thus , we also estimated metallicities using the [ nii]/h@xmath0 ratios for a comparison , with : @xmath39}{h\alpha}\ ] ] ( kewley & dopita 2002 ) .
the metallicity of the yscc estimated using this relation is about 8.45 , consistent with that derived from the r@xmath38 value .
the metallicity of the nucleus in arp 24 was also estimated using the [ nii]/h@xmath0 ratio , and the value is about 8.79 , much higher than that of the yscc .
the total star formation rate ( sfr ) of arp 24 was calculated using infrared luminosity derived from _ spitzer _
mips ( 24 , 70 , 160@xmath5 m ) and the equation of dale & helou ( 2002 ) : @xmath40 then the sfr@xmath41 can be estimated by multiplying the l@xmath42 ( in ergs s@xmath15 ) by a conversion factor of 4.5@xmath1010@xmath43 m@xmath4yr@xmath15 ( kennicutt 1998 ) @xmath44 .
the total sfr in arp 24 is about 1.05m@xmath4yr@xmath15 .
the sfrs of the yscc and k1 , k2 , k3 were estimated using 24@xmath5 m dust emissions , which are thought to be good measures of the sfrs of galaxies ( wu et al .
2005 ; calzetti et al . 2005 ; perez - gonzalez et al .
2006 ) , and equation ( 3 ) of wu et al .
( 2005 ) : @xmath45 the measured sfrs in the regions of the yscc , k1 , k2 , and k3 are about 0.10 , 0.03 , 0.04 , 0.03 m@xmath12yr@xmath15 , respectively .
the sfr in yscc is extremely high , considering its relatively small spatial size comparing with the entire galaxy .
the sfr per unit area of this complex was estimated to be 0.05 to 1.97 m@xmath12yr@xmath15kpc@xmath46 using 24@xmath5 m 6@xmath2 radius and _ hst _ 0.95@xmath2 radius , respectively .
it is comparable to or even much stronger than the definition of a starburst galaxy ( @xmath10.1 m@xmath12yr@xmath15kpc@xmath46 , kennicutt et al .
2005 ) , thus the yscc can be justified as a localized starburst ( efremov 2004 ) .
although using infrared emission to estimate sfrs is more indirect than other young star tracers such as uv and h@xmath0 line emissions , it suffers relatively minor extinction effects which are difficult to correct . from the simulations of the performance of star - formation indicators in the presence of dust , jonsson ( 2004 )
found the infrared luminosity is more reliable than h@xmath0 and fuv luminosities which suffer severely from dust attenuation and the situation can only partially be remedied by dust corrections . for comparison we calculated the sfrs based on h@xmath0 line and fuv luminosities using the relations : sfr@xmath47(m@xmath4yr@xmath15 ) = 7.9@xmath1010@xmath48 l@xmath47(erg s@xmath15 ) ( kennicutt 1998 ) and log sfr@xmath49(m@xmath4yr@xmath15 )
= log l@xmath49(l@xmath4 ) - 9.51 ( iglesias - pramo et al .
the sfr@xmath47 and sfr@xmath49 ( no reddening corrected ) for each region are about 0.14 , 0.03 , 0.04 , 0.04 m@xmath12yr@xmath15 and 0.11 , 0.03 , 0.05 , 0.04 m@xmath12yr@xmath15 , respectively , a bit higher than that derived from the 24@xmath5 m luminosities . a better way for estimating sfrs may be based on a combination of the observed infrared and ultraviolet / optical luminosities as suggested by some authors recently ( e.g. , kennicutt 2006 ; iglesias - pramo et al .
we adopted the relations : sfr(m@xmath4yr@xmath15 ) = 4.5 l@xmath50 + 7.1 l@xmath49 ( 10@xmath51w ) ( dale et al . 2006 ) and log l(tir ) = log l(24 ) + 0.908 + 0.793 log [ l@xmath52(8)/l@xmath52(24 ) ] ( calzetti et al . 2005 ) , then the sfr@xmath53 for each region are about 0.14 , 0.05 , 0.07 , 0.06 m@xmath12yr@xmath15 , respectively , slightly higher than previous estimations . due to the relatively small variations of sfrs derived from different sfr indicators , we adopted the values of sfr@xmath54 for further analysis .
the stellar masses for the old stellar population in different regions of arp 24 were estimated based on the sdss photometries following bell et al .
( 2003 ) , @xmath55\ ] ] where the coefficients @xmath56 ( -0.306 ) and @xmath57 ( 1.097 ) are taken from table 7 of bell et al .
the estimated stellar masses in the regions of the yscc , k1 , k2 , and k3 are about 10@xmath58 , 10@xmath59 , 10@xmath58 , 10@xmath60 m@xmath12 , respectively .
we also compared the measured stellar masses with the monochromatic irac 3.6@xmath5 m luminosities ( which are around 10.57 , 8.55 , 8.10 , 6.81 @xmath1010@xmath61ergs s@xmath15 , respectively ) , which are thought to be approximate measures of the stellar masses in galaxies ( e.g. , smith et al .
we found the regions in arp 24 followed a similar scaling relation between mass@xmath62 and logl(3.6@xmath5 m ) to that in the arp 82 system ( hancock et al .
the specific star formation rate ( ssfr , sfr per unit stellar mass , in units of gyr@xmath15 ) for each region are 0.32 , 0.08 , 0.14 , 0.13 gyr@xmath15 , respectively .
the ssfrs of these regions in arp 24 are higher than that of most of the local star - forming galaxies ( with values between 0.03 and 0.2 gyr@xmath15 ; bell et al . 2005 ) , but are much lower than that of luminous infrared galaxies ( lirgs ) in the local universe ( lie between 1.2 and 10 gyr@xmath15 ; wang et al .
however , the ssfrs estimated here will only be a lower limit for the ssfrs within or around the yscs in themselves , due to the contamination of star lights ( dominated by old population ) from their parent galaxy .
peculiar one - time events and special places that have extraordinary high energy inputs was suggested by elmegreen ( 2004 ) as one of the possible mechanisms for triggering the formation of yscs .
he suggested that yscs in galaxies undergoing interactions may formed by large part or local shock compressions and collapse from cloud impacts or colliding supershells .
bastian et al .
( 2006a ) studied the young star cluster complexes in antennae and suggested that if we assume that the grouping of complexes formed out of the same gmc , then the star formation is triggered by an external perturbation .
cannon et al .
( 2005 ) studied the infrared properties of the supergiant shell region of the dwarf galaxy ic 2574 using _
spitzer _ and demonstrate that the expanding shell is affecting its surroundings by triggering star formation and heating the dust .
the morphology of arp 24 indicates that it may formed in a minor merger scenario ( interaction between a gas - rich late - type spiral and a small companion ) or from cosmological accretion of gas on galactic disks ( bournaud et al .
shaping / reshaping of galaxies is the biggest effect generally attribute to mergers ( cox 2004 ) .
the tidal stripping of a satellite can produce features such as long tails , and the accretion introduces new stellar populations into the galactic disk ( walker et al .
1996 ) , i.e. , promotes a high sfr ( e.g. , mihos & hernquist 1994,1996 ; see also cox 2004 ) in arp 24 .
the yscs may form as a consequence of the high sfr per unit area in the galaxy ( larsen & richtler 2000 ) .
arp 24 shows a strong asymmetry in g - i colors ( see fig .
8) through the galactic disk , the regions near the yscc are relatively bluer than that near the nucleus
. this asymmetry may be due to the differences in star formation , stellar populations , gas / dust contents of different regions caused by galaxy interactions .
furthermore , the one - armed appearance and the broad fan - like structure seen in the primary disk of arp 24 indicate that this galaxy may have formed by a retrograde fly - by encounter ( see , e.g. , barnes & hernquist 1996 ; struck & smith 2003 ; and a review by struck 1999 , and references therein ) .
the companion galaxy in the east seems nearly edge - on ( also a bit warp ) , with a weak bridge connects to the primary disk and faint knots near the outer edges of its disk ( see figs .
this could indicates the existence of an edge - on ring ( or other waves ) , which would be consistent with a small angle between the orbital plane of the companion and the disk plane of the primary , if the closest approach was on the west side .
it is also possible that the primary disk might have rotated by about half a turn in the time since the closest approach .
thus , the region of the yscc in arp 24 might be in the part of the disk that was closest to the companion and most perturbed at that time .
if we adopt a rotation velocity of about 100 km s@xmath15 ( derived from the rotation curve of m 51 @xmath63 ) and the distance between yscc and the nuclear star cluster ( @xmath1 2.7kpc ) as the radius of rotation , then such half a turn will takes about 8@xmath1010@xmath32yrs ; and moreover , if we assume a circular orbit of the encounter and a relative velocity of @xmath1 400 km s@xmath15 between the companion and the primary s core ( eneev et al .
1973 ) , then the time since the closest approach in the west is roughly about 10@xmath30yrs , consistent with the rotational time estimated above .
this time scale may linked to the processes of yscc formation ( e.g. , modes and mechanisms of the induced star formation ; feedbacks include radiation / thermal pressures , stellar winds and supernova blasts , etc ) .
the existence of a large amount of young ( a - type ) stars in the nucleus of arp 24 ( see @xmath133.3 ) may also be related to the age of interaction in this system .
however , such a formation scenario still remains speculative , and needs to be confirmed by hydrodynamical modeling ( e.g. , struck et al .
2005 ) and numerical simulations of star formation in galaxy mergers ( e.g. , struck & smith 2003 ; cox 2004 ) .
recent studies have shown that young star clusters tend to form in large complexes instead of isolation ( bastian et al .
according to efremov ( 2004 ) , these complexes ( or so called clusters of clusters ) are often the massive bound clusters formed from a single gas supercloud , and within high pressure surroundings .
zhang et al . (
2001 ) made a multi - wavelength study of the mass , age , and space distributions of young star clusters in the antennae galaxy ( ngc 4038/39 ) , and found the young clusters have a clumpy space distribution and located in regions of high interstellar density .
and the cluster formation rate ( @xmath64 ) is correlated with the interstellar medium ( ism ) density ( @xmath65 ) . from the spectroscopic studies of an unusual star complex in ngc 6946 ( larsen et al .
2002 ) using the sao 6 m and keck 10 m telescopes , efremov et al .
( 2002 ) found this complex resembles a circular bubble 600pc in diameter with a young super star cluster near its center . and the intensities of the emission lines within and around the complex indicate that shock excitation makes a significant contribution to the emission from the most energetic region .
larsen ( 2004b ) found a strong correlation between cluster age and crowding of the environment , with most of the crowded clusters having young ages ( @xmath66 10@xmath32yr ) .
chen et al .
( 2005 ) found a tight group of clusters in the very luminous giant h ii region : ngc 5461 in spiral galaxy m 101 .
bastian et al .
( 2005 ) found several young star cluster formed in larger groupings / complexes in spiral galaxy m 51 .
these complexes are all young ( @xmath6710 myr ) , have sizes between @xmath185 and @xmath1240 pc , and have masses between 3 - 30@xmath1010@xmath27m@xmath4 , similar to the yscc in arp 24 .
elmegreen et al .
( 1993 ) suggested that the local turbulent velocity dispersion plays an important role in setting the mass scale of the star - forming clouds .
the larger mass clouds which formed in more turbulent conditions , are more resistant to the disruptive effects of young stars , i.e. , can sustain a higher star - forming efficiency .
clark et al .
( 2005 ) investigated the formation of star clusters in unbound giant molecular clouds ( gmcs ) , where the cloud tend to form a series of star clusters and disperse quickly , within a rapid star formation process ( @xmath110myr ) .
they also proposed that the clusters that form in the unbound gmcs may be progenitors of the ob associations .
zhang et al . (
2001 ) found the youngest star clusters in the antennae galaxy are associated with molecular cloud complexes with characteristic radii of about 1 kpc .
wilson et al . ( 2003 ) studied properties of the supergiant molecular complexes in the antennae using co emission , and suggested that young massive star clusters formed from dense cores within the observed supergiant molecular complexes .
the star clusters in the yscc of arp 24 may formed in a single gmc simultaneously , since they have similar ages and masses , and reside in a small region of @xmath1127pc ( see @xmath133.2 ) .
then they may disperse quickly and dissipate to become field population before reaching ages of 100myr according to wilson et al .
however , this scenario still needs to be confirmed by deep spectroscopic observations with high dispersion and high spatial resolution , for measuring the velocity dispersions and obtaining chemical abundances of individual clusters in the yscc ( see , e.g. , larsen et al .
2006 ; bastian et al .
there is no evidence of merging of star clusters in the center of the yscc , which was found in many complexes in m 51 ( bastian et al .
2005 ) . alternatively , perhaps the yscc is a dwarf galaxy which have fallen into arp 24 previously during a merger event .
its blue color ( @xmath68 , see also fig .
8) is similar to those of irregular galaxies ( fukugita et al .
1995 ) , and its absolute magnitude ( @xmath69 ) is located at the faint end of the luminosity function of extremely low - luminosity galaxies ( blanton et al .
its metallicity is low ( see @xmath133.3 ) , and is comparable to those of local dwarf irregular galaxy ( e.g. , ngc 6822 ; lee et al .
its warm infrared color ( high f@xmath54/f@xmath70 ratio ; see next subsection ) is also consistent with that of irregular galaxies ( e.g. , ddo 053 ) in the _ spitzer _ infrared nearby galaxies survey ( sings , kennicutt et al .
2003 ; dale et al .
2005 ) . from the analysis of seds of the yscs in arp 24 ( see @xmath133.1 )
, we find the yscc have higher 24@xmath5 m emission relative to the starlight ( 3.6@xmath5 m ) and pah emission ( 8@xmath5 m ) than the nucleus and other infrared bright knots ( fig . 9 ) .
regions associated with yscs ( or sscs ) often show warm infrared colors ( high 24@xmath5 m emission relative to that at shorter wavelengths ) in several interacting and starburst galaxies .
thuan et al .
( 1997 ) discovered six super - star clusters ( sscs ) with ages @xmath7125 myr and within a region of @xmath1520 pc diameter in the extremely metal poor ( z@xmath1z@xmath4/41 ) blue compact dwarf galaxy ( bcd ) sbs 0335 - 052 . mid - infrared spectrum with _ iso _ and _ spitzer _ show no sign of pah emissions and a relatively flat continuum between 5 and 20 @xmath5 m ( thuan et al . 1999 ; houck et al .
2004 ) , which can be explained by the effect of the destruction of the pah molecules by high uv radiation field .
jarrett et al .
( 2006 ) found the tidal tail supermassive star clusters ( with m @xmath1 10@xmath72m@xmath4 ) in the tadpole galaxy ( ugc 10214 ) have exceptionally strong 24@xmath5 m emission relative to the starlight , hot dust continuum , and pah emissions .
elmegreen et al . ( 2006 ) studied the infrared clumpy emissions in the interacting galaxy pair ic 2163 and ngc 2207 , and found the brightest giant infrared clump ( feature i ) which contains two sscs , numerous other star clusters and associations , several dark dust clouds , and supernova 1999ec , have extremely high 24@xmath5 m emission relative to 3.6@xmath5 m and 8@xmath5 m emissions .
they suggest that this giant clump may formed by gravitational instabilities in the compressed gas of the oval and spiral arms . in the western spiral arm of the well - studied nearby spiral galaxy m 51 ( ngc 5194 ) ,
several star - forming regions are associated with luminous yscs and show high f@xmath54/f@xmath70 ratios ( e.g. , region 05 - 01 shown in calzetti et al .
2005 , associated with the ysc # 180 shown in larsen 2000 ; see fig .
special conditions in these infrared warm regions can lead to the formation of extremely young and massive star clusters , with strong uv radiation fields which can reduce the pah feature strength and enhance the warm dust emissions .
the high f@xmath54/f@xmath70 ratio in the yscc may be due to the weakness or lack of the pah emission bands , which are caused by low pah abundance as consequences of pah destructions by strong uv radiation field and/or other mechanisms ( e.g. , shocks ) within this region ; or pah - dust competition , i.e. , pah molecules ca nt be excited by uv photons due to the presence of the dust grains ( martin - hernandez et al .
the elevation of the 24@xmath5 m dust emission indicates that the vsgs have been heated to a high temperature by their nearby luminous young stars .
the hardness of the interstellar radiation field may play a major role in the destruction of pahs in the center of star formation regions ( e.g. , madden et al .
2006 ; bendo et al .
2006 ) , and cause the warm infrared colors .
the very blue optical color of the yscc region in arp 24 ( see @xmath133.1 ) indicates that shocks induced by strong stellar winds and/or supernovae explosions could also be a mechanism for destroying pahs ( ohalloran et al .
low pah abundance can also originate from low pah formation rates due to different dust formation processes in a low metallicity environment ( e.g. , hogg et al . 2005 ; engelbracht et al .
gordon et al .
( 2006 ) found the 6.2 and 7.8 + 8.6@xmath5 m pah features in m 101 h ii regions are weak or absent at a metallicity 12+log(o / h)@xmath18.0 .
thus , the low o / h abundance in the region of the yscc compared with that in the nuclear region ( see @xmath133.3 ) indicates that metallicity may play a role in setting infrared colors in arp 24 . however , bolatto et al .
( 2006 ) showed that the f@xmath70/f@xmath54 ratios of the regions in the small magellanic cloud ( smc ) are regulated primarily by the local interstellar radiation field rather than metallicity , and suggested that photo - destruction of pahs may be primarily responsible for the variation of the measured f@xmath70/f@xmath54 ratios in the smc .
dust can be heated by emissions arising from older and cooler stars ( e.g. , b stars , peeters et al .
2004 ) , besides ionizing photons from young hot stars .
the strength of the 24@xmath5 m continuum emission from vsgs is more sensitive to the radiation energy density and hardness than the 8@xmath5 m pahs ( roussel et al .
thus , the f@xmath54/f@xmath70 ratio also could link to the mean age of the stellar populations responsible for dust excitation .
regions associated with yscs do not always show warm infrared colors .
cannon et al .
( 2006a ) found a lower f@xmath54/f@xmath70 ratio and modest warm dust emission of the ssc in the center of nearby dwarf starburst galaxy ngc 1705 .
they suggested that this may due to the dust removal by multi - phase outflow , or evacuation by high uv flux from the ssc .
the k1 region in arp 24 , which is also associated with a young ( t@xmath13 - 5 myr ) massive ( m@xmath110@xmath3m@xmath4 ) star cluster , does nt show a high f@xmath54/f@xmath70 ratio like that of the yscc .
this indicates the region - to - region variations of mid - infrared colors may depend on several physical parameters such as dust composition , dust temperature , and pah molecular abundance . on the other hand ,
not all the regions with warm infrared colors are associated with optically bright yscs .
for example , in ngc 4038/39 , the brightest knot in the 15@xmath5 m map ( mirabel et al . 1998 ) has only a faint i - band counterpart ( object # 80 ; whitmore & schweizer ( 1995 ) ) , but was found to be associated with a molecular complex ( sgmc4 - 5 ; wilson et al .
we found that this knot has infrared colors even warmer than that of the yscc in arp 24 , based on archival _ spitzer _ data ( pi : g. fazio ; see also wang et al .
its very red near - infrared color ( @xmath73 ) indicates that it suffers from heavy dust extinction ( brandl et al .
thus , we speculate that some star clusters in galaxies which are very young and optically faint caused by heavy dust extinctions may also have warm infrared colors like the yscc we found in arp 24 . and
this may also caused by the strong uv radiation fields in these regions which may destroy the pah molecules and enhance the vsg emissions at 24@xmath5 m .
firm conclusions must await a quantitative multi - wavelength analysis of a large , well - defined , and unbiased sample of yscs in galaxies .
the precise roles of the pah destruction and the enhanced 24@xmath5 m dust emission in setting the f@xmath54/f@xmath70 ratios of the yscs in arp 24 could be well - determined by using infrared spectroscopic observations .
ground - based mid - infrared imaging and spectroscopy of yscs with high spatial resolution ( subarcsecond ; e.g. , snijders et al .
2006 ) will help for studying the spatial distributions and originations of pah and warm dust emissions .
in this work , we present a study of young massive clusters in the interacting galaxy arp 24 using images from _ hst _ , _ spitzer _ , sdss , _ galex _ , h@xmath0 image , and optical spectroscopy .
our major findings are the following : \(i ) from the _ hst _ wfpc2 u and i - band images , we found the brightest infrared knot in arp 24 is associated with a complex of young massive star clusters ( the yscc ) .
the ages and masses of the star clusters in the yscc and other three infrared bright knots ( k1 , k2 , k3 ) were estimated using _
hst _ color - magnitude diagrams and comparing with the starburst99 synthesis models .
the clusters in the yscc and k1 region are very young ( within ages of around 3 - 5 myr ) and massive ( m@xmath110@xmath3 m@xmath4 ) , while the clusters in k2 and k3 regions are relatively less massive ( m@xmath14@xmath1010@xmath27 m@xmath4 ) but have similar ages ( @xmath1 4 - 6 myr ) to that in the yscc and k1 region .
the masses of the star clusters in the yscc and k1 are consistent with that of young massive star clusters or super star clusters observed in other interacting galaxies .
\(ii ) using standard optical emission line ratio diagnostic diagrams and comparing with the mappings iii code results , we computed photo - ionization models and estimated the ionization parameter of the yscc .
the metallicity of the yscc was estimated using the emission line ratios , and the star formation rates of the yscc and k1 , k2 , k3 regions were calculated using monochromatic 24@xmath5 m , fuv , and h@xmath0 line luminosities .
the sfr density of the region of the yscc is high and comparable to or even much stronger than the definition of a starburst galaxy .
the stellar masses for the old stellar population in different regions were estimated based on the sdss photometries and irac 3.6@xmath5 m luminosities .
\(iii ) we speculate that the formation of the yscs in arp 24 is triggered by galaxy interaction , and this galaxy may formed by a retrograde fly - by encounter indicated by its one - armed appearance and fan - like structure .
the clusters in the yscc may formed in a single giant molecular cloud simultaneously since they have similar ages and masses , and reside in a small region of @xmath1127pc .
\(iv ) from the ultraviolet to mid - infrared spectral energy distributions ( seds ) , we found that the region of the yscc is relatively bluer in optical and has higher 24@xmath5 m dust emission relative to the emission at 8@xmath5 m .
we speculate that this is primarily due to strong uv radiation field ( or shocks ) within this region which may destroy the pahs and enhance the vsgs emission at 24@xmath5 m .
however , firm conclusions must await a well - defined sample and upcoming infrared spectroscopic observations .
we are gratefully to the referee : dr c. struck for his very careful review of this paper and many constructive advice and very useful suggestions .
we thank r. kennicutt , r. de grijs , j. wang , z .- g .
deng for advice and helpful discussions , and z. wang , j .- s .
huang , c .-
wang , f .- s .
liu for their capable help and assistance throughout the process of _ spitzer _ data reductions .
special thanks go to the staff at xinglong observatory of national astronomical observatories , chinese academy of sciences ( naoc ) for their instrumental and observational helps .
this project is supported by nsfc no.10273012 , no.10333060 , no.10473013 , no.10373008 .
this work is based in part on observations made with the nasa / esa _ hubble space telescope _
, obtained from the data archive at the space telescope science institute ( stsci ) .
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
the observations of arp 24 are associated with program ids 8599 ( t. bker ) and 9124 ( r. a. windhorst ) .
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 nasa contract 1407 .
the _ galaxy evolution explorer _ ( _ galex _ ) is a nasa small explorer .
the mission was developed in cooperation with the centre national detudes spatiales of france and the korean ministry of science and technology .
funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s .
department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england .
the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions .
the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , cambridge university , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington .
asplund , m. , grevesse , n. , sauval , a. j. , allende prieto , c. , & kiselman , d. 2004 , , 417 , 751 barnes , j. e. , & hernquist , l. 1996 , , 471 , 115 bastian , n. , gieles , m. , efremov , y. n. , & lamers , h. j. g. l. m. 2005 , , 443 , 79 bastian , n. , emsellem , e. , kissler - patig , m. , & maraston , c. 2006a , , 445 , 471 bastian , n. , saglia , r. p. , goudfrooij , p. , kissler - patig , m. , maraston , c. , schweizer , f. , & zoccali , m. 2006b , , 448 , 881 bell , e. f. 2003 , , 586 , 794 bell , e. f. , mcintosh , d. h. , katz , n. , & weinberg , m. d. 2003 , , 149 , 289 bell , e. f. , et al .
2005 , , 625 , 23 bendo , g. j. , et al .
2006 , , 652 , 283 blanton , m. r. , lupton , r. h. , schlegel , d. j. , strauss , m. a. , brinkmann , j. , fukugita , m. , & loveday , j. 2005 , , 631 , 208 bker , t. , laine , s. , van der marel , r. p. , sarzi , m. , rix , h .- w .
, ho , l. c. , & shields , j. c. 2002 , , 123 , 1389 bker , t. , stanek , r. , & van der marel , r. p. 2003 , , 125
, 1073 bolatto , a. d. , et al . 2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0608561 bontemps , s. , et al .
2001 , , 372 , 173 bournaud , f. , combes , f. , jog , c. j. , & puerari , i. 2005 , , 438 , 507 brandl , b. r. , et al .
2005 , , 635 , 280 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 calzetti , d. 2001 , , 113 , 1449 calzetti , d. , et al .
2005 , , 633 , 871 cannon , j. m. , et al .
2005 , , 630 , l37 cannon , j. m. , et al . 2006a , , 647 , 293 cannon , j. m. , et al .
2006b , , 652 , 1170 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 chandar , r. , bianchi , l. , ford , h. c. , & salasnich , b. 1999 , , 111 , 794 chen , c .- h .
r. , chu , y .- h . , & johnson , k. e. 2005 , , 619 , 779 clark , p. c. , bonnell , i. a. , zinnecker , h. , & bate , m. r. 2005 , , 359 , 809 cox , t. j. 2004 , ph.d .
thesis daigle , o. , carignan , c. , amram , p. , hernandez , o. , chemin , l. , balkowski , c. , & kennicutt , r. 2006 , , 367 , 469 dale , d. a. , et al .
2005 , , 633 , 857 dale , d. a. , et al . 2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0610688 de grijs , r. , lee , j. t. , clemencia mora herrera , m. , fritze - v .
alvensleben , u. , & anders , p. 2003
, new astronomy , 8 , 155 de grijs , r. 2003 , astronomy and geophysics , 44 , 14 de grijs , r. 2004 , asp conf .
ser . 322 : the formation and evolution of massive young star clusters , 322 , 29 efremov , y. n. 1995 , , 110 , 2757 efremov , y. n. , et al . 2002 , , 389 , 855 efremov , y. n. 2004 , arxiv astrophysics e - prints , arxiv : astro - ph/0410702 elmegreen , b. g. , kaufman , m. , & thomasson , m. 1993 , , 412 , 90 elmegreen , b. g. , sundin , m. , kaufman , m. , brinks , e. , & elmegreen , d. m. 1995 , , 453 , 139 elmegreen , b. c. 2004 , asp conf .
ser . 322 : the formation and evolution of massive young star clusters , 322 , 277 elmegreen , d. m. , elmegreen , b. g. , kaufman , m. , sheth , k. , struck , c. , thomasson , m. , & brinks , e. 2006 , , 642 , 158 eneev , t. m. , kozlov , n. n. , & sunyaev , r. a. 1973 , , 22 , 41 engelbracht , c. w. , gordon , k. d. , rieke , g. h. , werner , m. w. , dale , d. a. , & latter , w. b. 2005 , , 628 , l29 falco , e. e. , et al . 1999 , , 111 , 438 fall , s. m. 2004 , asp conf .
322 : the formation and evolution of massive young star clusters , 322 , 399 fazio , g. g. , et al .
2004 , , 154 , 10 fukugita , m. , shimasaku , k. , & ichikawa , t. 1995 , , 107 , 945 fukugita , m. , ichikawa , t. , gunn , j. e. , doi , m. , shimasaku , k. , & schneider , d. p. 1996
, , 111 , 1748 gil de paz , a. , et al .
2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0606440 hancock , m. , smith , b. j. , struck , c. , giroux , m. l. , appleton , p. n. , charmandaris , v. , & reach , w. t. 2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0610421 harris , j. , calzetti , d. , gallagher , j. s. , iii , smith , d. a. , & conselice , c. j. 2004 , , 603 , 503 helou , g. , et al .
2004 , , 154 , 253 hill , j. k. , et al .
1995 , , 438 , 181 houck , j. r. , et al .
2004 , , 154 , 211 huang , j .- s .
, et al .
2004 , , 154 , 44 iglesias - pramo , j. , et al .
2006 , , 164 , 38 jarrett , t. h. , et al .
2006 , , 131 , 261 johnson , k. e. , indebetouw , r. , & pisano , d. j. 2003 , , 126 , 101 jonsson , p. 2004 , ph.d .
thesis kennicutt , r. c. 1998 , , 36 , 189 kennicutt , r. c. , et al .
2003 , , 115 , 928 kennicutt , r. c. , lee , j. c. , akiyama , s. , funes , j. g. , & sakai , s. 2005 , aip conf .
proc . 783 : the evolution of starbursts , 783 , 3 kennicutt , r. c. 2006 , iau symposium , 237 , 53 kewley , l. j. , dopita , m. a. , sutherland , r. s. , heisler , c. a. , & trevena , j. 2001 , , 556 , 121 kewley , l. j. , & dopita , m. a. 2002 , , 142 , 35 kravtsov , v. v. 2006 , , 132 , 1248 larsen , s. s. 1999 , , 139 , 393 larsen , s. s. 2000 , , 319 , 893 larsen , s. s. , efremov , y. n. , elmegreen , b. g. , alfaro , e. j. , battinelli , p. , hodge , p. w. , & richtler , t. 2002 , , 567 , 896 larsen , s. s. 2004a , asp conf .
ser . 322 : the formation and evolution of massive young star clusters , 322 , 19 larsen , s. s. 2004b , , 416 , 537 larsen , s. s. , origlia , l. , brodie , j. p. , & gallagher , j. s. 2006 , , 368 , l10 lee , h. , skillman , e. d. , & venn , k. a. 2006 , , 642 , 813 leitherer , c. , et al . 1999 , , 123 , 3 lonsdale , c. j. , et al .
2003 , , 115 , 897 ma , j. , et al .
2006a , , 636 , l93 ma , j. , de grijs , r. , yang , y. , zhou , x. , chen , j. , jiang , z. , wu , z. , & wu , j. 2006b , , 368 , 1443 madden , s. c. , galliano , f. , jones , a. p. , & sauvage , m. 2006 , , 446 , 877 martin , d. c. , et al .
2005 , , 619 , l1 mayya , y. d. , bressan , a. , carrasco , l. , & hernandez - martinez , l. 2006 , , 649 , 172 mihos , j. c. , & hernquist , l. 1994 , , 425 , l13 mihos , j. c. , & hernquist , l. 1996 , , 464 , 641 mirabel , i. f. , et al . 1998 , , 333 , l1 moustakas , j. , kennicutt , r. c. , jr . ,
& tremonti , c. a. 2006 , , 642 , 775 ohalloran , b. , satyapal , s. , & dudik , r. p. 2006
, , 641 , 795 osterbrock , d. e. 1989 , astrophysics of gaseous nebulae and active galactic nuclei ( mill valley : university science books ) pahre , m. a. , ashby , m. l. n. , fazio , g. g. , & willner , s. p. 2004
, , 154 , 235 peeters , e. , spoon , h. w. w. , & tielens , a. g. g. m. 2004 , , 613 , 986 prez - gonzlez , p. g. , et al .
2006 , , 648 , 987 rieke , g. h. , et al .
2004 , , 154 , 25 rossa , j. , van der marel , r. p. , bker , t. , gerssen , j. , ho , l. c. , rix , h .- w . , shields , j. c. , & walcher , c .- j .
2006 , , 132 , 1074 roussel , h. , et al .
2001 , , 372 , 406 silva , l. , granato , g. l. , bressan , a. , & danese , l. 1998 , , 509 , 103 smith , b. j. , struck , c. , appleton , p. n. , charmandaris , v. , reach , w. , & eitter , j. j. 2005 , , 130 , 2117 smith , b. j. , struck , c. , hancock , m. , appleton , p. n. , charmandaris , v. , & reach , w. t. 2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0610562 snijders , l. , van der werf , p. p. , brandl , b. r. , mengel , s. , schaerer , d. , & wang , z. 2006 , , 648 , l25 stoughton , c. , et al .
2002 , , 123 , 485 struck , c. 1999 , , 321 , 1 struck , c. , & smith , b. j. 2003 , , 589 , 157 struck , c. , kaufman , m. , brinks , e. , thomasson , m. , elmegreen , b. g. , & meloy elmegreen , d. 2005 , , 364 , 69 surace , j. a. , et al .
2005 , the swire data release 2 : image atlases and source catalogs for elais - n1 , elais - n2 , xmm - lss , and the lockman hole vacca , w. d. , & conti , p. s. 1992 , , 401 , 543 walcher , c. j. , et al .
2005 , , 618 , 237 walcher , c. j. , bker , t. , charlot , s. , ho , l. c. , rix , h .- w . , rossa , j. , shields , j. c. , & van der marel , r. p. 2006
, , 649 , 692 walker , i. r. , mihos , j. c. , & hernquist , l. 1996 , , 460 , 121 wang , z. , et al .
2004 , , 154 , 193 wang , j. , & wei , j. y. 2006 , , 648 , 158 wang , j. l. , xia , x. y. , mao , s. , cao , c. , wu , h. , & deng , z. g. 2006 , , 649 , 722 werner , m. w. , et al .
2004 , , 154 , 1 whitmore , b. c. , & schweizer , f. 1995 , , 109 , 960 whitmore , b. c. , zhang , q. , leitherer , c. , fall , s. m. , schweizer , f. , & miller , b. w. 1999 , , 118 , 1551 wilson , c. d. , scoville , n. , madden , s. c. , & charmandaris , v. 2000 , , 542 , 120 wilson , c. d. , scoville , n. , madden , s. c. , & charmandaris , v. 2003 , , 599 , 1049 wilson , c. d. , harris , w. e. , longden , r. , & scoville , n. z. 2006 , , 641 , 763 wu , h. , et al .
2002 , , 123 , 1364 wu , h. , cao , c. , hao , c .-
n . , liu , f .- s . ,
wang , j .- l . , xia , x .- y . , deng , z .- g . , & young , c. k .- s .
2005 , , 632 , l79 york , d. g. , et al .
2000 , , 120 , 1579 zhang , q. , fall , s. m. , & whitmore , b. c. 2001 , , 561 , 727 zheng , z. , et al . 1999 , , 117 , 2757 m and mips 24@xmath5 m bands images of arp 24 .
the continuum - subtracted h@xmath0 image was also plotted to emphasize the good correspondence of the peaks between h@xmath0 and 24@xmath5 m emissions .
k0 , k1 , k2 , k3 denote the four infrared bright knots in arp 24 selected from the 24@xmath5 m image as we described in the text ( see @xmath133.1 ) .
[ fig2 ] ] m flux .
the yscc is similar to the prototype starburst galaxy m 82 ( dotted line ) , and the ir knots are comparable to late - type scd galaxy ngc 6946 ( dashed line ) .
note the flux densities of the ir knots have been divided by a factor 10 ( 1 dex ) for clarity .
[ fig3],title="fig : " ] m flux .
the yscc is similar to the prototype starburst galaxy m 82 ( dotted line ) , and the ir knots are comparable to late - type scd galaxy ngc 6946 ( dashed line ) .
note the flux densities of the ir knots have been divided by a factor 10 ( 1 dex ) for clarity .
[ fig3],title="fig : " ] radius apertures used in the sed studies ( see @xmath133.2 ) , and the green ones shown in the right image correspond to 0.2@xmath2 radius apertures on the f300w image .
[ fig4],title="fig : " ] radius apertures used in the sed studies ( see @xmath133.2 ) , and the green ones shown in the right image correspond to 0.2@xmath2 radius apertures on the f300w image .
[ fig4],title="fig : " ] ) between 0.1 and 120 m@xmath12 and metallicities z=0.02 & 0.008 are also shown in this figure .
evolutionary lines are drawn for cluster masses of 10@xmath3 and 4@xmath1010@xmath27 m@xmath4 .
model ages along the evolutionary lines are indicated by * @xmath74 * ( z=0.02 ) and * + * ( z=0.008 ) placed in t=1 myr intervals and begin from 1 myr on the left .
[ fig5 ] ] ( 5500)=1 ) , which indicates a mixture of populations of different ages .
strong h@xmath29 absorption indicates the existence of a large number of evolved a - type stars ( age @xmath1 10@xmath30yr ) , and strong h@xmath0 emission traces ongoing star formation .
a template spectrum corresponds to an instantaneous - burst model with a young population of 100 myr and metallicity z=0.02 plus an old population of 11 gyr was also plotted for comparison ( in green ; derived from bruzual & charlot 2003 ; note the template has been convolved to match the observed spectrum ) .
[ fig6 ] ] notes.@xmath75the line fluxes of the high resolution spectrum ( high res ) has been scaled to the low resolution one ( low res ) using h@xmath0 line fluxes .
this correction will not affect much on the line ratios used in @xmath133.3 . | we made a multi - wavelength study of young massive star clusters ( yscs ) in the interacting galaxy arp 24 , using the optical and ultraviolet images from _ hubble space telescope _
( _ hst _ ) , sloan digital sky survey , and _ galaxy evolution explorer _ ; the mid - infrared images from _ spitzer space telescope _ ; and the narrow - band h@xmath0 image and optical spectra from the naoc 2.16 m telescope . based on the _ hst _ images
, we found that the brightest infrared knot in arp 24 is associated with a complex of five young massive star clusters , within a region of @xmath1 0.95@xmath2 radius ( 127pc ) in size .
the ages and masses of the star clusters in this complex and other regions were estimated using _
hst _
broadband photometries and the starburst99 synthesis models .
the star clusters in this complex are very young ( within ages of @xmath1 3 - 5 myr ) and massive ( masses of @xmath110@xmath3m@xmath4 ) .
the ionization parameter and metallicity of the complex were estimated using the emission line ratios , and the star formation rates were calculated using monochromatic 24@xmath5 m , fuv , and h@xmath0 line luminosities .
we speculate that arp 24 may formed by a retrograde fly - by encounter indicated by its one - armed appearance and fan - like structure , and the formation of the yscs in this galaxy is triggered by the interaction .
the clusters in the ysc complex may formed in a single giant molecular cloud simultaneously . from the ultraviolet to mid - infrared spectral energy distributions , we found that the region of the ysc complex is relatively bluer in optical and has higher 24@xmath5 m dust emission relative to the starlight and 8@xmath5 m emission .
this warm infrared color may due to strong uv radiation field or other mechanisms ( e.g. , shocks ) within this region which may destroy the polycyclic aromatic hydrocarbons and enhance the small grain emission at 24@xmath5 m . |
since its recent discovery , graphene a monolayer of graphite has become the subject of intense research due to its peculiar electronic properties @xcite .
the unusual properties of graphene are largely a result of the linear dispersion of electron bands at low energies and its two - dimensionality .
this allows to describe the charge carriers in graphene as a two - dimensional gas of massless dirac - fermions .
the unusual electronic properties and the fact that its chemical potential can easily be tuned by a gate voltage make graphene an ideal experimental testing ground for exotic physics as well as a promising basis for novel nano - electronics devices . for recent reviews of this rapidly developing field and additional references , see e.g. refs . .
graphene is also a good testing ground for exotic kondo physics@xcite : first , the linearly vanishing dos of graphene favours the formation of a local moment @xcite which is a necessary condition for kondo physics . also vacancies and edge defects in graphene have been predicted to give rise to magnetic moments @xcite .
very recently , the kondo effect has actually been observed at local defects in graphene providing evidence that defects in graphene are indeed magnetic @xcite . on the other hand , as a result of the linearly vanishing dos , a magnetic moment coupled to a graphene layer can only display a kondo effect if the strength of the coupling is strong enough@xcite .
furthermore because of the existence of two inequivalent dirac cones , it has multiple channels which can in principle screen the impurity spin and could therefore lead to a multichannel kondo effect and hence non - fermi liquid behaviour @xcite .
kondo effect and the possibilty of multichannel kondo effect in graphene have been discussed in a number of recent publications @xcite . moreover , it is possible to probe the kondo effect of magnetic adatoms on graphene by scanning tunneling microscopy ( stm ) .
similar to stm spectroscopy of magnetic adatoms and molecules on metallic surfaces @xcite , the kondo effect could show up as fano lineshapes in the conductance vs. bias voltage characteristics of the stm or as additional temperature dependent peaks in the stm spectra .
but in comparison with normal metallic surfaces the fano lineshapes for graphene are expected to be more asymmetric and strongly dependend on the symmetry of the @xmath1-orbitals of magnetic atom @xcite .
the goal of this work is to perform a realistic study of the kondo effect for a single co adatom on an otherwise perfect graphene sheet by using a recently developed approach for nanoscopic systems that combines _ ab initio _ electronic structure calculations on the level of density functional theory ( dft ) calculations with a full dynamical treatment of the strongly interacting @xmath0-electrons @xcite .
a realistic treatment allows us to investigate under which conditions the kondo effect takes place in a given orbital and the detailed spectral functions that arise from the kondo effect in a semi - metallic substrate .
this paper is organized as follows : in sec .
[ sec : method ] we briefly describe our recently developed gga+oca
method@xcite for the quasi _
ab initio _ description of the electronic structure of nanoscopic systems fully taking into account dynamical correlations by combining dft calculations with a dynamical treatment of the strongly interacting @xmath0-electrons in the so - called one - crossing approximation ( oca ) . in sec .
[ sec : dft - results ] we present dft electronic structure calculations in the generalized gradient approximation ( gga ) for determining the most favourable adsorption site for the co atom on the graphene sheet . in agreement with previous results by other groups@xcite
we find that the hollow site is the most favourable adsorption site .
using hybrid functional calculations which generally predict better geometries for molecules we have then optimized the adsorption height of the co atom at the hollow site of the graphene . using gga
we have also calculated the charge and spin of the co atom in dependence of the distance to the graphene sheet for the three different adsorption sites .
the plots indicate that depending on the distance the co atom can be either in the local moment regime or in weak correlation regime .
since the potential energy curves are very shallow it seems feasible that the correlation regime be controlled via the distance of the co atom to the graphene substrate in an actual experiment . in sec .
[ sec : oca - results ] we study the impact of dynamical correlations of the co @xmath0-electrons on the elctronic structure of the co atom and the graphene sheet using the gga+oca method for nanoscopic systems as described in sec . [ sec : method ] .
we find that the dynamical correlations of the co @xmath0-electrons give rise to strongly temperature - dependent peaks near the fermi level in the spectral density of the co @xmath0-shell but also in the spectral density of the @xmath2-orbitals of nearby carbon atoms .
furthermore , we find that the co @xmath0-spectra depend quite strongly on the exact energies of the co @xmath0-levels . accordingly
, the application of a gate voltage which changes the chemical potential of the graphene substrate has a similar effect on the spectra .
hence the dynamic correlations and the kondo effect can be controlled by the application of a gate voltage . finally , in sec .
[ sec : conclusions ] we conclude this work with a general discussion of the results in the light of kondo physics .
we consider a _ single _
co adatom adsorbed at the hollow site of an otherwise perfect graphene sheet as shown in fig .
[ fig : graphene+co ] . to proceed we divide the system into two parts .
the central region c which contains the co adatom and a number of carbon atoms is embeded into the host material h given by a perfect graphene sheet .
the central region c has to be chosen big enough so that the electronic structure outside of the region can be assumed to be that of a perfect graphene sheet . as a first step a mean - field description of the electronic structure of the system
is obtained from dft calculations on the level of gga using the crystal06 _ ab initio _ electronic structure code@xcite together with the 6 - 31 g gaussian basis set : by setting up a 2d lattice of supercells of the c region we obtain an effective one - body hamiltonian @xmath3 for the c region from the kohn - sham hamiltonian of the converged gga calculation .
analogously , we obtain an effective mean - field description of the host material from a gga calculation for a supercell of clean graphene corresponding to the c region .
the green s function of the central unit cell c is then given by : @xmath4 @xmath5 is the _ embedding self - energy _ which describes the dynamic hybridization between the central region c and the host material h. the embedding self - energy is obtained from the gga calculation of the perfect graphene sheet as explained in app .
[ app : embedding ] .
@xmath6 is the electronic self - energy describing the dynamic correlations of the strongly interacting @xmath0-electrons .
the strong interactions of the co @xmath0-shell are captured by adding a hubbard - like interaction term to the one - body hamiltonian within the correlated @xmath0-subspace : @xmath7 @xmath8 are the matrix elements of the effective coulomb interaction of the @xmath0-electrons which is different from the bare coulomb interaction due to the screening by the conduction electrons . here
we take a model interaction taking into account only the direct coulomb repulsion @xmath9 between electrons and a hund s rule coupling @xmath10 . for @xmath0 transition metal elements in bulk materials
the repulsion @xmath9 is around 2 - 3 ev and @xmath10 is around 0.9 ev @xcite . due to the lower coordination of the co atom on the graphene sheet the screening
should be reduced compared to its bulk value . here
we take @xmath11 ev and @xmath12 ev as in our previous work@xcite .
the coulomb interaction within the correlated @xmath0 subspace has already been taken into account on a static mean - field level in the effective ks hamiltonian of the central unit cell c. therefore the ks hamiltonian within the correlated subspace @xmath13 has to be corrected by a double - counting correction term , i.e. @xmath14 . here
we use the standard expression , @xmath15\;\times\;\mathbf{1}_{3d}\ ] ] where @xmath16 is the identity matrix in the co @xmath0-subspace , and @xmath17 is the occupation of the impurity @xmath0-shell @xcite . in order to compute the electronic self - energy @xmath18 we have to solve the generalized anderson impurity problem given by the strongly interacting co @xmath0-electrons .
the impurity problem is completely determined by the interacting hamiltonian of the co @xmath0-shell , i.e. @xmath19 , and by the so - called hybridization function @xmath20 which describes the dynamic hybridization of the co @xmath0-shell with the rest of the system ( i.e. the bath )
. the hybridization function can be extracted from the non - interacting green s function @xmath21^{-1 } \equiv [ \mathbf{g}_{\rm c}]^{-1}+\mathbf\sigma_{3d}(\omega)$ ] of the central region c as follows : @xmath22^{-1}\ ] ] @xmath23 is the non - interacting green s function projected onto the @xmath0-shell of the co atom , i.e. @xmath24 . solving the generalized anderson impurity problem is a difficult task , and at present there is no universal impurity solver that works efficiently and accurately in all parameter regimes . here
we make use of impurity solvers based on an expansion in the hybridization strength given by @xmath25 around the atomic limit .
the starting point is an exact diagonalization of the ( isolated ) impurity subspace i.e. the co @xmath0-shell in our case given by the interacting hamiltonian @xmath26 .
the hybridization of the impurity subspace with the rest of the system ( given by the hybridization function @xmath25 ) is then treated perturbatively .
the so - called non - crossing approximation ( nca ) is a self - consistent perturbation expansion to lowest order in the hybridization strength .
nca only takes into account the most simple diagrams describing simple hopping processes where an electron or hole hops into the impurity at some time and then out at a later time ( see fig .
[ fig : nca+oca ] in app .
[ app : nca+oca ] ) . the so - called one - crossing approximation ( oca )
improves upon the nca by taking into account second order diagrams where two additional electrons ( holes ) are acommodated on the impurity at the same time in addition to the nca diagrams as shown in fig .
[ fig : nca+oca ] .
oca is the lowest order self - consistent approximation that is exact up to first order in the hybridization @xmath27 for all physical quantities while nca is only exact to first order in @xmath28 for the conduction electron self - energy .
oca improves considerably many of the shortcomings of nca : it substantially improves the width of the kondo peak and hence the kondo temperature which now are only slightly underestimated .
it also corrects the asymmetry of the kondo peak . for very low temperatures ( @xmath29 ) , however , the height of the kondo peak is still overestimated , and the fermi liquid behaviour at zero temperature is not recovered .
hence , oca is a reasonable approximation for solving the generalized impurity problem as long as the temperatures are not too low ( i.e. more than one order of magnitude below @xmath30 ) . in app .
[ app : nca+oca ] we give a brief introduction to the nca and oca impurity solvers .
a detailed description of the nca and oca methods can be found e.g. in refs . .
first , we have calculated the adsorption curves for a co atom on a graphene sheet with density - functional calculations within the generalized gradient approximation ( gga ) with the crystal06 @xcite _ ab initio _ electronic structure program for periodic systems employing an elaborate all - electron gaussian basis set ( 6 - 31 g ) . [ cols="^ " , ] due to the c6-symmetry of the system the co @xmath0-shell splits into three symmetry groups : the non - degenerate a@xmath31 group consisting of the rotationally invariant @xmath32-orbital only , the doubly degenerate e@xmath31 group consisting of the @xmath33- and @xmath34-orbitals and the e@xmath35 group consisting of the @xmath36- and @xmath37-orbitals . the right panel of fig .
[ fig : graphene+co ] shows the hybridization functions for each symmetry group .
as can be seen at low energies the @xmath32-orbital ( a@xmath31 ) has an essentially flat and zero hybridization with the graphene sheet .
this can be understood by the fact that the coupling of the @xmath32 on the hollow site to the @xmath38-orbitals of the graphene sheet is zero due to symmetry considerations @xcite .
the e@xmath35-group has the ( overall ) strongest coupling to the c @xmath38-orbitals in that geometry as can be seen from the corresponding hybridization function .
note that the hybridization function for both the e@xmath31 and e@xmath35 symmetry group vanish linearly near the fermi energy reflecting the linear dos of the graphene sheet .
interestingly , the slopes for the e@xmath31- and e@xmath35-hybridization functions are different for energies below and above the fermi level .
morover , for negative energies the e@xmath31-hybridization function has a steeper slope than the e@xmath35-hybridization , while for positive energies it is the other way round .
this implies that e@xmath31 couples predominantly to the negative energy band of graphene while @xmath39 couples predominantly to the positive energy band . in table .
[ tab : crystal - fields ] , we show the energy levels , the real part of the hybridization function at the fermi level , and the effective energy levels of the co @xmath0-shell . while the splitting of the bare energy levels @xmath40 is around 0.2 ev , the splitting of the effective energy levels @xmath41 taking into account the ( real part of the ) hybridization with the graphene is somewhat bigger with around 0.4 ev .
taking into account the real part of the hybridization changes also the ordering of the energy levels .
for the ordering of the effective energy levels we obtain : @xmath42 . note that this level ordering is similar to the one obtained by wehling _
et al._@xcite .
we now solve the generalized impurity problem of the co @xmath0-shell coupled to the graphene sheet by using the oca impurity solver as described in sec .
[ sec : method ] and in app .
[ app : nca+oca ] .
we take the coulomb repulsion @xmath9 between @xmath1-orbitals to be somewhat higher than in bulk due to the supposedly smaller screening : @xmath11 ev .
the hund s rule coupling is taken to be the same as in bulk , i.e. @xmath12 ev since it is less affected by screening effects .
the dcc was calculated by the standard expression given by eq .
( [ eq : hdcc ] ) and taking the gga value of about 7.5 for @xmath17 . with oca , the total occupation of the impurity @xmath0-shell is now between 7.1 and 7.0 at low temperatures compared to the gga value of 7.5 .
this result is very typical for correlated calculations : the correlations push the system towards integer occupation numbers to lower the energy . fig .
[ fig : oca - results ] shows the spectral functions of the co @xmath0-electrons for different temperatures calculated with oca . at high temperatures we find peaks near the fermi level for all three symmetry groups .
however , the temperature dependencies of the individual peaks are quite different from each other : for the a@xmath31-symmetry consisting of the @xmath43-orbital the peak becomes smaller at increasing temperatures and finally develops a step - like behaviour at low temperatures .
the peaks in the e@xmath31 and e@xmath35-spectra on the other hand become more pronounced with decreasing temperature more or less as expected for kondo effect .
the form of the spectral functions and the temperature behaviour can be understood by looking at the occupation numbers for the different symmetry groups .
the insets of figs .
[ fig : oca - results](a - c ) show the occupation numbers as a function of temperature for each of the three symmetry groups . for the a@xmath31-symmetry
the occupation changes from @xmath441.6 at high temperatures to almost 2 at low temperatures .
hence , at low temperatures the @xmath43-orbital is practically completely filled , and excitations near the fermi level are suppressed . for the e@xmath31-symmetry group composed of the @xmath45- and @xmath46-orbitals the occupation changes from 2.8 at high temperatures to 2.1 at low temperatures .
hence at low temperatures this doubly - degenerate channel is essentially half - filled , corresponding to a spin of almost 1 in this channel due to hund s rule coupling .
the peak near the fermi level in the e@xmath31-spectral function is hence due to a spin-1 kondo effect in this channel . for the e@xmath35-symmetry composed of the @xmath47- and @xmath48-orbitals we find an occupation of 2.8 for high temperatures , and of almost 3 for low temperatures . hence the doubly - degenerate e@xmath35-channel accomodates 3 electrons with a total spin-1/2 so that the peak in the e@xmath35-spectral function is due to a spin-1/2 kondo effect .
accordingly , the peaks near the fermi level in the @xmath0-spectral function at low temperature are due to a spin-3/2 kondo effect in the @xmath0-levels .
this interpretation is further corroborated by the fact that the main contribution to the electronic configuration of the @xmath0-shell is an eightfold degenerate state with @xmath49 and @xmath50 .
the very fact that such a high spin state can be screened by the graphene substrate is an evidence of the multi - channel character of the graphene as a conduction electron bath . in fig .
[ fig : oca - results](e ) we compare the @xmath0-spectral density calculated with the gga+oca ( at @xmath51 k ) method on the one hand and with spin - polarized gga on the other hand .
as can be seen the spectral densities calculated by the two methods differ significantly from each other . in the gga spectrum , the peak just below the fermi level ( see also the inset of fig . [
fig : oca - results](e ) ) corresponds to a completely filled @xmath52-shell with down - spin as can be seen by comparison with fig .
[ fig : graphene+co](c ) .
the peak above ( and also a little farther from ) the fermi level corresponds to a completely empty spin - down @xmath39-shell .
both peaks carry essentially the full spectral weight of the correpsonding orbitals . in contrast
, the two peaks near the fermi level in the gga+oca spectra which are also of @xmath52- and @xmath39-symmetry , do not carry the full spectral weight of the corresponding orbitals due to the renormalization of the quasi - particle by the electron - electron interactions which can not be captured by a static mean - field calculation like gga .
[ fig : oca - results](d ) shows that the correlations in the co @xmath0-shell also affect the projected dos of the @xmath2-orbitals on the carbon atoms that are directly coupled to the co atom ( i.e. the 6 carbon atoms nearest to the co atom ) .
this can be understood as follows : due to the coupling between the carbon @xmath2- and the co @xmath0-orbitals , the full @xmath53-green s function @xmath54 is related to the full green s of the @xmath0-orbitals @xmath55 by : @xmath56 where @xmath57 is the green s function of the ( unperturbed ) graphene @xmath2-orbitals without the coupling to the co @xmath0-orbitals and @xmath58 is the coupling between the @xmath2- and individual @xmath0-orbitals .
the full @xmath0-green s function @xmath55 takes into account the correlations given by @xmath6 _ and _ the hybridization with the @xmath2-orbitals given by @xmath25 , i.e. @xmath59 represents the full @xmath60-matrix for the @xmath2-electrons .
one can see a temperature - dependent peak roughly at the position of the peaks in the e@xmath31- and e@xmath35-spectral density . for comparison
we also show the @xmath2-dos when the self - energy of the co @xmath0-electrons @xmath18 is set to zero . in this case
the @xmath0-levels are far below the fermi level and hence can not affect the @xmath2-dos near the fermi level .
hence the sharp resonance in the @xmath2-dos vanishes .
this shows that the kondo peaks in the co @xmath0-spectra can also be detected indirectly by spectroscopy of the nearby carbon atoms .
-energy level dependence .
( a - c ) spectral functions of co @xmath0-levels for different energy shifts @xmath61 of the @xmath0-levels ( as indicated in the legend of ( c ) ) at @xmath62 k and for the three different symmetries . ( d , e ) orbital occupancies of the co @xmath0-levels as a function of the energy shift for each of the three symmetries and for @xmath62 k. ] the double - counting correction term @xmath63 given by eq .
( [ eq : hdcc ] ) which substracts out the coulomb interaction within the correlated @xmath0-subspace that has already been taken into account on a static mean - field level in the lda or gga calculation is not known exactly .
hence the exact position of the @xmath0-energy levels is not known . here
we explore how shifting the @xmath0-energy levels changes the results
. figs .
[ fig : ed - dependence](a - c ) show the co @xmath0-spectral densities for each of the different symmetries and for different shifts @xmath61 of the @xmath0-levels with respect to their original position , and figs . [
fig : ed - dependence]d , e the corresponding occupancies .
quite obviously there is a drastic change in the correlations when the impurity - levels are lowered by more than 0.2 ev . at that point , a substantial peak appears in the a@xmath31-spectra , while the peak in the e@xmath31-spectra disappears .
the peak in the e@xmath35-spectra on the other hand becomes qualitatively different for energy shifts @xmath64 ev , i.e. their spectral weight increases considerably
. the changes in the spectra are accompanied by changes in the occupancies : most importantly , the occupancy of the e@xmath35-levels increases abruptly by about 0.6 for @xmath65 ev .
the changes in the @xmath0-levels with a@xmath31- and e@xmath31-symmetry are less pronounced but also occur at the same point .
also the overall occupation of the impurity levels changes quite abruptly by about 0.5 electrons from @xmath66 to @xmath67 .
this indicates a strong change in the electronic configuration of the co @xmath0-shell .
indeed we find that for an energy shift around -0.2 ev , another @xmath0-shell state with @xmath68 and @xmath69 becomes similar in energy to the state with @xmath49 and @xmath50 .
hence , at this point the system enters the mixed - valence regime .
this explains the strong increase of the spectral weight in the e@xmath35-peak for @xmath64 ev since the mixed - valence regime is characterized by a strong peak near the fermi level . on the other hand ,
the integer occupation of 2 in the e@xmath31-levels , and the associated spin-1 , essentially does not change upon entering the mixed - valence regime .
in fact , the occupation is now even closer to 2 than it was in the kondo regime .
also the occupation of the a@xmath31-level remains essentially constant upon entering the mixed - valence regime .
hence the mixed - valence is actually only induced in the e@xmath35-levels since the occupations of the a@xmath31-level and the e@xmath31-levels in the @xmath68 state are the same as in the @xmath49 state .
the vanishing of the peak in the e@xmath31-spectra upon entering the mixed - valence regime is due to the weaker hybridization with the graphene for negative energies ( compare fig .
[ fig : graphene+co ] ) : as the kondo peak in the e@xmath31-levels moves beyond the dirac point as the impurity - levels are lowered in energy , the kondo coupling in the e@xmath31-levels becomes weaker and the kondo temperature much smaller so that the kondo effect is not observed anymore at the temperatures considered here .
it is not quite clear what triggers the peak in the a@xmath31-spectra upon entering the mixed - valence regime .
one possibility is that the increased charge fluctuations in the mixed valence regime also lead to charge fluctuations in this level via the hund s rule coupling to the other @xmath0-levels .
the slight decrease in the occupation of the a@xmath31-level in the mixed - valence regime supports this view .
note that the position of a peak ( if present ) changes according to the energy shift .
this also explains the change in the width of the peaks for the e@xmath31- and e@xmath35- symmetries : as a peak moves away from the fermi level the broadening increases due to the increased hybridization with the graphene substrate .
the a@xmath31-level on the other hand does not couple at all to the graphene substrate ( zero hybridization ) so that the changes in the shape of the a@xmath31-peak are purely correlation effects . hence tuning the energies of the co @xmath0-levels
changes the correlations in these levels due to changes in the occupancy .
more precisely , the system is driven from the kondo regime to the mixed - valence regime when the energies of the @xmath0-levels are lowered by more than 0.2 ev with respect to their original energies .
the change in the correlations is reflected by corresponding changes in the @xmath0-spectral functions . in the previous subsection
we have seen that changing the exact position of the co @xmath0-levels with respect to the fermi level has a considerable effect on the electronic correlations and hence on the spectral density of the @xmath0-levels .
although changing the impurity levels directly is not feasible in an experiment , one can instead easily tune the chemical potential of graphene by application of a gate voltage .
the chemical potential then is shifted with respect to both , the impurity levels and the graphene bands .
hence it should be possible to control the electronic correlations and in particular the kondo effect by application of a gate voltage .
the change in correlations can then be detected in the @xmath0-spectra for example by stm spectroscopy .
more precisely , a gate voltage @xmath70 shifts the chemical potential @xmath71 with respect to the graphene bands and the impurity levels according to @xmath72 where @xmath73 denotes the elementary charge and @xmath74 is the fermi level of neutral graphene , i.e. the energy of the dirac points ( see inset of fig .
[ fig : gate - voltage]a ) .
overall the results for shifting the chemical potential presented in fig .
[ fig : gate - voltage ] look very similar to the previous ones for shifting the impurity levels in fig .
[ fig : ed - dependence ] . at the hollow site only the co @xmath0-levels of e@xmath31- and e@xmath35-symmetry couple to the graphene sheet for low energies while the a@xmath31-level is completely decoupled .
hence only the spectra of e@xmath31- and e@xmath35-levels can be measured by an stm . therefore in fig .
[ fig : gate - voltage](a ) , we show the total spectral density of the co e@xmath31- and e@xmath35-levels for different values of the gate voltage . for gate voltages smaller than -0.1 ev
there is only one peak in the spectral function which is due to the e@xmath35-levels . for gate voltages @xmath75 ev
a second peak appears which originates from the e@xmath31-levels .
the e@xmath35-peak and also the e@xmath31-peak ( when present ) move with respect to the chemical potential @xmath71 according to the change in gate voltage .
hence the peaks in the spectral function are not pinned to the chemical potential as expected for normal kondo effect , but are rather pinned to the dirac points of the graphene sheet . for completeness
we also show the spectral density of the a@xmath31-level in fig .
[ fig : gate - voltage]b for different gate voltages .
a peak in the spectral density at low temperature is only observed for negative bias voltages .
the peak disappears for zero and positive gate voltage .
also the resonance in the a@xmath31-level appears to be pinned to the dirac point of graphene rather than to the chemical potential . as can be seen in figs .
[ fig : gate - voltage]d , e also the individual occupations of the @xmath0-levels as a function of the applied gate voltage show a similar behaviour as before as a function of the energy shift @xmath61 of the @xmath0-levels .
in fact , the explanation for the behaviour of the spectra and occupancies as a function of the energy shift of the @xmath0-levels also applies to their behaviour as a function of the gate voltage . here , when the gate voltage becomes smaller than -0.1 v the system leaves the kondo regime with an @xmath49 and @xmath50 state as the principal contribution to the electronic configuration of the @xmath0-shell , and enters the mixed - valence regime with an equal contribution coming from an @xmath68 and @xmath69 state .
we can estimate a kondo temperature from the width of the resonance in the kondo regime .
[ fig : gate - voltage]c shows the kondo temperature of the e@xmath35-levels estimated from the resonance width @xmath76 using the relation @xmath77 known from the fermi liquid theory of the anderson model where @xmath78 is wilson s number 0.4128 ( see e.g. the book by hewson@xcite , chap .
5 ) . in fig .
[ fig : gate - voltage]c we have also included `` kondo temperatures '' estimated from the width of the peaks for negative gate voltages where the system is not in the kondo regime anymore . hence outside the kondo regime it is only a measure for the width of the resonance near the chemical potential . note that in the kondo regime ( i.e. for @xmath79 ) the width of the resonance and hence the kondo temperature are essentially constant .
this behaviour is different from the one before where we only had shifted the impurity levels .
there the width of the e@xmath31- and e@xmath35-peaks increased as they moved away from the fermi level due to the linear increase of the hybridization with the graphene substrate . here , however , the peak is pinned to the dirac point of graphene , i.e. it moves together with the graphene bands with respect to the chemical potential .
hence the hybridization with the graphene conduction electrons does not change here .
the strong dependence of the width of the resonance on the gate voltage for negative gate voltages on the other hand is due to the changes in the electronic correlations as the system is driven out of the kondo regime .
using dft electronic structure calculations we have found that a single co atom most likely adsorbs at the hollow site of an otherwise clean graphene sheet .
furthermore the dft calculations show , that the magnetic moment and the electronic configuration of the @xmath0-shell of the co atom depends quite strongly on the adsorption height over the graphene sheet . at large distantes
the formation of a local moment is favoured while at short distances the magnetic moment is quenched due to strong hybridization with the substrate .
since the adsorption energy curves are quite shallow , it might be possible in an experiment to push the co atom from the local moment to the weakly correlated regime .
using our gga+oca correlated electronic structure method for nanoscopic systems , we have then studied the impact of dynamical correlations due to the strongly interacting co @xmath0-electrons on the electronic structure .
because of the strong crystal field splitting of the @xmath0-levels by graphene substrate , the behaviour of each of the co @xmath0-levels strongly depends on its orbital symmetry .
we find that the dynamic correlations can give rise to a spin-3/2 kondo effect .
the spin-3/2 is composed of a spin-1 due to two electrons within the doubly - degenerate orbitals of e@xmath31-symmetry and a spin-1/2 due to a hole within the doubly - degenerate orbitals of the e@xmath35-symmetry .
the kondo effect gives rise to corresponding peaks near the fermi level in the e@xmath31- and e@xmath35-spectral function .
the a@xmath31-channel on the other hand is nearly completely filled and hence in the kondo regime , it does not give rise to a peak near the fermi level .
the realization of a spin-3/2 kondo effect points to the multi - channel character of graphene .
hence a spin-1/2 impurity could in principle give rise to multi - channel kondo physics and non - fermi liquid behaviour .
the kondo effect can be controlled by a gate voltage which changes the chemical potential and thereby the filling of the @xmath0-levels .
the change in the correlations can be observed by a stark change in the spectra as the system is driven out of the kondo regime and into the mixed valence regime .
another remarkable result is that the kondo peaks appear to be pinned to the dirac point of the graphene substrate rather than to the chemical potential when a gate voltage is applied .
this last finding is actually in agreement with very recent stm spectroscopy measurements of co atoms on graphene which find peaks in the di / dv curves that move according to the applied gate voltage@xcite .
but it has not yet been pointed out by other theoretical work considering the dependence of the kondo effect in graphene on a gate - voltage @xcite .
we are grateful to sasha balatzky , kostyantyn kechedzhi and rgis decker for useful discussions , and to kristjan haule for providing us with the oca impurity solver .
part of this research was carried out at the kitp in the program `` towards material design with correlated electrons '' .
gk was supported by nsf - dmr-0906943 .
here we show how to compute the embedding self - energy @xmath80 given in eq .
( [ eq : gc ] ) that describes the dynamic hybridization of the central region c with the rest of the system ( i.e. the graphene sheet ) .
first , we obtain an effective mean - field description of the host material from a gga calculation for a supercell of clean graphene corresponding to the c region . from the kohn - sham hamiltonian @xmath81 of the host material in reciprocal space
we calculate the @xmath82-dependent green s function : @xmath83 the local ks hamiltonian of the supercell is obtained by summing up @xmath81 over all k - points within the brillouin zone : @xmath84 similarly , the local green s function of the supercell is given by summing up @xmath85 over all k - points within the brillouin zone : @xmath86 on the other hand , we can write the local green s function also in terms of the local hamiltonian @xmath87 of the supercell and the embedding self - energy ( which describes the dynamic hyrbidization with the rest of the graphene ) as : @xmath88^{-1}$ ] . hence the embedding self - energy is given by : @xmath89^{-1}\ ] ]
the general multi - orbital anderson impurity model can be written in the following form : @xmath90 where in order to keep the notation simple we have combined the spin- and orbital degrees of freedom into one index for each impurity level @xmath91 and each band @xmath92 . the non - crossing approximation ( nca ) and the one - crossing approximation ( oca ) both solve the anderson impurity model by expansion in the hybridization strength around the atomic limit .
the starting point is an exact diagonalization of the impurity subspace , i.e. of the co @xmath0-shell in our case , including the hubbard - like interaction term : @xmath93 where @xmath94 are the many - body eigenstates of @xmath26 and @xmath95 the respective eigen energies .
one now introduces auxiliary fields @xmath96 , @xmath97 ( called pseudo - particles ) such that each impurity state is represented by a corresponding pseudo - particle : @xmath98 where @xmath99 is the pseudo - particle vacuum .
the completeness of the impurity eigenstates imposes the following constraint : @xmath100 the physical electron operators @xmath101 can now be expressed by the pp operators : @xmath102 where @xmath103 are the matrix elements of the impurity - electron creation operator . for later convenience
we also define the corresponding matrix elements of the impurity - electron destruction operator as : @xmath104 .
the anti - commutation rules for the physical electron operators then require that the pp @xmath96 is a boson ( fermion ) if the corresponding state @xmath94 contains an even ( odd ) number of electrons . in the pp representation we can now rewrite the hamiltonian of the generalized anderson impurity model as follows : @xmath105 where we have included the constraint @xmath106 into the hamiltonian .
the corresponding lagrange multiplier @xmath107 can be interpreted as a ( negative ) chemical potential for the pps . in the pp picture , the hybridization with the bath electrons given by the last term in eq .
( [ eq : pp - aim ] ) becomes now the interaction for the pps .
because of the fermionic and bosonic commutation rules for the pps , one can now develop a diagrammatic perturbation expansion in the pp interaction .
the pp propagators can then be written as @xmath108 where @xmath109 is the pp self - energy describing the dynamic interaction with the other pps .
the non - crossing approximation consists in taken into account the diagrams shown in the first row of fig .
[ fig : nca+oca ] for a certain pseudo - particle .
the nca diagrams describe processes where a single electron ( hole ) jumps from the bath to the impurity and back thereby temporarily creating a pp with n+1 ( n-1 ) electrons .
the nca equations correspond to a self - consistent perturbation expansion to lowest order in the hybridization function @xmath110 .
since the fermionic self - energies depend on the dressed bosonic propagators , and vice versa , the nca equations have to be solved self - consistently .
once the nca equations are solved the physical quantities can be calculated from the pp self - energies .
oca takes into account second order diagrams where two bath electron lines cross as shown in the right column of fig .
[ fig : nca+oca ] .
the self - energies for the fermions ( bosons ) again depend on the full propagators for bosons ( fermions ) , and hence the oca equations have to be solved self - consistently .
the explicit expressions for the oca self - energies are second in the bath hybridization function @xmath111 .
the oca is the lowest order self - consistent approximation that is exact up to first order in the hybridization @xmath27 for all physical quantities while nca is only exact to first order in @xmath28 in the pp and conduction electron self - energies .
further details of the nca and oca impurity solver can be found e.g. in refs . .
0 k. s. novoselov , a. k. geim , s. v. morozov , d. jiang , y. zhang , s. v. dubonos , i. .v .
grigorieva , and a. a. firsov , science * 306 * , 666 ( 2004 ) .
k. s. novoselov , a. k. geim , s. v. morozov , d. jiang , m. i. katsnelson , i. .v .
grigorieva , s. v. dubonos , and a. a. firsov , nature * 438 * , 197 ( 2005 ) .
, y. zhang , y .- w . tan , h. l. stormer , and p. kim , nature * 438 * , 201 ( 2005 ) .
k. s. novoselov , e. mccann , s. v. morozov , v. i. falko , m. i. katsnelson , u. zeitler , d. jiang , f. schedin , and a. k. geim , nat .
* 2 * , 177 ( 2006 ) . c. w. j. beenakker , rev .
phys . * 80 * , 13371354 ( 2008 ) .
a. h. castro neto , f. guinea , n. m. r. peres , k. s. novoselov , and a. k. geim , rev .
phys . * 81 * , 109162 ( 2009 ) .
m. a. h. vozmediano , m. i. katsnelson , f. guinea , arxiv:1003.5179v1 , unpublished .
a. c. hewson , _ the kondo problem to heavy fermions _ ( cambridge university press , 1993 ) .
b. uchoa , v. n. kotov , n. m. r. peres , and a. h. castro neto , phys .
* 101 * , 026805 ( 2008 ) .
p. o. lehtinen , a. s. foster , y. ma , a. v. krasheninnikov , and r. m. nieminen , phys .
lett . * 93 * , 187202 ( 2004 ) .
j. j. palacios , j. fernandez - rossier , and l. brey , phys .
b * 77 * , 195428 ( 2008 ) .
j. h. chen , w. g. cullen , e. d. williams , and m. s. fuhrer , unpublished ( 2010 ) , arxiv:1004.3373 .
d. withoff and e. fradkin , phys .
64 , 1835 ( 1990 ) .
m. hentschel and f. guinea , phys .
b * 76 * , 115407 ( 2007 ) .
k. sengupta and g. baskaran , phys .
b * 77 * , 045417 ( 2008 ) .
p.d . sacramento and p. schlottmann , phys .
b * 43 * , 13294 ( 1991 ) .
p.d . sacramento and p. schlottmann , adv . phys . * 42 * , 641 ( 1993 ) .
d. l. cox and a. zawadowski , adv .
phys . * 47 * , 599 ( 1998 ) .
d. l. cox and m. jarell , j. phys . : condens .
matter * 8 * , 9825 ( 1996 ) . c. m. varma , z. nussinovb , and w. van saarloos , phys .
* 361 * , 267 ( 2002 ) .
s. cornaglia , g. usaj , and c. a. balseiro , phys .
lett . * 102 * , 046801 ( 2009 ) .
zhu , k .- h .
ding and j. berakdar , unpublished ( 2009 ) , arxiv:0912.0182v1 .
l. dellanna , j. stat .
p01007 ( 2010 ) .
t. o. wehling , h. .p .
dahal , a. .i .
lichtenstein , m. i. katsnelson , h. c. manoharan , and a. v. balatsky , phys .
b. * 81 * , 085413 ( 2010 ) .
t. o. wehling , a. v. balatsky , m. i. katsnelson , a. i. lichtenstein , and a. rosch , phys .
b * 81 * , 115427 ( 2010 ) .
m. vojta , l. fritz and r. bulla , unpublished ( 2010 ) , arxiv:1002.2215 .
v. madhavan , w. chen , t. jamneala , m. f. crommie , n. s. wingreen , science * 280 * , 567 ( 1998 ) .
n. nel , j. krger , l. limot , k. palotas , w. a. hofer , and r. berndt , phys .
lett . * 98 * , 016801 ( 2007 ) .
l. vitali , r. ohmann , s. stepanow , p. gambardella , k. tao , r. huang , v. s. stepanyuk , p. bruno , and k. kern , phys .
* 101 * , 216802 ( 2008 ) n. nel , j. krger , r. berndt , t. o. wehling , a. i. lichtenstein , and m. i. katsnelson , phys .
* 101 * , 266803 ( 2008 ) .
d. jacob , k. haule , and g. kotliar , phys .
* 103 * , 016803 ( 2009 ) .
r. dovesi _ et al .
_ , crystal06 , release 1.0.2 , theoretical chemistry group - universita di torino - torino ( italy ) .
g. kotliar , s. y. savrasov , k. haule , v. s. oudovenko , o. parcollet , and c. a. marianetti , rev .
78 * , 865 ( 2006 ) .
k. haule , c .- h .
yee , and k. kim , unpublished ( 2009 ) , arxiv:0907.0195 .
k. haule , s. kirchner , j. kroha , and p. wlfle , phys .
b * 64 * , 155111 ( 2001 ) .
a. grechnev , i. di marco , m. i. katsnelson , a. i. lichtenstein , j. wills , and o. eriksson , phys . rev .
b 76 , 035107 ( 2007 ) . a. g. petukhov , i. i. mazin , l. chioncel and a. i. lichtenstein , phys .
b * 67 * 153106 ( 2003 ) .
v. w. brar , r. decker , h .-
solowan , y. wang , l. maserati , k. t. chan , h. lee , c. o. girit , a. zettl , s. g. louie , m. l. cohen , and m. f. crommie , unpublished ( 2010 ) , arxiv:1006.1014v1 .
y. mao , j. yuan , and j. zhong , j. phys . : condens .
matter * 20 * 115209 ( 2008 ) .
m. j. frisch _
gaussian 03 , revision b.01 , gaussian , inc . , pittsburgh pa , 2003 . a. d. becke , j. chem .
phys . * 98 * , 5648 ( 1993 ) .
unfortunately , we were not able to converge hybrid functional calculations with crystal06 for co atoms on graphene sheets . | we have studied the effect of dynamical correlations on the electronic structure of single co adatoms on graphene monolayers with a recently developed novel method for nanoscopic materials that combines density functional calculations with a fully dynamical treatment of the strongly interacting @xmath0-electrons .
the coupling of the co @xmath0-shell to the graphene substrate and hence the dynamic correlations are strongly dependent on the orbital symmetry and the system parameters ( temperature , distance of the adatom from the graphene sheet , gate voltage ) . when the kondo effect takes place , we find that the dynamical correlations give rise to strongly temperature - dependent peaks in the co @xmath0-spectra near the fermi level . moreover , we find that the kondo effect can be tuned by the application of a gate voltage .
it turns out that the position of the kondo peaks is pinned to the dirac points of graphene rather than to the chemical potential . |
the creation of directional states of molecules represents an important tool to control and tailor the rotational degree of freedom .
when a molecule is oriented the molecular fixed axes are confined along laboratory fixed axes and its dipole moment is pointing in a particular direction . experimentally , the availability of oriented molecules provides a wealth of interesting applications in a variety of molecular sciences , such as in chemical reaction dynamics @xcite , photoelectron angular distributions @xcite , or high - order harmonic generation @xcite . due to this broad interest
, special efforts have been undertaken to create samples of oriented molecules and techniques based in the application of inhomogeneous @xcite , and homogeneous @xcite electric fields as well as homogeneous magnetic fields @xcite have been used .
a major breakthrough came with the proposal by friedrich and herschbach @xcite of enhancing the orientation of polar molecules by exposing them to combined weak electrostatic and strong non - resonant radiative fields .
this theoretical prediction was done within an adiabatic picture assuming that the switching on time of the laser pulse is larger than the molecular rotational period . for linear molecules ,
a linearly polarized laser field produces a double - well potential along the polarization direction . in the pendular limit , this double - well potential contains nearly degenerate pairs of states with opposite parity forming tunneling doublets . if the molecules possess a permanent electric dipole moment , a strong pseudo - first - order stark effect is induced by coupling the tunneling doublets with an additional electrostatic field . due to this coupling , the two levels in a pendular doublet are efficiently oriented but with their effective electric dipole moments pointing in opposite directions .
as a consequence of this oriented and antioriented states pairing , the orientation is small in a molecular ensemble with the population thermally distributed .
therefore , the first experimental measures of the orientation of a molecular beam were indeed reduced to small values @xcite .
a significant improvement was gained by using a quantum - state selected molecular beam , which allowed the creation of unprecedented degree of orientation for complex asymmetric tops @xcite .
a first theoretical study of the mixed - field orientation experiment of asymmetric top molecules , pointed out that a fully adiabatic description of the process does not reproduce the experimental observations @xcite .
recently , we have experimentally and theoretically investigated the mixed - field orientation of the carbonyl sulfide molecule @xcite .
our analysis has proven that a time - dependent description of the mixed - field orientation process is required to explain the experimental results .
we have shown how the non - adiabatic coupling of the levels forming the quasi - degenerate doublets as the laser intensity is increased , gives rise to the reduction of the orientation and , therefore , to the disagreement with the predictions of the adiabatic theory @xcite .
herein , we provide a detailed theoretical analysis on the dynamics of a linear molecule exposed to an electrostatic field combined with a non - resonant laser pulse . in the framework of the rigid rotor approximation ,
we solve the time - dependent schrdinger equation using experimental field configurations , i.e. , a gaussian laser pulse and a weak electrostatic field that is turned on at constant speed .
as prototype example , we consider the carbonyl sulfide molecule ( ocs ) . for several rotational states ,
we investigate the mixed - field orientation dynamics under different field - configurations by varying either the laser peak intensity , the duration of the gaussian pulse , the dc field strength or the angle between both fields .
hence , we demonstrate that for some field configurations , the field - dressed dynamics is non - adiabatic and provide a detailed account of the sources of non - adiabaticity and the field regimes at which they appear . for parallel fields ,
the dynamics is characterized by the population transfer between adiabatic states when the pendular doublets are formed .
whereas for non - parallel fields , we encounter additional non - adiabatic effects when the states from the same @xmath0-manifold , having now the same symmetry , are driven apart as the laser intensity is increased on the weak field regime . for different field configurations , we identify and discuss the experimental conditions needed to achieve an adiabatic molecular dynamics .
the paper is organized as follows : in we describe the hamiltonian of the system and its symmetries for various field configurations .
the results for the energy , alignment and orientation predicted by the adiabatic theory are analyzed in . in , we focus on the molecular dynamics when the fields are parallel .
in particular , we explore how the time - dependent orientation varies as the field parameters are modified , and indicate the experimental conditions under which an adiabatic orientation would be achieved .
a similar study is performed for tilted fields in , where we show that the conditions for an adiabatic mixed - field orientation are more difficult to fulfill . in , we assume that once the pulse is turned on its peak intensity is kept constant , and investigate the dynamics in this regime . in , first the laser pulse
is switched on and then an electric field is applied . in this field configuration
, we analyze the orientation of the ground state and provide the field parameters for an adiabatic orientation .
the conclusions are given in .
we consider a polar linear molecule exposed to an homogeneous static electric field and a non - resonant linearly polarized laser pulse .
the field configuration is illustrated in : the polarization of the laser field lies along the @xmath1-axis of the laboratory fixed frame ( lff ) @xmath2 , and the dc field is contained in the @xmath3-plane forming an angle @xmath4 with the @xmath1-axis .
the @xmath5-axis of the molecule fixed frame @xmath6 is defined by the permanent dipole moment @xmath7 of the molecule .
these two frames are related by the euler angles @xmath8 , cf .
the description of this system is done within the rigid rotor approximation , assuming that the vibrational and electronic dynamics are not affected by the fields .
thus , the rigid rotor hamiltonian reads @xmath9 where @xmath10 is the field - free hamiltonian @xmath11 with @xmath12 being the total angular momentum operator and @xmath13 the rotational constant .
the terms @xmath14 and @xmath15 stands for the interactions with the static and laser fields , respectively .
the dipole coupling with the static field reads @xmath16 with @xmath17 , and @xmath18 being the electrostatic field strength .
the angle between the dipole moment @xmath7 and this field is @xmath19 , cf , and @xmath20 .
we consider a non - resonant laser field linearly polarized along the @xmath1-axis of the lff , @xmath21 , with @xmath22 being its frequency , @xmath23 the peak field strength , and @xmath24 the pulse envelope . assuming that @xmath25 is much shorter than the pulse duration and the rotational period
, we average over the rapid oscillations of the non - resonant field .
this causes the coupling of this field with the permanent dipole moment to vanish @xcite .
thus , the non - resonant laser field molecule interaction can be written as @xmath26 where @xmath27 is the polarizability anisotropy , @xmath28 is the intensity of the laser , @xmath29 is the speed of light and @xmath30 is the dielectric constant .
note that we have neglected the term @xmath31 , which represents only a shift in the energy . in this work ,
the field configurations are chosen based on the mixed - field orientation experiments @xcite .
initially , the molecule is in field - free space , then the electrostatic field is switched on , and its strength is increased linearly with time . at @xmath32
the maximum strength is achieved and kept constant afterwards .
this time @xmath33 is chosen long enough to ensure the adiabaticity of this turning - on process .
for the laser pulse , we use a linearly polarized gaussian pulse with a full width half maximum ( fwhm ) @xmath34 on the nanosecond range .
the intensity is given by @xmath35 , with @xmath36 being the peak intensity , and @xmath37 is related with the fwhm @xmath38 .
numerically , the non - resonant laser field is turned on in such a way that the interaction due to this field is much weaker than coupling with the dc field .
.action of the symmetry operations on the euler angles . [ cols="<,<,<",options="header " , ] the time - dependent schrdinger equation associated to the hamiltonian is solved by means of a second - order split - operator technique @xcite , combined with the discrete - variable and finite - basis representation methods for the angular coordinates @xcite . for reasons of addressability
, we will label the time - dependent states as @xmath39 and @xmath40 for @xmath41 and @xmath42 , respectively , with @xmath43 and @xmath44 indicating even or odd parity with respect to the @xmath3-plane .
the labels @xmath0 and @xmath45 refer to field - free quantum number to which they are adiabatically connected .
note that the labeling of the states depends on the way the fields are turned on @xcite .
the time - dependent wave function depends on the time @xmath46 , the peak intensity @xmath36 , the fwhm @xmath34 , the electrostatic field strength @xmath47 , and the angle @xmath4 . for the sake of simplicity
, we have not made explicit these dependences , but the field configuration is clearly indicated through the text . to get a better physical insight on the field - dressed dynamics , the time - dependent results will be compared to those from the adiabatic theory . for this system , we take the adiabatic limit by using a constant electrostatic field @xmath47 and constant laser intensity @xmath48 in the hamiltonian . the time - independent schrdinger equation associated to this hamiltonian is solved by expanding the wave function in a basis that respects the symmetries .
the adiabatic states are labeled as @xmath49 and @xmath50 for @xmath41 and @xmath42 , respectively , and we have not made explicit their dependence on the field parameters .
the field - dressed eigenfunctions of this time - independent hamiltonian form a basis , which is used to analyze the time - dependent wave function @xmath51 by means of the following expansion @xmath52 with @xmath53 , @xmath54 and @xmath55 including all the labels identifying these levels . for computational reasons , we have only considered the lowest - lying @xmath56 adiabatic levels , and always ensured that the contributions of highly excited states are negligible .
let us remark that for each time @xmath46 , the expansion of the wave function is performed in a different adiabatic basis obtained by solving the time - independent schrdinger equation using the static field strength and laser intensity at time @xmath46 , i.e. , @xmath18 and @xmath28 .
in this work , we use the ocs molecule ( see fig . [
fig : fig_1 ] ) as benchmark to illustrate our results .
the rotational constant of ocs is @xmath57 @xmath58 , the permanent dipole moment @xmath59 d and the polarizability anisotropy @xmath60 @xmath61 . and
( c ) @xmath62 as a function @xmath48 of the adiabatic states @xmath63 ( red thick solid ) , @xmath64 ( gold thin solid ) , @xmath65 ( orange thick short - dashed ) , @xmath66 ( dark blue long - dashed ) , @xmath67 ( blue dot - short - dashed ) , @xmath68 ( purple dotted ) , @xmath69 ( green thin short - dashed ) , and @xmath70 ( pink dot - long - dashed ) .
the insets show the relevant energy and intensity ranges where the formation of the near - degenerate doublets occurs .
( d ) polar plots of the square of their wave functions at @xmath71 . @xmath72 and @xmath41 for all data . ]
we start by analyzing the adiabatic limit .
we restrict this study to the following eight states : @xmath63 , @xmath64 , @xmath65 , @xmath66 , @xmath67 , @xmath68 , @xmath69 , and @xmath70 . for @xmath41 , they adiabatically correspond to the states forming the four first doublets .
note that they well represent the main physical features observed in the overall molecular dynamics , and similar behavior and properties are , therefore , obtained for states in other irreducible representations . for @xmath72 and @xmath41 ,
the energies and the expectation values @xmath73 and @xmath62 of these levels are plotted versus the laser intensity in ( a ) , ( b ) and ( c ) , respectively .
the weak static field breaks the field - free degeneracy in the magnetic quantum number and , as the laser intensity is increased , these states become high field seekers . in the strong laser field regime ,
once the pendular regime is reached , pairs of quasi - degenerate states with the same symmetry are formed .
the insets in this panel show how these doublets appear .
the gap in energy in a doublet goes as @xmath74 , where @xmath75 is the effective dipole moment of the state @xmath76 in the doublet , which is of opposite sign for @xmath77 . within a doublet ,
the two levels are characterized by the same hybridization of the angular motion @xmath78 and alignment @xmath73 , see ( b ) .
in contrast , they possess opposite orientation @xmath62 , one being oriented and the other antioriented , cf .
( c ) . this opposite orientation is also illustrated in ( d ) by the polar plots of the square of their wavefunctions for @xmath71 .
the larger is the field - free rotational quantum number of the levels , i.e. , their field - free energy , the stronger is the laser intensity needed to achieve a significant orientation .
indeed , the states in the third and fourth doublets are not aligned for @xmath79 and , therefore , not oriented . once the pendular regime is achieved , the orientation of these states @xmath80 approaches to @xmath81 as @xmath48 is enhanced .
if the laser field is sufficiently strong , this adiabatic orientation is independent of the dc field strength , and of the angle between both fields .
in this section , we investigate the rotational dynamics in a parallel configuration : a dc - field of @xmath82 and a gaussian pulse with fwhm @xmath83 ns and several peak intensities . for the ground state @xmath84 , the expectation value @xmath62 is presented in ( a ) as a function of @xmath28 up until the peak intensity @xmath36 is reached . for comparison ,
the adiabatic results are also shown .
since the fwhm is 125.31 times larger than the rotational period , one would expect that the rotational dynamics follows the adiabatic limit .
however this is not the case , and there are significant discrepancies between the time - dependent and adiabatic results .
in contrast to what is predicted by the adiabatic theory , the final orientation decreases as the peak intensity of the laser pulse is increased . for @xmath85 , @xmath62
initially resembles the adiabatic behavior , but it achieves a maximum value @xmath86 for @xmath87 . for @xmath88 , @xmath89 , and @xmath90 , the orientation shows a qualitatively similar but quantitatively different behavior : in the weak laser field regime , @xmath62 monotonically increases following the adiabatic limit , but for @xmath91 it reaches a plateau behavior being the orientation smaller than the adiabatic value .
for instance , @xmath92 for @xmath93 whereas the adiabatic value is @xmath94 . as a function of @xmath28 for gaussian pulses of @xmath83 ns and peak intensities @xmath85 ( red thick solid ) , @xmath95 (
orange dashed ) , @xmath89 ( gold dotted ) and @xmath90 ( green dot - dashed ) .
the adiabatic results for @xmath62 ( thin solid ) are also included .
( b ) the squares of the projections of the time - dependent wave functions onto the adiabatic pendular state @xmath63 .
( c ) adiabatic criteria @xmath96 as a function of @xmath28 .
the field configuration is @xmath72 and @xmath41 . ]
a first physical insight into the non - adiabatic dynamics could be gained by analyzing the characteristic times of the molecule .
when the states in a pendular doublet are quasi - degenerated , the energy gap between them , @xmath97 , defines a time scale of this system @xcite .
note that we have assumed @xmath98 , which holds in the strong laser field regime . for @xmath72 and @xmath41 ,
the energy separation within the first doublet formed by @xmath63 and @xmath64 is @xmath99 @xmath58 giving a time scale of @xmath100 ps , which is larger than the rotational period @xmath101 ps .
thus , only long enough pulses compared to this pendular time would ensure an adiabatic orientation of the molecule .
since the static field strength is so weak , its impact on the rotational dynamics is very small , and before the pulse the levels could be considered as field - free rotor states . as the laser intensity increases
, the states are hybridized by the combined action of the both fields , and the doublets of nearly - degenerate states are formed in the strong laser field regime , as it is shown in ( a ) .
when the energy splitting of this pendular doublet approaches the coupling of the two sublevels due to the pseudo - first - order stark interaction , these states can mix because they have the same symmetry for @xmath102 . as a consequence ,
there is a population transfer between the oriented and anti - oriented states , which results in a decrease of the final orientation compared to the adiabatic limit . for this field configuration ,
the dynamics can be analyzed by means of the adiabatic states forming this pendular doublet , because their couplings to states in neighboring doublets is much smaller than the energy difference between them .
note that these adiabatic states are the eigenstates of the hamiltonian at fixed time @xmath46 . under a time - dependent interaction , i.e. , in our case the interaction with the laser field @xmath103 , the dynamics could be considered as adiabatic if and only if the following condition @xcite @xmath104{\left\langle i \left|\cfrac{\partial h_\textup{l}(t)}{\partial t}\right| j\right\rangle}{_{\textup{p } } } \right|}{\left|e_i - e_j\right|^2}\ll 1\ ] ] is fulfilled , with @xmath105 and @xmath106 being the eigenstates of the hamiltonian in the adiabatic limit , and @xmath107 and @xmath108 their energies .
according to this criterion , the probability for mixing , corresponding to a transfer of population from one state of the doublet to the other , is determined by the rate of change of the laser field interaction and the energy separation between the states .
thus , as the laser intensity is increased the population transfer between the two states in a doublet takes place because the criterion is not satisfied . to illustrate this phenomenon
, we show the contribution of the adiabatic ground state @xmath109 to the time - dependent wave function of @xmath84 , ( b ) , and the adiabatic parameter @xmath96 when @xmath110 , ( c ) . note that @xmath111 . in these four cases ,
the dynamics is initially adiabatic , i.e. , @xmath109 remains equal to @xmath81 and @xmath112 .
as @xmath28 is increased , the energy splitting of the doublet decreases and , moreover , it becomes comparable or even larger than the rate of turning - on the pulse ; thus , @xmath96 is close to @xmath81 , and the population transfer takes place .
this region where @xmath96 is not negligible corresponds to the formation of the quasi - degenerate doublet .
once the doublet is formed , @xmath113 reaches a small value and slowly decreases as @xmath28 is enhanced ; but the two states are oriented in opposite directions and their wave functions do not overlap .
therefore , the coupling due to the alignment laser is much smaller than @xmath114 , @xmath115 and the population transfer does not take place any longer , i.e. , @xmath109 remains constant as @xmath28 is enhanced .
the larger is this population transfer , the smaller is the orientation compared to the adiabatic prediction . since these adiabatic states contributing to the dynamics are quasi - degenerated and have very close values of the alignment and hybridization of the angular motion , the lack of adiabaticity is not reflected on the time evolution of the energy , @xmath73 or @xmath116 .
for this field configuration , the molecular dynamics of excited states present analogous features for @xmath62 , @xmath73 and @xmath116 as those encountered here for the ground state .
the adiabaticity of the field - dressed dynamics is determined by the rate of change of the laser field interaction compared to the largest time scale of the system . in the pendular regime ,
the energy splitting in a doublet goes as @xmath117 ; then , the population transfer decreases if @xmath47 is increased . on the other hand , by increasing the fwhm augments the time scale on which the pendular doublets are formed , and facilitates the adaptation of the molecule to this field .
that is , the mixed - field orientation will be more adiabatic when either longer pulses or stronger static electric fields are used .
let us remark that with the expression _ being the dynamics more adiabatic _ we mean that for a certain state , the weight of its corresponding adiabatic state in the time - dependent wave function is closer to one during the dynamics . here
, we investigate the orientation at the maximum of the laser pulse , as it is done in most of the experiments @xcite .
the rate of change of the laser field and the adiabatic parameter depend linearly on @xmath36 .
then , for a gaussian pulse with fixed fwhm , the dynamics will be more diabatic if @xmath36 is increased .
in this section , we consider the states @xmath84 , @xmath118 , @xmath119 , @xmath120 , @xmath121 , @xmath122 , @xmath123 , and @xmath124 . their orientation at @xmath125 , i.e. , @xmath62 for @xmath126 , is plotted as a function of @xmath36 in ( a ) and ( b ) for @xmath72 and @xmath127 , respectively .
the fields are parallel , and the fwhm of these pulses is @xmath128 ns . at @xmath125 as a function of the peak intensity @xmath36 for the states @xmath84 ( red thick solid ) , @xmath118 ( gold thin solid ) @xmath119 ( orange thick short - dashed ) , @xmath120 ( dark blue long - dashed ) @xmath121 ( blue dot - short - dashed ) , @xmath122 ( purple dotted ) , @xmath123 ( green thin short - dashed ) and @xmath124 ( pink dot - long - dashed ) , for @xmath41 and ( a ) @xmath72 and ( b ) @xmath127 .
the fwhm of the laser pulses is @xmath128 ns . ] for @xmath72 , the orientation of the low - lying level in a doublet increases as @xmath36 is enhanced , reaching a maximum and smoothly decreasing thereafter .
this is counterintuitive to what is expected in the adiabatic limit ; namely , a larger orientation when the laser intensity is increased .
the maximum in the orientation is achieved with an alignment pulse that already gives rise to a non - adiabatic dynamics .
however , due to the coupling between the populated adiabatic states in the pendular pair @xmath129 , the orientation is enhanced compared to what happens at the adiabatic limit . by further increasing @xmath36 ,
the population transferred between the two states is enhanced , but now the coupling between them is very small or even zero due to their opposite orientation . as a consequence , the final orientation decreases as @xmath36 is increased . for a certain pendular doublet
, the upper state is antioriented , and @xmath62 shows the opposite behavior as a function of @xmath36 .
regarding the third and fourth doublets , the states are not oriented nor aligned for ac pulses with @xmath130 .
compared to low - lying states , their orientation is smaller and the maximum of @xmath62 appears at larger peak intensities . by increasing the static field strength to @xmath127 the energy gap of the pendular pair is also increased , being the characteristic time scale of the system reduced .
thus , for the same laser pulse , the dynamics is more adiabatic , i.e. , less population is transferred , and the final orientation is increased , see ( b ) .
the orientation ( antiorientation ) of the pendular states also achieves a maximum ( minimum ) , but it is so shallow that it is hardly appreciated on the scale of this figure . for the states @xmath84 ( red thick solid ) , @xmath119 ( orange thick
short - dashed ) , @xmath121 ( blue dot - short - dashed ) and @xmath123 ( green thin short - dashed ) , for ( a ) @xmath72 and ( b ) @xmath127 .
we use @xmath128 ns laser pulses and @xmath41 . ] to illustrate the field - dressed dynamics , we plot in the weights of the adiabatic states associated to the oriented levels in these pendular doublets . for the corresponding antioriented levels , the contributions of its associated adiabatic state are identical to the one presented here , e.g. , for the ground state , we present the contribution of the adiabatic ground state @xmath131 , which is identical to the weight @xmath132 for @xmath133 . in an adiabatic molecular dynamics , these coefficients are equal to one .
note that in the considered regime only the two adiabatic states of the pendular doublet contribute to the dynamics .
for all these levels , @xmath134 decreases , i.e. , the dynamics is less adiabatic , as @xmath36 is enhanced . by increasing @xmath47 ,
@xmath114 is increased ; thus , under the same gaussian pulse the population transfer is reduced ,
i.e. , @xmath134 is closer to one , and the range of peak intensities that could be considered as adiabatic is increased . the duration of the gaussian pulse plays an important role in the molecular dynamics .
it has been shown that an alignment pulse of @xmath128 ns is not enough to achieve an adiabatic mixed - field orientation for molecules such as ocs , benzonitrile and iodobenzene @xcite .
it has been pointed out the need of increasing the rising time of the laser pulses to achieve the highest possible orientation @xcite . by increasing the fwhm ,
the time needed to form the pendular doublets is also increased . for a given field configuration , at the point where the pulse reaches a certain intensity
the adiabatic parameter is reduced if @xmath34 is increased .
hence , the molecular dynamics becomes more adiabatic and , therefore , the population transfer to other pendular states is reduced .
here , we investigate how the directional properties of ocs depends on the laser - pulse fwhm .
for the same set of states as in the previous section , shows the orientation at @xmath125 as a function of @xmath34 .
the fields are parallel , and we consider the peak intensities @xmath85 and @xmath89 , and a dc field of @xmath72 . at @xmath125 as a function of @xmath34 for ( a ) @xmath85 and ( b ) @xmath135 . the fields are parallel and @xmath72 .
the states and their labels are the same as in . ]
the degree of orientation of the two states in a given pendular pair shows the same behavior as a function of @xmath34 , but with their dipole moment pointing in opposite directions . in the first two doublets and @xmath85
, @xmath80 increases with @xmath34 till it reaches a plateau - like behavior . the second pair satisfies that @xmath136 for @xmath137 ns .
since the states in the third and fourth doublets have not achieved the pendular regime for @xmath85 , increasing the pulse duration does not have any impact on their orientation , and @xmath62 keeps a constant value close to zero as @xmath34 is increased . for @xmath135 , the degree of orientation of all the states increases and
approaches the adiabatic limit as @xmath34 is enhanced .
for the ground state and @xmath138 ns , we obtain @xmath139 , which is very close to the adiabatic value @xmath140 .
these results show that for parallel fields , the mixed - field orientation dynamics of any state could be adiabatic if a sufficiently long pulse and sufficiently strong fields are used . for a @xmath141 ns gaussian pulse with @xmath135 , and @xmath72 ,
the dynamics can be considered as adiabatic for all these states , with @xmath142 .
since the energy splitting in a pendular doublet is proportional to the static field strength , the degree of adiabaticity in the molecular orientation should increase if @xmath47 is enhanced , i.e. , the characteristic time scale of the system is reduced . in
we present the final orientation at @xmath125 of these eight states versus @xmath47 .
we have considered two laser pulses of @xmath83 ns with peak intensities @xmath85 and @xmath89 , and @xmath41 . at @xmath125 as a function of @xmath47 for ( a ) @xmath85 and ( b ) @xmath135 .
we use a @xmath128 ns laser pulses and @xmath41 .
the states and their labels are the same as in . ] for the lowest laser intensity , the orientation of the @xmath84 and @xmath118 states is constant and independent of the static field for @xmath143 with @xmath144 . the orientation of the levels @xmath119 and @xmath120 monotonically increases as @xmath47 is enhanced , and we obtain @xmath145 for @xmath146 .
this peak intensity is not large enough for the states in the third and fourth doublets to be in the pendular regime .
thus , these pairs are weakly oriented even if a strong dc field is used , e.g. , for @xmath146 , @xmath147 and @xmath148 for the third and fourth doublets , respectively . for the strong peak intensity ,
all the states are in the pendular regime , and their @xmath80 increases as @xmath47 is increased reaching a constant value for sufficiently strong static fields .
their orientation approaches the adiabatic limit and for @xmath149 , @xmath150 for the states of the first doublet , and @xmath151 of their population is on the corresponding adiabatic pendular state . for @xmath146 , the states in the fourth doublet satisfy @xmath152 and @xmath153 of their population
is on the corresponding adiabatic level . in conclusion , by combining sufficiently strong electrostatic fields with standard gaussian pulses , i.e. , with experimentally accessible peak intensities of @xmath89 and @xmath128 ns fwhm , a significant orientation is obtained even for excited rotational levels .
it is worth remarking that the fields have to be parallel ; then , techniques such us the ion imaging method @xcite could not be used to measure the degree of orientation ; whereas techniques , such as time of flight @xcite are feasible .
in this section we investigate the rotational dynamics when the electrostatic field forms an angle @xmath42 with the linearly polarized laser pulse
. the azimuthal symmetry is lost , and the number of irreducible representations is reduced to two , see .
thus , states with different field - free magnetic quantum numbers are now coupled by the electrostatic field .
the field - free wave function of the initial state is constructed as an eigenstate of the operators @xmath154 and @xmath155 , see ; i.e. , @xmath156=@xmath157@xmath156@xmath158 , where @xmath157 is the rotation operator of an angle @xmath4 around the lff
@xmath159-axis @xcite .
this ensures that this wave function has the correct symmetries , and that its time evolution corresponds , in the adiabatic limit , to an eigenstate of the field - dressed hamiltonian at any time .
before the pulse is turned on , an important feature of the ground state is that its energy gap to the next state with the same symmetry is proportional to the rotational constant @xmath13 , which is much larger than the coupling in the weak laser field regime . in this regime
, it evolves as an isolated state , and its interaction to the neighboring levels could be considered negligible .
hence , analogously to the parallel field configuration , the formation of the doublets in the pendular regime is the only source of non - adiabatic effects in its field - dressed dynamics .
note that the lowest lying level of the odd irreducible representation will show the same behavior . and ( b ) @xmath160 as a function of @xmath28 of a @xmath128 ns gaussian pulse .
the field configurations are @xmath161 ( thick ) and @xmath162 ( thin ) with peak intensities
@xmath85 ( dashed ) , @xmath95 ( dot - dashed ) , and @xmath89 ( dotted ) . the static electric field is fixed to @xmath72 .
the adiabatic results ( solid ) are also included . ] for the ground state , the time evolution of the expectation values @xmath62 and @xmath160 are presented as a function of @xmath28 till the peak intensity is reached in ( a ) and ( b ) .
the gaussian pulse has @xmath128 ns fwhm and peak intensities @xmath85 , @xmath95 , and @xmath89 .
we consider the inclination angles @xmath161 and @xmath162 , and a dc field of @xmath72 . for comparison ,
the adiabatic results are also included , with @xmath62 being independent of @xmath4 . for a certain laser pulse , increasing the inclination angle towards @xmath163 implies a decrease of the energy splitting in the first doublet , @xmath164 , and , therefore , an increase of the adiabatic parameter .
note that in the energy splitting we have not considered the component of the dc field along the lff @xmath165-axis .
compared to the @xmath41 configuration , the dynamics could be considered as less adiabatic being characterized by a larger population transfer to the other adiabatic state in this pendular pair .
the final orientation is significantly decreased as @xmath4 is increased , e.g. , for @xmath135 , @xmath166 and @xmath167 , and the contribution of the adiabatic ground state is @xmath168 and @xmath169 for @xmath161 and @xmath162 , respectively . for a certain angle @xmath4 , the orientation achieved at @xmath125 decreases as @xmath36
is increased , cf . .
since the molecular dynamics of the ground state is restricted to the two lower pendular adiabatic states , its time dependent results for its energy and expectation values @xmath73 and @xmath116 resemble the adiabatic ones . to illustrate the rotational dynamics of excited states , we show in ( a ) the orientation cosine @xmath62 as a function of @xmath28 for @xmath170 .
the field configurations are the same as in .
the adiabatic model predicts a sharp wrong - way orientation .
in contrast , this state presents a weak or even zero orientation , and in addition , a larger peak intensity does not imply a larger orientation .
when the peak intensity is reached , this level shows a weak right - way orientation for @xmath161 : @xmath171 and @xmath172 , for @xmath88 and @xmath89 , respectively . for @xmath173 and the peak intensities @xmath88 and @xmath89 , due to the non - adiabatic dynamics @xmath170 is not oriented
let us analyze in detail these results .
for highly excited states , the dynamics is more complicated .
apart from the doublet formation , there is another physical phenomenon at weak laser intensities which causes loss of adiabaticity . in the presence of only a weak static field
, the @xmath45-degeneracy of the states with the same field - free @xmath0 is broken due to the quadratic stark effect , i.e. , the splitting goes as @xmath174 .
as the pulse is switched on , the energy gap between two states of this @xmath0-manifold is much smaller than the rate of their coupling due to the laser field , i.e. , @xmath96 is larger than one . for @xmath161 ,
the adiabatic parameter @xmath96 between the states @xmath175 and @xmath176 , both contributing to the dynamics of @xmath170 , is presented in ( b ) , and it achieves large values for @xmath177 . as the states in this @xmath0-manifold are driven apart by the laser field , the process is non - adiabatic and there is a population transfer between them .
the projections of the time - dependent wave function in terms of the adiabatic states @xmath178 , @xmath175 , @xmath176 , and @xmath179 is presented in ( d ) for @xmath161 and @xmath135 . under these diabatic conditions , @xmath180 decreases as @xmath28 is increased , whereas @xmath181 increases . by further increasing @xmath28 , the coupling between these states becomes very small or even zero and their energy separation increases , so that @xmath96 decreases and the population transfer is stopped .
this process is so diabatic that the wave function does not change , but its projections on the adiabatic basis are modified because the basis varies with time .
for instance , the field - free state is @xmath170@xmath182@xmath65@xmath183@xmath64 , which belongs to the proper irreducible representation . after swichting on the static field ,
its wave function could be approximated by the same expression because this field is very weak .
once the splitting of this @xmath0-manifold is finished , i.e. , for @xmath184 , the contributions of the states @xmath175 and @xmath176 are approximately @xmath185 and @xmath186 , respectively .
note that the states @xmath187 and @xmath188 are not related adiabatically .
state , expectation value @xmath62 as a function of @xmath28 of a @xmath128 ns gaussian pulse .
the field configurations are @xmath161 ( thick ) and @xmath162 ( thin ) , with @xmath85 ( dashed ) , @xmath95 ( dot - dashed ) and @xmath89 ( dotted ) . the adiabatic results ( solid line ) are also included . for @xmath135 and @xmath161 , ( b )
square of the projections of the time - dependent wave function on the adiabatic pendular states @xmath175 ( dot - dashed ) , @xmath176 ( dotted ) , @xmath178 ( long - dashed ) , and @xmath179 ( short - dashed ) , and ( c ) adiabatic parameter between the pendular states @xmath175 and @xmath176 ( dot - dashed ) , @xmath175 and @xmath178 ( dashed ) , and @xmath176 and @xmath179 ( dotted ) . the dc field is fixed to @xmath72 . ] in contrast to the ground state , the wave function of any excited level has contributions from adiabatic states which correspond to different pendular doublets .
as the laser intensity is increased , the molecular dynamics is affected by the formation of these pendular doublets .
thus , the final orientation could be significantly reduced compared to the parallel fields result .
for instance , the time - dependent @xmath170 state has contributions from the adiabatic levels @xmath175 and @xmath176 , which correspond to the first and second pendular doublets , respectively . in ( b )
we show how the adiabatic parameters @xmath96 between the pairs @xmath178-@xmath175 and @xmath176-@xmath179 , which form the first and second doublets , respectively , achieve values close to @xmath81 .
the final population of the state @xmath170 is @xmath189 , @xmath190 , @xmath191 , and @xmath192 , which gives rise to a small orientation as a consequence of this population redistribution to other pendular doublets , features of the system such as energy , alignment and hybridization of the angular motion do not resemble the adiabatic results .
in particular , since the levels on the second pendular doublet possess a smaller alignment , the adiabatic result is larger than the time - dependent one .
for instance , for @xmath161 and @xmath135 , once the time evolution is finished the alignment of this state @xmath170 is @xmath193 , compared to @xmath194 obtained for the adiabatic level @xmath175 . for @xmath173 , despite the fact that the @xmath170 level is significantly aligned , @xmath195
, it is not oriented with @xmath196 for @xmath135 .
this state does not gain any orientation if stronger peak intensities are used .
this is a consequence of the population redistribution explained above .
indeed , this level could be considered as a _ dark state _ for the mixed - field orientation dynamics .
this physical phenomenon is not restricted to this state and field configuration .
we show below that other levels also behave as _ dark states_. it is worth noting that if in a mixed - field orientation experiment these dark states form part of the molecular beam , the degree of orientation will be diminished .
the population redistribution to other pendular doublets significantly affects the expectation value @xmath160 . to @xmath160 contribute terms which mix up adiabatic states with different magnetic quantum numbers since their wave functions could spatially overlap ,
their coupling matrix elements do not vanish , and @xmath160 oscillates as @xmath46 is increased .
the phenomenon of population redistribution at weak laser intensities also occurs for highly excited rotational levels , and for them , more adiabatic states would be involved in it . before the gaussian pulse
is turned on , the stark separation of the states in a certain @xmath0-manifold is increased if the electrostatic field strength is enhanced .
then , the adiabatic parameter @xmath96 is reduced , and the process of splitting of this @xmath0-manifold becomes less diabatic .
indeed , for sufficiently strong dc - fields , the dynamics would be adiabatic without population transfer between the states with the same field - free @xmath0 .
for instance , the mixed - field dynamics of the @xmath170 level can be considered as adiabatic on the weak laser field regime for @xmath197 and @xmath161 .
let us remark that the excited states could also suffer avoided crossings with adjacent levels , having different field - free magnetic quantum number @xmath45 , as the pulse intensity is varied .
the rotational dynamics in most of these crossings will be non - adiabatic @xcite .
analogously to the parallel - field configuration , we investigate now the impact of the laser peak intensity on the orientation .
to do so , we restrict this study to the following eight states : @xmath198 , @xmath133 , @xmath170 , @xmath199 , @xmath200 , @xmath201 , @xmath202 , and @xmath203 . note that they are related to the ones analyzed in the parallel fields configuration , by a rotation of @xmath4 around the lff @xmath159 axis .
their orientation at @xmath125 , i.e. , @xmath62 for @xmath126 , is plotted as a function of @xmath36 in for @xmath161 , panels ( a)-(b ) , @xmath173 , panels ( c)-(d ) , and @xmath204 , panels ( e)-(f ) , and @xmath72 and @xmath127 , respectively .
the fwhm of these gaussian pulses is fixed to @xmath83 ns . at @xmath125 as a function of the peak intensity @xmath36 for @xmath198 ( red thick solid ) , @xmath133 ( gold thin solid ) @xmath170 ( orange thick short - dashed ) , @xmath200 ( dark blue long - dashed ) @xmath199 ( blue dot - short - dashed ) , @xmath202 ( purple dotted ) , @xmath201 ( green thin short - dashed ) and @xmath203 ( pink dot - long - dashed ) .
the field configurations are ( a)-(b ) @xmath161 , ( c)-(d ) @xmath173 and ( e)-(f ) @xmath204 , with @xmath72 and @xmath127 , respectively .
the fwhms of the gaussian pulses are fixed to @xmath128 ns . ]
let us start analyzing the results for the ground state .
for all field configurations , @xmath62 shows a qualitatively similar behavior as a function of the peak intensity : initially increases , reaches a maximum and decreases thereafter . at the peak intensity where the maximum of @xmath62 takes place ,
the dynamics of this state is non - adiabatic , but due to the coupling of both states the orientation increases with respect to the adiabatic result . for a fixed peak intensity and electric field strength , @xmath62 decreases as @xmath4 is increased towards @xmath163 . for @xmath204 and @xmath205 , the ground state achieves a moderate maximal orientation , @xmath206 and @xmath207 for @xmath72 and @xmath127 , respectively .
the population transfer taking place at weak and strong laser intensities leaves its finger - prints in the dynamics of the excited states .
compared to the parallel field results , cf , their orientation is reduced for any inclination angle @xmath4 and the pendular pairs are not any longer formed by a right- and wrong - way oriented states .
whereas for most of the field configurations , the ground state possesses the largest orientation , the levels @xmath133 or @xmath208 could achieve a similar or even larger orientation , e.g. , for @xmath173 and @xmath209 , @xmath210 and @xmath72 . for @xmath161 @xmath211 ,
the degree of orientation is moderate for most of the states .
several dark states are found for @xmath173 : @xmath170 , @xmath133 , @xmath202 , and @xmath203 , see ( c)-(d ) .
for instance , the levels @xmath170 and @xmath133 are strongly aligned with @xmath212 for @xmath213 and @xmath211 , whereas they are not orientated with @xmath214 and @xmath215 , respectively . for @xmath204 ,
when the peak intensity of the gaussian pulse is reached most of the states present a weak orientation , only the levels @xmath198 and @xmath170 have a large orientation for small values of @xmath36 .
these results indicate that with a @xmath128 ns alignment pulse , strong dc fields and small inclination angles are required to reach a moderate orientation for excited states . for the same set of states as in the previous section , we analyze here how their directional properties depend on the fwhm of the gaussian pulse . in ( a ) and ( b )
we show @xmath62 at @xmath125 as a function of @xmath34 for @xmath161 and @xmath162 , respectively . the static electric field is fixed to @xmath72 , and the peak intensity to @xmath135 . at @xmath125 as a function of @xmath34 .
the field configurations are ( a ) @xmath161 and ( b ) @xmath173 , with @xmath135 and @xmath72 . the states and their labels are the same as in .
] the orientation of the ground state increases approaching the adiabatic limit as @xmath34 is increased , and it reaches it with a @xmath141 ns pulse .
we encounter several excited states presenting a moderate or weak orientation , and their @xmath80 monotonically increases as @xmath34 is enhanced , e.g. , for @xmath161 the levels @xmath133 , @xmath170 , @xmath199 , @xmath200 , @xmath201 and @xmath202 and for @xmath173 @xmath199 , @xmath200 and @xmath201 .
for all of them , a @xmath216 ns pulse is not enough to achieve the adiabatic regime .
in contrast , other excited levels present a very small , almost zero , orientation independently of the pulse duration .
some of these levels behave as dark states being strongly aligned but not oriented independently of the pulse duration , e.g. , the @xmath203 state has @xmath217 and @xmath218 for @xmath219 and any value of @xmath34 .
an analogous behavior is found for the levels @xmath133 , @xmath170 , @xmath202 and @xmath203 at @xmath173 .
as described above , this phenomenon is due to the non - adiabatic dynamics at weak laser intensities when the levels of the @xmath0-manifold are driven apart , and it takes places even for @xmath141 ns pulses . in this section , we consider two inclination angles , and investigate the impact of the electrostatic field on the mixed - field orientation dynamics of the same states .
figures [ fig : fig_12 ] ( a ) and ( b ) illustrate the behavior of @xmath62 at @xmath125 as a function of @xmath47 for @xmath161 and @xmath162 , respectively .
the laser pulse has @xmath83 ns and @xmath135 . at @xmath125 as a function of @xmath47 for the field configurations @xmath83 ns , @xmath135 , and ( a ) @xmath161 and ( b ) @xmath173 . the labeling of the states is done as in . ] the ground state presents the largest orientation , which increases as @xmath47 is enhanced , being strongly oriented for sufficiently large fields , e.g. , @xmath220 for @xmath221 and @xmath161 . regarding the excited states , their orientation strongly depends on the inclination angle . for @xmath161 , @xmath80
monotonically increases till it reaches a plateau - like behavior , and they show a moderate orientation . indeed , for @xmath161 , @xmath135 and @xmath146
, we obtain at the maximum of the gaussian pulse @xmath222 and @xmath223 for the states @xmath133 and @xmath170 , respectively . for @xmath173 , the level @xmath200 presents a large orientation : @xmath224 for @xmath225 .
there are some darks states for @xmath173 , which are not oriented even when dc fields of @xmath226 are used , e.g. , @xmath202 and @xmath203 . for non - parallel fields , a strong dc field
does not ensure a large orientation for excited rotational states
. if the aim is a strongly oriented molecular ensemble , then this should be as pure as possible in the ground state . in the hamiltonian ,
the term @xmath227 is responsible for the mixing of states with different field - free magnetic quantum numbers . on the weak dc field regime ,
the mixing between these states is so small that @xmath45 could be considered as conserved , and this term could be neglected . by increasing @xmath47 ,
this coupling between levels with different field - free @xmath45 becomes important , and this should affect the molecular dynamics .
thus , the questions that remain open is how important is the @xmath165-component of the electrostatic field to the dynamics , and for which electric field regime , we could only consider its @xmath1-component @xmath228 .
as indicated above , even for tilted fields , the dynamics of the ground state can be described by a two state model .
its energy separation to the next state with @xmath229 is of the order of @xmath13 and larger than the dc field coupling to these levels .
thus , for @xmath230 , the dynamics considering the dc field is equal to the one obtained when only its @xmath1-component is included . for the excited states , the answer to these questions depends on how the initial wave function , before the fields are switched on , is constructed .
the first option is to proceed as indicated at the beginning of this section ; the field - free @xmath231 and @xmath41 wave functions are related by a rotation of @xmath4 around the @xmath159-axis @xmath156=@xmath157@xmath156@xmath158 . in this case , for the level @xmath170 , some differences in its orientation are observed for @xmath149 with @xmath80 being larger if the two components of @xmath232 are considered .
these differences are augmented as @xmath47 is increased , e.g. , for a @xmath128 ns laser pulse with @xmath233 , @xmath173 and @xmath234 , we obtain at @xmath125 @xmath235 compared to @xmath236 if only the @xmath1-component of @xmath232 is included . by increasing @xmath47
this state will achieve an adiabatic dynamics only if both components of the static field are present .
the second option is to construct the field - free @xmath231 wave function equal to the field - free @xmath41 one . in this case , the results resemble those of the parallel field configuration taking into account @xmath237 as scaling factor for the static field strength . the symmetries of the rotational hamiltonian , see , and , therefore , the rotational dynamics strongly depend on the angle between the fields . in this section ,
we investigate in detail the impact of the inclination angle in the mixed - field orientation dynamics . for the ground state @xmath198 , the orientation cosines @xmath62 and @xmath160 are plotted in , as a function of @xmath4 , together with the adiabatic results . for a weak dc field and strong laser field ,
the following relation @xmath238 is satisfied within the adiabatic limit . in @xmath160
the term @xmath239 has been neglected , which can be done as far as the mixing between states with different field - free @xmath45 is very small . by increasing the electrostatic field strength ,
a regime would be encountered where this approximation does not hold any longer .
an analogous relation is satisfied between the time - dependent orientation cosines of the ground state . for @xmath85
, its orientation @xmath62 shows a plateau - like behavior till @xmath240 , which is very close to the adiabatic limit . by further increasing @xmath4 , @xmath62 decreases and approaches to zero .
for @xmath241 , the states in a pendular doublet have different symmetry and are not coupled by the dc field , thus they might be strongly aligned but not oriented . for @xmath135 , @xmath62
monotonically decreases as @xmath4 is increased towards @xmath242 , and its value is always smaller than for @xmath85 . for both laser fields , @xmath243 decreases as @xmath4 is increased .
( thick solid line ) and @xmath160 ( thin solid line ) at @xmath125 as a function of @xmath4 for the ground state .
the peak intensities are ( a ) @xmath205 and ( b ) @xmath244 .
the adiabatic results for @xmath62 ( thick dashed line ) and @xmath160 ( thin dashed line ) are also presented .
the fwhm of the laser pulse is fixed to @xmath83 ns and the dc field to @xmath72 . ] in figs .
[ fig : fig_14 ] ( a ) , ( b ) , ( c ) and ( d ) , we present the orientation cosine of the pairs @xmath198-@xmath133 , @xmath170-@xmath200 , @xmath201-@xmath203 and @xmath199-@xmath202 , respectively , as a function @xmath4 .
the static field strength is @xmath72 and we consider two gaussian pulses of @xmath128 ns fwhm and peak intensities @xmath205 and @xmath89 . due to the complicated field - dressed dynamics of excited states for @xmath42 with contributions from several pendular pairs , in @xmath160
the term @xmath245 can not be neglected .
then , the simple relation @xmath246 does not hold for these levels . at @xmath125 as a function of @xmath4 for the states ( a ) @xmath198 ( black ) , and @xmath133 ( blue ) , ( b ) @xmath170 ( black ) , and @xmath200 ( blue ) , ( c ) @xmath199 ( black ) , and @xmath202 ( blue ) , ( d ) @xmath201 ( black ) , and @xmath203 ( blue ) .
the field configuration is @xmath72 and @xmath205 ( solid lines ) and @xmath89 ( dashed lines ) .
the fwhm of the laser pulse is fixed to @xmath83 ns . ] based on the adiabatic theory , the ground state and the level @xmath133 , should present the same orientation but with opposite directions .
however , this is only satisfied for @xmath41 . due to the non - adiabatic effects at weak laser intensities , its @xmath80 is smaller than the corresponding value of @xmath198 for @xmath247 . for the second doublet , cf .
( b ) , @xmath62 oscillates as @xmath4 is varied , and the orientation even changes its direction . both states could present a moderate orientation at a certain value of @xmath4 .
the pendular regime is not achieved by the third and fourth pairs with a @xmath128 ns laser pulse and @xmath85 , and their orientation is either zero or very small independently of @xmath4 . for @xmath135 and @xmath41 ,
these four states show a moderate orientation , which is reduced for any other angle , being small for @xmath248 . at the strong peak intensity @xmath135 ,
in all pendular doublets one of the two levels presents the dark behavior with respect to the mixed - field orientation dynamics at a certain angle @xmath4 .
these results show that if the molecular beam is rotationally cold , a small inclination angle will optimize the degree of orientation observed in the experiment .
let us investigate the dynamics for @xmath249 assuming that the laser peak intensity , reached at @xmath125 , and the dc field strength are kept constant for @xmath249 ; i.e. , @xmath250 and @xmath251 for @xmath249 . at @xmath125
, the time - dependent wave function can be expressed in terms of the corresponding adiabatic basis . since the hamiltonian is time - independent for @xmath249 , the contribution of each adiabatic state remains constant as @xmath46 is increased . for a certain state @xmath252
, the expectation value of an operator @xmath253 in this adiabatic basis reads as @xmath254{\langle \gamma_j|\hat{a}|\gamma_j\rangle}{_{\textup{p}}}\\ \label{eqn : oscillate_cosine } & + & 2\sum_{j
< k}|c_j(0)||c_k(0)|\tensor[_{\textup{p}}]{\langle\gamma_j|\hat{a}|\gamma_k\rangle}{_{\textup{p}}}\cos\left(\cfrac{\delta e_{jk}t}{\hbar}+\delta_{jk}\right ) \nonumber , \end{aligned}\ ] ] with @xmath255 being the weight at @xmath125 of the adiabatic state @xmath256 to the wave function of @xmath252 , @xmath257 the energy splitting between the adiabatic levels @xmath256 and @xmath258 , and @xmath259 the phase difference of @xmath255 and @xmath260 . based on the results presented above
, the time - dependent wave function could have contributions from : i ) only the adiabatic levels forming a pendular doublet , or ii ) several adiabatic levels from at least two pendular doublets .
all the states for @xmath41 , and the ground state for @xmath261 could belong to the first case .
whereas , the second one refers to all excited states when @xmath42 , unless the static field is very strong . .
for the state @xmath170 , ( a ) expectation value @xmath262 with @xmath85 and @xmath41 ( solid ) , @xmath263 ( dashed ) and @xmath162 ( dotted ) ; and ( b ) @xmath243 for @xmath161 and @xmath85 ( solid ) , @xmath95 ( dashed ) and @xmath89 ( dotted ) .
the dc field is @xmath264 . ]
let us first analyze the case when the dynamics takes place within a pendular doublet . if the adiabatic states are not fully oriented , the coupling term in is non - zero and this expectation value oscillates for @xmath249 with the frequency equal to the energy splitting of the corresponding pendular doublet . for the @xmath170 state ,
this behavior is shown for the time evolution of @xmath62 in ( a ) , with @xmath85 , @xmath265 and @xmath41 .
an analogous behavior is obtained for the ground state and @xmath42 . by further increasing the peak intensity , the orientation of the adiabatic states increases , the coupling terms are reduced approaching zero , and these regular oscillations will disappear . when two pendular doublets participate in the dynamics , this oscillatory behavior becomes irregular , because the frequencies associated with the energy separations within each pendular doublet and between two of them do not form a commensurable set .
as an example , we show in ( a ) these irregular oscillations of @xmath62 for @xmath170 with @xmath161 and @xmath162 , @xmath265 and @xmath85 . by increasing @xmath36 ,
the dynamics of this state still has contributions from different pendular doublets , but the two states in a pendular pair are not populated . as a consequence ,
the coupling terms are reduced and the oscillation decreases or even disappears . for @xmath42 ,
the time evolution of @xmath160 is dominated by the couplings of adiabatic levels from doublets with @xmath266 .
this is illustrated in ( b ) , for the state @xmath170 , @xmath85 , @xmath95 and @xmath89 and @xmath161 .
independently of @xmath36 , in this time scale @xmath160 oscillates with the largest frequency given by the energy gap between of the two pendular doublets involved , which is similar for the three peak intensities . on a larger time scale , the frequencies due to the states in a doublet will modulate the oscillations of @xmath243 in the weak field regime .
in previous sections , the field configuration was based on the mixed - field orientation experiments @xcite . here
, we investigate the molecular dynamics when the temporal order of the fields is inverted : the gaussian pulse is switched on first , its peak intensity is kept constant , and then the static electric field is turned on . while the laser field is switched on ,
the pendular doublets of quasi - degenerate states with opposite parity are formed .
this process is adiabatic @xcite , and for a sufficiently large peak intensity , these two levels are strongly aligned but not oriented . by turning on the static field , these states have the same symmetry and they should be oriented due to their interaction with this field . for this field configuration ,
we check now the validity of the adiabatic predictions @xcite by comparing them to a time dependent analysis .
the peak intensity of the gaussian pulse is reached at @xmath125 and kept constant afterwards . at this point ,
if @xmath36 is large enough , the energy gap between the states in a pendular doublet is much smaller than the energy gap with the neighboring doublet .
then , for a certain pendular level , its rotational dynamics for @xmath249 , i.e. , when the static field is switched on , could be approximated by a two - state model involving the two levels forming the corresponding pendular doublet @xcite . at @xmath125 ,
i.e. , @xmath267 and @xmath126 , the pendular states are @xmath268 with @xmath43 and @xmath44 indicating even or odd parity . under this approximation
, we assume that the levels @xmath269 and @xmath270 are right- and wrong - way oriented , respectively .
the two - state - model hamiltonian yields as @xmath271 where we have taken @xmath272 with @xmath273 and @xmath33 being the switching on speed and time , respectively .
this time @xmath33 is chosen so that if these states are exposed only to this field , the turning - on process is adiabatic .
note that we have taken @xmath274 , @xmath275 , and @xmath276 .
the time - dependent schrdinger equation associated to this hamiltonian admits a scaling factor .
that is , when the dynamics is adiabatic using @xmath277 for a pendular doublet with energy splitting @xmath114 at @xmath126 , then , for a peak intensity @xmath278 and @xmath279 , the dynamics is adiabatic for @xmath280 . and ( b ) weight of the adiabatic ground state on its time - dependent wave function as a function of @xmath18 , for turning on speeds @xmath281 ( orange dashed ) , @xmath282 ( red solid ) , @xmath283 ( blue dotted ) and @xmath284 v / scm ( pink dot - dashed ) , and adiabatic results ( thin solid ) the fields are parallel and the gaussian pulse has @xmath83 ns and @xmath85 . ] for the sake of simplicity , we focus on the ground state in a parallel - field configuration . for several switching on speeds , ( a ) and ( b ) display the directional cosine and the population of the adiabatic ground state , respectively , as a function of @xmath18 .
the gaussian pulse has @xmath128 ns fwhm and peak intensity @xmath85 . before the dc field
is turned on , the alignment of the ground state is @xmath285 , the energy separation within this pendular pair is @xmath286 @xmath58 , and there are @xmath287 @xmath58 to the next pendular doublet . for @xmath288 , the coupling term is @xmath289 @xmath58 with @xmath290 . in an adiabatic picture ,
the energy gap @xmath114 can not be neglected , and as @xmath18 is increased the energy of the ground state does not increase linearly with @xmath47 @xcite . for @xmath281
v / scm , the adiabatic parameter @xmath291 , the rotational dynamics is non - adiabatic and there is a population transfer between the two states in this doublet .
we note that for this process , the adiabatic parameter @xmath96 is defined as in but replacing the laser field interaction @xmath103 by the dipole term @xmath14 . the ground state presents a moderate orientation bellow the adiabatic limit due to the contributions of the adiabatic states @xmath178 and @xmath176 , @xmath109 decreases until a minimum value showing a smooth oscillation afterwards , cf .
( b ) . due to the coupling term
, @xmath62 oscillates as @xmath18 is increased , and its frequency is equal to the energy separation between the adiabatic levels @xmath178 and @xmath176 . a similar behavior is observed for @xmath292 v / scm , but the orientation of the ground state oscillates around a value closer to the adiabatic limit because the process is more adiabatic and @xmath293 for @xmath294 . for @xmath295 and @xmath296
v / scm , the dynamics can be considered as adiabatic , with @xmath109 being larger than @xmath297 . however , for @xmath295 , @xmath62 still oscillates around the adiabatic value . by increasing the peak intensity of the laser pulse ,
the energy splitting of the levels in a pendular doublet is decreased , but their coupling due to the dc field is not significantly modified .
thus , the rotational dynamics becomes more diabatic , and larger turning - on times are needed to achieve the adiabatic limit . for @xmath88 ,
the ground state is separated by @xmath298 @xmath58 from @xmath176 , and by @xmath299 @xmath58 from the next pendular doublet .
the coupling due to the dc field is @xmath300 @xmath58 for @xmath288 and with @xmath301 . according to the scaling law of the time - dependent schrdinger equation ,
this process would be adiabatic for a speed of @xmath302 v / scm . for @xmath135 and the ground state ,
we find @xmath303 @xmath58 , @xmath304 @xmath58 to the second doublet , and @xmath305 @xmath58 for @xmath288 and with @xmath306 . within an adiabatic framework , as @xmath47
is increased the ground state energy can be approximated by the pseudo - first - order stark linear effect @xmath307@xmath308@xmath309@xmath176@xmath310 @xcite .
note that @xmath114 is smaller than the dc filed coupling even for @xmath311 .
based on the scaling law of the two - state model schrdinger equation , the dc field should be turned on very slowly , @xmath312 v / scm , to achieve the adiabatic limit . for larger
turning - on speeds ,
the dynamics is so diabatic that the @xmath198 wave function does not change , and its projections on the adiabatic states @xmath178 and @xmath176 are close to the field - free values even for @xmath313 v / scm .
in this work , we have investigated the mixed - field orientation dynamics of linear molecules .
the richness and variety of the field - dressed rotational dynamics has been illustrated by analyzing in detail the directional properties of several low - lying states .
in particular , we have explored the degree of orientation as the peak intensity and fwhm of the gaussian pulse , the electrostatic field strength and the angle between both fields are varied . by considering prototypical field configurations used in current mixed - field orientation experiments ,
we have proven that the assumption of a fully adiabatic dynamics is incorrect . for parallel fields ,
a non - adiabatic transfer of population takes place when the quasi - degenerated pendular doublets are formed as the laser intensity is increased . as a consequence ,
the time - dependent results for the degree of orientation are smaller than the predictions of the adiabatic theory .
using current available experimental peak intensities , longer laser pulses or stronger static fields will increase the degree of orientation even for highly excited states . in particular
, we have provided the field parameters under which the mixed - field orientation dynamics will be fully adiabatic .
we have also shown that the field - dressed dynamics is more complicated if both fields are tilted . apart from the non - adiabatic effects when the pendular doublets are formed , at weak laser intensities
there is also population transfer due to the splitting of the states within a @xmath0-manifold having now the same symmetry . for non - parallel fields
, we have shown that the ground state is strongly oriented , whereas excited states might only present a moderate orientation , and , furthermore , some of them could behave as dark states to the mixed - field orientation dynamics .
the requirements for an adiabatic dynamics are now more difficult to satisfy for excited levels than for the ground state .
again , we have indicated the field configuration that will give rise to an adiabatic mixed - field - orientation .
if the peak intensity is kept constant after turning on the pulse , we have shown that the orientation of the states might oscillate with time due to the non - adiabatic dynamics .
finally , we have investigated the molecular dynamics when the temporal order of the fields is inverted .
we have shown that once the ground state is adiabatically aligned , the switching on of the dc field has to be very slow to achieve a significant orientation .
although our study is restricted to the ocs molecule , we stress that the above - observed physical phenomena are expected to occur in many other polar molecules .
indeed , the hamiltonian can be rescaled , and the above results used to describe another polar linear molecule .
in addition , due the complexity of their rotational level structure of asymmetric tops , these non - adiabatic effects should have a negative impact in their mixed - field orientation experiments @xcite .
we would like to thank b. friedrich , j. kpper , j. h. nielsen , p. schmelcher and h. stapelfeldt for fruitful discussions .
financial support by the spanish project fis2011 - 24540 ( micinn ) , the grants fqm-2445 and fqm-4643 ( junta de andaluca ) , and andalusian research group fqm-207 is gratefully appreciated . j.j.o .
acknowledges the support of me under the program fpu . | we present a theoretical study of the impact of an electrostatic field combined with non - resonant linearly polarized laser pulses on the rotational dynamics of linear molecules . within the rigid rotor approximation
, we solve the time - dependent schrdinger equation for several field configurations . using the ocs molecule as prototype
, the field - dressed dynamics is analyzed in detail for experimentally accessible static field strengths and laser pulses .
results for directional cosines are presented and compared to the predictions of the adiabatic theory .
we demonstrate that for prototypical field configuration used in current mixed - field orientation experiments , the molecular field dynamics is , in general , non - adiabatic , being mandatory a time - dependent description of these systems .
we investigate several field regimes identifying the sources of non - adiabatic effects , and provide the field parameters under which the adiabatic dynamics would be achieved . |
geographical data visualization is a fundamental tool for communicating results related to geospatial analyses , and for generating hypotheses during exploratory data analysis @xcite .
the constantly increasing availability of geolocated data from social media , mobile devices and spatial databases implies that we need new tools for exploring , mining and visualizing large - scale spatial datasets .
the python programming language @xcite has been gaining attention as a data analysis tool in the scientific community @xcite thanks to the clarity and simplicity of its syntax , and due to an abundance of third - parties libraries e.g. within many disciplines including scientific computing @xcite , machine learning @xcite , bayesian modeling @xcite , neuroscience @xcite , and bioinformatics @xcite . currently , however , there is limited support for geographical visualization . here , we introduce geoplotlib , a python toolbox for visualizing geographical data .
geoplotlib provides a simple yet powerful api to generate geographical visualizations on openstreetmap @xcite tiles .
we release geoplotlib as open - source software @xcite , accompanied by a rich set of examples and documentation . in the remainder of this paper
, we discuss existing tools for geographical visualization and document the geoplotlib functionalities in detail , and finally we evaluate the computational performance on a large - scale dataset .
in this section we compare existing tools for visualizing geographical data using python . we divide the related work into three categories : pure python packages , html - based packages and geographical information system plug - ins .
the matplotlib @xcite library has become the de - facto standard for data visualization in python and provides a large array of visualization tools including scatter and line plots , surface views , 3d plots , barcharts , and boxplots , but it does not provide any support for visualization on a geographical map by default . the basemap @xcite and cartopy @xcite packages support multiple geographical projections , and provide several visualizations including point plots , heatmaps , contour plots , and shapefiles .
pysal @xcite is an open - source library of spatial analysis functions written in python and provides a number of basic plotting tools , mainly for shapefiles .
these libraries however do not allow a user to draw on map tiles , and have limited support for custom visualizations , interactivity , and animation .
there is a very rich ecosystem for data visualization for the web .
a number of frameworks allow users to generate plots and charts : we cite as representative protoviz @xcite , d3 @xcite , google charts @xcite , sigmajs @xcite .
there is also a large number of libraries for displaying online tile maps , including google maps @xcite , bing maps @xcite , leaflet @xcite , openlayers @xcite , modestmaps @xcite , polymaps @xcite . in order to generate a html visualization from python code ,
it is needed to generate the html and javascript code that maps the data to the graphical elements .
a number of libraries attempt to automate the conversion , such as folium @xcite , vincent @xcite and mplleaflet @xcite .
this process however is often complex , error - prone and time consuming
. the complexity can be even greater if some support for animation or interaction is needed .
finally , the javascript rendering performance may not be adequate for plotting very large datasets .
geographic information systems ( gis ) such as qgis @xcite , grassgis @xcite , arcgis @xcite , mapinfo @xcite provide very powerful tools for spatial data analysis and visualization .
gis tools usually provide some support for python scripting , although the availability varies from one to another .
the main limitation of gis products is their complexity , requiring a significant amount of training to be used effectively , and as discussed before , the need to export the data from python .
an overview of the geoplotlib architecture is given in fig . [
fig : conceptual ] .
geoplotlib builds on top of numpy @xcite and scipy @xcite for numerical computations , and opengl / pyglet @xcite for graphical rendering .
geoplotlib implements the map rendering , the geographical projection , the user interface interaction and a number of common geographical visualizations .
geoplotlib is designed according to three key principles : * _ simplicity _ : geoplotlib tries to minimize the complexity of designing visualizations by providing a set of built - in tools for the most common tasks such as density visualization , spatial graphs , and shapefiles .
the geoplotlib api is inspired by the matplotlib @xcite programming model and syntax , the de - facto standard for data visualization in python ; this makes it easier for matplotlib users to get started . *
_ integration _ : geoplotlib visualizations are standard python scripts , and may contain any arbitrary python code and use any other package .
there is no need to export to other formats ( e.g. shapefiles , html ) or use external programs .
this supports a complete integration with the rich python data analysis ecosystem such as scientific computing , machine learning and numerical analysis packages .
the visualization can even run within an ipython @xcite session , supporting interactive data analysis and facilitating the iterative design for visualizations . * _ performance _ : under the hood , geoplotlib uses numpy / scipy for fast numerical computations , and pyglet / opengl for hardware - accelerated graphical rendering .
this allows the visualizations to scale to millions of datapoints in realtime .
a simple geoplotlib script looks like this : data = read_csv(data / bus.csv ) geoplotlib.dot(data ) geoplotlib.show ( ) this script launches the geoplotlib window and shows a dot map of the data points , in this example the location of bus stops in denmark ( fig . [ fig : dot ] ) .
geoplotlib automatically determines the map bounding box , downloads the map tiles , perform the geographical projection , draws the base map and the visualization layers ( the dots in this example ) .
the map is interactive and allows a user to zoom and pan with mouse and keyboard .
as discussed above , the usage of the geoplotlib api is very similar to matplotlib .
the visualization canvas is initially empty , and each command adds a new layer of graphics .
the geoplotlib window is displayed when ` show ( ) ` is called .
alternatively , the map can be rendered to image file using ` savefig('filename ' ) ` , or displayed inline in an ipython notebook using ` inline ( ) ` .
the geoplotlib package provides several common geographical visualizations in form of layers .
the api provides convenient methods for quickly adding a new visualization layer . in this section
we provide a summary of the built - in visualizations .
the data for all examples is available on the project website @xcite .
an elementary operation in geographical visualization is to display `` what is where '' , that is to place a graphic element on the map for each of the objects in consideration .
this provides an immediate idea of the absolute and relative locations of objects .
moreover , the density of points directly maps to the density of objects on geographical surface , identifying zones of higher and lower density .
an example of dot map is shown in fig .
[ fig : dot ] .
the dot map shows the spatial distribution of bus stops in denmark at a glance .
the zones of higher density corresponding to the copenhagen metropolitan area and to the other major cities are immediately recognizable .
the ` dot ` method allows users to configure points size , color and transparency , and optionally to attach a dynamic tooltip to each point .
one limitation of dot maps is that it is hard to distinguish between areas of high density , as the number of point is so high that they uniformly cover the visualization canvas .
a more direct visualization of density is to compute a 2d histogram of point coordinates .
a uniformly spaced grid is placed on the map , and the number of samples within each cell is counted .
this value is an approximation of the density , and can be visualized using a color scale . in geoplotlib
we can generate the 2d histogram of the data using ` hist ` : data = read_csv(data / opencellid_dk.csv ) geoplotlib.hist(data , colorscale=sqrt , binsize=8 ) geoplotlib.show ( ) here ` binsize ` refers to the size in pixels of the histogram bins .
the example above loads some data related to cell tower positions in denmark , and then generates a histogram with a specific colorscale and bin size ( fig .
[ fig : hist ] ) . compared to the dot map example
, the histogram provides a clearer depiction of the density distribution .
the main deficiency of histogram visualizations is that they are discrete approximations of a ( effectively continuous ) density function .
this creates a dependence on the bin size and offset , rendering histograms sensitive to noise and outliers . to generate a smoother approximation ,
a kernel density estimator approximates the true density function applying kernel functions in a window around each point @xcite .
the size of this window depends on the bandwidth parameter : a smaller bandwidth will produce more detailed but also noisier estimation , while a larger bandwidth will produce a less detailed but smoother estimation . a kernel estimation function
can then be visualized by a surface where the color encodes the density value ( this visualization is often called a `` heatmap '' ) . in geoplotlib ,
the ` kde ` method generates a kernel density estimation visualization : data = read_csv(data / opencellid_dk.csv ) geoplotlib.kde(data , bw=[5,5 ] ) geoplotlib.show ( ) fig .
[ fig : kde ] shows the kernel density estimation applied to the cell tower data . comparing the histogram from fig .
[ fig : hist ] with the kernel density estimation in fig .
[ fig : kde ] , it is evident how the latter produces a smoother and consequently clearer visualization of density .
the kernel bandwidth ( in screen coordinates ) can be configured to regulate the smoothness .
the density upper bound can be set to clip density values over a threshold .
also the density lower bound can be set , to avoid rendering areas of very low density : # lowering clip_above changes # the max value in the color scale geoplotlib.kde(data , bw=[5,5 ] , cut_below=1e-6 , clip_above=1 ) # different bandwidths geoplotlib.kde(data , bw=[20,20 ] , cmap=coolwarm , cut_below=1e-6 ) geoplotlib.kde(data , bw=[2,2 ] , cmap=coolwarm , cut_below=1e-6 ) # linear colorscale geoplotlib.kde(data , bw=[5,5 ] , cmap=jet , cut_below=1e-6 , scaling=lin ) ) .,scaledwidth=100.0% ] in some cases it is useful to represent objects on the map using custom symbols with specific meaning .
the ` markers ` method allows a user to place customs markers on the map : metro = read_csv(./data / metro.csv ) s_tog = read_csv(./data / s - tog.csv ) geoplotlib.markers(metro , data / m.png , f_tooltip = lambda r : r[name ] ) geoplotlib.markers(s_tog , data / s - tog.png , f_tooltip = lambda r : r[name ] ) geoplotlib.show ( ) fig .
[ fig : markers ] shows an example of custom markers for metro and train stops in copenhagen .
markers graphics can be any common raster format ( png , jpeg , tiff ) , and can be rescaled to a custom size .
optionally a dynamic tooltip can be attached to each marker .
spatial graphs are a special type of graphs where nodes have a well - defined spatial configuration .
examples includes transport networks ( bus routes , train tracks , flight paths ) , supply chain networks , phone call networks and commute networks . in geoplotlib `
graph ` renders a spatial graph : data = read_csv(./data / flights.csv ) geoplotlib.graph(data , src_lat=lat_departure , src_lon=lon_departure , dest_lat=lat_arrival , dest_lon=lon_arrival , color=hot_r , alpha=16 , linewidth=2 ) geoplotlib.show ( ) fig .
[ fig : graph ] shows the resulting spatial graph of airport locations , where each node represents an airport and each edge represents a flight connection .
edges are colored using a colormap encoding the edge length .
a voronoi tessellation @xcite is a partition of space into regions induced by some seed points , so that each region ( called a voronoi cell ) consists of all points closer to a specific seed than to any others .
the analysis of voronoi tessellation is used in numerous fields including ecology , hydrology , epidemiology , mining and mobility studies . in geoplotlib `
voronoi ` can be used to generate a voronoi tessellation visualization .
voronoi cell fill , shading and colors can be configured .
data = read_csv(data / bus.csv ) geoplotlib.voronoi(data , line_color=b ) geoplotlib.show ( ) fig .
[ fig : voronoi ] provides an example of voronoi tessellation of bus stops in denmark .
voronoi cells provide a measure of the space closer to one stop than any others .
the density of points is also captured by the size of voronoi cells , as smaller cells indicate more densely covered areas .
a delaunay triangulation @xcite is a convenient method for generating triangles meshes from a set of points . in geoplotlib
the ` delaunay ` method can be used for this purpose .
the edge color can be configured to a fixed value , or to encode the length of the edges .
data = read_csv(data / bus.csv ) geoplotlib.delaunay(data , cmap=hot_r ) geoplotlib.show ( ) fig .
[ fig : delaunay ] shows the delaunay triangulation of bus stops , with edges colored according to length .
a convex hull @xcite of a set of finite points is the smallest convex polygon that contains all the points .
convex hulls can be used for example to visualize the approximate area corresponding to a set of points . in geoplotlib : geoplotlib.convexhull(data , color , fill = true ) fig .
[ fig : convexhull ] shows the bus stops points split into 6 groups , and each group is represented by a differently colored convex hull .
shapefile @xcite is a popular file format for describing vector graphics for geographical information systems .
geoplotlib uses pyshp @xcite to parse the shapefiles .
the line color can be configured and an optional tooltip can be attached to each shape . in the following example we display the _ kommuner _ administrative regions in denmark ( fig .
[ fig : shapefiles ] ) : geoplotlib.shapefiles(data / dk_kommune / dk_kommune , f_tooltip = lambda attr : attr[stednavn ] , color=[0,0,255 ] ) geoplotlib.show ( ) geojson @xcite is a human - readable format for encoding geographical data , such as polygons and lines .
geoplotlib can render shapes from the geojson format , and shape color and tooltip can be dynamically altered to encode data . for instance geojson shapes can be used to generate a choropleth where each geographic unit is colored to encode a continuous variable . in the following example ( fig . [
fig : choropleth ] ) we generate a choropleth of unemployment in usa @xcite : def get_color(properties ) : key = str(int(properties[state ] ) ) key + = properties[county ] if key in unemployment : return cmap.to_color(unemployment.get(key ) , .15 , lin ) else : return [ 0 , 0 , 0 , 0 ] with open(data / unemployment.json ) as fin : unemployment = json.load(fin ) cmap = colormap(blues , alpha=255 , levels=10 ) geoplotlib.geojson(data
/ gz_2010_us_050_00_20m.json , fill = true , color = get_color , f_tooltip = lambda properties : properties[name ] ) geoplotlib.geojson(data / gz_2010_us_050_00_20m.json , fill = false , color=[255 , 255 , 255 , 64 ] ) geoplotlib.show ( )
the ` dataaccessobject ` class is the fundamental interface between the raw data and all the geoplotlib visualizations .
a ` dataaccessobject ` is conceptually similar to a table with one column for each field and one row for each sample .
this paradigm is very common in data analysis terminology , and is equivalent to ndarrays in numpy , and dataframes in pandas and r. a ` dataaccessobject ` can be initialized by reading a comma - separated values ( csv ) file with the built - in ` read_csv ` method , or can be constructed from a python dict , or from a pandas @xcite dataframe : dao1 = dataaccessobject(field1 : somevalues , field2 : othervalues ) dao2 = dataaccessobject(mydataframe ) dao3 = read_csv(somefile.csv ) the only two fields required are ` lat ` and ` lon ` , which represent to the geographic coordinates .
most of the built - in visualization implicitly refer to these two fields to locate entities in space . `
dataaccessobject ` also provides a few method for basic data wrangling , such as filtering , grouping , renaming and deleting rows and columns .
any openstreetmap tile server can be configured using the `
tile_provider ` method ( users are kindly asked to check the tile usage policy for the selected server , and make sure to provide attribution as needed ) . a number of common free tiles providers are supported , including stamen watercolor and toner @xcite , cartodb positron and darkmatter @xcite .
the built - in visualizations provide various commonly used tools for geographical data visualization .
multiple layers can be combined into a single visualization for richer display . for even more complex visualizations
, geoplotlib allows users to define custom layers . in order to generate a new visualization
, a new class extending ` baselayer ` must be defined .
the custom layer must at least define an ` invalidate ` and a ` draw ` method .
the ` invalidate ` method is called each time the map projection must be recalculated , which typically happens each time that the map zoom - level changes . the ` invalidate ` method receives a ` projection ` object , which provides methods for transforming the data points from the geographic coordinates to screen coordinates .
the screen coordinates can then be passed to a ` batchpainter ` object for the rendering .
a ` batchpainter ` can efficiently draw opengl primitives such as points , lines and polygons .
the ` draw ` method is called at each frame , and typically calls the ` batch_draw ` method of the painter prepared during ` invalidate ` .
the following is a complete example of a custom layer , which simply draws samples as points : class customlayer(baselayer ) : def _ _
init__(self , data ) : self.data = data def invalidate(self , proj ) : x , y = proj.lonlat_to_screen(self.data[lon ] , self.data[lat ] ) self.painter = batchpainter ( ) self.painter.points(x , y ) def draw(self , proj , mouse_x , mouse_y , ui_manager ) : self.painter.batch_draw ( ) the final step needed is to add the layer to the visualization using ` add_layer ` , then call ` show ` : geoplotlib.add_layer(customlayer(mydata ) ) geoplotlib.show ( ) a custom layer can be also used for creating animated visualizations .
each time the draw method is called , the custom layer can update its state to the next frame . as an example , let us imagine having data containing the position of an object over time .
a simple animation can use a frame counter , and at each frame render only the datapoint at the current instant : class animatedlayer(baselayer ) : def _ _
init__(self , data ) : self.data = data self.frame_counter = 0 def invalidate(self , proj ) : self.x , self.y = proj.lonlat_to_screen ( self.data[lon ] , self.data[lat ] ) def draw(self , proj , mouse_x , mouse_y , ui_manager ) : self.painter = batchpainter ( ) self.painter.points(self.x[self.frame_counter ] , self.y[self.frame_counter ] ) self.painter.batch_draw ( ) self.frame_counter + = 1 notice that in this case we do not initialize the ` batchpainter ` inside ` invalidate ` , but we create a new one at each frame .
we also keep track of the current frame with the ` frame_counter ` variable .
even this very simple code is able to visualize a non - trivial animation of an object moving over time . to produce a movie from the animation , individual frames
can be captured using the ` screenshot ` method , and then combined together .
colors can be used as additional mapping for encoding information into a visualization .
continuous variables ( for example points density or the edges distances ) can be mapped to a continuous color scale .
the ` colormap ` class allows a user to perform this conversion .
a ` colormap ` object is constructed by passing any of the matplotlib colorscales , and optionally an alpha value and a number of discretization levels .
the ` to_color ` method performs the conversion from real value to color : # hot colormap cmap = colormap(hot ) # reds colormap with transparency cmap = colormap(reds , 128 ) # coolwarm colormap with 4 levels cmap = colormap(coolwarm , levels=4 ) # linear scaling cmap.to_color(10 , 100 , lin ) # logarithmic scaling cmap.to_color(10 , 100 , log ) # square - root scaling cmap.to_color(10 , 100 , sqrt ) discrete variables such as categories can be represented using categorical colormaps .
the ` colorbrewer ` method provides access to the colorbrewer @xcite colors .
categorical colormaps can be also generated from regular colormaps using using ` create_set_cmap ` : cmap1 = colorbrewer([1,2,3 ] ) cmap2 = create_set_cmap(hsv , [ 1,2,3 ] ) the map view is determined by the projection parameters : the latitude offset , the longitude offset and the zoom level . by default ,
the projection is chosen so to fit all selected points , with the maximum zoom level possible .
the view can changed to a specific portion of the map by passing a ` boundingbox ` object to the ` set_bbox ` method .
a ` boundingbox ` object defines the map view boundaries , and can be constructed in multiple ways .
the most direct way is to specify two ranges of latitudes and longitudes .
alternatively , a ` boundingbox ` can be constructed to fit a subset of points using the ` from_points ` methods .
bbox1 = boundingbox(north=51.3 , west=-124.3 , south=14.8 , east=-56.8 ) bbox2 = boundingbox.from_points(lons , lats ) finally , geoplotlib allows users to create interactive visualizations by provides support for rendering a user interface , and dynamically changing the visualization on user input : * on - screen text such as information or status can be added using the ` uimanager ` class .
* mouseover tooltips can be configured on arbitrary graphical elements or screen regions using the ` hotspotmanager ` class .
* layers can be configured to react to specific key presses by defining a ` on_key_release ` method
we test the performance of geoplotlib by generating some of the described visualization on a dataset consisting of one million samples , using the default visualization parameters .
all tests consider only the time needed for the actual rendering of the visualization , excluding the time for loading the data .
the measurements are repeated 10 times for each visualization type .
the experiments were performed on a macbook pro 2012 with an intel 2.3 ghz i7 cpu , 8 gb ram and nvidia geforce gt 650 m gpu .
table [ tab : performance ] shows that in all cases the visualizations require only a few seconds , thus demonstrating that geoplotlib is suitable even for large - scale datasets .
.execution time for one million samples [ cols="<,>,>",options="header " , ]
we have presented geoplotlib , a python toolbox for generating geographical visualizations .
we demonstrated how geoplotlib provides a simple yet powerful api to visualize geographical data , greatly facilitating exploratory data analysis of geographical information .
we believe that geoplotlib can become a powerful tool in the data analyst toolbox , both for analyzing complex spatial patterns and for communicating results in forms of geographical visualizations .
future work includes the addition of more visualization tools , and the integration of spatial analysis methods .
this work is funded in part by the high resolution networks project ( the villum foundation ) , as well as social fabric ( university of copenhagen ) . | we introduce geoplotlib , an open - source python toolbox for visualizing geographical data .
geoplotlib supports the development of hardware - accelerated interactive visualizations in pure python , and provides implementations of dot maps , kernel density estimation , spatial graphs , voronoi tesselation , shapefiles and many more common spatial visualizations .
we describe geoplotlib design , functionalities and use cases . |
controlling the propagation of acoustic or elastic waves is of fundamental interest for many applications ranging from imaging the living and detecting hazardous components , to information processing and structural health monitoring . owing to this context
, the concept of negative refraction has received a great deal of attention for the last fifteen years . in a negative index material , the energy flow
as dictated by the poynting vector should be in the opposite direction to the wave vector @xcite .
this property implies that at an interface between a positive and a negative index material sound is bent the unusual way relative to the normal . in this paper
, we consider a related phenomenon referred to as negative reflection .
a negative reflecting mirror is an interface at which sound is retro - reflected .
there is a strong analogy with a phase conjugating mirror @xcite , except that , in our case , the negative reflecting mirror does not involve any nonlinear process .
+ until now , negative refraction of elastic waves has been mainly implemented either using the concept of phononic crystals @xcite , or opting for metamaterials @xcite , an arrangement of tailored subwavelength building blocks from which the material gains unusual macroscopic properties .
however such manmade materials usually rely on resonating structures , a feature that induces strong losses .
more recently , an alternative route has been proposed for the negative refraction of guided elastic waves .
an elastic plate actually support an ensemble of modes , the so - called lamb waves , which exhibit complex dispersion properties .
in particular , some lamb modes , often referred to as backward propagating modes in the literature , naturally display a negative phase velocity @xcite .
these modes originate from the repulsion between two dispersion branches having close cut - off frequencies , corresponding to a longitudinal and a transverse thickness mode of the same symmetry .
the lowest branch exhibits a minimum corresponding to a zero - group velocity ( zgv ) point @xcite . above this resonance
, there is a coexistence of a negative phase velocity ( backward ) mode and a positive phase velocity ( forward ) mode .
this peculiar property has been taken advantage of to achieve negative refraction @xcite through mode conversion between forward and backward propagating modes ( or vice versa ) at a step - like thickness discontinuity . in a pionneering experiment , few years ago , germano _ et al .
_ showed that this conversion from forward to backward lamb waves also occurs at a simple free edge of an elastic plate , giving rise to negative reflection @xcite .
more recently , in a concurrent study by veres _
_ , the negative reflection of lamb waves has been investigated above the zgv resonance through numerical simulations and laser - ultrasound experiments @xcite .
+ in this paper , we investigate this phenomenon in - depth both theoretically and experimentally . from the theoretical side ,
we compute the reflection coefficient of a lamb mode impinging on the free edge of the plate as a function of frequency and angle of incidence .
several studies addressed this problem in the past under normal @xcite and oblique @xcite incidence but none of them considered the case of backward lamb modes .
our main result is that , close to the zero - group velocity point , the forward and backward modes are strongly coupled .
they convert preferentially into each other , giving rise to negative reflection over the whole angular spectrum .
these theoretical results are then illustrated through an experiment conducted on a duralumin plate .
the incident mode is selectively generated by a transducer array in contact , and the field reflected at the plate edge is probed by laser interferometry . whereas the pioneering study of germano _ et al . _
@xcite only dealt with the negative reflection of plane waves , a cylindrical incident wave - field is here considered . the free edge is shown to back - focus the incident wave - field at different depths according to the wave number mismatch between the forward and backward propagating modes . in the vicinity of the zgv - point
, this mismatch vanishes and the reflected wave - field back - focuses exactly on the point - source : the free edge then acts as a phase conjugating mirror .
the interaction of an incident lamb wave with the free edge of a semi - infinite plate has been widely studied under normal incidence , using modal decomposition and collocation techniques @xcite .
the case of an oblique incidence has been considered more recently .
et al . _
@xcite and wilcox _
et al . _
@xcite addressed this problem in the low frequency regime using a modal decomposition and a frequency - domain finite element method , respectively .
more recently , santanham _ et al . _
@xcite , and feng _ et al . _
@xcite have investigated the reflection of lamb modes in a higher frequency regime , using a modal decomposition .
nevertheless , in the above mentionned references , the question of backward modes is not addressed as they consider frequency bands over which backward propagation does not occur . in this paper
, we consider the reflection of a lamb wave at the free edge of a plate in the vicinity of a zgv resonance at which forward and backward propagating modes coexist . to that aim
, we extend the modal decomposition method derived by pagneux @xcite under normal incidence to a 3d polarization .
the plate modes are first derived in the right - handed coordinate system @xmath2 whose @xmath3axis lies along the propagation direction of the wave and @xmath4axis along the normal of the plate .
considering a homogeneous , isotropic plate at the pulsation @xmath5 , the elasticity equations are given by @xmath6 where @xmath7 is the density of the material , @xmath8 is the displacement field and @xmath9 $ ] is the stress tensor . considering an isotropic material , the strain - stress relations can be written as @xmath10 } \delta_{ij } + 2 \mu \epsilon_{ij}$ ] , with @xmath11 the lam parameters , @xmath12 the kronecker symbol and @xmath13 $ ] the infinitesimal strain tensor .
the boundary conditions correspond to the cancellation of the stress tensor on the plate surfaces , @xmath14 , where @xmath15 is the normal to the surface boundary . considering the geometry of the problem ,
solutions are in the form : @xmath16 dimensionless variables are obtained by normalizing the components of the displacement field @xmath17 by @xmath18 , the coordinates @xmath19 by @xmath18 and the stress tensor @xmath20 by @xmath21 .
the resulting equations are @xmath22 } \delta_{ij } + 2\epsilon_{ij},\\ \label{eq : elasticity_v2 } \boldsymbol{\sigma}\left(x_2 = \pm 1\right)\cdot\mathbf{n_2 } & = & 0 , \label{eq : elasticity_v4}\end{aligned}\ ] ] with @xmath23 , the dimensionless frequency , @xmath24 , @xmath25 and @xmath26 , the bulk longitudinal and shear velocities , and @xmath27 the normal to the plane @xmath28 .
+ in this system the equations associated with the different displacement components are uncoupled .
this decoupling distinguishes two different families of modes , shear horizontal ( sh ) waves with polarization @xmath29 and lamb waves , polarized in the propagation plane @xmath30 .
both families are composed of an infinite number of modes called propagative for a real wave number , evanescent for an imaginary wave - number and inhomogeneous for a complex wave number . at a given frequency
, there is only a finite number of propagating sh and lamb modes , whereas an infinite number of evanescent sh modes and inhomogeneous or evanescent lamb modes exists .
considering the symmetries of the displacement field with respect to the plane @xmath31 , both lamb and sh modes can be separated in two independent subfamilies of symmetrical and antisymmetrical modes .
symmetric / antisymmetric sh modes correspond to an even / odd @xmath32 polarization along the plate thickness .
symmetric / antisymmetric lamb modes have an even / odd in - plane component @xmath33 combined with an odd / even transverse component @xmath34 .
as the plate edge is symmetrical with respect to the @xmath35 plane , the scattering at a free edge preserves the mode s symmetry . as a consequence , a symmetrical lamb mode
is reflected into symmetrical lamb and sh modes .
+ right - going ( respectively left - going ) propagating modes correspond to a positive ( respectively negative ) group velocity @xmath36 , whereas right - going ( respectively left - going ) evanescent and inhomogeneous modes correspond to wave numbers with strictly positive ( respectively negative ) imaginary parts . in the following ,
each ( lamb and sh ) right - going mode , @xmath37 , is associated with an index @xmath38 .
these modes are ranged in ascending order of their wave numbers imaginary part and descending order of their real part .
each right - going mode , @xmath37 , of wave number @xmath39 can be associated with a left going mode , @xmath40 , of wave number @xmath41 .
dictated by the conservation of the wave - vector component along @xmath42 .
here we show the case of an incident mode @xmath43 that is positively reflected into two forward modes , @xmath44 and @xmath45 , and negatively reflected into a backward mode @xmath46 .
for the sake of clarity , evanscent and inhomogeneous modes are not represented in the figure but note that they are taken into account in our calculation .
note that the poynting vector @xmath47 of the negatively reflected mode @xmath46 is anti - parralel to its wave vector @xmath48 . ]
the plate is now assumed to be semi - infinite , occupying the region @xmath49 , bounded by two parallel planes at @xmath50 , as depicted in fig .
let us consider an incident right - going lamb mode , of wave number @xmath51 , carrying an unit energy flux towards the free edge of the plate with an angle of incidence @xmath52 with respect to @xmath53 , the normal to the free edge , as depicted in fig .
[ fig1 ] . in order to satisfy the stress - free condition , this incident lamb mode
is reflected into an infinite combination of left - going lamb and sh modes of wave numbers @xmath41 .
thus , it is necessary to consider not only the propagating modes but also the different evanescent and inhomogeneous lamb and sh modes . for a given mode @xmath38 , the reflection angles @xmath54 are determined by the conservation of the component @xmath55 of the wave vector along the edge direction .
@xmath56 note that for an incident forward mode associated with a positive angle , the reflection angle is positive for a forward mode and negative for a backward mode .
the strain - displacement field @xmath57 of a reflected mode expressed in the coordinate system @xmath58 , associated with the plate edge , can be obtained from the strain - displacement field @xmath37 expressed in the coordinate system @xmath2 using the following equations : @xmath59 where @xmath60 is the rotation matrix , @xmath61.\ ] ] the stress - free condition at the edge of the plate ( @xmath62 ) can be written as the cancellation of the sum of the incident mode and the reflected modes weighted by their complex reflection coefficients @xmath63 : @xmath64 the solution of this problem is calculated following the method developed by pagneux @xcite and extending it to a 3d polarization .
it combines a collocation discretization along the @xmath65-coordinate and a modal approach .
this approach , fully described in the appendix , allows to derive the amplitude reflection coefficient @xmath66 for any incoming mode and incident angle @xmath52 .
note that , due to reciprocity , @xmath67 . in practice , in the frequency band
starting from @xmath68 to @xmath69 , we have considered 199 lamb modes and 100 sh modes , which allows to satisfy the conservation of energy , since @xmath70 , where the index p denotes a sum over the propagating modes .
our theoretical approach is now applied to the case of a duralumin plate , with a density @xmath71 , a longitudinal wave velocity @xmath72 , and a transverse wave velocity @xmath73 .
the dispersion curves of the first propagating symmetrical lamb and sh modes are shown in fig .
[ fig2a ] . for the sake of clarity , the sh and lamb modes are labeled @xmath74 and @xmath75 , respectively . as highlighted by previous studies
@xcite , over the frequency range from the zgv resonance to the cut - off frequency , a right - going @xmath76 lamb mode is associated with a negative wave number , _
i.e. _ , a negative phase velocity @xmath77 and a positive group velocity @xmath36 resulting in a backward propagation .
this mode is labelled @xmath1 in order to differentiate it from the forward propagating @xmath78 mode .
another crucial feature is the occurrence of a minimum for the @xmath0 and @xmath1 dispersion branches at the same frequency @xmath79 and opposite wave numbers @xmath80 .
this minimum corresponds to a zero - group velocity point , where the group velocity vanishes while the phase velocity remains finite @xcite .
+ figure [ fig2b ] displays the frequency - dependent reflection coefficients for a normally incident @xmath0 wave on a free edge in the frequency band starting from @xmath81 to @xmath69 . at normal incidence , due to the decoupling between lamb and sh modes ,
the incident mode is converted into a combination of lamb modes only . a remarkable feature in fig .
[ fig2b ] is the perfect conversion from the forward @xmath0 mode into the backward @xmath1 mode in the vicinity of the zgv - point .
it arises from the equality between the wave numbers of the incident forward @xmath0 mode and the reflected @xmath1 mode . as a consequence
, these two modes are associated with similar stress - displacement fields .
the only difference lies in their opposite poynting vector directions .
thus , the stress - free boundary condition at the edge of the plate can be satisfied with a simple combination of @xmath0 and @xmath1 modes stress field , leading to a reflection coefficient @xmath82 .
+ nevertheless , this picture is only valid under a normal incidence since it implies a perfect decoupling between the sh and lamb modes .
figure [ fig3a ] shows the evolution of the reflection coefficients @xmath83 with the angle of incidence of the incoming @xmath0 lamb mode at @xmath84 .
far from the zgv - point , the @xmath0 mode is simultaneously reflected into @xmath85 , @xmath0 , @xmath1 and @xmath86 modes . as @xmath87 , the conversion of @xmath0 into @xmath1 only occurs below the critical angle @xmath88 .
interestingly , this conversion remains almost constant for @xmath89 . the @xmath1 mode is thus excited over the whole angular spectrum .
this feature holds close to the zgv - point ( see fig .
[ fig3b ] ) , moreover the conversion is now nearly perfect @xmath90 over the whole angular spectrum @xmath91 .
besides , as @xmath92 , the conversion of an incident @xmath1 mode into @xmath0 is also nearly perfect .
close to the zgv resonance , the free edge thus acts as a nearly perfect negative reflecting mirror for both @xmath0 and @xmath1 modes .
this striking behavior is now investigated experimentally in the next section .
the conversion between forward and backward modes at the free edge of a plate results in negative reflection .
this effect is a direct consequence of their oppositely signed wave numbers and of the conservation of the parallel component of the wave vectors at the edge of the plate . in the first two parts
, we will take advantage of negative reflection to experimentally demonstrate a tunable focusing mirror above the zgv resonance . in a third part
, the free edge will be shown to act as a phase conjugating mirror in the vicinity of the zgv point .
the principle of the focusing plane mirror for an incident @xmath0 lamb mode is depicted in fig .
[ fig4 ] for a plate of thickness @xmath93 mm .
considering a point source @xmath94 emitting selectively the @xmath0 mode at @xmath95mhz [ fig .
[ fig4a ] ] and @xmath96mhz [ fig .
[ fig4b ] ] , the network of the @xmath1 reflected rays corresponding to different angles of incidence cross the principal axis at different distances @xmath97 from the free edge ,
@xmath98 at the critical angle of incidence @xmath99 , the intersection with the principal axis always occurs on the edge . on the contrary , in the limit of small incident angles
, it occurs at a maximal distance @xmath100 given by @xmath101 where @xmath102 is the distance from the source to the edge .
the non - linearity of @xmath103 as a function of the incident angle @xmath52 [ eq . [ eq : d_d_tan ] ] causes a non - uniform deposit of energy along the principal axis between the edge and the point located at a distance @xmath100 .
[ fig4c ] displays the dependence of @xmath104 with the incident angle @xmath105 at three frequencies .
the concavity of the different curves at each frequency indicates that the density of rays increases gradually on the principal axis as the angle of incidence decreases .
thus , most energy is focused at a distance @xmath100 from the edge .
this is all the more true as the frequency approaches @xmath106 .
+ in order to illustrate the focusing effect induced by the plate edge , the @xmath0 mode generation should be made as selective as possible .
experimentally , this is performed by means of an array of 64 transducers placed at a distance @xmath107 cm from the free edge , as shown in fig .
[ fig5 ] ( see appendix for details ) .
the plate dimensions are chosen so that only the facing free edge causes reflection during the recording time of the experiment .
the out - of - plane displacement is measured from the transducers array to the plate edge with a heterodyne interferometer .
the measurement is made over a grid of points that maps @xmath108@xmath109 of the plate surface , with a pitch of @xmath110 mm .
signals detected by the optical probe are fed into a digital sampling oscilloscope and transferred to a computer . a spatio - temporal discrete fourier transform ( dft ) of the recorded wavefronts
is then performed from @xmath111 to @xmath112 mhz for wave numbers ranging from @xmath113 to @xmath114 mm@xmath115 .
then , the experimental dispersion curves are obtained by angularly integrating the spatial fourier plane at each frequency .
+ mode is generated selectively in the duralumin plate of thickness @xmath116 mm by an array of 64 transducers .
the out - of - plane component of the local vibration is measured with a heterodyne interferometer over a grid of point that maps @xmath108@xmath109 of the plate surface .
the grid pitch is 0.5 mm . ] to characterize the wave - field generated by the array of transducers , a dispersion curve is first computed over a time range of @xmath117s that excludes the echoes due to reflections on the free edge .
the result is displayed in fig .
[ fig6a ] .
it appears that the @xmath0 mode is efficiently generated , unlike the @xmath1 mode .
the antisymmetric @xmath118 mode is also significantly excited .
nevertheless , this mode is only reflected into anti - symmetric ones at the free edge for symmetry reasons .
hence , its excitation does not affect the analysis of symmetrical modes .
the angular dependence of the @xmath0 mode excitation can be evaluated at each frequency by considering the spatial fourier plane . fig .
[ fig6b ] highlights the anisotropy and angular inhomogeneities of our generation scheme at the frequency @xmath119mhz .
this can be accounted for by the finite size of the source and the imperfect coupling between the transducers array and the plate surface . nevertheless , the angular range is sufficient to access the critical angle @xmath120 , hence providing a reflection of the @xmath1 mode over the whole angular spectrum .
[ fig6c ] depicts the dispersion curve computed over a time range from @xmath121 to @xmath122s that includes the waves reflected by the free edge .
as expected theoretically , a significant part of the incident right - going @xmath0 mode is converted into a left - propagating @xmath1 mode at the free edge , resulting in negative reflection .
the left - going backward propagating @xmath1 mode is also associated with a positive wave number @xmath123 , because of its negative phase velocity . figure [ fig6d ] shows the corresponding spatial fourier plane at @xmath124 mhz .
it brings to light the wide angular distribution of the reflected @xmath1 wave that will give rise to the focusing of the reflected wavefront . to demonstrate this focusing ,
the @xmath1 reflected wave - field is extracted from the data by filtering the fourier plane at each frequency with a spatial low pass filter with a cut - off @xmath125 . fig .
[ fig7](a)-(d ) shows the reflected wave - field at different frequencies .
given the position of the source , the focusing distance from the free edge is determined by the ratio @xmath126 as depicted in eq .
( [ eq : d_d ] ) . as this
ratio increases with frequency , the focal spot goes closer to the free edge . in the limit of infinite wavelength for the mode @xmath1 ( _ i.e._at its cut - off frequency ) , the focusing occurs directly on the free edge .
to conclude , fig .
[ fig7 ] is a nice illustration of the tunable focusing property that can exhibit a simple free edge thanks to negative reflection .
the principle of the focusing plane mirror is now studied for an incident @xmath1 lamb mode . slightly above the zgv point , the incident mode is expected to be exclusively reflected into the @xmath0 lamb mode at the free edge of the plate .
the principle of the focusing plane mirror for an incident @xmath1 lamb mode emitted selectively at @xmath127 is depicted in fig .
[ fig8a ] for a plate of thickness @xmath128 mm .
contrary to the case of an incident @xmath0 lamb mode , as @xmath129 , the focusing occurs at a distance further than the distance of the source from the edge . in the limit of small angles ,
it occurs at a minimal distance @xmath100 as given by eq .
( [ eq : d_d ] ) , whereas , in the limit of @xmath130 , it occurs at an infinite distance from the edge . however , as in the previous part , the non - linearity of @xmath131 as a function of @xmath52 causes most of the energy to be focused at a distance @xmath100 from the edge .
+ in this subsection , the transducers array can not be used as it would perturb the reflected wave - field .
hence , the source consists in a piezo - electric transducer of @xmath132 mm - diameter placed at a distance @xmath133 mm from the edge of a duralumin plate of thickness @xmath134 mm . a @xmath135s chirp signal spanning the frequency range @xmath136 mhz
is sent to the transducer which generates a cylindrical incident wavefront in the plate .
as in the previous experiment , the vibration of the plate is measured with a heterodyne interferometer over a grid of points that maps @xmath137@xmath109 of the plate surface .
the dimension of the plate is large enough to avoid reflections from the other edges during the recording time of the experiment .
+ a spatial dft of the recorded wave - field is performed at @xmath127 , over a time range from @xmath121 to @xmath138s [ see fig.[fig8b ] ] . in the spatial fourier plane ,
the inner circle accounts for the cylindrical generation of the backward @xmath1 mode , whereas the contribution from the @xmath0 mode ( outer circle ) only occurs for negative @xmath139 values , meaning it only originates from the reflection on the free edge .
the selective excitation of the @xmath1-mode is explained by the matching of its wavelength with the transducer diameter .
contrary to the previous experiment , as @xmath129 , the reflection is angularly limited to @xmath140 , resulting in an arc of a circle for the reflected @xmath0 mode in the spatial fourier plane .
the corresponding wave - field is investigated in fig .
[ fig8](c)-(e ) . fig .
[ fig8c ] displays the wave - field resulting from the superposition of the backward @xmath1 mode generated by the transducer and the forward @xmath0 mode reflected by the free edge . in order to clarify our interpretation of data , the contribution from each mode
can be extracted by spatial filtering in the fourier plane .
[ fig8d ] represents the incident @xmath1 mode whose excitation is almost perfectly isotropic , as expected from its circular distribution in the fourier plane [ fig .
[ fig8b ] ] .
the reflected wave - field is shown in fig .
[ fig8e ] . as expected from eq .
( [ eq : d_d_tan ] ) , the focusing occurs at a distance greater than @xmath102 .
moreover , as the @xmath0 mode is reflected over a limited angular range [ see fig . [ fig8b ] ] , the focusing operation is not as efficient as in sec .
[ sec3a ] . the reflected wave - field
is also disturbed by the presence of the transducer whose position is represented by a black dashed line in figs .
[ fig8](c)-(e ) . at last , it is important to note that , whereas , in the first experiment , all the reflected rays were of finite length [ fig .
[ fig4 ] ] , paths of infinite length ( @xmath141 ) here can not be collected during the recording time of our measurements . as shown in sec .
[ sec2 ] , the conversion between the @xmath0 and @xmath1 modes at a free edge is almost perfect in the vicinity of the zgv point .
moreover , for a point - like source , the negatively reflected wave - field should focus back exactly at the initial source location ( see fig .
[ fig4 ] ) .
hence the free edge should act as a phase conjugating mirror near the zgv resonance .
the aim of this subsection is to investigate experimentally this striking behavior .
+ the experimental configuration is described in the previous subsection .
the wave is generated at the frequency - thickness product @xmath142 , assimilated to the zgv point .
a spatial dft of the recorded wave - field is performed over the first @xmath122s including the waves reflected by the free edge .
the associated spatial fourrier transform is displayed in fig .
[ fig9a ] .
two circles associated with close @xmath123 values correspond to the forward @xmath0 ( outer circle ) and the backward @xmath1 ( inner circle ) modes .
an inverse dft yields the corresponding wave - field in the real space [ see fig .
[ fig9b ] ] .
contrary to the wave - field measured above the zgv resonance [ fig .
[ fig8c ] ] , the cylindrical incident and reflected wave - fronts are here impossible to discriminate as they exhibit the same curvature .
+ to extract the contribution of the incident wave - field , a spatial dft is performed on the first @xmath143s .
this time window excludes the reflection on the free edge of most slow modes [ see fig .
[ fig9c ] ] . as @xmath144 in the vicinity of the zgv resonance
, the generation can not be selective and both modes are excited .
however , it appears that the @xmath1 mode is still preferentially and isotropically generated than the @xmath0 mode .
the corresponding wave - field is displayed in the real space in fig .
[ fig9d ] .
as expected from its circular distribution in the fourier plane , the incident wave - field is almost perfectly isotropic . + the reflected contribution is now isolated by computing the dft of the recorded wave - field over a time window from @xmath145 to @xmath122s [ see fig .
[ fig9e ] ] .
however , because of their near zero group velocities , the incident and reflected waves can not be perfectly separated . the angular range labelled ( 1 ) in fig .
[ fig9e ] ( small angles of reflection ) corresponds to the reflected wave - field whereas the angular range labelled ( 2 ) in fig .
[ fig9e ] ( large angles ) corresponds to the residual incident wave - front .
[ fig9e ] shows that , in agreement with our theoretical calculations , the incident @xmath1 mode is efficiently reflected into the right - going @xmath0 mode @xmath146 .
an inverse dft of the signal in the angular range labelled ( 1 ) in fig .
[ fig9e ] yields the reflected wave - field in the real space [ see fig . [ fig9f ] ] .
as expected , the reflected wave focuses back on the source location whose position is represented by a black dashed circle in fig .
[ fig9f ] .
the free edge is shown to act as a phase conjugating mirror .
note that the presence of the transducer pertubs the reflected wave - field and prevents us from observing its propagation at later times .
+ this experiment demonstrates that the free edge of the plate acts as as a passive phase conjugating mirror in the vicinity of the zgv resonance .
the negative reflection phenomenon thus offers numerous possibilities to control the propagation of lamb waves in the vicinity of a free edge .
in summary , we have studied semi - analytically and experimentally the interaction of lamb waves with the free edge of a plate . in particular , the conversion between forward and backward lamb modes , that gives rise to negative reflection , has been investigated in details .
interestingly , it has been theoretically shown that this conversion is uniform over most of the angular spectrum .
experimentally , this has been taken advantage of to show how a simple free edge can act as an efficient tunable focusing mirror .
last but not least , close to the zgv resonance , the conversion between forward and backward modes has been shown to be total and the plate edge shown experimentally to act as a phase conjugating mirror .
this striking behavior opens new perspectives for the manipulation of lamb waves .
in particular , negative reflection gives rise to enhanced back - scattering that can be taken advantage of for the detection of defects in non destructive testing .
besides , an association of the phase conjugation phenomenon with the sensitivity of the zgv resonance @xcite can be fruitful and may lead to a new type of acoustic sensors .
the authors wish to thank t. w. murray for drawing our attention to the topic and v. pagneux for fruitful discussions .
the authors are grateful for funding provided by labex wifi ( laboratory of excellence within the french program investments for the future , anr-10-labx-24 and anr-10-idex-0001 - 02 psl * ) and by the agence nationale de la recherche ( anr-15-ce24 - 0014 - 01 , research project coppola ) .
b.g . acknowledges financial support from the french `` direction gnrale de larmement''(dga ) .
the numerical method presented in this appendix is adapted from the one implemented by pagneux @xcite .
the elasticity equations ( [ eq : elasticity_v1 ] ) and ( [ eq : elasticity_v2 ] ) can be written @xmath147 where @xmath148 , @xmath149 , @xmath150 and where @xmath151 , @xmath152 and @xmath153 are the operators , @xmath154 \end{array}\right ) , \label{eq : f}\ ] ] @xmath155 @xmath156 the boundary condition expressed in eq .
( [ eq : elasticity_v4 ] ) can be expressed in term of @xmath157 by writing @xmath158 then , the whole problem as the boundary conditions can be expressed in term of @xmath159 , @xmath157 and @xmath160 .
the problem is solved numerically by using the matlab differentiation suite developed by weideman and reddy @xcite .
this suite can be used to solve ordinary differential equations using a chebyshev collocation spectral method .
equations are spatially discretized along the @xmath65 axis .
any function @xmath161 on the domain @xmath162 $ ] may also be approximated as @xmath163 , where , @xmath164 , the functions @xmath165 are the chebyshev polynomials of order @xmath123 and the interpolation points are the chebyshev - gauss - labatto points @xmath166 , for @xmath167 .
then , the spatial derivatives of @xmath168 along the @xmath65 coordinate can be evaluated at the collocation points by the matrix multiplication with the differentiation matrices @xmath169 . to solve our problem ,
the displacement and stress tensor components are written respectively as @xmath170 the implementation of the boundary conditions @xmath171 has to be performed carefully .
the stress - free condition @xmath172 leads to @xmath173 .
it can also be implemented by considering @xmath174 as a vector of dimension @xmath175 .
the boundary condition @xmath176 is implemented with the use of eq .
( [ eq : boundary_r_y ] ) , leading to the direct relation between @xmath177 and @xmath178 , for @xmath179 and @xmath180 , @xmath181 \mathbf{l_i^t\mathbf{\hat{u}_2}}$ ] , where @xmath182 is the vector for the @xmath183 row of the differentiation matrix @xmath184 .
the stress - free condition @xmath185 is implemented by calculating the corresponding differentiation matrix @xmath186 incorporating robin conditions ( see ref .
@xcite for details ) . with this discretization , eqs .
( [ u_i_app ] ) and ( [ sigma_i_app ] ) become @xmath187 @xmath188 @xmath189 with @xmath190 @xmath191 where @xmath192 ( respectively @xmath193 ) is the first ( respectively the last ) column of the differentiation matrix @xmath184 and @xmath194 is the identity matrix .
each matrix has to be understood as having the dimension in agreement with the vectors it links .
+ considering @xmath195 an even number , the problem can be reduced to the symmetric ( respectively to the antisymmetric ) problem by imposing the corresponding parity of the different displacement and stress tensor components .
for example , in case of a symmetric problem , @xmath196 , @xmath29 , @xmath197 and @xmath198 are even functions whereas @xmath199 and @xmath200 are odd functions .
the diagonalization of this system of ordinary differential equations yields @xmath201 eigenvalues and eigenvectors , corresponding to the @xmath202 lamb modes and @xmath203 sh modes , which halves correspond to right - going modes and left - going modes . considering an incident propagating right - going lamb mode with an angle @xmath52 with respect to the normal direction , one can define the reflection angles @xmath204 corresponding to the different left - going modes , associated with the wave numbers @xmath41 using eq .
( [ eq : angle_reflection ] ) . the strain - displacement field associated with each reflected mode
is deduced by applying a rotation operator to the corresponding eigenvector ( eqs.([u_rotation])-([sigma_rotation ] ) ) @xmath205 @xmath206 where @xmath207 can be deduced from @xmath159 , @xmath208 $ ] .
finally , the complex reflection coefficient can be computed using eq .
( [ eq : boundary_condition ] ) . considering an obliquely incident
right - going lamb mode , there are @xmath209 coefficients to determine the @xmath210 left - going lamb modes and the @xmath211 left - going sh modes .
these coefficients satisfy @xmath212 equations derived from the stress - free condition at the plate edge ( eq . ( [ eq : boundary_condition ] ) ) .
the cancellation of @xmath213 , @xmath214 and @xmath215 gives @xmath211 , @xmath216 and @xmath211 independent conditions , respectively .
the computation is done with @xmath217 .
the selective generation of the @xmath0 mode is achieved by using an array of 64 transducers .
the transmitted signals are calculated using a 2d inverse dft of the selected part of the dispersion curve @xmath218 , priorly convoluted with a gaussian structuring element .
the @xmath0 mode was selected between @xmath219 and @xmath220 mhz .
the transducer array is driven by a lecoeur electronics 64-channel programmable device which allows for an @xmath221 mhz sampling frequency . the b - scan corresponding to the spatio - temporal excitation by the array of transducers
is displayed in fig .
[ appendix_b ] .
the array element number 64 is the closest to the free edge .
b. morvan , a. tinel , a .- c .
hladky - hennion , j. vasseur , and b. dubus , experimental demonstration of the negative refraction of a transverse elastic wave in a two - dimensional solid phononic crystal , _ http://dx.doi.org/10.1063/1.3302456[appl .
phys . lett . ]
_ * 96 * , 101905 ( 2010 ) . c. cronne , e. manga , b. morvan , a. tinel , b. dubus , j. vasseur , and a .- c .
hladky - hennion , negative refraction of longitudinal waves in a two - dimensional solid - solid phononic crystal , _ http://dx.doi.org/10.1103/physrevb.83.054301[phys .
b. ] _ * 83 * , 054301 ( 2011 ) .
j. pierre , o. boyko , l. belliard , j. vasseur , and b. bonello , negative refraction of zero order flexural lamb waves through a two - dimensional phononic crystal , _ http://dx.doi.org/10.1063/1.3491290[appl .
lett .. ] _ * 97 * , 121919 ( 2010 ) . m. dubois , m. farhat , e. bossy , s. enoch , s. guenneau , and p. sebbah , flat lens for pulse focusing of elastic waves in thin plates , _
phys . lett . ]
_ * 103 * , 071915 ( 2013 ) .
r. zhu , x. liu , g. hu , c. sun , and g. huang , negative refraction of elastic waves at the deep - subwavelength scale in a single - phase metamaterial , _ http://dx.doi.org/10.1038/ncomms6510[nat . com . ] _ * 5 * ( 2014 ) .
c. prada , o. balogun , and t. murray , laser - based ultrasonic generation and detection of zero - group velocity lamb waves in thin plates , _
phys . lett . ]
_ * 87 * , 194109 ( 2005 ) .
s. bramhavar , c. prada , a. a. maznev , a. g. every , t. b. norris , and t. w. murray , negative refraction and focusing of elastic lamb waves at an interface , _ http://dx.doi.org/10.1103/physrevb.83.014106[phys .
b. ] _ * 83 * , 014106 ( 2011 ) .
m. germano , a. alippi , a. bettucci , and g. mancuso , anomalous and negative reflection of lamb waves in mode conversion , _
b. ] _ * 85 * , 012102 ( 2012 ) . i. a. veres , c. grnsteidl , d. m. stobbe , and t. w. murray , broad - angle negative reflection and focusing of elastic waves from a plate edge , _ http://dx.doi.org/10.1103/physrevb.93.174304[phys .
b. ] _ * 93 * , 174304 ( 2016 ) .
b. a. auld and e. m. tsao , a variational analysis of edge resonance in a semi - infinite plate , _ http://dx.doi.org/10.1109/t-su.1977.30952[ieee transactions on sonics and ultrasonics ] _ * 24 * , 317326 ( 1977 ) . b. morvan , n. wilkie - chancellier , h. duflo , a. tinel , and j. duclos , lamb wave reflection at the free edge of a plate , _ http://dx.doi.org/10.1121/1.1539521[j .
soc . am . ] _ * 113 * , 14171425 ( 2003 ) .
p. d. wilcox , a. velichko , b. w. drinkwater , a. j. croxford , and m. d. todd , scattering of plane guided waves obliquely incident on a straight feature with uniform cross - section , _
soc . am . ] _ * 128 * , 27152725 ( 2010 ) . | the paper studies the interaction of lamb waves with the free edge of a plate .
the reflection coefficients of a lamb mode at a plate free edge are calculated using a semi - analytical method , as a function of frequency and angle of incidence .
the conversion between forward and backward lamb modes is thoroughly investigated .
it is shown that , at the zero - group velocity ( zgv ) frequency , the forward @xmath0 lamb mode fully converts into the backward @xmath1 lamb mode at normal incidence .
besides , this conversion is very efficient over most of the angular spectrum and remains dominant at frequencies just above the zgv - point .
this effect is observed experimentally on a duralumin plate .
firstly , the @xmath0 lamb mode is selectively generated using a transducer array , secondly the @xmath1 mode is excited using a single circular transducer .
the normal displacement field is probed with an interferometer .
the free edge is shown to retro - focus the incident wave at different depths depending on the wave number mismatch between the forward and backward propagating modes . in the vicinity of the zgv - point ,
wave numbers coincide and the wave is retro - reflected on the source . in this frequency range ,
the free edge acts as a perfect phase conjugating mirror . |
the log - gamma polymer was introduced and studied by sepplinen @xcite .
let @xmath3 be a positive real .
a random variable @xmath4 has _ inverse - gamma distribution with parameter @xmath5 _ if it is supported on the positive reals where it has distribution @xmath6 we abbreviate this @xmath7 .
the _ log - gamma polymer partition function _ with parameter @xmath8 is given by @xmath9 where @xmath10 is an up / right directed lattice path from the euclidean point @xmath11 to @xmath12 and where @xmath13 . in @xcite
it was proved that @xmath14 where @xmath15 and @xmath16 is a large constant . here
@xmath17'(x)$ ] is the digamma function .
the scale of the variance upper - bound is believed to be tight , since directed polymers at positive temperature should have kpz universality class scalings ( see e.g. the review @xcite ) .
moreover , it is believed that , when centered by @xmath18 and scaled by @xmath0 , the distribution of the free energy @xmath19 should limit to the gue tracy - widom distribution @xcite .
we presently prove this form of kpz universality for the log - gamma polymer for @xmath20 for some @xmath21 .
this assumption is purely technical and comes from the asymptotic analysis .
it is likely that this assumption can be removed following the approach of @xcite , where a similar assumption was removed in the case of the semi - discrete polymer . for
this model @xmath22 plays a role akin to temperature .
[ thm : fluct ] there exists @xmath21 such that the log - gamma polymer free energy with parameter @xmath23 has limiting fluctuation distribution given by @xmath24 where @xmath15 , @xmath25 and @xmath26 is the gue tracy - widom distribution function .
we give the proof of this theorem in section [ sec : asym ] .
there are two ingredients in the proof , and then some asymptotic analysis .
the first ingredient is the @xmath2-fold integral formula given in @xcite for the laplace transform of the polymer partition function .
this is given below as proposition [ coszform ] .
the second ingredient in the proof is a general identity between a class of @xmath2-fold contour integrals and a class of fredholm determinants .
this is given below as theorem [ thm : equiv ] .
applying this identity to proposition [ coszform ] yields corollary [ cor : coszdet ] which is a new fredholm determinant expression for the laplace transform of the log - gamma polymer partition function .
this formula lends itself to straightforward asymptotic analysis , as is done in section [ sec : asym ] .
the log - gamma polymer may be generalized , as done in @xcite , so that the distributions of the @xmath27 depend on two collections of parameters .
discrete directed polymer partition functions , under _ intermediate disorder _ scaling @xcite , converge to the solution of the multiplicative stochastic heat equation ( whose logarithm is the kpz equation ) . if the two collections of parameters determining the distributions of the @xmath27 are tuned correctly , then the initial data for the limiting stochastic heat equation is determined by two collections of parameters as well .
the fredholm determinant formula of corollary [ cor : coszdet ] should limit to an analogous formula for the laplace transform of the stochastic heat equation with this general class of initial data which would be a finite temperature analog of the results of @xcite .
when only one of the collections of parameters is tuned , this formula was computed in @xcite via a similar limit of the fredholm determinant formula for the laplace transform of the semi - discrete polymer partition function ( see also @xcite in the case where only a single parameter is tuned ) , and this is a finite temperature analog of the results of @xcite . the _ sklyanin measure _
@xmath28 on @xmath29 is given by @xmath30 the following result is taken from @xcite , theorem 3.8.ii .
[ coszform ] fix @xmath31 , and choose parameters @xmath32 for @xmath33 and @xmath34 for @xmath35 such that @xmath36 .
consider the log - gamma polymer partition function where @xmath37 .
then for all @xmath38 with @xmath39 @xmath40 = \int_{({{\rm i}}{\ensuremath{\mathbb{r}}})^n } s_n(dw_1,\ldots , dw_n ) \prod_{i , j=1}^{n } \gamma(a_j - w_i ) \prod_{j=1}^{n } \frac{f(w_j)}{f(a_j)},\ ] ] where @xmath41 until now , there has not been progress in extracting asymptotics from this formula .
the following theorem , however , transforms this integral formula into a fredholm determinant for which we can readily perform asymptotic analysis .
this identity should be considered the main technical contribution of this paper , of which theorem [ thm : fluct ] is essentially a corollary ( after some asymptotic analysis ) .
we introduce the following contours : @xmath42 is a positively oriented circle around the origin with radius @xmath43 ; @xmath44 is a line parallel to the imaginary axis from @xmath45 to @xmath46 ; @xmath47 is similarly the contour from @xmath48 to @xmath49 ; and @xmath50 is the horizontal line segment going from @xmath51 to @xmath52 . for any simple smooth contour @xmath22 in @xmath53 we will write @xmath54 to mean the space @xmath55 where @xmath56 is the path measure along @xmath22 divided by @xmath57 . [ thm : equiv ] fix @xmath58 , @xmath59 and @xmath60 such that @xmath61
. suppose @xmath62 is a meromorphic function such that all its poles have real part strictly larger than @xmath63 , @xmath62 is non - zero along and inside @xmath64 , and for all @xmath65 @xmath66{}0.\label{eq : decayassum}\ ] ] then @xmath67 where @xmath68 is the fredholm determinant of @xmath69 with @xmath70 the proof of this theorem is given in section [ sec : equiv ] .
the argument uses only the andrief identity @xcite ( sometimes referred to as the generalized cauchy - binet identity ) , the cauchy determinant , the fact that @xmath71 and some simple contour shifts and residue computations to go from the fredholm determinant formula on the right hand side of to the integral formula on the left hand side .
going from the integral formula to the fredholm determinant formula ( even knowing that it should be true ) is a much more challenging path , since one has to undo certain cancelations to discover determinants . in the context of the related semi - discrete polymer ( see the end of the introduction ) oconnell describes ( after corollary 4.2 of @xcite ) another fredholm determinant expression for the laplace transform of the partition function .
that expression and the one which arises from the application of theorem [ thm : equiv ] are different .
the condition in theorem [ thm : equiv ] on the @xmath61 can be relaxed at the cost of more complicated choices of contours . instead of taking the @xmath72 contour to be a vertical line @xmath73 one could , in order to accommodate larger values of the @xmath74 , shift the vertical line horizontally to the right by some positive integer @xmath75 , and then augment the @xmath72 contour with a collection of sufficiently small contours around the positive integers up to and including @xmath75 .
see section 5 of @xcite for an example of this sort of procedure
. we may apply theorem [ thm : equiv ] to show the following .
[ cor : coszdet ] fix @xmath31 and any @xmath76 such that @xmath58 and @xmath59 .
given a collection of @xmath77 for @xmath33 , and @xmath78 for @xmath79 , set @xmath80 and consider the log - gamma polymer partition function with weights @xmath37 .
then for all @xmath38 with @xmath81 @xmath82 = \det(i+k_u)_{l^2(c_{\delta_1})},\ ] ] where @xmath68 is the fredholm determinant of @xmath69 with @xmath83 and @xmath84 as given in .
we start with the laplace transform formula given in proposition [ coszform ] ( if some @xmath85 we simply shift the @xmath86 integration contour slightly to the left ) . in order to apply theorem [ thm : equiv ]
we must check that all the conditions are satisfied .
the function @xmath84 has no poles with real part less than @xmath87 and it is non - zero in the entire complex plane .
this implies that given @xmath88 and @xmath63 as specified in the hypothesis , the conditions on the poles and zeros of @xmath62 are satisfied .
it remains to check the decay condition .
this , however , is immediate from the estimate on the gamma function as given in .
note that the condition @xmath31 becomes important in this case ( actually @xmath89 would do ) .
the conditions having been satisfied , we may apply theorem [ thm : equiv ] to arrive at the corollary .
the asymptotic analysis of this fredholm determinant is performed in section [ sec : asym ] and yields the proof of theorem [ thm : fluct ] .
the existence of such an identity as in theorem [ thm : fluct ] did not arise out of the blue .
let us briefly explain the two results which suggested this identity ( though only in the special case of @xmath84 as in ) .
an _ up / right path _ in @xmath90 is an increasing path which either proceeds to the right or jumps up by one unit . for each sequence
@xmath91 we can associate an up / right path @xmath92 from @xmath93 to @xmath94 which jumps between the points @xmath95 and @xmath96 , for @xmath97 , and is continuous otherwise .
fix a real vector @xmath98 and let @xmath99 for @xmath100 be independent standard brownian motions such that @xmath101 has drift @xmath74 .
define the _ energy _ of a path @xmath92 to be @xmath102 then the _ oconnell - yor semi - discrete directed polymer partition function _
@xmath103 is given by @xmath104 where the integral is with respect to lebesgue measure on the euclidean set of all up / right paths @xmath92 ( i.e. , the simplex of jumping times @xmath91 ) .
due to the invariance principle , the semi - discrete polymer is a universal scaling limit for discrete directed polymers when @xmath105 is fixed and @xmath2 goes to infinity ( and temperature is suitably scaled see for instance @xcite ) . as such
, the semi - discrete polymer inherits the solvability of the log - gamma polymer .
in fact , before the work of @xcite on the log - gamma polymer , oconnell @xcite showed ( see corollary 4.2 ) that @xmath106 $ ] was given by the left hand side of with @xmath107 on the other hand , the theory of macdonald processes was developed in @xcite . using macdonald difference operators and the cauchy identity for macdonald polynomials @xcite computes expectations of certain observables of the macdonald processes , which
, when formed into generating functions , lead to fredholm determinants .
after a particular limit transition , the macdonald processes become the whittaker processes introduced in @xcite and the generating functions become a fredholm determinant expression for @xmath106 $ ] exactly in the form of the one on the right hand side of .
in particular the following result appeared in @xcite as theorem 5.2.10 ( a change of variables @xmath108 brings it exactly to the form of ) .
fix @xmath109 and a drift vector @xmath110 .
fix @xmath58 , and @xmath111 such that @xmath61 .
then for @xmath112 , @xmath113 = \det(i+ k_{u})\ ] ] where @xmath114 is the fredholm determinant of @xmath115 for @xmath116 a positively oriented contour containing @xmath117 and such that for all @xmath118 , we have @xmath119 .
the operator @xmath120 is defined in terms of its integral kernel @xmath121 this provides a very indirect proof of the identity given in , in the very particular case of @xmath84 as in .
an obvious question this development raised was to provide a direct proof of this identity and to understand how general it is .
this is what theorem [ thm : equiv ] accomplishes .
it is worth noting that macdonald processes also have a limit transition to the whittaker processes described in @xcite , which are connected to the log - gamma polymer .
the fredholm determinant given in corollary [ cor : coszdet ] was anticipated in @xcite via the formalism of macdonald processes yet its rigorous proof was so far lacking because of the nontriviality of certain decay estimates required by that approach .
if one only cares about the gue tracy - widom asymptotics then the approach given here provides a direct , though non - obvious and rather ad hoc , route from the integral formula of @xcite to the fredholm determinant .
ab was partially supported by the nsf grant dms-1056390 .
ic was partially supported by the nsf through dms-1208998 as well as by the clay research fellowship and by microsoft research through the schramm memorial fellowship .
dr was partially supported by the natural science and engineering research council of canada , by a fields - ontario postdoctoral fellowship and by fondecyt grant 1120309 .
dr is appreciative for mit s hospitality during the visit in which this project was initiated .
let us first recall two useful lemmas . in performing steepest descent analysis on fredholm determinants ,
the following allows us to deform contours to descent curves .
[ twprop1 ] suppose @xmath122 is a deformation of closed curves and a kernel @xmath123 is analytic in a neighborhood of @xmath124 for each @xmath125 .
then the fredholm determinant of @xmath126 acting on @xmath127 is independent of @xmath125 .
[ problemma1 ] consider a sequence of functions @xmath128 mapping @xmath129 $ ] such that for each @xmath2 , @xmath130 is strictly decreasing in @xmath131 with a limit of @xmath132 at @xmath133 and @xmath134 at @xmath135 , and for each @xmath136 , on @xmath137 $ ] @xmath138 converges uniformly to @xmath139 . define the @xmath140-shift of @xmath138 as @xmath141 .
consider a sequence of random variables @xmath142 such that for each @xmath143 , @xmath144 \to p(r)\ ] ] and assume that @xmath145 is a continuous probability distribution function .
then @xmath142 converges weakly in distribution to a random variable @xmath4 which is distributed according to @xmath146 . consider the function @xmath147 and define @xmath148 .
observe that this sequence of functions meets the criteria of lemma [ problemma1 ] .
setting @xmath149 observe that @xmath150 by lemma [ problemma1 ] , if for each @xmath143 we can prove that @xmath151 = p_{\gamma}(r)\ ] ] for @xmath152 a continuous probability distribution function , then it will follow that @xmath153 as well .
now , the starting point of our asymptotics is the fredholm determinant formula given in corollary [ cor : coszdet ] for @xmath154 $ ] .
observe that by setting @xmath155 and @xmath156 , we can choose @xmath88 and @xmath63 as necessary for the corollary to hold .
in particular , that implies that @xmath157 = \det(i+k_{u(n , r,\gamma)})\ ] ] where @xmath158 is the fredholm determinant of @xmath69 with @xmath159 + r n^{1/3}(v - w)\right ) \frac{dw}{w - v'},\ ] ] and @xmath160 we have rewritten the kernel in the form necessary to perform steepest descent analysis .
let us first provide a critical point derivation of the asymptotics .
we will then provide the rigorous proof . besides standard issues of estimation
, we must be careful about manipulating contours due to a mine - field of poles ( this ultimately leads to our present technical limitation that @xmath20 ) .
the idea of steepest descent is to find critical points for the argument of the function in the exponential , and then to deform contours so as to go close to the critical point .
the contours should be engineered such that away from the critical point , the real part of the function in the exponential decays and hence as @xmath2 gets large , has negligible contribution .
this then justifies localizing and rescaling the integration around the critical point .
the order of the first non - zero derivative ( here third order ) determines the rescaling in @xmath2 ( here @xmath0 ) which in turn corresponds with the scale of the fluctuations in the problem we are solving .
it is exactly this third order nature that accounts for the emergence of airy functions and hence the gue tracy - widom distribution .
let us record the first three derivatives of @xmath161 : @xmath162 define @xmath163 such that @xmath164 ; that is , @xmath165 .
notice that @xmath166 was defined such that @xmath167 as well .
around @xmath168 , the first non - vanishing derivative of @xmath169 is the third derivative .
notice that @xmath170 .
this indicates that near @xmath168 , @xmath171 where h.o.t . denotes higher order terms in @xmath172
this cubic behavior suggests rescaling around @xmath163 by the change of variables @xmath173 clearly the steepest descent contour for @xmath174 from @xmath163 departs at an angle of @xmath175 whereas the contour for @xmath72 departs at angle @xmath176 .
the @xmath72 contour must lie to the right of the @xmath174 contour ( so as to avoid the pole from @xmath177 ) .
as @xmath2 goes to infinity , neglecting the contribution away from the critical point , the point - wise limit of the kernel becomes @xmath178 where the kernel acts on the contour @xmath179 ( oriented from negative imaginary part to positive imaginary part ) and the integral in @xmath180 is on the ( likewise oriented ) contour @xmath181 for any horizontal shift @xmath136 .
this is owing to the fact that @xmath182 where the @xmath183 came from the jacobian associated with the change of variables in @xmath174 and @xmath184 .
another change of variables to rescale by @xmath185 results in @xmath186 one now recognizes that the fredholm determinant of this kernel is one way to define the gue tracy - widom distribution ( see for instance lemma 8.6 in @xcite ) .
this shows that the limiting expectation @xmath187 which shows that it is a continuous probability distribution function and thus lemma [ problemma1 ] applies .
this completes the critical point derivation .
the challenge now is to rigorously prove that @xmath158 converges to @xmath188 where the two operators act on their respective @xmath189 spaces and are defined with respect to the kernels above in and .
we will consider the case where @xmath20 for @xmath190 small enough and perform certain estimates given that assumption .
let us record some useful estimates : for @xmath191 close to zero we have ( @xmath192 is being used to represent the euler mascheroni constant @xmath193 ) @xmath194 from this we can immediately observe that for @xmath22 small , @xmath195 with the change of variables @xmath196 we may estimate @xmath197 where the error is uniform for @xmath198 in any compact domain and @xmath199 our approach will be as follows : step 1 : we will deform the contour @xmath64 on which @xmath174 and @xmath184 are integrated as well as the contour on which @xmath72 is integrated so that they both locally follow the steepest descent curve for @xmath161 coming from the critical point @xmath163 and so that along them there is sufficient global decay to ensure that the integral localizes to the vicinity of @xmath163 .
step 2 : in order to show this desired localization we will use a taylor series with remainder estimate in a small ball of radius approximately @xmath22 around @xmath163 , and outside that ball we will use the estimate for the @xmath174 and @xmath184 contour , and a similarly straightforward estimate for the @xmath72 contour .
step 3 : given these estimates we can show convergence of the fredholm determinants as desired .
* step 1 : * define the contour @xmath200 which corresponds to a steep descent contour , we say that @xmath22 is a steep descent path if ( 1 ) @xmath201 reaches the maximum at some @xmath202 : @xmath203 for @xmath204 , and ( 2 ) @xmath201 is monotone along @xmath22 except at its maximum point @xmath205 and , if @xmath22 is closed , at a point @xmath206 where the minimum of @xmath207 is reached .
] of the function @xmath208 leaving @xmath209 at an angle of @xmath210 and returning to @xmath209 at an angle of @xmath211 , given a positive orientation .
in particular we will take @xmath200 to be composed of a line segment from @xmath209 to @xmath212 , then a circular arc ( centered at 0 ) going counter - clockwise until @xmath213 and then a line segment back to @xmath209 .
it is elementary to confirm that along @xmath214 , @xmath207 achieves its maximal value of 0 at the unique point @xmath209 and is negative everywhere else .
the contour @xmath215 will serve as our steep descent contour for @xmath216 ( see figure [ contour1 ] ) .
one should note that for any @xmath217 , for @xmath22 small enough , the contour @xmath215 is such that for @xmath174 and @xmath184 along it , @xmath218 . by virtue of this fact , we may employ proposition [ twprop1 ] to deform our initial contour @xmath64 to the contour @xmath215 without changing the value of the fredholm determinant .
likewise , we must deform the contour along which the @xmath72 integration is performed . given that @xmath219
now , we may again use proposition [ twprop1 ] to deform the @xmath72 integration to a contour @xmath220 which is defined symmetrically over the real axis by a ray from @xmath221 leaving at an angle of @xmath222 ( again , see figure [ contour1 ] ) .
it is easy to see that this deformation does not pass any poles , and with ease one confirms that due to the linear term in @xmath223 the deformation of the infinite contour is justified by suitable decay bounds near infinity between the original and final contours .
thus , the outcome of this first step is that the @xmath224 contour is now given by @xmath225 and the @xmath72 contour is now given by @xmath220 which is independent of @xmath174 . * step 2 : * we will presently provide two types of estimates for our function @xmath169 along the specified contours : those valid in a small ball around the critical point and those valid outside the small ball .
let us first focus on the small ball estimate .
[ lem : smallball ] there exists @xmath21 such that for all @xmath20 the following two facts hold :
\(1 ) there exists a constant @xmath226 such that for all @xmath174 along the straight line segments of @xmath215 : @xmath227 \leq { \ensuremath{\mathrm{re}}}\left [ -c_1\gamma^3 ( v-\bar{t}_{\gamma})^3 \right].\ ] ] \(2 ) there exists a constant @xmath228 such that for all @xmath72 along the contour @xmath220 at distance less than @xmath22 from @xmath163 : @xmath229 \geq { \ensuremath{\mathrm{re}}}\left [ -c_2\gamma^3 ( w-\bar{t}_{\gamma})^3 \right].\ ] ] recall the taylor expansion remainder estimate for a function @xmath230 expanded around @xmath231 , @xmath232 we may apply this to our function @xmath161 around the point @xmath163 giving @xmath233 let @xmath234 and also let @xmath235 . then for @xmath22 small
, we have the following estimate @xmath236 both parts of the lemma follow readily from this estimate and the comparison along the contours of interest of @xmath237 with the right hand side above .
we may now turn to the estimate outside the ball of size @xmath22 .
[ lem : circ ] there exists @xmath21 and @xmath238 such that for all @xmath174 along the circular part of the contour @xmath215 , the following holds : @xmath239 \leq -c.\ ] ] writing @xmath240 we may appeal to the second line of the estimate and the fact that along the circular part of the contour @xmath215 , the real part of function @xmath241 is strictly negative and the error in the estimate is @xmath242 .
the above bound suffices for the @xmath174 contour since it is finite length .
however , the @xmath72 contour is infinite so our estimate must be sufficient to ensure that the contour s tails do not play a role . [
lem : line ] there exists @xmath21 and @xmath238 such that for all @xmath72 along the contour @xmath220 at distance exceeding @xmath22 from @xmath163 , the following holds : @xmath243 \geq { \ensuremath{\mathrm{re}}}\!\left[c\gamma^{-1 } w\right].\ ] ] this estimate is best established in three parts .
we first estimate for @xmath244 between distance @xmath22 and distance @xmath245 from @xmath163 ( @xmath246 large ) .
second we estimate for @xmath72 between distance @xmath247 and distance @xmath248 from @xmath163 .
finally we estimate for all @xmath72 yet further . this third estimate is immediate from the first line of in which the linear term in @xmath191 clearly exceeds the other terms for @xmath22 small enough and @xmath249 . to make the first estimate we use the bottom line of .
the function @xmath208 has sufficient growth along this contour to overwhelm the @xmath250 error as long as @xmath22 is small enough .
the @xmath251 error in is only valid for @xmath198 in a compact domain though .
so , for the second estimate we must use the cruder bound that @xmath252 for @xmath191 along the contour and of size less than @xmath248 from @xmath163 . since in this regime of @xmath191 , @xmath208 behaves like @xmath253 , one sees that for @xmath22 small enough , the @xmath254 error is overwhelmed and the claimed estimate follows .
* step 3 : * we now employ the estimates given above to conclude that @xmath158 converges to @xmath188 where the two operators act on their respective @xmath189 spaces and are defined with respect to the kernels above in and .
the approach is standard ( see for instance @xcite ) so we just briefly review what is done .
convergence can either be shown at the level of fredholm series expansion or trace - class convergence of the operators . focusing on the fredholm series expansion ,
the estimates provided by lemmas [ lem : smallball ] , [ lem : circ ] and [ lem : line ] , along with hadamard s bound , show that for any @xmath255 there is a @xmath256 large enough such that for all @xmath2 , the contribution of the terms of the fredholm series expansion past index @xmath256 can be bounded by @xmath255 .
this localizes the problem of asymptotics to a finite number of integrals involving the kernels .
the same estimates then show that these integrals can be localized in a large window of size @xmath183 around the critical point @xmath163 .
the cost of throwing away the portion of the integrals outside this window can be uniformly ( in @xmath2 ) bounded by @xmath255 , assuming the window is large enough .
finally , after a change of variables to rescale this window by @xmath0 we can use the taylor series with remainder to show that as @xmath2 goes to infinity , the integrals coming from the kernel @xmath257 converge to those coming from @xmath258 .
this last step is essentially the content of the critical point computation given earlier .
let us first recall a version of the andrief identity @xcite ( sometimes referred to as the generalized cauchy - binet identity ) : assuming all integrals exist , @xmath259_{i , j=1}^{n } \det\left[g_i(z_j)\right]_{i , j=1}^{n } = \det\left[\int d\mu(z)\ , f_i(z)g_j(z ) \right]_{i , j=1}^{n}. \label{cauchybinet}\ ] ] we start working from the right hand side of the formula .
we regard the fredhom determinant as given by its fredholm series expansion ( the convergence of the series is a consequence of the identity we will prove ) .
note that our assumptions on @xmath88 and @xmath63 imply that @xmath260 , so the factors @xmath261 have no poles for @xmath262 .
they also imply that the factor @xmath263 in the denominator in is never zero on @xmath73 . to see that the integral is convergent , recall the asymptotics @xmath264 as @xmath265 for @xmath266 ( see ( 6.1.45 ) in @xcite ) , which implies that , for @xmath267 in some finite interval , @xmath268 as @xmath269 for some @xmath226 and @xmath270 .
we also have , under the same assumption on @xmath191 , @xmath271 for some @xmath228 .
then @xmath272 as @xmath273 for some @xmath274 , and hence the integral in is convergent by . observe that @xmath275 with @xmath276 and @xmath277 given by @xmath278 ( checking that @xmath279 and @xmath280 map @xmath281 to @xmath282 and vice - versa is easy , and uses in the case of @xmath279 ) .
let @xmath283 , which is then given by @xmath284 now recall in general that if @xmath279 and @xmath280 have integral kernels so that the integrals @xmath285 and @xmath286 both converge for @xmath287 in the right domains , then @xmath71 in the sense that the two convergent fredholm expansions coincide termwise ( this can be seen as an application of the andrief identity ) . using this
we deduce that @xmath288 recalling that @xmath289 , let @xmath290 and rewrite @xmath291 as @xmath292 by the assumption @xmath293 , @xmath294 and @xmath295 are never zero for @xmath262 and @xmath296 .
thus the only poles of the integrand inside @xmath64 are those coming from @xmath297 , which can only come from the factors @xmath298 in the definition of @xmath216 .
for simplicity we will assume that all the @xmath74 s are pairwise distinct , the general case follows likewise by taking limits .
then the integrand has simple poles at @xmath299 , @xmath300 , and thus @xmath301\\ & = \frac1{2\pi{{\rm i}}}\sum_{i=1}^ng(w')\frac{1}{w - a_i}\frac\pi{\sin(\pi(a_i - w'))}\frac1{f(a_i)}\prod_{m\neq i}\frac{\gamma(a_i - a_m+1)}{a_i - a_m}. \end{aligned}\ ] ] we rewrite this as @xmath302 with @xmath303 in particular , this means that @xmath304 where @xmath305 is given by @xmath306 and @xmath307 by @xmath308 , so that @xmath309 the integral is convergent for the same reason as the integral in , and hence using again the formula @xmath71 we deduce that @xmath310_{i , j=1}^n.\ ] ] our next goal is to shift the contour @xmath73 in the integral appearing on the right hand side of to @xmath311 .
note that , as we do this , we will cross all the @xmath74 s .
we have @xmath312 since all the poles of @xmath62 lie to the right of @xmath73 , the only singularities of @xmath313 we encounter as we shift the contour are those coming from the zeros of the sine , and hence we only see simple poles at @xmath314 in the case @xmath315 . on the other hand ,
the integral of @xmath313 along the segments going from @xmath316 to @xmath317 goes to 0 as @xmath318 by and as before .
the conclusion is that @xmath319+\frac1{2\pi{{\rm i}}}\int_{-{\ell}_{\delta_1}}dw\,f_i(w)g_j(w).\ ] ] since @xmath320=-c_if(a_i)\prod_{m\neq i}\frac{a_i - a_m}{\gamma(a_i - a_m+1)}=-1,\ ] ] we deduce from that @xmath321_{i , j=1}^n\\ & = \frac1{(2\pi{{\rm i}})^nn!}\int_{-{\ell}_{\delta_1}}\!\dotsm\int_{-{\ell}_{\delta_1}}dw_1\dotsm dw_n\,\det[f_i(w_j)]_{i , j=1}^n\det[g_i(w_j)]_{i , j=1}^n , \end{split}\ ] ] where the second equality follows from the andrief identity .
note that the contours in this last integral are exactly the ones appearing in the formula we seek to prove .
hence what is left is to compute the two determinants and show that the integrand coincides with the one appearing in .
we start by observing that @xmath322_{i , j=1}^n$ ] is a cauchy determinant , so @xmath323_{i , j=1}^n=\frac{\prod_{i < j}(a_j - a_i)(w_i - w_j)}{\prod_{i , j=1}^n(w_i - a_j)}.\ ] ] on the other hand @xmath324_{i , j=1}^n=\prod_{i=1}^ng(w_i)c_i\det\!\left[\frac\pi{\sin(\pi(a_i - w_j))}\right]_{i , j=1}^n.\ ] ] the determinant on the right hand side can be turned into another cauchy determinant by writing @xmath325 so that @xmath326_{i , j=1}^n=\prod_{i=1}^n\frac{2\pi{{\rm i}}}{e^{-{{\rm i}}\pi(a_i+w_i ) } } \det\!\left[\frac1{e^{2{{\rm i}}\pi a_i}-e^{2{{\rm i}}\pi w_j}}\right]_{i , j=1}^n,\ ] ] and therefore @xmath327_{i , j=1}^n=(2\pi{{\rm i}})^ne^{{{\rm i}}\pi\sum_i(a_i+w_i)}\prod_{i=1}^ng(w_i)c_i \frac{\prod_{i < j}(e^{2{{\rm i}}\pi a_j}-e^{2{{\rm i}}\pi a_i})(e^{2{{\rm i}}\pi w_i}-e^{2{{\rm i}}\pi w_j})}{\prod_{i , j=1}^n(e^{2{{\rm i}}\pi a_i}-e^{2{{\rm i}}\pi w_j})}.\ ] ] now we collect the factors involving only @xmath328 s , only @xmath74 s , and
cross - terms : from , and the definitions and of @xmath329 and @xmath330 we get @xmath331_{i , j=1}^n\det[g_i(w_j)]_{i , j=1}^n=(2\pi{{\rm i}})^nd_{a , a}d_{w , w}d_{a , w}\ ] ] with @xmath332 @xmath333 can be rearranged in the following way : @xmath334 where have used @xmath335 similar computations give @xmath336 hence @xmath337 and thus from and we deduce that @xmath338 putting this together with and the definition of @xmath28 yields . | we prove that under @xmath0 scaling , the limiting distribution as @xmath1 of the free energy of sepplinen s log - gamma discrete directed polymer is gue tracy - widom . the main technical innovation we provide is a general identity between a class of @xmath2-fold contour integrals and a class of fredholm determinants . applying this identity to the integral formula proved in @xcite for the laplace transform of the log - gamma polymer partition function , we arrive at a fredholm determinant which lends itself to asymptotic analysis ( and thus yields the free energy limit theorem ) .
the fredholm determinant was anticipated in @xcite via the formalism of macdonald processes yet its rigorous proof was so far lacking because of the nontriviality of certain decay estimates required by that approach . |
the present work is supported by the korea research foundation grant funded by the korean government(moehrd ) ( krf-2006 - 312-c00507 ) .
the work of s.n . is supported by the brain korea 21 ( bk21 ) project in center of excellency for developing physics researchers of pusan national university , korea .
the authors are grateful to m. m. musakhanov for critical comments and discussions .
99 m. a. shifman , a. i. vainshtein and v. i. zakharov , nucl .
b * 147 * ( 1979 ) 385 ; nucl .
b * 147 * ( 1979 ) 448 .
j. gasser and h. leutwyler , annals phys .
* 158 * ( 1984 ) 142 .
m. gell - mann , r. j. oakes and b. renner , phys .
* 175 * ( 1968 ) 2195 .
m. a. shifman , a. i. vainshtein and v. i. zakharov , nucl .
b * 147 * ( 1979 ) 448 .
b. l. ioffe , nucl .
b * 188 * ( 1981 ) 317 [ erratum - ibid .
b * 191 * ( 1981 ) 591 ] .
i. i. balitsky , d. diakonov and a. v. yung , z .
c * 33 * ( 1986 ) 265 .
v. l. chernyak and a. r. zhitnitsky , phys . rept .
* 112 * ( 1984 ) 173 .
v. m. belyaev and b. l. ioffe , sov .
jetp * 56 * ( 1982 ) 493 [ zh .
* 83 * ( 1982 ) 876 ] .
h. g. dosch , m. jamin and s. narison , phys . lett .
b * 220 * ( 1989 ) 251 .
a. e. dorokhov , s. v. esaibegian and s. v. mikhailov , phys . rev .
d * 56 * ( 1997 ) 4062 .
m .v .
polyakov and c. weiss , phys .
b * 387 * ( 1996 ) 841 .
t. doi , n. ishii , m. oka and h. suganuma , phys .
d * 67 * ( 2003 ) 054504 .
e. v. shuryak , nucl .
b * 203 * ( 1982 ) 93 .
d. diakonov and v. y. petrov , nucl .
b * 245 * ( 1984 ) 259 .
m. c. chu , j. m. grandy , s. huang and j. w. negele , phys .
d * 49 * ( 1994 ) 6039 .
j. w. negele , nucl .
. proc .
* 73 * ( 1999 ) 92 .
t. degrand , phys .
d * 64 * ( 2001 ) 094508 .
p. faccioli and t. a. degrand , phys .
* 91 * ( 2003 ) 182001 .
o. bowman , u. m. heller , d. b. leinweber , a. g. williams and j. b. zhang , nucl .
. proc . suppl .
* 128 * ( 2004 ) 23 .
d. diakonov and v. y. petrov , nucl .
b * 272 * ( 1986 ) 457 .
d. diakonov , m. v. polyakov and c. weiss , nucl . phys .
b * 461 * ( 1996 ) 539 .
d. diakonov , prog .
* 51 * ( 2003 ) 173 .
m. musakhanov , eur .
j. c * 9 * ( 1999 ) 235 .
m. musakhanov , hep - ph/0104163 .
m. musakhanov , nucl .
a * 699 * ( 2002 ) 340 .
m. a. nowak , j. j. m. verbaarschot and i. zahed , nucl .
b * 324 * ( 1989 ) 1 .
r. alkofer , m. a. nowak , j. j. m. verbaarschot and i. zahed , phys .
b * 233 * ( 1989 ) 205 .
m. kacir , m. prakash and i. zahed , acta phys .
b * 30 * ( 1999 ) 287 .
p. v. pobylitsa , phys .
b * 226 * ( 1989 ) 387 .
kim , m. m. musakhanov and m. siddikov , phys .
b * 633 * ( 2006 ) 201 .
t. schafer and e. v. shuryak , rev .
* 70 * ( 1998 ) 323 .
m. jamin , phys .
b * 538 * ( 2002 ) 71 .
a. di giacomo and y. a. simonov , phys .
b * 595 * ( 2004 ) 368 .
m. beneke and h. g. dosch , phys .
b * 284 * ( 1992 ) 116 .
k. aladashvili and m. margvelashvili , phys . lett .
b * 372 * ( 1996 ) 299 .
v. m. khatsimovsky , i. b. khriplovich and a. r. zhitnitsky , z. phys .
c * 36 * ( 1987 ) 455 .
a. a. ovchinnikov and a. a. pivovarov , sov .
j. nucl .
* 48 * ( 1988 )
721 [ yad . fiz .
* 48 * ( 1988 ) 1135 ] .
v. m. braun and a. lenz , phys .
d * 70 * ( 2004 ) 074020 .
s. i. nam , h .- ch .
kim , a. hosaka and m. m. musakhanov , phys .
d * 74 * ( 2006 ) 014019 . | we investigate the effects of flavor su(3)-symmetry breaking on the quark , gluon , and mixed quark - gluon condensates , based on the nonlocal effective chiral action from the instanton vacuum .
we take into account the effects of the flavor su(3 ) symmetry breaking in the effective chiral action , so that the dynamical quark mass depends on the current quark mass ( @xmath0 ) .
we compare the results of the present approach with those without the current quark mass dependence of the dynamical quark mass .
it is found that the result of the quark condensate is decreased by about 30 % as @xmath0 increases to 200 mev , while that of the quark - gluon mixed condensates is diminished by about @xmath1 .
we obtain the ratios of the quark and quark - gluon mixed condensates , respectively : @xmath2^{1/3}=0.75 $ ] and @xmath3^{1/5}=0.87 $ ] .
it turns out that the dimensional parameter @xmath4 . *
1 . *
understanding the qcd vacuum is very complicated , since both perturbative and non - perturbative fluctuations come into play . in particular ,
the quark and gluon condensates , being the lowest dimensional ones , characterize the non - perturbative structure of the qcd vacuum . the quark condensate is identified as the order parameter for spontaneous chiral - symmetry breaking ( s@xmath5sb ) which plays an essential role in describing low - energy phenomena of hadrons : in the qcd sum rule , these condensates come from the operator product expansion and are related to hadronic observables @xcite , while in chiral perturbation theory ( @xmath5pt ) , the free parameter @xmath6 is introduced in the mass term of the effective chiral lagrangian at the leading order @xcite measuring the strength of the quark condensate @xcite . on the other hand ,
the gluon condensate is not the order parameter but measures the vacuum energy density @xcite , which was first estimated by the charmonium sum rule @xcite .
while the quark and gluon condensates are well understood phenomenologically , higher dimensional condensates suffer from large uncertainty . though it is still possible to estimate dimension - six four - quark condensates in terms of the quark condensate by using the factorization scheme which is justified in the large @xmath7 limit , the dimension - five mixed quark - gluon condensate is not easily determined phenomenologically .
in particular , the mixed condensate is an essential parameter to calculate baryon masses @xcite , exotic hybrid mesons @xcite , higher - twist meson distribution amplitudes @xcite within the qcd sum rules .
moreover , the mixed quark condensate plays a role of an additional order parameter for s@xmath5sb , since the quark chirality flips via the quark - gluon operator .
thus , it is naturally expressed in terms of the quark condensate : @xmath8 with the dimensional parameter @xmath9 which was estimated in various works @xcite and with the definition @xmath10 .
the instanton picture allows us to study the qcd vacuum microscopically .
since the instanton picture provides a natural mechanism for @xmath11sb due to the delocalization of single - instanton quark zero modes in the instanton medium , the quark condensate can be evaluated .
the instanton vacuum is validated by the two parameters : the average instanton size @xmath12 and average inter - instanton distance @xmath13 .
these essential numbers were suggested by shuryak @xcite within the instanton liquid model and were derived from @xmath14 by diakonov and petrov @xcite .
these values were recently confirmed by lattice qcd simulations @xcite . in the present work ,
we want to investigate the effect of flavor su(3)-symmetry breaking on the above - mentioned three qcd condensates , _
i.e. _ quark , gluon , and mixed quark - gluon condensates , based on the instanton liquid model for the qcd vacuum @xcite .
the model was later extended by introducing the current quark masses @xcite . since we are interested in the effect of explicit flavor su(3)-symmetry breaking , we follow the formalism of ref
. @xcite in which the dependence of the dynamical quark mass on the current quark mass has been studied in detail . though the mixed quark - gluon condensate was already studied in the instanton vacuum @xcite , explicit su(3)-symmetry breaking was not considered .
hence , we want now to extend the work of ref .
@xcite , emphasizing the effect of flavor su(3)-symmetry breaking on the qcd vacuum condensates .
we will show that with a proper choice of the @xmath0 dependence of the dynamical quark mass @xcite the gluon condensate is independent of @xmath0 .
the corresponding results are summarized as follows : the ratio @xmath15^{1/3}= 0.75 $ ] , @xmath16^{1/5 } = 0.87 $ ] , @xmath17 , and @xmath18 , with isospin symmetry assumed .
* 2 . *
we start with the effective low - energy qcd partition function from the instanton vacuum with su(3)-symmetry breaking taken into account @xcite : @xmath19\left[\frac{y_{+}^{n_f}}{vm^{n_f}}\right]^{n_{+ } } \left[\frac{y_{-}^{n_f}}{vm^{n_f}}\right]^{n_{- } } , \label{partition } \end{aligned}\ ] ] where @xmath20 and @xmath21 denote the quark fields and @xmath0 stands for the current quark mass for a given flavor .
we consider here the strict large @xmath7 expansion expansion , is not small @xcite .
however , a proper adjustment of the parameters @xmath22 and @xmath23 may yield approximately the same results as those without the meson - loop corrections . ] .
@xmath24 and @xmath25 stand for the space - time volume and for the dimensional parameter , respectively .
note that , by dropping the current quark mass @xmath0 in the first bracket in the r.h.s . , eq .
( [ partition ] ) turns out to be the same as that derived in the chiral limit @xcite .
@xmath26 in eq .
( [ partition ] ) represents a t hooft - type @xmath27-quark interaction generated by instantons : @xmath28[2\pi\rho f_f(p_f)]\nonumber\\&\times&\exp\left [ -x\cdot\left(\sum_{f}k_f-\sum_{f}p_f\right)\right]\left [ u^{\alpha}_{i'}(u^{j'}_{\beta})^{\dagger}\epsilon^{ii'}\epsilon_{jj ' } \right]_f\left[i\psi_f^{\dagger}(k_f)_{\alpha i}\frac{1\pm\gamma_5}{2 } \psi_f(p_f)^{\beta j}\right ] , \label{vertex } \end{aligned}\ ] ] where @xmath29 denotes the instanton size and @xmath30 represents the color orientation matrix .
since we are interested in the vacuum condensates , it is enough to consider the case of @xmath31 in which the integration over @xmath30 becomes trivial .
we employ the instanton distribution of a delta - function type , so that we get a simple quark - quark interaction for a given flavor @xmath32 : @xmath33 ^ 2\left[\psi_f^{\dagger}(k)\frac{1\pm\gamma_5}{2 } \psi_f(k)\right ] .
\label{vertex1}\ ] ] @xmath34 is the fourier transformation of the fermionic zero mode solution @xmath35 .
we implicitly assumed that @xmath34 is a function of the current - quark mass .
this assumption will be verified in what follows .
@xmath35 satisfies the following dirac equation under the ( anti)instanton effects @xmath36 : @xmath37 \phi_{i\bar{i}}=0.\ ] ] instead of computing @xmath35 directly from eq .
( [ eq : zeromode ] ) , being equivalently , we follows the course suggested in ref .
@xcite in which pobylitsa used a elaborated and systematic expansion of the quark propagator @xmath38 . by doing this ,
one can immediately obtain analytical form of @xmath34 . here , we make a brief explanation on this method .
diakonov _
et al .
_ made a zero - mode approximation for the quark propagator in the instantons @xcite : @xmath39 we note that this zero - mode approximation causes one difficult problem of breaking the conservation of vector and axial - vector currents .
to amend this current conservation problem , pobylitsa expand the quark propagator as follows .
having summed up the planar diagrams , one arrives the following integral equation for a quark propagator : @xmath40^{-1 } + ( i\to\bar{i})\right ] , \label{eq : pobyl } \end{aligned}\ ] ] where @xmath41 indicates the collective coordinate for the ( anti)instanton .
then the inverse of the quark propagator can be approximately written with an arbitrary scalar function to be determined : @xmath42 ^ 2}{n_c}. \label{inpro2}\ ] ] note that we choose the scalar function to be @xmath43 for later convenience .
the value of @xmath44 will be determined by the self - consistent equation ( so called `` saddle - point equation '' ) of the model . to obtain @xmath34 , inserting eq.([inpro2 ] ) into eq.([eq : pobyl ] ) , we obtain an integral equation for @xmath45 : @xmath46 .
\label{inpro3}\ ] ] having carried out a straightforward manipulation , finally , we arrive at : @xmath47 = m_0f^2_f(k)f(m_f),\,\,\,\,d=\sqrt{\frac{0.08385}{2n_c } } \frac{8\pi\bar{\rho}}{r^2 } \label{fa } \nonumber\\ f_f(k ) & = & 2t\left[i_0(t)k_1(t)-i_1(t)k_0(t)-\frac{1}{t}i_1(t)k_1(t)\right ] \simeq0.198\,{\rm gev},\,\,\,\,t=|k|\bar{\rho}/2 .
\label{ff1}\end{aligned}\ ] ] thus , we have exact form of @xmath34 in terms of explicitly broken flavor su(3 ) symmetry ( @xmath48 ) .
now , we are in position to discuss the saddle - point equation of the model derived from the effective qcd partition function . introducing the lagrange multipliers @xmath49 @xcite , we are able to exponentiate the partition function of eq .
( [ partition ] ) : @xmath50\nonumber\\ & \times&\exp\left[n_+\left(\ln\frac{n_+}{\lambda_+vm^{n_f}}-1\right ) + \lambda_+y_{+}\right]\exp\left[n_-\left(\ln\frac{n_-}{\lambda_- vm^{n_f}}-1\right)+\lambda_-y_{-}\right ] .
\end{aligned}\ ] ] in the following , we assume that @xmath51 and @xmath52 so that the partition function can be simplified as follows : @xmath53.\nonumber\\ \label{partion3}\ ] ] it is also interesting to compare eq .
( [ partion3 ] ) with the partition function proposed by diakonov _
et al . _
: @xmath54 .
\label{partion4}\ ] ] one can easily find the difference between the effective actions of eq .
( [ partion3 ] ) and of eq .
( [ partion4 ] ) ; in eq .
( [ partion3 ] ) , the current quark mass @xmath0 appears as a mass term of the quark fields accounting for the explicit breaking of chiral symmetry , while the @xmath0 is placed in the quark - instanton vertex in eq .
( [ partion4 ] ) .
we note that this configuration of the current quark mass is deeply related to the parity - violating quark mass term ( @xmath55 ) in terms of the instanton number fluctuation @xcite . concentrating on the first and third brackets in eq .
( [ partion3 ] ) , we get the fermionic trace log which relates to the quark propagator as follows : @xmath56 ^ 2\right ] } { \left[-\rlap{/}{k}+im_f\right]}\nonumber\\&&\hspace{2cm}=n_cv\int \frac{d^4k}{(2\pi)^4}{\rm tr}_{\gamma}\ln\frac{\left[-\rlap{/}{k}+im_f + im_f(k)\right ] } { \left[-\rlap{/}{k}+im_f\right ] } , \end{aligned}\ ] ] where the subscripts @xmath57 and @xmath58 denote the color and dirac - spin spaces .
thus , we obtain the partition function in a compact form as follows : @xmath59 } { \left[-\rlap{/}{k}+im_f\right]}\right].\nonumber\\ \label{effectivez}\ ] ] now , we perform a functional variation of the partition function with respect to @xmath60 : @xmath61}{k^2+[m_f+m_f(k])^2 } , \nonumber\\\frac{n}{v}&=&4n_c\int\frac{d^4k } { ( 2\pi)^4}\frac{m_f(k)[m_f+m_f(k)]}{k^2+[m_f+m_f(k)]^2}. \label{speq } \end{aligned}\ ] ] final formula in eq .
( [ speq ] ) is called the saddle - point equation of the model .
we employ the standard values of the instanton ensemble , @xmath62 mev@xmath63 and @xmath64 fm @xmath65 mev@xmath66 for the numerical calculations . from these values , we obtain @xmath67 gev , which is determined self - consistently by the saddle - point equation . using this value , we obtain @xmath68 mev@xmath69 and pion weak decay constant @xmath70 mev .
we note that this saddle - point equation is closely related to the gluon - condensate @xcite ; @xmath71 . as for an arbitrary @xmath72
, it becomes rather difficult to obtain the saddle - point equation , since we need to perform a complicated integration over the color orientation .
however , keeping only the leading order in the large @xmath7 limit and additional variation over the hermitian @xmath73 flavor matrix ( @xmath74 in ref .
@xcite ) , we have the same form of the saddle - point equation for each flavor as shown in eq .
( [ speq ] ) .
the similar argument is also possible for the quark and mixed condensates @xcite .
being similar to the saddle - point equation ( the gluon condensate ) , the quark condensate for each flavor can be obtained by the following functional derivative with respect to the current quark mass @xmath0 : @xmath75}{k^2+[m_f+m_f(k)]^2}-\frac{\rlap{/}{k}+im_f } { k^2+m_f^2}\right]\nonumber\\&=&4n_c\int\frac{d^4k}{(2\pi)^4}\left [ \frac{m_f+m_f(k)}{k^2+[m_f+m_f(k)]^2}-\frac{m_f}{k^2+m_f^2}\right ] .
\label{qq } \end{aligned}\ ] ] one can immediately see that when @xmath76 , eq .
( [ qq ] ) gets equal to the well - known expression for the quark condensate in the chiral limit .
now , we discuss how to calculate the mixed condensate , @xmath77 .
actually , the local operator inside this condensate corresponds to the quark - gluon interaction of a yukawa type .
however , in the present work , the gluon field strength ( @xmath78 ) can be expressed in terms of the quark - instanton interaction @xcite .
first , the one flavor quark and one instanton interaction in eq .
( [ vertex ] ) can be rewritten as a function of space - time coordinates @xmath79 and color orientation matrix @xmath30 : @xmath80[2\pi\rho f_f(k_2)]e^{-ix\cdot(k_1-k_2)}\nonumber\\ & \times&\left[u^{\alpha}_{i'}(u^{j'}_{\beta})^{\dagger}\epsilon^{ii ' } \epsilon_{jj'}\right]_f\left[i\psi_f^{\dagger}(k_1)_{\alpha i } \frac{1\pm\gamma_5}{2}\psi_f(k_2)^{\beta j}\right ] .
\label{vertex2 } \end{aligned}\ ] ] here , we assume again the delta function - type instanton distribution .
we define then the field strength @xmath81 in terms of the instanton configuration as a function of @xmath82(@xmath83 ) position and orientation matrix @xmath30 : @xmath84g^b_{\pm\mu\nu}(x'-x ) .
\label{field}\ ] ] @xmath85 stands for the field strength consisting of a certain instanton configuration . using eqs .
( [ vertex2 ] ) and ( [ field ] ) , we define the field strength in terms of the quark - instanton interaction : @xmath86 following the method in ref .
@xcite , we finally obtain the quark - gluon mixed condensate as follows : @xmath87 ^ 2 ] [ k^2_2+[m_f+m_f(k_2)]^2]},\nonumber\\ \label{mc}\ ] ] where @xmath88 and @xmath89 are defined as follows : @xmath90\right ] , \nonumber\\ n(k_1,k_2)&=&\frac{1}{(k_1-k_2)^2}\left[8k^2_1k^2_2 - 6(k^2_1+k^2_2)k_1\cdot k_2 + 4(k_1\cdot k_2)^2\right ] \end{aligned}\ ] ] with @xmath91 . the functions @xmath92 and @xmath93 stand for the modified bessel functions of the second kind of order @xmath94 and @xmath95 , respectively .
if we consider for arbitrary @xmath72 the mixed condensate , the situation may be somewhat different from the cases of the gluon and quark condensates .
we note that , though the mixed condensate has been calculated for the case of @xmath31 , the same formula of eq .
( [ mc ] ) still holds for each flavor with arbitrary @xmath72 as discussed previously @xcite .
* 3 . *
we first discuss the numerical results for the gluon condensate , @xmath96 .
as mentioned before , we use @xmath97 mev@xmath63 @xcite . in fig .
[ fig2 ] , we show the numerical results of the gluon condensates with and without the @xmath0-correction factor @xmath98 in eq .
( [ fa ] ) . as shown in fig .
[ fig2 ] , the result with @xmath98 is noticeably different from that without it .
since the gluon condensate is related to the vacuum energy density , it should be independent of the current quark mass @xmath0 .
however , if we turn off the @xmath0-correction factor , the gluon condensate increases almost linearly .
it indicates that it is essential to have the @xmath0-correction factor @xmath98 in order to consider su(3 ) flavor symmetry breaking properly . in table
[ table1 ] we list the values of the gluon condensate when @xmath99 mev and @xmath100 mev with and without @xmath98 . [ cols="^,^ " , ] the value of @xmath9 increases as @xmath0 does , which implies that the mixed condensate is less sensitive to the @xmath0 than the quark condensate .
the values of @xmath9 are in the range of @xmath101 for the strange quark and of @xmath102 for the up quark .
thus , the strange @xmath103 turns out to be larger than the nonstrange @xmath104 by about @xmath105 .
we now consider the scale dependence of the condensates .
@xmath106^{\gamma_1/b}\langle\mathcal{o}_1(\mu_b ) \rangle,\,\,\,\langle\mathcal{o}_2(\mu_a)\rangle=\left [ \frac{\alpha_s(\mu_b)}{\alpha_s(\mu_a)}\right]^{\gamma_2/b } \langle\mathcal{o}_2(\mu_b)\rangle , \label{norm}\ ] ] where the subscripts @xmath95 and @xmath107 indicate different generic operators corresponding to the condensates in which we are interested .
the @xmath108 and @xmath109 denote two different renormalization scales of the operators .
@xmath58 is the corresponding anomalous dimension for the condensates .
@xmath110 is defined as @xmath111 and becomes @xmath112 in the present work .
taking the ratio of the operators @xmath95 and @xmath107 , we have : @xmath113^ { ( \gamma_1-\gamma_2)/b}\frac{\langle\mathcal{o}_1(\mu_b)\rangle } { \langle\mathcal{o}_2(\mu_b)\rangle}. \label{norm1}\ ] ] if @xmath114 and @xmath115 stand for the mixed and quark condensates respectively , we have @xmath116 and @xmath117 resulting in @xmath118 . on the contrary , we have @xmath119 and @xmath120 .
thus , @xmath121 shows relatively strong dependence on the scale , whereas the quark and mixed condensates are less influenced by the scaling .
thus , @xmath9 depends more strongly on the scale .
finally , we would like to mention that the relation between the mixed condensate and the expectation value of the transverse momentum for the leading - twist light - cone distribution amplitude .
this relation can be written as follows : @xmath122 considering @xmath121 for the light quarks in table [ table4 ] ( @xmath123 ) , we get @xmath124 . note that the value estimated here is in qualitatively good agreement with that calculated directly from the pseudoscalar meson light - cone distribution amplitude within the same framework @xcite ; @xmath125 . in the present work
, we investigated the various qcd vacuum condensates within the framework of the instanton liquid model , emphasizing the effects of flavor su(3)-symmetry breaking .
the modified improved action elaborated by musakhanov was used for this purpose . in the modified improved action ,
the current quark mass appeared explicitly in the denominator of the quark propagator as well as in the dynamical quark mass .
thus , we were able to take into account the current quark mass effects to the qcd condensates . in order to consider the @xmath126 dependence of the dynamical quark mass
, we employed the @xmath0-correction factor .
it arises from the resummation of the qcd planar loops in the large @xmath7 limit and is parameterized to satisfy the saddle - point equation @xcite .
we observed that the gluon condensate is almost independent of the current quark mass when the strange current quark correction factor is introduced , whereas it increases monotonically without it .
the quark and mixed condensates were also calculated .
the results were consistent with those from other model calculations as well as phenomenological values . in particular , the ratios of the condensates between the strange and up quarks were also investigated : @xmath127^{1/5}$ ] and @xmath128^{1/3}$ ] .
it turned out that the results are again compatible with other theoretical calculations .
the dimensional quantity @xmath129 was also studied : @xmath130 and @xmath131 . in general , the quark and mixed condensates decrease as the current quark mass increases .
however , the @xmath9 increases as the current quark mass does , which indicates that the mixed condensate is less sensitive to the current quark mass than the quark condensate .
in addition , we also presented the relation of the @xmath9 to the expectation value of the transverse momentum for the leading - twist pion light - cone distribution amplitude .
though we considered the effects of flavor su(3)-symmetry breaking on various condensates , we did not take into account those from meson - loop corrections which are known to be of importance .
the corresponding investigation is under way . |
the results presented below are based on the publicly available data of cyg x1 , gx3394 and gs1354644 observations with the proportional counter array aboard the rossi x - ray timing explorer ( @xcite ) from 19961998 during the low ( hard ) spectral state of the sources . in total
our sample included @xmath4 observations of these sources .
the energy and power density spectra were averaged for each individual observation .
the energy spectra were extracted from the `` standard mode 2 '' data and arf and rmf were constructed using standard rxte ftools v.4.2 tasks .
the `` vle '' and `` q6 '' models were used for the background calculation with the preference being given to the `` vle '' model when possible . a uniform systematic error of 0.5% was added quadratically to the statistical error in each energy channel .
the power spectra were calculated in the 216 kev energy band and in the @xmath5 hz frequency range following the standard x ray timing technique and normalized to units of squared fractional rms .
the energy spectra were fit with a model consisting of a power law with a superimposed reflected continuum ( pexrav model in xspec , @xcite ) and an intrinsically narrow line at @xmath6 kev .
the centroid energy of the line was a free parameter of the fit .
for all sources the binary system inclination was fixed at @xmath7 degrees , the iron abundance at the solar value of @xmath8 and the low energy absorption at @xmath9 @xmath10 .
the effects of ionization were not included . in order to approximately include in the model the smearing of the reflection features due to motion in the accretion flow the reflection continuum and line were convolved with a gaussian ,
which width was a free parameter of the fit .
the spectra were fitted in the 420 kev energy range .
this spectral model is identical to that applied by @xcite to the cyg x-1 data , except that in the present analysis we let the line centroid energy be a free parameter of the fit .
strength of the reflected component is characterized in the pexrav model by the reflection scaling factor @xmath11 , which approximately measures the solid angle subtended by the reflecting media as seen from the source of the primary radiation , @xmath12 , so that @xmath13 for an isotropic point source above an infinite optically thick slab .
the spectral model is obviously oversimplified . moreover ,
several important parameters , such as the binary system inclination angle and the iron abundance were fixed at fiducial values .
therefore the best fit values do not necessarily represent the exact values of the physically interesting parameters .
particularly subject to the uncertainties due to the choice of the spectral model is the reflection scaling factor @xmath14 .
this might explain the values of @xmath11 exceeding unity obtained for some of the spectra .
however , as was shown by @xcite the model correctly ranks the spectra according to the strength of the reflected component and the slope of the underlying power law .
cm 9.0 cm the power density spectra in units of @xmath15 were fit in the 0.00515 hz frequency range with a model , consisting of two broken constant components and a lorentzian profile @xmath16 .
we use centroid frequency of the lorentzian profile as a characteristic noise frequency .
its dependence on the break frequency of the band limited noise ( the first broken constant component in the model ) is shown in fig.[pds ] .
the results of the energy and power spectra approximation are presented in fig .
[ refl_slope ] and [ refl_qpo ] showing dependence of the slope of the primary emission and characteristic noise frequency on the amplitude of the reflected component . the noise frequency and the spectral parameters
are clearly correlated .
an increase of the amplitude of the reflected component is accompanied with a steepening of the spectrum of the primary emission and an increase of the noise frequency .
the photon index @xmath17 of the comptonized radiation is governed by the compton amplification factor @xmath18 which equals to the ratio of the energy deposited into the hot electrons to the influx of the soft seed photons to the comptonization region .
the concrete shape of the @xmath19 relation depends on the ratio @xmath20 of the temperatures of the seed photons and the electrons , the thomson optical depth and the geometry . in the simplest
although not unique scenario the correlation between the amplitude of the reflected component @xmath11 and the slope of the primary radiation @xmath17 ( fig.[refl_slope ] ) could be understood assuming that there is a positive correlation between the solid angle subtended by the reflecting media , @xmath21 , and the influx of the soft photons to the comptonization region ( @xcite ) .
existence of such correlation is a strong argument in favor the reflecting media being the primary source of the soft seed photons to the comptonization region . in the absence of strong beaming effects
a correlation between @xmath21 and the seed photons flux should be expected since an increase of the solid angle of the disk seen by the hot electrons ( @xmath22 ) should generally lead to the increase of the fraction of the disk emission reaching the comptonization region .
cm in order to demonstrate that such geometrical effects can explain the observed dependence @xmath23 we consider two idealized models having different cause of change of the @xmath21 . in the first , disk - spheroid model , an optically thin uniform hot sphere with radius @xmath24
is surrounded with an optically thick cold disk with an inner radius @xmath25 ( e.g. @xcite ) , the @xmath21 depending on the ratio @xmath26 .
propagation of the disk towards / inwards the hot sphere ( decrease of @xmath26 ) leads to an increase of the reflection scaling factor , a decrease of the the compton amplification factor @xmath18 and a steepening of the comptonized spectrum . in such context
the model was first studied by @xcite and we used their results to calculate the relation between the reflection scaling factor @xmath11 and the compton amplification @xmath18 . in the second , plasma ejection model , value of the @xmath21 is defined by the intrinsic properties of the emitting hot plasma , particularly by its bulk motion with mildly relativistic velocity towards or away from the disk , which itself remains unchanged ( @xcite ) . in the case of an infinite disk , values of reflection below and above unity
correspond to the hot plasma moving respectively away and towards the disk .
both models predict relation between reflection @xmath11 and compton amplification factor @xmath18 which can be translated to @xmath23 given a dependence @xmath19 of the photon index of comptonized spectrum on the amplification factor .
the latter was approximated by : @xmath27 this formula is based on a representation of the comptonized spectrum by a power law in the energy range @xmath28 and agrees with the results of the monte - carlo calculations with reasonable accuracy for optical depth @xmath29 and @xmath30 ( fig.[enh_slope ] ) . \9 .
cm the expected @xmath23 relations are shown in fig.[refl_slope ] . with a proper tuning of the parameters both models can reproduce the observed shape of the @xmath23 dependence and in this respect are virtually indistinguishable .
the observed range of the reflection @xmath31 and the slope @xmath32 can be explained assuming variation of the disk radius in from @xmath33 to @xmath34 in the disk - spheroid model or variation of the bulk motion velocity from @xmath35 away from the disk to @xmath36 in the plasma ejection model .
finally , it should be emphasized that these idealized models do not include a number of important effects which might affect the particular shape of the @xmath23 dependence .
the spectrum of emission , reflected from a keplerian accretion disk is expected to be modified by special and general relativity effects ( e.g. fabian et al . , 1989 ) .
if increase of the reflection is caused by a shift of the inner boundary of the accretion disk closer to the compact object , a correlation between the reflection and relativistic smearing should be expected .
such a correlation is a generic prediction of the model and might be used to discriminate between different assumptions about geometry of the accretion flow .
the energy resolution of rxte / pca is not adequate to study relativistic smearing of the reflection features with sufficient degree of confidence .
however , the data shown in fig.[refl_gsmo ] might present an evidence in favor of correlated behavior of the reflection and relativistic smearing .
if confirmed , it would prove that the spectral and timing properties in the low state of the black hole binaries are defined mainly by the position of the inner boundary of the optically thick accretion disk .
in the low spectral state of black hole candidates the amplitude of reflection @xmath11 , slope of the comptonized radiation @xmath17 and characteristic noise frequency change in a correlated way . the correlation between the slope and reflection indicates that the accretion disk is the primary source of the soft seed photons for comptonization .
we considered two idealized models , the disk spheroid and plasma ejection model , and showed that with appropriate tuning of the parameters either model can explain the observed @xmath23 dependence .
the rxte data also hints at a correlated change of the width of relativistic smearing of the reflection features .
if confirmed by observations with better energy resolution , this correlation may help to distinguish between different geometries of the accretion flow .
this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center .
the work was done in the context of the research network `` accretion onto black holes , compact objects and protostars '' ( tmr grant erb - fmrx - ct98 - 0195 of the european commission ) .
m.revnivtsev acknowledges hospitality of the max
planck institute for astrophysics and a partial support by rbrf grant 961596343 and intas grant 933364ext . | based on the analysis of a large sample of rxte / pca observations of several black hole binaries in the low spectral state we show that a correlation exists between the spectral parameters and characteristic noise frequency . in particular , the amplitude of reflection increases and the slope of comptonized radiation steepens as the noise frequency increases .
we consider possible implications of these results on the accretion flow models and discuss a possible observational test aimed to discriminate between different geometries of the accretion flow .
accretion , accretion disks black hole physics stars : binaries : general stars : individual : cygnus x-1 , gx339 - 4 , gs1354 - 644 x - rays : general x - rays : stars comptonization of soft seed photons in a hot , optically thin electron cloud near the compact object is thought to produce the hard x ray radiation in the low spectral state of accreting black holes ( @xcite , @xcite ) .
the slope of the comptonized spectrum is governed by the ratio of the energy deposited into the electrons and the influx of the soft radiation into the comptonization region ; the lower the ratio the steeper the comptonized spectrum ( e.g. @xcite , @xcite , @xcite ) .
reflection of the comptonized radiation from neutral or partially ionized matter , presumably the optically thick accretion disk , leads to appearance of characteristic features in the spectra of x ray binaries ( fig.[plrat ] ) .
the main signatures of the emission reflected from cold neutral medium are well known
the fluorescent k@xmath0 line of iron at 6.4 kev , iron k - edge at 7.1 kev and a broad hump above @xmath1 kev ( basko , sunyaev & titarchuk 1974 , george & fabian 1991 ) .
the exact shape of these spectral features in the x - ray binaries depends on the ionization state of the reflecting medium ( @xcite ) and might be modified by the strong gravity effects and intrinsic motions in the reflector ( e.g. fabian et al . , 1989 ) .
the amplitude of the reflection signatures depends primarily on the ionization state and on the solid angle of the reflector as seen from the source of the primary radiation .
based on the analysis of a large sample of seyfert agns and several x
ray binaries @xcite ( see also @xcite ) found a correlation between the amplitude of reflection and the slope of the underlying power law .
they concluded that the existence of such a correlation implies a dominant role of the reflecting medium as the source of seed photons to the comptonization region .
8 cm the power density spectra of the black hole binaries in the low state ( see @xcite for a review ) plotted in the units of frequency@xmath2power usually appear as a superposition of two or more rather broad humps containing most of the power of aperiodic variations below several tens hz ( fig.[pds ] ) .
these humps define several characteristic noise frequencies of which the lowest usually corresponds to the break frequency @xmath3 of the band limited noise component . although a number of theoretical models was proposed to explain the power spectra of x ray binaries ( e.g. @xcite , @xcite ) , the nature of the characteristic noise frequencies is still unclear . despite the fact that the characteristic noise frequencies vary from source to source and from epoch to epoch
, they are correlated with the break frequency of the band limited noise ( fig.[pds ] , @xcite ) .
@xcite showed that a tight correlation exists between the characteristic noise frequencies and the spectral parameters .
based on rxte / pca observations of cyg x1 from 19961998 they found that the amplitude of reflection increases and the spectrum of primary radiation steepens as the noise frequency increases . an identical behavior was found for another black hole candidate , gx339 - 4 , by @xcite . interpreting these correlations
@xcite suggested that increase of the noise frequency might be caused by the shift of the inner boundary of the optically thick accretion disk towards the compact object .
the related increase of the solid angle , subtended by the disk , and of the influx of the soft photons to the comptonization region lead to an increase of the amount of reflection and steepening of the comptonized spectrum .
below we investigate further the relation between the spectral parameters and characteristic noise frequencies based on a larger sample of objects and observations and discuss implications of these results on the geometry of the accretion flow in the low spectral state of black hole binaries . |
much evidence indicates that massive black holes ( bhs ) reside in the centers of many or most galaxies ( e.g. @xcite ) .
mergers of galaxies are likely to form massive binary black holes ( bbhs ) .
theoretical estimation shows that the bbh lifetime is not much shorter than the hubble time and many bbhs should be still in the centers of galaxies ( @xcite ) .
the existence of bbhs in the universe will not only provide a laboratory to test gravitation radiation theory and bh physics , but also probe of the hierarchical structure model of galaxy and large - scale structure formation . currently , there is no systematic and unambiguous method to identify bbhs .
bbhs stay at a separation in the range @xmath0 cm ( e.g. @xmath1 arcsec at 10mpc ) during the slowest evolution period ( @xcite ) , and thus it is hard to resolve a bbh two very close galactic nuclei in the image with current telescope resolution .
the shallow cusps in the inner surface brightness profiles of some nearby giant galaxies may be produced from steep cusps by ejecting stars from their inner regions during the hardening of bbhs ( @xcite ) , but there is still no proof of a currently existing bbh in those galaxies . to identify a bbh
, we have to find some other effect of bbhs on their nearby environment and/or some manifestation of the motion in a two - body system , such as jet precession ( @xcite ) , double - peaked balmer lines ( @xcite ) , or quasi - periodic radio , optical , x - ray or @xmath2-ray variation ( e.g. oj287 : sillanp et al .
1988 , valtaoja et al .
2000 ; mkn501 : rieger & mannheim 2000 ) . some active galactic nuclei ( agns )
have been claimed to be detected as bbh candidates by those methods , but all of them are controversial because of other explanations for the same phenomena or some inconsistency with other observational evidence .
if there is sufficient gas with some angular momentum close to a bbh , we may expect that the gas in the vicinity of each bh is accreting onto the bh in the form of a disc rather than in the form of spherical accretion , which may make the system appear as an agn .
if the binary separation is small ( say , much less than the scale of the broad line region ) , the two accretion discs may be warped at outer parts and connected with an outer large circumbinary accretion disc .
if the separation is large enough , each bh is probably accompanied by its own disc and broad line region .
the spin axes of the two bhs are very likely to be misaligned ( @xcite ) and the discs in this two - accretion - disc ( tad ) system associated with the bbh can also have different inclination angles to the line of sight .
agns are observed to be copious x - ray emitters .
this x - ray emission is believed to originate from the very inner accretion disc around a massive bh .
the broad skewed iron k@xmath3 line profile found by _ ginga _ and confirmed by _ asca _ is believed to result from a combination of gravitational broadening and doppler shift in an accretion disc ( @xcite ) .
so far , alternative models have failed to account for this profile ( @xcite ) , which offers one of the strongest lines of evidence for the existence of massive bhs .
x - ray spectroscopy also promises a powerful tool to detect strong - gravitation - field relativistic effects in the vicinity of a massive bh .
the observed line profile is significantly affected by the inclination angle of the disc to the observers line of sight ( @xcite ) . in a bbh system
, the observed line profile can be a combination of the line profiles from two discs with very different inclination angles to the line of sight .
motivated by this observation , we propose a method to probe bbhs in agns by line profiles .
we expect that line profiles will become an efficient way to probe bbhs in agns .
consider a bbh containing bhs of mass @xmath4 and @xmath5 rotating around their center of mass ( fig . 1 )
, the relative orbit of the two bhs is assumed to be circular with separation @xmath6 .
the keplerian orbital period of the binary is @xmath7 and their maximum orbital velocities relative to the center of mass in the line of sight are @xmath8\sin \theta_{\rm orb } \ \ \ \ ( i=1,2 ) \label{eq : vel}\ ] ] where @xmath9 is the orbital inclination to the line of sight .
a moderate orbital eccentricity of the bbh orbit will not change the above values much .
the orbital motion of each bh is ignored in the later calculation of line profiles since it only causes a slight overall doppler shift in the profiles . for a single rapidly rotating bh with mass @xmath10 , the bardeen - petterson effect ( @xcite ) can cause the gradual transition of the disc into the bh equatorial plane in the region between radii @xmath11 and @xmath12 ( note that there are still controversies on the location of the transition radii , c.f .
nelson & papaloizou 2000 and references therein ) , where @xmath13 .
thus a planar disc could be formed in the equatorial plane of the bh .
the timescale for the bh to align with the outer disc angular momentum can not be estimated precisely , but is of the order of the overall lifetime of agns ( rees 1978 , nelson & papaloizou 2000 ; however , a much smaller realignment timescale of the order of @xmath14 yr is gotten by natarajan & pringle 1998 ) . in a bbh system
, it is likely that both bhs have large spins and the spin axes can be in very different directions , especially in the early phase of the nuclear activity .
this difference may not be changed in a time much shorter than the lifetime of the nuclear activity .
therefore , the two accretion discs associated with a bbh can have different inclination angles to the line of sight . even if only one of the bhs has a large spin , it is still very likely that its spin axis is different from the disc direction of the other ( schwarzschild ) bh . the roche radius of the smaller bh is about @xmath15 if @xmath16 .
if the two bhs have comparable masses and a typical separation of @xmath17 , the inner region of each disc , at least within @xmath18 , is not easily disrupted by tidal forces from the other bh . in a bbh system
, the spin axis of each bh will undergo geodesic precession about its total angular momentum as discussed by begelman , blandford & rees ( 1980 ) .
when the spin angular momentum of the binary is small compared with its orbital angular momentum ( @xmath19 if @xmath16 ) , the spins will precess in a cone with the orbital angular momentum as axis .
their precession periods are given by : @xmath20 the precession may cause the two accretion discs to be tilted to their bhs equatorial plane , but the two disc inclination angles to our line of sight will not change in a short time
. detailed study of the dynamics and stability of the tads in bbhs is beyond the scope of this paper .
we have argued that there are good reasons for the existence of tads with different inclination angles to the line of sight in bbhs .
we shall assume that both discs are cold thin accretion discs .
the observed line profile is then the summation of the two components from the tads .
the combined line profile is mainly controlled by the inclination angles of the tads and the relative strength of the two components .
the relative strength is related to the mass ratio of the two bhs and the accretion rates onto them ( if the two accretion systems have almost the same dimensionless accretion rate @xmath21 , where @xmath22 is the accretion rate and @xmath23 is eddington accretion rate , the relative strength of the two components will be given by the mass ratio @xmath24 ) . using the ray - tracing technique and elliptic integrals ( @xcite )
, we follow the photons from each accretion disc to a distant observer , and calculate the corresponding redshift of the photons and the resulting line profiles .
the spins of bhs are both set to be @xmath25 .
the line photons are assumed to be isotropically emitted in the frame moving with the accretion disc material , and the surface emissivity is given by the power - law @xmath26 for the region @xmath27 .
we adopt the averaged line emissivity exponent index of @xmath28 which is obtained from the fitting of line profiles for a sample of agns observed by _ asca _
( nandra et al .
we set the inner radius to be the marginally stable orbit @xmath29 ( few line photons come from the region inside the marginally stable orbit for kerr bh with spin 0.998 ) , and the outer radius @xmath30 ( the line profiles are not sensitive to the value of @xmath31 since most line photons come from the inner region for a typical line emissivity law with @xmath28 ) .
an example of the resulting spectral line is shown in figure 1 , for discs with inclination angles @xmath32 and @xmath33 , and @xmath34 .
this unusual line profile is double - peaked , asymmetric and has two ` edge'-like feature .
the peak with the smaller central energy , near the rest frame energy of line 6.4kev , ( or less than 6.4kev due to gravitational redshift for extremely low inclination disc ) , comes from disc d@xmath35 with a small inclination angle , has the characteristic features of emission from a relativistic accretion disc : a skewed red wing and a sharp `` blue '' edge ( note that the energy of the `` blue '' edge is mainly determined by the inclination of a disc , and the red wing extent is sensitive to the inner radius of the line emission region ) .
the peak with the smaller central energy is narrower than the broad component , which comes from disc d@xmath36 with a high inclination angle .
if the spin of the bh @xmath4 is smaller , the `` narrow component '' can be even narrower .
the `` narrow '' component can be stronger or weaker than the broad component depending on the relative emissivity .
another important feature is that these two components should both have _ short - term variability _ of intensity and shapes on the timescale of @xmath37s , as suggested by the variation in some seyfert galaxies ( @xcite ) , and their variation patterns can be totally different and unrelated with each other since they come from two different discs .
all of the above features make it easy to distinguish the bbh model from the other alternative models producing a two - component line shape .
not all of the emergent line profiles from tad systems are so distinguishable from the profiles produced by only one disc .
if the relative emissivity of the two discs is too small or too large , the component emitted from one accretion disc in tads system will be drowned by the other . only when @xmath38 is about in the range @xmath39 ( or @xmath40 is in a similar range if luminosity is proportional to eddington luminosity ) will the line profiles from tads be significantly different from the one emitted by a single bh accretion disc system .
if @xmath38 is a little less than @xmath41 , the line profile may be misunderstood as a relativistic line from a single disc plus an absorption feature ; and if @xmath38 is a little larger than @xmath42 , the line profile may be misunderstood as a relativistic line from a single disc plus a high ionization line ( e.g. fe k@xmath43 or ni k@xmath3 line ) . if the difference between @xmath44 and @xmath45 is too small , the two line components will be blended , and thus difficult to distinguish from the profile produced by a single bh accretion disc system .
the line profiles in tad systems can be very complicated with double peaks , three peaks and even four peaks depending on the inclination angles of the two accretion discs and the relative emissivity ( fig .
when @xmath46 and @xmath47 , the emergent line profile will clearly exhibit two distinct components ( see fig .
2 ; note that the line from a disc with very high inclination angle ( say , @xmath48 ) may be strongly affected by the limb - darkening effect of the outer layer of accretion disc ) .
if the inner disc is somewhat ionized , or the emissivity law is somewhat different , the line profile should remain similar .
the probability that the combined line from two randomly oriented equal discs is in the shape similar to those shown in figure 2 , is about 20% ( the difference between the inclination angles is larger than @xmath49 and both the disc inclination angles are not larger than @xmath50 ) .
the amounts of agn sources expected to harbor bbh systems with comparable bh masses , which could be identified by line profiles , are relevant to the process of structure formation and the merger history of galaxies .
the line shapes from the tad systems shown in figure 1 and figure 2 are examples chosen from many idealizations .
we have neglected such complication as the actual dynamics of the system , the possible warp of the outer disc , the ionization of the accretion discs , the real line emissivity law , contamination from non - disc emitters , the irradiation of one disc by the x - ray photons from the other , and the absorption of intervening gas and dust etc .
the complication certainly affects the line profile quantitatively , but will not make much difference qualitatively .
there are other possible models that can produce two - component line profiles , but they are not difficult to differentiate from the bbh model . first , off - axis x - ray flares above a single disc can strongly affect the line profiles ( off - axis - flare model ; c.f .
yu & lu 2000 ) . in 3 ,
the disc emissivity is assumed to be axisymmetric .
in reality , x - ray flares can be off the disc rotation axis and the emissivity law of the disc is non - axisymmetric because the disc region just under the flares generally receive more illumination .
thus , the non - axisymmetric emissivity may strongly affect the line profiles since different energy parts of line profiles stem from different regions of a disc .
for example , a cold accretion disc is illuminated by two local flares : one is atop the approaching side of the disc , where the energy of line photons are blue shifted by doppler effect ; and the other is atop the disc region towards or backwards us which corresponds to the line emission around 6.4kev .
this situation can make line shapes like the one from tads .
the x - ray flares are probably produced by some thermal instability in the disc or magnetic reconnection and their locations should be randomly distributed .
the appearance of the line profile arising from flares can be very complex .
however , the average line profile over several flares should be consistent with the profile from a single disc ( with a single inclination angle ) , as suggested by the observations of mcg-6 - 30 - 15 ( @xcite ) , which is quite different from the combined line profiles from tads .
second , a two - component line shape can be produced if a narrow line component is emitted from a molecular torus or broad line region ( blr ) clouds , and a broadened line component is emitted from the inner region of a highly inclined accretion disc ( blr / torus - ad model , c.f .
@xcite ) . in this scenario ,
the central energy of the narrow component should be around 6.4kev or a little blue - shifted by the outflow motion , and the line width caused by the velocity dispersion of the clouds should be several thousand .
in contrast , in the bbh model , the `` narrow '' component is redshifted to energies less than 6.4kev if the inclination is very low ( due to gravitational redshift ) . with higher inclination ,
the narrow component is centered around 6.4kev or higher , but with a large width .
more importantly , in the blr / torus - ad model , the narrow component should remain constant on the timescale of several days or more ( the light - travel time across the blr or the torus ) rather than varying on timescales @xmath51s , as predicted by the bbh model .
third , a two - component line shape can also be produced by the combination of a broadened iron line from a highly inclined accretion disc plus a component which is scattered into our line of sight by an efficient electron - scattering material ( scatter - ad model , c.f .
wang et al .
it is unlikely that this electron - scattering material is close to the disc because this configuration is unlikely to produce a line with two distinct components : first , the scattered line component shape will be different from the one observed from a disk with a small inclination angle ; second , photons from the inner accretion disc can not be seen directly because the disc is covered by the scattering material .
if the scattering layer is high enough above the disc , the scattered line profile would be somewhat like the one observed at a small inclination angle ; however , the covering factor would normally be small and the equivalent width of this component should therefore be much smaller than the broad one since the scattered photons are redistributed in all directions .
in contrast , in the bbh model , the equivalent width of the `` narrow '' component can be larger than that of the broad component .
furthermore , the temporal variations of the two components in the scatter - ad model must be strongly correlated with a time delay which reflects the distance between the scattering material and the disc .
to use profiles to identify a bbh with tads , we need to resolve both the narrow component and broad component of the line profiles and study the variation of the line profiles and intensity with time . over the past decade ,
a number of seyfert galaxies and qsos have been shown to have broad lines ( @xcite ) .
the line profile of ngc4151 is fitted by two disc components with inclinations of @xmath52 and @xmath53 better than by a one - disc line model ( @xcite ) . the component with inclination @xmath52
was explained by being scattered into our line of sight ; however , those two components can also be explained as coming from possible tads .
future observations with higher resolution are needed to check if ngc4151 is a possible bbh candidate .
the unique profile of mark205 ( see fig . 2 in reeves et al .
2000 ) , recently revealed by _ xmm - newton _ , is somewhat like the line shape established in figure 1 . its broad line component can be fitted by an accretion - disc line with a high inclination .
unfortunately , the narrow component is not fully resolved .
the present observations can be explained by assuming that the narrow component comes from neutral matter at large distant from a central bh and the broad one is emitted from a highly ionized relativistic accretion disc ( @xcite ) .
so , further observations and variability studies are needed to check if this object is a bbh candidate or not .
stimulated by those special line shapes observed in some objects , we believe that some bbhs in agns , if any , can be identified by searching the unusual iron line profile with current and future x - ray satellites , such as _ xmm - newton , constellation - x and xeus_. if any one of agns is revealed to have the typical line shapes shown in figure 2 as well as short - term variability , it will provide one of the strongest lines of evidence for the existence of bbhs and a laboratory to investigate the dynamics in strong gravitation field ( e.g. the bardeen - petterson effect ) .
bardeen j. m. , petterson j. a. , 1975 , apjl , 195 , 65 begelman m. c. , blandford r. d. , rees m. j. , 1980 , nature , 287 , 307 fabian a. c. , rees m. j. , stella l. , white n. e. , 1989 , mnras , 238 , 729 fabian a. c. , nandra k. , reynolds c. s. , brandt w. n. , otani c. , tanaka y. , inoue h. , iwasawa k. , 1995 , mnras , 277 , l11 fabian a. c. , iwasawa k. , reynolds c. s. , young a. j. , 2000 , pasp , 112 , 1145 faber s. m. et al .
, 1997 , aj , 114 , 1771 gaskell c. m. , 1996 , apjl , 464 , 107 ivanov p. b. , papaloizou j. c. b. , polnarev a. g. , 1999 , mnras , 307 , 79 iwasawa k. et al . , 1996 , mnras , 282 , 1038 iwasawa k. , fabian a. c. , young a. j. , inoue h. , matsumoto c. , 1999 , mnras , 306 , l19 laor a. , 1991 , apj , 376 , 90 magorrian j. et al . , 1998 , aj , 115 , 2285 nandra k. , george i. m. , mushotzky r. f. , turner t. j. , yaqoob t. , 1997 , apj , 477 , 602 nandra k. , george i. m. , mushotzky r. f. , turner t. j. , yaqoob t. , 1999 , apjl , 523 , 17 natarajan p. , pringle j. e. , 1998 , apj , 506 , l97 nelson r. p. , papaloizou j. c. b. , 2000 , mnras , 315 , 570 quinlan g. d. , hernquist l. 1997 , new astron .
, 2 , 533 rauch k. p. , blandford r. d. , 1994 , apj , 421 , 46 rees , m. j. 1978 , nature , 275 , 516 reeves j. n. , turner m. j. l. , pounds k. a. , obrien p. t. , boller th . , ferrando p. , kendziorra e. , vercellone s. , 2000 , a&a , 365 , l134 rieger f. m. , mannheim k. , 2000 , a&a , 359 , 948 sillanp a , haarala s. , valtonen m. j. , sundelius b. , byrd g. g. , 1988 , apj , 325 , 628 sulentic j. w. , marziani p. , dultzin - hacyan d. , 2000 , ara&a , 38 , 521 tanaka y. et al .
, 1995 , nature , 375 , 659 valtaoja e. , tersranta h. , tornikoski m. , sillanp a. , aller m. f. , aller h. d. , hughes p. a. , 2000 , apj , 531 , 744 wang j. , zhou y. , wang t. , 1999 , apjl , 523 , 129 yaqoob t. , geroge i. m. , nandra k. , turner t. j. , serlemitsos p. , mushotzky r. f. , 2001 , apj , 546 , 759 yaqoob t. , serlemitsos p. , 2000 ,
apjl , 544 , l95 yu q. , lu y. , 2000 , mnras , 311 , 161 yu q. , 2002 , chapter 1 in ph.d . thesis ( princeton university ) , in preparation | hierarchical mergers of galaxies can form binary black holes ( bbhs ) since many or most galaxies have central massive black holes ( bhs ) .
it is possible that some bbhs exist in active galactic nuclei ( agns ) .
we argue that each bh may be surrounded by an accretion disc with a different inclination angle to the line of sight ( due to different bh spin directions and the bardeen - petterson effect ) .
the observed line profile from a bbh system is a combination of the lines from the inner regions of the two discs , which is significantly affected by the inclination angles of the two discs .
the line profile associated with bbhs may have an unusual shape with double or more peaks as well as short - term variability , which can be distinguished from the line properties of some other possible models .
we suggest that with the improvement of resolution in x - ray astronomy , line profiles be a potential tool to probe the existence of massive bbhs in agns . the line profile associated with bbhs
may also provide a tool to investigate the dynamics in strong gravitation field ( e.g. providing evidence of the bardeen - petterson effect ) . |
a popular and exactly soluble model in quantum optics is the jaynes - cumming model ( jcm ) .
it describes the interaction of a two - level atom with a single - mode of the electromagnetic field @xcite . over the last two decades
various extensions of the ordinary jcm have been used in various directions , e.g. , as adapted to : ( i ) the study of interaction of a three - level atom with a two - mode squeezed vacuum @xcite ; ( ii ) the study of atom - field interaction in the presence of a cavity damping @xcite ; ( iii ) the same as in ( i ) , including an additional ( nonlinear ) kerr medium @xcite ; ( iv ) the two - level atoms inside a cavity acted upon by an external field control @xcite ; ( v ) study of the nonlinear dynamical evolution of a driven two - photon jaynes - cummings model @xcite ; ( vi ) the study of a generalized jaynes - cummings models , including dissipation 15,16,16a and multiphoton interactions @xcite ; etc . in all these cases , with interest either on the field or on atomic properties , the theoretical approach traditionally assumes the atom - field coupling as a constant parameter .
comparatively , the number of works in the literature is very small when one considers such coupling and the atomic frequency as time dependent parameters @xcite , including time dependent amplitudes @xcite .
however , this scenario is also relevant ; for example , the state of two qubits ( qubits stand for quantum bits ) with a desired degree of entanglement can be generated via a time dependent atom - field coupling @xcite .
actually , such coupling can modify the dynamical properties of the atom and the field , with transitions that involve a large number of photons @xcite .
in general , these studies are simplified by neglecting the atomic decay from an excited level .
theoretical treatments taking into account this complication of the real world may employ a modified jcm . in these case , as expected , the state describing the system decoheres , since the presence of dissipation destroys the state of a system as time flows .
in the present work we extend what we have learned from the jcm applied to the atom - field interaction to investigate a more advantageous system , from the experimental viewpoint ( faster response , better controllability , and useful scalability for quantum computation @xcite ) by considering a nanomechanical resonator ( nr ) interacting with a cooper pair box ( cpb ) 17,16,18 .
such nanodevice has been explored extensively in the literature , e.g. , to investigate : ( i ) quantum nondemolition measurements @xcite , ( ii ) decoherence of nonclassical states , as fock states and superposition states describing mesoscopic systems @xcite , etc .
the fast advance in the technique of fabrication in nanotechnology has implied great interest in the study of the nr system in view of its potential applications , as a sensor - to be used in biology , astronomy , quantum computation , and in quantum information @xcite , to implement the quantum qubit ak3 and in the production of nonclassical states , as fock state akk , bb4 , schrdinger s cat state @xcite , squeezed states a34 , clusters states @xcite , etc . in particular , when accompanied by superconducting charge qubits , the nr has been used to prepare entangled states @xcite .
zhou et al.@xcite have proposed a scheme to prepare squeezed states using a nr coupled to a cpb qubit ; in this proposal the nr - cpb coupling is under an external control while the connection between these two subsystems plays an important role in quantum computation .
such a control is achieved via convenient change of the system parameters , which can set on " and off " the interaction between the nr and the cpb , on demand . in this report
we will investigate the cpb excitation inversion , its control , and the average photon number in the nr .
we will consider dissipation in the cpb due to a decay rate from excited to ground states .
we will also verify in which way the time dependence of the cpb - nr coupling modifies these two properties .
to this end we must solve the time evolution of the whole cpb - nr system , via the approach presented in the following section .
a josephson charge qubit system has been used to couple with a nr . here
we study a modified model where a cpb is coupled to a nr , as shown in fig .
cooper below .
the scheme is inspired by the works of jie - qiao liao et al .
@xcite and zhou et al .
@xcite where we have substituted each josephson junction by two of them .
this creates a new configuration that includes a third loop .
a superconducting cpb charge qubit is adjusted via a voltage @xmath0 at the system input and a capacitance @xmath1 .
we want that the scheme attains an efficient tunneling effect for the josephson energy . in fig .
[ cooper ] we observe three loops : one great loop between two small ones .
this makes it easier controlling the external parameters of the system since the control mechanism includes the input voltage @xmath0 plus three external fluxes @xmath2 @xmath3 and @xmath4 . in this way
one can induce small neighboring loops .
the great loop contains a nr which is modeled as a harmonic oscillator with a high - q mode of frequency @xmath5 and its effective area in the center of the apparatus changes as the nr oscillates , which creates an external flux @xmath4 that provides the cpb - nr coupling . in pursuing the quantum behavior of a macro scale object the nano scale mechanical resonator plays an important role . at sufficiently low temperature
the zero - point fluctuation of the nr will be comparable to its thermal brownian motion .
the detection of zero - point fluctuations of the nr can give a direct test of heisenberg s uncertainty principle . with a sensitivity up to 10 times the amplitude of the zero - point fluctuation ,
lahaye et al @xcite have experimentally detected the vibrations of a 20 mhz mechanical beam of tens of micrometres size . for a 20mhz mechanical resonator
its temperature must be cooled below 1mk to suppress the thermal fluctuation . for a ghz mechanical resonator
a temperature of 50mk is sufficient to effectively freeze out its thermal fluctuation and let it enter the quantum regime .
this temperature is already attainable in dilution refrigerators . in this work
we will assume the four josephson junctions being identical , with the same josephson energy @xmath6 , the same being assumed for the external fluxes @xmath7 and @xmath3 , i.e. , with same magnitude but opposite sign : @xmath8 .
this interaction actually couples the two subsystems .
together with the free hamiltonian of flux qubit and nr , the hamiltonian of the whole system reads @xmath9 where @xmath10 is the creation ( annihilation ) operator for the nr excitation , with frequency @xmath5 and mass @xmath11 ; @xmath6 and @xmath12 are respectively the energy of each josephson junction and the charge energy of a single electron ; @xmath1 and @xmath13 are the input capacitance and the capacitance of each josephson tunnel , respectively .
@xmath14 is the quantum flux and @xmath15 is the charge number in the input with the input voltage @xmath0 .
we have used the pauli matrices to describe our system operators , where the states @xmath16 and @xmath17 represent the number of extra cooper pairs in the superconducting island .
we have : @xmath18 , @xmath19 and @xmath20 the magnetic flux can be written as the sum of two terms , @xmath21where the first term @xmath22 is the induced flux , corresponding to the equilibrium position of the nr and the second term describes the contribution due to the vibration of the nr ; @xmath23 represents the magnetic field created in the loop .
we have assumed the displacement @xmath24 described as @xmath25 , where @xmath26 is the amplitude of the oscillation . substituting the eq.(a4 ) in eq.([a1 ] ) and controlling the flux @xmath22 we can adjust @xmath27 , and making the approximation @xmath28 the above hamiltonian results as ( in rotating wave approximation ) , @xmath29where the constant coupling @xmath30 and the effective energy @xmath31 an important advantage of this coupling mechanism is its easy and convenient controllability .
next , we will extend the previous approach to a more general scenario by substituting @xmath32 and @xmath33 $ ] @xcite ; in addition we assume the presence of a constant decay rate @xmath34 in the cpb ; @xmath35 is the transition frequency of the cpb and @xmath36 stands for the cpb - nr coupling . @xmath37 and @xmath38 are the cpb transition and excitation inversion operators , respectively ; they act on the hilbert space of atomic states and satisfy the commutation relations
@xmath39 = \hat{\sigma}_{z}$ ] and @xmath40 = \pm \hat{\sigma}_{\pm } $ ] .
as well known , the coupling parameter @xmath41 is proportional to @xmath42 , where the time dependent quantization volume @xmath43 takes the form @xmath44 $ ] @xcite .
accordingly , we obtain the new ( _ non hermitian _ ) hamiltonian @xmath45 it is worth remembering that non hermitian hamiltonians ( nhh ) are largely used in the literature . to give some few examples we mention : ref .
@xcite , where the authors use a nhh and an algorithm to generalize the conventional theory ; ref .
@xcite , using a nhh to get information about entrance and exit channels ; ref .
@xcite , using non hermitian techniques to study canonical transformations in quantum mechanics ; ref .
@xcite , solving quantum master equations in terms of nhh ; ref .
@xcite , using a new approach for nhh to study the spectral density of weak h - bonds involving damping ; ref .
@xcite , studying nhh with real eigenvalues ; ref .
@xcite , using a canonical formulation to study dissipative mechanics exhibing complex eigenvalues ; ref .
@xcite , studying nhh in non commutative space , and more recently : ref .
@xcite , studying the optical realization of relativistic nhh ; ref .
@xcite , studying the evolution of entropy of atom - field interaction ; ref .
@xcite , using a damping jc - model to study entanglement between two atoms , each of them lying inside different cavities
the state that describes this time dependent system can be written in the form @xmath46 where @xmath47 @xmath48 represents the cpb in its excited state @xmath49 ( ground state @xmath50 ) .
taking the cpb initially prepared in its excited state @xmath17 and the nr in a coherent states @xmath51 , and expanding coherent state component in the fock s basis , i.e. , @xmath52 , we have @xmath53 . assuming the nr and cpb decoupled at @xmath54 and the initial conditions @xmath55 and @xmath56 we may write the eq .
( [ b2 ] ) as @xmath57 the time dependent schrdinger equation for this system is @xmath58 with the hamiltonian @xmath59 given in eq .
( [ b1 ] ) .
_ _ _ _ substituting eq.([b1 ] ) in eq.([b5 ] ) we get the ( coupled ) equations of motion for the probabilitity amplitudes @xmath60 and @xmath61:@xmath62the numerical solutions of the coefficients @xmath60 , @xmath61 furnish the quantum dynamical properties of the system , including the cpb - nr entanglement .
as well known , in the presence of decay rate @xmath34 in the cpb the state of the whole cpb - nr system becomes mixed . in this case
its description requires the use of the density operator @xmath63 , which describes the entire system . to obtain the reduced density matrix describing the cpb ( nr ) subsystem we must trace over variables of the nr ( cpb ) subsystem . for example , @xmath64 : @xmath65 \left\vert n\right\rangle \left\langle n\prime \right\vert .
\label{c2}\ ] ]
the cpb excitation inversion , here denoted as @xmath66 , is an important observable of two level systems .
it is defined as the difference of probabilities of finding the system in the excited and ground state , as follows@xmath67 .
\label{joao}\ ] ] the eq .
( [ joao ] ) allows one to look at the time evolution of the cpb excitation inversion .
first , we assume the resonant case @xmath68 for different values of the decay rate @xmath34 , with @xmath69 and @xmath70 and assuming the nr initially in a coherent state with the average number of excitations @xmath71 as shown in fig .
[ inversao1 ] . with the exceptions of amplitudes , the plots ( a ) , ( b ) and ( c ) of fig .
[ inversao1 ] show identical for collapse - revival effects : the higher the decay rate the lower is the amplitude of oscillation . in presence of detuning , where @xmath72 , where @xmath73 , @xmath5 , we see that the excitation inversion in fig . [
inversao2](a ) occurs within the interval @xmath74 , whereas in fig .
[ inversao2](b ) it occurs in the range @xmath75 and in fig .
[ inversao2](c ) the excitation inversion goes to zero rapidly . considering the case of variable detuning , @xmath76
, we see in the plots ( a ) , ( b ) and ( c ) of fig .
inversao3 that the excitation inversion occurs frequently in fig .
inversao3(a ) and desapears when the parameter @xmath77 increases , as shown in the plots ( b ) and ( c ) . now ,
even considering the worst results obtained in the off - resonant cases , with detuning @xmath78 , as shown in fig .
[ inversao2](c ) and fig .
[ inversao3](c ) we see that we can recover the collapse - revival effects via the increase of the parameter @xmath79 as shown in the plots ( a ) , ( b ) and ( c ) of fig .
so , the parameter @xmath80 plays a fundamental role in the control of collapse and revival effect .
[ nmf1 ] shows plots of the mean value of the nr excitations in the presence of cpb decay rate for various values of amplitude of oscillations ( parameter @xmath77 )
. plots ( * d ) * and ( * h ) * are for the resonant case : in ( * d ) * the decay rate is greater than in ( * h)*. plots ( * a ) * , ( * b ) * , and ( * c ) * are for constant decay rates , with ( time independent ) detuning that increases from ( * c ) * @xmath81 * * ( b ) * * @xmath81 * * ( a)**. finally , plots ( * e ) * , ( * f ) * and ( * g ) * are for time dependent detuning , with the parameter @xmath77 increasing from ( * e ) * @xmath81 * * ( f ) * * @xmath81 * * ( g)**. we note that the three plots for time dependent detunings ( * e ) * , ( * f ) * and _ _ ( _ _ * g * ) are better than those for constant detunings ( * a ) * , ( * b ) * and ( * c * ) : despite all plots are concerned with the same decay rate , the first group is more robust against decay . for example , comparing the plots ( * a * ) and ( * g ) * : although in * ( a ) * the fixed detuning is @xmath82 and in ( * g ) * maximum detuning is @xmath83 we see that in the last case the average value of the nr decays more slowly .
one observes in the plot ( * c ) * of fig .
[ inversao4 ] that the interval @xmath84 , where the presence of the time dependent detuning recovers the collapse - revival effect , coincides with that in the plot ( * g ) * of fig .
[ nmf1 ] where the mean value of excitation is around 5 times greater than in the case of constant detuning ( plot ( * a ) * of same figure ) . , @xmath85 , for @xmath86 ( resonance ) and different values of decay rates @xmath87 : ( a ) @xmath88 , ( b ) @xmath89 and ( c ) @xmath90.__,width=340,height=302 ] for @xmath89 and different values of detunings @xmath91cf .
@xmath92 : ( a ) @xmath93 , ( b ) @xmath94 , ( c ) @xmath95.__,width=340,height=302 ] for @xmath89 and different time - dependent detunings _ _ ( cf .
@xmath96 ) : ( a ) @xmath97 and @xmath98 , ( b ) @xmath99 and @xmath100 ( c ) @xmath101 and @xmath102__,width=340,height=302 ] for @xmath89 and different time - dependent detunings _ _ ( cf .
@xmath103 ) : ( a ) @xmath101 and @xmath104 , ( b ) @xmath101 and @xmath105 ( c ) @xmath101 and @xmath106.__,width=340,height=302 ] , @xmath85 , for the following cases : ( a ) off - resonance _
@xmath89 , @xmath107 _ _ ( b ) off - resonance _ _
@xmath89 , @xmath108 _ _ ( c ) off - resonance _ _
@xmath89 , @xmath109 _ _ ( d ) on resonance _ _ @xmath110 _ _ ( e ) off - resonance _ _ @xmath111 @xmath97 and @xmath100 _ _ ( f ) off - resonance _ _ @xmath111 @xmath99 and @xmath100__(g ) off - resonance _ _ @xmath111 @xmath101 and @xmath100 _ _ ( h ) on resonance _ _ @xmath89 .
, width=340,height=302 ]
we have considered a hamiltonian model that describes a cpb - nr interacting system to study the cpb excitation inversion , @xmath112 and the average excitation number of the nr , @xmath113 .
we have also considered the off - resonant case , with various values of the detuning parameter @xmath114(@xmath115 and @xmath116 ) and in the presence of cpb decay ( about 10 times greater than the ( neglected ) nr decay ) .
these properties are characteristics of the entangled state that describes this coupled system for various values of the parameters involved .
we have assumed the cpb initially in its excited state and the nr initially in a coherent state ( see preparation in @xcite ) .
so , the following three scenarios were treated : ( _ i _ ) both subsystems in resonance ( detuning @xmath117 ) ; ( _ ii _ ) off - resonance , with a constant detuning @xmath118 , and ( _ iii _ ) with a time dependent detuning @xmath119 . the results were discussed in the previous section : in resume , concerning the cpb excitation inversion , an interesting result emerged : although the presence of a constant detuning destroys the collapse and revivals of the excitation inversion , these effects are restituted by the action of convenient time dependent detunings - even in the presence of damping in the cpb ; concerning the nr average excitation number , another interesting result appeared : convenient choices of the time dependent detuning @xmath120 makes the nr subsystem more robust against the decay affecting the cpb subsystem . for constant values of detuning ,
our numerical results are similar to others in the literature using a master equation ( see , e.g. , ref .
@xcite ) .
finally we emphasize that the change in magnetic flux @xmath121 ( cf .
[ cooper ] ) , due to the presence of an external force upon the nr , is the responsible for controlling the parameters @xmath122 and @xmath41 .
the authors thank the fapeg ( cv ) and cnpq ( ata , bb ) , brasilian agencies , for partially supporting this paper .
99 jaynes e t and cummings f w 1963 proc . | we investigate the action of time dependent detunings upon the excitation inversion of a cooper pair box interacting with a nanomechanical resonator .
the method employs the jaynes - cummings model with damping , assuming different decay rates of the cooper pair box and various fixed and t - dependent detunings .
it is shown that while the presence of damping plus constant detunings destroy the collapse / revival effects , convenient choices of time dependent detunings allow one to reconstruct such events in a perfect way .
it is also shown that the mean excitation of the nanomechanical resonator is more robust against damping of the cooper pair box for convenient values of t - dependent detunings . |
in the introduction to pauli s collected scientific papers , the editors , ralph kronig and victor weisskopf , make the following statement : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is always hard to look for a leading principle in the work of a great man , in particular if his work covers all fundamental problems of physics .
pauli s work has one common denominator : his striving for symmetry and invariance .
[ ... ] the tendency towards invariant formulations of physical laws , initiated by einstein , has become the style of theoretical physics in our days , upheld and developed by pauli during all his life by example , stimulation , and criticism . for pauli , the invariants in physics where the symbols of ultimate truth which must be attained by penetrating through the accidental details of things .
the search for symmetry and general validity transcend the limits of physics in pauli s work ; it penetrated his thinking and striving throughout all phases of his life , in all fields of philosophy and psychology._(@xcite , vol.1 , p.viii ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ indeed , if i were asked to list those of pauli s scientific contributions which make essential use of symmetry concepts and applied group theory , i would certainly include the following , which form a substantial part of pauli s scientific vre : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ relativity theory and weyl s extension thereof ( 1918 - 1921 ) , the hydrogen atom in matrix mechanics ( 1925 ) , exclusion principle ( 1925 ) , anomalous zeeman effect and electron spin ( 1925 ) , non - relativistic wave - equation for spinning electron ( 1927 ) , covariant qed ( 1928 , jordan ) , neutrino hypothesis ( 1930 ) , kaluza - klein theory and its projective formulation ( 1933 ) , theory of @xmath0-matrices ( 1935 ) , poincar - invariant wave equations ( 1939 , fierz ) , general particle statistics and lorentz invariance ( 1940 , belinfante ) , spin - statistics ( 1940 ) , once more general relativity and kaluza - klein theory ( 1943 , einstein ) , meson - nucleon interaction and differential geometry ( 1953 ) , cpt theorem ( 1955 ) , @xmath1-decay and conservation of lepton charge ( ` pauli group ' , 1957 ) , unifying non - linear spinor equation ( collaboration with heisenberg , 1957 - 58 ) , group structure of elementary particles ( 1958 , touschek ) .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ amongst the theoretical physicists of his generation , pauli was certainly outstanding in his clear grasp of mathematical notions and methods .
he had a particularly sober judgement of their powers as well as their limitations in applications to physics and other sciences .
let us once more cite kronig and weisskopf : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ pauli s works are distinguished by their mathematical rigour and by a thorough and honest appraisal of the validity of assumptions and conclusions .
he was a true disciple of sommerfeld in his clear mathematical craftsmanship .
by example and sharp criticism he constantly tried to maintain a similarly high standard in the work of other theoretical physicists . he was often called the living conscience of theoretical physicists . _
( @xcite , vol.1 , p.viii ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it seems plausible that this critical impregnation dates back to his school - days , when young pauli read , for example , ernst mach s critical analysis of the historical development of the science of mechanics , a copy of which pauli received as a present from his godfather ( mach ) at around the age of fourteen .
mach s `` mechanik '' , as this book is commonly called , starts out with a discussion of archimedes law of the lever , thereby criticising the following symmetry consideration ( @xcite , p.11 - 12 ) : imagine two equal masses , @xmath2 , and a perfectly stiff and homogeneous rod of length @xmath3 , both being immersed into a static homogeneous vertical gravitational field , where the rod is suspended at its midpoint , @xmath4 , from a point @xmath5 above ; see fig.[fig : lawoflever ] .
what happens if we attach the two equal masses to the ends of the rod and release them simultaneously without initial velocity ?
an immediate symmetry argument suggests that it stays horizontal ; it might be given as follows : everything just depends on the initial geometry and distribution of masses , which is preserved by a reflection at the plane perpendicular to the rod through @xmath5 and @xmath4 .
suppose that after release the rod dropped at one side of the suspension point @xmath4 , then the mirror image of that process would have the same initial condition with the rod dropping to the other side .
this is a contradiction if the laws governing the process are assumed to be reflection symmetric and deterministic ( unique outcome for given initial condition ) .
this argument seems rigorous and correct .
now , how does one get from here to the law of the lever ?
the argument criticised by mach is as follows : assume that the condition for equilibrium depends only on the amount of mass and its suspension point on the rod , but not on its shape .
then we may replace the mass to the left of @xmath4 by two masses of half the amount each on a small rod in equilibrium , as shown in the second ( upper right ) picture .
then replace the suspension of the small rod by two strings attached to the left arm of the original rod , as shown in the third ( lower left ) picture , and observe that the right one is just under the suspension point @xmath4 , so that it does not disturb the equilibrium if it were cut away as in the last ( lower right ) picture of fig.[fig : lawoflever ] .
( -42.5,7)@xmath2 ( -112.5,7)@xmath2 ( -77,27)@xmath4 ( -75,68)@xmath5 @xmath6 + @xmath7 the weak point in the argument is clearly the transition from the second to the third picture : there is no _ global _ symmetry connecting them , even though locally , i.e. regarding the small rod only , it connects two equilibrium positions .
it is easy to see that , in fact , the assumption that a global equilibrium is maintained in this change is _ equivalent _ to archimedes law of the lever .
this example shows ( in admittedly a fairly trivial fashion ) that alleged symmetry properties can work as a _ petitio principii _ for the law to be derived .
this is essentially the criticism of mach .
the reason why we consider this ` derivation ' of the law of the lever to be a _
petitio principii _ is that we have other , physically much more direct ways to actually derive it from _ dynamical _ first principles . from that point of view
the alleged symmetry is to be regarded as an artifact of the particular law and certainly not vice versa .
the observed symmetry requires an explanation in terms of the dynamical laws , which themselves are to be established in an independent fashion .
this is how we look upon , say , the symmetry of crystals or the symmetric shape of planetary orbits . on the other hand ,
all fundamental dynamical theories of 20th century physics are motivated by symmetry requirements .
they are commonly looked at as particularly simple realisations of the symmetries in question , given certain _ a priori _ assumptions .
it is clear that , compared to the previous example , there are different concepts of symmetry invoked here .
however , there also seems to be a shift in attitude towards a more abstract understanding of ` physical laws ' in general .
what makes pauli an interesting figure in this context is that this shift in attitude can be traced in his own writings .
consider special relativity as an example , thereby neglecting gravity .
one may ask : what is the general relation between the particular symmetry ( encoded by the poincar group ) of spacetime and that very same symmetry of the fundamental interactions ( weak , strong , and electromagnetic , but not gravity ) ?
is one to be considered as logically prior to the other ?
for example , if we take einstein s original operationalist attitude , we would say that the geometry of spacetime is defined through the behaviour of ` rods ' and ` clocks ' , which eventually should be thought of as physical systems obeying the fundamental dynamical laws .
in fact , einstein often complained about the fact that rods and clocks are introduced as if they were logically independent of the dynamical laws , e.g. , in a discussion remark at the 86th meeting of the gesellschaft deutscher naturforscher und rzte in bad nauheim in 1920 : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is a logical shortcoming of the theory of relativity in its present form to be forced to introduce measuring rods and clocks separately instead of being able to construct them as solutions to differential equations . _
( @xcite , vol.7 , doc.46 , p.353 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ from that viewpoint , symmetry properties of spacetime are nothing but an effective codification of the symmetries of the fundamental laws
. consequences like ` length contraction ' and ` time dilation ' in special relativity are then only _ effectively _ described as due to the geometry of spacetime , whereas a fundamental explanation clearly has to refer to the dynamical laws that govern clocks and rods .
this was clearly the attitude taken by h.a.lorentz and h.poincar , though in their case still somehow afflicted with the idea of a material ther that , in principle , defines a preferred rest frame , so that the apparent validity of the principle of relativity must be interpreted as due to a ` dynamical conspiracy . ' in his famous article on relativity for the encyclopedia of mathematical sciences , the young pauli proposes to maintain this view , albeit without the idea on a material ther .
he writes : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ should one , then , in view of the above remarks , completely abandon any attempt to explain the lorentz contraction atomistically ?
we think that the answer to this question should be no .
the contraction of a measuring rod is not an elementary but a very complicated process .
it would not take place except for the covariance with respect to the lorentz group of the basic equations of electron theory , as well as those laws , as yet unknown to us , which determine the cohesion of the electron itself .
we can only postulate that this is so , knowing that then the theory will be capable of explaining atomistically the behaviour of moving measuring rods and clocks . _
( @xcite , p.15 . ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ very recently , this traditional view has once more been defended under the name of ` physical relativity ' @xcite against todays more popular view , according to which special relativity is about the symmetry properties of spacetime itself .
clearly , the latter view only makes sense if spacetime is endowed with its own ontological status , independently of the presence of rods and clocks . this shift in emphasis towards a more abstract point of view
is also reflected in pauli s writings , for example in the preface to the english edition of his ` theory of relativity ' of 1956 , where the abstract group - theoretic properties of dynamical laws are given an autonomous status in the explanation of phenomena : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the concept of the state of motion of the ` luminiferous ther ' , as the hypothetical medium was called earlier , had to be given up , not only because it turned out to be unobservable , but because it became superfluous as an element of a mathematical formalism , the group - theoretical properties of which would only be disturbed by it . by the widening of the transformation group in general relativity the idea of a distinguished inertial coordinate system
could also be eliminated by einstein , being inconsistent with the group - theoretical properties of the theory .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ pushed to an extreme , this attitude results in the belief that the most fundamental laws of physics are nothing but realisations of basic symmetries .
usually this is further qualified by adding that these realisations are the most ` simple ' ones , at least with respect to some intuitive measure of simplicity .
such statements are well known from einstein s later scientific period and also from heisenberg in connection with his ` unified theory ' of elementary particles , for which he proposed a single non - linear differential equation , whose structure was almost entirely motivated by its symmetry properties .
heisenberg made this point quite explicitly in his talk entitled _ _
planck s discovery and the foundational issues of atomism _ _ , delivered during the celebrations of max planck s 100th anniversary at which occasion wolfgang pauli received the max - planck medal in absentia , where he also talked about his own ` unified theory ' : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the mentioned equation contains , next to the three natural units [ @xmath8 , merely mathematical symmetry requirements .
these requirements seem to determine everything else .
in fact , one should just regard this equation as a particularly simple representation of the symmetry requirements , which form the actual core of the theory .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ pauli , who briefly collaborated with heisenberg on this project , did not at all share heisenberg s optimism that a consistent quantum - field theory could be based on heisenberg s non - linear field equation .
his objections concerned several serious technical aspects , overlayed with an increasing overall dislike of heisenberg s readiness to make premature claims , particularly when made publicly .
however , i think it is fair to say that the overall attitude regarding the heuristic rle and power of symmetry principles in fundamental physics , expressed by heisenberg in the above quote , was also to a large extent shared by pauli , not only in his later scientific life .
this is particularly true for symmetry induced conservation laws , towards which pauli had very strong feelings indeed .
examples from his later years will be discussed in later sections ( e.g. sect.[sec : irritations ] ) .
an example from his early scientific life is his strong resistance against giving up energy - momentum conservation for individual elementary processes , while keeping it on the statistical average .
such ideas were advocated in the `` new radiation theory '' of bohr , kramers , and slater of early 1924 @xcite and again by bohr in connection with @xmath1-decay , which pauli called _ spiritual somersaults _ in a letter to max delbrck . a week after his famous letter suggesting the existence of the neutrino , pauli wrote to oskar klein in a letter dated dec .
12th 1930:-zerfall wissen ) , wenn ersterer versagt .
zweitens mte bei einer verletzung des energiesatzes auch mit dem _ gewicht _
etwas sehr merkwrdiges passieren . [ ... ] dies widerstrebt meinem physikalischen gefhl auf das uerste !
denn es mu dann sogar auch fr das gravitationsfeld , das von dem ganzen kasten ( samt seinem radioaktiven inhalt ) selber _ erzeugt _ wird ( ... ) , angenommen werden , da es sich ndern kann , whrend wegen der erhaltung der ladung das nach auen erzeugte
elektrostatische feld ( beide felder scheinen mir doch analog zu sein ; das wirst du ja brigens auch aus deiner fnfdimensionalen vergangenheit noch wissen ) unverndert bleiben soll . ''
( @xcite , vol.ii , doc.[261 ] , p.45 - 46 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ first it seems to me , that the conservation law for energy - momentum is largely analogous to that for electric charge , and i can not see a theoretical reason why the latter should still be valid ( as we know empirically from @xmath1-decay ) if the former fails .
secondly , something strange should happen to the _ weight _ if energy conservation fails .
[ ... ] this contradicts my physical intuition to an extreme ! for
then one has to even assume that the gravitational field _ produced _
[ ... ] by the box ( including the radioactive content ) can change , whereas the electrostatic field must remain unchanged due to charge conservation ( both fields seem to me analogous ; as you will remember from your five - dimensional past ) . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this is a truly remarkable statement .
not many physicists would nowadays dare suggesting such an intimate connection between the conservation laws of charge and energy - momentum .
what pauli hints at with his last remarks in brackets is the kaluza - klein picture , in which electric charge is interpreted as momentum in an additional space dimension in a five - dimensional spacetime .
it is not difficult to find explicit commitments from pauli s later scientific life expressing his belief in the heuristic power of symmetry considerations .
let me just select two of them .
the first is from his introduction to the international congress of philosophers , held in zrich in 1954 , where pauli states : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` it seems likely to me , that the reach of the mathematical group concept in physics is not yet fully exploited . ''
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the second is from his closing remarks as the president of the conference `` 50 years of relativity '' held in berne in 1955 , where with respect to the still unsolved problem of whether and how the gravitational field should be described in the framework of quantum - field - theory he remarks : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it seems to me , that the heart of the matter [ the problem of quantising the gravitational field ] is not so much the linearity or non - linearity , but rather the fact that there is present a more general group than the lorentz group . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this , in fact , implicitly relates to much of the present - day research that is concerned with that difficult problem . before we can discuss specific aspects of ` symmetry ' in pauli s work in section[sec : specificcomments ] , we wish to recall various aspects of symmetry principles as used in physics .
the term ` symmetry ' is used in such a variety of meanings , even in physics , that it seems appropriate to recall some of its its main aspects .
one aspect is that which mathematicians call an ` automorphism ' and which basically means a ` structure preserving self - map ' .
take as an example ( conceptually not an easy one ) the modern notion of spacetime .
first of all it is a set , @xmath2 , the members of which are events , or better , ` potential events ' , since we do not want to assume that every spacetime point to be an actual physical event in the sense that a material happening is taking place , or at least not one which is dynamically relevant to the problem at hand .
that set is endowed with certain structures which are usually motivated through operational relations of actual physical events .
one such structure could be that of a preferred set of paths , which represent inertial ( i.e. force free ) motions of ` test bodies ' , that is , localised objects which do not react back onto spacetime structure .
this defines a so - called ` path - structure ' ( compare @xcite@xcite ) , which in the simplest case reduces to an _ affine structure _ in which the preferred paths behave , intuitively speaking , like ` straight lines ' .
this can clearly be said in a much more precise form ( see , e.g. , @xcite ) . under very mild technical assumptions ( not even involving continuity ) one
may then show that the only automorphisms of that ` inertial structure ' can already be narrowed down to the inhomogeneous galilei or lorentz groups , possibly supplemented by constant scale transformations ( cf .
@xcite@xcite ) .
another structure to start with could have been that of a causal relation on @xmath2 .
that is , a partial order relation which determines the pairs of points on spacetime which , in principle , could influence each other in form of a propagation process based on ordinary matter or light signals .
the automorphism group of that structure is then the subgroup of bijections on @xmath2 that , together with their inverse , preserve this order relation .
for example , in case of minkowski space , where the causal relation is determined by the light - cone structure , it may be shown that the most general automorphism is given by a poincar transformation plus a constant rescaling@xcite@xcite . since , according to klein s _ erlanger programm _
@xcite , any geometry may be characterised by its automorphism group , the geometry of minkowski space is , up to constant rescalings , entirely encoded in the causal relations .
the same result can be arrived at through topological considerations .
observers ( idealised to be extensionless ) move in spacetime on timelike curves .
take the set @xmath9 of all ( not necessarily smooth ) timelike curves which are continuous in the standard ( euclidean ) topology @xmath10 of minkowski spacetime @xmath2 .
now endow @xmath2 with a new topology , @xmath11 , called the _ path topology _ , which is the finest topology on @xmath2 which induces the same topology on each path in @xmath9 as the standard ( euclidean ) topology @xmath10 .
the new topology @xmath11 is strictly finer than @xmath10 and has the following remarkable property : the automorphism group of @xmath12 .
] , i.e. the group of bijections of @xmath2 which , together with their inverses , preserve @xmath11 , is just the poincar group extended by the constant rescalings @xcite .
this is possibly the closest operational meaning one could attribute to the topology of spacetime , since in @xmath11 a set in spacetime is open if and only if every observer `` times '' it to be open .
all this is meant to illustrate that there are apparently different ways to endow spacetime with structures that are , physically speaking , more or less well motivated and which lead to the same automorphism group .
that group may then be called the group of _ spacetime symmetries_. so far , this group seems to bear no direct relation to any dynamical law .
however , the physical meaning of such statements of symmetry is tight to an ontological status of spacetime points .
we assumed from the onset that spacetime is a _ set _ @xmath2 .
now , recall that georg cantor , in his first article on transfinite set - theory @xcite , started out with the following definition of a set : von bestimmten wohlunterschiedenen objecten @xmath4 unserer anschauung oder unseres denkens ( welche die ` elemente ' von @xmath2 genannt werden ) zu einem ganzen . ''
( @xcite , p.481 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ by a ` set ' we understand any gathering - together @xmath2 of determined well - distinguished objects @xmath4 of our intuition or of our thinking ( which are called the ` elements ' of @xmath2 ) into a whole . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
hence we may ask : is a point in spacetime , a ` potential event ' as we called it earlier , a `` determined well - distinguished object of our intuition or of our thinking '' ?
this question is justified even though modern axiomatic set theory is more restrictive in what may be called a set ( for otherwise it runs into the infamous antinomies ) and also stands back from any characterisation of elements in order to not confuse the axioms themselves with their possible _
interpretations_. however , applications to physics require interpreted axioms , where it remains true that elements of sets are thought of as definite as in cantors original definition .
but it is just this definiteness that seems to be physically unwarranted in application to spacetime .
the modern general - relativistic viewpoint takes that into account by a quotient construction , admitting only those statements as physically meaningful that are invariant under the group of ( differentiable ) permutations of spacetime points .
this is possible only because all other structures on spacetime , in particular the metric and with it the causal structure , are not fixed once and for all but are subsumed into the dynamical fields .
hence no non - dynamical background structures remain , except those that are inherent in the definition of a differentiable manifold .
the group of automorphisms is therefore the whole diffeomorphism group of spacetime , which , in some sense , comes sufficiently close to the group of all permutations .- point transitive , that is , given any two @xmath13-tuples of mutually distinct spacetime points , @xmath14 and @xmath15 , there is a diffeomorphism @xmath16 such that for all @xmath17 we have @xmath18 ; this is true for all positive integers @xmath13 . ]
what is the relation between spacetime automorphisms and symmetries of dynamical laws ? before we can answer this , we have to recall what a symmetry of a dynamical law is . for definiteness ,
let us restrict attention to dynamical laws in classical ( i.e. non - quantum ) physics .
the equations of motion generally take the form of systems of differential equations , which we here abbreviate with @xmath19 ( @xmath20quation of @xmath21otion ) .
these equations involve two types of quantities : 1 ) background structures , collectively abbreviated here by @xmath22 , and 2 ) dynamical entities , collectively abbreviated here by @xmath23 .
the former will typically be represented by geometric objects on @xmath2 ( tensor fields , connections , etc ) , which are taken from a somehow specified set @xmath24 of ` admissible backgrounds ' .
typical background structures are external sources , like currents , and the geometry of spacetime in non - general - relativistic field theories .
dynamical entities typically involve ` particles ' and ` fields ' , which in the simplest cases are represented by maps to and from spacetime , [ eq : particlefield ] @xmath25 and were @xmath26 is usually some vector space . in order to state the equations of motion , one has to first specify a set of so - called*kinematically possible trajectories * out of which
the dynamical entities @xmath23 are taken and solutions to the equations of motion are sought .
usually this involves particle trajectories which are sufficiently smooth ( typically piecewise twice continuously differentiable ) and fields which are sufficiently smooth and in addition have a sufficiently rapid fall - off at large spatial distances , so as to give rise to finite quantities of energy , angular - momentum , etc .
this space of kinematically possible trajectories will be denoted by @xmath27 . according to the discussion above , the equation of motion takes two arguments , one from @xmath24 the other from @xmath27 , and is hence written in the form @xmath28 where the zero on the right - hand side may be a many - component object . equation ( [ eq : eom ] )
should be read as a selection criterion on the set @xmath27 , depending on the externally specified values of @xmath22 .
we shall sometimes write @xmath29 for @xmath30 to denote the particular equation of motion for @xmath23 corresponding to the choice @xmath22 for the background structures .
in general , the sets of solutions to ( [ eq : eom ] ) for variable @xmath22 are @xmath22-dependent subset @xmath31 , whose elements are called the * dynamically possible trajectories*. we can now say more precisely what is usually meant by a symmetry : [ def : symmetrygroup ] an abstract group @xmath32 is called a * symmetry group * of the equations of motion iff the following conditions are satisfied : 1 .
there is an effective ( see below ) action @xmath33 of @xmath32 on the set of kinematically possible trajectories , denoted by @xmath34 .
2 . this action leaves the subset @xmath31 invariant ; that is , for all @xmath35 in @xmath32 we have : @xmath36 recall that an action is called effective if no group element other than the group identity fixes all points of the set it acts on .
effectiveness is required in order to prevent mathematically trivial and physically meaningless extensions of @xmath32 .
what really matters are the orbits of @xmath32 in @xmath27 , that is , the subsets @xmath37 for each @xmath38 .
if the action were not effective , we could simply reduce @xmath32 to a smaller group with an effective action and the same orbits in @xmath27 , namely the quotient group @xmath39 , where @xmath40 is the normal subgroup of elements that fix all points of @xmath27 .
it should be noted that this definition is still very general due to the fact that no further condition is imposed on the action of @xmath32 , apart from the obvious one of effectivity .
for example , for fields one usually requires the action to be ` local ' , in the sense that for any point @xmath5 of spacetime , the value @xmath41 of the @xmath35-transformed field should be determined by the value of the original field at some point @xmath42 of spacetime , and possibly _ finitely many _ derivatives of @xmath43 at @xmath42 . if there are no dependencies on the derivatives , the action is sometimes called ` ultralocal ' .
note that the point @xmath42 need not be identical to @xmath5 , but it is assumed to be uniquely determined by @xmath35 and @xmath5 .
a striking example of what can happen if locality is not imposed is given by the vacuum maxwell equations ( no external currents ) , which clearly admit the poincar group as ultralocally acting symmetry group .
what is less well known is the fact that they also admit the inhomogeneous galilei group as symmetry group and the electric displacement - field @xmath44 on one side , and between the magnetic induction - field @xmath45 and the magnetic field @xmath46 on the other ; see e.g. @xcite and @xcite .
] , albeit the action is non - local ; see @xcite or chap.5.9 of @xcite .
( there are also other non - local symmetries of the vacuum maxwell equations @xcite . ) to be strictly distinguished from the notion of symmetry is the notion of covariance , which we define as follows : [ def : covariancegroup ] an abstract group @xmath32 is called a * covariance group * of the equations of motion iff the following conditions are satisfied : 1 .
there is an effective action @xmath33 of @xmath32 on the set of kinematically possible trajectories , denoted by @xmath34 .
2 . there is also an action ( this time not necessarily effective ) @xmath47 of @xmath32 on the set of background structures , likewise denoted by @xmath48 .
3 . the solution - function @xmath49 from @xmath24 into the subsets of @xmath27 is @xmath32-equivariant .
this means the following : if @xmath50 denotes the set @xmath51 , then , for all @xmath35 in @xmath32 , we have @xmath52 an alternative way to say this is that the relation that @xmath19 establishes on @xmath53 via ( [ eq : eom ] ) is @xmath32 invariant , that is , for all @xmath35 in @xmath32 , we have @xmath54 the obvious difference between ( [ eq : defsymmetry ] ) and ( [ eq : defcovariance2 ] ) is that in the former case the background structure is not allowed to change .
the transformed dynamical entity is required to satisfy the _ very same _ equation as the untransformed one , whereas for a covariance it is only required to satisfy a suitably changed set of equations . here ` changed ' refers to the fact that @xmath55 is generally different from @xmath22 .
hence it is clear that a symmetry group is automatically also a covariance group , by just letting it act trivially on the set @xmath24 of background structures .
the precise partial converse is as follows : given a covariance group @xmath32 with action on @xmath24 , then for each @xmath56 define the ` stabiliser subgroup ' of @xmath22 in @xmath32 as the set of elements in @xmath32 that fix @xmath22 , @xmath57 then the subgroup @xmath58 of the covariance group is also a symmetry group of the equation of motion @xmath29 .
the requirement of covariance is a rather trivial one , since it can always be met by suitably taking into account all the background structures and a sufficiently general action of @xmath32 on @xmath24 . to see how this works in a specific example , consider the ordinary ` heat equation ' for the temperature field @xmath59 ( @xmath60 is a dimensionful constant ) : @xmath61 let @xmath62 be the 7-parameter group of euclidean motions ( rotations and translations in @xmath63 ) and time translations , whose defining representation on spacetime ( @xmath64 ) is denoted by @xmath65 , then @xmath32 acts effectively on the set of temperature fields via @xmath66 ( the inverse being just introduced to make this a _ left _
it is immediate from the structure of ( [ eq : heateq1 ] ) that this implements @xmath32 as symmetry group of this equation .
the background structures implicit in ( [ eq : heateq1 ] ) are : a ) a preferred split of spacetime into space and time , 2 ) a preferred measure and orientation of time , and c ) a preferred distance measure on space .
there are many ways to parametrise this structure , _ depending on the level of generality one starts from_. if , for example , we start from special relativity , we only list those structural elements that we need _ on top of _ the minkowski metric @xmath67 in order to write down ( [ eq : heateq1 ] ) .
they are given by a single constant and normalised timelike vector field @xmath13 , by means of which we can write ( [ eq : heateq1 ] ) in the form @xmath68 in the special class of inertial reference frames in which @xmath69 equation ( [ eq : heateq2 ] ) reduces to ( [ eq : heateq1 ] ) .
from the structure of ( [ eq : heateq2 ] ) it is obvious that this equation admits the whole poincar group of special relativity as covariance group .
however , the symmetry group it contains is the stabiliser subgroup of the given background structure .
the latter is given by the vector field @xmath13 , whose stabiliser subgroup within the poincar group is just @xmath70 , the same as for ( [ eq : heateq1 ] ) .
had we started from a higher level of generality , in which no preferred coordinate systems are given to us as in special relativity , we would write the heat equation in the form @xmath71 where now @xmath13 as well as @xmath35 feature as background structures .
@xmath13 is again specified as unit timelike covariant - constant vector field , @xmath35 as a flat metric , and @xmath72 as the unique covariant derivative operator associated to @xmath35 ( i.e. torsion free and preserving @xmath35 ) . since @xmath72 is here taken as a unique function of @xmath35 , it does not count as independent background structure .
once again it is clear from the structure of ( [ eq : heateq3 ] ) that the covariance group is now the whole diffeomorphism group of spacetime .
however , the symmetry group remains the same as before since the stabiliser subgroup of the pair @xmath73 is @xmath70 .
this example should make clear how easy it is to almost arbitrarily inflate covariance groups by starting from higher and higher levels of generality and adding the corresponding extra structures into ones list of background structures .
this possibility is neither surprising nor particularly disturbing .
slightly more disturbing is the fact that a similar game can be played with symmetries , at least on a very formal level .
the basic idea is to simply declare background structures to be dynamical ones by letting their values be determined by equations .
we may do this since we have so far not qualified ` equations of motion ' as any special sort of equations . for example , in the special relativistic context we may just take ( [ eq : heateq2 ] ) and let @xmath13 be determined by @xmath74 then ( [ eq : heateq2 ] ) and ( [ eq : heateq2-suppl ] ) together define a background free ( from the special relativistic point of view ) system of equations for @xmath75 which has the full poincar group as symmetry group .
its symbolic form is @xmath76 where the @xmath77 on the right - hand side has now 18 components : one for ( [ eq : heateq2 ] ) , one for the first equation in ( [ eq : heateq2-suppl ] ) , and 16 ( @xmath78 ) for the second equation in ( [ eq : heateq2-suppl ] ) .
but note that its @xmath59-sector of solution space is _ not _ the same as that of ( [ eq : heateq1 ] ) , as it now also contains solutions for different @xmath13 .
however , as the equations ( [ eq : heateq2-suppl ] ) for @xmath13 do not involve @xmath59 , the total solution space for @xmath79 can be thought of as fibred over the space of allowed @xmath13 , with each fibre over @xmath13 being given by the solutions @xmath59 of ( [ eq : heateq2 ] ) for that given @xmath13 .
each such fibre is a faithful image of the original solution space of ( [ eq : heateq1 ] ) , suitably transformed by a lorentz boost that relates the original @xmath13 in ( [ eq : heateq1 ] ) ( i.e. @xmath80 ) to the chosen one .
even more radically , we could take ( [ eq : heateq3 ] ) and declare @xmath13 and @xmath35 to be dynamical entities obeying the extra equations @xmath81=0\,,\ ] ] where @xmath82 is the riemann curvature tensor of @xmath35 , so that the last equation in ( [ eq : heateq3-suppl ] ) just expresses flatness of @xmath35 .
the system consisting of ( [ eq : heateq3 ] ) and ( [ eq : heateq3-suppl ] ) has no background structures and admits the full diffeomorphism group as symmetry group .
it is of the symbolic form @xmath83 which now comprises 36 components : the 16 as above and an additional set of 20 for the independent components of @xmath82 .
again , note that the @xmath59-sector of solution space of ( [ eq : heateq3 ] ) is now much bigger that of ( [ eq : heateq1 ] ) of or ( [ eq : heateq2 ] ) . with any solution @xmath59 it also contains its diffeomorphism - transformed one , @xmath84 , where @xmath85 .
again , since the equations for @xmath13 and @xmath35 do not involve @xmath59 , the total solution space is fibred over the allowed @xmath13 and @xmath35 fields , with each fibre corresponding to a faithful image of the original solution space for ( [ eq : heateq1 ] ) . finally we remark that , in principle , constants appearing in equations of motion could also be addressed as background structures whose values might eventually be determined by more general dynamical theories .
for example , one might speculate ( as was done some time ago in the so - called brans - dicke theories ) that the gravitational constant is actually the value of some field that only in the present epoch of our universe has settled to a spatially constant and quasi - static value , but whose value at much earlier times was significantly different .
another example from quantum field theory concerns the idea that masses of elementary particles are dynamically generated by the so - called higgs field ( whose existence is strongly believed but not yet experimentally confirmed ) . in any case ,
the important message from the considerations of this subsection is the following : symmetries emerge or disappear if , respectively , background structures become dynamical ( @xmath86 ) or dynamical structures ` freeze ' ( @xmath87 ) . within the concept of symmetry
as explained so far , an important distinction must be made between _
observable _ symmetries on one hand , and _ gauge _ symmetries on the other .
an observable symmetry transforms a state or a history of states ( trajectory ) into a _
different _ , that is , _ physically distinguishable _ state or history of states . on the other hand ,
a gauge symmetry transforms a state or a history of states into a _ physically indistinguishable _ state or a history of states . in this case
there is a redundancy in the mathematical description , so that the map from mathematical labels to physical states is not faithful .
this is usually associated with a group , called the _ group of gauge transformations _ , denoted by @xmath88 , which acts on the set of state labels such that two such labels correspond to the same physical state iff they lie in the same orbit of @xmath88 .
it is clear that the notion of ` distinguishability ' introduced here refers to the set of observables , i.e. functions on state space that are physically realisable in the widest sense .
assuming for the moment that this was well defined , we could attempt a definition as follows : [ def : observablesymmetry ] let @xmath32 be a symmetry group in the sense of definition[def : symmetrygroup ]
. then @xmath89 is called an * observable or physical symmetry * iff there exists a @xmath90 and a physical observable that separates @xmath91 from @xmath23 .
if no such observable exists , @xmath35 is called a * gauge symmetry*. it is clear that for a theoretician the stipulation of what functions on state space correspond to physically realisable observables is itself of hypothetical nature . however , what is important for us at this point is merely that _ relative _ to such a stipulation the distinction between observables and gauge symmetries makes sense . in the mathematical practice gauge symmetries
are often signalled by an underdeterminedness of the equations of motion , which sometimes simply fail to restrict the motion in certain degrees of freedom which are then called ` gauge degrees of freedom ' . in that case ,
given any solution @xmath90 , we can obtain another solution , @xmath92 , by just changing @xmath23 in those non - determined degrees of freedom in an arbitrary way .
for example , if the equations of motion are obtained via an action principle , such spurious degrees of freedom will typically reveal their nature through the property that motions in them are not associated with any action . as a result , the equations of motion , which are just the condition for the stationarity of the action , will not constrain the motion in these directions .
conversely , if according to the action principle the motion in some degree of freedom costs action , it can hardly be called a redundant one . in this sense
an action principle is not merely a device for generating equations of motion , but also contains some information about observables .
the combination of observable and gauge symmetries into the total symmetry group @xmath32 need not at all be just that of a semi - direct or even direct product .
often , in field theory , the gauge group @xmath88 is indeed a subgroup of @xmath32 , in fact an invariant ( normal ) one , but the observable symmetries , @xmath93 , are merely a quotient and not a subgroup of @xmath32 . in standard group
theoretic terms one says that @xmath32 is a @xmath94extension of @xmath93 .
this typically happens in electromagnetism or more generally in yang - mills type gauge theories or general relativity with globally charged configurations . in this case
only the ` gauge transformations ' with sufficiently rapid fall - off at large spatial distances are proper gauge transformations in our sense , whereas the long ranging ones cost action if performed in real time and therefore have to be interpreted as elements of @xmath93 ; see e.g. @xcite and chap.6 of @xcite .
this ends our small excursion into the realm of meanings of ` symmetry ' .
we now turn to the discussion of specific aspects in pauli s work .
the usage of symmetry concepts in pauli s work is so rich and so diverse that it seems absolutely hopeless , and also inappropriate , to try to present them in a homogeneous fashion with any claim of completeness .
rather , i will comment on various subjectively selected aspects without in any way saying that other aspects are of any lesser significance . in fact , i will not include some of his most outstanding contributions , like , for example , the formulation of the exclusion principle , the neutrino hypothesis , or his anticipation of yang - mills gauge theory for the strong interaction .
there exist excellent reviews and discussions of these topics in the literature .
specifically i wish to refer to bartel van der waerden s contribution _ exclusion principle and spin _ to the pauli memorial volume ( @xcite , pp.199 - 244 ) , norbert straumann s recent lecture on the history of the exclusion principle @xcite , pauli s own account of the history of the neutrino ( in english : @xcite , pp.193 - 217 ; in german : @xcite , vol.2 , pp
. 1313 - 1337 and @xcite , p.156 - 180 ) , chien - shiung wu s account _ the neutrino _ in the pauli memorial volume ( @xcite , pp.249 - 303 ) , and the historical account of gauge theories by lochlainn oraifeartaigh and norbert straumann @xcite . a non - technical overview concerning _
pauli s belief in exact symmetries _ is given by karl von meyenn @xcite .
last , but clearly not least , i wish to mention charles enz s fairly recent comprehensive scientific biography @xcite of wolfgang pauli , which gives a detailed discussion of his scientific vre . in this contribution
i rather wish to concentrate on some particular aspects of the notion of symmetry that are directly related to the foregoing discussion in sections[sec : dynamicalsymmetriesversuscovariance ] and [ sec : physicalversusgaugesymmetries ] , as i feel that they are somewhat neglected in the standard discussions of symmetry . in january 1926
pauli managed to deduce the energy spectrum for the hydrogen atom from the rules of matrix mechanics .
for this he implicitly used the fact that the mechanical problem of a point charge moving in a spherically symmetric force - field with a fall - off proportional to the square of the inverse distance has a symmetry group twice as large ( i.e. of twice the dimension ) as the group of spatial rotations alone , which it contains .
hence the total symmetry group is made half of a ` kinematical ' part , referring to space , and half of a ` dynamical ' part , referring to the specific force law ( @xmath95 fall - off ) .
their combination is a proper physical symmetry group that transforms physically distinguishable states into each other . in the given quantum - mechanical context
one also speaks of ` spectrum generating ' symmetries .
let us recall the classical problem in order to convey some idea where the symmetries and their associated conserved quantities show up , and how they may be employed to solve the dynamical problem .
consider a mass - point of mass @xmath4 and position coordinate @xmath96 in the force field @xmath97 , where @xmath98 is the length of @xmath96 , @xmath99 , and @xmath100 is some dimensionful constant .
then , according to newton s 3rd law ( an overdot stands for the time derivative ) , @xmath101 next to energy , there are three obvious conserved quantities corresponding to the three components of the angular - momentum vector ( here written per unit mass ) @xmath102 but there are three more conserved quantities , corresponding to the components of the following vector ( today called the lenz - runge vector ) , @xmath103 conservation can be easily verified by differentiation of ( [ eq : lenzrungevect ] ) using ( [ eq : eqmotsphsympot ] ) and @xmath104 . hence on has ( @xmath105 = length of @xmath106 ) @xmath107 from which the classical orbit immediately follows : setting @xmath108 , the last equation ( [ eq : conicsection1 ] ) reads @xmath109 which is the well known equation for a conic section in the plane perpendicular to @xmath106 , focus at the origin , eccentricity @xmath110 (= length of @xmath111 ) , and _
latus rectum _ @xmath112 .
the vector @xmath111 points from the origin to the point of closest approach ( periapsis ) .
the few steps leading to this conclusion illustrate the power behind the method of working with conservation laws which , in turn , rests on an effective exploitment of symmetries .
the total energy per unit mass is given by @xmath113 .
a simple calculation shows that @xmath114 which allows to express @xmath115 as function of the invariants @xmath116 and @xmath117 .
this is the relation which pauli shows to have an appropriate matrix analogue , where it allows to express the energy in terms of the eigenvalues of the matrices for @xmath117 and @xmath116 which pauli determines , leading straight to the balmer formula .
from a modern point of view one would say that , for fixed energy @xmath118 , one obtains a hamiltonian action of @xmath119 . ]
the state space of this problem carries a hamiltonian action of the lie algebra @xmath120 , generated by the 3 + 3 quantities @xmath106 and @xmath111 .
quantisation then consists in the problem to represent this lie algebra as a commutator algebra of self - adjoint operators and the determination of spectra of certain elements in the enveloping algebra .
this is what pauli did , from a modern point of view , but clearly did not realise at the time .
in particular , even though he calculated the commutation relations for the six quantities @xmath106 and @xmath111 , he did not realise that they formed the lie algebra for @xmath120 , as he frankly stated much later ( 1955 ) in his address on the occasion of hermann weyl s 70th birthday : , p.402 ) note that , in modern terminology , pauli actually refers to a representation of the _ lie algebra _ of the orthogonal group . ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ similarly i did not know that the matrices which i had derived from the new quantum mechanics in order to calculate the energy values of the hydrogen atom were a representation of the 4-dimensional orthogonal group .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this may be seen as evidence for pauli s superior instinct for detecting relevant mathematical structures in physics . much later ,
in a cern - report of 1956 , pauli returned to the representation - theoretic side of this problem @xcite .
the first big impact of group theory proper on physics took place in quantum theory , notably through the work or eugene wigner @xcite and hermann weyl @xcite .
while in atomic spectroscopy the usage of group theory could be looked upon merely as powerful mathematical tool , it definitely acquired a more fundamental flavour in ( quantum ) field theory .
according to a dictum usually attributed to wigner , every elementary system ( particle ) in special - relativistic quantum theory corresponds to a unitary irreducible representation of the poincar group .
in fact , all the poincar invariant linear wave equations on which special - relativistic quantum theory is based , known by the names of klein & gordon , weyl , dirac , maxwell , proca , rarita & schwinger , bargmann & wigner , pauli & fierz , can be understood as projection conditions that isolate an irreducible sub - representation of the poincar group , or sometimes an extension thereof by the discrete transformations of space and time reversal . ] within a reducible one that is easy to write down .
more concretely , the latter is usually obtained as follows : take a field @xmath43 on spacetime @xmath2 with values in a finite - dimensional complex vector space @xmath26 .
let @xmath121 be a finite - dimensional irreducible representation of the ( double cover of the ) lorentz group @xmath122 on @xmath26 .
is never unitary , simply because the lorentz group has no non - trivial finite - dimensional unitary irreducible representations .
but it will give rise to an infinite - dimensional representation on the linear space of fields @xmath43 which will indeed be unitary . ]
it is uniquely labelled by a pair @xmath123 of two positive integer- or half - integer - valued numbers . in the standard terminology
, @xmath124 corresponds to the number of unprimed , @xmath125 to the number of primed spinor indices of @xmath43 .
the set of such fields furnishes a linear representation of the ( double cover of the ) poincar group , @xmath126 , where the action of the group element @xmath127 is given by @xmath128 or for the fourier transform @xmath129 , @xmath130 one immediately infers from ( [ eq : actiononfield - ft ] ) that irreducibility implies that @xmath129 must have support on a single @xmath122 orbit in momentum space .
here one usually restricts to those orbits consisting of non - spacelike @xmath5 ( those with spacelike @xmath5 give rise to the tachyonic representations which are deemed unphysical ) , which are labelled by @xmath131 with non - negative @xmath4 . for @xmath43
this means that it obeys the klein - gordon equation @xmath132 .
this is already half the way to an irreducible representation , insofar as it now contains only modes of fixed mass .
but these modes still contains several spins up to the maximal value @xmath133 . a second and last step
then consists of projecting out one ( usually the highest ) spin , which gives rise to the equations named above . in this fashion
the physical meanings of _ mass _ and _ spin _ merge with the abstract mathematical meaning of mere labels of irreducible representations .
mass and spin are the most elementary attributes of physical objects , so that objects with no other attributes are therefore considered _ elementary_.
as just described , these elementary attributes derive from the representation theory of a group whose significance is usually taken to be that it is the automorphism group of spacetime .
however , as already discussed in sections[sec : genintro ] and [ sec : symmetryspacetime ] , this point of view presupposes a hierarchy of physical thinking in which spacetime ( here minkowski space ) is considered an entity prior to ( i.e. more fundamental than ) matter , which may well be challenged .
a more consistent but also more abstract point of view would be to think of the abstract poincar group as prior to the matter content _ as well as _ the spacetime structure and to derive both simultaneously . here ` deriving ' a spacetime structure ( geometry ) from a group would be meant in the sense of klein s _ erlanger programm _
@xcite .
we have already discussed in section[sec : genintro ] pauli s shift in emphasis towards a more abstract point of view as regards spacetime structure . but also as regards to matter he was , next to wigner , one of the proponents to put symmetry considerations first and to derive the wave equations of fundamental fields as outlined above . based on previous work by fierz on the theory of free wave equations for higher spin
@xcite , fierz and pauli published their very influential paper _ on relativistic equations for particles of arbitrary spin in an electromagnetic field _
( @xcite , vol.2 , pp.873 - 894 ) which is still much cited today . in fact , much earlier , in his 1927 paper _
_ quantum mechanics of the magnetic electron _ _ , pauli succeeded to implement the electron s spin into non - special - relativistic quantum mechanics in an entirely representation - theoretic fashion as regards the ( lie algebra of ) spatial rotations .
in contrast to the other ( translational ) degrees or freedom , spin does not appear as the quantisation of an already existent classical degree of freedom .
this must have appeared particularly appealing to pauli , who never wanted the electron s ` spin ' to be understood as an intrinsic angular momentum due to a spatial rotation of a material structure . when pauli introduced the new spin quantum - number for the electron in his 1924 paper _
_ on the influence of the velocity dependence of the electron mass on the zeeman effect _
_ he deliberately stayed away from any model interpretation and cautiously referred to it as _ _ a peculiar , classically indescribable disposition of two - valuedness of the quantum - theoretic properties of the light - electron__. at that time an understandable general scepticism against possible erroneous prejudices imposed by the usage of classical models had already firmly established itself in pauli s ( and others ) thinking .
as much justified as this is in view of quantum mechanics , this had also led to overstatements to the effect that spin has no classical counterpart and that any classical model is even classically contradictory in the sense of violating special relativity .
as regards the second point , which was also pushed by pauli , we refer to the detailed discussion in @xcite . to the first point we first wish to mention that composite models with half - integer angular momentum states exist in ordinary quantum mechanics ( without spin ) , as , e.g. , pointed out by bopp & haag in 1950 @xcite .
this is possible if their classical configuration space contains the whole group @xmath134 of spatial rotations .
pauli himself showed in his 1939 paper _ _ on a criterion for single- or double - valuedness of the eigenfunctions in wave mechanics _ _ the possibility of double - valued wavefunctions , which are the ones that give rise to half - integer angular momentum states .
moreover , in classical mechanics there is also a precise analog of wigner s notion of an elementary system .
recall that the space of states of a mechanical system is a symplectic manifold ( phase space ) .
the analog of an irreducible and unitary representation of the group of spacetime automorphisms is now a transitive and hamiltonian action of this group on the symplectic manifold .
it is interesting to note that this classical notion of an elementary system was only formulated much later than , and in the closest possible analogy with , the quantum mechanical one .
an early reference where this is spelled out is @xcite .
the classification of elementary systems is now equivalent to the classification of symplectic manifolds admitting such an action .
an early reference where this has been done is @xcite . here ,
as expected , an intrinsic angular momentum shows up as naturally as it does in quantum mechanics .
what makes it slightly unusual ( but by no means awkward or even inconsistent ) is the fact that it corresponds to a phase space that is not the cotangent bundle ( space of momenta ) over some configuration space of positions .
pauli s later writings also show this strong inclination to set the fundamentals of ( quantum ) field theory in group - theoretic terms . in his survey
_ relativistic field theories of elementary particles _ ( @xcite , vol.2 , pp.923 - 952 ) , written for the 1939 solvay congress , pauli immediately starts a discussion of `` transformation properties of the field equations and conservation laws '' . his posthumously published notes on _ continuous groups in quantum mechanics _ @xcite focus exclusively on lie - algebra methods in representation theory .
today we are used to _ define _ physical quantities like energy , momentum , and angular momentum as the conserved quantities associated to spacetime automorphisms via noether s theorem . here ,
too , pauli was definitely an early advocate of this way of thinking .
reviews on the subject written shortly after pauli s death show clear traces of pauli s approach ; see e.g. @xcite .
pauli s proof of the spin - statistics correlation @xcite ( also @xcite , vol.2 , pp.911 - 922 ) , first shown by markus fierz in his habilitation thesis @xcite , is a truly impressive example for the force of abstract symmetry principles . here
we wish to recall the basic lemmas on which it rests , which merely have to do with classical fields and representation theory .
we begin by replacing the proper orthochronous lorentz group by its double (= universal ) cover @xmath122 in order to include half - integer spin fields .
we stress that everything that follows merely requires the invariance under this group .
no requirements concerning invariance under space- or time reversal are needed !
we recall from the previous section that any finite - dimensional complex representation of @xmath122 is labelled by an ordered pair @xmath123 , where @xmath5 and @xmath135 may assume independently all non - negative integer or half - integer values .
@xmath124 and @xmath125 correspond to the numbers of ` unprimed ' and ` primed ' spinor indices , respectively .
the tensor product of two such representations decomposes as follows : @xmath136 where and this is the important point in what follows the sums proceed in _ integer _ steps in @xmath98 and @xmath137 . with each @xmath138
let us associate a ` pauli index ' , given by @xmath139 this association may be extended to sums of such @xmath138 proceeding in integer steps , simply by assigning to the sum the pauli index of its terms ( which are all the same ) .
then we have is a homomorphism of semigroups .
one semigroup consists of direct sums of irreducible representations proceeding in integer steps with operation @xmath140 , the other is @xmath141 , which is actually a group . ]
@xmath142 according to their representations , we can associate a pauli index with spinors and tensors .
for example , a tensor of odd / even degree has pauli index @xmath143/@xmath144 .
the partial derivative , @xmath145 , counts as a tensor of degree one .
now consider the most general linear ( non interacting ) field equations for integer spin ( here and in what follows @xmath146 simply stands for `` sum of terms of the general form @xmath147 '' ) : @xmath148 these are invariant under @xmath149 next consider any current that is a polynomial in the fields and their derivatives : @xmath150 then one has @xmath151 this shows that for any solution of the field equations with charge @xmath152 for the conserved current @xmath153 ( @xmath152 being the space integral over @xmath154 ) there is another solution ( the @xmath155 transformed ) with charge @xmath156 .
it follows that charges of conserved currents can not be sign - definite in any @xmath122-invariant theory of non - interacting integer spin fields . in the same fashion
one shows that conserved quantities , stemming from divergenceless symmetric tensors of rank two , bilinear in fields , can not be sign - definite in any @xmath157 invariant theory of non - interacting half - integer spin fields .
in particular , the conserved quantity in question could be energy !
an immediate but far reaching first conclusion ( not explicitly drawn by pauli ) is that there can not exist a relativistic generalisation of schrdinger s one- particle wave equation .
for example , for integer - spin particles , one simply can not construct a non - negative spatial probability distribution derived from conserved four - currents .
this provides a general argument for the need of second quantisation , which in textbooks is usually restricted to the spin - zero case . upon second quantisation the celebrated spin - statistics connection for free fields
can now be derived in a few lines .
it says that integer spin fields can not be quantised using anti - commutators and half - integer spin field can not be quantised using commutators . here
the so - called jordan - pauli distribution plays a crucial role in the ( anti)commutation relations , which ensures causality ( observables localised in spacelike separated regions commute ) .
also , the crucial hypothesis of the existence of an @xmath157 invariant stable vacuum state is adopted .
pauli ends his paper by saying : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in conclusion we wish to state , that according to our opinion the connection between spin and statistics is one of the most important applications of the special relativity theory . _
( @xcite , p.722 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it took almost 20 years before first attempts were made to generalise this result to the physically relevant case of interacting fields by lders & zumino @xcite .
general covariance is usually presented as _ the _ characteristic feature of general relativity .
the attempted meaning is that a generally covariant law takes the ` same form ' in all spacetime coordinate systems .
however , in order to define the ` form ' of a law one needs to make precisely the distinction between background entities , which are constitutive elements of the law , and the dynamical quantities which are to be obey the laws so defined ( cf .
section[sec : symmetryspacetime ] ) . in the language we introduced above , ` general covariance ' can not just mean simple covariance under all smooth and invertible transformations of spacetime points , i.e. that the spacetime diffeomorphism group is a covariance group is the sense of definition[def : covariancegroup ] , for that would be easily achievable without putting any restriction on the intended law proper , as was already pointed out by erich kretschmann in 1917 @xcite .
einstein agreed with that criticism of kretschmann s , which he called `` acute '' ( german original : `` scharfsinnig'')(@xcite , vol.7 , doc.4 , pp.38 - 41 ) , and withdrew to the view that the principle of general covariance has at least some heuristic power in the following sense : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ between two theoretical systems which are compatible with experience , that one is to be preferred which is the simpler and more transparent one from the standpoint of the absolute differential calculus .
try to bring newton s gravitational mechanics in the form of generally covariant equations ( four dimensional ) and one will surely be convinced that principle a ) is , if not theoretically , but practically excluded . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ but the _ principle of general covariance _ is intended as a non - trivial selection criterion . hence modern writers often characterise it as the requirement of diffeomorphism invariance , i.e. that the diffeomorphism group of spacetime is a symmetry group in the sense of definition[def : symmetrygroup ] .
but then , as we have seen above , the principle is open to trivialisations if one allows background structures to become formally dynamical .
this possibility can only be inhibited if one limits the amount of structure that may be added to the dynamical fields .
the reason why i mention all this here is that pauli s relativity article is , to my knowledge , the only one that seems to address that point , albeit not as explicitly as one might wish . after mentioning kretschmann s objection ,
he remarks ( the emphases are pauli s ) : _ allein _ beschrieben wird , und da diese nicht unabhngig von der materie gegeben , sondern selbst durch feldgleichungen bestimmt sind .
erst deshalb knnen die @xmath158 als _ physikalische zustandsgren _ bezeichnet werden .
( @xcite , p.181 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the generally covariant formulation of the physical laws acquires a physical content only through the principle of equivalence , in consequence of which gravitation is described _
solely _ by the @xmath158 and the latter are not given independently from matter , but are themselves determined by the field equations .
only for this reason can the @xmath158 be described as _ physical quantities_. _ ( @xcite , p.150 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ note how perceptive pauli addresses the two central issues : 1 ) that one has to limit the the amount of dynamical variables and 2 ) that dynamical structures have to legitimate themselves as physical quantities through their back reaction onto other ( matter ) structures .
it is by far the best few - line account of the issue that i know of , though perhaps a little hard to understand without the more detailed discussion given above in section[sec : dynamicalsymmetriesversuscovariance ] .
most modern textbooks do not even address the problem .
see @xcite for more discussion . in this section
i wish to give a brief but illustrative example from pauli s work for the non - trivial distinction between observable physical symmetries on one hand , and gauge symmetries on the other ( cf . section[sec : physicalversusgaugesymmetries ] ) .
the example i have chosen concerns an argument within the ( now outdated ) attempts to understand elementary particles as regular solutions of classical field equations .
pauli reviewed such attempts in a rather detailed fashion in his relativity article , with particular emphasis on weyl s theory , to which he had actively contributed in two of his first three published papers in 1919 .
the argument proper says that in any ` generally covariant ' theory , which allows for regular static solutions representing charged particles , there exists for any solution with mass @xmath4 and charge @xmath110 another such solution with the same mass but opposite charge @xmath159 .
pauli s proof looks like an almost trivial application of general covariance and runs as follows : let @xmath160 and @xmath161 represent the gravitational and electromagnetic field respectively .
the hypothesis of staticity implies that coordinates ( and gauges for @xmath162 ) can be chosen such that all fields are independent of the time coordinate , @xmath163 , and that @xmath164 as well as @xmath165 for @xmath166 .
may , in fact , be relaxed .
] now consider the orientation - reversing diffeomorphism @xmath167 .
it maps the gravitational field to itself while reversing the sign of @xmath168 and hence of the electric field .
general covariance assures these new fields to be again solutions with the same total mass but opposite total electric charge .
pauli presents this argument in his second paper addressing weyl s theory , entitled _ _ to the theory of gravitation and electricity by hermann weyl _ _
( @xcite , vol.2 , pp.13 - 23 , here p.18 ) and also towards the end of section67 of his relativity article .
the idea of this proof is due to weyl who communicated it ( without formulae ) in his first two letters to pauli ( @xcite , vol.1 , doc.[1 ] and [ 2 ] ) , as pauli also acknowledges in his paper ( @xcite , vol.2 , p.18 , footnote2 ) .
it is interesting to note that einstein rediscovered the very same argument in 1925 and found it worthy of a separate communication @xcite . at the time it was common to all , weyl , pauli , and einstein , to regard the argument a nuisance and of essentially destructive nature
this was because at this time antiparticles had not yet been discovered so that the apparent asymmetry as regards the sign of the electric charges of fundamental particles was believed to be a fundamental property of nature .
already in his first paper on weyl s theory ( @xcite , vol.2 , pp.1 - 9 ) , entitled _ _ perihelion motion of mercury and deflection of rays in weyl s theory of gravitation _ _ , pauli emphasised : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the main difficulty [ with weyl s theory ] is apart from einstein s objection , which appears to me not yet sufficiently disproved that the theory can not account for the asymmetry between the two sorts of electricity . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ now , there is an interesting conceptual point hidden in this argument that relates to our discussions in sections[sec : dynamicalsymmetriesversuscovariance ] and [ sec : physicalversusgaugesymmetries ] .
first of all , the two solutions are clearly considered physically distinct , otherwise the argument could not be understood as contradicting the charge asymmetry in nature .
hence the diffeomorphism involved can not be considered a gauge transformation but rather corresponds to a proper physical symmetry . on the other hand , we know that diffeomorphisms within bounded regions must be considered as gauge transformations , for otherwise one would run into the dilemma set by the so - called `` hole argument '' be a bounded region in spacetime which is disjoint from a spacelike hypersurface @xmath22 .
consider two solutions to the field equations which merely differ by the action of a diffeomorphism @xmath16 with support in @xmath169 .
if they are considered distinct , then the theory can not have a well posed initial - value problem , since then for any @xmath22 distinct solutions exist with identical data on @xmath22 .
this is a rephrasing of einstein s original argument ( @xcite , vol.4 , doc.25 , p.574 , doc.26 , p580 , vol.6 , doc.2 , p.10 ) , which did not construct a contradiction to the existence of a well posed _
initial - value problem _ , but rather to the requirement that the gravitational field be determined by the matter content ( more precisely : its energy momentum tensor ) . but this requirement is clearly never fulfilled in any generally covariant theory in which the gravitational field has its own degrees of freedom , independent of whether one regards diffeomorphisms as gauge . slightly later
he rephrased it so as to construct a contradiction to the existence of a well posed _
boundary - value problem _
( @xcite , vol.6 , doc.9 , p.110 ) , which is also not the right thing to require from equations that describe the propagation of fields with own degrees of freedom . ] .
hence one faces the problem of how one should characterise those diffeomorphisms which are not to be considered as gauge transformations ( cf .
section[sec : physicalversusgaugesymmetries ] ) .
it is conceivable that this question is not decidable without contextual information .
( see e.g. @xcite and chapter6 of @xcite for more discussion of this point . )
the historical sources have almost nothing to say about this , though there are suggestions by all three mentioned authors how to circumvent the argument by adding more non - dynamical structures , as a result of which general covariance is lost .
einstein , being most explicit here , suggested the existence of a timelike vector field which fixes a time orientation .
at least the so - defined time orientation would then have to be considered as non - dynamical structure of type @xmath22 ( cf .
section[sec : dynamicalsymmetriesversuscovariance ] ) in order to break the symmetry group down to the stabiliser group of @xmath22 .
the time - orientation - reversing transformation used above would then not be a symmetry anymore .
similar suggestions were made by weyl , who also hinted at a structure to distinguish past and future : halte ich , i m gegensatz zu den meisten physikern , fr eine tatsache von noch viel fundamentalerer bedeutung als der wesensunterschied zwischen positiver und negativer elektrizitt . ''
( @xcite , vol.1 , doc.[2 ] , p.6 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ their essential difference [ of past and future ] i take , contrary to most physicists , to be a fact of much more fundamental meaning than the essential difference between positive and negative charge . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in the last ( 5th ) edition of _ raum zeit materie _ , hermann weyl writes regarding his unified theory ( the emphases are weyl s ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the theory gives no clue as regards the _ disparity of positive and negative electricity_. but that can not be taken as a reproach against the theory . for that disparity
is based without doubt on the fact that of both fundamental constituents of matter , the electron and the hydrogen nucleus , the positively charged one is tight to another mass then the negatively charged one ; it originates from the nature of matter and not of the field .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ given that weyl is talking about his unified field - theory of gravity and electricity , whose original claim was to explain all of matter by means of field theory , this statement seems rather surprising .
it may be taken as a sign of weyl s beginning retreat from his once so ambitious programme .
one issue that attracted much attention during the 1960s was , whether the observed particle multiplets could be understood on the basis of an all embracing symmetry principle that would combine the poincar group with the internal symmetry groups displayed by the multiplet structures .
this combination of groups should be non - trivial , i.e. , not be a direct product , for otherwise the internal symmetries would commute with the spacetime symmetries and lead to multiplets degenerate in mass and spin ( see , e.g. , @xcite ) .
subsequently , a number of no - go theorems appeared , which culminated in the now most famous theorem of coleman & mandula @xcite .
this theorem states that those generators of symmetries of the @xmath170-matrix belonging to the poincar group necessarily commute with those belonging to internal symmetries .
the theorem is based on a series of assumptions ) s - matrix which depends analytically on @xmath137 ( the squared centre - of - mass energy ) and @xmath171 ( the squared momentum transfer ) ; ( 2 ) the mass spectrum of one - particle states consists of ( possibly infinite ) isolated points with only finite degeneracies ; ( 3 ) the generators ( of the lie algebra ) of symmetries of the @xmath170-matrix contains ( as a lie - sub algebra ) the poincar generators ; ( 4 ) some technical assumptions concerning the possibility of writing the symmetry generators as integral operators in momentum space . ] involving the crucial technical condition that the @xmath170-matrix depends analytically on standard scattering parameters . what is less visible here is that the structure of the poincar group enters in a decisive way .
this result would not follow for the galilean group , as was explicitly pointed out by coleman & mandula ( @xcite , p.159 ) .
one way to avoid the theorem of coleman & mandula is to generalise the notion of symmetries .
an early attempt was made by golfand & likhtman @xcite , who constructed what is now known as a super - lie algebra , which generalises the concept of lie algebra ( i.e. symmetry generators obeying certain commutation relations ) to one also involving anti - commutators . in this way it became possible for the first time to link particles of integer and half - integer spin by a symmetry principle .
it is true that supersymmetry still maintains the degeneracy in masses and hence can not account for the mass differences in multiplets .
but its most convincing property , the symmetry between bosons and fermions , suggested a most elegant resolution of the notorious ultraviolet divergences that beset quantum field theory .
it is remarkable that the idea of a cancellation of bosonic and fermionic contributions to the vacuum energy density occurred to pauli . in his lectures _ selected topics in field quantization _ , delivered in 1950 - 51 ( in print again since 2000 , @xcite ) , he posed the question _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ .. whether these zero - point energies [ from bosons and fermions ] can compensate each other . _ ( @xcite , p.33 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ he tried to answer this question by writing down the formal expression for the zero point energy density of a quantum field of spin @xmath172 and mass @xmath173 ( pauli restricted attention to spin @xmath77 and spin @xmath174 , but the generalisation is immediate ) : @xmath175 cancellation should take place for high values of @xmath176 .
the expansion @xmath177 shows that the quartic , quadratic , and logarithmic terms must cancel in the sum over @xmath172 for the limit @xmath178 to exist .
this implies that for @xmath179 one must have @xmath180 pauli comments that _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ these requirements are so extensive that it is rather improbable that they are satisfied in reality . _
( @xcite , p.33 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ unless enforced by an underlying symmetry _ ,
one is tempted to add !
this would have been the first call for a supersymmetry in the year 1951 .
however , the real world does not seem to be as simple as that .
supersymmetry , if at all existent , is strongly broken in the phase we live in .
so far no supersymmetric partner of any existing particle has been detected , even though some of them ( e.g. , the neutralino ) are currently suggested to be viable candidates for the missing - mass problem in cosmology .
future findings ( or non - findings ) at the large hadron collider ( lhc ) will probably have a decisive impact on the future of the idea of supersymmetry , which whether or not it is realised in nature is certainly very attractive ; and pauli came close to it . ever since its first formulation in 1921 , pauli as well as einstein were much attracted by the geometric idea of theodor kaluza and its refinement by oskar klein , according to which the classical theories of the gravitational and the electromagnetic field could be unified into a single theory , in which the unified field has the same meaning as einsteins gravitational field in general relativity , namely as metric tensor of spacetime , but now in five instead of four dimensions .
the momentum of a particle in the additional fifth direction ( which is spacelike ) is now to be interpreted as its charge .
charge is conserved because the geometry of spacetime is _ a priori _ restricted to be independent of that fifth direction .
the combined field equations are exactly the five - dimensional analog of einstein s equations for general relativity .
a natural question to address in this unified classical theory was whether it admits solutions that could represent particle - like objects .
more precisely , the solution should be stationary , everywhere regular , and possess long - ranging gravitational and electromagnetic fields ( usually associated with aspects of mass and charge ) .
pauli , who was very well familiar with this theory since its first appearance , kept an active interest in it even after the formulations of quantum mechanics and early quantum electrodynamics , which made it unquestionable for him that the problem of matter could not be adequately addressed in the framework of a classical field theory , unlike einstein , who maintained such a hope in various forms until the end of his life in 1955 .
it is therefore remarkable that in 1943 einstein and pauli wrote a paper in which they proved the non - existence of such solutions .
the introduction contains the following statement : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ when one tries to find a unified theory of the gravitational and electromagnetic fields , he can not help feeling that there is some truth in _
kaluza s _ five - dimensional theory .
( @xcite , p.131 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in fact , einstein and pauli offered a proof for the more general situation with an arbitrary number of additional space dimensions , fulfilling the generalised kaluza - klein `` cylinder - condition '' that the gravitational field should not depend on any of these extra directions .
note that this extra condition introduces non - dynamical background structures , so that of the 5-dimensional diffeomorphism group only those diffeomorphisms preserving this condition can act as symmetries , a point pauli often emphasised as a deficiency regarding the kaluza - klein approach . restricting attention to five dimensions ,
the explicitly stated hypotheses underlying the proof were these ( @xcite , p.131 ; annotations in square brackets within quotations are mine ) : * `` the field is stationary ( i.e the @xmath158 [ the five - dimensional metric ] are independent of @xmath181 [ the time coordinate ] ) . '' clearly
, @xmath158 is also assumed to be independent of the fifth coordinate @xmath182 . * `` it [ the field @xmath158 ] is free from singularities . ''
* `` it is imbedded in a euclidean space ( of the minkowski type ) , and for large values of @xmath98 ( @xmath98 being the distance from the origin of the spatial coordinate system ) @xmath184 has the asymptotic form @xmath185 , where @xmath186 . ''
the last condition is meant to assure the non - triviality of the solution , i.e. that there really is an attracting object at the spatial origin .
this becomes clear if one recalls that in the lowest weak - field and slow - motion approximation @xmath187 just corresponds to the newtonian gravitational potential .
unfortunately , the other statement : `` it is imbedded in a euclidean space ( of the minkowski type ) '' seems ambiguous , since the solution is clearly not meant to be just ( a portion of ) 5-dimensional flat minkowski space .
hence the next closest reading is presumably that the underlying five - dimensional spacetime manifold is ( diffeomorphic to ) @xmath188 , with some non - flat metric of minkowskian signature @xmath189 . the elegant method of proof makes essential use of the fact that the suitably restricted group of spacetime diffeomorphisms ( to those preserving the cylinder condition ) is a symmetry group for the full set of equations in the sense of ( [ eq : defsymmetry ] ) of definition[def : symmetrygroup ] .
more precisely , two types of diffeomorphisms from that class are considered separately by einstein and pauli : * arbitrary ones in the three coordinates @xmath190 which leave invariant the @xmath191 coordinates . *
linear ones in the @xmath191 coordinates , leaving invariant the @xmath190 .
now , as a matter of fact , this innocent looking split introduces a further and , as it turns out , crucial restriction , over and above the hypotheses h1-h3 .
the point is that the split and , in particular , the set d2 of diffeomorphisms simply do not exist unless the spacetime manifold , which in h3 was assumed to be @xmath188 , globally splits into @xmath192 such that the first factor , @xmath193 , corresponds to the @xmath194-planes of constant spatial coordinates @xmath190 and the second factor , @xmath63 , corresponds to the @xmath195-spaces of constant coordinates @xmath191 . but
this need not be the case if h1-h3 are assumed .
the identity derived by einstein and pauli from the requirement that transformations of the field induced by diffeomorphisms of the type d2 are symmetries are absolutely crucial in proving the non - existence of regular solutions .
we now know that this additional restriction is essential to the non - existence result : there do exist solutions of the type envisaged that satisfy h1-h3 , but violate the extra ( and superfluous ) splitting condition .
coordinate parametrises circles which combine with the 2-spheres ( polar coordinates @xmath196 ) of constant spatial radius , @xmath98 , into 3-spheres ( hopf fibration ) which are parametrised by @xmath197 , now thought of as euler angles .
the radii of these 3-spheres appropriately shrink to zero as @xmath98 tends to zero , so that @xmath198 define , in fact , polar coordinates of @xmath199 .
together with time , @xmath181 , we get @xmath188 as global topology . now
the submanifolds of constant @xmath190 are those of constant @xmath200 and have a topology @xmath201 rather than @xmath193 , so that the linear transformations d2 in the @xmath194 coordinates do not define diffeomorphisms of the kaluza - klein spacetime manifold .
] they are called _ kaluza - klein monopoles _ @xcite@xcite and carry a gravitational mass as well as a magnetic charge .
it is hard to believe that pauli as well as einstein would not have been much impressed by those solutions , though possibly with different conclusions , had they ever learned about them .
it is also conceivable that these solutions could have been found at the time , had real attempts been made , rather than possibly discouraged by pauli s and einstein s result .
in fact , kurt gdel , who was already in princeton when pauli visited einstein , found his famous cosmological solution @xcite in 1949 by a very similar geometric insight that also first led to the kaluza - klein monopole @xcite . in the kk case , @xmath202 in gdels case .
this simplifies the calculations considerably . ]
one of pauli s major interests were discrete symmetries , in particular the transformation of space inversion , @xmath203 , also called _ parity _ transformation . given a linear wave equation which is symmetric under the proper orthochronous (
i.e. including no space and time inversions ) poincar group , one may ask whether it is also symmetric under space and time inversions . for this to be a well defined question one has to formulate conditions on how these inversions interact with poincar transformations .
let us focus on the operation of space inversion .
if this operation is implementable by an operator @xmath204 , it must conjugate each rotation and each time translation to their respective self , and each boost and each space translation to their respective inverse .
this follows simply from the geometric meaning of space inversion .
hence , generally speaking , we need to distinguish the following three possible scenarios ( recall the notation from section[sec : dynamicalsymmetriesversuscovariance ] ) : * @xmath204 acts on @xmath27 and is a symmetry , i.e. leaves @xmath31 invariant ; * @xmath204 acts on @xmath27 and is no symmetry , i.e. leaves @xmath31 not invariant ; * @xmath204 is not implementable on @xmath27 .
it is clear that when one states that a certain equation is not symmetric under @xmath204 one usually addresses situation ( b ) , though situation ( c ) also occurs , as we shall see .
consider now the field of a massless spin-@xmath205 particle , that transforms irreducibly under the proper orthochronous poincar group .
the field is then either a two - component spinor , @xmath206 , which in the absence of interactions obeys the so - called weyl equation invariant 2-form .
an overbar denotes the map into the complex - conjugate vector space . unless stated otherwise , my conventions are those of @xcite . ]
@xmath207 alternatively , one may also start from a four - component dirac spinor , @xmath208 which carries a reducible representation of the proper orthochronous poincar group : if @xmath206 transforms with @xmath209 then @xmath210 transforms with @xmath211 ( being an element of the complex - conjugate dual space ) , so that the space of the upper two components @xmath206 of @xmath43 and the space of the lower two components @xmath210 of @xmath43 are separately invariant .
one may then eliminate two of the four components by the so - called majorana condition , which requires the state @xmath43 to be identical to its charge - conjugate , @xmath212 , where @xmath213 hence for a majorana spinor one has @xmath214 and the interaction - free dirac equation reads @xmath215 one can now either regard ( [ eq : weyleq ] ) or ( [ eq : diraceq ] ) as the interaction - free equation for a neutrino . here
i wish to briefly recall a curious discussion between pauli and fierz on whether or not these two equations describe physically different state of affairs .
superficially this discussion is about a formal and , mathematically speaking , rather trivial point .
but , as we will see , it relates to deep - lying preconceptions in pauli s thinking about issues of symmetry .
this makes it worth looking at this episode in some detail .
first note that there is an obvious bijection , @xmath1 , between two - component spinors and majorana spinors , given by @xmath216 note also that the set of majorana spinors is _ a priori _ a _ _ real _ _ vector space , though it has a complex structure , @xmath172 , given by @xmath217 with respect to which the bijection ( [ eq : spinorbijection ] ) satisfies @xmath218 , where here @xmath219 stands for the standard complex structure ( multiplication with imaginary unit @xmath219 ) in the space @xmath220 of two - component spinors
. however , regarded as a map between complex vector spaces , the bijection @xmath1 is _ not _ linear .
now , pauli observed already in 1933 ( see quotation below ) that the weyl equation ( [ eq : weyleq ] ) is not symmetric under parity .
hence he concluded it could not be used to describe nature . in fact
, what is actually the case is that parity can not even be implemented as a _ linear _ map on the space of two - component spinors ( case ( c ) above ) .
this is easy to see and in fact true for any irreducible representation of the lorentz group that stays irreducible if restricted to the rotation group ( i.e. for purely primed or purely unprimed spinors ) . as stated above , the geometric meaning of space inversion
requires that the parity operator ( if existent ) commutes with spatial rotations and conjugates boosts to their inverse .
the first requirement implies ( via schur s lemma ) that it must be a multiple of the identity in any irreducible representation that stays irreducible when restricted to the rotation subgroup , which contradicts the second requirement .
hence it can not exist in such representations , which are precisely those with only unprimed or only primed indices . ] on the other hand , the dirac equation _ is _ symmetric under space inversions .
indeed , the spinor - map corresponding to the inversion in the spatial plane perpendicular to the timelike normal @xmath13 is given by @xmath221 where @xmath222 and where @xmath223 is a complex number of unit modulus , called the _ intrinsic parity _ of the particular field @xmath43 .
it is easy to see that @xmath204 is a symmetry of ( [ eq : diraceq ] ) for any @xmath223 .
note that @xmath224 so that @xmath225 , since for spinors one only requires @xmath226 ( rather than @xmath227 ) .
it is also easy to verify that @xmath204 commutes with @xmath228 iff @xmath229 .
so if we assign imaginary parity to the majorana field , the operator @xmath204 also acts on the subspace of majorana spinors .
we conclude that the free majorana equation _ is _ parity invariant .
hence it seems at first that the weyl formulation and the majorana formulation differ since they have different symmetry properties .
but this is not true . using the bijection ( [ eq : spinorbijection ] ) , we can pull - back the parity map ( [ eq : diracparity ] ) to the space of two - component spinors , where it becomes ( now either @xmath230 or @xmath231 ) @xmath232 which is now an anti - linear map on the space of two - component spinors .
all this was essentially pointed out to pauli by markus fierz in a letter dated february 6th 1957 ( @xcite vol.iv , partiva , doc.[2494 ] , p.171 ) in connection with lee s and yang s two - component theory of the neutrino .
fierz correctly concluded from this essential equivalence that the 2-component theory as such ( i.e. without interactions ) did not warrant the conclusion of parity violation ; only interactions could be held responsible for that .
this was a relevant point in the theoretical discussion at the time , as can be seen from the fact that there were two independent papers published in _ the physical review _ shortly after fierz s private letter to pauli , containing the very same observation .
the first paper was submitted on february 13th by mclennan @xcite , the second on march 25th by case @xcite .
in fact , serpe made this observation already in 1952 @xcite and emphasised it once more in 1957 @xcite .
one might be worried about the anti - linearity of the transformation in ( [ eq : weylparity ] ) . in that respect , also following fierz , an illuminating analogy may be mentioned regarding the vacuum maxwell equations , which can be written in the form @xmath233 where @xmath234 is a complex combination of the electric and magnetic field . both equations ( [ eq : complexmaxwelleq1 ] ) are clearly equivalent of the full set of maxwell s equations .
it can be shown that spatial inversions can not be implemented as complex - linear transformations on the complex - valued field @xmath235 . )
are equivalent to @xmath236 , where @xmath237 is the unprimed spinor equivalent of the tensor @xmath238 for the electromagnetic field strength .
parity can not be linearly implemented on this purely unprimed spinor , for reasons already explained in footnote[foot : parityimplementability ] . ] but , clearly , we know that maxwell s equations are parity invariant , namely if we transform the electric field as @xmath239 ( ` polar ' vector - field ) and the magnetic field as @xmath240 ( ` axial ' vector - field ) , where @xmath241 .
this corresponds to an _
_ anti__linear symmetry of ( [ eq : complexmaxwelleq1 ] ) , given by @xmath242 .
coming back to fierz s ( and other s ) original observation for the spinor field , they were accepted without much ado by others .
for example , in her survey on the neutrino in the pauli memorial volume , madame wu states that `` it is the interaction and the interaction only that violates parity '' ( @xcite , footnote on p.270 . ) . in note 25c of that paper
she explicitly thanks fierz for `` enlightening discussions '' on the two - component theory of the neutrino .
clearly fierz expected his observation to be of interest to pauli , who had already in the 1933 first edition of his handbook article on wave mechanics propagated the view that weyl s two - component equations are _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ... not invariant under reflections ( interchange of left and right ) and , as a consequence , not applicable to the physical reality .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ but instead , pauli reacts with a surprising plethora of ridiculing remarks : wenn dieser brief ankommt ( ihren rahme ich ein ! ) , wissen sie wohl schon alles ! '' ( @xcite , vol.iv , partiva , doc.[2497 ] , p.179 ) . ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ dear mr .
your letter from the 6th is the biggest blunder you ever commited in your life ! ( probably this afternoon you will send a correction ) .
have only read the first paragraph of your letter which originated in the asylum and was shaking with laughter .
[ ... ] when this letter arrives ( yours i will frame ! )
you probably will already know everything .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ personal irritations emerged which lasted about one week through several exchanges of letters and a phone - call .
finally pauli essentially conceded fierz s point in a long letter of february 12th 1957 that also contains first hints at pauli s psychological resistances ( the emphasis is pauli s ) : , p.197 ) . ]
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ your presentation creates in me a feeling of `` formal boredom '' , to which the fusillade of laughter was of a _ compensatory _ nature . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this is a curious episode and not easy to understand .
pauli s point seems to have been that he wanted to maintain the particle - antiparticle distinction _ independently _ of parity , whereas fierz pointed out that the two - component theory provided no corresponding structural element : in weyl s form the operations @xmath228 and @xmath204 simply do not exist separately , in the majorana form @xmath204 exists and @xmath228 is the identity ( hence not distinguishing ) . psychologically speaking ,
pauli s point becomes perhaps more understandable if one takes into account the fact that since the fall of 1956 he was thinking about the question of lepton - charge conservation .
intuitively he had therefore taken as self - evident that opposite helicities also corresponded to the particle - antiparticle duality ( cf .
@xcite , vol.iv , partiva , doc.[2497 ] ) , even though this mental association did not correspond to anything in the equations . in a letter dated february 15th 1957 he offered the following in - depth psychological explanation to fierz ( the emphases are pauli s ) : - ( @xmath243 majorana @xmath244 vertauschung von elektron und positron ) invarianz ist mir auch wichtig , aber weniger wichtig als die erhaltung der leptonladung .
es ist _ sicher wahr _ , da ` mein platonischer spiegelkomplex angestochen ' war .
teilchen und antiteilchen sind das symbol fr jene allgemeine spiegelung ( wie weit sie speziell _ platonisch _
ist , dessen bin ich nicht sicher ) .
[ ... ] offenbar hat der ` spiegelungskomplex ' bei mir etwas mit tod und unsterblichkeit zu tun .
_ daher _ die angst !
wre die beziehung zwischen dem schlafenden spiegelbild und dem wachenden gestrt , oder wren sie gar identisch ( majorana ) , so gbe es , psychologisch gesprochen , weder leben ( geburt ) noch tod . ''
( @xcite , vol.iv , partiva , doc.[2517 ] , p.225 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ well , the fusillade of laughter occurred with the expression `` majorana theory '' of your first letter .
after this catchword i could not go on reading .
the immediate association with majorana clearly has been this : `` aha , particles and antiparticles should no longer exist , these one intends to take away from me ( as one takes away a symbol from somebody ) ! ''
this causes me _
anxiety_. i also know that since last fall the conservation of lepton charge in physics was tremendously important to me looked upon rationally probably _ too important_. i am _ anxious
_ it could turn out to be incorrect and , psychologically speaking , `` discontentedness '' is a euphemism for anxiety . the @xmath245 ( @xmath243 majorana @xmath244 exchange between electron and positron ) invariance is also important to me , but less so than the conservation of lepton charge .
it is _ certainly true _ that it `` hit upon my platonic mirror complex '' .
particles and antiparticles are the symbol for that more general mirroring ( i am not sure to what extent it is particularly _ platonic _ ) .
[ ... ] mirroring is also a _ gnostic symbol _ for life and death .
there light is _ extinguished _ at birth and _ lightened up _ at death . [ .. ] obviously , for me the `` mirroring complex '' has something to do with death and immortality . _
hence _ the anxiety ! if the relation between the sleeping mirror image and the one awake would be disturbed , or if they would even be identical ( majorana ) , then , psychologically speaking , there would neither be life ( birth ) nor death .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ fierz later commented on that episode in a personal letter to norbert straumann , parts of which are quoted in @xcite . in 1955 a collection of essays by distinguished physicists appeared to celebrate niels bohr s 70th birthdays @xcite .
pauli s contribution ( @xcite , p.30 - 51 ) is entitled `` exclusion principle , lorentz group and reflection of space - time and charge '' , whose introduction contains the following remarks : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ after a brief period of spiritual and human confusion , caused by provisional restriction to `` anschaulichkeit '' , a general agreement was reached following the substitution of abstract mathematical symbols , as for instance psi , for concrete pictures .
especially the concrete picture of rotation has been replaced by mathematical characteristics of the representations of the group of rotations in three dimensional space .
this group was soon amplified to the lorentz group in the work of dirac . [ ... ] the mathematical group was further amplified by including the reflections of space and time .
[ ... ] i believe that this paper also illustrates the fact that a rigorous mathematical formalism and epistemological analysis are both indispensable in physics in a complementary way in the sense of niels bohr .
while i try to use the former to connect all mentioned features of the theory with help of a richer `` fullness '' of plus and minus signs in an increasing `` clarity '' , the latter makes me aware that the final `` truth '' on the subject is still `` dwelling in the abyss '' . _ _
( @xcite , p.30 - 31 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in some sense this paper of pauli s can be seen as a follow - up to his spin - statistics paper already discussed above , the main difference being that pauli now considers _ interacting _ fields .
pauli now _ assumes _ ( 1 ) the validity of the spin - statistics correlation for interacting fields ( for which there was no proof at the time ) , ( 2 ) invariance under ( the universal cover of ) the proper orthochronous lorentz group @xmath122 ( as in the spin - statistics paper ) , and ( 3 ) locality of the interactions ( i.e. involving only finitely many derivatives ) .
then pauli shows that this suffices to derive the so - called cpt theorem that states that the combination of charge conjugation ( c ) and spacetime reflection ( pt ) is a symmetry . at the time ( 1955 )
pauli wrote his paper it was not known whether any of the operations of @xmath246 , @xmath247 , or @xmath59 would separately _ not _ be a symmetry .
this changed when in january 1957 through the experiments of madame wu _ et al .
_ , in which explicit violations of @xmath247 and @xmath246 were seen in processes of beta - decay , following a suggestion that this should be checked by lee and yang in mid 1956 @xcite .
pauli had still offered a bet that this would not happen on january 17th 1957 ( the emphases are pauli s ) : , p.82 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ i do _ not _ believe _ that god is a weak left - hander _ and would be prepared to bet a high amount that the experiment will show a symmetric angular distribution of the electrons ( mirror symmetry ) . for i can not see a logical connection between the _ strength _ of an interaction and its mirror symmetry .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in view of this firm belief in symmetry the following is remarkable : in his cpt paper pauli takes great care to writes down the most general ultralocal ( i.e. no derivatives ) four - fermion interaction ( for the neutron , proton , electron and neutrino ) , which is _ not _
@xmath247 invariant . in contains 10 essentially different terms with ten coupling constants @xmath248 , only the first five of which are parity invariant ( scalars ) , whereas the other five are pseudoscalars , i.e change sign under spatial inversions .
apparently this he did just for the sake of mathematical generality without any physical motivation , as he explicitly stated in a letter to madame wu dated january 19th 1957 ( the emphases are pauli s ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ when i considered such formal possibilities in my paper in the bohr - festival volume ( 1955 ) , i did not think that this could have something to do with nature .
i considered it merely as a mathematical play , and , as a matter of fact , i did not believe in it when i read the paper of yang and lee .
[ ... ] what prevented me _ until now _ from accepting this formal possibility is the question why this restriction of mirroring appears only in the ` weak ' interactions , not in the strong ones . _
theoretically _ , i do not see any interpretation of this fact , which is empirically so well established . _
( @xcite , vol.iv , partiva , doc.[2460 ] , p.89 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ lee and yang took this possibility more serious : in an appendix to their paper they also write down all ten terms for the full , parity non - invariant interaction ( @xcite , p.258 ) , without any citation of pauli .
pauli first learnt that the experiments by madame wu _ et al .
_ had led to an asymmetric angular distribution from a letter by john blatt from princeton , dated january 15th 1957 .
there blatt writes : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ i do nt know whether anyone has written you as yet about the sudden death of parity .
miss wu has done an experiment with beta - decay of oriented co nuclei which shows that parity is _ not _ conserved in @xmath1 decay .
[ ... ] we are all rather shaken by by the death of our well - beloved friend , parity . _ ( @xcite , vol.iv , partiva , doc.[2451 ] , p.74 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ pauli , too , was shocked as he stated in his famous letter to weisskopf dated january27/28 1957 ( @xcite , vol.iv , partiva , doc.[2476 ] ) . in that very same letter
pauli already started speculating how symmetry could be restored by letting the constants @xmath249 become dynamical field , scalar fields for @xmath250 and pseudo - scalar ones for @xmath251 : mit einem skalarfeld @xmath252 , die terme mit @xmath253 mit einem pseudo - skalarfeld @xmath254 multipliziert .
fr den herrgott , der das vorzeichen von @xmath254 umdrehen kann , wre eine solche theorie natrlich rechts - links - invariant nicht aber fr uns sterbliche menschen , die wir gar nichts wissen ber jenes hypothetische neue feld , auer da es praktisch auf der erde raum - zeitlich konstant ( statisch - homogen ) ist , und die wir noch kein mittel haben , es zu ndern . ''
( @xcite , vol.iv , partiva , doc.[2476 ] , pp.122 - 123 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ let us imagine , for example , the terms with @xmath255 being multiplied with a scalar field @xmath252 , the terms @xmath253 multiplied with a pseudo - scalar field @xmath254 . for god himself , who can change the sign of @xmath254 , such a theory would be left - right - invariant not for us mortal men , however , who do not know anything about that new hypothetical field , except that it is practically constant in space and time on earth ( static - homogeneous ) , and that we do not yet have any means to change it . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the mechanism envisaged here to restore symmetry is just that discussed in section[sec : dynamicalsymmetriesversuscovariance ] , where non - dynamical backgrounds structures , @xmath22 , are ( formally ) turned into dynamical quantities , @xmath23 . as already mentioned , since fall of 1956 pauli s thinking about beta - decay was dominated by the lepton - charge conservation . in a paper submitted on march 14th 1957 , entitled _ on the conservation of lepton charge _
( @xcite , vol.2 , pp.1338 - 1349 ) , pauli once more showed his mastery of symmetry considerations while keeping everything at the largest possible degree of generality .
he starts by considering the most general ultralocal four - fermion interactions ( not necessarily preserving parity or lepton charge ) in which the neutrino field is represented by a dirac 4-spinor , @xmath43 . for what follows it
is convenient to think of the four components of @xmath43 as comprising the following four particle states ( per momentum ) : a left - handed neutrino , @xmath256 , a right - handed neutrino , @xmath257 , and their antiparticles @xmath258 and @xmath259 respectively .
note that this means @xmath260 and that accordingly @xmath258 is right- and @xmath259 is left - handed . here
we follow the convention of @xcite .
next pauli considers a four - parameter group of canonical transformations ( i.e. they leave the anticommutation relations between the fermion fields invariant ) of the neutrino field , henceforth called the _ pauli group _ , whose interpretation will be given below .
these transformations define a symmetry of the interaction - free equations of motions ( assuming a massless neutrino throughout ) , but will generally _ not _ define a symmetry once the interaction is taken into account . rather , the following is true ( cf .
@xcite ) : suppose that the general interaction depends on a finite number of coupling constants @xmath261 for @xmath262 and that the equations of motion follow from an action principle with lagrange density @xmath263 , where @xmath22 represents the array of coupling constants ( we notationally ignore other non - dynamical structures here ) and @xmath23 the dynamical fields .
then the pauli group acts as covariance in a slightly stronger sense than ( [ eq : defcovariance2 ] ) , namely so that @xmath264 this means that on the level of the lagrange density ( or the hamiltonian ) , and hence in particular at the level of the equations of motion , the transformation of the dynamical fields can be compensated for by a transformation of the coupling constants . a large part of pauli s paper is actually devoted to the determination of that compensating action of the pauli group on the array of coupling constants .
next suppose the initial state is chosen to be invariant under the pauli group , i.e. @xmath265 for all @xmath35 .
then ( [ eq : paulisymmetry ] ) implies that its evolution with interaction parametrised by @xmath22 ( the array of @xmath261 s ) is identical to the evolution parametrised by @xmath55 for any @xmath35 .
hence the outcome of the evolution can only depend on the @xmath261 s through their pauli - invariant combinations .
in particular , since the neutrinoless double beta - decay simply has no initial neutrino , this reasoning can be applied to it .
if this lepton - charge - conservation violating process is deemed impossible , the corresponding pauli - invariant combination of coupling constants to which the scattering probability is proportional must vanish .
this , in turn , gives the sought - after constraint on the possible four - fermion interaction . for ( massless ) neutrinos in majorana representation pauli finally arrived at the result that either only the left- or the right - handed component enters the interaction .
it should be added that this clever sort of reasoning was shortly before used by pursey in a less general setting @xcite in which the interaction was specialised _ a priori _ to conserve lepton charge . part . ]
more on the history of the search for the right form of the four - fermion interaction may be found in @xcite .
it should also be mentioned that the possibility of neutrinoless double beta - decays is currently still under active experimental investigation at the national gran sasso laboratory , where the 2003 - 2005 cuoricino experiment set upper bounds for the majorana mass of the electron neutrino well below one @xmath266 .
the upcoming next - generation experiment , cuore , is designed to lower this bound to @xmath267 ; compare @xcite .
what is the interpretation of the pauli group ?
mathematically it is isomorphic to @xmath268 , the group of @xmath269 unitary matrices acting on a two - dimensional complex vector space .
here there are two such spaces ( per 4-momentum ) in which it acts : the ` left - handed subspace ' that is spanned by the two left - handed components @xmath256 and @xmath259 , and the ` right - handed subspace ' that is spanned by the two right - handed components @xmath257 and @xmath258 .
the two actions of @xmath268 in these spaces are complex conjugate to each other ( see equation ( [ eq : pauligroupaction ] ) ) .
usually one thinks of the pauli group as @xmath270 , which is a double cover of @xmath268 , so that the four real parameters are written as a phase @xmath271 , parametrising @xmath272 , and two complex parameters @xmath273 satisfying @xmath274 , which give three real parameters when split into real and imaginary part and which parametrise a 3-sphere that underlies @xmath275 as group manifold . in this parametrisation the action of the pauli group reads ( an asterisk stands for complex conjugation ) : as @xmath276 , where @xmath277 is the charge conjugate field .
but this is easily seen to be equivalent to ( [ eq : pauligroupaction ] ) if one sets @xmath278 and @xmath279 .
the more explicit form ( [ eq : pauligroupaction ] ) is better suited for the interpretational discussion ; cf .
the two - to - one homomorphism from @xmath270 to @xmath268 is given by @xmath280 whose kernel is @xmath281 . ]
[ eq : pauligroupaction ] @xmath282 invariance under the pauli group is now seen to correspond to an ambiguity in the particle - antiparticle distinction .
this ambiguity would only be lifted by interactions that allowed to distinguish the two left and the two right states respectively . in the absence of such interactions the various definitions of ` particle ' and ` antiparticle ' are physically indistinguishable , so that the pauli group acts by _ gauge _ symmetries in the sense of section[sec : physicalversusgaugesymmetries ] .
also , the different presentations of the two - component theory , already discussed in section[sec : irritations ] can be seen here .
the majorana condition reads @xmath283 , which in terms of the four components introduced above leads to @xmath284 and @xmath285 .
this can be read in two different ways , depending on whether one addresses @xmath286 or @xmath287 as independent basic states .
in the first case one would say that there is a left - handed neutrino and its right - handed antiparticle , whereas in the second case one regards the tuple @xmath288 as respectively the left- and right - handed components of a single particle which is identical to its antiparticle . beyond weak interaction and beta - decay , the pauli group played a very important rle in pauli s brief participation in heisenberg s programme for a unified field theory .
it was pauli who first showed that the ( so far classical ) non - linear spinor equation proposed by heisenberg was invariant under the pauli group ( cf .
heisenberg s account in his letter to zimmermann from jan.7th 1958 in @xcite , vol.iv , partivb , p.779 ) . in this new context the @xmath272 part of the pauli group was connected to conservation of baryon charge and
the @xmath275 part acquired the meaning of isospin symmetry .
the central importance of isospin for this programme may already be inferred from the title of the proposed common publication by heisenberg and pauli , which reads : _ on the isospin group in the theory of elementary particles_. however , due to pauli s later retreat from this programme , the manuscript ( cf .
@xcite , vol.iv , partivb , pp.849 - 861 ) for this publication never grew beyond the stage of a preprint . in his last paper on the subject of discrete symmetries , entitled _ _ the violation of mirror - symmetries in the laws of atomic physics _ _
( @xcite , vol.2 , pp.1368 - 1372 ) , pauli comes back to the question which bothered him most : how is the strength of an interaction related to its symmetry properties ? he says that having established a violation of @xmath246 and @xmath247 symmetry for weak interactions , we may ask why they are maintained for strong and electromagnetic interactions , and whether the reason for this is to be found in particular properties of these interactions .
he ends with some speculations on possible connections between violations of @xmath246 and @xmath247 symmetry in the laws of microphysics on one hand , and properties of theories of gravitation and its cosmological solutions on the other : um bei der frage des zusammenhangs zwischen dem kleinen und dem grossen ber vage spekulationen hinauszugelangen , fehlen daher noch wesentlich neue ideen .
hiermit soll jedoch nicht die unmglichkeit eines solchen zusammenhanges bestimmt behauptet werden . ''
( @xcite , vol.2 , p.1371 ) ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ second , one can try to find and justify a connection between symmetry violation in the small with properties of the universe in the large .
but this exceeds the capabilities of the presently known theory of gravity . [ ... ] new ideas are missing to go beyond vague speculations .
but this shall not be taken as definite expression for the impossibility of such a connection .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it may be of interest to contrast this expression of a certain open - mindedness for speculations concerning the physics of elementary particles on one side and large - scale cosmology on the other , with a more critical attitude from pauli s very early writings . in section65 of his relativity article , where pauli discussed weyl s attempt for a unifying theory of gravity and electromagnetism ( to which pauli himself actively contributed )
, he observes that in weyl s theory ( as well as in einstein s own attempts from that time ) it is natural to suspect a relation between the size of the electron and the size ( mean curvature radius ) of the universe .
but then he comments somewhat dismissively that this _ _ might seem somewhat fantastic__(@xcite , p.202 ) .
i have tried to display some of the aspects of the notion of symmetry in the work of wolfgang pauli which to me seem sufficiently interesting in their own right . in doing this
i have drawn freely from pauli s scientific vre , irrespectively of whether the particular part is commonly regarded as established part of present - day scientific knowledge or not .
pauli s faith in the explanatory power of symmetry principles clearly shows up in all corner of his vre , but it also appears clearly rooted beyond the limits of his science . in the editorial epilogue to the monumental collection of pauli s scientific correspondence , karl von meyenn reports that many physicists he talked to at the outset of his project spoke against the publication of those letters that contained ideas which did not stand the test of time ( @xcite , vol.iv , partivb , p.1375 ) . leaving aside that this must clearly sound outrageous to the historian ,
it is , in my opinion , also totally misguided as far as the scientific endeavour is concerned .
science is not only driven by the urge to _ know _ but also , and perhaps most importantly , by the urge to _
understand_. no one who as ever actively participated in science can deny that .
one central aspect of scientific understanding , next to offering as many as possible alternative and complementary explanations for the actual occurrences in nature , is to comprehend why things could not be different from what they appear to be . the insight into a theoretical or an explanatory failure can be as fruitful as an experimental failure .
what makes pauli a great scientist , amongst the other most obvious reasons , is not that he did not err such mortals clearly do not exist , but that we can still learn much from where he erred and how he erred . in that sense , let me end by the following words from johann wolfgang von goethe s _ maximen und reflexionen _ ( # 1292 ) : * acknowledgements : * i sincerely thank harald atmanspacher and hans primas for the invitation to talk at the conference on _ wolfgang pauli s philosophical ideas and contemporary science _ on the monte verita in ascona , switzerland .
i also thank norbert straumann for comments and suggestions for improvement .
finally i wish to express my strongest appreciation to the editor of pauli s scientific correspondence , karl von meyenn , without whose admirable editorial work we would not be in the position to share many of wolfgang pauli s wonderful insights . thank you ! domenico giulini .
algebraic and geometric structures in special relativity . in claus lmmerzahl and jrgen ehlers , editors ,
_ special relativity : will it survive the next 101 years ? _ , volume 702 of _ lecture notes in physics _ , pages 45111 .
springer verlag , berlin , 2006 .
online available at @xmath291arxiv.org / pdf / math - ph/0602018@xmath292 .
domenico giulini . some remarks on the notions of general covariance and background independence . in erhard seiler and ion - olimpiu stamatescu ,
editors , _ approaches to fundamental physics _ , volume 721 of _ lecture notes in physics _ , pages 105120 .
springer , berlin , 2007 .
online available at @xmath291arxiv.org / pdf / gr - qc/0603087@xmath292 .
hendrik antoon lorentz . , volume 1 of _ beihefte zur zeitschrift fr mathematischen und naturwissenschaftlichen unterricht , w.lietzmann and e. grimsehl ( editors)_. b.g
. teubner , leipzig and berlin , 1914 . three lectures given at teyler s foundation at haarlem , edited by w.h.keesom .
karl von meyenn .
pauli s belief in exact symmetries . in manuel doncel , armin hermann , louis michel , and abraham pais , editors , _ symmetries in physics ( 1600 - 1980 ) _ , pages 331360 , barcelona , 1987 .
servei de publicacions , universitat autnoma de barcelona .
proceedings of the 1st international meeting on the history of scientific ideas held at sant feliu de guixols , catalonia , spain , september 20 - 26 , 1983 .
wolfgang pauli .
continuous groups in quantum mechanics . in g.
hhler , s. flgge , f. hund , and f. trendelenburg , editors , _ ergebnisse der exakten naturwissenschaften _ , volume 37 , pages 85104 .
springer verlag , berlin , 1965 .
first published as cern - report 56 - 31 .
wolfgang pauli . .
springer verlag , berlin , 2000 .
reprint of the original ` encyclopdie - artikel ' from 1921 with various additions , including pauli s own supplementary notes from 1956 . edited and annotated by domenico giulini . | `` symmetry '' was one of the most important methodological themes in 20th - century physics and is probably going to play no lesser role in physics of the 21st century .
as used today , there are a variety of interpretations of this term , which differ in meaning as well as their mathematical consequences .
symmetries of crystals , for example , generally express a different kind of invariance than gauge symmetries , though in specific situations the distinctions may become quite subtle .
i will review some of the various notions of `` symmetry '' and highlight some of their uses in specific examples taken from pauli s scientific vre .
this paper is based on a talk given at the conference _
wolfgang pauli s philosophical ideas and contemporary science _ , may 20.-25 .
2007 , at monte verita , ascona , switzerland
. 10000 |
the transport properties of quark - gluon plasma ( qgp ) have attracted a lot of interest since nearly perfect liquid behavior has been discovered in the relativistic heavy - ion collision experiments at the relativistic heavy ion collider ( rhic ) @xcite .
further investigations in the heavy - ion collisions at higher energy are ongoing at the large hadron collider ( lhc ) .
the discovery of nearly perfect liquid behavior seems to contradict the notion of weakly interacting qgp , and it rather suggests a strongly interacting nature of the qgp . indeed , the universal lower bound of the ratio of shear viscosity to entropy density @xmath3 is proposed in the strongly interacting quantum field theories based on the conjectured duality between the gauge theory and string theory , or the anti - de - sitter space and conformal field theory ( ads / cft ) correspondence @xcite .
this is close to the value @xmath4 , which is phenomenologically extracted from the experimental data by hydrodynamic simulations for the heavy - ion collisions @xcite . in the heavy - ion collisions , not only the bulk collective dynamics but also hard probes , such as heavy quarks and jets ,
reveal independent aspects of the transport properties of the qgp . for example , medium modification of heavy quark momentum spectra offers an opportunity to study drag force of the qgp acting on a test particle with color @xcite .
phenomenological studies of the heavy quark probe in the heavy - ion collisions are summarized in ref .
@xcite . since there is a kinematical hierarchy between the heavy quark and the qgp , the heavy quark dynamics is slow compared to the correlation time of matter constituents so that the effects of interaction between them are averaged out .
this enables one to use a simple effective description of the heavy quark using the drag and fluctuation forces .
there have been various efforts to calculate the heavy quark transport coefficients by the perturbation theory @xcite , by the lattice qcd simulations @xcite , and by applying the ads / cft correspondence @xcite . in this paper ,
i shed light on the dynamics in the heavy quark color space .
the langevin dynamics is a classical description of the brownian motion . unlike kinetic variables ,
the time scale of the heavy quark color does not get long even with its heavy mass . typically , the time scale of color diffusion is about @xmath5 while the duration of a soft scattering is about @xmath6 .
therefore , it is only in the weak coupling regime that one can expect the validity of langevin description that couples with the heavy quark color degrees of freedom . in the case of the heavy quark with color , description in the phase space is classical while that in the color space is quantum . here
the heavy quark color is in the fundamental representation of color @xmath7 . instead of describing it classically by wong s equation @xcite in terms of a @xmath8-dimensional vector ,
i treat a color state as a quantum state in a hilbert space with @xmath9 dimensions . indeed , the latter description is a direct consequence of the quantum description of the heavy quark brownian motion @xcite .
the main finding of this paper is that an analog of the schrdinger s cat state appears in the heavy quark langevin dynamics due to the non - abelian interaction of qcd .
the classical momentum corresponds to the cat and couples to quantum states in the color space .
it is well known that decoherence of such a macroscopic superposition state is essential to have a classical picture of the brownian motion @xcite .
consequently , the discretization time scale of the classical langevin dynamics must come out from the time scale of decoherence in the heavy quark sector .
the discretization time scale turns out to be @xmath10 .
it depends on the heavy quark mass @xmath1 and on the heavy quark momentum diffusion constant @xmath2 ( up to some factors ) .
phenomenological implication of the discretization time scale @xmath10 is intriguing .
typically , the drag force extracted from the experimental data corresponds to @xmath11 @xcite . for a charm ( bottom )
quark in the qgp with @xmath12 , the discretization time scale is @xmath13 . in the heavy - ion collisions , typical lifetime of the qgp
is @xmath14 .
therefore the values of @xmath15 might indicate that the macroscopic superposition state remains , at least partially , until the freeze - out stage of the heavy - ion collisions .
also the heavy quark hadronization time scale @xmath16 is shorter than @xmath15 and thus in any case the freezeout process may be able to resolve the macroscopic superposition state .
this would enhance the possibility of recombining heavy quark@xmath17antiquark pairs into heavy quarkonia and might support the statistical hadronization models @xcite when heavy quarks are produced in abundance .
this paper is organized as follows . in sec .
[ sec : hq_open_system ] , i introduce and analyze the basic properties of the heavy quark master equation .
one can see that ehrenfest equations show kinetic equilibration of heavy quarks . in sec .
[ sec : derivation_langevin ] , i derive the langevin equation of heavy quarks with color .
the langevin equation is not closed in the heavy quark phase space : the momentum update depends on the heavy quark color state . in sec .
[ sec : langevin_decoherence ] , i discuss how to interpret the coupling between the dynamics in the heavy quark phase space and that in the color space . also the appropriate discretization time scale for classical langevin description is discussed .
section [ sec : summary ] is devoted to a summary . throughout this paper ,
i adopt the natural units , @xmath18 , and operators in hilbert and fock spaces are denoted by bold fonts .
when one describes a heavy quark in the quark - gluon plasma ( qgp ) as an open quantum system @xcite , reduced density matrix @xmath19 is a basic dynamical quantity . here @xmath20 denotes a density matrix of a total system of the heavy quark and the qgp .
the total hilbert space is composed of a direct product of the heavy quark and the qgp hilbert spaces @xmath21 . by taking the trace over the hilbert space for the qgp ( @xmath22 ) ,
one gets the reduced density matrix @xmath23 that operates in the heavy quark hilbert space .
the master equation for the reduced density matrix @xmath24 was derived by the influence functional formalism @xcite . in the leading order of the qcd coupling @xmath25 and in the nonrelativistic limit , in ref .
it is @xmath26 here @xmath27_{ij } = \langle \vec x , i|\bm\rho_{\rm q}(t)| \vec y , j\rangle$ ] is a reduced density matrix in the position and color spaces .
the matrix @xmath28 is in the fundamental representation of the color @xmath7 algebra and @xmath29 . @xmath30 and @xmath31
are given in terms of a function @xmath32 as @xmath33 here @xmath32 is defined by a thermal two - point function of gluons @xmath34 as is clear from the definition , @xmath32 is an even function of @xmath35 . at @xmath36 , the hard thermal loop resummed perturbation theory gives at leading order @xmath37 here
@xmath38 is the debye screening mass @xmath39 for qcd with @xmath40 light flavors . from the master equation
, one can derive time - evolution of thermal averaged quantum expectation values @xmath41 , or the ehrenfest equations , such as @xmath42 reproducing the _ consequences _ of classical langevin dynamics ( up to @xmath43 ) .
the drag force parameter @xmath44 and the correction to equipartition @xmath43 @xmath45 are given by @xmath46 the master equation has several important properties , which i summarize here .
the terms with @xmath47 describe thermal fluctuation and that with @xmath48 describes dissipation . in the ehrenfest equations and ,
the thermal fluctuation @xmath49 derives from @xmath47 whereas the damping terms @xmath50 and @xmath51 are from @xmath48 . by taking the trace in the color space , the color averaged master equation for @xmath52
is derived as @xmath53 close to equilibrium , thermal de broglie wavelength for a heavy quark @xmath54 is much shorter than typical correlation length of ( electric ) gluons @xmath55 , where for @xmath56 , @xmath57 .
therefore one can take a short distance approximation for @xmath58 . in this approximation , @xmath59 . by this expansion ,
one gets the caldeira - leggett master equation @xcite . in the following ,
i derive the langevin equation itself , instead of being satisfied with reproducing the @xmath60 of its dynamics .
first , let us investigate more closely how the @xmath47 terms in the master equation are related to the thermal fluctuation .
suppose an infinitesimal time step @xmath61 under a random background @xmath62 , which rotates a heavy quark color state by @xmath63 $ ] .
the following statistical property of the random field is assumed : @xmath64 hereafter , i call average over the random field @xmath65 the noise average and denote it as @xmath66
. then the density matrix in the random field , which i denote as @xmath67 , evolves as @xmath68^{\dagger}.\end{aligned}\ ] ] noting that @xmath69 , @xmath70 can be expanded as @xmath71 ^ 2 + \mathcal o(dt^{3/2 } ) \nonumber \\ & \simeq&1-idt\xi^a(t,\vec x ) \hat t^a-\frac{dt}{2}c_{\rm f}f_1(\vec 0).\end{aligned}\ ] ] in the second line of eq . , i approximate @xmath72 ^ 2\simeq dt^2\langle\left[\xi^a(t,\vec
x ) \hat t^a\right]^2\rangle_{\xi}=dtc_{\rm f}f_1(\vec 0)$ ] because the fluctuation around the average is @xmath73 .
such a fluctuation can be neglected , for it does not contribute to the master equation after taking the noise average .
therefore one can always substitute @xmath74 .
the evolution equation then becomes @xmath75 .
\nonumber\end{aligned}\ ] ] taking the noise average for eq . and interpreting the noise - averaged density matrix @xmath76 as the original density matrix @xmath77
, one can reproduce the thermal fluctuation terms ( @xmath47 terms ) in the master equation . adding the kinetic term and the dissipation terms ( @xmath48 terms ) to eq .
, i derive the following stochastic master equation : @xmath78.\nonumber\end{aligned}\ ] ] although i construct the stochastic master equation quite intuitively , there also exists a formal derivation .
one can formulate the problem using the influence functional formalism @xcite .
following the notation of ref .
@xcite , the influence functional @xmath79 can be obtained as an expansion @xmath80 .
the thermal fluctuation @xmath81 and a part of the dissipation @xmath82 can be expressed using a stochastic variable equivalent to @xmath83 here .
see appendix [ sec : thermal_fluctuation ] for details
. using the stochastic master equation , one can derive stochastic ehrenfest equations .
since @xmath84 depends on the random field @xmath83 , it corresponds to quantum expectation value in a particular realization of @xmath83 , which i denote as @xmath85 or simply as @xmath86 if there is no confusion .
the relation to the thermal average of quantum expectation value is @xmath87\rangle_{\xi}$ ] .
the following stochastic ehrenfest equations are derived : @xmath88 with @xmath89 in the derivation , i assume that the heavy quark is localized as a wave packet . to be explicit , the spatial extension of the wave packet should be much smaller than the correlation length of gluons @xmath90 .
this is the classical ( point particle ) limit for the heavy quark .
in such a limit , it is natural to assume that the density matrix is factorized in the color and the configuration spaces @xmath91 .
, it is defined only in the @xmath92 representation . ] in terms of @xmath93 , @xmath94 .
the random force @xmath95 is the force in the random potential @xmath83 evaluated at the position of wave packet @xmath96 : @xmath97 the random force strength follows from the definitions and . unlike the conventional langevin equation , the ones derived here in eqs . and are not in a closed form .
one needs to know the dynamics of @xmath98 . using the same assumption
as made for @xmath99 , it is derived as @xmath100 with @xmath101 here @xmath102 is the random field evaluated at the position of the wave packet @xmath103 .
note also that @xmath104 which follows from the definitions of @xmath95 and @xmath102 .
the parameter is @xmath105 and @xmath106 is structure constant of the color @xmath7 algebra .
hereafter , i write @xmath107 instead of @xmath108 .
the stochastic equation has an interesting property : it conserves @xmath109 . if and only if @xmath110 is a pure state density matrix , @xmath111 .
therefore @xmath110 stays a pure state density matrix if it was initially .
this feature can also be understood concisely by reconstructing @xmath112_{ij}\equiv\langle i|{\bm\rho}_{\rm color}(t;\zeta)|j\rangle$ ] with @xmath113 \hat t^a.\end{aligned}\ ] ] equation is equivalent to @xmath114\nonumber\\ & & - \ i\left [ \zeta^a(t)\hat t^a,\hat\rho_{\rm color}(t;\zeta ) \right].\end{aligned}\ ] ] this color space master equation is very similar to eq . with substitution @xmath115 .
therefore eq . can be derived from random rotation in the color space @xmath116^{\dagger}$ ] with @xmath117 $ ] .
since the random rotation @xmath118 is a unitary evolution , one obtains @xmath119 ^ 2\right\}={\rm tr}_{\rm color}\left\{\left[{\bm \rho}_{\rm color}(t;\zeta)\right]^2\right\}$ ] , which gives the conservation of @xmath109 . note that @xmath120 ^ 2\right\}\leq { \rm tr}_{\rm
color}\left\{{\bm \rho}_{\rm color}(t;\zeta)\right\}=1 $ ] sets the upper limit of @xmath109 . because of the conservation of @xmath120 ^ 2\right\}$ ]
, it is natural to demand ( only initially ) that @xmath107 be a pure state density matrix .
let us summarize what i have obtained so far . taking the classical point particle limit for the heavy quark ,
i have derived the coupled langevin equations in the phase space and in the color space : @xmath121 the statistical properties of the noises are @xmath122 the initial condition for @xmath98 should satisfy the pure state constraint @xmath123 . in the above equations ,
the noises @xmath95 and @xmath102 derive from @xmath83 .
therefore i still use the notation of @xmath66 for taking the noise average over @xmath95 and @xmath102 .
typical magnitudes of the parameters are @xmath124 physically , the drag parameter @xmath2 comes from scatterings with both soft and hard momentum exchanges while @xmath125 is dominated by soft scatterings that rotate the heavy quark color .
having derived the langevin equation , let us investigate its property in more detail .
hereafter , i regard @xmath127 and @xmath128 as classical variables @xmath129 and @xmath130 . it is not trivial whether to identify @xmath131 as a classical variable @xmath132 satisfying the constraint @xmath133 .
note that @xmath132 is now a ( @xmath134)-dimensional vector and is different from a matrix @xmath28 .
actually , this identification contradicts the equipartition derived from the master equation because it results in @xmath135 with @xmath136 .
therefore with the identification @xmath137 as a classical variable , the heavy quarks can not get thermalized correctly .
one easy solution , which turns out to be correct , is to impose a constraint @xmath138 , independently of the identification @xmath139 .
the problem of this quick solution is that the connection to @xmath107 is seemingly lost .
i will see how to interpret the constraint @xmath140 . to reach a deeper understanding ,
let us recall the physical meaning of @xmath141 . by definition
, @xmath141 is a quantum expectation value of the heavy quark color charge and it couples to the time evolution of _ classical _ or _ macroscopic _ variable @xmath129 .
this indicates that @xmath129 can be regarded as a macroscopic variable in the so - called _
cat state_. to determine the momentum update , the actual value for @xmath142 should be the one that is as if measured .
otherwise , the noise term @xmath142 only describes the thermal fluctuation and lacks the quantum fluctuation in the color space . in order to take into account the quantum fluctuation ,
i introduce a notation @xmath143_{\rm meas}$ ] which takes one of the possible values of the observable @xmath144 in a measurement with respect to a pure state @xmath107 . here
i examine how @xmath145 is measured .
first , let us parametrize @xmath146 with normalization @xmath147 . to be consistent with the statistical property of @xmath95 , the independent white noises @xmath148 and @xmath149 must satisfy @xmath150 and @xmath151 . then @xmath152_{\rm meas}$ ] takes values as @xmath153_{\rm meas } = \vec f(t)\times \left\{\frac{1}{2 } , 0 , \cdots , 0 , -\frac{1}{2 } \right\}.\end{aligned}\ ] ] there are @xmath154 zero modes for @xmath145 in the fundamental representation .
the probability that a pure state @xmath107 shrinks to each eigenstate depends on the details of both the pure state and the eigenstate .
the quantum fluctuation can also be described using the wigner function .
see appendix [ sec : wigner ] for a brief summary . by interpreting the momentum update as @xmath155_{\rm meas},\end{aligned}\ ] ] one can take into account the quantum fluctuation in the color space . in this interpretation , the noise strength due to the quantum fluctuation is evaluated as @xmath156 together with the thermal fluctuation , the noise strength becomes @xmath157 and the equipartition of heavy quarks is realized correctly . in an original derivation ,
@xmath158 with @xmath159 being a _ mathematical _ discretization time scale . the discretization time scale @xmath159 is shorter than any physical time scales of the heavy quark and is thus sent to @xmath160 .
one will see , however , that there is an appropriate discretization time scale @xmath15 at which the measurement " makes sense .
therefore , physically , one should take @xmath161 . in the above
, the factorization of @xmath146 is essential for the simultaneous measurements of @xmath162 , @xmath163 , and @xmath164 . in general
, @xmath165 does not factorize in such a way .
however , i need to assume the factorization only at the time scale @xmath15 , during which , as one will see shortly , several soft scatterings take place .
therefore , the factorization is assumed only for the statistical property of the total momentum kick accumulated during @xmath15 .
[ fig : decoherence ] here i show that the quantum fluctuation of @xmath145 becomes relevant as a consequence of decoherence . for a moment , let us get back to the quantum master equation for the heavy quark .
the second line of eq .
describes decoherence of heavy quark wave function . when the heavy quark wave function is localized compared to the debye screening length @xmath166 , @xmath167 can be taylor expanded and the decoherence term can be approximated as @xmath168 this term tells us that a wave function of size @xmath169 loses coherence in a time scale @xmath170 .
if the reduced planck constant @xmath171 is recovered , @xmath172 indicating the quantum nature of decoherence .
let us apply this decoherence time formula to the case of the decoherence of the macroscopic superposition in the momentum space .
the decoherence time gives an estimate for an appropriate time step @xmath173 in the classical langevin dynamics .
figure [ fig : decoherence ] depicts hierarchy of the length scales i consider . when the momentum difference of the wave packets in the macroscopic superposition state is @xmath174 , it propagates to the distance between them @xmath175 in time @xmath15 .
the momentum difference due to the noise is @xmath176 .
therefore i get @xmath177 for the distance between the wave packets . as the time step @xmath15 lengthens ,
the distance between them @xmath178 increases and thus the decoherence time @xmath179 decreases .
the time step @xmath15 of the classical langevin dynamics should be comparable to the decoherence time scale @xmath179 during which the quantum interference between the wave packets is destroyed .
equating @xmath180 , i obtain @xmath181 with explicit dependence on @xmath171 , one can see the quantum origin of the time scale @xmath182 . in this time scale @xmath15
, the decoherence turns the quantum superposition in the schrdinger s cat state into a probabilistic mixture of the classical states .
there is an alternative view to the time scale @xmath15 from the uncertainty principle @xcite . here
let us recover the reduced planck constant @xmath171 .
starting from the same phase space point , _ the classical trajectories in superposition _ are close to each other by @xmath175 and @xmath183 after time @xmath15 .
those trajectories get separated by @xmath184 at @xmath182 and occupy different phase space volume elements .
the appropriate time step @xmath15 depends on the mass and the momentum diffusion constant of heavy quark . with the hard collision time scale @xmath185 , which provides a conventional lower limit of time discretization . neglecting the logarithms @xmath186
, @xmath187 corresponds to @xmath188 .
] this is much shorter than the relaxation time of the heavy quark momentum @xmath189 and thus is consistent with the above analysis in which the dissipative term with @xmath48 is ignored .
using this time scale , the distance between the heavy quark wave packets in superposition is @xmath190 so that @xmath191 also holds consistently .
moreover , the wave function is much more extended than the thermal de broglie length @xmath192 , which is a typical size of a wave packet .
the appropriate discretization time scale for the classical langevin dynamics is @xmath15 .
however , the typical time scale for the color space dynamics is much shorter @xmath193 , which is the interval of soft scatterings .
therefore , in a time scale @xmath194 , @xmath195 is updated , in effect , by a random @xmath7 matrix . replacing @xmath196 by a random pure state , which subsequently shrinks to one of the eigenstates of @xmath197 , the heavy quark color degrees of freedom are no more dynamical degrees of freedom . for a random pure state
, @xmath198_{\rm meas}$ ] takes values @xmath199 with equal probabilities .
this allows us to treat @xmath198_{\rm meas}$ ] as a stochastic variable with variance @xmath200 .
in this way , the normal langevin equation without color degrees of freedom is derived at time scale of @xmath0 : @xmath201
in this paper , i studied the langevin dynamics of heavy quarks with internal color degrees of freedom .
the langevin dynamics is not obtained in a closed form in the phase space but as a coupled dynamics in the phase space and in the color space .
the coupling causes the macroscopic superposition state , or the so - called schrdinger s cat state , in the updates of heavy quark momentum .
the classical variable @xmath129 , which corresponds to the cat , couples to the quantum state in the color space .
the quantum interference in the superposition state is destroyed in a time scale @xmath0 by the decoherence and the momentum update should be regarded as one of the possible outcome by a `` measurement '' .
note that the macroscopic superposition and the decoherence in the langevin dynamics are unique to the non - abelian interaction . in this way ,
i am naturally led to take @xmath0 for a discretization time scale of the langevin dynamics .
it is interesting because the physical quantities of the classical langevin equation ( @xmath1 and @xmath2 ) determine the discretization time scale @xmath15 for solving it . at this time scale @xmath15
, the color degrees of freedom are expected to be randomized .
there have been extensive phenomenological studies on the heavy quark energy loss and the drag force parameter @xmath202 has been extracted from the experimental data @xcite .
the decoherence of a charm ( bottom ) quark in the qgp with @xmath12 proceeds in a time scale @xmath203 .
these values are not very small compared to the typical lifetime of the qgp in the heavy - ion collisions , @xmath14 .
moreover , it is longer than the time scale of hadronization , which i roughly estimate to be @xmath16 .
this indicates that not only might the classical langevin equation be inappropriate to describe heavy quarks in the fireball , but also the freezeout process might be able to resolve a heavy quark wave function in the macroscopic superposition state .
such a wave function is extended about @xmath204 for a charm ( bottom ) quark and may enhance the probability of recombining heavy quark@xmath205antiquark pairs into heavy quarkonia . to establish a better ( semi)classical description for heavy quarks in the heavy - ion collisions , it would be desirable to derive a novel kinetic theory for heavy quarks in which ( i ) the time scale below @xmath10 can be resolved and ( ii ) the superposition of wave packets and their decoherence are effectively described . the kadanoff - baym equation for 4-dimensional wigner function can be helpful for this project @xcite .
finally , let us remark on the similarity between the discretization time scale @xmath0 and the formation time of induced gluon radiation off an energetic parton @xmath206 @xcite . here
@xmath207 and @xmath208 are gluon energy and transverse momentum , and @xmath209 is a transport coefficient characterizing transverse momentum kicks which the energetic parton experiences in the medium . in both cases , decoherence determines the time scales . for future prospect , it would be an interesting challenge to simulate the stochastic master equation , possibly in a simplified version without @xmath48 terms . without the @xmath48 terms
, the stochastic master equation is equivalent to the stochastic schrdinger equation @xcite .
the natural time step of the stochastic master equation is @xmath210 while the classical langevin dynamics is expected to emerge at longer time scale @xmath194 .
starting from a localized wave packet state , one may ask the following key questions : 1 .
does the thermal fluctuation with color @xmath83 generate a superposition state of wave packets ? 2 .
how is the interference between the wave packets destroyed by the decoherence ? 3 .
how does @xmath67 acquire the effect of decoherence in the corresponding classical phase space distribution ?
the answers to these questions will help bridge a gap between the quantum and classical descriptions of heavy quarks in the high - temperature quark - gluon matter .
i thank jean - paul blaizot , tetsufumi hirano , chiho nonaka , and derek teaney for fruitful discussions in an early stage of this study .
i also thank bin wu for suggesting an alternative interpretation of the decoherence time scale , during the molecule - type workshop selected topics in the physics of the quark - gluon plasma and ultrarelativistic heavy ion collisions , " at yukawa institute for theoretical physics , kyoto university .
in the markov limit , the influence functional can be expanded as @xmath80 @xcite .
the influence functional gives a part of path - integral weight @xmath211 for the propagation of the reduced density matrix .
explicit forms of @xmath81 and @xmath82 are @xmath212 \left ( \begin{array}{c } \rho^{a}_1\\ \rho^{a}_2 \end{array } \right)_{(t,\vec y)},\\ is_{\rm diss}&=&\frac{i}{4t}\int_{t_0 } dt\int d^3xd^3y\vec\nabla_xd(\vec
x-\vec y)\nonumber\\ & & \ \cdot\left ( \rho^{a}_1,\ \rho^{a}_2 \right)_{(t,\vec x ) } \left [ \begin{array}{cc } -1 & -1 \\ 1 & 1 \end{array } \right ] \left ( \begin{array}{c } \vec j^{a}_1\\ \vec j^{a}_2 \end{array } \right)_{(t,\vec y)}.\end{aligned}\ ] ] here @xmath213 is nonrelativistic heavy quark color current and labels 1 and 2 denote the schwinger - keldysh contour . integrating by parts , the spatial derivatives in @xmath214
, @xmath82 contains terms of similar structure to @xmath81 .
let us call this contribution @xmath215 .
then @xmath216 \left ( \begin{array}{c } \rho^{a}_1\\ \rho^{a}_2 \end{array } \right)_{(t,\vec y)}.\nonumber\end{aligned}\ ] ] since @xmath30 is positive definite for @xmath217 , the thermal fluctuation and a part of the dissipation @xmath218 can be expressed using a gaussian white noise @xmath83 as @xmath219 \right\rangle_{\xi } , \nonumber \\ & & \langle \xi^a(t,\vec x)\xi^b(s,\vec y)\rangle_t = f_1(\vec x-\vec y)\delta(t - s)\delta^{ab}.\end{aligned}\ ] ] this gives nothing but the stochastic propagation in eq . .
i investigated how to reconcile the kinetic equilibration of heavy quarks with the langevin equation .
i found that the quantum fluctuation in the color space is essential and proposed an interpretation . using the wigner function
, i can also describe the same physics and discuss the effect of quantum fluctuation . the stochastic master equation with explicit @xmath171
is obtained by substituting @xmath220 , @xmath221 , and @xmath222 .
the wigner function is defined by @xmath223 using @xmath224 , the color - averaged wigner function is given by @xmath225 .
taking the limit @xmath226 in the master equation , i obtain a kramers equation for @xmath227 : @xmath228 with @xmath229 .
the kramers equation is not obtained as a closed form of @xmath227 but rather depends on the color state of @xmath224 .
let us make the equivalent assumptions which have been made previously : 1 .
@xmath227 is localized in the phase space .
this allows us to replace @xmath230 by @xmath231 .
@xmath224 can be factorized into the wigner function in the phase space and that in the color space : @xmath232 .
the momentum kick at the position of the heavy quark can also be factorized : @xmath146 . in order to take the quantum fluctuation into account
, the last term of the kramers equation can be understood as @xmath233_{\rm meas } \cdot\frac{\partial}{\partial\vec p}\overline w_{\rm q}(t,\vec r,\vec p;\xi),\end{aligned}\ ] ] where @xmath234_{\rm meas}$ ] is evaluated by the color state of @xmath235 .
the dynamics of @xmath235 is the same with that of @xmath236 in eq . .
d. a. teaney , in _ quark gluon plasma 4 _ , edited by r. c. hwa , x .-
wang ( world scientific , singapore , 2010 ) [ arxiv:0905.2433 [ nucl - th ] ] .
k. kovtun , d. t. son and a. o. starinets , phys .
lett . * 94 * , 111601 ( 2005 ) [ hep - th/0405231 ] .
h. song , nucl .
a * 904 - 905 * , 114c ( 2013 ) [ arxiv:1210.5778 [ nucl - th ] ] .
g. d. moore and d. teaney , phys .
c * 71 * , 064904 ( 2005 ) [ hep - ph/0412346 ] .
r. rapp and h. van hees , in _ quark gluon plasma 4 _ , edited by r. c. hwa , x .-
wang ( world scientific , singapore , 2010 ) [ arxiv:0903.1096 [ hep - ph ] ] .
recent developments are summarized in a. beraudo , nucl .
a * 931 * , 145 ( 2014 ) [ arxiv:1407.5918 [ hep - ph ] ] .
s. caron - huot and g. d. moore , phys . rev .
* 100 * , 052301 ( 2008 ) [ arxiv:0708.4232 [ hep - ph ] ] ; j. high energy phys . *
0802 * , 081 ( 2008 ) [ arxiv:0801.2173 [ hep - ph ] ] ; s. caron - huot , m. laine and g. d. moore , _
ibid_. * 0904 * , 053 ( 2009 ) [ arxiv:0901.1195 [ hep - lat ] ] ; y. burnier , m. laine , j. langelage and l. mether , _
ibid_. * 1008 * , 094 ( 2010 ) [ arxiv:1006.0867 [ hep - ph ] ] .
p. petreczky and d. teaney , phys .
d * 73 * , 014508 ( 2006 ) [ hep - ph/0507318 ] ; h. t. ding , a. francis , o. kaczmarek , f. karsch , h. satz and w. soeldner , _
ibid_. * 86 * , 014509 ( 2012 ) [ arxiv:1204.4945 [ hep - lat ] ] ; d. banerjee , s. datta , r. gavai and p. majumdar , _
ibid_. * 85 * , 014510 ( 2012 ) [ arxiv:1109.5738 [ hep - lat ] ] ; a. francis , o. kaczmarek , m. laine , m. mller , t. neuhaus and h. ohno , pos lattice * 2013 * , 453 ( 2014 ) [ arxiv:1311.3759 [ hep - lat ] ] ; a. francis , o. kaczmarek , m. laine , t. neuhaus and h. ohno , arxiv:1508.04543 [ hep - lat ] .
c. p. herzog , a. karch , p. kovtun , c. kozcaz and l. g. yaffe , j. high energy phys . * 0607 * , 013 ( 2006 ) [ hep - th/0605158 ] ; s. s. gubser , phys .
d * 74 * , 126005 ( 2006 ) [ hep - th/0605182 ] ; j. casalderrey - solana and d. teaney , _ ibid_. * 74 * , 085012 ( 2006 ) [ hep - ph/0605199 ] . s. k. wong , nuovo cim .
a * 65 * , 689 ( 1970 ) .
y. akamatsu , phys .
d * 91 * , 056002 ( 2015 ) [ arxiv:1403.5783 [ hep - ph ] ] ; * 87 * , 045016 ( 2013 ) [ arxiv:1209.5068 [ hep - ph ] ] .
a. o. caldeira and a. j. leggett , phys .
a * 31 * , 1059 ( 1985 ) .
a concise review is w. h. zurek , phys . today *
44*(10 ) , 36 ( 1991 ) ( updated version available at arxiv : quant - ph/0306072 ) .
p. braun - munzinger and j. stachel , phys .
b * 490 * , 196 ( 2000 ) [ nucl - th/0007059 ] ; r. l. thews , m. schroedter and j. rafelski , phys .
c * 63 * , 054905 ( 2001 ) [ hep - ph/0007323 ] .
breuer and f. petruccione , _ the theory of open quantum systems _ ( oxford university press , new york , 2002 ) . | a langevin equation of heavy quarks in high - temperature quark - gluon plasma is derived .
the dynamics of heavy quark color is coupled with the phase space dynamics and causes a macroscopic superposition state of heavy quark momentum .
decoherence of the superposition state allows one to use a classical description .
the time scale of decoherence gives an appropriate discretization time scale @xmath0 for the classical langevin equation , where @xmath1 is heavy quark mass and @xmath2 is heavy quark momentum diffusion constant . |
in this paper , we analyze point - to - point communication of energy harvesting nodes from an information - theoretic perspective .
we focus on wireless networking applications where nodes ( e.g. , sensors nodes ) can harvest energy from nature through various different sources , such as solar cells , vibration absorption devices , water mills , thermoelectric generators , microbial fuel cells , etc . in such systems ,
energy that becomes available for data transmission can be modeled as an exogenous recharge process .
therefore , unlike traditional battery - powered systems , in these systems , energy is not a deterministic quantity , but is a random process which varies stochastically in time at a scale on the order of symbol duration .
the transmission can be interrupted due to lack of energy in the battery . on the other hand ,
excess energy can be buffered in the battery before consumption for transmission .
this model requires a major shift in terms of the power constraint imposed on the channel input compared to those in the existing literature . to illustrate , in information - theoretic approaches ,
there are two widely used input constraints on the channel inputs of continuous - alphabet channels : average power constraint and amplitude constraint .
if the input is average power constrained , then any codeword should be such that while each symbol can take any real value , the average power of the entire codeword should be no more than the power constraint .
on the other hand , if the input is amplitude constrained , then every code symbol should be less than the constraint in amplitude .
it is clear that in an energy harvesting model , the constraint imposed on the channel input is different than these constraints , in that , while code symbols are instantaneously amplitude constrained , energy can be saved in the battery for later use .
this amounts to an unprecedented input constraint on the channel input . in this context
, the main goal of this paper is to investigate the effect of stochastic energy arrivals on the communication in an information - theoretic framework . in particular , we augment an energy buffer to the classical awgn system and study information - theoretically achievable rates . first , we consider the setting where energy arrives at the transmitter as a stochastic process , which varies on the order of symbol duration .
the problem is posed as design of a codebook that complies with all of the input constraints .
the input constraint imposed by stochastic energy arrivals is a random one and in this sense it generalizes classical deterministic average or amplitude power constraints .
the recharge process together with past code symbols determine the allowable range of inputs in each channel use .
we start with showing that the capacity of the awgn channel with an average power constraint equal to the average recharge rate is an upper bound for the capacity in the energy harvesting system .
then , we develop a scheme called _ save - and - transmit _ scheme that achieves this upper bound and hence the capacity .
the _ save - and - transmit _ scheme relies on sending zero code symbols in a portion of the total block length , which becomes negligible as the block length gets larger .
next , based on feasibility to send the code symbol in a channel use , we provide an alternative scheme called the _ best - effort - transmit _
scheme for achieving the capacity .
whenever the available energy is sufficient to send the code symbol , it is put to the channel , while a zero is put to the channel if there is not enough energy in the battery .
we show that this scheme achieves rates arbitrarily close to the capacity .
secondly , we address a typical behavior in certain energy harvesting sensors , such as solar - powered sensors , where the recharge process is not ergodic or stationary . in this case
, we assume that the average recharging rate is not a constant in time , but rather fosters time variation in a scale much larger than the time scale that communication takes place .
accordingly , we extend the formulation to the case in which recharge process has a mean value that is varying in sufficiently long time and we call each such sufficiently long time a slot .
we derive the optimal power control for maximum average throughput in this case , and provide a geometric interpretation for the resulting power allocation .
we illustrate the advantage of the optimal power allocation in a numerical study . to the best of our knowledge
, this work is the first attempt to analyze an energy harvesting communication system from an information - theoretic perspective .
there are many motivating works in the networking literature . in @xcite , lei _
address replenishment in one hop transmission . formulating transmission strategy as a markov decision process ,
@xcite uses dynamic programming techniques for optimization of transmission policy under replenishment . in @xcite , gatzianas _
_ extend classical wireless network scheduling results to a network with users having rechargeable batteries .
each battery is considered as an energy queue , and data and energy queues are updated simultaneously in an interaction determined by rate versus power relationship .
stability of data queues is studied using lyapunov techniques . in @xcite , in a similar energy harvesting setting , a dynamic power management policy is proposed and is shown to stabilize data queues . in each slot , energy spent is equal to the average recharge rate . moreover , under a linear approximation , some delay - optimal schemes are proposed .
the capacity of scalar awgn channel has been extensively studied in the literature under different constraints on the input signal .
average power ( or the second moment ) constraint on the input yields the well - known result that the capacity achieving input distribution is gaussian with variance equal to the power constraint .
smith @xcite considers amplitude constraints in addition to average power constraints and concludes that the capacity achieving input distribution function is a step function with finite number of increase points .
moreover , shamai and bar - david @xcite extend smith s result to amplitude constrained quadrature gaussian channel for which the optimal input distribution is concentrated to finite number of uniform phase circles within the amplitude constraint .
we consider scalar awgn channel characterized by the input @xmath0 , output @xmath1 , additive noise @xmath2 with unit normal distribution @xmath3 and a battery ( see fig . [ sys_mod ] ) . input and output alphabets
are taken as real numbers .
energy enters the system from a power source that supplies @xmath4 units of energy in the @xmath5th channel use where @xmath6 .
@xmath7 is the time sequence of supplied energy in @xmath8 channel uses .
@xmath4 is an i.i.d .
sequence with average value @xmath9 , i.e. , @xmath10=p$ ] , for all @xmath5 .
we assume that the energy stored and depleted from the battery are for only communication purposes ( for example , the energy required for processing is out of the model ) .
existing energy in the battery can be retrieved without any loss and the battery capacity is large enough so that every quanta of incoming energy can be stored in the battery .
this assumption is especially valid for the current state of the technology in which batteries have very large capacities compared to the rate of harvested energy flow @xcite : @xmath11 .
the battery is initially empty and energy needed for communication of a message is obtained from the arriving energy during transmission of the corresponding codeword subject to causality . in particular , @xmath4 units of energy is added to the battery and @xmath12 units of energy is depleted from the battery in the @xmath5th channel use .
this brings us to the following cumulative power constraints on the input based on causality : @xmath13 this is illustrated in fig .
[ cw ] , where in each channel use , @xmath4 amount of energy arrives and @xmath12 amount of energy is used . note that the constraints in ( [ constraints ] ) are upon the support set of the random variables @xmath14 .
the first constraint restricts the support set of @xmath15 to @xmath16 $ ] .
the second constraint is @xmath17 . in general , letting @xmath18 denote @xmath19^+$ ] , in channel use @xmath5 , the symbol @xmath14 is subject to the constraint @xmath20 .
the input constraints in ( [ constraints ] ) are challenging because @xmath4 are random and these constraints introduce memory ( in time ) in the channel inputs .
randomness in @xmath4 makes the problem similar to fading channels in that state of the recharge process ( i.e. , low or high @xmath4 ) affects instantaneous quality of communication .
moreover , this time variation in recharge process allows opportunistic control of energy as in fading channels .
however , recharge process can be collected in battery unlike a fading state .
in fact , we will see that , this nature of energy arrivals makes it more advantageous to save energy in the battery for future use when a peak occurs in the recharge process , as opposed to opportunistically _ ride _ the peaks .
we will invoke the general capacity formula of verdu and han @xcite . for fixed @xmath8 ,
let @xmath21 be the joint density function of random variables @xmath22 and let @xmath23 be the set of @xmath8 variable joint density functions that satisfy the constraints in ( [ constraints ] ) . since awgn is an information - stable channel @xcite , the capacity of the channel in fig .
[ sys_mod ] with constraints in ( [ constraints ] ) is : @xmath24 in general , capacity achieving input distribution is in the form of product of marginal distributions ( independent distribution ) @xcite .
however , note that the power constraints create dependence among the random variables .
the constraint on @xmath25 is dependent on given value of @xmath26 , @xmath27 .
though independent processes achieve higher mutual information than the ones with the same marginal distribution but with correlation @xcite , the capacity that we seek in this problem does not let the process be independent .
this problem falls in the family of problems of finding capacity under dependence constraints on code symbols which is by itself interesting and less studied .
, @xmath4 denotes the energy arriving , and @xmath14 denotes the channel input .
therefore , @xmath12 is the energy used at time @xmath5 . ]
an upper bound for @xmath28 is the corresponding awgn capacity with average power @xmath9 , as @xmath29 and by the i.i.d .
nature of @xmath4 , invoking strong law of large numbers @xcite , @xmath30 with probability 1 .
therefore , each codeword satisfying constraints in ( [ constraints ] ) automatically satisfies @xmath31 with probability 1 .
however , if a codeword satisfies average power constraint , it does not necessarily satisfy the constraints in @xmath32 .
hence , we get the following bound : @xmath33 next , we state and prove that the above upper bound can be achieved .
[ cap ] the capacity of the awgn channel under i.i.d .
random energy arrivals @xmath4 , where @xmath34=p$ ] , is independent of the realizations of @xmath4 and equal to the channel capacity under average power constraint @xmath9 : @xmath35 to prove the theorem , we need an achievable scheme .
achieving the capacity requires design of an @xmath36 codebook with encoding function @xmath37 and decoding function @xmath38 where @xmath8 is the code length , @xmath39 is the code size and @xmath40 is the probability of error such that @xmath41 and @xmath42 .
there are two separate causes of error .
the first one is that any codeword does not satisfy the input constraints . in this case
, the encoder does not send that codeword but sends an all zero codeword .
hence , all zero codeword is assumed to be in the codebook and it is decoded to message @xmath43 at the receiver .
if the decoder maps received signal to message @xmath43 , it is accepted as an error .
second cause of error is the actual decoding error at the receiver .
if the received signal is decoded to a message that is different from the message sent , then an error occurs .
accordingly , the error event is defined as union of these two events . while designing the codebook and the encoding / decoding rule
, a first approach can be to optimize the codebook design subject to the input constraints .
therefore , the occurrence of the first type of error ( i.e. , insufficient energy ) is eliminated from the beginning .
however , we will show that , it is not necessary to solve this difficult optimization problem .
instead , the asymptotic behavior of the constraints allows us to design codes that do not obey the input constraints for a particular @xmath8 with a small probability but the requirement is that small probability goes to zero asymptotically . for the error probability to go to zero , we need to average out the error due to randomness introduced by the channel and due to the randomness in the energy arrivals . to do this
, we propose a scheme that implements a _ save - and - transmit _ principle in that error due to randomness in energy is averaged out first and then the channel coding is performed .
in addition , it is possible to maintain error - free communication even if some code symbols can not be put to the channel correctly .
therefore , we also investigate achievable rates using a _ best - effort - transmit _ scheme , which can interestingly approach @xmath44 neighborhood of the capacity for any @xmath45 .
we now present an achievable scheme .
let @xmath46 such that @xmath47 . here
, @xmath48 denotes the class of functions @xmath49 such that @xmath50 . consider the sequence of codes with code length @xmath8 and rate @xmath51 such that the first @xmath52 symbols of each codeword is zero and the remaining @xmath53 codewords are chosen as independent random variables from the ( capacity achieving ) gaussian distribution with variance @xmath9 where @xmath9 is the average recharge rate .
that is , for @xmath54 , we have , the encoding function , @xmath55 for all @xmath56 . for @xmath57 , @xmath58 comes from realizations of a gaussian distributed random variable of variance @xmath9 for all @xmath56 . as @xmath8 grows large , by strong law of large numbers , at time index @xmath52 , about @xmath59 amount of energy
is collected in the battery with very high probability . since no energy is consumed up to this time , the codebook is feasible until time @xmath52 .
after time @xmath52 , the probability that a code symbol requires more energy than the energy available in the battery goes to zero and energy consumed is accounted for the energy arrived . in the next @xmath52 channel uses , @xmath59 units of energy is collected in the battery and another @xmath59 is consumed for data transmission .
we repeat this procedure for a total of @xmath60 times , where in each interval of @xmath52 symbols , we collect about @xmath61 units of energy and use about @xmath61 units of energy collected in the previous block of @xmath52 symbols .
the first collected amount guarantees that the codewords are always feasible ( see fig .
[ sys_mod3 ] ) .
more specifically , as the length of the codewords gets large , by strong law of large numbers , we have @xmath62 almost surely for all @xmath63 . for a precise proof of these arguments ,
we need to make use of marcinkiewicz - zygmund type strong laws for sums of i.i.d . random variables
. please see @xcite for a complete proof .
the achievable rate for this scheme is @xmath64 in other words , communication can start with @xmath52 amount of delay and the capacity of a corresponding average power constrained awgn channel can be achieved . since
, this scheme is asymptotically feasible , this proves the achievability part and completes the proof of theorem [ cap ] .
it should be emphasized that the above achievable scheme is obtained without using causal or noncausal information of energy arrivals . for any realization of @xmath4
, we introduce a reasonable delay @xmath52 to save energy and afterwards transmit with average power @xmath9 .
this scheme obeys the input constraint with probability 1 and achieves the upper bound in ( [ bound ] ) .
the ability to collect energy in the battery allows designer to apply a save - and - transmit strategy so that the uncertainty in the energy arrivals is eliminated first and then the uncertainty in the channel is dealt with by means of appropriate channel coding .
code symbols are identically zero for all codewords .
remaining @xmath65 code symbols are selected as i.i.d .
gaussian distributed .
evolution of a sample energy arrival and energy expenditure for a particular codeword is illustrated .
collected energy in the first @xmath52 channel uses makes it asymptotically impossible to deem any codeword infeasible . ]
let @xmath66 be a codeword of length @xmath8 where @xmath14 is the code symbol to be sent in channel use @xmath5 and the codebook be @xmath67 .
the codebook that two parties agree upon is determined by generating independent gaussian distributed random variables with mean zero and variance @xmath68 for each code symbol .
let @xmath69 be the battery energy just before the @xmath5th channel use starts . in the best - effort - transmit scheme
, the code symbol @xmath14 can be put to the channel if @xmath70 .
otherwise , transmitter puts a code symbol @xmath43 to the channel as battery does not have sufficient energy to send symbol @xmath14 .
hence , the input to the channel is @xmath71 and therefore there is a possible mismatch .
yet , we will show in the following that the mismatch does not affect the correct decoding of the message at the receiver .
the battery energy is updated as follows @xmath72 the energy updates in ( [ updates ] ) are analogous to the queue updates in classical slotted systems @xcite where the notion of a slot is replaced with channel use
. therefore , battery acts like an energy queue . unlike in the classical systems with packet queues ,
our aim is to keep the energy queue unstable so that the availability of the energy to send a code symbol is asymptotically guaranteed .
the key to the updates in ( [ updates ] ) is to put @xmath43 symbol on the channel as it guarantees that infeasible code symbols are observed only finitely many times , which is stated in the following theorem , and detailed proof is again skipped for brevity and can be found in @xcite .
[ fmthm ] let @xmath73 be a codeword such that @xmath74 is an i.i.d . real random sequence with mean zero and variance @xmath75 for some @xmath45 sufficiently small .
let @xmath4 be the energy arrivals and @xmath69 be the battery energy which is updated as in ( [ updates ] ) .
then , only finitely many code symbols are infeasible .
it is well known that we can pack about @xmath76 codewords with block length @xmath8 , each having average power less than or equal to @xmath68 , in the @xmath77 space such that , whenever one of them is sent by the transmitter , a joint typicality decoder @xcite which checks whether a codeword is jointly typical with the received signal vector , can reconstruct the codeword at the receiver with probability of error approaching zero . in the _ best - effort - transmit _ scheme
, the received signal vector will still be jointly typical with the sent codeword , since the number of infeasible code symbols is only finitely many due to theorem [ fmthm ] .
hence , the same decoder can recover the codeword if the channel input is @xmath78 in each channel use .
therefore , the code rate @xmath79 is achievable , and hence by continuity of @xmath80 , @xmath81 can be achieved .
we have seen that classical awgn capacity with average power constraint can be achieved with @xmath48 delay for fueling the battery with energy if the recharge process is i.i.d .
however , the recharge process can deviate from its i.i.d .
characteristic in a large time scale .
in particular , the mean value of the recharge process may vary after a long time . in the classical example of sensor nodes fueled with solar power , mean recharge rate changes depending on the time of the day . as an example , the mean recharge rate may vary in one - hour slots and the sensor may be on for twelve hours a day , in which case , a careful management of energy expenditure in each slot will be required to optimize the average performance during the day .
we generalize the system model for @xmath82 large time slots ( see fig .
[ sys_mod4 ] ) .
we assume that the duration of each slot is @xmath83 ( large enough ) . for each slot @xmath84 ,
average recharge rate is @xmath85 and @xmath86 units of power is allocated for data transmission
. in slot @xmath5 , @xmath87 units of energy enters the battery and @xmath88 units of energy is spent for communication . if @xmath89 , @xmath90 units of energy is saved , or otherwise @xmath91 units of energy is depleted from the battery . assuming zero initial energy in the battery and large enough battery capacity , every unit of incoming energy is saved in the battery .
the causality of energy arrivals requires @xmath92 around @xmath93 bits of data are sent in slot @xmath5 .
before the communication starts , suppose the designer knows the mean recharge rates @xmath85 for all @xmath5 , calculates @xmath86 and adjusts the average power of codewords in slot @xmath5 to @xmath86 during transmission works as well .
this can also be interpreted as a codebook with dynamic power allocation @xcite in slow time variation . ] .
we allocate transmit power to each slot subject to causality constraint so that average throughput in @xmath82 slots is optimized : @xmath94 we will denote the solution of the above optimization problem as @xmath95 $ ] .
note that the slot duration @xmath83 does not appear in the optimization problem .
the rest of this section is devoted to characterizing @xmath96 . after developing necessary concepts ,
we present an algorithm to find @xmath96 and illustrate it for several cases . in order to understand how the optimal solution may look like
, we will first pose the simplest version of the problem .
suppose there is @xmath97 amount of energy available in the battery where @xmath98 is a constant and the recharge process is zero .
then , the optimal power allocation strategy @xmath96 is the one that maximizes @xmath99 subject to @xmath100 . by jensen s inequality @xcite ,
optimal strategy is @xmath101 .
that is , if @xmath102 for @xmath103 , then we can always achieve higher average throughput by setting @xmath104 .
returning to the original problem , above observation translates to a general optimality criterion for the problem , stated in the following theorem
. please see @xcite for a proof of this theorem .
large time slots . in each slot ,
sufficiently large time passes to achieve awgn capacity with average power constrained to allocated power in that slot . ]
let @xmath105 $ ] be any power allocation such that @xmath106 , for @xmath107 .
assume @xmath108 be another power allocation such that @xmath109 , @xmath107 and for some @xmath110 with @xmath111 @xmath112 then , @xmath113 .
this theorem proposes a method to optimize the power allocation .
it should be performed such that variation in power from slot to slot is avoided as much as possible and the energy should be consumed in as smooth a way as possible .
we will now make this more precise by describing the method of finding the solution @xmath96 .
it is inferred that @xmath96 must take a constant value ( not necessarily the same value ) in each slot .
in addition , we can partition the whole interval into disjoint intervals over which @xmath114 is constant as follows : @xmath115}{(n_{k+1 } - n_k)t_s}\ ] ] for all @xmath5 in @xmath116 for some subset @xmath117 of @xmath118 with @xmath119 .
thus , the problem can also be cast as finding the right subset @xmath120 that optimizes the average throughput .
note that @xmath43 and @xmath82 are always in this subset .
hence , @xmath121 and the element with last order is @xmath82 .
define the cumulative energy arrival rate as @xmath122 for all @xmath123 and reset @xmath124 . in each point
, we have to use the knowledge of the amount of energy that is going to arrive in the remaining time interval and determine feasible constant power strategies and find the optimal one among them .
we have @xmath125 and next @xmath126 are determined for optimal strategy as follows @xcite : @xmath127 then , the optimal power allocation is as follows : @xmath128 an illustration of the operation of the algorithm is presented in fig .
[ opalg ] .
the operation of the algorithm has the following geometric structure . at each point @xmath5 ,
lines are drawn from cumulative energy point @xmath129 to future points @xmath130 for all @xmath131 and the one with minimum slope is chosen , say @xmath132 .
that slope is the allocated power for all slots between @xmath5 and @xmath132 .
the optimal power management algorithm takes the arrival power vector @xmath133 $ ] and outputs the transmit power vector @xmath134 $ ] .
we let the arrival rates of energy in all slots , @xmath85 , follow an i.i.d . exponential distribution .
a benchmark algorithm is simply no power management algorithm , i.e. , @xmath135 . in this simple scheme ,
the energy arrival rate in each slot is taken as the communication power in that slot .
this scheme yields an average throughput @xmath136 which is a lower bound .
however , if the designer has the information of arrival rates in future slots , then optimal power management algorithm can improve the average throughput .
it is clear that an upper bound for the average throughput is @xmath137 the comparison of performances of optimal power management with the upper bound @xmath138 and the lower bound @xmath139 ( no power management ) is given in fig .
[ numres ] for a @xmath140 slot system .
we observe that as the variance of the arrival rates increases , the advantage of optimal power management becomes more apparent with respect to no power management .
because , the power is more peaky as the variance is increased yet for better throughput , transmitter should not ride the peaks but rather save the energy for future use .
another observation is that the difference between upper bound and average throughput with optimal power management also increases as the standard deviation of the arrival rate is increased .
hence , the causality constraint becomes more restrictive as the variation in the arrival rate is increased .
slots is plotted for different values of mean / standard deviation of the arrival rate . as the variation in the arrival rate increases , optimal power management algorithm yields higher advantage over no power management . ]
we studied communication in an awgn channel under random energy arrivals using an information - theoretic framework .
we showed that the capacity of awgn under energy harvesting is the same as the capacity of the awgn channel , where the average power is constrained to the average recharge rate .
this upper bound can be achieved by save - and - transmit and best - effort - transmit schemes .
next , we addressed time varying recharge rates in large time scales .
we obtained an algorithm to find the optimal power management for maximum average throughput .
we illustrated its operation and provided a numerical study to visualize the advantage of the algorithm .
j. lei , r. yates , and l. greenstein , `` a generic framework for optimizing single - hop transmission policy of replenishable sensors , '' _ ieee transactions on wireless communications _ ,
vol . 8 , pp . 547551 , feb .
2009 .
v. sharma , u. mukherji , v. joseph , and s. gupta , `` optimal energy management policies for energy harvesting sensor nodes , '' _ ieee transactions on wireless communications _
, vol . 9 , pp .
13261336 , april 2010 . | in energy harvesting communication systems , an exogenous recharge process supplies energy for the data transmission and arriving energy can be buffered in a battery before consumption .
transmission is interrupted if there is not sufficient energy .
we address communication with such random energy arrivals in an information - theoretic setting . based on the classical additive white gaussian noise ( awgn ) channel model , we study the coding problem with random energy arrivals at the transmitter . we show that the capacity of the awgn channel with stochastic energy arrivals is equal to the capacity with an average power constraint equal to the average recharge rate .
we provide two different capacity achieving schemes : _ save - and - transmit _ and _ best - effort - transmit_. next , we consider the case where energy arrivals have time - varying average in a larger time scale .
we derive the optimal offline power allocation for maximum average throughput and provide an algorithm that finds the optimal power allocation . |
measurements of the spin - dependent structure function of the nucleon , @xmath0 , continue to generate enormous interest .
not only has the precision with which the ellis - jaffe sum - rule is known improved , but with the addition of neutron data from slac and smc the bjorken sum - rule has also been tested @xcite .
our aim is to provide a necessarily brief review of the situation with respect to the sum - rules .
( for those interested in more detail we suggest some recent reviews @xcite . )
we then argue that the emphasis in future should be much more on the shape and momentum dependence of @xmath0 .
there has recently been substantial progress in understanding the flavour and spin dependence of the parton distributions in terms of the quark models familiar from low energy spectroscopy and we briefly recall the main ideas . as nature
has not seen fit to give us a free neutron target , nuclear and particle physicists must work together in order to extract @xmath1 .
some very recent progress in the treatment of deep - inelastic scattering from the deuteron is reviewed .
finally we highlight some important open questions which need experimental or theoretical attention .
the spin structure function @xmath0 is determined by the difference in the cross section for a polarised , virtual photon to be absorbed by a nucleon with its spin anti - parallel and parallel to that of the photon g_1_-_. as the photon is absorbed by a spin-@xmath2 quark , in the former case the absorbing parton must have spin parallel to that of the nucleon @xmath3 while in the latter case it must be anti - parallel @xmath4 .
it is usual to define the net spin ( helicity ) carried by quarks ( and anti - quarks ) of a given flavour as u(x , q^2)=u^-u^+^- ^ , etc .
( quarks and anti - quarks have the same charge squared and contribute with the same sign . ) thus one easily finds g_1p(x , q^2)=\{u(x , q^2)+ d(x , q^2)+s(x , q^2)+ } , and if we define the isovector , octet and singlet flavour combinations in the usual way q_3(x , q^2)&=&u(x , q^2)-d(x , q^2 ) , + q_8(x , q^2)&=&u(x , q^2)+d(x , q^2)-2s(x , q^2 ) , + q_0(x , q^2)&=&u(x , q^2)+d(x , q^2)+s(x , q^2 ) , we find g_1p(n)(x , q^2)=q_3+q_8 + q_0 .
using current algebra the integrals of @xmath5 and @xmath6 can be related to the axial charges in neutron and hyperon @xmath7-decay q_3&=&_0 ^ 1dx q_3(x , q^2)=1.2570.003 , + q_8&=&_0 ^ 1dx q_8(x , q^2)0.590.02 , where the question mark indicates some residual uncertainty over the use of flavour su(3 ) symmetry . from the definition in equ.(4 ) it is clear that the integral of @xmath8 is just the fraction of the spin of the nucleon carried by its quarks , @xmath9 : = _ 0 ^ 1dx q_0(x , q^2 ) .
within qcd there are also perturbative corrections which vanish logarithmically as @xmath10 , so that the integrals of the proton and neutron spin structure functions s^p(n)(q^2)=_0 ^ 1 dx g_1p(n)(x , q^2 ) , become @xcite : s^p(n)(q^2)&=&(1 - - 3.58()^2- 20.2()^3 ) ( q_3 + ) + & + & ( 1 - - 1.1()^2 ) , ( for three flavours ) .
the bjorken sum - rule involves the difference of the integrals for the proton and neutron s^bj&=&s^p - s^n=(1 - - 3.6()^2)q_3 , + & = & 0.185atq^2=10 gev^2 .
this particular sum - rule is extremely important as a failure would represent a failure of current algebra for the quarks @xcite .
the ellis - jaffe sum - rule @xcite requires a dynamical assumption , namely that the polarisation of the strange sea of the nucleon vanishes , @xmath11 . with this reasonable assumption
we find @xmath12 and hence s_ej^p&=&0.1710.004,(0.1610.004 ) , + s_ej^n&&=-0.0140.004,(-0.0100.004 ) at @xmath13 ( @xmath14 ) .
the latest experimental information was presented in detail at this meeting @xcite and therefore we simply state the results .
the original emc measurement @xcite gave @xmath15 at @xmath16 .
this corresponds to @xmath9 of only @xmath17 ( 12% to 0(@xmath18 ) ) .
it is clearly compatible with _
none of the nucleon spin being carried by its quarks _ , which was the origin of the term ` spin crisis ' .
this led to great interest in alternatives to the quark model , such as the skyrme model @xcite .
indeed the question has been asked whether the quark model would have been taken seriously if this had been known in the 60 s rather than the famous prediction @xmath19 . on the other hand
, a cautious observer might add that the result is only 2 - 3 standard deviations from the ellis - jaffe prediction . at this meeting
new values were presented from the smc and slac groups .
for the proton , smc finds @xmath20 at 10 @xmath21 , while a combined analysis with earlier proton data yields @xmath22 @xcite . the preliminary result reported from slac e143 was @xmath23 at @xmath14 , while for the deuteron and neutron they found @xmath24 and @xmath25 , respectively @xcite .
these values are summarised in table 1 . in all cases the experimental values are significantly below the theoretical expectation and it is common to conclude that the two experiments agree and that @xmath9 is of order one third . on the other hand ,
we know that the integral of @xmath26 changes sign at @xmath27 ( to give the dhg sum - rule @xcite ) , so that there is a dramatic variation in the sum - rule , due to non - perturbative effects , between @xmath27 and 3 @xmath21 .
vector meson dominance has been used by ioffe and collaborators to provide a physically reasonable , smooth interpolation between the low-@xmath28 and asymptotic regimes @xcite . at 3 and 10
gev@xmath29 this approach suggests a correction to the measured sum - rule for the proton ( neutron ) of 29% ( 20% ) and 8% ( 6% ) respectively . while there is , as yet , no consensus on these values @xcite we find the sign and magnitude very reasonable .
applying them to the results quoted above , we find the values for the integrals and corresponding spin fractions shown in table 1 .
.integrals of @xmath30 and @xmath1 with and without a correction for the non - perturbative dependence on @xmath28 following refs.@xcite . [ cols="^,^,^,^,^",options="header " , ] after correction for non - perturbative effects the experiments are consistent with an average value of the spin carried by the quarks of @xmath31 somewhat higher than the value of one third found without the correction .
we believe that @xmath31 is probably the best estimate at the present time .
it is roughly one standard deviation below the value of 60% implicit in the ellis - jaffe sum - rule .
notice that this expectation is not 100% for the same reason that @xmath32 is not 5/3 the lower component of the quark wave function is p - wave and the quark spin most often points opposite to the total angular momentum .
this is a simple , well understood , relativistic correction .
clearly the data is no longer consistent with @xmath33 and a 1 @xmath34 deviation is not a crisis .
nevertheless the physics is far from clear and in many ways the serious study of the problem is just beginning !
much of the theoretical interest in the ellis - jaffe sum - rule , and most of the papers , have been concerned with the role of the u(1 ) axial anomaly in the flavour singlet piece , @xmath35 . following the work of efremov and teryaev and altarelli and ross @xcite ( etar )
see also ref.@xcite it is common to correct the naive parton model ( npm ) value for @xmath36 and @xmath37 by u(x , q^2 ) & & u(x , q^2)|_npm-(x , q^2 ) , + d(x , q^2 ) & & d(x , q^2)|_npm-(x , q^2 ) , + s(x , q^2 ) & & s(x , q^2)|_npm-(x , q^2 ) , where ( x , q^2)= _
x^1c^g(,q^2 ) g(y , q^2 ) dy , and _ 0 ^ 1(x , q^2)=g . here
@xmath38 is the fraction of the proton spin carried by gluons . in this picture _
npm-3 , [ sigeqn ] and if ( following ellis and jaffe ) one sets @xmath37 to zero in the npm we find @xmath39 through the anomaly .
clearly the residual discrepancy in the ellis - jaffe sum - rule could be fixed with @xmath40 , which corresponds to @xmath41 at @xmath13 , or @xmath42 at a quark model scale ( @xmath43 ) .
the latter agrees rather too well with the simple estimate of @xmath38 associated with one - gluon - exchange by brodsky and schmidt @xcite .
certainly this result is far more reasonable that the values of @xmath44 required when @xmath9 was compatible with zero !
we would like to emphasise that one wins in two ways through radiation of polarised gluons .
firstly , as we have just seen , the point - like coupling to the quarks in the gluon reduces @xmath9 . secondly , however , these gluons carry spin away from the quarks . this correction was shown to amount to 5 - 10% in @xmath9 @xcite but as it involves non - perturbative qcd has generally been overlooked by the same groups who use equ.([sigeqn ] ) . as usual with the anomaly
there has been considerable controversy .
the separation between @xmath45 and @xmath46 is not gauge invariant at the operator level .
indeed this lies at the heart of the anomaly @xcite , which arises precisely because one can not satisfy chiral symmetry gauge invariance . in the operator product expansion ( ope )
there is no operator corresponding to @xmath38 . from this a number of authors , notably bodwin and qiu , have argued that the first moment of @xmath47 should vanish @xcite .
there has also been much debate over the infrared sensitivity of the separation of @xmath48 and @xmath46 [ 4,5,24 - 27 ] . in the end , for the gauge usually used in the parton model ( @xmath49 ) ,
the correction is probably reliable .
however this raises another problem , the shape of @xmath50 .
as emphasised in ref.@xcite , the gluonic correction effects @xmath50 only at very small x ( well below 0.1 ) .
this is well below the region where the calculations of @xmath51 deviate from the data .
the study of the shape of @xmath50 , and particularly the mechanism for transfering strength from intermediate x to low x , is therefore a clear priority for future work . with this in mind , in the next section we present some simple ideas which help us to understand the shape of @xmath30 and @xmath1 . before closing
we should mention that not everyone is convinced that the effects of the anomaly are confined to small x , and any hint of ozi violation in the purely valence region would be most interesting indeed @xcite .
formally the twist - two parton distributions can be written @xcite q^(2)(x,^2)=_ndp |np|_+(0)|n|^2(m(1-x)-p_n^+ ) , with p_n^+=+p_z>0 .
here @xmath52 are a complete set of states of momentum @xmath53 and rest mass @xmath54 and equ.(16 ) guarantees that the distribution vanishes for @xmath55 .
this formal expression has been used extensively to calculate parton distribution for the mit bag model by the adelaide group @xcite .
an important conclusion of that work is that the dominant contribution ( at the quark model scale @xmath56 c.f .
refs.@xcite ) arises from a two - quark intermediate state , n=2 . for a simple nucleon wave function , with three quarks in the 1s - state
, the matrix element @xmath57 should peak near @xmath58 .
hence the maximum value of @xmath59 should occur at @xmath60 from the @xmath46-function and the dominant contribution at large x will be the one associated with the lowest value of @xmath61 .
this was first realised by close and thomas who stressed the importance for both the flavour and the spin dependence of the parton distribution @xcite .
for example , using the standard su(6 ) wave function for the proton one easily finds that a valence d - quark has only an s=1 pair as spectator , while a u - quark has a 50 - 50 chance of s=0 or s=1 .
whatever mechanism is used to generate the n-@xmath62 mass splitting
e.g. gluon exchange , pion exchange or instantons gives typically @xmath63 .
hence the u - quarks should dominate at large x
i.e. one expects @xmath64 as @xmath65 , as found experimentally . for a polarised proton ( neutron ) only the valence @xmath66
is found with an @xmath67 pair and hence @xmath66 should dominate at large x. from equ.(2 ) we see that @xmath30 must be positive , as must @xmath1 , at intermediate and large x. note that all this follows from the _ same physics _ required in low energy spectroscopy ( and a simple su(6 ) wave function ) .
it works because deep inelastic scattering involves a light - cone correlation function and the energy of the quark is as important as its momentum .
of course , a full calculation should also include the configuration mixing in the wave function , but this is not the _ leading _ effect .
in passing we note that this same approach to calculating parton distributions leads to an unexpectedly large violation of charge symmetry for the minority valence quarks ( @xmath68 ) @xcite .
it is somewhat disappointing that the focus on sum - rules means that we still do not know whether @xmath1 does become positive at intermediate and large x. although the approach just described is unambiguous , the inclusion of the correction associated with the axial anomaly could change this and it is very important to know .
figure 1 illustrates this by showing the prediction for @xmath1 from the bag model ( with a bag radius r=0.6 fm ) @xcite , with and without a phenomenological singlet term added to fit the proton data @xcite .
the dashed curve was then a prediction for @xmath1 published before the data .
recall that the earlier phenomenology of carlitz - kaur @xcite and schfer @xcite used counting rules to parameterize the effect of admixing polarised gluons in the proton wave function .
as we have already noted , this complements the work of etar - see also refs.@xcite .
it would be desirable to have a full microscopic calculation , including this effect and the close - thomas mechanism , within a model which is consistent with low energy hadronic properties . as a final consideration in calculating the shape of the spin distributions we must also mention the role of chiral symmetry .
chiral quark models , like the cloudy bag@xcite , have proven very successful in dealing with low - energy hadronic properties .
the charge form - factor of the neutron , @xmath69 , is a famous example , where the positive core and negative tail of the corresponding charge density is a first - order effect in the perturbative dressing of the nucleon bag through the process @xmath70@xcite .
( indeed , within the cloudy bag model an accurate measurement of the zero in @xmath71 would determine the radius within which the quarks are confined . ) for the spin problem the role of the pion cloud is more complicated .
for the proton the dominant , pionic correction is @xmath72 , while for the neutron it is @xmath73 . combined with the reduction in the bare nucleon contribution through the wave function renormalisation constant , z , both of these effects tend to reduce the positive values of @xmath26 at large x @xcite .
the only complete investigation of pionic corrections has been carried out for the ellis - jaffe sum - rule @xcite , where it gives a small reduction ( @xmath74 ) .
this reduction would be larger were it not for an @xmath75 interference term .
the x - dependence of the latter is a completely new term which has never been calculated .
in order to study quark - parton distributions in the neutron , one must deal with either the deuteron or some heavier nucleus . in the spin - dependent case
the two nuclei studied are polarised deuterium ( @xmath76 ) and @xmath77 .
the simplest theoretical approach for deuterium is to approximate it as a free neutron and proton with polarisation ( @xmath78 ) - where @xmath79 is the deuteron d - state probability . at the next level of sophistication
we have a convolution of the free nucleon structure function with the light - cone momentum distribution of nucleons in the nucleus @xcite : g_1d(x , q^2)=_n = n , p_x f_n / d(y ) g_1n(,q^2 ) .
strangely the situation for polarised structure functions is better than for the unpolarised case , where data is still often analysed @xcite using smearing functions " pre - dating @xcite the original emc effect @xcite - and therefore ignoring binding corrections @xcite ( see however ref.@xcite ) .
tokarev found that for the polarised case the convolution gave a result very close to that obtained with a constant depolarisation for @xmath80 @xcite - see also kaptari and umnikov @xcite .
reassuring as these results may be , the only way to test the reliability of the convolution is to do better . even within the impulse approximation ( ia )
one may ask , for example , whether off - shell effects might spoil the convolution form . for the unpolarised case this was studied by melnitchouk , thomas and schreiber @xcite who found that even in the bjorken limit the dirac structure of the nucleon involves not one but three functions ^=p_t^(_0(p , q)+_1(p , q)+_2(p , q ) ) . thus , whereas the free structure function involves one linear combination f_2^freem_0+m^2_1+pq_2 , obtained by tracing @xmath81 with @xmath82 , in a nucleus , where the nucleon propagator is @xmath83 , we find f_2^boundb_0+ap_1+aq_2 , which is proportional to @xmath84 . to study the practical significance of this breakdown , one needs a covariant description of the @xmath85 vertex as well as the off - shell nucleon structure function . for the former we can use the vertex of refs.@xcite . for the latter melnitchouk , piller and
thomas @xcite have employed a covariant version of the physics presented in sect .
3 . that is ,
they constructed a set of phenomenological functions for the @xmath86 ( di - quark ) vertices which were fitted to free nucleon data .
as before , the di - quark states were either scalar or pseudovector .
after considerable dirac algebra one finds that the anti - symmetric nucleon tensor , surviving for massless quarks , in the bjorken limit is _ a - s^=i ^ \{q_p__5 g_(p ) + q_p_g_(q ) + q___5 g _ ( ) } , so that g_1n^freepq . for the deuteron case one must trace @xmath87 with a term involving @xmath88 , with the result : g_1d(x)_n = n , pdy dp^2\{aq + pq ap g_(p)^n}. the last term is not present in the free case , so that in principle the convolution must fail .
figure 2 shows the numerical results obtained by melnitchouk et al .
@xcite for the ratio of the deuteron to isoscalar nucleon spin structure function . the full calculation ( solid line ) is very close to the simple approximation of a constant depolarisation ( dotted line ) for @xmath89 .
this agrees with the result found in earlier work using the convolution approach .
indeed we see from the dashed curve that an appropriate convolution is very close to the full off - shell calculation for all @xmath90 . in practice the additional term in equ.(24 )
is of order @xmath91 and negligible for the deuteron . in the end
the major uncertainty in extracting the neutron structure function is not the failure of convolution or even the assumption of a constant depolarisation , but the old chestnut of the lack of knowledge of the deuteron d - state probability .
this ranges between 4 and 7% , with modern , non - local potentials preferring @xmath92 . at @xmath93 , for example , a variation of 1% in @xmath79 corresponds to an error of roughly 10% in @xmath1 @xcite . at present
this is smaller than the experimental errors but will soon be significant . to conclude this discussion , we must note that the ia is by no means secure . for heavier nuclei
it has been shown to significantly overestimate the effect of binding because deep - inelastic scattering determines the energy and momentum distribution of quarks nucleons @xcite ! this has not yet been investigated for the deuteron where the ia may not be so bad .
theoretical studies of @xmath94 have so far not gone beyond an essentially non - relativistic convolution because of the absence of a relativistic model of the necessary vertices . in view of the deuteron
results it would be surprising if there were a major problem , nevertheless it should be checked .
finally we note that the actual structure of the bound nucleon may change .
for the unpolarised case this has been shown to be a very small effect @xcite within the guichon model @xcite .
however , spin structure functions involve the lower components of the quark wave functions more directly and these change most in nuclear matter .
it has been suggested in ref.@xcite , for example , that @xmath95 differs from the free value by about 5% .
it is therefore an urgent matter to check this issue .
as we have seen there has been remarkable progress in the experimental determination of the spin structure functions of the proton and the neutron .
the bjorken sum - rule has been verified at the 10% level ( at 1@xmath34 ) . in view of its significance
it is very important to continue to reduce this error . with respect to the ellis - jaffe sum - rule
it is now clear that there is no longer a ` crisis ' in the sense that none of the nucleon spin is carried by quarks . with a reasonable estimate of the non - perturbative corrections
we find that all experiments are consistent with @xmath96 compared with the expectation of about 60% of the nucleon spin carried by quarks .
however , _ the end of the crisis is the beginning of some very challenging and fascinating physics_. it is now known that the regge behaviour usually assumed in extrapolating to @xmath97 is not appropriate to the flavour singlet component , which may have logarithmic corrections @xcite .
bass and landshoff suggest these might contribute as much as 25% to @xmath9 @xcite , while close and roberts have suggested recently that the error might be even larger@xcite .
there is as yet no consensus on the size of the higher twist corrections .
both of these issues need thorough experimental study of the @xmath98 and @xmath28 variation of the structure functions as well as more theoretical work .
the shape of the @xmath98 distribution of the neutron is still not well known . in view of the beautiful link between low energy hadron structure and parton distributions outlined in sect .
3 it is very important to determine whether @xmath1 does become positive at intermediate and large @xmath98 . related to this is the question of whether strength is shifted from intermediate @xmath98 to the region below 0.1 , and if so what mechanism is responsible .
the role of the axial anomaly is particularly fundamental , and can in principle be determined by comparing @xmath26 with a c - odd structure function like @xmath99 @xcite .
the use of light nuclear targets to extract information on the neutron seems to be well under control .
we have seen that an appropriate convolution formula is very accurate when compared with a full off - shell calculation . the major problem for the deuteron is the uncertainty over its d - state probability .
this may also be a problem for @xmath77 .
shadowing and meson exchange corrections have been investigated in detail for the unpolarised case @xcite but much more must be done for the polarised case @xcite .
there is so far no estimate of the effect on @xmath26 of a possible change in the internal nucleon structure . throughout our discussion
we have followed the conventional approach of ignoring charm quarks ( i.e. @xmath100 is set to zero ) .
since most data points are actually above charm threshold , this needs more investigation @xcite .
a related problem concerns the suggestion of using neutrino proton elastic scattering to measure @xmath37 . as the neutral current couples to a non - singlet axial current
it is less than obvious that such an experiment can yield information on a singlet quality like @xmath46 .
thus one may learn about @xmath101 rather than @xmath37 @xcite .
the breadth of open questions in this field is enormous - ranging from issues of quark structure in nuclei , to pomerons and the axial anomaly .
it is truly an appropriate meeting ground for nuclear and particle physicists and an appropriate beginning to the first joint spin conference .
i am grateful to the organising committee(s ) for the invitation to present these thoughts .
it is a pleasure to thank my colleagues , particularly s. bass , w. melnitchouk , g. piller , a. schreiber and f. steffens , who have taught me a great deal about the matters presented here .
i would also like to thank a. rawlinson and a. shaw for their assistance in the production of this manuscript .
this work was supported by the australian research council .
000 j. ashman et al .
( emc ) , phys .
b206 ( 1988 ) 364 ; nucl phys .
b328 ( 1990 ) 1 .
r. windmolders ( smc ) , proceedings spin 94 ; d. adams et al .
( smc ) , cern ppe/94 - 57 ; b. adeva et al .
b320 ( 1994 ) 400 .
d. day ( e132 & e143 ) , proceedings spin 94 .
ioffe , preprint itep-61 ( 1994 ) , hep - ph/9408291 .
s. bass and a.w .
thomas , prog .
nucl . phys .
33 ( 1994 ) 449 ; j. phys .
g9 ( 1993 ) 925 .
r. windmolders , int . j. mod
a7 , ( 1992 ) 639 .
j. ellis , in proceedings panic xiii , ed .
a. pascolini ( world scientific , singapore,1994 ) p.48 .
g. veneziano , okubofest lecture in from symmetries to strings " , ed .
a. das ( world scientific , singapore , 1990 ) .
j. kodaira et al . , phys .
d20 ( 1979 ) 627 ; nucl .
( 1980 ) 129 .
ahmed and g.g .
ross , nucl .
b111 , ( 1976 ) 441 .
larin and j.a.m .
vermaseren , phys .
lett . b259 ( 1991 )
345 . c. itzykson and j .- b .
zuber , quantum field theory " ( mcgraw - hill , new york,1980 ) .
j. ellis and r.l .
jaffe , phys .
d9 ( 1974 ) 1444 ; ibid .
d10 ( 1974 ) 1669 .
brodsky , j. ellis and m. karliner , phys .
b206 ( 1988 ) 309 .
s.d . drell and a.c .
hearn , phys .
( 1966 ) 908 ; s.b .
gerasimov , yad .
fiz . 2 ( 1965 ) 598 .
burkert and b.l .
ioffe , phys .
b296 ( 1992 ) 223 ; m. anselmino , b.l . ioffe and e.
leader , sov .
49 ( 1989 ) 136 . j. ellis and m. karliner , phys .
b313 ( 1993 ) 131 ; i.i .
balitsky et al . , phys .
b242 ( 1990 ) 245 ; ( e ) b318 ( 1993 ) 648 .
efremov and o.v .
teryaev , dubna preprint jinr - e2 - 88 - 287 ( 1988 ) ; g. altarelli and g.g .
ross , phys .
b212 ( 1988 ) 391 .
carlitz , j.c .
collins and a.h .
mueller , phys .
b214 ( 1988 ) 229 .
s.j . brodsky and i. schmidt ,
b234 , ( 1990 ) 144 .
f. myhrer and a.w .
thomas , phys .
d38 ( 1988 ) 1633 . h. hogassen and f. myhrer , z. phys .
c48 ( 1990 ) 295 .
crewther , acta physica austriaca suppl .
xix ( 1978 ) 47 . g.t . bodwin and j. qiu ,
d41 ( 1990 ) 2755 .
jaffe and a. manohar , nucl .
b337 ( 1990 ) 509 .
bass et al . , j. moscow phys .
soc . 1 ( 1991 ) 317 .
l. mankiewicz , phys .
d43 ( 1991 ) 64 .
bass , n.n .
nikolaev and a.w .
thomas , adp-90 - 133/t80 ( 1990 ) .
s.d . bass and a.w .
thomas , nucl .
a527 ( 1991 ) 519 .
bass , z. phys .
c55 ( 1992 ) 653 ; ibid .
c60 ( 1993 ) 343 .
jaffe , nucl .
b229 ( 1983 ) 205 .
a.i . signal and a.w .
thomas , phys .
d40 ( 1989 ) 2832 .
schreiber , a.w .
thomas and j.t .
londergan , phys .
rev . d42 ( 1990 ) 2653 .
schreiber , a.i .
signal and a.w .
thomas , phys . rev .
d44 ( 1991 ) 2653 .
r. parisi and g. petronzio , phys .
lett . b62 ( 1976 ) 331 .
thomas , prog .
phys . 20 ( 1988 ) 21 .
f.e . close and a.w .
thomas , phys .
b212 ( 1988 ) 227 .
e. sather , phys .
b274 ( 1992 ) 433 .
e. rodionov , a.w .
thomas and j.t .
londergan , mod .
a9 ( 1994 ) 1799 .
s.d . bass and a.w .
thomas , phys .
b312 ( 1993 ) 345 .
r. carlitz and j. kaur , phys .
38 ( 1977 ) 673 .
a. schfer , phys .
b208 ( 1988 ) 175 .
thomas , adv .
( 1984 ) 1 ; g.a.miller , int . rev . nucl . phys . 1 ( 1984 ) 189 ; s. thberge et al .
d22 ( 1980 ) 2838,(e ) d23 ( 1981 ) 2106 .
s. thberge et al . , can . j. phys . 60 ( 1982 ) 59 .
schreiber et al . , phys .
d45 ( 1992 ) 3069 .
a.w . schreiber and a.w .
thomas , phys .
b215 ( 1988 ) 141 .
woloshyn , nucl .
phys . a496 ( 1989 ) 749 . c. ciofi degli atti , e. pace and g. salme , phys .
c46 ( 1992 ) r1591 .
whitlow et al . , phys .
b282 ( 1992 ) 475 .
frankfurt and m.i .
strikman , phys .
76 ( 1981 ) 215 ; ibid . 160 ( 1988 ) 235 .
aubert et al .
b123 ( 1983 ) 275 ; nucl .
b293 ( 1987 ) 740 .
bickerstaff and a.w .
thomas , j. phys .
g15 ( 1989 ) 1523 .
w. melnitchouk , a.w . schreiber and a.w .
thomas , phys .
d49 ( 1994 ) 1183 .
w. melnitchouk , a.w . schreiber and a.w .
thomas , phys . lett .
b335 ( 1994 ) 11 .
kaptari and a.yu .
umnikov , phys .
b259 ( 1991 ) 155 .
tokarev , phys .
b318 ( 1993 ) 559 .
kaptari and a.yu .
umnikov , phys .
b240 ( 1990 ) 203 .
buck and f. gross , phys .
d20 ( 1979 ) 2361 ; r.g .
arnold , c.e .
carlson and f. gross , phys .
c21 ( 1980 ) 1426 .
w. melnitchouk , g. piller and a.w .
thomas , adelaide preprint adp-94 - 16/t157 .
thomas et al .
lett . b233 ( 1989 ) 43 .
k. saito , a. michels and a.w .
thomas , phys .
c46 ( 1992 ) r2149 ; k. saito and a.w .
thomas , nucl .
a547 ( 1994 ) 659 .
guichon , phys .
b200 ( 1988 ) 235 .
close and g.g .
roberts , phys .
lett . 60 ( 1988 ) 1471 .
s.d . bass and p.v .
landshoff , cavendish preprint hep-94/4 .
f.e . close and r.g .
roberts , ral report ral-94 - 071 ( 1994 ) .
n.n . nikolaev and v.i .
zakharov , phys .
b55 ( 1975 ) 397 ; n.n .
nikolaev and v.r .
zoller , z. phys .
c56 ( 1992 ) 623 ; b. badalek and j. kwiecinski , nucl .
b370 ( 1992 ) 278 ; w. melnitchouk and a.w .
thomas , phys .
d47 ( 1993 ) 3783 ; phys .
b317 ( 1993 ) 437 .
kaptari et al . ,
b321 ( 1994 ) 271 . h. khan and p. hoodbhoy , phys .
b298 ( 1993 ) 181 .
j. ellis , m. karliner and s. sachrajda , phys .
lett . b231 ( 1989 ) 497 .
s.d.bass and a.w .
thomas , phys .
b293 ( 1992 ) 457 . | we review recent progress in the understanding of the spin structure of the nucleon . for the free nucleon
the issues addressed include the status of the bjorken and ellis - jaffe sum - rules and the role of the axial anomaly .
we outline recent work connecting the quark models familiar from hadron spectroscopy to the spin and flavour dependence of the parton distributions .
finally we review the current understanding of nuclear spin structure functions and particularly the extraction of the neutron spin structure function from deuteron data .
= 5.5truein = 8.5truein adp-94 - 21/t161 + invited paper at spin 94 + bloomington , september 1994 |
social networks have certainly become an important center of attention in our modern information society by transforming human relationships into a huge interchange of , very often , _ sensitive _ data .
there are many truly beneficial consequences when social network data are released for justified mining and analytical purposes .
for example , researchers in sociology , economics and geography , as well as vendors in service - oriented systems and internet advertisers can certainly benefit and improve their performances by a fair study of the social network data .
but , such benefits are definitely _ not _ free of cost as dishonest individuals or organizations may compromise the _ privacy _ of its users while scrutinizing a public social network and may deliberately use such privacy violations for harmful or other unfair commercial purposes . a common way to handle this kind of unwelcome intrusion on the user s privacy
is to somehow _ anonymize _ the data by removing most potentially identifying attributes .
however , even after such anonymization , often it may still be possible to infer many sensitive attributes of a social network that may be linked to its users , such as node degrees , inter - node distances or network connectivity , and therefore _ further _ privacy - preserving methods need to be investigated and analyzed .
these additional privacy - preserving methods of social networks are based on the concept of @xmath0-anonymity introduced for microdata in @xcite , aiming to ensure that _ no _ record in a database can be re - identified with a probability higher than @xmath1 .
crucial to modelling a social network anonymization process are of course the adversary s background knowledge of any object and the structural information about the network that is available . for example , assuming the involved social network as a simple graph in which individuals are represented by nodes and relationships between pairs of individuals are represented by edges , the adversary s background knowledge about a target ( a node ) could be the node degree @xcite , the node neighborhood @xcite , _ etc_. in such scenarios , it frequently suffices to develop attacks to re - identify the individuals and their relationships .
such attacks are usually called _ passive _ ( see @xcite for more information ) .
some examples of passive attacks and the corresponding privacy - preserving methods for social networks can be found in references @xcite .
in contrast , backstrom _
_ introduced the concept of the so - called _ active _ attacks in @xcite .
such attacks are mainly based on creating and inserting in a network some nodes ( the `` attacker nodes '' ) under control by the adversary .
these attacker nodes could be newly created accounts with pseudonymous or spoofed identities ( commonly called sybil nodes ) , or existing legitimate individuals in the network which are in the adversary s proximity .
the goal is then to establish links with some other nodes in the network ( or even links between other nodes ) in order to create some sort of `` fingerprints '' in the network that will be further released .
clearly , once the releasing action has been achieved , the adversary could retrieve the fingerprints already introduced , and use them to re - identify other nodes in the network .
et al . _ in @xcite showed that @xmath2 attacker nodes in a network could in fact seriously compromise the privacy of any arbitrary node . in recent years
, several research works have appeared that deal with decreasing the impact of these active attacks ( see , for instance , @xcite ) . for other related publications on privacy - preserving methods in social networks ,
see @xcite .
there are already many well - known active attack strategies for social networks in order to find all possible vulnerabilities .
however , somewhat surprisingly , not many prior research works have addressed the goal of measuring how resistant is a given social network against these kinds of active attacks to the privacy . to this effect ,
very recently a novel privacy measure for social networks was introduced in @xcite .
the privacy measure proposed there was called the _ @xmath3-anonymity _ , where @xmath0 is a number indicating a privacy threshold and @xmath4 is the _ maximum _ number of attacker nodes that can be inserted into the network ; @xmath4 may be estimated through some statistical methods - anonymity where @xmath4 represented the number of common neighbors of two nodes . ] .
trujillo - rasua and yero in @xcite showed that graphs satisfying @xmath3-anonymity can prevent adversaries who control at most @xmath4 nodes in the network from re - identifying individuals with probability higher than @xmath1 .
this privacy measure relies on a graph parameter called the @xmath0-metric anti - dimension .
consider a simple connected unweighted graph @xmath5 and let @xmath6 be the length ( number of edges ) of a shortest path between two nodes @xmath7 .
for an ordered sequence @xmath8 of nodes of @xmath9 and a node @xmath10 , the vector @xmath11 is called the _ metric representation _ of @xmath12 with respect to @xmath13 .
based on the above definition , a set @xmath14 of nodes is called a @xmath0-_anti - resolving set _ for @xmath9 if @xmath0 is the largest positive integer such that for every node @xmath15 there exist at least @xmath16 different nodes @xmath17 such that @xmath18 , _
i.e. _ , @xmath12 and @xmath19 have the same metric representation with respect to @xmath13 . the @xmath0-_metric anti - dimension _ of @xmath9 , denoted by @xmath20 , is then the minimum cardinality of any @xmath0-anti - resolving set in @xmath9 .
note that @xmath0-anti - resolving sets may _ not _ exist in a graph for every @xmath0 .
the connection between @xmath21-anonymity privacy measure and the @xmath0-metric anti - dimension can be understood in the following way .
suppose that an adversary takes control of a set of nodes @xmath13 of the graph ( _ i.e. _ , @xmath13 plays the role of attacker nodes ) , and the background knowledge of such an adversary regarding a target node @xmath12 is the metric representation of the node @xmath12 with respect to @xmath13 .
the @xmath3-anonymity privacy measure is then a privacy metric that naturally evolves from the adversary s background knowledge .
intuitively , if @xmath13 ( the attacker nodes of an adversary ) is a @xmath0-anti - resolving set then the adversary can not uniquely re - identify other nodes in the network ( based on the metric representation ) from these attacker nodes with a probability higher than @xmath1 ( based on uniform sampling of other nodes ) , and if the @xmath0-metric anti - dimension of the graph is @xmath4 then the adversary must use at least @xmath4 attacker nodes to get the probability of privacy violation down to @xmath1 .
there is a rich literature on theoretical investigations of privacy measures and privacy preserving computational models in several other application areas such as multi - party communications , distributed computing and game - theoretic settings ( _ e.g. _ , see @xcite ) .
however , none of these settings apply directly to our application scenario of active attack model for social networks .
the differential privacy model , introduced by dwork @xcite in the context of privacy preservation in statistical databases against malicious database queries , works by computing the correct answer to a query and adding a noise drawn from a specific distribution , and is quite different from the anonymization approach studied in this paper .
it is obviously desirable to know how secure a given social network is against active attacks .
this necessitates the study of computational complexity issues for computing @xmath3-anonymity .
currently known results only include some heuristic algorithms with no provable guarantee on performances such as in @xcite , or algorithms for very special cases .
in fact , it is not even known if any version of the related computational problems is @xmath22-hard . to this effect , we formalize three computational problems related to measuring the @xmath3-anonymity of graphs and present non - trivial computational complexity results for these problems . the rest of the paper is organized as follows : 1 . in section [ sec - termi ]
we review some basic terminologies and notations and then present the three computational problems that we consider in this paper . for the benefit of the reader
, we also briefly review some standard algorithmic complexity concepts and results that will be used later .
2 . in section [ sec - statementofresults ] ,
we state the results in this paper mathematically precisely along with some informal remarks .
we group our results based on the problem definitions and the expected size of the attacker nodes .
sections [ sec - pfb][sec - pfe ] are devoted to the proofs of the results stated in section [ sec - statementofresults ] .
we finally conclude in section [ sec - conc ] with some possible future research directions .
in this section , we first describe the terminologies and notations required to describe our computational problems , and subsequently describe several versions of the problems we consider . an example to illustrate the notations in section [ sec - termi-2 ] . ]
let @xmath5 be our _ undirected unweighted _ input graph over @xmath23 nodes @xmath24 .
we use @xmath25 to denote the distance ( number of edges in a shortest path ) between nodes @xmath26 and @xmath27 . for illustrating various notations
, we use the example in fig .
[ fignew ] . 1 . @xmath28 . for example , @xmath29 .
@xmath30 is the _ diameter _ ( length of a longest shortest path ) of the graph @xmath5 .
for example , @xmath31 .
@xmath32 is the ( open ) _ neighborhood _ of node @xmath33 in @xmath5 .
for example , @xmath34 .
4 . for a subset of nodes @xmath35 and any @xmath36
, @xmath37 denotes the metric representation of @xmath26 with respect to @xmath38 , _
i.e. _ , the vector of @xmath39 elements obtained from @xmath40 by deleting @xmath25 for every @xmath41 .
for example , @xmath42 .
@xmath43 for any @xmath44 .
for example , if @xmath45 then @xmath46 . 6 .
@xmath47 is a partition of @xmath48 if and only if @xmath49 and @xmath50 for @xmath51 .
partition @xmath52 is called a _ _ refinement__. ] of partition @xmath53 , denoted by @xmath54 , provided @xmath55 and @xmath56 can be obtained from @xmath53 in the following manner : 1 . for every node @xmath57 ,
remove @xmath26 from the set containing it in @xmath53 .
optionally _ , for every set @xmath58 in @xmath53 , replace @xmath58 by a partition of @xmath58 .
3 . remove empty sets , if any .
+ for example , if @xmath59 and @xmath60 then @xmath54 . 7 .
the equality relation over a set of vectors , all of same length , obviously defines an _
equivalence relation_. the following notations are used for such an equivalence relation over the set of vectors @xmath61 for some @xmath62 .
the set of equivalence classes , which forms a partition of @xmath61 , is denoted by @xmath63 .
for example , + @xmath64 .
2 . abusing terminologies slightly , two nodes @xmath65 will be said to belong to the _ same _ equivalence class if @xmath37 and @xmath66 belong to the same equivalence class in @xmath63 , and thus @xmath63 also defines a partition into equivalence classes of @xmath67 .
for example , @xmath68 will also denote @xmath69 .
the _ measure _ of the equivalence relation is defined as @xmath70 .
thus , if a set @xmath13 is a @xmath0-anti - resolving set then @xmath71 defines a partition into equivalence classes whose measure is _ exactly _ @xmath0 .
for example , @xmath72 and @xmath73 is a @xmath74-anti - resolving set .
it is obviously desirable to know how secure a given social network is against active attacks .
this necessitates the study of computational complexity issues for computing @xmath3-anonymity . to this effect , we formalize three computational problems related to measuring the @xmath3-anonymity of graphs for all the problem versions , let @xmath5 be the ( connected undirected unweighted ) input graph representing the social network under study .
[ prob1 ] given @xmath9 , find a subset of nodes @xmath38 that _ maximizes _ @xmath75 .
* notation related to problem [ prob1 ] * @xmath76 .
problem [ prob1 ] simply finds a @xmath0-anti - resolving set for the largest possible @xmath0 .
intuitively , it sets an absolute bound on the privacy violation probability of an adversary assuming that the adversary can use _ any _ number of attacker nodes . in practice , however , the number of attacker nodes employed by the adversary may be limited , which leads us to the second problem formulation stated below .
[ prob2 ] given @xmath9 and a positive integer @xmath0 , find a subset of nodes @xmath38 of _ minimum _ cardinality such that @xmath77 , if such a @xmath38 exists .
* notation and assumption related to problem [ prob2 ] * @xmath78 for some @xmath79 . if @xmath80 for _ no _
@xmath38 then we set @xmath81 and @xmath82 .
problem [ prob2 ] finds a @xmath0-anti - resolving set for largest @xmath0 while simultaneously minimizing the number of attacker nodes . the remaining third version of our problem formulation relates to a trade - off between privacy violation probability and the corresponding minimum number of attacker nodes needed to achieve such a violation .
to understand this motivation , suppose that @xmath9 has a @xmath0-metric anti - dimension of @xmath4 , a @xmath83-metric anti - dimension of @xmath84 , @xmath85 and @xmath86 .
then , this provides a trade - off between privacy and number of attacker nodes , namely we may allow a smaller privacy violation probability @xmath87 but the network can tolerate _ adversarial control _ of a _ fewer _ number @xmath84 of nodes or we may allow a larger privacy violation probability @xmath1 but the network can tolerate adversarial control of a larger number @xmath4 of nodes .
such a trade - off may be crucial for a network administrator in administering privacy of a network or for an individual in its decision to join a network . clearly , this necessitates solving a problem of the following type .
[ prob3 ] given @xmath9 and a positive integer @xmath0 , find a subset of nodes @xmath38 of _ minimum _ cardinality such that @xmath88 , if such a @xmath38 exists .
* notation and assumption related to problem [ prob3 ] * @xmath89 for some @xmath90 . if @xmath88 for _ no _
@xmath38 then we set @xmath91 and @xmath92 for the benefit of the reader , we summarize the following concepts and results from the computational complexity theory domain .
_ we assume that the reader is familiar with standard @xmath93 , @xmath94 , @xmath95 and @xmath96 notations used in asymptotic analysis of algorithms @xmath97_e.g . _ , see @xcite@xmath98_. an algorithm @xmath99 for a minimization ( resp . ,
maximization ) problem is said to have an _ approximation ratio _ of @xmath100 ( or is simply an _ @xmath100-approximation _ )
@xcite provided @xmath99 runs in polynomial time in the size of its input and produces a solution with an objective value _ no larger than _ @xmath100 times ( resp .
, _ no smaller than _ @xmath101 times ) the value of the optimum .
dtime@xmath102 refers to the class of problem that can be solved by a deterministic algorithm running in @xmath103 time when @xmath23 is the size of the input instance ; it is widely believed that @xmath104dtime@xmath103 . the minimum set - cover problem ( sc ) is a well - known combinatorial problem that is defined as follows @xcite .
our input is an universe @xmath105 of @xmath23 elements , and a collection of @xmath106 sets @xmath107 over this universe with @xmath108 .
a valid solution of sc is a subset of indices @xmath109 such that every element in @xmath110 is `` covered '' by a set whose index is in @xmath111 , _
i.e. _ , @xmath112 .
the objective of sc is to _ minimize _ the number @xmath113 of selected sets .
we use the notation @xmath114 to denote the size ( number of sets ) in an optimal solution of an instance of sc . on the inapproximability side ,
sc is @xmath22-hard @xcite and , assuming @xmath115dtime@xmath102 , sc does not admit a @xmath116-approximation for any constant @xmath117 @xcite . on the algorithmic side , sc admits a @xmath118-approximation using a simple greedy algorithm @xcite that can be easily implemented to run in @xmath119 time @xcite .
in this section we provide precise statements of our results , leaving their proofs in sections [ sec - pfb][sec - pfe ] .
[ thm1 ] + * ( a ) * both adim and adim@xmath120 can be solved in @xmath121 time . *
( b ) * both adim and adim@xmath120 can also be solved in @xmath122 time `` with high probability '' @xmath97_i.e .
_ , with a probability of at least @xmath123 for some constant @xmath124@xmath98 .
the randomized algorithm in theorem _ [ thm1](b ) _ runs faster that the deterministic algorithm in theorem _ [ thm1](a ) _ provided @xmath125 .
[ k - large ] + * ( a ) * adim@xmath126 is @xmath22-complete for any integer @xmath0 in the range @xmath127 where @xmath128 is any arbitrary constant , even if the diameter of the input graph is @xmath129 . *
( b ) * assuming @xmath130__dtime__@xmath131 , there exists a universal constant @xmath132 such that adim@xmath126 does not admit a @xmath133-approximation for any integer @xmath0 in the range @xmath127 where @xmath128 is any arbitrary constant , even if the diameter of the input graph is @xmath129 . *
( c ) * if @xmath134 for some constant @xmath135 then @xmath136 if a solution exists and adim@xmath126 can be solved in polynomial time .
[ rem - opt ] + * ( a ) * for @xmath137 , the inapproximability ratio in theorem _ [ k - large](a ) _ is asymptotically optimal up to a constant factor because of the @xmath138-approximation of adim@xmath139 in theorem _ [ k - one](a)_. * ( b ) * the result in theorem _ [ k - large](b ) _ provides a much stronger inapproximability result compared to that in theorem _ [ k - large](a ) _ at the expense of a slightly weaker complexity - theoretic assumption @xmath97_i.e . _ , @xmath140__dtime__@xmath131 vs. @xmath141@xmath98 .
note that even when @xmath137 adim@xmath126 is @xmath22-hard and even hard to approximate within a logarithmic factor due to theorem [ k - large ] .
we show the following algorithmic results for adim@xmath126 when @xmath137 .
[ k - one ] + * ( a ) * adim@xmath142 admits a @xmath143-approximation in @xmath144 time . *
( b ) * if @xmath9 has at least one node of degree @xmath74 then @xmath145 and thus adim@xmath142 can be solved in @xmath144 time . *
( c ) * if @xmath9 does not contain a cycle of @xmath146 edges then @xmath147 and thus adim@xmath142 can be solved in @xmath144 time .
* ( a ) * we first consider the claim for adim@xmath120 .
we begin by proving some structural properties of valid solutions for adim@xmath120 .
[ prop1 ] consider two subsets of nodes @xmath148 .
let @xmath149 be two nodes such that they do not belong to the same equivalence class in @xmath150 .
then @xmath26 and @xmath27 do not belong to the same equivalence class in @xmath151 also .
since @xmath26 and @xmath27 are not in the same equivalence class in @xmath150 , we have @xmath152 which in turn implies ( since @xmath153 ) @xmath154 which implies @xmath26 and @xmath27 are not in the same equivalence class in @xmath151 . [ cor1 ] proposition [ prop1 ] implies @xmath155 . note that @xmath155 in corollary [ cor1 ] does not necessarily imply that @xmath156 .
the following proposition gives some condition for this to happen .
[ prop2 ] consider two subsets of nodes @xmath148 , and let @xmath157 be the only @xmath158 equivalence classes @xmath97subsets of nodes@xmath98 in @xmath150 such that @xmath159 .
then , _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 1 .
@xmath160 implies @xmath161 , and 2 . if @xmath162 for some @xmath163 then @xmath164 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ since @xmath165 , there exists a node @xmath166 such that @xmath167 and @xmath168 . similarly , since @xmath169 , there exists a node @xmath170 such that @xmath171 and @xmath172 . by corollary [ cor1 ] , @xmath155 and thus the following implications
hold : * if @xmath173 then @xmath151 contains an equivalence class ( subset of nodes ) @xmath174 such that @xmath175 .
this implies @xmath176 . *
if there exists a @xmath177 such that @xmath162 then @xmath151 contains an equivalence class @xmath178 with @xmath179 .
this implies @xmath180 .
based on the above structural properties , we design algorithm i for adim@xmath120 as shown below .
r l l + & + [ 5pt ] * 2 .
* & + [ 5pt ] * 3 . * & + [ 5pt ] * 3.1 * & & @xmath181 ; @xmath182 + [ 5pt ] * 3.2 * & & * while * @xmath183 @xmath184
@xmath185 @xmath186 * do * + [ 5pt ] * 3.2.1 * & & compute @xmath187 + [ 5pt ] * 3.2.2 * & & * if * @xmath188 and @xmath189 + * 3.2.3 * & & * then*@xmath190 @xmath191 ; @xmath192 ; @xmath193 + [ 5pt ] * 3.2.4 * & & * else*@xmath194 let @xmath195 be the _ only _ @xmath158 equivalence classes ( subsets of nodes ) + & & in @xmath63 such that @xmath196 + [ 5pt ] * 3.2.5 * & & @xmath197 + [ 5pt ] * 4 . * & + algorithm
i returns an optimal solution for adim@xmath120 .
assume that @xmath198 since otherwise obviously our returned solution is correct .
fix any optimal solution ( subset of nodes ) @xmath199 of measure @xmath200 and select any arbitrary node @xmath201 . consider the iteration of the * for * loop in step 3 when @xmath26 is equal to @xmath33 .
we now analyze the run of _ this particular iteration_. let @xmath202 be the @xmath203 subsets of nodes that were assigned to @xmath38 in _ successive _ iterations of the * while * loop in step 3.2 .
we have the following cases to consider .
case 1 : @xmath204 for some @xmath205 .
: : then , our solution is a set @xmath206 such that + @xmath207 and @xmath208 .
case 2 : @xmath209 for any @xmath205 . : : since @xmath210 and @xmath211 for any @xmath205 , only one of the following cases is possible : + case 2.1 : there exists @xmath212 such that @xmath213 but @xmath214 . ; ; let + @xmath215 be all the @xmath216 equivalence classes ( subsets of nodes ) in @xmath217 such that @xmath218 .
now we note the following : + * by step 3.2.5 , @xmath219 . * thus , @xmath213 and @xmath214 implies @xmath220 , and therefore there exists an index @xmath221 such that @xmath222 .
let @xmath223 ( @xmath224 could be empty ) .
then , for some @xmath225 , @xmath226 is an equivalence class in @xmath227 implying @xmath228 since @xmath229 , we have @xmath230 + thus , @xmath231 and @xmath232 , contradicting the optimality of @xmath233 . case 2.2 : @xmath234 .
; ; if @xmath235 was set to @xmath236 at the last iteration of the * while * loop , then @xmath237 and @xmath238 , contradicting the optimality of @xmath233 .
thus , @xmath235 must have remained @xmath239 after the last iteration of the * while * loop , which implies @xmath240 .
let @xmath241 be all the @xmath216 equivalence classes ( subsets of nodes ) in @xmath242 such that @xmath243 . since @xmath244 , we have @xmath245
thus , @xmath246 contradicting our assumption of @xmath240 . algorithm i runs in @xmath121 time .
there are @xmath23 choices for the * for * loop in step 3 . for each such choice
, we analyze the execution of the * while * loop in step 3.2 .
the running time in each iteration of the * while * loop is dominated by the time taken to compute @xmath247 from @xmath248 .
suppose that @xmath249 . by corollary [ cor1 ] , @xmath250
thus , it follows that the _ total _ time to execute _ all _ iterations of the * while * loop _ for a specific choice _ of @xmath26 in step 3 is of the order of @xmath23 times the time taken to solve a problem of the following kind : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ for a subset of nodes @xmath251 , given @xmath252 and a node @xmath253 , compute @xmath254 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ since @xmath254 is a refinement of @xmath252 by corollary [ cor1 ] , we can use the following simple strategy .
for every set @xmath255 , we split @xmath256 into two or more parts , if needed , by doing a bucket - sort ( with @xmath23 bins ) in @xmath257 time on the sequence of values @xmath258 .
the total time taken for all sets in @xmath248 is thus @xmath259 .
this completes the proof for adim@xmath120 .
now we consider the claim for adim .
obviously , adim can be solved in @xmath260 time by solving adim@xmath120 for @xmath261 in this order and selecting the largest @xmath0 as @xmath262 for which @xmath263 .
however , we can modify the steps of algorithm i directly to solve adim in @xmath121 time , as shown in algorithm ii .
r l l + + * 1 . * & + [ 5pt ] * 2 . * & + [ 5pt ] * 3 . * & + [ 5pt ] * 3.1 * & & @xmath181 + [ 5pt ] * 3.2 * & & * while * @xmath264
* do * + [ 5pt ] * 3.2.1 * & & compute @xmath187 + [ 5pt ] * 3.2.2 * & & * if * @xmath265 + * 3.2.3 * & & * then * @xmath266 @xmath267 ; @xmath192 + [ 5pt ] * 3.2.4 * & & * else*@xmath194 let @xmath195 be the _ only _
@xmath158 equivalence classes ( subsets of nodes ) + & & in @xmath63 such that @xmath196 + [ 5pt ] * 3.2.5 * & & @xmath197 + [ 5pt ] * 4 . * & + the proof of correctness is very similar ( and , in fact simpler due to elimination of some cases ) to that of adim@xmath120 . * ( b ) * our solution is the obvious randomization of algorithm ii ( for adim@xmath120 ) or algorithm - ii ( for adim ) as shown below .
l l l l + * 1 . * & + [ 5pt ] * 2 .
* & + [ 3pt ] & or & & + [ 3pt ] & + [ 5pt ] * 3 . * & + [ 5pt ] & * 3.1 * & + [ 5pt ] & * 3.2 * & execute step 3.1 and step 3.2 ( and its sub - steps ) of algorithm i & ( for adim@xmath120 ) + [ 3pt ] & & or & + [ 3pt ] & & execute step 3.1 and step 3.2 ( and its sub - steps ) of algorithm ii & ( for adim ) + [ 5pt ] * 4 . * & + the success probability @xmath268 is given by @xmath269 \\ = 1- \pr\left[\!\!{\hbox{\rm \ $ v_i\notin { v_{\mathrm{opt}}^{\geq k } } $ in \emph{each } of the $ \left\lceil\frac{2 n\ln n}{k}\right\rceil$ iterations\ \/}}\!\!\right ] \geq 1- \left ( 1 - \frac{k}{n } \right)^ { \left\lceil\frac{2n\ln n}{k}\right\rceil } > 1 - \frac{1}{{{\mathbf{e}}}^{2\ln n } } = 1 - \frac{1}{n^2}\end{gathered}\ ] ]
* ( a ) * adim@xmath126 trivially belongs to @xmath22 for any @xmath0 , thus we need to show that it is also @xmath22-hard .
the standard @xmath22-complete _ minimum dominating set _ ( mds ) problem for a graph is defined as follows @xcite .
our input is a connected undirected unweighted graph @xmath5 .
a subset of nodes @xmath35 is called a _ dominating set _ if and only if every node in @xmath67 is adjacent to some node in @xmath38 .
the objective of mds is to find a dominating set of nodes of _ minimum _ cardinality .
let @xmath270 denote the cardinality of a minimum dominating set for a graph @xmath9 .
it is well - known that the mds and sc problems have precisely the same approximability via approximation - preserving reductions in both directions and , in particular , there exists a standard reduction from sc to mds as follows .
given an instance @xmath105 and @xmath107 of sc , we create the following instance @xmath271 of mds .
@xmath272 has an _ element node _
@xmath273 for every element @xmath274 and a _ set node _
@xmath275 for every set @xmath177 with @xmath276 .
there are two types of edges in @xmath277 .
every set node @xmath275 has an edge to every other set node @xmath278 and the collection of these edges is called the set of _ clique edges_. moreover , a set node @xmath275 is connected to an element node @xmath273 if and only if @xmath279 and the collection of these edges is called the set of _ membership edges_. a standard straightforward argument shows that @xmath280 is a solution of sc if and only if the collection of set nodes @xmath281 is a solution of mds on @xmath282 and thus @xmath283 . for the purpose of our @xmath22-hardness reduction , it would be more convenient to work with a restricted version of sc known as the _ exact cover by @xmath284-sets _ ( x3c ) problem . here
we have exactly @xmath23 elements and exactly @xmath23 sets where @xmath23 is a multiple of three , every set contains exactly @xmath284 elements and every element occurs in exactly @xmath284 sets .
obviously we need at least @xmath285 sets to cover all the @xmath23 elements . letting @xmath286 to denote the number of sets in an optimal solution of x3c , it is well - known that problem of deciding whether @xmath287 is in fact @xmath22-complete .
let @xmath288 be the real - valued solution of the quadratic equation @xmath289 .
note that since @xmath290 for some constant @xmath291 , we have @xmath292 , _
i.e. _ , @xmath23 and @xmath293 are `` polynomially related '' .
illustration of the @xmath22-hardness reduction in theorem [ k - large](a ) .
only a part of the graph @xmath9 is shown for visual clarity . ]
we assume without loss of generality that @xmath293 is an even integer , and start with an instance of x3c of @xmath294 elements and transform it to an instance graph @xmath271 having @xmath293 nodes of mds via the reduction outlined before . since @xmath294 is polynomially related to @xmath23 , such an instance of x3c is @xmath22-complete with respect to @xmath23 being the input size .
we reduce @xmath282 to an instance @xmath5 of adim@xmath126 in polynomial time as follows ( see fig . [ fig2 ] for an illustration ) : * we `` clone '' each element node @xmath295 to get @xmath296 copies , _
i.e. _ , every node @xmath297 is replaced by @xmath296 new nodes @xmath298 .
we refer to these nodes as _ clones _ of the element node @xmath297 ( or , sometimes simply as _ element - clone nodes _ ) .
there are precisely @xmath299 such nodes .
* we `` clone '' each set node @xmath300 to get @xmath296 copies , _
i.e. _ , every node @xmath275 is replaced by @xmath296 new nodes @xmath301 .
we refer to these nodes as _ clones _ of the set node @xmath275 ( or , sometimes simply as _ set - clone nodes _ ) .
there are precisely @xmath299 such nodes .
* we add @xmath0 new nodes @xmath302 .
we refer to these nodes as _
clique nodes_. * we add an edge between every pair of clique nodes @xmath303 and @xmath304 .
we refer to these edges as _
clique edges_. there are precisely @xmath305 such edges .
* we add an edge between every clique node and every non - clique node , _ i.e. _ , we add every edge in the set @xmath306 we refer to these edges as the _ partition - fixing _ edges .
there are precisely @xmath307 such edges .
* we add an edge between every pair of distinct element - clone nodes @xmath308 and @xmath309 .
we refer to these as the _ element - clone edges_. there are precisely @xmath310 such edges . * for every element @xmath311 and every set @xmath177 such that @xmath312 , we add the following @xmath313 edges : @xmath314 we refer to these edges as the _ non - member _ edges corresponding to the element node @xmath311 and the set node @xmath177 .
there are precisely @xmath315 such edges .
note that @xmath9 has precisely @xmath316 nodes and thus our reduction is polynomial time in @xmath23 .
since any clique node is adjacent to every other node in @xmath9 , it follows that @xmath317 .
we now show the validity of our reduction by showing that @xmath318 * proof of @xmath319 * consider an optimal solution @xmath320 of mds on @xmath282 with @xmath321 .
we now construct a solution @xmath35 of adim@xmath126 on @xmath9 by setting @xmath322 .
note that @xmath323 .
we claim that @xmath38 is a valid solution of adim@xmath126 by showing that ( _ a _ ) : : @xmath324 and ( _ b _ ) : : any other equivalence class in @xmath248 has at least @xmath0 nodes . to prove ( _ a _ ) ,
consider a clique node @xmath303 and any other non - clique node .
then , the following cases apply : * suppose that the non - clique node is a element - clone node @xmath325 for some @xmath326 and @xmath4 . since @xmath327 is a solution of mds on @xmath282 , there exists a set node @xmath328 such that @xmath329 and consequently @xmath330 .
this implies that there exists a node @xmath331 such that @xmath332 , and therefore @xmath308 _ can not _ be in the same equivalence class with @xmath303 . *
suppose that the non - clique node is a set - clone node @xmath333 .
pick any set - clone node @xmath334 . then ,
@xmath335 , and therefore @xmath336 _ can not _ be in the same equivalence class with @xmath303 . to prove ( _ b _ ) , note the following : * since @xmath317 , @xmath337 for any two _ distinct _ set - clone nodes @xmath338 and @xmath339 , and thus all the set nodes in @xmath67 belong together in the _ same _ equivalence class in @xmath248 .
there are at least @xmath340 such nodes in @xmath67 .
thus , any equivalence class that contains these set - clone nodes can not have less than @xmath0 nodes .
* consider now an equivalence class in @xmath248 that contains a copy @xmath341 of the element node @xmath273 for some @xmath342 and @xmath326 .
consider another copy @xmath343 of the element node @xmath273 for some @xmath344 .
for any set node @xmath331 , if @xmath345 then @xmath346 , whereas if @xmath347 then , since @xmath317 , it follows that @xmath348 .
thus , any equivalence class that contains at least one clone of an element node must contain all the @xmath349 clones of that element node and thus such an equivalence class can not have a number of nodes that is less than @xmath0 .
* proof of @xmath350 * since we know that @xmath351 is always at least @xmath352 , it suffices to show that @xmath353 . consider an optimal solution @xmath354 with @xmath355 .
since @xmath356 is a solution of adim@xmath126 on @xmath9 , there exists a subset of nodes , say @xmath357 , such that @xmath358 and @xmath359 .
[ no - element - set ] @xmath360 does not contain any set - clone or element - clone nodes and thus @xmath361 .
suppose that @xmath360 contains at least one element - clone node @xmath341 for some @xmath342 and @xmath326 .
but , @xmath362 contains at least @xmath363 other clones of the element node @xmath311 and all these clones must belong together with @xmath341 in the _ same _ equivalence class .
this implies @xmath364 , a contradiction .
similarly , suppose that @xmath360 contains at least one set - clone node @xmath365 for some @xmath342 and @xmath326 .
but , @xmath362 contains at least @xmath363 other clones of the set node @xmath366 and all these clones must belong together with @xmath365 in the _ same _ equivalence class .
this implies @xmath364 , a contradiction .
[ no - set - clone ] @xmath356 does not contain two or more clones of the same set node .
suppose that @xmath356 contains two set - clone nodes @xmath336 and @xmath339 of the same set node @xmath275 .
but , @xmath362 contains at least @xmath363 other clones of the element node @xmath311 and all these clones must belong together in the _ same _ equivalence class @xmath13 .
if we remove @xmath336 from @xmath356 then @xmath336 gets added to this equivalence class .
thus , such a removal produced another valid solution but with one node less than @xmath367 , contradicting the optimality of @xmath368 .
[ no - element - clone ] @xmath356 does not contain any element - clone node .
suppose that @xmath356 contains at least one element - clone node and thus at most @xmath369 set - clone nodes .
note that @xmath362 contains at least @xmath370 clones of every element node @xmath311 .
consider an element - clone node @xmath371 and a clique node @xmath304 .
since @xmath372 , there must be a node in @xmath356 such that the distance of this node to @xmath304 is different from the distance to @xmath373 .
such a node in @xmath356 can not be an element - clone node , say @xmath374 since @xmath375 .
since there is an edge between every set - clone node and every clique node , such a node must be a set - clone node , say @xmath376 for some @xmath377 and @xmath378 , such that @xmath379 , _
i.e. _ , @xmath380 .
since every set in x3c contains exactly @xmath284 elements and @xmath381 , there must then exist an element - clone node @xmath373 such that the distance of @xmath373 to any node in @xmath356 is exactly the same as the distance of @xmath304 to that node in @xmath356 .
this implies @xmath382 , contradicting proposition [ no - element - set ] . by proposition [ no - set - clone ] and proposition [ no - element - clone ]
, @xmath356 contains exactly one clone of a subset of set nodes . without loss of generality , assume that @xmath383 and let @xmath384 .
note that @xmath385 .
we are now ready to finish our proof by showing @xmath327 is indeed a valid solution of mds on @xmath282 .
suppose not , and let @xmath273 be an element - node that is not adjacent to any node in @xmath327 .
then , @xmath386 which contradicts proposition [ no - element - set ] . *
( b ) * the proof is similar to that of ( a ) but this time we start with a general version of sc as opposed to the restricted x3c version , and show that the reduction is approximation - preserving in an appropriate sense . in the sequel , we use the standard notation poly@xmath387 to denote a polynomial @xmath388 of @xmath23 ( for some constant @xmath124 ) .
we recall the following details of the inapproximability reduction of feige in @xcite .
given an instance formula @xmath389 of the standard boolean satisfiability problem ( sat ) , feige reduces @xmath389 to an instance @xmath390 of sc ( with @xmath391 ) in @xmath392 time such that the following properties are satisfied for any constant @xmath117 : * for some @xmath393 , either @xmath394 or @xmath395 . *
the reduction satisfies the following completeness and soundness properties : + ( completeness ) : : if @xmath389 is satisfiable then @xmath394 .
( soundness ) : : if @xmath389 is not satisfiable then @xmath396 . since @xmath391 , by adding duplicate copies of a set , if necessary , we can ensure that @xmath397 for some constant @xmath398 .
our reduction from sc to mds to adim@xmath126 is same as in ( a ) except that some details are different , which we show here .
* we start with an instance of sc as given by feige in @xcite with @xmath293 elements and @xmath399 sets , where @xmath400 is a real - valued solution of the equation @xmath401 .
note that since @xmath290 for some constant @xmath291 , we have @xmath402 , _
i.e. _ , @xmath23 and @xmath293 are polynomially related . *
we make @xmath403 copies of each element node and each set node as opposed to @xmath296 copies that we made in the proof of ( a ) .
note that @xmath9 has again precisely @xmath404 nodes .
* let @xmath132 be the constant given by @xmath405 .
our claim @xmath406 in the proof of ( a ) is now modified to + l @xmath406 + [ cols= " > , < " , ] * our proof of the _ completeness _ claim follows the `` proof of @xmath319 '' in the proof of ( a ) with the obvious replacement of @xmath352 by @xmath407 .
* note that our soundness claim is equivalent to its contra - positive @xmath408 and the proof of this contra - positive follows the `` proof of @xmath350 '' in the proof of ( a ) . in the proof , the quantity @xmath296 corresponding to the number of copies for each set and element node needs to be replaced by @xmath403 ; note that @xmath409 . * ( c ) * since @xmath134 for some constant @xmath135 , @xmath410 contains a single equivalence class @xmath35 such that @xmath411 .
thus , we can employ the straightforward exhaustive method of selecting every possible subset @xmath38 of @xmath0 nodes to be in @xmath248 and checking if the chosen subset of nodes provide a valid solution .
there are @xmath412 such possible subsets and therefore the asymptotic running time is @xmath413 which is polynomial in @xmath23 .
note that for this case @xmath136 if a solution exists .
* ( a ) * note that trivially @xmath414 and thus @xmath415 .
our algorithm , shown as algorithm v , uses the greedy logarithmic approximation of johnson @xcite for sc that selects , at each successive step , a set that contains the maximum number of elements that are still not covered .
l l l + * 1 . * & + [ 5pt ] * 2 . * & + [ 5pt ] * 3 . * & + [ 5pt ] & * 3.1 * & create the following instance of sc containing @xmath416 elements and @xmath416 sets : + & & @xmath417 , + & & @xmath418 for @xmath419 + [ 5pt ] & * 3.2 * & * if * @xmath420 * then * + [ 5pt ] & & * 3.2.1 * run the greedy approximation algorithm @xcite for this instance of sc + & & giving a solution @xmath421 + [ 5pt ] & & * 3.2.2 * @xmath422 + [ 5pt ] & & * 3.2.3 * * if * @xmath423 * then * @xmath424 ; @xmath425 + [ 5pt ] * 4 . * & + algorithm v returns a valid solution for adim@xmath142 .
suppose that our algorithm returns an invalid solution in the iteration of the * for * loop in step 3 when @xmath26 is equal to @xmath33 for some @xmath426 .
we claim that this can not be the case since @xmath427 .
indeed , since @xmath111 is a valid solution of the sc instance , for every @xmath428 , the following holds : @xmath429 and thus @xmath33 can not be together with any other node in any equivalence class in @xmath248 . algorithm v solves adim@xmath142 with an approximation ratio of @xmath430 .
fix any optimal solution @xmath431 . since @xmath432 , @xmath433 for some @xmath426 .
consider the iteration of the * for * loop in step 3 when @xmath26 is equal to @xmath33 .
we now analyze the run of _ this particular iteration _ , and claim that the set - cover instance created during this iteration satisfies @xmath434 . to see this , construct the following solution of the set - cover instance from @xmath435 containing exactly @xmath367 sets : @xmath436 to see that this is indeed a valid solution of the set - cover instance , consider any @xmath437 .
then , the following cases apply showing that @xmath438 belongs to some set selected in our solution of sc : * if @xmath439 then @xmath440 and @xmath441 is a selected set in the solution . * if @xmath442 then @xmath443 . using the approximation bound of the algorithm of
@xcite it now follows that the quality of our solution @xmath444 satisfies @xmath445 algorithm v runs in @xmath144 time .
there are a total of @xmath23 instances of set cover that we need to build in step 3.1 and solve by the greedy heuristic in step 3.2.1 .
building the set - cover instance can be trivially done in @xmath446 time by comparing @xmath25 for all appropriate pairs of nodes @xmath26 and @xmath27 .
since the set - cover instance in step 3.1 has @xmath416 sets each having no more than @xmath416 elements , each implementation of the greedy heuristic in step 3.2.1 takes @xmath446 time . *
( b ) * let @xmath26 be the node of degree @xmath74 .
let @xmath33 be the unique node adjacent to @xmath26 ( _ i.e. _ , @xmath447 ) .
consider the following solution of adim@xmath139 : @xmath448 .
we claim that is a valid solution of adim@xmath139 by showing that @xmath449 .
consider any node @xmath450 , then , @xmath451 .
illustration of the proof of theorem [ k - one](c ) .
edges marked by @xmath452 can not exist .
no node in @xmath453 can have an edge to _ both _ @xmath26 and @xmath27 . ] * ( c ) * since @xmath9 does not contain a @xmath146-cycle , @xmath454 .
thus , there exists two nodes @xmath455 such that @xmath456 .
let @xmath457 be a node at a distance of @xmath74 from both @xmath26 and @xmath27 on a shortest path between @xmath26 and @xmath27 ( see fig .
[ fig1 ] ) .
consider the following solution of adim@xmath139 : @xmath458 .
note that @xmath459 .
we claim that this is a valid solution of adim@xmath139 by showing that @xmath449 ( _ i.e. _ , no node @xmath460 can belong together with @xmath33 in the same equivalence class of @xmath248 ) in the following manner : * if @xmath461 then @xmath462 but @xmath463 since @xmath9 has no @xmath146-cycle ( see the edges marked @xmath452 in fig .
[ fig1 ] ) . *
if @xmath464 then @xmath465 but @xmath466 since @xmath9 has no @xmath146-cycle ( see the edges marked @xmath452 in fig .
[ fig1 ] ) . *
if @xmath467 then @xmath166 can not be adjacent to _ both _ @xmath26 and @xmath27 since @xmath9 does not contain a @xmath146-cycle .
this implies that @xmath468 but at least one of @xmath469 and @xmath470 is not equal to @xmath74 . * if @xmath166 is any node not covered by the above cases , then @xmath471 but @xmath465 .
prior to our work , known results for these privacy measures only included some heuristic algorithms with no provable guarantee on performances such as in @xcite , or algorithms for very special cases .
in fact , it was not even known if any version of these related computational problems is @xmath22-hard .
our work provides the first non - trivial computational complexity results for effective computation of these measures .
theorem [ thm1 ] shows that both adim and adim@xmath120 are _ provably _ computationally easier problems than adim@xmath126 .
in contrast , theorem [ k - large](a)(b ) and theorem [ k - one ] show that adim@xmath126 is in general computationally hard but admits approximations or exact solution for specific choices of @xmath0 or graph topology .
we believe that our results will stimulate further research on quantifying and computing privacy measures for networks .
in particular , our results raise the following interesting research questions : 1 .
we have only provided a logarithmic approximation algorithm for adim@xmath142 .
is it possible to design a non - trivial approximation algorithm for adim@xmath126 for @xmath472 ?
we conjecture that a @xmath473-approximation is possible for adim@xmath126for every fixed @xmath0 .
we have provided a logarithmic inapproximability result for adim@xmath126 for every @xmath0 _ roughly _ up to @xmath474 .
can this approximability result be further improved when @xmath0 is not a constant ?
we conjecture that the inapproximability factor can be further improved to @xmath475 for some constant @xmath117 when @xmath0 is around @xmath474 .
99 l. backstrom , c. dwork and j. kleinberg .
_ wherefore art thou r3579x ? :
anonymized social networks , hidden patterns , and structural steganography _ , proc .
@xmath476 international conference on world wide web , 181 - 190 , new york , ny , usa , 2007 .
s. mauw , r. trujillo - rasua and b. xuan .
_ counteracting active attacks in social network graphs _ ,
@xmath479 annual ifip wg 11.3 working conference on data and applications security and privacy , trento , italy , 2016 .
m. netter , s. herbst and g. pernul .
_ analyzing privacy in social networks an interdisciplinary approach _ , ieee @xmath480 international conference on privacy , security , risk and trust and ieee @xmath480 international conference on social computing , 1327 - 1334 , 2011 .
b. viswanath , m. mondal , k. p. gummadi , a. mislove and a. post .
_ canal : scaling social network - based sybil tolerance schemes _ , proc .
@xmath481 acm european conference on computer systems , 309 - 322 , new york , ny , usa , 2012 .
x. wu , x. ying , k. liu and l. chen . _ a survey of privacy - preservation of graphs and social networks _ , in c. c. aggarwal and h. wang ( eds . ) , managing and mining graph data , vol .
40 of advances in database systems , 421 - 453 .
springer , 2010 . | with the arrival of modern internet era , large public networks of various types have come to existence to benefit the society as a whole and several research areas such as sociology , economics and geography in particular . however , the societal and research benefits of these networks have also given rise to potentially significant privacy issues in the sense that malicious entities may violate the privacy of the users of such a network by analyzing the network and deliberately using such privacy violations for deleterious purposes .
such considerations have given rise to a new active research area that deals with the quantification of privacy of users in large networks and the corresponding investigation of computational complexity issues of computing such quantified privacy measures .
in this paper , we formalize three such privacy measures for large networks and provide non - trivial theoretical computational complexity results for computing these measures .
our results show the first two measures can be computed efficiently , whereas the third measure is provably hard to compute within a logarithmic approximation factor .
furthermore , we also provide computational complexity results for the case when the privacy requirement of the network is severely restricted , including an efficient logarithmic approximation . |
in the local universe , galaxies separate into two broad classes @xcite : ( 1 ) large disk - dominated galaxies show ongoing star formation rates ( sfrs ) comparable to their past averages as evidenced by their overall blue colors , and frequent spiral arms , and ( 2 ) spheroidal , or bulge - dominated galaxies with current sfrs much less than their past averages .
however , especially in the more distant universe , there exists a third class of galaxies which exhibits a large amount of stellar population diversity and irregular morphologies ( e.g. , @xcite ) .
the mechanism by which a galaxy transitions from a uniformly blue , star forming galaxy ( with either disk - dominated or irregular morphologies ) into a bulge - dominated , quiescent galaxy remains an open question .
many studies have added to a growing body of evidence that the irregular morphologies are due , at least in part , to heterogeneous ( clumpy ) star formation ( e.g. , @xcite )
. believed to be formed through gravitational instabilities in gas - rich disks ( e.g. , @xcite ) , these clumps typically contribute @xmath6 of the total galaxy light @xcite and most have bluer colors and elevated specific star formation rates ( ssfr ) compared to the surrounding regions @xcite .
this contrast between a young star forming region and the older underlying galaxy population may lead to a large amount of variation in the internal colors of a galaxy .
the ability to investigate the separation ( heterogeneity ) of the different stellar populations ( blue star forming clumps versus the older , redder stellar disk ) becomes increasingly difficult at high redshift , because a galaxy s morphology often depends heavily on the bandpass in which the galaxy is observed . for example , observed optical light ( rest - frame uv ) traces the distribution of unobscured star - forming regions @xcite which are often non - uniformly distributed .
recent observations of clumpy galaxies ( e.g. , @xcite ) find that clumps located toward galaxy centers are older and redder than clumps located toward the outskirts .
this suggests that highly star forming clumps may migrate toward the center where they could condense to form a psuedo - bulge ( e.g. , @xcite ) .
the formation of such a bulge , in turn , could be sufficient to stabilize the stellar disk against new star - formation @xcite , homogenizing the stellar populations and reducing the variations in internal color .
if this scenario is correct , we would expect to observe a relation between the internal color dispersion of a galaxy and the formation of a bulge . to investigate any potential relation , we require a tool to quantify the stellar population diversity of a galaxy .
traditional morphological classification methods : visual inspection , srsic profile fitting , concentration and asymmetry values @xcite , and gini coefficients @xcite are calculated in a single bandpass and , as such , do not provide enough insight into the stellar population variations .
@xcite developed a differential morphological measure , the internal color dispersion ( icd ) , which quantifies the heterogeneity of stellar populations ( such as with clumps ) between bandpasses ; galaxies with non - uniform distributions of young stellar populations give rise to a larger icd than a galaxy with a homogeneous or evenly mixed stellar population .
similarly , galaxies with variable amounts of extinction will also exhibit large icd values , due to the spatial variations in the reddening .
combining the icd with other morphological indicators that quantify the distribution of light in a galaxy ( e.g. srsic , gini-@xmath7 , concentration and asymmetry ) provides insight into galaxies physical properties such as clumpy star - formation and their evolution in both overall color and structure . in this paper
, we build upon previous studies ( e.g. , @xcite ) and we use the icd as the primary metric in a comparative study of the rest - frame uv and optical morphological properties of the largest sample yet ( @xmath8 ) of @xmath0 galaxies to date .
in addition to broader ir coverage , we combine optical and near - ir data from _ hst _ from the southern region of the great observatories origins deep survey and cosmic assembly near - infrared deep extragalactic survey , which also includes the deeper data in the hubble ultra deep field .
we investigate the presence of heterogeneous stellar populations in a sample of @xmath0 galaxies and interpret our results in the context of galaxy evolution where a galaxy transitions from a disk - dominated to a more bulge - dominated morphology while experiencing a suppression of the diversity in its stellar populations . in section [ sec : data ]
we discuss the data and sample selection .
we discuss the icd measuring technique in detail in section [ sec : the icd ] .
we compare icd and other galaxy properties in section [ sec : results ] , and discuss possible causes of high icd in section [ sec : high icd ] .
we discuss the role of star - forming clumps in galaxy evolution in section [ sec : on morphology ] , and finally summarize our results in section [ sec : summary ] . throughout this paper
, we use a concordance cosmological model ( @xmath9 , @xmath10 , and @xmath11 ) , assume a chabrier initial mass function ( imf ; @xcite ) , and use ab magnitudes @xcite .
for this work , we make use of deep imaging data taken as part of the cosmic assembly near - infrared deep extragalactic survey ( candels ; pis faber and ferguson ; @xcite ) of the southern region of the great observatories origins deep survey ( goods , hereafter referred to as goods - s , @xcite ) along with extremely deep observations of the hubble ultra deep field ( hudf ) .
the large variation in depth between the goods - s observations and the hudf ensures that we have access to galaxies with a large range in stellar mass .
we take advantage of the third version of the advanced camera for surveys ( acs ) mosaics in goods - s which includes images taken with f435w , f606w , f775w , and f850lp ( hereafter referred to as , , , and ) . in addition , we use imaging obtained from wide field camera 3 ( wfc3 ) in f098 m , f105w , f125w , and f160w ( hereafter as y@xmath12 , , , and ) .
the wfc3 imaging is split into two parts ; the deep portion covers the central 50% of goods - s with the wide portion covering the southern - most 25% to a depth approximately 1 magnitude shallower than the deep portion . to complete our wfc3 coverage of goods - s ,
we make use of the wfc3 science organizing committee s early release science program ( pi oconnell ; @xcite ) which covers the northern 25% in y@xmath12 , , and . in addition
, we include very deep imaging of the hudf taken with both acs @xcite and wfc3 as part of hudf09 ( pi illingsworth ; e.g. , @xcite ) and udf12 ( pi ellis ; @xcite ) .
both the acs and wfc3 datasets contain imaging using the previously mentioned seven bandpasses .
we use a multi - wavelength photometric catalog of the goods - s region provided by @xcite . in this catalog ,
the sources are detected in a mosaic drizzled to @xmath13pixel .
along with the source detections , photometry covering 0.48 ( vimos u , , , , , y@xmath12 , , , , issac ks , irac 3.6 , 4.5 , 5.8 , and 8) is provided .
the photometric data for the _ hst _ bands are computed by sextractor @xcite on images which have had their point spread function ( psf ) convolved to match the resolution , while the lower resolution ground based and _
spitzer_/irac imaging is processed by tfit @xcite .
see @xcite for full details on the construction of the photometric catalog .
we make use of the photometric redshifts for the galaxies in our sample derived by @xcite .
this photometric redshift catalog is constructed by combining redshifts derived by 11 different investigators using different photometric redshift codes e.g. , eazy @xcite , hyperz @xcite , lephare @xcite , etc .
these photometric redshift codes derive redshifts by fitting different combination of spectral energy distribution templates and priors including available spectroscopic redshifts .
we make use of spectroscopic redshifts where available ( see @xcite for the compilation of spectroscopic datasets ) .
the scatter between the photometric redshifts and corresponding spectroscopic redshifts , which span @xmath14 , is rms@xmath15 = 0.03 $ ] where @xmath16 .
@xcite showed that galaxies with heterogeneous stellar populations have high icd between their rest - frame uv and optical light .
for our galaxy sample with @xmath0 , the candels acs and wfc3 bands probe rest - frame uv and optical wavelengths respectively . using the photometric redshifts described previously ( section [ sec : photoz ] ) we select all galaxies with photometric or spectroscopic redshifts in this range .
we apply a magnitude limit of 25 mag in , which allows for a robust determination of morphological parameters and the icd values ( see section [ sec : icd vs ston ] for reasoning ) . to avoid stars ,
we remove objects with stellarity index @xmath17 ; stellarity derived using sextractor gives a likelihood of an object being a star . applying this set of criteria
yields a parent sample 3369 galaxies .
because the photometric redshifts are the result of the combination of of the probability density functions from many routines , @xcite show they provide more accurate estimates of the redshift .
the method employed by @xcite does not provide additional physical properties for the candels galaxies .
we derive stellar population parameters for the galaxies in our sample using fast @xcite , a stellar population synthesis modeling and fitting code .
we fix the galaxy s redshift at the photometric redshift from the @xcite catalog and fit 14-band galaxy photometry covering @xmath18 with model spectral energy distributions to estimate the stellar masses ( along with dust attenuation ) for the galaxies in the sample .
we used models for a range of stellar population properties from the 2003 version of the bruzual & charlot models ( @xcite ) stellar population synthesis models allowing for a range of attenuation ( @xmath19 mag ) using the @xcite extinction law .
we opt to use @xcite to facilitate the comparison to other studies and because recent work shows that the bruzual & charlot 2007 models may overestimate the contributions of late type giants to the near - ir ( e.g. , @xcite ) .
we assumed models with a delayed , exponentially declining ( log @xmath20 ) star formation history , solar metallicity and a @xcite imf . using a salpeter imf @xcite would to first order increase systematically the stellar masses by @xmath21 dex .
using different assumptions for the stellar population metallicities will affect the derived stellar masses by @xmath22 dex @xcite .
the average uncertainty in stellar mass for our sample is 0.11 dex , although errors from systematics in the metallicities , star - formation histories and other aspects of the spectral energy distribution fitting likely dominate ( e.g. , @xcite ) .
the errors on the photometric redshifts do not influence greatly our conclusions . while , photometric error uncertainties can affect the derivation of stellar population parameters , for small photometric redshift errors , like those for our sample , the contribution to the stellar mass uncertainty is minimal because small changes in redshift translate to small changes in distance ( see discussion in @xcite ) .
for example , our uncertainty of rms@xmath23 = 0.03 $ ] corresponds to a change in luminosity distance of @xmath245% for our sample . because the stellar mass scales quadratically with luminosity distance , the mass uncertainty from photometric redshift errors is similarly small , @xmath2410%
this error is much smaller than systematic uncertainties in stellar mass arising from model fitting ( see discussion in @xcite ) , and so we neglect errors from the photometric redshifts on the stellar masses . along with stellar masses
fast provides an estimate of the total attenuation ( @xmath25 ) due to dust in a galaxy .
we use this value in section [ sec : attenuation ] to understand better how the icd might be effected by reddening due to dust .
we use the structural catalog provided by @xcite which provides srsic index , @xmath26 , @xcite and the effective radius , @xmath27 ( measured along the major axis of the galaxy ) , for all sources in the goods - s catalog from @xcite .
these parameters are produced by galfit @xcite using the best - fitting srsic models for the objects in the imaging .
we also use the non - parametric indicators gini parameter ( @xmath28 ) and the second - order moment of the brightest 20% of the galaxy light , @xmath7 .
these indicators quantify the spatial distribution of light without assuming a specific functional form when classifying the morphologies of disturbed or irregular galaxies . for a complete description of this method see @xcite .
we use color gradient measurements from hilton et al .
( in prep ) , where hilton et al .
compute the color gradients between the core and outskirts by centering annular apertures on each galaxy .
they define concentric , elliptical , annular apertures such that the total light of each galaxy is divided equally between three annuli .
the galaxy core corresponds to the `` inner '' aperture and the galaxy outskirts correspond to the `` outer '' aperture .
the color gradient of the galaxy is denoted by @xmath29 where , @xmath30 and @xmath31 are the respective apparent magnitudes
. a galaxy where the core has a redder color than the outskirts will have a negative color gradient and a galaxy with a bluer core than outskirts will have a positive gradient . in this work
, we measure the total star - formation rate ( sfr ) as the sum of the sfr implied by the observed uv luminosity ( uncorrected for dust extinction , @xmath32 ) and by the ir ( as estimated from the 24 luminosity , @xmath33 ) .
we convert the observed 24 luminosity into a sfr using the empirical relationship from @xcite . for galaxies which are not detected in 24 ,
the sfrs are based on an attenuation - corrected uv flux using attenuation value estimated by fast , so @xmath34 . in both instances , we follow @xcite to convert the luminosity to sfr .
the @xcite conversion is salpeter - based , whereas the masses and other derived quantities are chabrier - based . to account for the difference in imf
, we divide the total sfr by 1.66 , @xmath35 in section [ sec : clumps ] we compare the internal color dispersion measurements to the frequency of discrete star forming regions ( colloquially `` clumps '' ) in galaxies using the clump catalog of @xcite . in order to facilitate clump finding ,
@xcite require galaxies to be approximately face - on ( b / a @xmath36 ) , reasonably extended ( @xmath37 ) , and with a sufficient brightness ( @xmath38 mag ) .
individual clumps are selected from the uv luminosity contrast between the individual clump and the total galaxy . from the catalog , we select clumps with uv luminosity @xmath39 of the total galaxy uv light ( @xmath40 ) contained within galaxies in our sample . while @xcite selects clumps with uv luminosity @xmath41 ; we lower this threshold to include more uv - faint clumps as they are redder and provide a more comprehensive view of the icd .
this provides clump measurements for 180 objects , containing between zero and eight clumps per galaxy .
first developed by @xcite , the internal color dispersion ( icd ) is a flux - independent statistic , defined as the ratio of the squared difference of the image flux - intensity values about the mean galaxy color to the square of the total image flux , or , @xmath42 the icd assesses the morphological difference of a galaxy between two passbands .
the rest - frame uv - optical icd quantifies the spatial homogeneity ( or lack thereof ) between young stellar populations ( dominating the rest - frame uv light ) and older stellar populations , contributing most of the rest - frame optical light . galaxies with
spatially segregated stellar populations with different colors would have large icd ( see @xcite ) .
whereas , a galaxy with a homogeneous stellar population throughout would have a low icd .
the terms in equation [ eqn : xi ] sum over @xmath43 with @xmath44 being the number of pixels associated with each galaxy , @xmath45 and @xmath46 are the image flux intensity values for each object obtained in each of two passbands and @xmath47 and @xmath48 are the pixel intensity values from a contiguous , similarly sized , region of the two passbands not associated with any object . to determine which pixels belong to a galaxy we define the aperture size as @xmath49 , where @xmath50 is the petrosian radius of @xmath46 as defined for the _ sloan digital sky survey _ @xcite , and @xmath51 is defined as the ratio of the galaxy surface brightness , @xmath52 , averaged over an annulus of radius @xmath53 , to the mean surface brightness within this radius , @xmath54 .
the petrosian radius depends on the surface
brightness distribution of the galaxy and is thus independent of redshift or systematics in the image calibration .
the chosen aperture encompasses most of the light from the galaxy without adding excessive background light .
therefore , the summations are over the individual pixels associated with the object or blank region of a similar size .
the scaling factor @xmath55 is the ratio of total fluxes between images @xmath46 and @xmath45 , while the linear offset @xmath56 adjusts ( if necessary ) for differences in the relative background levels of the two images . in practice , we calculate these terms by minimizing the statistic , @xmath57 ^ 2 $ ] , adjusting @xmath55 and @xmath56 as free parameters ( where @xmath58 represents the uncertainties on the @xmath46 flux intensity values ) .
the treatment of the background terms ( @xmath47 and @xmath48 ) are important to the icd , as the background itself can contain an intrinsic amount of internal color dispersion based on our definition .
therefore , to eliminate icd contribution from the background , @xmath47 and @xmath48 are selected for each individual galaxies to contain the same number of pixels ( @xmath44 ) at random locations excluding any preexisting objects , and include the same pixels in both image 1 and 2 . in order to minimize the icd variations due to background , we select the median icd value computed from using nine different background regions . however , simple fluctuations may not be the only source of icd . because , the wfc3 imaging psf has been convolved ( see section [ sec : data ] ) to match that of the acs , there exists the possibility of correlated noise in the background measurements .
this could have an effect if the background region is not selected in the same way as the galaxy ( e.g. , choosing individual pixels at random versus a single contiguous region ) .
we test for possible effects by choosing individual pixels not associated with an object at random from the imaging , choosing a single contiguous _
square _ region and contiguous _ circular _ region .
we find no effect based on the morphology of chosen background pixels .
previous studies @xcite noted that the icd produces unreliable results for galaxies with low signal - to - noise ( s / n ) .
we test the behavior of the icd using simulated galaxies with _ a priori _ known ( fixed ) total flux and icd .
each simulated galaxy is placed throughout the _ hst _ images at a random location and is then checked to insure no overlap with an actual object by comparing the location with the sextractor segmentation image .
the flux of the galaxy is slowly decreased to a level barely above the background .
we then derive an icd value for each simulated object using the definition in equation [ eqn : xi ] . in the _ bluer band
_ the median icd deviates more than 34% away from the true value .
therefore , we remove any object from our initial sample that have @xmath59 in the bluer band .
the s / n requirement for galaxies with intrinsically high icd could be set to a lower threshold , but we choose to constant criterion to keep the selection uniform.,scaledwidth=49.0% ] figure [ fig : ston_i ] shows the fractional change ( true minus observed divided by true ) in the icd as a function of @xmath60 , where @xmath61 represents the flux
intensity values , and @xmath62 is the image variance due to the noise of the background . the sum is taken over all pixels identified as belonging to the object , by computing the petrosian radius and comparing the enclosed pixels to the sextractor segmentation image .
the four simulated objects illustrate how different levels of icd ( 2% , 6% , 13% , and 26% ) are effected by the galaxy s s / n . the behavior of the fractional change is dramatically different depending on whether a galaxy has intrinsically `` low '' or `` high '' icd . for all galaxies with very high s
/ n we recover the expected icd to within 0.1%
. however , as the s / n decreases equation [ eqn : xi ] systematically _ underestimates _ the icd .
this bias is most apparent for galaxies with `` low '' intrinsic icd . because of the systematic underestimation of the icd it is important to clip the sample at a s / n which still provides accurate results .
we choose s / n = 10 such that the bias is does not impact our results
. for example , the bias is the fractional change ( @xmath63 ) . for objects with s / n = 10 and high intrinsic icd ( @xmath64 ) , the bias is small , with relative changes of @xmath65 , and the bias decreases for objects with increasing icd .
the bias increases for objects with lower intrinsic icd , such that the relative change is 100% for objects with s / n = 10 and @xmath66 .
this is acceptable because the icd increases by only a factor of 2 .
therefore , to our s /
n = 10 limit , we can be confident that objects with observed high icd have high intrinsic icd , and objects with low observed icd have low intrinsic icd . in principle
this s / n requirement applies to both bands used in equation [ eqn : xi ] .
however , for this particular study , the redder band ( namely ) s / n is much higher in the majority of galaxies analyzed , such that this requirement , in practice , only applies to the bluer imaging band .
it is also important to note that this s / n does not correspond to a single bluer band magnitude limit , because at fixed magnitude a larger object ( with lower surface brightness ) will have a lower s / n . for galaxies with an intrinsic icd very near zero , and low surface brightness in the bluer band , it is possible for the icd calculation to produce negative results when the background contributes significantly to the fluctuations .
galaxies with an intrinsically larger icd are more resistant to surface brightness dimming as the fluctuations of the galaxy dominate those of the background . after applying the s / n = 10 restriction in
, we have a final , primary sample of 2601 galaxies .
unfortunately , this s / n requirement leads to a selection effect where we prefer galaxies with some level of active star - formation , as passively evolving galaxies often do not have enough blue light to be included in this analysis .
the effect of this selection bias is discussed further in section [ sec : icdvmass ] .
@xcite defines the statistical error associated with the icd , assuming that the background pixels ( those not associated with any galaxy ) are normally distributed , as @xmath67 @xcite note that equation [ eq : icd error ] underestimates the uncertainties associated with the icd .
they estimated that the uncertainties for equation [ eqn : xi ] in their data required corrections of factors of 6 - 7 to account for correlated noise in the images .
@xcite find that even after the correction suggested by @xcite , the uncertainty on the icd obtained for bright objects ( those with low errors ) are underestimated by as much as a factor of three .
the dominant source of error being contributions from non - uniform sky backgrounds and psf mismatch between the two passbands . for this work ,
we use the observed icd values for different levels of intrinsic icd and galaxy s / n to estimate the error associated with each measurement . to estimate the systematic uncertainty we compare the median icd fractional change ( the dashed line in figure [ fig : ston_i ] ) to the true value , and to estimate the statistical uncertainty we compute the median absolute deviation . in all cases , galaxies with very high ( s / n @xmath68 )
s / n have low ( @xmath69 ) systematic uncertainties and very low ( @xmath70 ) statistical uncertainties .
the closer to our s / n cutoff ( s / n = 10 ) the systematic error , described in the previous section , increasingly dominates .
the statistical uncertainty remains very low @xmath71 and so we neglect any effect due to the statistical variation in the icd .
for the majority of our results we choose to use @xmath72 as our primary diagnostic . for the galaxies in our sample ,
these two filters span the balmer/4000 break over @xmath73 , and allow us to compare the rest - frame uv ( traced by the band ) to the rest - frame optical ( traced by the band ) .
this provides us the ability to investigate recent star formation ( with ) and past - averaged star formation ( with ) . in several sections (
section [ sec : icdvmass ] and section [ sec : redshift ] ) we make use of different versions of the icd , namely @xmath74 which allows us to compare slightly different parts of each galaxy s spectral energy distribution to better understand the results given by @xmath75 .
the median icd for the full sample is @xmath76 . in the sections that follow we refer to galaxies with icd value above and below this median as `` high '' and `` low '' icd , respectively stellar mass is one of the most fundamental properties of galaxies .
the stellar mass of a galaxy correlates with many of its physical properties including overall color @xcite and sfr ( e.g. , @xcite ) to @xmath77 ( and perhaps beyond e.g. , @xcite ) . as such , it is possible to gain much insight into the formation and evolution of galaxies through the rates of change of the icd as functions of stellar mass .
figure [ fig : icd_vs_mass ] shows the @xmath75 as a function of stellar mass for the galaxies in our sample .
we find the median @xmath75 values in the three low stellar mass bins are between @xmath78 , the median icd value in the moderate stellar mass bin is above 8% , and at stellar masses @xmath79 , we find that galaxies have a median icd value about 9% . if we take into account the systematic underestimations of low icd values , discussed in section [ sec : icd error ] , we find little relative change in the median @xmath75 between bins , in that the moderate stellar mass bins still show higher @xmath75 medians than the bins of lower stellar masses . to illustrate the icd , figure [ fig : montage ] shows galaxy images in the @xmath72-mass plane .
the galaxies are selected at random from all of the galaxies contained in each bin of mass and @xmath75 .
the rgb images are combinations of , and .
galaxies with elevated icd values ( top rows ) and in the moderate stellar mass range appear significantly different than the galaxies in other stellar mass bin or galaxies with lower icd values .
visually , galaxies with high icd have morphologies with well - separated regions of rest - frame uv ( blue ) and rest - frame optical ( red ) light .
galaxies with low icd values tend to have smooth morphologies and uniform colors , suggesting fewer regions of separated red and blue emission .
the highest stellar mass galaxies ( @xmath80 ) deserve special attention .
because the s / n threshold applies only to the bluer band used in the icd calculation , galaxies which are well detected in may not be well detected in .
specifically , massive galaxies often have the reddest @xmath81 colors , so we expect that many of them have high s / n and low s / n .
the concern is that these objects could have higher intrinsic icd , but are excluded because of our signal to noise requirement ( see section [ sec : icd vs ston ] ) . to investigate this possibility
, we compare the mean @xmath81 color of the high stellar mass galaxies included and excluded from our sample .
galaxies which have s / n@xmath82 have @xmath83 mag while galaxies with s / n@xmath84 have @xmath85 mag .
this suggests that the excluded high stellar mass galaxies simply do not have enough blue light to produce high icd values .
@xcite simulate just such a situation where the authors find that extremely red galaxies , even with the addition of highly star - forming regions produce nearly no icd as the older stellar populations simply overwhelm any diversity .
for the seven high stellar mass galaxies with @xmath86 mag in our sample , four have high @xmath72 ; however , a visual inspection provides additional insight into these objects ( discussed below ) .
a visual inspection of the 26 most massive galaxies , shown in figure [ fig : highmass_stamps ] shows that the majority of these objects have smooth light profiles in both , and .
there are several exceptions which fall into two broad categories : galaxies which appear to have faint , nearby objects detected in but not ( e.g. , 16375 , 17205 and 16934 ) . because serves as the object detection image , it is possible that these faint objects lie within the pertrosian radius used for the icd calculation , contributing significantly to the measured icd . while the origin of these nearby objects is unclear ,
possible scenarios include a chance alignment or a disrupted satellite . and , galaxies in the highest - mass subsample with extended structures , showing distinct red / blue regions and likely individual `` clumps '' of star - formation embedded within the galactic structures ( e.g. , 282 , 15639 and 15261 ) .
in both cases , these features contrast with the dominated central region producing the large @xmath75 value . the blue central regions in the @xmath81 color maps reflects small mismatches in the psf matching ( see @xcite for a description ) and do not contribute significantly to the icd as they are small compared to the total flux of the galaxy .
after a visual inspection , if we remove the galaxies with neighboring objects from the high stellar mass bin the median @xmath75 becomes @xmath87 .
however , if we instead place a more stringent s / n requirement ( s / n @xmath88 ) on the high stellar mass galaxies , we find the median @xmath75 further drops to 3.3% . because it is unclear whether or not the faint objects neighboring the primary galaxy are physically associated with the primary , or are chance alignments along the line of sight , we do not exclude these objects and do not adjust the s / n requirement .
instead , we note that these are objects are infrequent in the sample , and so do not affect our conclusions .
so while the median @xmath75 in the high stellar mass bin may appear to be very similar to the moderate stellar mass bins , it is likely that the median is being inflated due to inclusion of galaxies with neighboring objects or galaxies with low s / n and high s / n . many previous works ( e.g. , @xcite ) use the srsic index to classify a galaxy s morphology
. the srsic index parameterizes a galaxy s surface brightness profile , where @xmath89 corresponds to a classical bulge , @xmath90-law profile @xcite and @xmath91 corresponds to the exponential profile of a galactic disk .
we refer to galaxies with @xmath5 as disk - dominated , and those with @xmath92 as bulge - dominated .
figure [ fig : icd_vs_mass_vs_sersic ] shows the @xmath75 as a function of stellar mass , separated based on their srsic indices .
again , we divide the sample into mass bins , where each bin is 0.5 dex wide or larger to ensure there are at least 10 objects in each bin .
galaxies with high icd prefer a lower srsic index ( disk dominated ) . inside of the individual srsic bins , galaxies with intermediate stellar mass ( @xmath2 )
display the largest icd , but as the srsic index increases , the upper envelope of icd values decreases : bulge - dominated ( @xmath93 ) galaxies have low icd . for disk - dominated galaxies ( @xmath5 ) the median galaxy across all masses has a high icd value .
the systematic underestimation described in section [ sec : icd error ] , has very little effect on the preference of galaxies with low srsic index and intermediate stellar mass to have high @xmath75 .
this is because there is very little relative change in the medians of the bins due to the underestimation of the icd .
there is a well known trend between the srsic index and stellar mass , with the srsic index increasing with increasing stellar mass , this relationship is apparent both at low ( e.g. , @xcite ) and high ( e.g. , @xcite ) redshift .
we find a similar trend .
the three low srsic bins lack ( or have very few ) galaxies with masses greater than @xmath94 .
in fact , it is only the highest srsic bin which shows any significant number high stellar mass galaxies .
the icd is sensitive to heterogeneities in the stellar populations of galaxies .
one possible cause of these heterogeneities is unobscured , young stellar populations separated from older , passively evolving populations . because young populations are needed to produce measurable icd , the star formation rate ( sfr ) of a galaxy could be closely related to the icd .
figure [ fig : icd_sfr ] shows the specific star formation rate ( ssfr ) , the star formation rate normalized by the galaxy s stellar mass , as a function of @xmath75 . .
the gray points are the actual data values and the red line is the running median value .
galaxies which have been clipped for clarity are shown as black arrows .
the shaded region shows the width of the star - forming `` main sequence '' for a @xmath95 stellar mass galaxy at @xmath96 taken from @xcite .
we find that galaxies with the highest @xmath75 have very close to the median ssfr while galaxies with the highest and lowest ssfrs lack many galaxies with very high @xmath75 values.,scaledwidth=49.0% ] we find that the @xmath75 of a galaxy in not directly tied to its ssfr .
galaxies with low @xmath75 values show no correlation with ssfr , however , galaxies with high @xmath75 values show predominately `` normal '' ssfrs . for galaxies with low ssfrs
we expect low @xmath75 values as such galaxies have few young stellar populations producing the blue light needed to contrast the redder light from more evolved stellar populations .
galaxies showing very high ssfrs , often due to re - radiation by dust in the ir , also have low @xmath75 suggesting that dust variations play a small role in the production of icd .
again , this result is robust even after accounting for the systematic underestimate of the low @xmath75 galaxies .
shifting the lowest @xmath75 galaxies upward to higher @xmath75 values does not produce a trend with ssfr .
as discussed previously , the icd is sensitive to heterogeneous stellar population variations that give rise to differences in the uv - optical flux ratio within a galaxy . in this section
we discuss possible scenarios which could lead to high icd .
specifically , we investigate whether merger activity , the galaxy s physical size , intrinsic color gradients , effects due to redshift , or heterogeneous extinction drive increased icd . based on morphological asymmetries
, @xcite suggested that some of the galaxies with high icd are likely the result of merger activity . to investigate this
, we use non - parametric morphological indicators , specifically the gini coefficient , @xmath28 , and the second order moment of the brightest 20% of a galaxy s flux , @xmath7 , to better understand the role of galaxy mergers in producing high icd .
many studies demonstrate these parameter s usefulness in quantifying galaxy morphology in large samples at both low and high redshifts ( e.g. , @xcite ) . ) of objects in our sample . based on this classification scheme , objects above the dashed ,
green line are mergers , below the dashed , green line but above the blue , dotted line are elliptical galaxies and below both lines are spiral galaxies .
the majority of galaxies with @xmath97 are not classified as mergers.,scaledwidth=49.0% ] @xcite adopt the following classification scheme based on g-@xmath7 : @xmath98 figure [ fig : gini_m20 ] shows the @xmath28 and @xmath7 values for the galaxies in our sample broken into large bins of icd . as can be seen ,
mergers are not a dominant driver of the icd .
the majority of galaxies with high icd values ( @xmath99 ) prefer disk - dominated morphologies , consistent with our findings between the icd and sric index above .
this suggests that mergers are not a significant cause of elevated icd , except , possibly , at the highest levels ( @xmath100 ) where 21% of galaxies are classified as mergers .
given the small number of galaxies in the highest @xmath75 bin , we use a series of monte carlo simulations to estimate how often we would find a similar distribution simply by chance .
we assign each galaxy a morphology based on their location in the @xmath101 space while allowing their @xmath75 to be changed .
we then select galaxies with @xmath100 and record the fraction classified as mergers .
we find that galaxies in the highest icd bin are at least 21% mergers only 1.32% of the time . or , given just random chance , 98.7% of the time we find fewer mergers in the highest icd bin than what is actually measured .
the icd requires that we are able to resolve scales to a minimum of 0.5 kpc @xcite . below this
, the resolution of the image suppresses any icd that may be present .
here we also considered how the icd values change with increasing physical size .
as a function of the galaxy s effective radii ( semi - major axis ) .
gray points are the individual galaxies .
the red line is the median @xmath75 value and the blue dashed represent the interquartile range , 75% and 25% respectively .
the dashed vertical line shows the effective radius corresponding to the fwhm of the wfc3 psf ( fwhm = @xmath102 ) at @xmath103 .
we find no trend suggesting a preference of physical size for a galaxy with elevated ( @xmath104 ) @xmath75.,scaledwidth=49.0% ] figure [ fig : halflight ] shows @xmath75 as a function of semi - major axis , @xmath105 . for galaxies with @xmath105 below 3 kpc
the median icd remains low ( @xmath106 ) .
considering only the intermediate mass ( @xmath107 ) we find icd values remain low until @xmath108 kpc after which it increases . in either case , we find more galaxies with large icd at large @xmath105 . however , this is most likely a selection effect where we are more able to detect icd in extended objects .
furthermore , placing the galaxies on a size mass relation provides a similar result . at a fixed stellar mass
, we find more elevated @xmath75 galaxies with larger radii than at small radii .
one of our findings is that the icd values are low for lower stellar mass galaxies , @xmath109 . because these galaxies have lower @xmath105 compared to higher mass galaxies ( e.g. , @xcite ) we test if the trend between icd and @xmath105 affects this finding .
we remove all objects with @xmath110 kpc ( at all stellar masses ) and recompute how the median icd changes as a function of mass .
the median @xmath72 for @xmath111 remains low , with @xmath112 .
the median @xmath75 for moderate mass galaxies remains elevated , @xmath113 , for @xmath107 , and the @xmath75 declines at the highest masses .
therefore , the relation between icd and @xmath105 does not dominate our conclusions between icd and stellar mass .
galaxies show a radial color gradient when the color of the light in an annulus on the outskirts differs from the color inside a central region ( see section [ sec : data_colorgraident ] ) . for galaxies with large color gradients , this difference could potentially lead to large icd values .
@xcite found that the majority of a mass - selected sample from the hudf show negative color gradients ( redder cores and bluer outer regions ) .
we find a similar result for the galaxies in our sample , with an average color gradient @xmath114 and 66% of the sample showing a negative color gradient .
, as a function of @xmath75 .
gray points are the data and the red line is the median color gradient as a function of icd .
we find that galaxies with high icd have systematically lower color gradients , suggesting they have redder cores than outskirts.,scaledwidth=49.0% ] the relationship between color gradient and the icd is complex .
figure [ fig : icd_vs_color_grad ] shows @xmath115 as a function of @xmath75 .
we find that while galaxies with high icd typically have negative color gradients , there is no correlation between the strength of the color gradient and the icd .
therefore , while strong color gradients do not necessarily imply high icd values , galaxies with high icd appear to require larger color gradients , either positive or negative .
calculated with the center @xmath116 removed and the @xmath75 calculated for the entire galaxy as a function of @xmath75 for the entire galaxy . we find that as galaxies with their centers removed show an elevated amount of @xmath75 .
this is expected as the variations in the outskirts play a more significant role with the bright central region removed , and indicates the observed icd is not due to an underlying color gradient.,scaledwidth=49.0% ] to understand how much a color gradient contributes to the measured icd , we perform the following test .
we mask the central @xmath116 ( @xmath117 kpc ) of all galaxies and recompute the icd value .
figure [ fig : icd_icdcored ] shows the relative change in the icd due to this masking . in the majority of galaxies with small @xmath75 values , removing the nuclear region causes the icd to decrease , while in galaxies with larger icd values there is little change .
this is as expected as galaxies with small icd values , masking out the central regions also removes some of the variation .
therefore , while galaxies with strong color gradients may have high icd , this is not a requirement .
the observed color gradients alone are insufficient to produce the icd in the galaxies in our sample . .
the upper and lower quartiles are indicated by the blue boxes .
the medians of each bin is indicated by the red line . for clarity ,
only a portion of the possible icd range is shown .
top and bottom panel show @xmath75 and @xmath118 as a function of redshift , respectively . in both panels
we do not see a significant increase in the median values of icd .
the inclusion of the bluer bands ( and ) acts as a probe for bandpass shifting effects .
because neither shows a significant trend , this suggests that the lack of increased values is not simply due to bandpass shifting effects.,scaledwidth=49.0% ] the galaxies in our sample span a range of redshift from @xmath0 . as a result
, the rest - frame light probed by both and varies between galaxies .
for example , in galaxies at @xmath119 , probes rest - frame @xmath120 , while at higher redshifts , @xmath121 , probes light shortward of 4000 .
therefore , we have considered how the variation in rest - frame light could affect our icd measurement . for galaxies at @xmath122
, we measure the icd between and , as these bandpasses observe similar rest - frame wavelengths as and for galaxies at @xmath123 .
figure [ fig : icdvz ] shows icd as a function of redshift where we have binned the galaxies by redshift .
the top row of figure [ fig : icdvz ] shows @xmath75 and the bottom shows @xmath118 . for the comparable bins ( @xmath124 for @xmath118 and @xmath125 for @xmath75 )
we find very little ( @xmath126 ) difference between the median icd values .
this shows that as long as the imaging bands chosen to calculate the icd span the rest frame balmer/4000 break we calculate a similar icd .
for galaxies above @xmath127 both and lie blueward of the rest frame balmer/4000 break , and @xcite suggests that a sharp downturn would be observed as the icd probes similar populations .
we find no such downturn , suggesting that while the balmer/4000 break is moving through , the observed trends in @xmath75 as a function of stellar mass ( figure [ fig : icd_vs_mass ] ) is not simply due to the effect of redshift , but due to the individual characteristics of the galaxies .
as mentioned earlier , high icd could result from spatial variations of stellar ages or dust attenuation . in this section
, we investigate whether the icd is related to the total dust content of our galaxies .
the colored boxes show the average attenuation value ( @xmath128 ) of the points within the region . for the galaxies in our sample
, we find no evidence of a correlation between the measured stellar population diversity and the rest frame uv - optical color .
we do find that galaxies with large icd values also show large attenuation values.,scaledwidth=49.0% ] if inhomogeneous attenuation ( the reddening of selective parts of a uniformly blue population ) is the primary driver of the icd then we would expect to find galaxies with both very red colors and high icd values
. a uniformly blue galaxy would begin with a blue @xmath129 color and if we were to add a level of inhomogeneous dust then we would expect the icd to rise , as there is now red light with which to contrast the blue and the overall color to shift toward the red .
similarly if a galaxy with a heterogeneous stellar populations has the young star forming regions attenuated then the icd would go down as there is less blue light contrasting the red and the @xmath129 color again shifts toward the red .
finally , if a galaxy dominated by older stellar populations is dust attenuated then the red stellar populations become more red , shifting the @xmath129 color , but the icd would increase as there is now some contrast between stellar populations . in figure
[ fig : icd_vs_color ] we show the icd as a function of @xmath129 color and attenuation value ( @xmath128 ) as reported by fast ( see section [ sec : stellar masses ] ) .
there is little evidence for a correlations between either the icd or @xmath128 as we find high @xmath75 galaxies for a wide range of colors and @xmath128 .
the icd is low for galaxies with both high and low @xmath81 colors , but galaxies with moderate @xmath81 color ( 1 mag @xmath130 2 mag ) show the largest range in the icd .
therefore , it seems that galaxies with moderate @xmath81 colors are required for high icd , but color alone is not enough to predict high icd . @xcite
argue this lack of direct correlation supports the idea that the icd depends on spatial variations in the ages of the stellar populations and not dust attenuation .
, as a function of icd .
the colored boxes show the average attenuation value ( @xmath128 ) of the points within the region .
galaxies with ir dominated sfrs have smaller icds than those with a lower @xmath131 . at fixed @xmath131 ,
there is a large spread in derived attenuation .
this suggests that a mix of stellar population ages drives the icd and attenuation plays a minor role.,scaledwidth=49.0% ] as a second investigation , we compare the ratio of the total galaxy sfr ( @xmath132 ) , derived from the ir and uv emission ( see section [ sec : data_sfr ] ) , against the sfr inferred from the uv measurement ( uncorrected for dust attenuation ) . in this way , the ratio of @xmath133 is a measure of the amount of light from star formation that is attenuated by dust and reradiated in the ir .
figure [ fig : sfrratio_icd ] shows there is no trend in @xmath133 as a function of @xmath75 .
galaxies with high @xmath75 span the entire range of @xmath133 . taken together ,
this is evidence that spatial variations of dust is not the main driver of the icd , and as such , the icd requires spatial variations in stellar population ages in galaxies ( perhaps with contributions from spatial variations in dust attenuation ) .
this agrees with the conclusions from @xcite and @xcite .
both authors acknowledge that extinction variations could play a role in providing variations in the blue starlight , but conclude that the effect is smaller than those produced by variations in segregated stellar populations of different mass - weighted ages .
the icd appears to driven mostly by variations in the ages of stellar populations in galaxies .
@xcite suggest that in order to have icd @xmath134 , young star - forming disk in galaxies are required to have a surrounding older spheroid ; for higher icds ( @xmath135 ) , they suggest compact star - forming hii regions ( discrete star - forming regions or clumps ) are also required .
this picture is corroborated in a recent study by @xcite who found that star - forming galaxies at @xmath136 , with stellar masses @xmath137 have redder centers and bluer outer regions .
figure [ fig : clump_icd ] shows @xmath75 as a function of the number of clumps . using the catalog of @xcite , see section [ sec : data_clumps ] .
the average @xmath75 value steadily increases for increasing number of clumps but turns over after a galaxy contains at least four clumps .
the average number of clumps for a given @xmath75 value shows a similar increase with increasing @xmath75 but does not show a turnover . for galaxies with low ( @xmath138 ) icd values ,
the mean number of clumps per galaxy is 1.2 ; for galaxies with higher icds ( @xmath139 ) , the mean number of clumps is 2.1 therefore , we find that if a galaxy has at least two clumps , on the average , it will have a high @xmath75 value .
this trend continues , where galaxies with four or more clumps show median icd of greater than 10% .
therefore , the icd is sensitive to the number of star - forming clumps .
the clump catalog provides several clump detection levels ( see section [ sec : data_clumps ] ) .
when we increase in the detection threshold ( uv luminosity @xmath139 instead of @xmath140 ) we find the mean number of clumps remains roughly constant ( @xmath141 ) regardless of the @xmath75 value . also ,
while the total number of clumps per galaxy is reduced , we find the @xmath75 of galaxies with fewer than two clumps remains generally unchanged between the two uv luminosity thresholds . because the star - forming clumps are expected to form in gas - rich disks ( e.g.@xcite ) , the icd must trace galaxies which have a redder but still gaseous stellar disk .
these individual clumps are thought to be relatively short lived ( only @xmath142 myr ; @xcite ) , which would correspond well with the lifetimes of hot , blue stars contributing to the uv light associated with high icd .
for the galaxies in this study , we find that as both the stellar mass and srsic index increase , the icd decreases ( figure [ fig : icd_vs_mass_vs_sersic ] ) .
this is unsurprising as previous studies suggest galaxies with high srsic indices have smooth light profiles with little color variation across them ( e.g. @xcite ) and these stellar populations produce little color dispersion as any variation is small relative to the total galaxy flux @xcite .
however , the galaxies with high icd values have intermediate stellar mass and are disk - dominated ( @xmath5 ) .
as mentioned previously , @xcite found that for a galaxy to have high icd a red bulge surrounded by a blue disk with hii regions ( i.e. clumps ) is required . because high icd is most prominent in disk - dominated morphologies and associated with star - forming clumps , the icd is less sensitive to the red bulge component and more driven by the color variation between the underlying stellar disk and highly star - forming clump .
we see this in the previous comparison of the @xmath75 computed with and without the central region ( see section [ sec : color gradients ] ) , where we found that galaxies with their nuclear region removed show an increased @xmath75 ( figure [ fig : icd_icdcored ] ) .
interpreted another way , the decrease of @xmath75 with increasing srsic index suggests that there is a continual suppression of the mechanism which causes high icd ( perhaps clumps ) with spheroid formation .
this shows that _ if _ galaxies are to form quiescent bulges or spheroids along with a star - forming disk or clumps ( as is typical of classical hubble sequence galaxies ) it is only happening at this moderate ( @xmath143 ) stellar mass range and with relatively small bulge contributions .
once a bulge contributes significantly to the galaxy s morphology , the amount of icd decreases dramatically , suggesting that the bulge plays a role in the homogenization of the galaxy s stellar populations .
two different effects could be associated with the formation of a bulge and subsequent reduction in the observed icd .
one theory is the `` morphological quenching '' @xcite concept , by which the formation of the a bulge stabilizes the gaseous disk against further fracturing and star formation .
the second could be some feedback process associated with the bulge which removes any remaining gas from the disk , preventing new clump formation .
the first step in the quenching process is to form a strong bulge component .
one scenario for this process is for clumps to migrate through dynamical friction to coalesce in the nuclei , and merge to form a bulge ( e.g. , @xcite ) .
the galaxies with high icd values also show negative color gradients ( figure [ fig : icd_vs_color_grad ] ) suggesting the central stars are older ( or perhaps more attenuated by dust ) than the outer star - forming regions .
other studies ( e.g. , @xcite ) find that central clumps are both older ( by @xmath144 myrs ) and more attenuated than the clumps in the outskirts suggesting that migration plays a significant role .
of course , not all galaxies in the moderate stellar mass range have high icd values .
either only some galaxies have periods of high icd or all galaxies go through such a phase , but the period of elevated icd is very short lived . if all galaxies experience a period of elevated icd at some point during their evolution , then the characteristic time scale of high icd is the product of the time spanned by the look - back time across the redshift range ( 2.4 grys ) and the fraction of sources with high icd . in this work ,
21% of galaxies have @xmath145 . therefore
, if all galaxies go through a high icd phase , then the timescale of high icd is approximately 375 myrs .
this timescale is similar to the predicted clump survival timescale of @xmath146 myrs @xcite .
regardless of the mechanism at work , there is a distinct suppression in the icd of galaxies once a bulge is formed .
galaxies which contain highly star forming clumps match well the description of these high icd objects and clump migration could be a primary cause of icd suppression .
it will take further investigation in the future to better associate galaxies showing high icd with star forming clumps .
in this work , we investigate the dependence of the rest - frame uv and optical icd on other galaxy properties .
we define the icd as the ratio of the square difference of the image flux
intensity values about the mean galaxy color to the square of the total image flux .
it is a measure of the amount of variation in the stellar diversity of a galaxy ; that is separated red and blue stellar populations . using imaging from candels of goods - south and the hudf , we select galaxies between @xmath147 , a magnitude brighter than 25 mag and a requirement that the signal - to - noise be sufficient enough to provide an accurate measurement of the icd ( see section [ sec : qc and selection ] ) .
this selection yields 2601 galaxies .
we present our results from comparing the icd to the stellar mass of the galaxies ( section [ sec : icdvmass ] ) , the morphology based on srisc index ( section [ sec : sersic ] ) and star formation rates ( section [ sec : icdvssfr ] ) .
we then investigate possible causes of the icd including : mergers ( section [ sec : mergers ] ) , physical size ( section [ sec : size ] ) , color gradients ( section [ sec : color gradients ] ) , band - pass shifting effects ( section [ sec : redshift ] ) , attenuation ( section [ sec : attenuation ] ) and clumpy star forming regions ( section [ sec : clumps ] ) .
we attribute icd to heterogeneous stellar populations ( possibly clumps ) with a possible contribution from dust attenuation .
* galaxies have the largest amount of stellar population diversity , as measured by the icd , at stellar masses @xmath148 .
this stellar mass range corresponds well to the range in which local galaxies experience a decrease in their star formation efficiency , and galaxies at @xmath136 which are found to show star forming clumps .
therefore , the stellar population diversity driving the icd could be discrete , highly star - forming regions contrasted against an older , redder , underlying population . *
high icd is preferentially found in galaxies with extended light profiles ( srisc index @xmath1 ) .
this suggests that galaxies only have diverse stellar populations when there is little or no bulge present .
* there is a continual suppression in the mechanism which causes high icd with increasing srisc index .
this could be due formation of clumps , which initially increase the icd , and then migrate toward the center and coalescence into a quiescent bulge .
the presence of this bulge then stabilizes the disk again fragmentation and new clump formation , which suppresses the icd .
the results suggests that some fraction of galaxies with high icd are a result of the combination of structures such as quiescent bulges and star - forming ( perhaps clump ) disks in galaxies .
if this corresponds to the formation of the spiral galaxies indicative of the hubble sequence , then this phase occurs only in galaxies with stellar masses @xmath4 and when the bulge component remains relatively small ( @xmath5 ) .
the authors also wish to thank the anonymous referee whose comments and suggestions significantly improved both the quality and clarity of this work .
this work is based on observations taken by the candels multi - cycle treasury program with the nasa / esa hst , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
we also make use of partial support from nsf ast-0808133 .
we utilize the rainbow cosmological surveys database , which is operated by the universidad complutense de madrid ( ucm ) , partnered with the university of california observatories at santa cruz ( uco / lick , ucsc ) .
several open source resources are used to complete this study python @xcite along with matplotlib @xcite and ipython @xcite . | we use data from the cosmic assembly near - infrared deep extragalactic legacy survey to study how the spatial variation in the stellar populations of galaxies relate to the formation of galaxies at @xmath0 .
we use the internal color dispersion ( icd ) , measured between the rest - frame uv and optical bands , which is sensitive to age ( and dust attenuation ) variations in stellar populations .
the icd shows a relation with the stellar masses and morphologies of the galaxies .
galaxies with the largest variation in their stellar populations as evidenced by high icd have disk - dominated morphologies ( with srsic indexes @xmath1 ) and stellar masses between @xmath2 .
there is a marked decrease in the icd as the stellar mass and/or the srsic index increases . by studying the relations between the icd and other galaxy properties including sizes , total colors , star - formation rate , and dust attenuation ,
we conclude that the largest variations in stellar populations occur in galaxies where the light from newly , high star - forming clumps contrasts older stellar disk populations .
this phase reaches a peak for galaxies only with a specific stellar mass range , @xmath2 , and prior to the formation of a substantial bulge / spheroid .
in contrast , galaxies at higher or lower stellar masses , and/or higher srsic index ( @xmath3 ) show reduced icd values , implying a greater homogeneity of their stellar populations .
this indicates that if a galaxy is to have both a quiescent bulge along with a star forming disk , typical of hubble sequence galaxies , this is most common for stellar masses @xmath4 and when the bulge component remains relatively small ( @xmath5 ) . |
x - ray imaging observations have shown that the x - ray emitting , hot gas in a large fraction of all galaxy clusters reaches high enough densities in the cluster centers that the cooling time of the gas falls below the hubble time , and gas may cool and condense in the absence of a suitable fine - tuned heating source ( e.g. silk 1976 , fabian & nulsen 1977 ) . from the detailed analysis of surface brightness profiles of x - ray images of clusters obtained with the _ einstein _ , _ exosat _ , and _ rosat _ observatories , the detailed , self - consistent scenario of inhomogeneous , comoving cooling flows emerged ( e.g. fabian et al .
1984 , nulsen 1986 , thomas , fabian , & nulsen 1987 , fabian 1994 ) .
the main assumptions on which the cooling flow model is based and some important implications are : ( i ) each radial zone in the cooling flow region comprises different plasma phases covering a wide range of temperatures .
the consequence of this temperature distribution is that gas will cool to low temperature and condense over a wide range of radii .
( ii ) the gas features an inflow in which all phases with different temperature move with the same flow speed .
( iii ) there is no energy exchange between the different phases , between material at different radii , and no heating .
now the first analysis of high resolution x - ray spectra and imaging spectroscopy obtained with _ xmm - newton _ has shown to our surprise that the spectra show no signatures of cooler phases of the cooling flow gas below an intermediate temperature which constitutes a problem for the interpretation of the results in the conventional cooling flow picture ( e.g. peterson et al . 2001 , tamura et al .
another result is that the spectroscopic data are better explained with local isothermality in the cooling flow region ( e.g. bhringer et al .
2001a , matsushita et al .
2001 , molendi & pizzolato 2001 ) also in conflict with the inhomogeneous cooling flow model . here
, we discuss these new spectroscopic results and their implications and point out the way to a new possible model for this phenomenon . the results are mostly based on the detailed observations of the m87 x - ray halo .
a detailed description of this study is provided by bhringer et al .
( 2001b ) .
xmm reflection grating spectrometer ( rgs ) observations of several cooling core regions show signatures of different temperature phases ranging approximately from the hot virial temperature of the cluster to a lower limiting temperature , @xmath0 .
clearly observable spectroscopic features of even lower temperature gas expected for a cooling flow model are not observed .
a1835 with a bulk temperature of about 8.3 kev has @xmath0 around 2.7 kev ( peterson et al . 2001 ) and similar results have been derived for a1795 ( tamura et al .
these results are very well confirmed by _ xmm _ observations with the energy sensitive imaging devices , epn and emos , providing spectral information across the entire cooling core region , yielding the result that ( for m87 , a1795 , and a1835 ) single temperature models provide a better representation of the data than cooling flow models ( bhringer et al . 2001a , molendi & pizzolato 2001 ) also implying the lack of low temperature components . the very detailed analysis of m87 by matsushita et al .
( 2001 and the contribution to this workshop ) has shown that the temperature structure is well described locally by a single temperature over most of the cooling core region , except for the regions of the radio lobes and the very center ( @xmath1 arcmin , @xmath2 kpc ) . among the spectroscopic signatures which are sensitive to the plasma temperature in the relevant temperature range ,
the complex of iron l - shell lines is most important .
1 shows simulated x - ray spectra as predicted for the xmm epn instrument in the spectral region around the fe l - shell lines for a single - temperature plasma at various temperatures from 0.4 to 2.0 kev and 0.7 solar metallicity . there is a very obvious shift in the location of the peak making this feature an excellent thermometer . for a cooling flow with a broad range of temperatures
one expects a composite of several of the relatively narrow line blend features , resulting in a quite broad peak .
2a shows for example the deprojected spectrum of the m87 halo plasma for the radial range 1 - 2 arcmin ( outside the inner radio lobes ) and a fit of a cooling flow model with a mass deposition rate slightly less than 1 m@xmath3 yr@xmath4 as expected for this radial range from the analysis of the surface brightness profile ( e.g. stewart et al .
1984 , matsushita et al .
it is evident that the peak in the cooling flow model is much broader than the observed spectral feature . for comparison fig .
2b shows the same spectrum fitted by a cooling flow model where a temperature of 2 kev was chosen for the maximum temperature and a suitable lower temperature cut - off ( 1.44 kev ) was determined by the fit .
the very narrow temperature interval ( almost isothermality ) is well consistent with the narrow peak .
a similar result is obtained for other clusters , e.g. a1795 as shown in fig .
since this diagnostics of the temperature structure is essentially based on the observation of metal lines , an inhomogeneous distribution of the metal abundances in the cluster icm and a resulting suppression of line emission at low temperatures was suggested as a possible way to reconcile the above findings with the standard cooling flow model by fabian et al .
( 2001a and contribution in these proceedings ) . as shown by bhringer et al .
( 2001b ) such a scenario will still result in a relatively broad fe l - line feature and does not solve the problem in this case of m87 .
another possible attempt to obtain consistency is to allow the absorption parameter in the fit to adjust freely .
this is demonstrated in fig . 2c with the same observed spectrum where the best fitting absorption column density is selected in such a way by the fit that the absorption edge limits the extent of the fe - l line feature towards lower energies .
this is actually the general finding with asca observations which has shown two possible options for the interpretation of the spectra of cluster core regions : ( 1 ) an interpretation of the results in form of an inhomogeneous cooling flow model which than necessarily includes an internal absorption component ( e.g. allen 2000 , allen et al .
2001 ) , or ( 2 ) an explanation of the spectra in terms of a two - temperature component model ( e.g. ikebe et al .
1997 , 1999 , makishima 2001 ) where the hot component is roughly equivalent to the hot bulk temperature of the clusters and the cool component corresponds approximately to @xmath0 .
thus for the cooling flow interpretation to work and to produce a sharp fe - l line feature as observed , the absorption edge has to appear at the right energy and therefore values for the absorption column of typically around @xmath5 @xmath6 are needed ( e.g. allen 2000 and allen et al .
2001 who find values in the range @xmath7 @xmath6 ) .
it is therefore important to perform an independent test on the presence of absorbing material in the cluster cores .
thanks to _ chandra _ and _ xmm - newton _ we can now use central cluster agn as independent light sources for probing . using the nucleus and jet of m87 ( with _ xmm
_ , see fig .
3 ) and the nucleus of ngc1275 ( with _ chandra _ ) we find no signature of internal absorption . thus at least for these two cases it is difficult to argue for internal absorption to obtain consistency of the observations with the cooling flow model .
in view of these difficulties of interpreting the observations with the standard cooling flow model , we may consider the possibility that the cooling and mass deposition rates are much smaller than previously thought , that is reduced by at least one order of magnitude . to decrease
the mass condensation under energy conservation some form of heating is clearly necessary .
three forms of heat input into the cooling flow region have been discussed : ( i ) heating by the energy output of the central agn ( e.g. pedlar et al .
1990 , binney & tabor 1993 , mcnamara et al .
2000 , \(ii ) heating by heat conduction from the hotter gas outside the cooling flow ( e.g. tucker & rosner 1983 , bertschinger & meiksin 1986 ) , and ( iii ) heating by magnetic fields , basically through some form of reconnection ( e.g. soker & sarazin 1990 , makishima et al .
the latter two processes depend on poorly known plasma physical conditions and are thus more speculative .
the energy output of the central agn , however , can be determined as shown below .
a heating scenario can only successfully explain the observations if among others the two most important requirements are met : ( i ) the energy input has to provide sufficient heating to balance the cooling flow losses , that is about @xmath8 to @xmath9 erg in 10 gyr or on average about @xmath10 erg s@xmath4 , and ( ii ) the energy input has to be fine - tuned .
too much heating would result in an outflow from the central region and the central regions would be less dense than observed .
too little heat will not reduce the cooling flow by a large factor . therefore the heating process has to be self - regulated :
mass deposition triggers the heating process and the heating process reduces the mass deposition .
@xmath11 [ tab1 ] further constraints are discussed by bhringer et al .
( 2001b ) .
the total energy input into the icm by the relativistic jets of the central agn can be estimated by the interaction effect of the jets with the icm by means of the scenario described in churazov et al .
it relies on a comparison of the inflation and buoyant rise time of the bubbles of relativistic plasma which are observed e.g. in the case of ngc 1275 ( bhringer et al .
1993 , fabian et al .
2001b ) and requires as observational input parameters the bubble size , @xmath12 , the ambient pressure , @xmath13 , and the keplerian velocity at the bubble radius in the cluster , @xmath14 .
the parameters and the estimated total energy output is given in table 1 for three examples , m87 , perseus , and hydra a. these values for the energy input have to be compared with the energy loss in the cooling flow , which is of the order of @xmath15 erg s@xmath4 for m87 and about @xmath16 erg s@xmath4 for perseus .
thus in these cases the energy input is larger than the radiation losses in the cooling flow for at least about the last @xmath17 yr .
we have , however , evidence that this energy input continued for a longer time with evidence given by the outer radio halo around m87 with an outer radius of 35 - 40 kpc ( e.g. kassim et al . 1993 , rottmann et al . 1996 ) .
owen et al . ( 2000 ) give a detailed physical account of the halo and model the energy input into it .
they estimate the total current energy content in the halo in form of relativistic plasma to @xmath18 erg and the power input for a lifetime of about @xmath17 years , which is also close to the lifetime of the synchrotron emitting electrons , to the order of @xmath16 erg s@xmath4 , consistent with our estimate .
the very characteristic sharp outer boundary of the outer radio halo of m87 , noted by owen et al .
( 2000 ) , has the important implications , that this could not have been produced by magnetic field advection in a cooling flow .
thus , we find a radio structure providing evidence for a power input from the central agn into the halo region of the order of about ten times the radiative energy loss rate over at least about @xmath17 years ( for this representative example of m87 ) .
the energy input could therefore balance the heating for at least about @xmath19 years .
the observation of active agn in the centers of cooling flows is a very common phenomenon . e.g.
ball et al .
( 1993 ) find in a systematic vla study of the radio properties of cd galaxies in cluster centers , that 71% of the cooling flow clusters have radio loud cds compared to 23% of the non - cooling flow cluster cds .
therefore we can safely assume that the current episode of activity was not the only one in the life of m87 and its cooling flow .
the mechanism for a fine - tuned heating of the cooling flow region should most probably be searched for in a feeding mechanism of the agn by the cooling flow gas . the most simple physical situation would be given if simple bondi type of accretion from the inner cooling core region would roughly provide the order of magnitude of the power output that is observed and required .
using the classical formula for spherical accretion from a hot gas by bondi ( 1952 ) we can obtain a very rough estimate for this number . for the proton density near the m87 nucleus ( @xmath20 arcsec ) of about 0.1 @xmath21 , a temperature of about @xmath22 k ( e.g. matsushita et al . 2001 ) , and a black hole mass of @xmath23 m@xmath3 ( e.g. ford et al . 1994 ) we find a mass accretion rate of about 0.01 m@xmath3 yr@xmath4 and an energy output of about @xmath24 erg s@xmath4 , where we have assumed the canonical value of 0.1 for the ratio of the rest mass accretion rate to the energy output .
the corresponding accretion radius is about 50 pc ( @xmath25 arcsec ) .
this accretion rate is more than a factor of 1000 below the eddington value and thus no reduction effects of the spherical accretion rate by radiation pressure has to be expected .
small changes in the temperature and density structure in the inner cooling core region will directly have an effect on the accretion rate .
therefore we have all the best prospects for building a successful self - regulated agn - feeding and cooling flow - heating model .
several observational constraints have let us to the conclusion that the mass deposition rates in galaxy cluster cooling cores are not as high as previously predicted .
the new x - ray spectroscopic observations with a lack of spectral signatures for the coolest gas phases expected for cooling flows and the lower mass deposition rates indicated at other wavelength bands than x - rays are more consistent with mass deposition rates reduced by one or two orders of magnitude below the previously derived values . this can , however , only be achieved if the gas in the cooling flow region is heated .
the most promising heating model is a self - regulated heating model powered by the large energy output of the central agn in most cooling flows . most of the guidance and the support of the heating model proposed here ( based on concepts developed in churazov et al .
2000 , 2001 ) is taken from the detailed observations of a cooling core region in the halo of m87 and to a smaller part from the observations in the perseus cluster .
these observations show that the central agn produces sufficient heat for the energy balance of the cooling flow , that the most fundamental and classical accretion process originally proposed by bondi ( 1952 ) provides an elegant way of devising a self - regulated model of agn heating of the cooling flow , and that most of the further requirements that have to be met by a heating model to be consistent with the observations can most probably be fulfilled . since these ideas are mostly developed to match the conditions in m87 , it is important to extent such detailed studies to most other nearby cooling flow clusters . in this new perspective
the cooling cores of galaxy clusters become the sites where most of the energy output of the central cluster agn is finally dissipated .
strong cooling flows should therefore be the locations of agn with the largest mass accretion rates .
while in the case of m87 with a possible current mass accretion rate of about 0.01 m@xmath3 y@xmath4 the mass addition to the black hole ( with an estimated mass of about @xmath26 m@xmath3 ) is a smaller fraction of the total mass , the mass build - up may become very important for the formation of massive black holes in the most massive cooling flows , where mass accretion rates above 0.1 m@xmath3 y@xmath4 become important over cosmological times . | new x - ray observations with xmm - newton show a lack of spectral evidence for large amounts of cooling and condensing gas in the centers of galaxy clusters believed to harbour strong cooling flows . here , we explore these diagnostics of the temperature structure of cooling cores with xmm - spectroscopy .
we further find no evidence of intrinsic absorption in the center of the cooling flows of m87 and the perseus cluster .
to explain these findings we consider the heating of the core regions of clusters by jets from a central agn .
we find that the power of the agn jets as estimated by their interaction effects with the intracluster medium in several examples is more then sufficient to heat the cooling flows .
we explore which requirements such a heating model has to fulfill and find a very promising scenario of self - regulated bondi accretion of the central black hole . in summary
it is argued that most observational evidence points towards much lower mass deposition rates than previously inferred for cooling flow clusters .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
[ p. monsky , 1970 ] a square can not be cut illustrates this . ] into an odd number of triangles of equal areas .
the only known proof of this theorem was published by monsky in 1970 @xcite . the proof is based on two key ideas : the sperner s lemma and the coloring of the plane in three colors based on a 2adic valuation .
after that , several generalizations of monsky s results appear .
the first generalization was conjectured by stein and proved by monsky in 1990 @xcite .
it claims that a centrally symmetric polygon can not be cut into an odd number of triangles of equal areas .
though it is based on the same idea of 3-coloring , this proof is technically more challenging than the proof in the case of the square and uses a non - trivial homological technique . in 1994
bekker and netsvetaev proved similar statement in higher dimensions @xcite .
to state another generalization we need a definition .
let us call a finite union of squares of area 1 with integer coordinates of vertices a _
polyomino_. first , stein proved in 1999 @xcite that a polyomino of odd area can not be cut into an odd number of triangles of equal areas , and in 2002 praton @xcite proved the same for an even - area polyomino . in 2000
stein @xcite made a conjectural generalization of theorem 1 , see also @xcite .
let @xmath0 be a plane polygon with clockwise oriented boundary .
@xmath0 is called _ balanced _ if its edges can be divided into pairs so that in each pair edges are parallel , equal in length and have opposite orientation ( the edges are oriented , their orientation comes from the orientation of the boundary ) .
now we are ready to formulate the stein conjecture .
[ s. stein , 2000 ] [ stein conjecture ] a balanced polygon can not be cut into an odd number of triangles of equal areas . in this note
we will present a proof of a partial case of conjecture [ stein conjecture ] .
namely , we will prove the following theorem .
[ non - equidissectibility of a balanced lattice polygon ] [ main ] consider a balanced polygon @xmath0 of the integer odd area and assume that the coordinates of all the vertices are integer numbers
. then @xmath0 can not be cut into an odd number of triangles of equal areas .
for an example of a balanced lattice polygone of area 15 , see fig .
[ lattice ] .
the proof of nonexistence of equidissection of a balanced lattice polygon consists of several steps .
in section @xmath1 we review the coloring of the plane in three colors introduced by monsky . in section @xmath2
we introduce a notion of a _ degree _ of a broken line .
it is an integer that depends both on a coloring and a broken line .
we prove that if a polygon can be cut into triangles with nonnegative @xmath1adic valuations of araes , then its degree is @xmath3 . in section
@xmath4 previous results are applied to the case of a lattice polygon .
the proof of the nonexistence of equidissection of a balanced lattice polygon is finished in section @xmath5 . in the appendix
we show connections between tropical geometry and @xmath2-colorings of projective plane . .
my gratitude goes to sergei tabachnikov for inspiring me to write this article . also to sherman stein for proposing the conjecture and for his constructive criticism of my nascent ideas .
i am especially grateful to nikolai mnev , without whose guidance and support this article would not have been possible .
the main tool for us will be a special type of coloring of a plane in 3 colors . to begin with , let us recall the notion of a discrete valuation and sketch its basic properties .
a function @xmath6 is a 2adic valuation on the field of real numbers if for any two numbers @xmath7 the following properties hold : property 1 : @xmath8 , @xmath9 , property 2 : @xmath10 , property 3 : if @xmath11 then @xmath12 , property 4 : @xmath13 , property 5 : it extends standart 2adic valuation on rationals : @xmath14 the existence of such a function follows from the theorem of the extension of valuations , see @xcite .
this function is not unique and its construction is based on the axiom of choice .
our goal now is to construct a family of 3colorings of a plane ( we will call these colorings `` tropical '' ) with two properties : * ( p 1 ) * on any line points of only two colors occur . * ( p 2 ) * for any triangle with vertices having all 3 different colors its area has a negative 2adic valuation . let us color points in the plane in three colors @xmath15 according to the following rule : a point @xmath16 with coordinates @xmath17 is colored in color @xmath18 , if @xmath19 , in color @xmath0 , if @xmath20 , in color @xmath21 , if @xmath22 .
+ this defines a map from the plane to a three - element set @xmath23 on fig .
3 the way of coloring is presented in coordinates @xmath24 . for any area - preserving affine transformation @xmath25
, we can define another coloring @xmath26 by the rule @xmath27 this defines a family of 3colorings , which we will call tropical .
[ one ] for the 3coloring @xmath26 properties * p1 * and * p2 * hold .
* p2 * @xmath28 * p1*. + if there were three points of different colors on the same line , they would form a triangle of area @xmath3 . because @xmath29 , this is impossible . + * p2*. for a coloring @xmath26 we need to prove that for any triangle @xmath30 , whose image under @xmath31 has vertices of all three different colors , the following holds true : @xmath32 since @xmath31 is area - preserving , it is enough to prove that @xmath33 has area with negative valuation .
suppose that the triangle @xmath34 has vertex @xmath35 of color a , @xmath36 of color b and @xmath37 of color c. then , its area is equal to @xmath38 application of the properties of valuation and the definition of the coloring leads to @xmath39 @xmath40 @xmath41 therefore , @xmath42
given a tropical coloring @xmath26 , one can construct a degree map associated with it .
it assigns an integer number to any oriented broken line .
let @xmath43 be a complete graph with @xmath44 vertices considered as 1dimensional simplicial complex .
suppose that we are given a 1dimensional simplicial complex @xmath45 and a map @xmath46 sending vertices of @xmath45 to vertices of @xmath43 . then , this map can be extended to a continuous map from complex @xmath45 to @xmath43 according to the following rules : * vertex @xmath47 is sent to point @xmath48 . * edge @xmath49 is sent to edge @xmath50 by a linear map determined by its ends .
this map is , obviously , a continuous simplicial map from one simplicial complex to the other .
we will use the same letter for both the original map ( coloring ) and the extended one . for
the following , let us fix a tropical coloring @xmath26 .
, width=480 ] any closed broken line @xmath51 has a natural structure of a simplicial complex .
this complex is homeomorphic to a circle .
its vertices are 3colored by @xmath26 .
the extension of @xmath26 gives a continuous map from topological circle @xmath52 to topological circle @xmath53 .
we denote by @xmath54 its topological degree .
see fig . [ class ] . for a polygon @xmath55
we denote its boundary by @xmath56 .
[ two ] if a polygon @xmath55 can be cut into triangles , whose areas have non negative 2adic valuations , then for every tropical coloring @xmath26 @xmath57 suppose that polygon @xmath55 has a triangulation @xmath58 with each triangle having non - negative valuation of area .
this triangulation carries a natural structure of a 1dimensional simplicial complex , induced from the plane , with vertices colored by @xmath26 .
@xmath56 is a subcomplex of @xmath58 homeomorphic to a circle , so there is a class @xmath59 \in h_1(\mathcal{t},\mathbb{z})$ ] corresponding to @xmath56 .
the degree @xmath60 is equal to the image of the class @xmath59 $ ] under the map @xmath61 one can orient all triangles in the cut in a coherent way .
then the triangles sharing an edge will induce opposite orientations on this edge . adding up the classes of all triangles boundaries we obtain the class of @xmath56 @xmath62=\sum_{\triangle \in \mathcal{t}}[\partial \triangle]\ ] ] and applying @xmath26 we get : @xmath63)=\sum_{\triangle \in \mathcal{t}}\pi^\mathcal{a}([\partial \triangle]).\ ] ] since any @xmath64 has non negative valuation of area , at least two of its vertices are of the same color according to lemma [ one ] .
so , @xmath65)=0 $ ] for any triangle in @xmath58 and thus @xmath63)=deg(\partial m,\pi^\mathcal{a})=0.\ ] ]
points with integer coordinates in a plane form a two dimensional lattice in @xmath66 , we will call it @xmath67 .
we call a polygon @xmath55 or a closed broken line @xmath52 _ lattice _ if all its vertices have integer coordinates .
let us denote by @xmath68 a simplicial complex with four vertices labeled by elements of the group @xmath69 and edjes connecting each two of its vertices .
we can map @xmath67 to @xmath68 by the map @xmath70 for any lattice broken line @xmath52 we can consider its image under the map @xmath71 using the construction from the previous section .
this map induces a map on simplicial homology groups : @xmath72 the image of @xmath73 in the group @xmath74 will be denoted by @xmath75 and called the _ class _ of broken line @xmath76 . [
three ] if a lattice polygon m can be dissected into triangles , whose areas have non - negative 2adic valuations , then the map @xmath71 sends @xmath77 to @xmath3 .
let us denote the points of @xmath68 in the following way : @xmath78 the three cycles @xmath79 @xmath80 @xmath81 generate @xmath82 let us suppose that @xmath83 for an integer number its 2adic valuation is always nonnegative and it is equal to zero if and only if the number is odd . from the definition of the 2adic valuation it is clear that for a point @xmath84 @xmath85 @xmath86 @xmath87 @xmath88 this means that the map @xmath89 is correctly defined on @xmath90 and @xmath91 .
any area - preserving affine transformation @xmath92 acts on the four vertices of @xmath90 by permutation , a simple check shows that @xmath93 .
we will apply lemma [ two ] to the three colorings corresponding to the following area preserving affine transformations : @xmath94 @xmath95 @xmath96 by lemma [ two ] , @xmath97 .
@xmath98 @xmath99 @xmath100
so , @xmath101 .
similarly , applying the same procedure to affine transformations @xmath102 and @xmath103 , we get @xmath104 and @xmath105 .
for two vectors @xmath106 and @xmath107 their _ wedge product _ is defined as the oriented area of the parallelogram formed by these vectors .
it can be calculated as the determinant : @xmath108 for any closed broken line @xmath109 we define its generalised area by @xmath110 for a non selfintersecting broken line the notion defined above gives the oriented area of a polygon , which is bounded by the broken line .
[ three+ ] for a lattice parallelogram @xmath111 the following is true : * if area of @xmath111 is even , then @xmath112 * if area of @xmath111 is odd , then @xmath113 the parallelogram can be cut into two equal triangles of the integer area .
the application of lemma [ two ] gives the first statement . to prove the second statement
, we will show that if the parallelogram @xmath111 has odd area , then all its pairs of coordinates of vertices are different modulo @xmath1 . if the vertices of the parallelogram have coordinates @xmath114 , @xmath115 , then its area is equal to @xmath116 @xmath117 from this formula , it is clear that if there are two vertices with both @xmath118 and @xmath119 coordinates being conjugate modulo @xmath1 , then @xmath120 is even .
so , vertices of the parallelogram are colored in colors a , b , c , c. depending on the order in which these colors follow each other we obtain one of the cycles @xmath121 @xmath122 @xmath123
. the following lemma generalizes lemma [ three+ ] [ four ] if @xmath0 is a balanced lattice polygon , then the image of its boundary under the map @xmath71 is representing a class @xmath124 in the group @xmath125 , which is lying in a subgroup of index @xmath1 generated by @xmath126 : @xmath127 for some @xmath128 .
+ furthermore , @xmath129 parallelograms are basic examples of balanced polygons and we have seen that lemma [ four ] holds true for them .
now , we are going to show that any balanced polygon is built from parallelograms in some sense .
for this we need to describe an action of group @xmath130 on the set of broken lines . for a broken line @xmath109 ,
let us denote by @xmath131 the side vector of @xmath52 ( here @xmath132 ) .
any @xmath133 acts on the set of broken lines according to the rule : @xmath134 with @xmath135 , and for each @xmath136 @xmath137 this action sends balanced broken lines to balanced and lattice to lattice . let @xmath138 denote a transposition @xmath139 . it is well known that a set @xmath140 generates @xmath130 . one can check that @xmath141 and @xmath142 here p is a parallelogram @xmath143 , with @xmath144 these properties guarantee that broken lines @xmath52 and @xmath145 satisfy the conditions of lemma [ four ] simultaneously .
since lattice polygon @xmath0 is balanced , number @xmath44 of its vertices is even and sides of @xmath0 can be indexed by numbers @xmath146 and @xmath147 so that @xmath148 and sides with indices @xmath149 and @xmath150 are parallel , equal in length and inherit opposite orientations from the polygons boundary .
numbers @xmath149 and @xmath150 are just natural numbers from @xmath151 to @xmath148 , so one can consider permutation @xmath152 in the broken line @xmath153 after any side with an odd number goes the side parallel and equal to it and having the opposite direction . both the area and
the class of @xmath153 in @xmath154 is equal to @xmath3 .
since @xmath155 can be presented as a product of transpositions @xmath138 , @xmath156 satisfies the conditions of lemma [ four ] .
now we are ready to finish the proof of our main result .
suppose that for a balanced lattice polygon @xmath0 of integer odd area there exist a cut into an odd number of triangles of equal areas . if @xmath157 and the number of triangles is equal to @xmath158 , then the area of each of them is @xmath159 .
since @xmath160 and @xmath158 are odd numbers , @xmath161 according to lemma [ three ] , the class of the broken line @xmath156 in @xmath162 is equal to @xmath3 , and according to lemma [ four ] there exists @xmath128 for which @xmath163 therefore , @xmath164 and @xmath165 .
this contradicts the oddness of @xmath160 .
in the following section no new results are obtained , so the style will be rather informal . it is more natural to define tropical colorings on @xmath166 the real projective plane .
it is well - known that a point of @xmath166 is defined by its homogenious coordinates a triple of real numbers @xmath167 $ ] with not all @xmath168 equal to @xmath3 . for any nonzero @xmath169 triples
@xmath167 $ ] and @xmath170 $ ] define the same point .
one can define a * momentum map * from the projective plane to a triangle @xmath171 in the plane with vertices ( 1,0),(0,1 ) and ( 0,0 ) : @xmath172 defined by the formula @xmath173)=\frac{2^{-\nu_2(x ) } \big ( 1,0\big ) + 2^{-\nu_2(y ) } \big ( 0,1\big)+ 2^{-\nu_2(z ) } \big ( 0,0\big)}{2^{-\nu_2(x)}+2^{-\nu_2(y)}+2^{-\nu_2(z)}}.\ ] ] one can check that the image of any line in @xmath166 under the momentum map is a union of three segments sharing a common end . for each segment
the remaining end is lying on a side of triangle @xmath171 and the whole segment is contained within a line , passing through the vertex of @xmath171 opposite to the side . the image of the line @xmath174 cuts t into three pieces , whose points we color in three colors a , b and c.
now we can color each point in @xmath166 in the color of its image under the momentum map .
this coloring is the same as coloring @xmath89 constructed in the beginning of the paper if considered on the affine chart of @xmath166 with @xmath175 .
[ a. hales , e. straus , 1982 ] [ hs ] let @xmath21 be a set of algebraic curves in @xmath66 having the same newton polygon @xmath111 with @xmath44 integer points inside ( @xmath21 is a @xmath44dimensional linear system of algebraic curves ) .
then there exists a coloring of @xmath66 in @xmath176 colors such that no curve in @xmath21 contains all @xmath176 colors and no color is confined within a curve in @xmath21.dimensional linear systems of algebraic curves without based points . ] in @xcite such colorings are constructed algebraically ; here we will give a tropical explanation of this construction .
by @xmath177 we mean @xmath178 .
let us consider algebraic torus @xmath179 and `` a momentum map '' : @xmath180 where @xmath181 there exists a curve with newton polygon @xmath111 , whose image devides @xmath182 into @xmath183 regions .
we assign to them different colors .
it can be proved that the image of any other curve with newton polygon @xmath111 intersects at most @xmath44 regions .
now , we can color each point in @xmath184 in the color of the region of @xmath66 containing its image .
this coloring is the same as that constructed by hales and straus .
this illustration shows that colorings constructed in @xcite are natural from the perspective , suggested by tropical geometry .
unfortunatelly , it does not lead to an easier way of proving theorem [ hs ] , because of both combinatorial difficulties in analyzing the way in which two tropical curves intersect and algebraic difficulties in extending the colorings from @xmath185 to the whole @xmath186 . | in this paper we show that a lattice balanced polygon of odd area can not be cut into an odd number of triangles of equal areas .
first result of this type was obtained by paul monsky in 1970 .
he proved that a square can not be cut into an odd number of triangles of equal areas . in 2000
sherman stein conjectured that the same holds for any balanced polygon .
we also show connections between the equidissection problem and tropical geometry . |
analysis of the long - term behavior of flows can be broadly classified into geometric methods and statistical methods .
geometrical methods include the determination of fixed points , periodic orbits , and invariant manifolds .
invariant manifolds of fixed points or periodic orbits act as barriers to transport as trajectories may not cross the manifolds transversally .
statistical methods include determining the distribution of points in very long trajectories of a very large set of initial points ( i.e. a physical invariant measure @xcite , often possessing an invariant density ) and the identification of meta - stable or almost - invariant sets @xcite .
almost - invariant sets partition the phase space into almost dynamically disconnected regions and are important for revealing global dynamical structures that are often invisible to an analysis of trajectories .
these metastable dynamics also go under the names of persistent patterns , or strange eigenmodes , both of which are realisable as eigenfunctions of a transfer operator .
frequently , the boundaries of maximally almost - invariant sets ( those sets which are locally closest to invariant sets ) coincide with certain invariant manifolds @xcite .
our focus in the present paper is on statistical methods , although we demonstrate via case studies the relationships to geometrical methods .
the most commonly used tool for statistical methods is the transfer operator ( or perron - frobenius operator ) . fixed points of the transfer operator correspond to invariant densities , while eigenfunctions corresponding to real positive eigenvalues strictly less than one provide information on almost - invariant sets . in practice
, one typically constructs a finite - rank numerical approximation of a transfer operator and computes large spectral values and eigenfunctions for this finite - rank operator .
the construction of the finite - rank approximation requires the integration of many relatively short trajectories with initial points sampled over the domain of the flow .
it is this use of _ short _ trajectories that gives the transfer operator approach additional stability and accuracy when compared with computations based upon very long trajectories .
long trajectories continually accumulate small errors from imperfect numerical integration and finite computer representation of numbers ; these small errors quickly grow in chaotic flows .
while the transfer operator approach is very stable , it still requires the computation of many small trajectories which can be very time consuming in some systems .
the approach we describe in the present work obviates the need for any trajectory integration at all and works directly with the vector field .
our approach exploits the fact that the evolution of probability densities @xmath0 can be described by generalized solutions of the abstract cauchy problem @xmath1 .
the perron
frobenius operator is the evolution operator of this equation , and has the same eigenfunctions as the operator @xmath2 .
the operator @xmath2 is an unbounded hyperbolic ( if the underlying dynamics is deterministic ) or elliptic ( if the deterministic dynamics is perturbed by white noise ) partial differential operator .
standard techniques allow us to approximate the eigenmodes we are interested in : finite difference , finite volume , finite element and spectral methods yield such discretizations ; see @xcite and the references therein .
an outline of the paper is as follows . in section 2
we provide background on the infinitesimal operator arising from smooth vector fields and describe conditions under which the operator generates a semigroup of transfer operators . in order to obtain formal results
, we will require the addition of a small amount of diffusion to the deterministic flow . in the small diffusion setting we discuss existence of invariant densities and spectral results for the associated infinitesimal operator .
section 3 describes a spectral galerkin method for the approximation of the infinitesimal operator .
we apply results from the numerical analysis of advection - diffusion pdes to show that our galerkin method approximates the true eigenfunctions of the infinitesimal operator as our galerkin basis becomes increasingly refined .
we can also show that the convergence rate is spectral ; that is , faster than any polynomial .
section 4 describes an ulam - based galerkin method for approximating the infinitesimal operator .
this ulam - based approach is new and shares some similarities with finite - difference schemes . while we can not show convergence of this approximation scheme , the numerical results obtained are extremely fast and accurate .
sections 58 detail the practical application of our two approximation methods to flows in one- , two- , and three- dimensional domains . we demonstrate the spectral accuracy of our spectral galerkin method and compare with the accuracy of the ulam - based galerkin method .
we also compare the accuracy vs. computational effort of our two new approaches with standard transfer operator approaches .
the natural relationships between the outputs of our statistical methods and geometric objects such as the vector field and invariant manifolds are also elucidated in each case study .
we conclude in section 9 .
let the domain @xmath3 of our flow be a smooth compact manifold and @xmath4 the ( normalized ) lebesgue measure on @xmath5 .
denote by @xmath6 the vector field generating the flow and by @xmath7 , @xmath8 , the flow , i.e. @xmath9 represents the location of a trajectory beginning at @xmath10 after flowing for @xmath8 time units .
one has that @xmath11 .
note that provided that the components @xmath12 have continuous derivatives
the function @xmath7 is a diffeomorphism for every @xmath8 .
invariant sets are structures of dynamical interest .
a set @xmath13 is called _ invariant _ , iff @xmath14 for all @xmath15 .
also , one asks how the flow changes probability measures .
sample @xmath16 according to a probability measure @xmath17 ; the distribution of @xmath18 is then given by @xmath19 .
special attention is to be drawn to _
invariant measures _ , which do not change under the dynamics ( @xmath20 ) .
invariant measures @xmath17 are called _ ergodic _ if invariant sets have either zero or full measure , i.e. if @xmath13 satisfies @xmath14 then @xmath21 .
an even more restricted class of ergodic measures are the _ physically relevant _ ( or _ natural _ ) ones , satisfying @xmath22 for all @xmath23 continuous and @xmath24 with @xmath25 .
one has that absolutely continuous is absolutely continuous ( with respect to @xmath4 , if not specified further ) , if there is an @xmath26 such that for all @xmath4-measurable @xmath27 we have @xmath28 .
the function @xmath29 is called the density of @xmath17 . ]
ergodic measures are natural invariant measures . the density of an invariant measure is called the _ invariant density_.
when looking for invariant densities , one can rephrase the action of the flow on measures as an action on densities .
if @xmath29 denotes the density of @xmath17 , then @xmath29 is evolved by the flow as @xmath30 where @xmath31 denotes @xmath32 for a matrix @xmath33 .
the linear operator @xmath34 is known as a _ transfer operator _ or the _ perron - frobenius operator _ associated with the flow @xmath35 .
note that invariant densities are fixed points of @xmath36 .
for each @xmath8 we have 1 .
@xmath36 is linear , 2 .
@xmath37 if @xmath38 and , 3 .
@xmath39 for all @xmath38 , where @xmath40 is the @xmath41 norm .
i.e. @xmath36 is a _
markov operator_. moreover , these properties also imply @xmath42 , hence @xmath36 is a contraction and its eigenvalues lie inside the unit disc .
in fact , since the flow @xmath43 is a bijection , we have @xmath39 for all @xmath44 . to see this , note that @xmath45 for @xmath46 and @xmath13 measurable .
now write @xmath47 , with @xmath48 , then @xmath49 , where @xmath50 is the support of @xmath51 . now use properties ( ii ) , ( iii ) and ( [ eq : bijection ] ) with @xmath52 .
a consequence of this is that the eigenvalues of @xmath36 lie on the unit circle .
the transfer operator also inherits some ( semi)group properties of the flow @xmath43 .
let @xmath53 be a banach space .
a one parameter family @xmath54 of bounded linear operators @xmath55 is called a _ semigroup _ on @xmath56 , if * @xmath57 ( @xmath58 denoting the identity on @xmath56 ) , * @xmath59 for all @xmath60 .
further , if @xmath61 , the family is called a semigroup of contractions . if @xmath62 for every @xmath63 , then @xmath64 is a _ continuous semigroup _ ( _ @xmath65 semigroup _ ) . the transfer operator @xmath36 is a @xmath65 semigroup of contractions on @xmath41 , see @xcite for a proof ( in particular remark 7.6.2 for the continuity ) . for a semigroup @xmath64 we define the operator @xmath66 by @xmath67 with @xmath68 being the linear subspace of @xmath56 where the above limit exists . the operator @xmath2 is called the _ infinitesimal generator _ of the semigroup . for @xmath36 , the infinitesimal generator turns out to be ( provided the @xmath69 are continuously differentiable ) @xmath70 see @xcite . the following result ( see eg .
theorem 2.2.4 @xcite ) shows the connection between the eigenvalues of the semigroup operators and their infinitesimal generator : let @xmath64 be a @xmath65 semigroup and let @xmath2 be its infinitesimal generator .
then @xmath71 where @xmath72 denotes the point spectrum of the operator .
the corresponding eigenvectors are identical .
this has important consequences for invariant densities : the function @xmath29 is a invariant density of @xmath36 for all @xmath73 if and only if @xmath74 .
according to the discussion at the end of section [ subsec.transfer ] , we have the eigenvalues of @xmath75 lie on the imaginary axis .
in this section we discuss almost - invariant sets and escape rates via a spectral analysis of the infinitesimal generator .
as remarked near the end of section [ subsec.transfer ] , the @xmath41 spectrum of @xmath76 lies on the unit circle , and lacks a spectral gap . applying the spectral mapping theorem
, we see that the spectrum of @xmath77 must be pure imaginary . in the following discussion , to prove formal results , we will add a small random perturbation to @xmath76 .
later , in the numerics section , we will see that our numerical methods introduce a numerical diffusion that plays the role of creating a spectral gap . in many real world situations , a deterministic model of some physical system
is not appropriate .
rather , one should account for the fact that certain external perturbations are present which might be unknown or for which a detailed model would be overly complicated .
often , it is appropriate to account for these influences by incorporating a _ small random perturbation _ into the description . from a theoretical point of view
, this even facilitates the analysis of the system : under certain assumptions , the transfer operator becomes compact .
we therefore now leave the deterministic setting and assume that our dynamical system described by the ordinary differential equation @xmath78 is slightly stochastically perturbed by _
white noise_. an exact mathematical treatment of this topic would require tools which are beyond the scope of this work ( for an exact derivation see @xcite , chapter 11 ) .
we are primarily interested in the time evolution of probability densities .
the following material highlights the relevant formal statements and attempts to point out the intuition behind them . instead of an ordinary differential equation
, we now deal with a stochastic differential equation ( sde ) @xmath79 with @xmath80 and @xmath81 being a @xmath82-dimensional wiener process .
the solutions are time - dependent random variables @xmath83 with values in @xmath84 . just as in the deterministic case
, we look at the evolution of density functions ; now the density functions represent the distribution of random outcomes of the variables @xmath83 .
the density functions @xmath29 satisfy @xmath85 assuming the existence of such a density @xmath86 , where @xmath87 is given , we may again define the transfer operator as @xmath88 .
if the vector field @xmath89 is smooth enough we have following characterization of the density function , the _ fokker - planck _ or _ kolmogorov forward equation _ , cf .
@xcite , 11.6 : @xmath90 [ prop : aepsgenc0semigroups ] the operator @xmath91 ( with neumann boundary conditions ) is the infinitesimal generator of @xmath65 semigroups @xmath92 on @xmath41 , @xmath93 on @xmath94 , and @xmath95 on @xmath96 .
we note that the operator @xmath92 is identical to the transfer operator @xmath97 .
[ [ invariant - densities ] ] invariant densities + + + + + + + + + + + + + + + + + + + again , we are particularly interested in distributions ( densities ) which do not change under the evolution . those are again given by fixed points of @xmath92 or alternatively , by functions in the null space of @xmath91 .
one can show @xcite that @xmath92 is compact and thus the null space of @xmath91 is finite dimensional .
furthermore , due to the white noise , the support of the stochastic transition function associated to @xmath92 is unbounded , and the null space of @xmath91 is one - dimensional , i.e. there is a _
unique _ invariant density ( see again @xcite , theorem 1 ) , characterizing the long term dynamical behavior of ( [ eq : sde ] ) .
we call a set @xmath13 _ almost - invariant _ with respect to a ( not necessarily invariant ) probability measure @xmath98 ( cf .
@xcite ) , if @xmath99 for modest times @xmath15 .
the analogous expression for @xmath35 perturbed by a small random perturbation is @xmath100 for modest times @xmath15 .
we can alternatively characterize this property of a set using an infinitesimal representation of @xmath101 . to this end
let @xmath44 and consider a measurable set @xmath13 .
we define the functional @xmath102 by @xmath103 where @xmath104 is the linear subspace of @xmath41 where the above limit exists .
let @xmath105 .
[ prop : rate ] let @xmath106 be the density of the probability measure @xmath98
. then @xmath107 for measurable @xmath108 as @xmath109 .
we say `` @xmath110 as @xmath111 '' , if @xmath112 . ]
we have @xmath113 where the second equation follows from the fact that @xmath114 is distributed according to @xmath98 ( with density @xmath29 ) , and @xmath115 ; which gives the density @xmath116 for @xmath114 , see section [ ssec : stoch_pert ] .
thus for @xmath109 the claim follows .
correspondingly , a set @xmath108 will be almost invariant with respect to the measure with density @xmath44 , if @xmath117 .
there is a strong connection between almost - invariant sets and the spectrum of the generator .
the following theorem illustrates this with a simple heuristic for producing almost - invariant sets .
[ thm : ai ] suppose that the generator @xmath91 possesses a real eigenvalue @xmath118 with corresponding eigenfunction @xmath29
. then @xmath119 .
define @xmath120 , @xmath121 .
let @xmath122 .
then @xmath123 by proposition 5.7 in @xcite and the spectral mapping theorem we have that @xmath124 where @xmath98 is the probability measure with density @xmath125 .
using proposition [ prop : rate ] we obtain @xmath126 , i.e. @xmath127 and for @xmath109 we obtain the claim . for @xmath128 theorem [ thm : ai ]
yields @xmath129 , which means that the sets @xmath130 and @xmath131 will be almost - invariant with respect to the probability measure with density @xmath125 .
other techniques for extracting sets @xmath132 from the eigenfunction @xmath29 may be found in @xcite .
the papers @xcite also discuss almost - invariance with respect to physical _ invariant _ probability measures , a property that is particularly meaningful when studying typical dynamical behavior . in all cases , the basis for these methods are eigenfunctions of @xmath91 corresponding to ( real ) eigenvalues close to @xmath133 .
if @xmath91 has a complex eigenvalue with real part close to zero , then the corresponding complex eigenfunction may also be used to construct almost - invariant sets .
[ lem : realpart ] let @xmath134 with @xmath135 and let @xmath136 and @xmath137 denote the real and imaginary part of @xmath29 , respectively .
let @xmath46 be such that @xmath138
. then @xmath139 and @xmath140 . from the proof of theorem 2.2.4
@xcite we have @xmath141 .
note , that @xmath142 . by linearity
we have @xmath143 .
thus , @xmath144 the claim follows immediately . hence , if for a @xmath46 we have @xmath145 , then the real ( and imaginary ) part of the corresponding eigenfunction yields a decomposition of the phase space into almost invariant sets in the sense of theorem [ thm : ai ] .
we have seen in the previous section that there is a connection between eigenvalues of @xmath146 close to zero and almost - invariant sets .
there is also a strong connection between the _ escape rate _ of the sets @xmath130 , @xmath131 and the corresponding eigenvalue of @xmath146 . in the deterministic setting ( @xmath147 ) the upper ( resp .
lower ) escape rate of a set @xmath108 under the time-@xmath15 map of the flow @xmath43 is defined as : @xmath148 if both limits are equal , we call @xmath149 the escape rate from @xmath108 . the escape rate is the asymptotic exponential rate of loss of @xmath4-mass from the set @xmath108 .
one might expect that almost - invariant sets have low escape rate and vice - versa , but simple counterexamples in @xcite show that this is not the case .
this is because the notion of almost - invariance as defined above is a finite - time property , while escape rate is an asymptotic quantity .
the @xmath80 version of @xmath150 is @xmath151 where @xmath152 is the standard restriction of the operator @xmath153 to @xmath108 @xcite .
the following theorem provides a recipe for constructing sets with low escape rate from eigenfunctions of @xmath91 with real eigenvalues .
[ thm : er ] suppose that @xmath154 for some @xmath155 and @xmath156
. then @xmath157 where @xmath158 and @xmath159 .
let @xmath29 be scaled such that @xmath160 , and define the ( signed ) measure @xmath98 as @xmath161 .
then @xmath162 and @xmath163 are both probability measures .
for an event @xmath164 define @xmath165 . since @xmath29 is an eigenfunction of @xmath97 with eigenvalue @xmath166 , we have @xmath167 it follows for the summands @xmath168 in the above sum : @xmath169 where in the first equation the action of the transfer operator @xmath97 on the measure @xmath98 is defined through its action on the density @xmath29 of @xmath98 .
the second equation follows from @xmath98 being an eigenmeasure . clearly , @xmath168 is non - positive , and thus @xmath170 holds .
the rest of the proof follows the lines of the one of theorem 2.4 in @xcite . the recipe of theorem [ thm : er ] is in fact the same as in theorem [ thm : ai ] .
the difference is the measure used : in theorem [ thm : ai ] , almost - invariance is computed with respect to the particular measure @xmath98 with density @xmath125 where @xmath29 is the eigenfunction in question , whereas in theorem [ thm : er ] escape is computed with respect to lebesgue measure .
a more natural notion of escape rates in the time - continuous case would be the following : @xmath171 however , for this definition it is more complicated to obtain similar results to the one in theorem [ thm : er ] , and a fuller discussion will appear elsewhere .
one can view with @xmath172 as the `` time sampled version '' of this continuous - time definition .
having seen that certain eigenpairs of the transfer operator ( resp .
infinitesimal generator ) carry the information we are seeking , we describe in the sequel our proposed approximation of these eigenpairs . to this end
, we define finite dimensional approximation spaces and consider the eigenvalue problem projected onto these spaces . throughout this section
we assume that the underlying deterministic vector field @xmath89 is smooth .
we describe here the `` standard '' ulam approach ; see the surveys @xcite for more details .
we partition @xmath5 into @xmath82-dimensional connected , positive volume subsets @xmath173 .
typically , each @xmath174 will be a hyperrectangle or simplex to simplify computations . as an approximation space we consider the space @xmath175 of functions which are piecewise constant on the cells of the partition .
let @xmath176 , @xmath177 be the @xmath94-orthogonal projection onto @xmath178 .
we let @xmath179 , @xmath180 , be the approximate frobenius - perron operator .
note that @xmath181 i.e. the matrix representation @xmath182 of @xmath183 with respect to the basis @xmath184 and multiplication on the left is @xmath185 this matrix is easily constructed numerically using eg .
gaio @xcite .
in the stochastic setting , letting @xmath186 , @xmath187 , one obtains @xmath188 as above . in principle
one can use monte - carlo integration to compute these entries , however , this is not particularly efficient .
we partition @xmath5 as in the standard ulam s method .
we will see that the numerical scheme itself introduces some diffusion and we therefore consider the deterministic generator @xmath2 ( i.e. @xmath147 ) .
we wish to construct an operator @xmath189 that is close in some sense to the operator @xmath2 . motivated by ulam s method
, one would like to form @xmath190 , which unfortunately does not exist , because @xmath191 , cf .
( [ eq : generator ] ) . instead of differentiating w.r.t .
time and then doing the projection , we swap the order of these operations .
let us build the ulam approximation @xmath192 first , which will _ not _ be a semigroup any more , but for fixed @xmath15 it approximates @xmath36 . taking the time derivative , our candidate approximate operator is @xmath193 the following lemma emphasizes the intuition behind this definition : if @xmath192 is a sufficiently good approximation ( for small @xmath15 ) of the markov jump process generated by the dynamics on the sets @xmath174 , then @xmath194 will be the generator of this process . to be exact ,
the following is the case : [ ulamformula1lemma ] the matrix representation of @xmath195 with respect to the basis @xmath196 under multiplication on the left is @xmath197 we consider the action of @xmath36 on @xmath198 . @xmath199
thus under left multiplication we obtain ( [ ulamformula ] ) .
lemma [ ulamformula1lemma ] states that @xmath200 is the _ outflow rate of mass _ from @xmath174 into @xmath201 .
let @xmath202 denote the semigroup generated by @xmath194
. then @xmath203 and @xmath36 are near to each other in the following sense , cf .
@xcite : [ prop : rnearqp ] as @xmath109 @xmath204 for all @xmath205 .
we say `` @xmath206 as @xmath111 '' , if @xmath207 . ]
the following lemma allows us to construct @xmath208 without the computation of the flow @xmath43 .
[ ulamformula2lemma ] for @xmath209 , define @xmath210 to be the the unit normal vector pointing out of @xmath174 into @xmath201 if @xmath211 is a @xmath212-dimensional face , and the zero vector otherwise . the matrix representation of @xmath213 with respect to the basis @xmath196 under multiplication on the left is @xmath214 from ( [ ulamformula ] ) we have for @xmath209 that @xmath215 . denoting @xmath216
we have that @xmath217 where the prime denotes differentiation with respect to @xmath15 .
the quantity @xmath218 is simply the rate of flux out of @xmath174 through the face @xmath219 into @xmath201 and so @xmath220 . for the diagonal elements @xmath221 we have @xmath222 .
note that @xmath223 .
clearly @xmath224 modulo sets of lebesgue measure zero .
thus , @xmath225 .
now , by ( [ ulamformula ] ) , @xmath226 in one dimension , ( [ ulamformula2 ] ) has a particularly simple form . [ onedcorollary ]
let @xmath227 .
assume ( without loss and @xmath228 , we have one or more stable fixed points , and every trajectory converges to one of them .
hence , there is no interesting statistical behavior to analyze . ] ) that @xmath229 .
let @xmath230 and consider the partition @xmath173 with @xmath231 $ ] .
then @xmath232 [ rem : upwind ] clearly , @xmath194 is the spatial discretization from the so - called upwind scheme in finite volume methods ; cf .
the scheme is known to be stable .
stability of finite volume schemes is often related to `` numerical diffusion '' in them .
we demonstrate numerical diffusion on a simple example , for further details we refer to @xcite section 8.6.1 .
set @xmath227 and @xmath233 @xmath234 .
then @xmath235 and @xmath194 is the backward difference scheme .
consider @xmath236 as a vector of values .
for sufficiently smooth @xmath29 we have @xmath237 but @xmath238 hence @xmath239 is a better approximation of @xmath240 , with @xmath241 , and @xmath242 is the infinitesimal generator associated with the sde @xmath243 than of @xmath244 .
that is why one expects quantities computed by @xmath194 to reflect the actual behavior of @xmath242 .
since the above equations contain only _ local _ estimates , for a general nonconstant @xmath89 is the diffusion also spatially varying ; i.e. @xmath245 . note , that this notion of numerical diffusion is easily extended to multiple space dimensions .
the definition allows us to interpret the numerical diffusion in yet another way .
we showed in proposition [ prop : rnearqp ] that @xmath246 is the transition matrix of a markov process near the markov jump process generated by @xmath247 for small @xmath46 .
the discretized fpo @xmath246 can be related to a non - deterministic dynamical system , which , after mapping the initial point , adds some uncertainty to produce a uniform distribution of the image point in the box where it landed ; see @xcite .
thus , this uncertainty resulting from the numerical discretization , equivalent to the numerical diffusion in the upwind scheme , can be viewed as the reason for robust behavior stability .
finally , we show that our constructions ( [ ulamformula2 ] ) and ( [ ulam1dformula ] ) always provide a solution to the system @xmath248 for some @xmath249 .
[ existenceofzerond ] there exists a nonnegative , nonzero @xmath205 so that @xmath248 .
let @xmath250 and note that @xmath251 satisfies @xmath252 note that all row sums of @xmath253 equal zero .
let @xmath254 .
the matrix @xmath255 is nonnegative with all row sums equal to @xmath256 . by the perron - frobenius theorem @xcite ,
the largest eigenvalue of @xmath257 is @xmath256 ( of multiplicity is aperiodic ( there exists @xmath258 such that @xmath259 ) then the eigenvalue @xmath256 is simple and the corresponding eigenvector is positive ( see @xcite ) . ]
possibly greater than 1 ) and one of the corresponding left eigenvectors pair @xmath260 is nonnegative . clearly @xmath260 is a left eigenvector of @xmath253 corresponding to the eigenvalue 0 and thus @xmath261 is a nonnegative left eigenvector corresponding to 0 for @xmath247 .
the eigenfunctions of @xmath91 are smooth .
this motivates the use of smooth approximation functions , e.g. polynomials .
we here outline the general principles of spectral collocation methods ; for a more thorough presentation we refer to @xcite , @xcite or @xcite . choose a family of approximation spaces @xmath262 , such that @xmath263 for all @xmath264 . depending on the type of the phase space , we use two different approximation spaces .
we introduce them in one dimension ; the multidimensional ones can then be constructed by tensor products . in both cases ,
the approximation space comes with an associated set of collocation nodes : * _ periodic domain / uniform grid .
_ we have @xmath265 and restrict ourselves to odd values of @xmath264 .
then the basis we choose for @xmath266 is @xmath267 the associated collocation nodes are the uniform grid @xmath268 . * _ standard domain / chebyshev grid . _ here , @xmath269 $ ] .
the space @xmath266 is spanned by the monomials of order @xmath133 to @xmath264 .
we use chebyshev polynomials as basis functions : @xmath270 together with the chebyshev grid @xmath271 , as collocation nodes .
let @xmath272 and @xmath273 be the interpolation operator for the given collocation nodes .
we define the approximate generator by @xmath274 for both cases we have following : [ thm : spectral ] for @xmath275 let @xmath276 be the best approximation of @xmath29 in @xmath266 w.r.t . the supremum norm @xmath277 . then for each @xmath278 there is a @xmath279 such that @xmath280 convergence then follows from standard results on the analysis of galerkin methods for elliptic differential operators and spectral approximation ( cf .
also @xcite ) : let @xmath5 be a compact tensor product domain with smooth boundary .
let @xmath89 be a smooth vector field on @xmath5 and @xmath91 , @xmath281 defined as above , with neumann ( resp .
periodic ) boundary conditions .
then the eigenvalues and eigenfunctions of @xmath91 are approximated by the corresponding eigenvalues and eigenfunctions of @xmath281 with spectral accuracy .
[ [ why - neumann - boundary - conditions ] ] why neumann boundary conditions ?
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the objects that we are approximating are densities , so there should be no `` loss of mass '' as time goes on . in a closed physical system ,
the flow normal to the boundary is zero .
hence , there is no loss of mass through advection .
the physical meaning of diffusion is a flow of mass in the direction opposite to the gradient and proportional to its magnitude .
thus , no loss of mass via diffusion translates into neumann boundary conditions for the densities : their normal gradients have to be zero at the boundary .
equivalently , one may write as a continuity equation , @xmath282 with @xmath283 being the probability flow or probability current . the condition @xmath284 leads precisely to neumann boundary conditions .
[ ulamalg ] 1 .
partition @xmath5 into connected sets @xmath173 of positive volume .
typically , each @xmath174 will be a hyperrectangle or simplex .
2 . choose @xmath46 and compute the matrix @xmath285 or @xmath286 as described in section [ sec : ulam ] .
estimates of invariant densities for @xmath43 are given by fixed points of @xmath287 ( or @xmath288 ) : let @xmath289 .
then @xmath290 satisfies @xmath291 .
4 . similarly ,
eigenvectors of @xmath287 corresponding to real eigenvalues @xmath292 provide information about almost - invariant sets ( cf .
theorem [ thm : ai ] ) and sets with low escape rates ( cf .
theorem [ thm : er ] ) .
note that @xmath287 typically is a sparse matrix and the number of nonzero entries per column will be determined by lipschitz constants of @xmath43 .
[ ulamalg_gen ] 1 .
partition @xmath5 into connected sets @xmath173 of positive volume .
typically each @xmath174 will be a hyperrectangle or simplex .
2 . compute@xmath293 where one uses standard quadrature rules to estimate the integral .
estimates of invariant densities lie in the null space of @xmath247 .
let @xmath294 ( the existence of such a @xmath295 is guaranteed by lemma [ existenceofzerond ] ) , then @xmath290 satisfies @xmath248 .
4 . similarly
, eigenvectors of @xmath247 corresponding to large real eigenvalues @xmath155 provide information about almost - invariant sets and sets with low escape rates .
note that the discretized generator @xmath247 is a sparse matrix since @xmath296 if @xmath174 and @xmath201 do not share a boundary . 1
. set up a grid @xmath297 , which is the chebyshev grid @xmath298 , @xmath299 , if @xmath300 $ ] , or the equispaced grid @xmath301 , if @xmath227 .
( if the state space @xmath5 is an affine transformation of the above ones , the grid is transformed analogously . ) for multidimensional tensor product spaces the grid is the tensor product of the one dimensional grids .
2 . let @xmath302 denote the lagrange basis on the grid , i.e. all @xmath303 are polynomials of maximal degree @xmath264 with @xmath304 .
denote the interpolation on the grid by @xmath305 , and define the discretization matrix obtained by collocation as @xmath306 note , that the computation of these matrix entries is very simple .
if @xmath227 , we may switch between the evaluation space and the frequency space simply by the fast fourier transformation .
multiplication by the vector field @xmath89 is pointwise multiplication in the evaluation space , computing the derivative is a diagonal scaling in the frequency space .
this can be simply extended to @xmath307 $ ] ( cf . @xcite chapter 8) , as well as for a multidimensional phase space .
3 . compute the left eigenvector @xmath295 of @xmath308 at the eigenvalue of smallest magnitude .
then , @xmath309 approximates the invariant density .
4 . similarly , eigenvectors of @xmath308 corresponding to large real eigenvalues @xmath118 provide information about almost - invariant sets .
once the matrix approximation of the operator or the generator has been computed , we then have to solve the corresponding eigenvalue problem . in ulam s method one
seeks dominant eigenvalues , and standard arnoldi type methods easily provide the interesting part of the spectrum ( we use the ` eigs ` function in matlab ) . for the approximate generator , we look for eigenvalues with small modulus . here , the arnoldi iteration in ` eigs ` uses inverse iteration , i.e. repeatedly solves linear systems of the type @xmath310 .
since the matrix @xmath247 is singular , this is not a well posed problem and we might expect some difficulties with these iterative methods .
a possibility for overcoming them is to solve the eigenvalue problem for @xmath311 , with a suitable shift @xmath312 .
this merely introduces a shift of the spectrum and the eigenfunctions stay the same . for @xmath264 big enough
is obtained from the ulam - based approach , the spectrum of @xmath247 , which is the generator of a markov jump process , is in the left complex half plane .
this does not hold in general for the spectral collocation approach , here we rely on the fast convergence of eigenvalues . ] , the spectrum of @xmath313 is going to be well separated from zero .
the size @xmath264 of the linear system may be large , so a direct solution via lu factorization may not be feasible .
however , in ulam s method for the generator , @xmath247 is a sparse matrix , hence iterative methods ( e.g. gmres ) can be used . for spectral collocation ,
in general , the matrix @xmath247 will not be sparse . nevertheless , in the case where the domain is periodic , the matrix - vector product @xmath314 can be computed fast using the fft . for @xmath82 dimensional systems in general ,
shows us that setting up the approximate generator requires the computation of @xmath315- dimensional integrals , while to set up the ulam - matrix , we compute @xmath82-dimensional integrals . of course , the gain is biggest in one dimension ( zero dimensional integrals are one - point - evaluations ) .
equally importantly , does not require integration of the vector field . for ulam s method and ulam s method for the generator ,
the matrix entries should be computed to an accuracy @xmath316 ; otherwise we can not expect the approximate eigenfunctions to exploit the full potential of the approximation space ( which is an error of @xmath316 for @xmath264 partition elements in a @xmath82 dimensional space ) .
assuming the quadrature rule used to compute the entries to also suffer from the curse of dimension , we expect to need @xmath317 and @xmath318 function evaluations for each matrix entry in the ulam - matrix , and the matrix arising from ulam s method for the generator , respectively .
this explains the third column of table [ tab : error_sources ] below .
a collection of examples is presented in the following .
the examples range over three phase space dimensions , different boundary types , and chaotic and regular dynamics .
we find that the generator approaches are typically superior to the standard ulam approach , producing more accurate results with less computation time .
if the vector field is infinitely differentiable , the spectral collocation approach can be extremely accurate . in our final example
, we find that for flows on complicated attractors , the standard ulam method can take advantage of the attracting dynamics to produce better estimates of invariant densities than the generator approaches . in all cases ,
the operator based methods are far superior to direct trajectory integration for estimating the invariant density .
other structures such as almost - invariant sets can only be identified using the operator approaches .
we start with a one dimensional example , a flow on the unit circle .
the vector field is given by @xmath319 @xmath320 $ ] with periodic boundary conditions , and we wish to compute the invariant density of the system . recall that an invariant density @xmath321 needs to satisfy @xmath322 , where @xmath323 .
the unique solution to this equation is @xmath324 , @xmath325 being a normalizing constant ( i.e. such that @xmath326 ) .
the honours thesis @xcite numerically investigated the estimation of the invariant density for this flow , and other two - dimensional flows , by finding approximate eigensolutions of the infinitesimal generator using finite difference and finite element methods .
here we use the three methods discussed in the previous section to approximate @xmath327 and compare their efficiency relative to one another and a histogram of a long simulation : 1 . the classical method of ulam for the frobenius - perron operator , 2 .
ulam s method for the generator and 3 .
spectral collocation for the generator .
as we have a periodic domain and the vector field is infinitely smooth , we expect spectral collocation to perform very well .
figure [ fig : sineflow ] shows the true invariant density ( dashed line ) , together with its approximations by the four methods . in this 1d case ,
the three methods can each be realized in a few lines of matlab . for illustration purposes ,
we include the code here .
[ [ ulams - method ] ] ulam s method + + + + + + + + + + + + + .... f = @(x ) sin(4*pi*x)+1.1 ; % definition of the vector field n = 32 ; x = 1/(2*n^2):1/(n^2):1 - 1/(2*n^2 ) ; % nodes for spatial integration [ t , y ] = ode23(@(t , x ) f(x),[0 2/n],x ' ) ; % time integration of the nodes i = ceil(max(n*mod(y(end,:),1),1 ) ) ; % cell indices of image points j = reshape(ones(n,1)*(1:n),1,n*n ) ; % construction of transition matrix p = sparse(i , j,1/n , n ,
n ) ; [ v , d ] = eig(full(p ) ) ; % spectrum of transition matrix f_n = @(x ) abs(v(ceil(max(x*n,1)),1 ) ) ; % plot approximate inv . density l1 = quadl(f_n,0,1 ) ; fplot(@(x)f_n(x)/l1,[0,1 ] ) .... the error in ulam s method decreases like @xmath328 for smooth invariant densities @xcite .
thus , we need to compute the transition rates between the intervals to an accuracy of @xmath328 ( since otherwise we can not expect the approximate density to have a smaller error ) . to this end
, we use a uniform grid of @xmath264 sample points in each interval .
this leads to @xmath329 evaluations of the vector field .
for the numbers in figure [ fig : sineflow ] we only counted each point once , i.e. we neglected the fact that for the time integration we have to perform several time steps per point .
[ [ ulams - method - for - the - generator-1 ] ] ulam s method for the generator + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + .... f = @(x ) sin(4*pi*x)+1.1 ; % definition of the vector field n = 32 ; fx = f(0:1/n:1 - 1/n ) ; % evaluation on the boundary nodes a = n*sparse(1:n,1:n ,- fx ) + sparse(1:n,[2:n 1],fx ) ; % assembling the discrete generator [ v , d ] = eig(full(a ' ) ) ; % spectrum [ d , i ] = sort(diag(d ) ) ; f_n = @(x ) abs(v(ceil(max(x*n,1)),i(1 ) ) ) ; % plot approximate inv .
density l1 = quadl(f_n,0,1 ) ; fplot(@(x)f_n(x)/l1,[0,1 ] ) .... here , only one evaluation of the vector field per interval is needed as in one dimension the integration on the boundaries reduces to the evaluation of a single boundary point . on a partition with @xmath264 intervals
, this method then yields an accuracy of @xmath328 .
note that from corollary [ onedcorollary ] it follows that the vector @xmath295 with @xmath330 is a left eigenvector of the transition matrix ( [ ulam1dformula ] ) for the generator at the eigenvalue 0 .
this fact proves pointwise convergence of the invariant density of the discretization towards the real one .
thus ulam s method for the generator is very accurate and stable ; the @xmath41 error is @xmath331 as this is the rate at which approximants created from a basis of characteristic functions converge in @xmath41 to regular functions .
[ [ spectral - collocation ] ] spectral collocation + + + + + + + + + + + + + + + + + + + + .... f = @(x ) sin(4*pi*x)+1.1 ; % definition of the vector field n = 32 ; fx = f(0:1/n:1 - 1/n ) ; % evaluation in collocation nodes e = ifft(eye(n ) ) ; % basis fe = fft(diag(fx)*e ) ; % multiplication by vector field i = [ 0:n/2 - 1 -n/2:-1 ] ; % frequencies d = ( 2i*pi)*i ' ; % differentiation matrix a = -d*ones(1,n).*fe ; % discrete generator in frequency space [ v , lambda ] = eig(a ) ; % spectrum f_n = @(x ) real(exp(x'*i*2i*pi)*v(:,end ) ) ; % approximate density l1 = quadl(f_n,0,1 ) ; fplot(@(x)f_n(x)/l1,[0,1 ] ) % plot .... here , the vector field is evaluated once per grid point ( cf . the second line of the code ) . as predicted by theorem [ thm : spectral ] , the accuracy increases exponentially with @xmath264 ( cf . figure [ fig : sineflow2 ] ) . in figure
[ fig : sineflow2 ] ( left ) we compare the efficiency of the four methods in terms of how the @xmath41-error of the computed invariant density depends on the number of evaluations of the vector field .
the @xmath41-error was computed using an adaptive lobatto quadrature as implemented in , e.g. , matlab s ` quadl ` command .
these errors are due to the approximation space , the approximate numerical computation of matrix entries , and due to solving the discretized eigenvalue problem . to illustrate how these effects accumulate , in figure [ fig : sineflow2 ] ( right )
we compare the errors of three different approximations , for different ( uniform ) partitions : the approximate invariant densities obtained from the two ulam type methods , and the projection of the true invariant density .
we conclude that the error due to the approximation space dominates the total error .
-error of the approximate invariant density as a function of the number of evaluations of the vector field ( @xmath332 - @xmath333 partition elements , resp .
collocation points , @xmath334 - @xmath335 number of iterations ) .
right : @xmath41-error of the approximate invariant densities obtained by projection of the true invariant density on the underlying partition , by ulam s method , and by ulam s method for the generator.,title="fig:",scaledwidth=50.0% ] -error of the approximate invariant density as a function of the number of evaluations of the vector field ( @xmath332 - @xmath333 partition elements , resp .
collocation points , @xmath334 - @xmath335 number of iterations ) .
right : @xmath41-error of the approximate invariant densities obtained by projection of the true invariant density on the underlying partition , by ulam s method , and by ulam s method for the generator.,title="fig:",scaledwidth=49.0% ] .sources of computational cost .
the dimension of the approximation space is denoted by @xmath264 , @xmath82 denotes the dimension of state space .
in contrast to the other two methods , spectral collocation produces a full matrix .
one can reduce the cost for solving the eigenproblem in this case to @xmath336 by using an iterative solver ( e.g. gmres ) , and fft to compute matrix - vector products . [
cols="<,^,^,^,^ " , ] [ tab : qg - effort2 ] by ulam s method ( left ) with the one computed by ulam s method for the generator ( right , together with the two saddle equilibria and their invariant manifolds).,title="fig:",scaledwidth=42.0% ] by ulam s method ( left ) with the one computed by ulam s method for the generator ( right , together with the two saddle equilibria and their invariant manifolds).,title="fig:",scaledwidth=42.0% ] the regions enclosed by the stable / unstable manifolds of the two saddle equilibria shown in figure [ fig : qg_vector_field ] are obviously invariant for the noiseless flow .
the introduction of noise means that it is possible ( but with a rather low probability ) for stochastic trajectories to jump between these regions , making these regions almost - invariant sets ( cf .
section [ subsec : almost ] ) .
figure [ fig : qg - evs ] shows that these almost - invariant sets are highlighted in the 2nd , 3rd , and 4th eigenvectors as level sets at high gradient regions in the eigenvectors ( in this example , near the yellow bands ) .
the 2nd and 3rd eigenvectors highlight regions near different stable / unstable manifolds at the yellow bands and the 4th eigenvector shown in figure [ fig : qgseporbits ] shows a combination .
we refer the reader to @xcite for details on the connection between invariant manifolds and almost invariant sets .
as one sees in figure [ fig : qgseporbits ] , the eigenvectors from ulam s method for the generator are smoother compared to the ones from the standard ulam method .
ulam s method for the generator appears to be more accurate , at least for this system with a small amount of noise .
a possible explanation is that with ulam s method for the generator we formally average the diffusion , while with the standard ulam method we simulate a finite number of random trajectories , which incompletely sample the true noise distribution . in this section
we demonstrate the efficacy of our generator approach for a three - dimensional flow .
we consider the so - called abc - flow @xcite , given by @xmath337 on the 3 dimensional torus .
the flow is volume - preserving , so in order to make the computation of the spectrum well posed we need to add some artifical diffusion again .
the vector field is very smooth , and in numerical experiments a diffusion coefficient @xmath338 turned out to work well . for @xmath339 , @xmath340 and @xmath341 ,
there is numerical evidence that the flow exhibits complicated dynamics and invariant sets of complicated geometry @xcite .
we approximate the leading few eigenfunctions of @xmath91 with ulam s method for the generator on a @xmath342 grid .
we used gauss quadrature with 16 nodes on each face of a box in order to approximate the surface integrals , requiring @xmath343 evaluations of the vector field .
the @xmath41-error of the approximate invariant density is @xmath344 .
the second and fifth eigenfunctions are shown in figure [ fig : abc_u64_lev2,5 ] .
the regions highlighted in red and blue are invariant cylinders of the deterministic flow ; see @xcite for more details on how to extract invariant and almost - invariant sets from the eigenfunctions .
+ we also apply spectral collocation on a @xmath345 grid , requiring @xmath346 evaluations of the vector field .
the @xmath41-error of the approximate invariant density is @xmath347 .
we observe fast convergence ( after comparing with computations on coarser resolutions ) of not just the first nontrivial eigenfunctions , but also of higher order ones . + as a last example we consider a system where the effective dynamics is supported on a set of complicated geometry and fractional dimension , the well - known lorenz system @xmath348 where we make the standard parameter choices @xmath349 , @xmath350 and @xmath351 .
since the effective ( complicated ) dynamics happens on a lower dimensional attractor we do not expect spectral collocation to work well and we therefore apply ulam s method for the generator here .
a decade ago , numerical techniques were introduced for computing box coverings for attractors of complicated structure , cf .
these techniques make use of the fact that the set @xmath5 to be computed is an attractor , hence each trajectory starting in its vicinity will be pulled to @xmath5 in a fairly short time . in our approach time
is not considered , we only use the vector field . since the boundary of the box covering does not have to coincide with the boundary of the attractor , a tight box covering
might not show the desired results , because of relatively big outflow rates in boundary boxes .
the simplest idea is to use a big enough rectangle in our case @xmath352\times [ -30,30]\times [ -10,70]$ ] .
however , this set is not invariant , hence the question arises , which boundary conditions one should impose .
translating the condition `` no probability flow on the boundary '' in this ulam type approach leads to the condition that the flow rates at the boundary are ignored .
the attractor is then extracted by simple thresholding : regions where the approximate invariant density is nonzero are expected to lie in a small neighbourhood of the attractor .
the finer the resolution , the smaller the diffusion introduced by the discretization and thus the tighter this neighbourhood of the attractor is . having constructed the attractor , we may restrict the other eigenfunctions to this set in order to obtain almost invariant sets within the attractor . for our computation , we employed a grid of @xmath353 boxes and used gauss quadrature in @xmath354 nodes on the faces in order to set up the approximate generator .
the attractor is extracted by thresholding the approximate invariant density @xmath355 : we cut off at a threshold value @xmath356 such that 96% of the invariant measure is supported on @xmath357 .
figure [ fig : lorattr128 ] shows the corresponding covering as well as the sign structure of the second and third eigenvector .
we introduced the infinitesimal generator as a tool for studying long term behavior of flows without requiring any trajectory integration .
the ( typically unit dimensional ) null space of the generator is spanned by the invariant density of the flow ; the invariant density describes the asymptotic distribution of trajectories .
we showed that eigenfunctions of the generator corresponding to eigenvalues close to 0 had strong connections with almost - invariant or metastable behavior , and could be used to identify regions in the flow from which the escape rate is very low .
we proposed two new numerical approaches for the estimation of the generator ; one based on ulam s method and the other based on spectral collocation .
we tested our numerical approaches on `` full phase space '' flows in one , two , and three dimensions and found that both approaches significantly outperformed the standard ulam approach based on the perron - frobenius operator .
moreover , all operator / generator based methods were superior to standard trajectory integration for estimating the invariant density ; the identification of almost - invariant sets is essentially impossible without an operator / generator approach .
the spectral collocation approach worked particularly well in periodic domains when the vector field is infinitely smooth .
we thank yuri latushkin for pointing us to useful books , and nick trefethen for careful proofreading and helpful questions and suggestions . ,
_ the algorithms behind set oriented numerical methods for dynamical systems _ , in ergodic theory , analysis , and efficient simulation of dynamical systems , bernold fiedler , ed . , springer , berlin , 2001 , pp . 145174 , 805807 . | the long - term distributions of trajectories of a flow are described by invariant densities , i.e. fixed points of an associated transfer operator . in addition
, global slowly mixing structures , such as almost - invariant sets , which partition phase space into regions that are almost dynamically disconnected , can also be identified by certain eigenfunctions of this operator .
indeed , these structures are often hard to obtain by brute - force trajectory - based analyses . in a wide variety of applications ,
transfer operators have proven to be very efficient tools for an analysis of the global behavior of a dynamical system .
the computationally most expensive step in the construction of an approximate transfer operator is the numerical integration of many short term trajectories . in this paper , we propose to directly work with the _ infinitesimal generator _ instead of the operator , completely avoiding trajectory integration .
we propose two different discretization schemes ; a cell based discretization and a spectral collocation approach .
convergence can be shown in certain circumstances .
we demonstrate numerically that our approach is much more efficient than the operator approach , sometimes by several orders of magnitude . |
the topic of control over a communication channel has been extensively studied in the past decade , with issues such as the minimum data rate for stabilization @xcite and optimal quadratic closed - loop performance @xcite being the main focus .
other issues of interest concern effects of channel - induced packet drops and/or time - varying delays on closed - loop performance . the majority of papers concerned with control over networks regards the mechanism of information loss in the network as probabilistic but not strategic . in contrast , in the problem of control over an _ adversarial _ channel , the communication link is controlled by a rogue jammer whose intention is to mount a cyber attack on the system by actively jamming the communication link .
its objectives are to impose on the controller a control law which can not be expected under regular operating conditions in a packet - dropping network
. if the controller is unaware of the jammer s actions and continues to follow a control policy designed for a regular network , the system performance is likely to be inferior .
it is this situation that is considered as a scenario of a successful cyber attack by the jammer . a natural way to describe
the problem of control over an adversarial channel is to employ a game - theoretic formulation .
originally proposed in @xcite , this idea has been followed upon in a number of recent papers including @xcite . a zero - sum dynamic game between a controller performing a finite horizon linear - quadratic control task and a jammer , proposed in @xcite , specifically accounted for the jammer s strategic intentions and limited actuation capabilities , but was otherwise agnostic regarding the type of channel involved .
a startling conclusion of @xcite was that in order to maximally disrupt the control task , the jammer must act in a markedly different way than a legitimate , non - malicious , packet - dropping channel .
once this deterministic behavior is observed by the controller , it can establish with certainty that an attack has taken place . in @xcite
, we introduced a different model , which , while capturing the same fundamental aspects of the problem as in @xcite , modified the jammer s action space so that each jammer decision corresponded to a choice of channel rather than to passing / blocking transmission .
the corresponding one - step zero - sum game was shown to have a unique saddle point in the space of mixed jammer strategies . in turn , the controller s best response to the jammer s optimal randomized strategy was to act as if it was operating over a packet - dropping channel whose statistical characteristics were controlled by the jammer . since under normal circumstances
the controller can not be aware of these characteristics , and can not implement such a best response strategy , we regard the zero - sum game in @xcite as an example of a successful cyber attack . in this paper , we show that such a situation is not specific to the zero - sum game considered in @xcite .
we introduce a class of zero - sum stochastic games that generalize the model introduced in @xcite . for these games
we obtain necessary and sufficient conditions which guarantee the existence of optimal jammer s strategies whose nature suggests that the jammer must select its actions randomly , in order to make a maximum impact on the control performance .
our conditions are quite general , they apply to nonlinear systems and draw on standard convexity / coercivity properties of payoff functions .
furthermore , we specialize these conditions to the linear - quadratic control problem over a packet - dropping link considered in @xcite and show that our conditions allow for an express characterization of a set of plant s initial states for which optimal randomized jammer strategies exist ( this is in contrast to @xcite where a complete analysis of the state space had to be performed to determine such regions ) .
we also compute an optimal controller response to those strategies , which turns out to be nonlinear .
our analysis is restricted to one - step zero - sum games .
although such a formulation is admittedly simple , due to the general nature of the game under consideration , it can be thought of as reflecting a more general situation where one is dealing with a one - step hamilton - jacobi - bellman - isaacs min - max problem associated with a multi - step optimal control problem .
also , even a one - step formulation provides a rich insight into a possible scenario of cyber attacks on controller networks .
we believe that such an insight can be valuable as was the case , e.g. , in early studies of adversarial channels and multi - agent decision problems involving incomplete information .
we present our model in section [ sec_mod ] . the problem formulation ,
its assumptions and preliminary results are given in section [ sec_form ] .
the main result of the paper that gives a necessary and sufficient condition for the game under consideration to have a nontrivial saddle point is presented in section [ main ] .
next , in section [ examples ] we demonstrate an application of this result to the linear - quadratic static problem which is an extension of the problem in @xcite . in this problem ,
the jammer is offered an additional reward for undertaking actions concealing its presence .
conclusions are given in section [ conclusions ] .
the general model description is an extension of that in @xcite .
we consider a situation where a strategic jammer is attacking the link in the feedback loop between a controller and a plant .
the plant is a general discrete - time system described by @xmath0 with a given initial condition ; @xmath1 is the state , @xmath2 is a scalar input , @xmath3 is an @xmath4-valued function defined on @xmath5 .
the plant input @xmath2 and the control signal @xmath6 are related by the equation @xmath7 which describes the transmission of information from the controller to the plant over a packet - dropping communication link . here
, @xmath8 is a discrete random variable taking value in @xmath9 , that describes the transmission state of the link .
the value of @xmath8 depends on actions of the jammer and the state of the communication link , as explained below .
the communication link consists of a finite set @xmath10 of channels ( with @xmath11 ) out of which the jammer can draw with certain probability a channel to replace the currently active channel so as to optimally disrupt the control task .
each channel can be either in passing or blocking state , and the transmission states of _ all _ channels randomly change _ after _ one of them is selected as a replacement .
hence , each channel @xmath12 represents a binary channel with the state space @xmath9 , as pictured in figure [ fig_2 ] . to describe the probability model of channel transitions ,
let @xmath13 denote the vectors of transmission states of _ all _ channels before and after the jammer has selected one of them to replace the current one , respectively , with the @xmath14th component @xmath15 , @xmath16 denoting the corresponding transmission state of channel @xmath17 .
the probability of channel @xmath17 to become `` passing '' after the replacement is selected , given its and all other channels previous transmission states , is then @xmath18 the jammer strategy is to choose a probability distribution over @xmath10 , indicating which channel it desires to switch to .
we denote this distribution by a vector @xmath19 in the unit simplex @xmath20 of @xmath21 .
that is , the jammer s strategy is to influence the selection of a channel linking the controller to the plant .
let the index of the selected channel be @xmath22 , then @xmath22 is a discrete random variable taking values in @xmath23 , distributed in accordance with the vector @xmath19 .
the latter depends on the information set available to the jammer which includes the current state of the plant @xmath24 , the index of the channel occupying the link @xmath25 , and the vector of transmission states of all channels @xmath26 which jammer observes before the link switches from channel @xmath27 to a new channel : @xmath28 in addition , if the control input @xmath6 is available to the jammer , the vector @xmath19 may depend on @xmath6 as well .
after the jammer has made its decision , the random variable @xmath22 is realized and the link switches to the channel @xmath29 . after that , the transmission state of all channels including @xmath29 changes , according to ( [ q_j ] ) .
thus , the jammer can not predict the transmission state of the channels in @xmath10 when selecting the probability vector @xmath19 . in accordance with this channel switching mechanism
, the transmission state of the link between the controller and the plant is determined by the binary random variable @xmath30 , i.e. , @xmath31 , which takes value 1 with probability @xmath32 and value 0 with probability @xmath33 .
clearly , @xmath8 and @xmath22 are statistically dependent .
all random variables considered in this paper will be adapted to the joint conditional probability distribution of @xmath22 and @xmath8 , given @xmath24 , @xmath25 and @xmath26 .
the expectation with respect to this conditional probability distribution is denoted @xmath34 $ ] .
1.5ex is a binary channel .
shown here is channel @xmath17 .
[ fig_2 ] ]
we now introduce a general two - player stochastic one - step zero - sum game as follows . in this game , we assume that the initial state of the plant @xmath24 , the initial vector of transmission states @xmath26 and the channel that initially occupies the link @xmath27 are known to both the jammer and controller . let @xmath35 be a scalar function of @xmath36 .
this function will determine the payoff of the game played by the controller and the jammer .
the standing assumptions regarding this function are summarized below : [ sigma.a1 ] for all @xmath12 , @xmath37 .
[ sigma.a3 ] for each @xmath12 and @xmath38 , the functions @xmath39 and @xmath40 are coercive . [ u ] under assumptions [ sigma.a1 ] and [ sigma.a3 ] ,
for every @xmath41 there exists a compact set @xmath42 with the properties : a. for all @xmath43 , @xmath44 } & & \nonumber \\ & & = \inf_{u\in u(x)}\max_{j } \mathbb{e}[\sigma(x^+,u , f_s)|s = j ) ] .
\label{max.property}\end{aligned}\ ] ] b. @xmath45=\inf_{u\in u(x)}\mathbb{e}[\sigma(x^+,u , f_s)].\ ] ] the proof is omitted for the sake of brevity .
it proceeds by first proving that the coercivity of the functions involved ensures that the infima on the left hand side of ( [ max.property ] ) and ( [ infu.1 ] ) exist .
in particular , @xmath46 , @xmath47 , @xmath48 $ ] .
next we show that a suitably defined set @xmath49 , with a sufficiently large @xmath50 can be chosen as @xmath51 .
we now define the stochastic zero - sum min - max game of interest for the plant ( [ plant ] ) . in this game ,
the controller is a minimizing player who selects the control input @xmath52 based on @xmath24 , @xmath25 and @xmath26 . also , the jammer is the maximizing player who chooses a probability distribution vector @xmath53 for the ` channel selection ' random variable @xmath22 , as the function of @xmath24 , @xmath25 , @xmath26 and possibly @xmath6 , as in ( [ p_j ] ) .
the controller s best action is determined by computing @xmath54 .
\label{inf_sup } \end{aligned}\ ] ] while the jammer s best action is obtained by computing @xmath55 .
\label{sup_inf } \end{aligned}\ ] ] our goal is to show that @xmath56 , i.e. , that the corresponding zero - sum game has a value .
lemma [ u ] allows to reduce the minimization over @xmath57 in ( [ inf_sup ] ) and ( [ sup_inf ] ) to minimization over a compact set @xmath51 .
indeed , the cost function of the inner maximization problem ( [ inf_sup ] ) is linear in @xmath19 , therefore using claim ( i ) of lemma [ u ] leads to the conclusion that @xmath58 \nonumber \\ & = & \inf_{u\in u(x)}\max_j \mathbb{e } \left[\sigma(x^+,u , f_j)\right ] \label{inf_u_max } \\ & = & \inf_{u\in u(x)}\max_{p \in \mathcal{s}_{n-1 } } \mathbb{e } \left[\sigma(x^+,u , f_s)\right ] .
\label{inf_u_sup}\end{aligned}\ ] ] also , it follows from claim ( ii ) of lemma [ u ] that for every @xmath19 , the inner minimization problem in ( [ sup_inf ] ) can be carried out over @xmath51 .
thus , @xmath59 . \label{sup_inf_u}\end{aligned}\ ] ] we make an additional assumption about the set @xmath51 .
[ sigma.a4 ] the set @xmath51 is connected . under this assumption ,
the set @xmath51 is a closed bounded interval , hence it is a convex set .
of course , this can be guaranteed when @xmath60 is chosen so that each @xmath61 is convex .
[ bo.t4 ] under assumptions [ sigma.a1 ] , [ sigma.a3 ] , and [ sigma.a4 ] , the value of the game ( [ inf_sup ] ) exists , i.e. , @xmath62 furthermore , the game has a ( possibly non - unique ) saddle point .
it is not unreasonable to assume that in the game ( [ inf_sup ] ) the jammer , who observes the controller action @xmath6 , can rank all the channels according to the contribution they make towards the payoff and order them accordingly .
it can do so by comparing the conditional expected cost values @xmath63 $ ] .
[ sigma.a2 ] for any two channels @xmath64 , @xmath65 , if @xmath66 then @xmath67\ge \mathbb{e}[\sigma(x^+,u , f_ss)|s = k ] \\
\nonumber \forall u\in u(x ) .
\end{aligned}\ ] ] assumption [ sigma.a2 ] generalizes the situation considered in @xcite where all channels were ranked according to the probability of becoming passing , @xmath68 . in section [ examples ]
we will show that such a natural ranking leads to satisfaction of assumption [ sigma.a2 ] . according to this assumption , the jammer who seeks a higher value of payoff should favour channels with lower numbers , since a larger reward is associated with these channels .
in contrast , the controller actions should be directed towards forcing the jammer into utilizing channels with higher numbers .
also , the channel @xmath27 is excluded from this ranking .
this is done to allow the jammer to consider contributions to payoff other than those based on blocking / passing .
these considerations may either discourage the jammer from switching , or conversely encourage it to undertake a denial - of - service attack .
such decisions can be influenced by a number of factors that are not related to channel properties .
the cost of channel switching is one reason as to why the jammer may decide not to change the channel . under another scenario , the jammer
may be offered a reward for remaining stealthy , and may choose this reward over disrupting the control loop .
for instance , when the controller monitors the link , an anomaly in the channel transition probabilities could signal the attack . in this case , rewarding the jammer for not defaulting to the most blocking channel unless it is absolutely necessary will provide it with an incentive for not revealing itself . in yet another class of problems , the jammer s decision could be based on the knowledge that the system is prepared to tolerate service disruptions as long as the cost of such disruptions is below the cost of rectifying them .
we defer detailed analyses of these situations to section [ examples ] .
it should be stressed that jammer decisions in each of these scenarios will depend on the plant state @xmath24 , the channel @xmath27 and the channel ranking ( the latter may require knowledge of @xmath6 ) . using the channel ranking introduced in assumption [ sigma.a2 ]
, the value and saddle points of the game ( [ inf_sup ] ) can be characterized by solving a game over a reduced jammer strategy space .
this reduced game focuses on two channels , namely the channel that currently occupies the link and the channel that delivers the highest payoff to the jammer when it seeks to block communications between the controller and the plant .
the latter channel is indexed as channel @xmath69 , by assumption [ sigma.a2 ] .
let us introduce the reduced jammer action vector @xmath70 , @xmath71 , @xmath72 .
also , consider payoffs associated with selecting channel @xmath69 and keeping the current channel @xmath27 : @xmath73 , \\
\tilde h_2(u)&= & \mathbb{e}\left[\sigma(x^+,u , f_s)|s = j^-\right ] , \end{aligned}\ ] ] and define @xmath74 . consider the following ` reduced ' two - player game with upper value @xmath75 and lower value @xmath76 [ only.two ] suppose assumptions [ sigma.a1]-[sigma.a2 ] are satisfied .
then @xmath77 furthermore , the zero - sum game ( [ inf_sup.red ] ) has a ( possibly non - unique ) saddle point .
also , if @xmath78 is such a saddle point , then @xmath79 is a saddle point of the game ( [ inf_sup ] ) , where @xmath80
lemma [ only.two ] allows the jammer to constrain its actions to the set @xmath81 . among these actions
there are two trivial actions : choose the most blocking channel ( channel @xmath69 in our notation ) by using @xmath82 and @xmath83 , or stay put by allocating @xmath84 and @xmath85 , so that the controller continues communicating with the plant over channel @xmath27 .
however , the question arises as to whether there exist _ optimal mixed strategies _ in @xmath86 for the jammer to undertake , i.e. , optimal policy vectors @xmath19 such that @xmath87 , @xmath88 .
the first main result of this paper provides necessary and sufficient conditions for the existence of nontrivial saddle points .
these conditions characterize the controller - jammer games in which the jammer randomizes its choice of optimal strategies .
as we will see , this will force the controller to respond in a non - obvious manner in order to remain optimal , which ultimately represents a signature of an attack on the communication link .
[ saddle.point ] suppose @xmath89 , @xmath90 are strictly convex functions of @xmath6 for all @xmath24 . for every @xmath24 ,
the zero - sum game ( [ inf_sup.red ] ) admits a nontrivial saddle point @xmath78 if and only if there exists @xmath91 such that @xmath92 and one of the following conditions hold : either @xmath93 or @xmath94
in this section , we specialize theorem [ saddle.point ] to the controller - jammer game where the plant is linear , @xmath95 and the performance cost is quadratic . in this game the jammer is rewarded for remaining stealthy .
we show that in this game , there is a region in the plant state space where the jammer s optimal policy is to randomize its channel selection .
furthermore , an optimal control response to this optimal jammer action is nonlinear .
consider a controller - jammer game for the plant ( [ plant.lin ] ) with the quadratic payoff @xmath96 here , @xmath97 is the kronecker symbol , and @xmath98 is the constant ` reward for stealthiness ' which the jammer receives if the channel does not change as a result of its action . as explained earlier , the rationale here is to reward the jammer for keeping the current channel in the link when excessive switching may reveal its presence , or may drain its resources .
this controller - jammer game was analyzed in @xcite ( for a one - dimensional plant ) , where the region in the state - space was found where the game has a unique saddle point corresponding to a jammer s nontrivial strategy .
such a region was found by computing the game value directly , which required a quite tedious analysis . here
, we revisit this result of @xcite from a more general perspective , using conditions of theorem [ saddle.point ] .
the corresponding function @xmath60 in this case is @xmath99 clearly , the function @xmath60 defined in ( [ lq.cost ] ) satisfies assumptions [ sigma.a1 ] and [ sigma.a3 ] . also in this case , the functions @xmath61 have the form @xmath100 \nonumber \\ & = & \gamma_j(x)+u^2+r_jq_ju(u+2\beta(x ) ) , \label{hj.lau3}\end{aligned}\ ] ] where @xmath101 @xmath102 , and @xmath103 for all @xmath14 .
also , the available channels are assumed to be ordered according to their probability to become passing , that is , @xmath104 [ u.lq ] under condition ( [ qq ] ) , the set @xmath105 verifies properties ( i ) and ( ii ) stated in lemma [ u ] , and also satisfies assumptions [ sigma.a4 ] and [ sigma.a2 ] . under the above assumptions , the payoff functions @xmath89 and @xmath90 for the reduced zero - sum game become @xmath106 with these definitions , condition ( [ f = f ] )
reduces to the equation @xmath107 which admits real solutions if @xmath108 .
also , condition ( [ df.df<0 ] ) reduces to the condition @xmath109 let @xmath110 the analysis of conditions ( [ f = f.lq ] ) , ( [ df.df<0.lq ] ) shows that only one of the solutions of equation ( [ f = f.lq ] ) , @xmath111 satisfies ( [ df.df<0.lq ] ) provided @xmath112 condition ( [ z.lau3a ] ) describes the region in the state space in which the jammer s optimal policy is to choose randomly between the channel @xmath27 currently in use and the most blocking channel @xmath69 .
observe that in the case where the plant ( [ plant.lin ] ) is scalar and @xmath113 , we recover the exactly same condition as that obtained in @xcite by direct computation .
that is , theorem [ saddle.point ] confirms the existence of the nontrivial optimal jammer s strategy for this region .
we refer the reader to @xcite for the exact value of the optimal vector @xmath114 ; the calculation for the multidimensional plant ( [ plant.lin ] ) follows the same lines , and is omitted for the sake of brevity .
we also point out that the optimal controller s policy ( [ u * ] ) is nonlinear .
hence , any linear feedback policy that controller may employ assuming that its signals are transmitted over a _ bona fide _ packet dropping channel will lead to an inferior control performance .
we interpret this situation as a signature of a successful dos attack by the jammer .
in this paper we have analyzed a class of control problems over adversarial channels , in which the jammer actively attempts to disrupt communications between the controller and the plant .
we have posed the problem as a static game , and have given necessary and sufficient conditions for such a game to have a nontrivial saddle point .
the significance of these conditions is to allow a characterization of a set of plant s initial states for which a dos attack can be mounted that requires a nontrivial controller s response . for instance , in the linear quadratic problem analyzed in the paper the optimal control law is nonlinear .
this gives the jammer an advantage over any linear control policy in those problems .
the jammer achieves this outcome by randomizing its choice of a packet - dropping channel rather than operating packet dropping facility directly . on the other hand ,
the part of the state space where the jammer randomizes is determined by the jammer s cost of switching ( reward for not switching ) and transition probabilities of the current and the most blocking channels .
if these parameters can be predicted / estimated by the controller , it has a chance of mitigating the attack by either eliminating those regions , or steering the plant so that it avoids visiting those regions .
future work will be directed to further understanding conditions for dos attacks , with the aim to consider dynamic / multi - step control problems .
another interesting question is whether associating a distinct payoff with one of the channels is necessary for the jammer to resort to randomization .
e. garone , b. sinopoli , and a. casavola , lqg control over lossy tcp - like networks with probabilistic packet acknowledgements , international journal of systems , control and communications , vol .
55 - 81 , 2010 .
a. gupta , c. langbort , and t. baar , optimal control in the presence of an intelligent jammer with limited actions , in proc . of 49th ieee conference on decision and control ( cdc ) ,
1096 - 1101 , december 2010 .
t. baar and y - w .
wuh , a complete characterization of minimax and maximin encoder - decoder policies for communication channels with incomplete statistical description , ieee transactions on information theory , vol .
31 , no . 4 , 1985 | we revisit a one - step control problem over an adversarial packet - dropping link .
the link is modeled as a set of binary channels controlled by a strategic jammer whose intention is to wage a ` denial of service ' attack on the plant by choosing a most damaging channel - switching strategy .
the paper introduces a class of zero - sum games between the jammer and controller as a scenario for such attack , and derives necessary and sufficient conditions for these games to have a nontrivial saddle - point equilibrium . at this equilibrium ,
the jammer s optimal policy is to randomize in a region of the plant s state space , thus requiring the controller to undertake a nontrivial response which is different from what one would expect in a standard stochastic control problem over a packet dropping channel . |
in regge phenomenology , high mass single diffraction implies a non - zero triple - pomeron coupling , which in turn implies , through the optical theorem , double pomeron exchange , d@xmath8e , : @xmath9 . here
@xmath10 means a proton or antiproton , the final state protons may be quasi - elastic or they may dissociate ( @xmath11 ) , and @xmath12 represents a large rapidity gap @xmath13 with no hadrons .
see ref.@xcite for a review . by
exclusive " we mean that the central state @xmath14 is simple and fully measured . at low masses , in the resonance region @xmath15 3 gev , d@xmath8e is non - perturbative and qcd ( or qcd - inspired ) calculations are challenging ; there are new efforts by the durham @xcite and cracow @xcite groups .
the quantum numbers of @xmath14 are restricted to be mostly @xmath16 , so @xmath6-channel resonances @xmath17 and @xmath18 are allowed .
resonances with a high gluon content will be favored , especially in comparison with @xmath19 . for the @xmath20 and @xmath21 states perturbative calculations of @xmath22 are applicable , related to the very interesting channels @xmath23 @xcite and @xmath24 higgs .
so we have several motivations : improving our understanding of the pomeron , meson ( especially glueball ) spectroscopy , and testing the qcd physics of exclusive production ( especially @xmath25 and higgs ) .
the cdf detector at the fermilab tevatron is well - known .
for this study we used data not only at the usual @xmath1 = 1960 gev , but also at 900 gev in a special run .
we only used bunch crossings with a single interaction , i.e. no pile - up , and we required all the cdf detectors , covering @xmath26 to be empty , except for two oppositely - charged tracks and their corresponding calorimeter hits .
the trigger for these events was @xmath27 calorimeter showers with @xmath28 0.5 gev , with a veto on beam shower counter hits ( @xmath29 ) , cherenkov luminosity counters ( @xmath30 = 3.7 - 4.7 ) and forward calorimeters ( @xmath30 = 1.32 - 3.64 ) .
we had 22 m ( 90 m ) triggers at @xmath1 = 900 ( 1960 ) gev .
off - line we required the central calorimeters ( @xmath31 1.3 ) to be also empty , apart from the trigger clusters .
importantly , we simultaneously recorded a large sample of 0-bias ( bunch crossing ) triggers . dividing these into interaction " and non - interaction " samples , as in ref .
@xcite , allowed us to determine the noise levels in all the detectors , and to measure the total visible cross section @xmath32 , which is the inelastic cross section @xmath33 times the fraction @xmath34 of inelastic events with particles in @xmath35 5.9 , estimated to be [email protected] ( [email protected] ) . at 1960
gev @xmath32 agreed with global fits ; at 900 gev we used @xmath32 to normalize our cross sections , as the luminosity counters were not calibrated .
the total delivered luminosity at the two energies was 0.056 ( 7.12 ) pb@xmath37 .
the effective no - pile - up " luminosity was 0.0435 ( 1.18 ) pb@xmath37 , determined by counting empty 0-bias events as a function of the bunch luminosity .
off - line we required exactly two well - measured opposite - charge tracks with @xmath38 and @xmath39 0.4 gev / c .
the pair @xmath40 ( @xmath41-masses assumed ) was required to have @xmath42 1.0 , and @xmath43 0.8 gev to have acceptance down to @xmath44 .
we calculated the acceptance and efficiencies for the above fiducial region , and with the effective luminosity calculated the differential cross section @xmath45 , assuming an isotropic ( s - wave ) @xmath46 distribution .
for two particles , assumed to be @xmath0 , in the stated kinematic region , between two rapidity gaps @xmath474.6 , at @xmath1 = 1960 gev.,width=472 ] fig .
[ fig : mass_corr_1960 ] shows the differential cross section integrated over @xmath48 as a function of @xmath49 , and fig .
2a shows the low mass region on a linear scale , and at both energies .
a small @xmath4 signal is seen , and a dominant @xmath5 ( also dominant in @xmath50 ) . a possible shoulder on the high mass side ( @xmath51 ? )
is followed by a distinct change of slope at 1500 mev , which was also seen at lower energies @xcite . while the cross section shapes are similar at the two energies , they differ in detail as seen in the ratio plot fig .
in addition to any @xmath6-dependence of the @xmath52 cross section ( expected from regge to be @xmath53ln@xmath54 ) there is more rapidity available for proton dissociation at 1960 gev , the beam rapidities being 6.87 and 7.64 while the detector extends to @xmath55 = 5.9 in both cases .
we observe that the ratio is lower in the region of the @xmath5 than it is below 1 gev , expected to be dominated by s - wave .
we also find that the mean @xmath56 has a minimum in the @xmath5 region , and rises abruptly at 1.5 gev .
we previously observed @xcite exclusive @xmath57 production in the mode @xmath58 , but could not distinguish the three @xmath20 states .
the @xmath0 and @xmath59 channels have larger branching fractions and enough resolution to separate the @xmath60 states .
we do not see significant signals in this data , and give upper limits ( 90% c.l . ) on @xmath61 = [email protected](syst.)nb ( in @xmath0 ) and [email protected](syst.)nb ( in @xmath59 ) .
this implies that @xmath62% of the @xmath63 events were @xmath64 .
even though the @xmath18 may have a much smaller production cross section its branching fraction is 17@xmath65 larger .
we studied the cos @xmath66 distributions of the @xmath67 in the @xmath14-frame relative to the incoming @xmath10-direction .
the data are consistent with isotropy up to 1.5 gev , above which they become progressively more forward - backward peaked .
isotropy is expected if any polarization at production is washed out after integration over the unseen protons or @xmath68-dissociations .
the durham " and cracow " groups @xcite have predicted the differential cross section with the same cuts as fig .
[ fig : mass_corr_1960 ] , but with no dissociation .
theoretical uncertainties in the region @xmath69 gev are about @xmath70 , but the data are within these uncertainties .
l.a.harland-lang , v.a.khoze , m.g.ryskin , and w.j.stirling , phys . lett . *
b725 * 316 ( 2013 ) .
p.lebiedowicz , r.pasechnik , and a.szczurek , nucl.phys.proc.suppl . * 219 - 220 * 284 ( 2011 ) .
( cdf collaboration ) , phys .
lett . * 108 * , 081801 ( 2012 ) .
et al . , _ ( cdf collaboration ) , phys .
102 * , 242001 ( 2009 ) .
see e.g. t.kesson _ et al . _
( afs collaboration ) , nucl.phys .
* b264 * , 154 ( 1986 ) ; a.breakstone _ et al . _
( abcdhw collaboration ) , z.phys * c31 * 185 ( 1986 ) . | using the collider detector at fermilab , cdf , we have measured exclusive @xmath0 production at @xmath1 = 900 gev and 1960 gev .
the @xmath0-pair is central , @xmath2 1.0 , and there are no other particles detected in @xmath3 .
we discuss the mass spectrum , showing @xmath4 and @xmath5 resonances , @xmath6-dependence , @xmath7-dependence , and angular distributions . |
the main goal of this paper is to exhibit a new family of simply connected minimal symplectic and infinitely many non - symplectic @xmath1-manifolds that is homeomorphic but not diffemorphic to @xmath12 for any @xmath13 and @xmath9 .
our construction is a generalization of the first author s work in @xcite , where he used yukio matsumoto s genus two lefschetz fibrations on @xmath3 over @xmath4 , arising from the elliptic involution of the genus two surface with two fixed points , along with the fake symplectic @xmath5 construction in @xcite obtained via knot surgery along the fibered knots . in the aforementioned paper @xcite , using the symplectic connected sum along the genus two surfaces , the first author has constructed the first known exotic minimal symplectic @xmath0 and minimal symplectic @xmath1-manifold that is homeomorphic but not diffeomorphic to @xmath2
. a straightforward generalization of these symplectic examples to an exotic @xmath14 ( for @xmath15 ) , using matsumoto s genus two fibration , can be found in @xcite , where the geography problem for simply connected minimal symplectic @xmath1-manifolds with signature less than equal @xmath16 studied in details . considering that mustafa kormaz s and yusuf gurtas higher genus lefschetz fibrations on @xmath6 , arising from the involution of the genus @xmath17 surface , are the generalizations of matsumoto s genus two fibration , it was a natural question whether the similar construction can be carried out using these higher genus lefschetz fibrations .
the fundamental group of these lefschetz fibrations are not abelian if @xmath7 and the relations in the fundamental group coming from the vanishing cycles are far more complicated , thus the fundamental group computations become more difficult and delicate than in @xcite . in this article , by carefully analysing the fundamental group ,
we overcome these difficulties and prove the following theorem : [ thm : main ] let @xmath18 be @xmath12 for any @xmath13 and @xmath9 . there exists a new family of smooth closed simply - connected minimal symplectic @xmath1-manifold and an infinite family of non - symplectic @xmath1-manifolds that are homeomorphic but not diffeomorphic to @xmath18 .
one difference , though not essential , with the construction in @xcite ( see also @xcite ) is that we use the luttinger surgery instead of the knot surgery .
the luttinger surgery will have some advantages for us in the computation of the fundamental groups , since the presentation of the fundamental groups look much simpler via luttinger surgery .
this is due to the fact that the relations introduced by our luttinger surgeries all involves a single commutator relation expression , while the fundamental groups via knot surgery is more challenging to work with .
our construction can be also interpreted using the knot surgery .
the second goal of our paper is to construct the symplectic @xmath1-manifolds with various fundamental groups ( and with a small size ) by applying the luttinger surgeries to a certain family of lefschetz fibrations .
more general and stronger results along these lines are obtained in @xcite and @xcite .
let @xmath19 be an orientable @xmath20-dimensional closed and connected surface of genus @xmath21 .
let @xmath22 denote the group of all orientation - preserving diffeomorphisms @xmath23 and @xmath24 be the subgroup of @xmath25 consisting of all orientation - preserving diffeomorphisms @xmath26 that are isotopic to the identity . _
the mapping class group _ @xmath27 of @xmath19 is defined to be the group of isotopy classes of orientation - preserving diffeomorphisms of @xmath19 , i.e. , @xmath28 let @xmath29 be a simple closed curve on @xmath19 .
right handed dehn twist _ @xmath30 about @xmath29 is the isotopy class of a self - diffeomophism of @xmath19 obtained by cutting the surface @xmath19 along @xmath29 and gluing the ends back after rotating one of the ends @xmath31 to the right .
let @xmath32 denote be the set of all isotopy classes of simple closed curves in @xmath19 .
the _ intersection number _ of two classes @xmath33 $ ] and @xmath34 $ ] is @xmath35 , [ \beta ] ) = min \
{ |a \cap b| | a \in [ \alpha ] , b \in [ \beta ] \}$ ] .
let us mention a few basic facts about dehn twists .
first , notice that the conjugate of a dehn twist is again a dehn twist .
indeed , if @xmath36 is an orientation - preserving diffeomorphism , then we can easily see that @xmath37 .
second , if @xmath29 and @xmath38 are two simple closed curves with @xmath35 , [ \beta ] ) = k \geq 0 $ ] , then @xmath39 . in section [ sbb ] , we will use some relations that hold in the mapping class group @xmath27 . in order to provide a right framework for our discussion in section [ sbb ] ,
we briefly review the theory of lefschetz fibrations on @xmath1-manifolds below .
[ defn : lef.fibration ] let @xmath40 be a compact , connected , oriented , smooth @xmath1-manifold .
lefschetz fibration _ on @xmath40 is a smooth map @xmath41 , where @xmath42 is a compact , oriented , smooth @xmath20-manifold of genus @xmath43 , such that @xmath44 is surjective and each critical point of @xmath44 has an orientation preserving chart on which @xmath45 is given by @xmath46 .
it is a consequence of the sard s theorem that @xmath44 is a smooth fiber bundle away from finitely many critical points .
let us denote the critical points of @xmath44 by @xmath47 .
the genus of the regular fiber of @xmath44 is defined to be the genus of the lefschetz fibration . if a fiber of @xmath44 passes through critical point set @xmath48
, then it is called a singular fiber which is an immersed surface with a single transverse self - intersection .
a singular fiber of the genus @xmath49 lefschetz fibration can be described by its monodromy , i.e. , an element of the mapping class group @xmath27 .
this element is a right - handed ( or a positive ) dehn twist along a simple closed curve on @xmath50 , called the _ vanishing cycle_. if this curve is a nonseparating curve , then the singular fiber is called _ nonseparating _ , otherwise it is called _ separating_. for a genus @xmath49 lefschetz fibration over @xmath51 , the product of right handed dehn twists @xmath52 along the vanishing cycles @xmath53 , for @xmath54 , determines the global monodromy of the lefschetz fibration , the relation @xmath55 in @xmath27 .
conversely , such a relation in @xmath27 determines a genus @xmath49 lefschetz fibration over @xmath51 with the vanishing cycles @xmath56 .
[ lemma : pi1 ] @xmath57@xcite@xmath58 let @xmath59 be a genus @xmath49 lefschetz fibration with global monodromy given by the relation @xmath55 .
suppose that @xmath44 has a section .
then the fundamental group of @xmath40 is isomorphic to the fundamental group of @xmath50 divided out by the normal closure of the simple closed curves @xmath60 , considered as elements in @xmath61 .
in particular , there is an epimorphism @xmath62 the following example is helpful for illustrating the discussion in the proof of our main theorem .
let @xmath63 , @xmath64 , .... , @xmath65 , @xmath66 denote the collection of simple closed curves given in figure [ fig : hyper ] , and @xmath67 denote the right handed dehn twists @xmath68 along the curve @xmath53 .
it is well - known that the following relations hold in the mapping class group @xmath69 : @xmath70 lefschetz fibration on @xmath71 , ] the monodromy relation given above , corresponding to the genus @xmath49 lefschetz fibration over @xmath51 , has total space @xmath71 , the complex projective plane blown up at @xmath72 points .
furthermore , it is well known that for @xmath73 , the above fibration on @xmath74 admits @xmath75 disjoint @xmath76-sphere sections ( see @xcite for a proof of this fact using a mapping class group argument or @xcite for a geometric argument ) .
let @xmath77 and @xmath78 are closed , oriented , smooth @xmath1-manifolds , and @xmath79 are @xmath20-dimensional , smooth , closed , connected submanifolds in them .
suppose that @xmath80^{2 } + [ f_{2}]^{2 } = 0 $ ] and the genera of @xmath81 and @xmath82 are equal .
we choose an orientation - preserving diffemorphism @xmath83 and lift it to an orientation - reversing diffemorphism @xmath84 between the boundaries of the tubular neighborhoods of @xmath85 . using @xmath86 , we glue @xmath87 and @xmath88 along the boundary .
this new oriented smooth @xmath1-manifold @xmath89 is called the _ connected sum _ of @xmath77 and @xmath78 along @xmath81 and @xmath82 , determined by @xmath86 .
let @xmath77 and @xmath78 be smooth @xmath1-manifolds as above .
then @xmath90 where @xmath49 is the genus of the surface @xmath91 .
the proofs of the above formulas simply are an easy consequence of the formulas @xmath92 since @xmath93 and @xmath94 .
if @xmath77 , @xmath78 are symplectic manifolds and @xmath81 , @xmath82 are symplectic submanifolds then according to theorem of gompf @xcite @xmath89 admits a symplectic structure .
we will use the following theorem of m. usher @xcite to show that the symplectic manifolds constructed in sections 4 and 5 are minimal .
@xcite * ( minimality of sympletic sums ) * let @xmath95 be sympletic fiber sum of manifolds @xmath77 and @xmath78 . then : \(i ) if either @xmath96 or @xmath97 contains an embedded sympletic sphere of square @xmath98 , then @xmath99 is not minimal .
\(ii ) if one of the summands @xmath100 ( say @xmath77 ) admits the structure of an @xmath101-bundle over a surface of genus @xmath49 such that @xmath102 is a section of this fiber bundle , then @xmath99 is minimal if and only if @xmath78 is minimal .
\(iii ) in all other cases , @xmath99 is minimal .
@xmath103 in this subsection , we will briefly review a luttinger surgery .
for the details , we refer the reader to @xcite and @xcite .
luttinger surgery has been very effective tool recently for constructing exotic smooth structures on @xmath1-manifolds .
let @xmath40 be a symplectic @xmath1-manifold with a symplectic form @xmath104 , and the torus @xmath105 be a lagrangian submanifold of @xmath40 with self - intersection @xmath106 .
given a simple loop @xmath107 on @xmath105 , let @xmath108 be a simple loop on @xmath109 that is parallel to @xmath107 under the lagrangian framing . for any integer @xmath110 , the @xmath111 _ luttinger surgery _
on @xmath40 will be @xmath112 , the @xmath113 surgery on @xmath105 with respect to @xmath107 under the lagrangian framing . here
@xmath114 denotes a gluing map satisfying @xmath115 ) = m[{\lambda ' } ] + [ \mu_{\lambda}]$ ] in @xmath116 , where @xmath117 is a meridian of @xmath105 .
it is shown in @xcite that @xmath118 possesses a symplectic form that restricts to the original symplectic form @xmath104 on @xmath119 .
the following lemma is easy to verify and the proof is left as an exercise to the reader .
1 . @xmath120 .
@xmath121 and @xmath122 .
in this section , we collect symplectic building blocks that are needed in our construction of exotic @xmath1-manifolds .
our building blocks will be the total space of the lefschetz fibrations over @xmath20-sphere constructed by m. korkmaz @xcite and y. gurtas @xcite , along with the symplectic building blocks that were constructed by the first author in @xcite .
the symplectic @xmath1-manifolds in @xcite were obtained via knot surgery along the fibered knots , but they can be constructed via the sequence of luttinger surgeries . in this paper
we use luttinger surgery , which will have some advantages for us in the computation of the fundamental groups . for the convenience of the reader , we provide necessary
background and state main the results in @xcite .
let us consider the case when @xmath123 .
recall that the four manifold @xmath124 is the total space of the well known genus @xmath49 lefschetz fibration over @xmath4 @xcite .
this was shown by y. matsumoto for @xmath125 , and in the case @xmath7 by m. korkmaz , by factorizing the _ vertical _ involution @xmath126 of the genus @xmath127 surface ( see figure [ fig : inv ] ) .
of the genus @xmath127 surface ] the branched - cover description of the above lefschetz fibrations can be given as follows : take a double branched cover of @xmath128 along the union of two disjoint copies of @xmath129 and two disjoint copies of @xmath130 ( see figure [ fig : rele ] ) .
the resulting branched cover has four singular points , corresponding to the number of the intersections points of the horizontal spheres and the vertical genus @xmath131 surfaces in the branch set .
next step is to desingularize this singular manifold to obtain @xmath132 . observe that a generic fiber of the horizontal fibration is the double cover of @xmath4 , branched over two points .
this gives a sphere fibration on @xmath133 .
also , a generic fiber of the vertical fibration is the double cover of @xmath134 , branched over two points .
thus , a generic fiber is a genus @xmath49 surface . according to @xcite ,
each of the two singular fibers of the vertical fibration can be perturbed into @xmath135 lefschetz type singular fibers . ]
perhaps this is a good place to remark that when @xmath136 , the total space of the corresponding lefschetz fibration is @xmath137 . in this case , the appropriate branched - cover description of the lefschetz fibrations is given as follows : take a double branched cover of @xmath128 along the union of @xmath1 disjoint horizontal copies of @xmath129 and two disjoint vertical copies of @xmath130 .
the resulting branched cover has @xmath138 singular points . by desingularize this manifold ,
we obtain @xmath139 .
a generic fiber of the vertical fibration is the double cover of @xmath134 , branched over @xmath1 points .
thus , a generic fiber has genus @xmath140 .
see subsection [ gfb ] for more details , where this case occurs if we set @xmath141 and @xmath9 .
in fact , the following theorem was proved in @xcite , which computes the global monodromy of the given lefschetz fibration for both an even and an odd @xmath49 .
[ thm : theta ] let @xmath126 denote the vertical involution of the genus @xmath49 surface with @xmath20 fixed points . in the mapping class group @xmath27 ,
the follwing relations between right dehn twists hold : a ) @xmath142 if @xmath49 is even , b ) @xmath143 if @xmath49 is odd . where @xmath144 , @xmath145 , @xmath146 , @xmath147 are the simple closed curves defined as in figure [ fig : relse ] .
let us denote by @xmath148 a regular fiber as the given lefschetz fibration and the standard generators of fundamental group of @xmath148 under the inclusion as @xmath149 and @xmath150 . using the homotopy exact sequence for a lefschetz fibration
, we have @xmath151 according to @xcite , we have the following identification of the fundamental group of @xmath152 :
@xmath153 it follows that the fundamental group of @xmath152 has a presentation with the generators @xmath154 , @xmath155 , @xmath156 , @xmath157 , @xmath158 , @xmath159 , @xmath160 and the relations @xmath161[a_{2 } , b_{2 } ] \cdots [ a_{g } , b_{g}]=1 $ ] , @xmath162 .
it was shown in @xcite the following identities hold @xmath163 @xmath164 , @xmath165 @xmath166 @xmath167 @xmath168 @xmath169[a_{2 } , b_{2}]\cdots [ a_{r } , b_{r}]$ ] now , we prove a lemma which we use in the sequel .
[ sixl ] the following relations hold in the fundamental group of @xmath152 @xmath170,\
\ \cdots , \\
\nonumber b_{i+1}b_{i+2 } \cdots
b_{g - i } = [ a_{2k - i+1},b_{2k - i+1 } ] \cdots [ a_{2k-1},b_{2k-1}][a_{2k},b_{2k}].\\ \nonumber\end{gathered}\ ] ] using the relations @xmath171 , @xmath172 , and @xmath173 in the fundamental group of @xmath152 , we easily see that @xmath174 .
next , using the relations @xmath175
= @xmath176 , @xmath177 = @xmath178 , and @xmath179 , we obtain @xmath180 = @xmath181 = @xmath182 . by continuing in this fashion ( i.e. using the relations @xmath183 = @xmath184 , @xmath185 = @xmath186 = @xmath182 , and @xmath187 ) , we have @xmath188 for any @xmath189 between @xmath182 and @xmath131 .
furthemore , by considering the relations @xmath190
= @xmath191 , @xmath192^{-1}$ ] , and @xmath179 , we have @xmath182 = @xmath175 = @xmath193 = @xmath194 = @xmath195 = @xmath182 . consequently , using the relations @xmath196
= @xmath197 , and @xmath198 , we have @xmath199 \cdots [ a_{g-1},b_{g-1}][a_g , b_g]$ ] for any @xmath189 between @xmath182 and @xmath131 . using the lemma above
, we see that the fundamental group of @xmath133 is isomorphic to the surface group @xmath200 , generated by loops @xmath201 ,
@xmath155 , @xmath158 , @xmath202 , and @xmath203 .
furthemore , the fundamental group of the complement of @xmath204 is also @xmath205 .
the normal circle @xmath206 to @xmath148 can be deformed using an exceptional sphere section , thus trivial in @xmath207 . in @xcite ,
yusuf gurtas generalized the constructions in @xcite even further . in @xcite
he presented the positive dehn twist expression for a new set of involutions in the mapping class group @xmath208 of a compact , closed , oriented @xmath20-dimensional surface @xmath209 .
these involutions were obtained by gluing the horizontal involution on a surface @xmath210 and the vertical involution on a surface @xmath211 , where @xmath212 is a positive even number .
let @xmath126 denote the involution under consideration on the surface @xmath213 , as shown in figure [ fig : f4 ] below .
according to gurtas @xcite , @xmath126 can be expressed as a product of @xmath214 positive dehn twists .
observe that if we set @xmath215 and @xmath216 , we recover the family of lefschetz fibrations in @xcite . of the surface @xmath217 in the above notation ,
let us set @xmath218 and @xmath219 , for reasons which will be clear soon .
let @xmath220 denote the total space of the lefschetz fibration defined by the word @xmath221 in the mapping class group @xmath222 .
the manifold @xmath220 has a genus @xmath223 lefschetz fibration over @xmath224 with @xmath225 singular fiberes and the vanishing cycles all are about nonseparating curves @xcite .
the euler characteristic of the symplectic @xmath1-manifold @xmath220 can be computed using the following formula @xmath226 .
the signature of the lefschetz fibration described by the word @xmath221 was computed in @xcite : @xmath227 ( see also related work in @xcite ) .
now , using the formulas @xmath93 , @xmath228 , we compute : @xmath229 and @xmath230 .
it is not too difficult to determine what should be the appropriate branched - cover description of the above lefschetz fibrations .
it is given as follows : take a double branched cover of @xmath128 along the union of @xmath231 disjoint copies of @xmath129 and two disjoint copies of @xmath130 ( see figure [ fig : gbc ] ) .
the resulting branched cover has @xmath232 singular points , corresponding to the number of the intersections points of the horizontal spheres and the vertical genus @xmath131 surfaces in the branch set .
we desingularize this manifold to obtain @xmath233 .
a generic horizontal fiber is the double cover of @xmath4 , branched over two points .
thus , we have a sphere fibration on @xmath234 .
a generic fiber of the vertical fibration is the double cover of @xmath134 , branched over @xmath231 points .
thus , a generic fiber of the vertical fibration has genus @xmath235 .
furthermore , two complicated singular fibers of the vertical fibration can be perturbed into @xmath236 lefschetz type singular fibers . ]
next , we recall the main theorem of @xcite , which will be needed in the proof of our lemma [ sixl ] .
we refer the reader to @xcite for any unexplained notation .
[ mainthm ] the positive dehn twist expression for the involution @xmath126 is given by @xmath237 now , we prove two lemmas which we use in the sequel .
[ sixl ] the following relations hold in the fundamental group of @xmath220 @xmath238 notice that the first set of relations simply follows from the fact that the dehn twists along the curves @xmath239 , @xmath240 , @xmath158 , @xmath241 appear in the factorization of @xmath126 . to prove the second set of relations , we use the relations @xmath242 , @xmath243 , @xmath158 , @xmath244 , @xmath245 , @xmath158 , @xmath246 , and @xmath247 in the fundamental group of @xmath220 .
the proof is identical to the proof of lemma [ sixl ] and therefore is omitted .
[ sixl ] the genus @xmath248 lefschetz fibration on @xmath220 admits at least @xmath232 disjoint @xmath98 sphere sections .
we first observe that @xmath220 is the symplectic sum @xmath249 and @xmath250 along the spheres @xmath251 and the sphere fiber of @xmath252 . since a generic sphere fiber of @xmath252 intersects a generic genus @xmath253 fiber at two points , after the symplectic sum we obtain a genus @xmath17 fibration on @xmath220 .
since for @xmath10 , the genus @xmath253 fibration on @xmath252 admits @xmath232 disjoint @xmath76-sphere sections ( see @xcite ) , these @xmath98 sphere sections extends to @xmath220 . the following two family of symplectic building blocks will be used in our construction .
they are obtained from @xmath254 and @xmath255 by performing a sequence of luttinger surgeries along the lagrangian tori @xcite .
the first family has @xmath256 , and the second has @xmath257
. one can relate these symplectic building blocks @xmath258 , @xmath259 and @xmath260 , @xmath261 in @xcite , where @xmath262 and @xmath263 are genus @xmath182 and genus @xmath253 fibered knots respectively . in order to see this connection
, one should view the knot surgery manifold @xmath264 via the sequence of @xmath265 luttinger surgeries on @xmath266 , where @xmath262 is a fibered genus @xmath49 knot in @xmath267 .
this was carefully explained in @xcite for a trefoil knot .
we also refer the reader @xcite ( pages 225 - 226 ) .
to construct the first family of examples , we proceed as follows .
let us fix integers @xmath268 , @xmath269 and @xmath270 , where @xmath271 .
let @xmath272 denote symplectic @xmath1-manifold obtained by performing the following @xmath273 luttinger surgeries on @xmath254 .
these @xmath274 surgeries comprise of the following @xmath138 surgeries @xmath275 together with the following @xmath276 additional luttinger surgeries @xmath277 here , @xmath278 ( @xmath279 ) and @xmath280 ( @xmath281 ) denote the standard loops that generate @xmath282 and @xmath283 , respectively .
see figure [ fig : lagrangian - pair ] for a typical lagrangian tori along which the luttinger surgeries are performed . and
@xmath284 the euler characteristic of @xmath272 is @xmath285 and its signature is @xmath106 .
the fundamental group @xmath286 is generated by @xmath287 ( @xmath279 and @xmath281 ) and the following relations hold in @xmath286 : @xmath288=a_1,\ \ [ a_1^{-1},d_1]=b_1,\ \ [ b_2^{-1},d_2^{-1}]=a_2,\ \ [ a_2^{-1},d_2]=b_2,\\ \nonumber [ d_1^{-1},b_2^{-1}]=c_1^{p_1},\ \ [ c_1^{-1},b_2]=d_1^{q_1},\ \ [ d^{-1}_2,b^{-1}_1]=c_2^{p_2},\ \ [ c_2^{-1},b_1]=d_2^{q_2},\\ \nonumber [ a_1,c_1]=1 , \ \ [ a_1,c_2]=1,\ \ [ a_1,d_2]=1,\ \ [ b_1,c_1]=1,\\ \nonumber [ a_2,c_1]=1 , \ \ [ a_2,c_2]=1,\ \ [ a_2,d_1]=1,\ \ [ b_2,c_2]=1,\\ \nonumber [ a_1,b_1][a_2,b_2]=1,\ \ \prod_{j=1}^n[c_j , d_j]=1,\\ \nonumber [ a_1^{-1},d_3^{-1}]=c_3^{p_3 } , \ \ [ a_2^{-1 } , c_3^{-1 } ] = d_3^{q_3 } , \
\dots , \ [ a_1^{-1},d_n^{-1}]=c_n^{p_n } , \ \ [ a_2^{-1 } , c_n^{-1 } ] = d_n^{q_n},\\ \nonumber [ b_1,c_3]=1,\ \ [ b_2,d_3]=1,\ \dots , \ [ b_1,c_n]=1,\ \ [ b_2,d_n]=1.\end{gathered}\ ] ] the surfaces @xmath289 and @xmath290 in @xmath291 descend to surfaces in @xmath272 .
they are symplectic submanifolds in + @xmath272 .
we will denote their images by @xmath292 and @xmath293 .
note that @xmath294 ^ 2=[\sigma_n]^2=0 $ ] and @xmath294\cdot[\sigma_n]=1 $ ] .
now , we consider a slightly different family .
let us fix a quadruple of integers @xmath10 , @xmath295 , @xmath296 and @xmath297 .
let @xmath298 denote smooth @xmath1-manifold obtained by performing the following @xmath231 torus surgeries on @xmath299 : @xmath300 here , @xmath278
( @xmath301 ) and @xmath302 denote the standard generators of @xmath303 and @xmath304 , respectively .
since all the torus surgeries above are luttinger surgeries when @xmath305 , @xmath306 is a minimal symplectic 4-manifold .
the fundamental group of @xmath298 is generated by @xmath278 ( @xmath307 ) and @xmath302 , and the following relations hold in @xmath308 : @xmath309=a_1,\ \ [ a_1^{-1},d]=b_1,\ \ [ b_2^{-1},d^{-1}]=a_2,\ \ [ a_2^{-1},d]=b_2,\\ \nonumber \cdots , \ \ \cdots , \\
\nonumber [ b_{n-1}^{-1},d^{-1}]=a_{n-1},\ \ [ a_{n-1}^{-1},d]=b_{n-1},\ \ [ d^{-1},b_{n}^{-1}]=c^p,\ \ { [ c^{-1},b_{n}]}^{-m}=d^q,\\ \nonumber [ a_1,c]=1,\ \ [ b_1,c]=1,\ \ [ a_2,c]=1,\ \ [ b_2,c]=1,\\ \nonumber [ a_3,c]=1,\ \ [ b_3,c]=1,\\ \nonumber \cdots , \ \ \cdots ,
\\ \nonumber [ a_{n-1},c]=1,\ \ [ b_{n-1},c]=1,\\ \nonumber [ a_{n},c]=1,\ \ [ a_{n},d]=1,\\ \nonumber [ a_1,b_1][a_2,b_2 ] \cdots [ a_n , b_n]=1,\ \ [ c , d]=1.\end{gathered}\ ] ] let @xmath310 be a genus @xmath311 surface that desend from the surface @xmath312 in @xmath255 .
in this section we prove the following theorem : [ thm : main ] let @xmath18 be @xmath12 for any @xmath13 and @xmath9 .
there exists a new family of smooth closed simply - connected minimal symplectic @xmath1-manifold and an infinite family of non - symplectic @xmath1-manifolds that is homeomorphic but not diffeomorphic to @xmath18
. we will accomplish the proof of this theorem in several steps .
first , we provide the construction which uses a combination of the luttinger surgery @xcite and symplectic fiber sum operations @xcite .
since the order of these operations do not influence the outcome , our construction can be obtained by performing luttinger surgeries on lefschetz fibrations over higher genus surfaces .
next , we show that the fundamental groups are trivial , and determine the homeomorphism types of our examples . in the final step ,
we study the diffeomorphism types of our manifolds and distinguish them from @xmath12 by computing their seiberg - witten invariants and applying theorem @xcite
. our first building block will be the symplectic @xmath1-manifold @xmath313 with a genus @xmath314 symplectic submanifold @xmath315 , a regular fiber of the lefschetz fibration which we discussed section [ sbb ] .
here we endowed @xmath313 with the symplectic structure induced from the given lefschetz fibration .
the other building block will be the symplectic @xmath1-manifold @xmath316 along the symplectic submanifold @xmath317 ( see section [ sbb ] ) , where we set @xmath318 and @xmath319 .
we denote by @xmath320 the symplectic @xmath1-manifold obtained by forming the symplectic fiber sum of @xmath220 and @xmath316 along the surfaces @xmath321 and @xmath322 .
recall from section [ sbb ] that loops @xmath154 ,
@xmath155 , @xmath158 , @xmath323 , and @xmath150 generate the inclusion - induced image of @xmath324 inside @xmath325 .
this is due to the fact that the loops @xmath326 , @xmath327 , @xmath158 , @xmath328 , @xmath241 and the normal circle to @xmath329 to @xmath321 are all nullhomotopic in @xmath330 .
as before , let @xmath331 , @xmath332 , @xmath158 @xmath333 , @xmath334 and @xmath335 $ ] generate @xmath336 in @xmath337 .
we choose the gluing diffeomorphism @xmath338 that maps the generators of the fundamental groups as follows : @xmath339 by gompf s theorem in @xcite , @xmath320 is symplectic .
[ lemma : pi_1(x)=1 ] @xmath320 is simply - connected . by van kampen s theorem , we have @xmath340 since the loops @xmath326 , @xmath327 , @xmath158 , @xmath341 , @xmath328 and the normal circle to @xmath329 are all nullhomotopic in @xmath342 , we get the following presentation for the fundamental group of @xmath320 . @xmath343=a_1,\ \ [ a_1^{-1},d]=b_1 , \\
\nonumber & & \cdots , \ \
\cdots , \\ \nonumber & & [ b_{2k}^{-1},d^{-1}]=a_{2k},\ \ [ a_{2k}^{-1},d]=b_{2k},\ \ [ d^{-1},b_{2k}^{-1}]=c,\ \ [ c^{-1},b_{2k}]=d,\\ \nonumber & & [ a_1,c]=1,\ \ [ b_1,c]=1,\ \ [ a_2,c]=1,\ \ [ b_2,c]=1,\\ \nonumber & & [ a_3,c]=1,\ \ [ b_3,c]=1,\\ \nonumber & & \cdots , \ \ \cdots , \\ \nonumber & & [ a_{2k-1},c]=1,\ \ [ b_{2k-1},c]=1,\\ \nonumber & & [ a_{2k},c]=1,\ \ [ a_{2k},d]=1,\\ \nonumber & & [ a_1,b_1][a_2,b_2 ] \cdots [ a_{2k},b_{2k}]=1,\ \ [ c , d]=1 , \\
\nonumber & & a_{1}a_{2k } = 1,\ \
a_2a_{2k-1}=1,\ \
\cdots , \ \
a_{k}a_{k+1}=1 , \\
\nonumber & & b_{1}b_{2}\cdots b_{2k } = 1,\
\ b_{2}b_{3}\cdots b_{2k-1}= [ a_{2k},b_{2k } ] , \\
\nonumber & & \cdots , \cdots , \cdots \\ \nonumber & & b_{i+1}b_{i+2 } \cdots b_{2k - i } = [ a_{2k - i+1},b_{2k - i+1 } ] \cdots [ a_{2k-1},b_{2k-1}][a_{2k},b_{2k } ] , \\ \nonumber & & [ a_1,b_1][a_2,b_2 ] \cdots [ a_{k},b_{k}]=1 \rangle.\end{aligned}\ ] ] to prove @xmath344 , it is enough to prove that @xmath345 , which in turn will imply that all other generators are trivial . using the last set of identities
, we have @xmath346 , @xmath158 , @xmath347 .
now , we rewrite the relation @xmath348=b_{1}$ ] as @xmath349 = b_{1}$ ] .
since @xmath349 = 1 $ ] , we obtain @xmath345 , which in turn imply that @xmath350 using the relations @xmath351=a_1 $ ] , @xmath352 , @xmath353 , @xmath354 $ ] , @xmath355=c$ ] and @xmath356=d$ ] . since @xmath357=a_{i}$ ] and @xmath358=b_{i}$ ] , for any @xmath359 , we have @xmath360 .
thus , @xmath361 is trivial .
@xmath362 , @xmath363 , @xmath364 , and @xmath365 .
since @xmath366 , @xmath367 , @xmath368 , and @xmath369 .
we easily compute that @xmath370 , @xmath371 , @xmath372 , and @xmath373 . since @xmath374 , @xmath375 , @xmath376 and @xmath377 , which proves the lemma . using the freedman s classification theorem for simply - connected 4-manifolds ( cf .
@xcite ) , we conclude that @xmath320 is homeomorphic to @xmath12 .
since @xmath320 is symplectic , by taubes s theorem ( cf .
@xcite ) @xmath378 , where @xmath379 is the canonical class of @xmath320 .
next , by we applying the connected sum theorem ( cf .
@xcite ) for the seiberg - witten invariant , we deduce that the seiberg - witten invariant of @xmath12 is trivial . since the seiberg - witten invariant is a diffeomorphism invariant ,
we conclude that @xmath320 is not diffeomorphic to @xmath12 .
all @xmath232 exceptional spheres @xmath380 , @xmath381 , @xmath158 , @xmath382 , and @xmath383 , which are the sections of the genus @xmath17 fibration on @xmath313 , meet with the fiber @xmath384 .
furthemore , any embedded symplectic @xmath98 sphere in @xmath220 has the form @xmath385 , thus intersect non - trivially with the fiber @xmath384 . using the usher s theorem ,
the symplectic sum manifold @xmath320 is a minimal symplectic @xmath1-manifold . since symplectic minimality implies smooth minimality ( cf .
@xcite ) , @xmath320 is also smoothly minimal . in order to produce an infinite family of exotic @xmath12 s
, we replace the building block @xmath316 with @xmath386 , where @xmath387 .
let us denote the resulting smooth manifolds as @xmath388 . in the presentation of the fundamental group
, this amounts replacing a single relation @xmath389 = d$ ] with @xmath390}^{-m}=d$ ] .
notice that chainging this relation does not affect our calculation of @xmath391 .
hence all the fundamental group calculations follow the same lines and result in the trivial group . moreover , using the argument similar as in @xcite ( section 4 , pages 12 - 18 ) , we see that @xmath320 has one basic class up to sign , the canonical class @xmath392 .
let @xmath393 denote the symplectic @xmath1-manifold obtained by performing the following luttinger surgery on : @xmath394 .
it is easy to see that @xmath395 and the canonical class @xmath396 + \sum_{j=1}^{4k}[r_j]$ ] , where @xmath397 are tori of self - intersection @xmath98 .
furthermore , the basic class @xmath398 of @xmath388 corresponding to the canonical class @xmath399 satisfies @xmath400 .
thus , every @xmath388 with @xmath401 is nonsymplectic .
in this section we modify the above construction to obtain symplectic @xmath1-manifolds with the various finitely generated fundamental groups , such as the free groups of rank @xmath11 , and the finite free products of cyclic groups , using the _
luttinger surgery_. our construction can be generalized further to obtain symplectic @xmath1-manifolds with arbitrarily finitely presented fundamental groups and with _ small size _ , but we will not pursue this here .
we refer the reader to the articles in @xcite and @xcite where this problem has been thoroughly studied using the _ luttinger
surgery_. our first building block will be the symplectic @xmath1-manifold @xmath313 along with a regular fiber of the genus @xmath314 lefschetz fibration on @xmath220 ( see section [ sbb ] ) .
we equip @xmath313 with the symplectic structure induced from the given lefschetz fibration
. the second building block will be the symplectic @xmath1-manifold @xmath402 discussed in the section [ l ] , along with the symplectic submanifold @xmath403 ( see section [ sbb ] ) , where we set @xmath404 . to simplify the notation
, we set @xmath405 and @xmath406 throughout this section . let @xmath407 denote the symplectic @xmath1-manifold obtained by forming the symplectic fiber sum of @xmath220 and @xmath402 along the surfaces @xmath321 and @xmath408 .
let @xmath409 , @xmath410 , @xmath158 , @xmath411 , @xmath412 , and @xmath413[a_{2},b_{2}]$ ] generate @xmath414 in @xmath415 ( see section [ sbb ] ) .
we choose the gluing diffeomorphism @xmath416 that maps the generators of the fundamental groups as follows : @xmath417 by gompf s theorem in @xcite , @xmath418 is symplectic .
by van kampen s theorem , we have @xmath419 since the loops @xmath326 , @xmath327 , @xmath158 , @xmath341 , @xmath328 and the normal circle to @xmath329 are all nullhomotopic in @xmath342 , we get the following presentation for the fundamental group of @xmath420 . @xmath421=a_1,\ \ [ a_1^{-1},d_1]=b_1,\ \ [ b_2^{-1},d_2^{-1}]=a_2,\ \ [ a_2^{-1},d_2]=b_2,\\ \nonumber & & [ d_1^{-1},b_2^{-1}]=c_1^{p_1},\ \ [ c_1^{-1},b_2]=d_1^{q_1},\ \ [ d_2^{-1},b_1^{-1}]=c_2^{p_2},\ \ [ c_2^{-1},b_1]=d_2^{q_2},\\ \nonumber & & [ a_1,c_1]=1 , \ \ [ a_1,c_2]=1,\ \ [ a_1,d_2]=1,\ \ [ b_1,c_1]=1 , \\
\nonumber & & [ a_2,c_1]=1 , \ \ [ a_2,c_2]=1,\ \ [ a_2,d_1]=1,\ \ [ b_2,c_2]=1 , \\
\nonumber & & [ a_1,b_1][a_2,b_2]=1,\ \ \prod_{j=1}^{2k}[c_j , d_j]=1 , \\
\nonumber & & [ a_1^{-1},d_3^{-1}]=c_3^{p_3 } , \ \ [ a_2^{-1 } , c_3^{-1 } ] = d_3^{q_3 } , \\
\nonumber & & \cdots , \ \ ,
\cdots , \\
\nonumber & & [ a_1^{-1},d_{2k}^{-1}]=c_{2k}^{p_l } , \ \ [ a_2^{-1 } , c_{2k}^{-1 } ] = d_{2k}^{q_l},\\ \nonumber & & [ b_1,c_3]=1,\ \ [ b_2,d_3]=1,\ \ \cdots , [ b_1,c_{2k}]=1,\ \ [ b_2,d_{2k}]=1 , \\
\nonumber & & c_{1}c_{2k } = 1,\ \
c_2c_{2k-1}=1,\ \
\cdots , \ \
c_{k}c_{k+1}=1 , \\
\nonumber & & d_{1}d_{2}\cdots d_{2k } = 1,\ \ d_2d_{3}\cdots d_{2k-1}= [ c_{2k},d_{2k}],\ \
\cdots , \\
\nonumber & & d_{i+1}d_{i+2 } \cdots d_{2k - i } = [ c_{2k - i+1},d_{2k - i+1 } ] \cdots [ c_{2k-1},d_{2k-1}][c_{2k},d_{2k } ] \rangle . \\
\nonumber\end{aligned}\ ] ] to realize the free group of rank @xmath11 as the fundamental groups , we simply set @xmath422 , @xmath423 in the above presentation . using the identity @xmath424 , we rewrite the relation @xmath425=d_{2k}$ ] as @xmath426 = d_{2k}$ ] . since @xmath427 = 1 $ ]
, we obtain @xmath428 .
@xmath428 in turn implies @xmath429 using the relations @xmath430 $ ] and @xmath431 .
next , using @xmath432=a_1 $ ] , @xmath433=b_1 $ ] and @xmath434=c_1 $ ] , we obtain @xmath435 .
since @xmath436=d_2 $ ] and @xmath437=c_2 $ ] , we have @xmath438 , which in turn lead @xmath439 . next , using the relations @xmath440=d_{i}$ ] for any @xmath441 , we have @xmath442 . since @xmath443 for any @xmath444 and @xmath445 ,
we conclude that @xmath446 is a free group of rank @xmath447 generated by @xmath448 , @xmath158 , @xmath449 . to realize the finite free products of cyclic groups as the fundamental groups , we simply set @xmath450 , @xmath451 , and let @xmath452 vary arbitrarily in the above presentation .
a. akhmedov was partially supported by nsf grants frg-0244663 and dms-1005741 .
n. saglam was partially supported by nsf grants frg-0244663 . | in @xcite , the first author constructed the first known examples of exotic minimal symplectic @xmath0 and minimal symplectic @xmath1-manifold that is homeomorphic but not diffeomorphic to @xmath2 .
the construction in @xcite uses y. matsumoto s genus two lefschetz fibrations on @xmath3 over @xmath4 along with the fake symplectic @xmath5 construction given in @xcite .
the main goal in this paper is to generalize the construction in @xcite using the higher genus versions of matsumoto s fibration constructed by mustafa korkmaz and yusuf gurtas on @xmath6 for any @xmath7 and @xmath8 , and @xmath9 and @xmath10 , respectively . using our building blocks
, we also construct symplectic @xmath1-manifolds with the free group of rank @xmath11 and various other finitely generated groups as the fundamental group . |
in this work we focus on the efficiency limits of crystalline silicon ( c - si ) solar cells . in this regard , we consider both ideal devices ( perfect material and interfaces ) , as well as more realistic conditions , including defect - related recombinations , parasitic losses , and non - optimal light management .
it is currently possible to fabricate c - si solar cells with absorber thickness values that differ by two orders of magnitude . on the one end of the thickness range
there are conventional wafer - based c - si solar cells with the absorber thickness of the order of a few hundreds of micrometers @xcite .
epitaxial growth allows fabricating solar cells with the thickness of a few tens of micrometers @xcite .
finally , epitaxy - free fabrication@xcite makes possible fabricating c - si solar cells with the absorbing layer as thin as @xmath0 . for a proper design strategy ,
it is important to determine the optimal absorber thickness in terms of the energy conversion efficiency .
we address this question using a model single - junction silicon solar cell
. it is well know that reducing the silicon thickness leads to the reduction of total absorption , and thus of the photocurrent .
this has to be compensated by implementing an appropriate light - trapping scheme .
yet , it should be emphasized that even when the optimal light trapping is applied ( i.e. , perfect anti - reflection action combined with a lambertian scatterer ) , the maximum achievable absorption still decreases with decreasing material thickness @xcite . on the other hand
, @xmath3 generally tends to decrease with increasing thickness . for a given material quality
, this leads to an optimal thickness that maximizes the conversion efficiency @xcite . in this context , the goal of this work is twofold : first , to establish the most suitable thickness range for c - si solar cells to approach the efficiency limits .
second , to identify the parameters that have to be improved in order to increase the energy conversion efficiency beyond the current record of 25.6% @xcite .
we use efficiency as a figure of merit to assess different solar cell structures .
this is motivated by the fact that the cost of electricity is mainly determined by the efficiency rather than by the cost of the active material ( which , in the case of silicon , is constantly decreasing ) . in this regard , it is particularly important to estimate the optimal absorber thickness range that maximizes the efficiency . in our analysis
we include intrinsic auger recombination , as well as defect - based bulk and surface recombination mechanisms .
we also introduce a simple approach that allows us to consider parasitic losses .
we compare our results with the measured performance of state - of - the - art silicon solar cells , showing the room for improvement in terms of current , voltage , and fill factor .
we pay particular attention to parasitic losses , which we believe are a major roadblock in reaching efficiency above the current record of 25.6% @xcite .
a review of the efficiency limits of silicon solar cells is given in ref .
the limits reported in the literature @xcite are usually calculated using the idealized diode equation @xcite .
this approach has the following limitations : 1 .
the diode equation gives less accurate results when the cell thickness is decreasing .
we attribute this inaccuracy to the assumptions underlying the treatment of the space - charge region ( scr ) in the idealized diode formalism . in practice
, idealized diode equation tends to significantly overestimate efficiency for thin cells .
we further discuss the accuracy of the results obtained using the idealized diode equation in the appendix .
solar cells require selective contacts , which can be achieved using a @xmath4-@xmath5 junction . yet ,
the junction is not explicitly considered in the ideal diode approach .
therefore , ideal diode equation gives unrealistic results that overestimate efficiency for undoped silicon . in our approach
, we calculate the efficiency limits overcoming these limitations : we analytically obtain the photogeneration rate ( ranging from double - pass absorption to lambertian light trapping ) , explicitly consider a @xmath4-@xmath5 junction , and numerically solve the drift - diffusion equations .
this allows us to calculate more realistic efficiency limits of silicon solar cells in a wide range of cell thicknesses and doping levels .
the paper is organized as follows : in sec .
[ sec : num_approach ] we describe our approach , based on analytical photogeneration rate and numerical solution of drift - diffusion equations . in sec .
[ sec : limits ] we calculate the efficiency limits for c - si solar cells . in sec .
[ sec : lt ] we discuss the effects of incomplete light trapping in approaching the efficiency limits . in sec .
[ sec : srh ] we cover the impact of bulk material imperfections on the cell performance , and introduce a simple approach to include parasitics losses in the analysis . in sec .
[ sec : sr ] we discuss the role of surface recombination . in sec .
[ sec : requirements ] we quantify the requirements , in terms of bulk and surface material quality , to achieve a given efficiency level .
conclusions are given in sec .
[ sec : conclusions ] .
finally , in the appendix we compare the results of our numerical treatment with those obtained using the idealized diode equation .
investigated solar cell structure consisting of a @xmath6 thick @xmath5-type emitter ( the blue region ) , @xmath4-type base ( the green region ) , and a perfect back reflector ( br ) , which also serves as a back contact .
the front surface is textured so that the incident light is scattered and trapped within the absorber.,scaledwidth=60.0% ]
let us consider the structure sketched in fig .
[ fig : fig1 ] .
it consists of a @xmath6 thick @xmath5-type emitter and a @xmath4-type base of variable thickness .
such a thin emitter allows to minimise recombination losses in this heavily doped layer . to calculate the efficiency limits we assume no reflection at the front interface , a perfect back reflector ( br ) , and a lambertian light trapping @xcite . in fig .
[ fig : fig1 ] we schematically show the lambertian scatterer at the front , yet we note that the photogeneration rate is calculated analytically and in the electrical calculations the structure is assumed to be flat .
finally , we assume full - area contacts : the carriers are collected at the silicon / br and emitter / front contact interfaces .
the photogeneration rate corresponding to the lambertian limit is calculated as in ref .
@xcite : @xmath7 where @xmath8 is the refractive index of c - si @xcite and @xmath9 is the photon flux density corresponding to the am1.5 g solar spectrum @xcite .
the effective absorption coefficient @xmath10 is related to the light path enhancement in textured cells according to ref .
@xcite . in the literature
we can find a number of different strategies that allow approaching the lambertian limit , including ordered @xcite and quasi - ordered @xcite photonic structures , as well as random textures @xcite .
the photogeneration rate calculated using eq .
[ eq : g_ll ] is integrated with respect to energy ( over the solar spectrum ) and used as the generation term in the drift - diffusion equations , which are solved using finite - element method ( fem ) .
we use fem implementation in the commercial device simulator silvaco atlas @xcite .
this methodology is similar to the one described in our previous works @xcite .
we start by considering solar cells limited by intrinsic auger recombination .
we treat auger recombination using the parametrization reported in ref . @xcite .
we also consider band gap narrowing ( bgn ) according to the model by schenk @xcite .
finally , we neglect free carrier absorption , which is a second - order effect @xcite .
silicon is an indirect band gap material , and thus we also neglect losses related to radiative recombination , which may give an appreciable effect only for very thick cells . yet , for the thick cells the probability of photon recycling increases @xcite , i.e. , radiatively emitted photons are reabsorbed .
these two effects are likely to compensate each other . in the appendix
we show that radiative recombination together with photon recycling have a negligible effect on the cell performance .
( a ) the limiting efficiency of c - si solar cells as a function of the absorber thickness .
( b ) @xmath3 and @xmath11 , and ( c ) fill factor ( ff ) corresponding to the calculated efficiency limits .
the solid lines are calculated by optimizing the base doping for each thickness , whereas the dashed lines are calculated assuming the constant doping @xmath12 .
the triangles denote the performance of the record - efficiency hit cells @xcite.,scaledwidth=55.0% ] in fig .
[ fig : fig2](a ) we show the limiting efficiency of c - si solar cells as a function of the absorber thickness , calculated for the structure sketched in fig .
[ fig : fig1 ] . for each thickness , we have simultaneously optimized the emitter and base doping concentrations ( the doping profile in each layer is assumed to be constant ) . the efficiency as a function of emitter doping
@xmath13 has a wide maximum , and the optimal value @xmath14 does not change with the absorber thickness , as the emitter thickness itself is kept constant . regarding the base doping @xmath15 , the solid lines in fig .
[ fig : fig2 ] are calculated by optimizing @xmath15 for each thickness . on the other hand ,
the dashed lines are calculated assuming a constant doping @xmath12 , which is the optimal doping for the optimal absorber thickness equal to @xmath16 .
the optimal base doping is highest for thinner cells ( reaching @xmath17 for the @xmath0 thick cell ) and decreases with increasing thickness .
the maximum efficiency is equal to @xmath18 , and it is obtained for the @xmath16 thick cell . this is lower than the efficiency limit of 29.43%
reported recently in the literature @xcite .
the main reason is that in our calculations , the base doping increases auger losses and therefore reduces the efficiency . in the contrary ,
the limit reported in the literature has been obtained using the idealized diode equation and assuming undoped silicon , as explained in sec .
[ sec : intro ] and in the appendix .
we note that the efficiency as a function of the absorber thickness has a broad maximum , and even a small change in the input parameters can substantially shift the nominal optimal thickness . in the range between 40 and @xmath1 ,
the calculated efficiency differs by no more than @xmath19 ( relative units ) from @xmath20 .
this somehow arbitrary interval shows that for thickness values that differ considerably ( by one order of magnitude ) , the maximum achievable efficiency is very similar .
finally , for an absorber thickness below @xmath2 , the efficiency drops significantly .
these results are compared with the efficiencies of the state - of - the - art hit cells @xcite ( the triangles ) .
we note that the efficiency of the hit solar cell has been recently increased to @xmath21 @xcite . yet , the thickness of the record cell is not specified , and therefore we were unable to include it in fig .
[ fig : fig2 ] .
this analysis shows that the theoretical margin for improvement of single - junction silicon solar cells is around 3% ( absolute units ) .
efficiency as a function of the base doping and absorber thickness .
the dashed line indicates the optimal doping.,scaledwidth=60.0% ] in fig .
[ fig : fig2](b ) we show @xmath3 and @xmath11 corresponding to the calculated efficiency limits .
the base doping optimization allows to slightly improve @xmath3 , whereas @xmath11 is practically the same . comparing these results with the performance of the hit cells
, we can see that for both values of thickness ( 98 and @xmath22 ) , the measured value is around 99% of the limiting @xmath3 , around 94.6% of the limiting ff , and around 91.6% of the limiting @xmath11 . therefore , these cells are limited by , in decreasing order of significance , @xmath11 , ff , and @xmath3 .
this shows that optimizing light trapping and , especially , minimizing optical parasitic losses ( affecting @xmath11 ) , as well as improving the contacts ( affecting ff ) is the key to further increase the efficiency towards the limiting values .
to further elaborate the issue of the base doping optimization , in fig .
[ fig : fig3 ] we show the efficiency as a function of the base doping and absorber thickness .
this plot can be compared with the results presented in fig .
4 reported in ref .
@xcite , which are obtained using the idealized diode equation . in the case of the diode equation
, solar cells approach the limiting efficiency for a practically undoped silicon . yet , in our calculations we can see a clear maximum around @xmath23 ( for the thickness range around 100 @xmath24 )
. then , efficiency decreases with decreasing doping .
this difference is due to the fact that the junction is not explicitly considered in the ideal diode approach .
therefore , the requirement of selective contacts , which can be achieved using a @xmath4-@xmath5 junction , is not included .
this leads to unrealistic results for very lightly doped materials . for the simplicity of the analysis , in the rest of this paper we assume the constant base doping @xmath12
in the calculations above we have assumed a lambertian light trapping , which is often taken as a benchmark in the optical design of solar cells . yet , it is difficult to fulfil this assumption in realistic devices
. therefore , let us now focus on the role of light trapping in achieving the efficiency limits .
to do so , we consider solar cells with incomplete light trapping : the photogeneration rate is taken as a weighted average of the photogeneration @xmath25 corresponding to the double - pass absorption ( i.e. , unstructured cell ) and the photogeneration @xmath26 corresponding to the lambertian limit .
@xmath25 is calculated as @xmath27 where @xmath28 is the absorption coefficient of c - si @xcite .
as previously , also in the double - pass case we assume a perfect back reflector and anti - reflection action .
we note that there are no interference effects included in eq .
( [ eq : gg_2p ] ) .
when we consider single wavelengths , the interference effects are profound . yet , when we integrate the photogeneration rate over solar spectrum , the interference peaks are smeared out , which justifies the approximation used in eq .
( [ eq : gg_2p ] ) .
the resulting weighted photogeneration rate is calculated as @xmath29 , \label{eq : phgen}\ ] ] where @xmath30 is the light - trapping factor : @xmath31 corresponds to the double - pass case , whereas @xmath32 corresponds to the lambertian limit .
@xmath33 is a function of depth @xmath34 .
finally , @xmath35 is a factor related to reflection at the front interface and to parasitic optical losses . for simplicity , we assume that @xmath35 does not depend on wavelength . ( a ) efficiency , ( b ) short - circuit current @xmath11 , ( c ) open - circuit voltage @xmath3 calculated as a function of the absorber thickness for different values of the light - trapping factor @xmath30.,scaledwidth=60.0% ] we begin by assuming no parasitic losses , that is @xmath36 . in fig .
[ fig : fig4](a ) we demonstrate that light trapping increases the maximum achievable efficiency .
this is not the case for semiconductors with a direct energy band - gap , like gaas , where surface texturing does not increase the maximum efficiency @xcite .
therefore , for silicon solar cells light trapping is an essential element required to approach the efficiency limits , even for very thick cells
. as expected , light trapping significantly increases @xmath11 , which is shown in fig .
[ fig : fig4](b ) .
@xmath11 for the thickest cells does not saturate because of the small absorption below the energy band - gap
. moreover , fig .
[ fig : fig4](c ) shows that light trapping also slightly improves @xmath3 . yet , this effect is appreciable only for cell thicknesses below @xmath37 , that is in a thickness range which is not particularly promising for achieving high efficiency .
@xmath3 as a function of the thickness for the double - pass case exhibits a gentle maximum
. this may be partly because the base doping is not optimized for thin cells .
although the importance of light trapping for maximizing the conversion efficiency of c - si solar cells is well known , the present results allow to quantify the effects of incomplete light trapping on the efficiency .
moreover , eq . ( [ eq : phgen ] ) represents a simple model that can be also used to study the effects of parasitic losses , which are different from the effects related to incomplete light trapping .
we shall discuss it in more details in the next section .
so far , we have considered an idealized material , and therefore the cell performance was limited by intrinsic auger recombination .
let us now consider extrinsic losses related to defect - base srh recombination and parasitic optical losses . with the emitter as thin as @xmath6
, the cell efficiency is likely to be limited by the diffusion length @xmath38 of the minority carriers ( electrons ) in the base . in fig .
[ fig : fig5](a ) we show the efficiency as a function of the absorber thickness and @xmath38 , calculated for the structure with lambertian light trapping .
the diffusion length of holes in the highly - doped emitter is taken equal to @xmath39 .
it can be seen that thicker cells are more sensitive to srh recombination . in this
regard , thinner cells with the absorber thickness of around @xmath2 can reach nearly the same @xmath20 as the thicker cells , but the conditions in terms of absorber quality are relaxed . finally , reducing the absorber quality ( i.e. , decreasing @xmath38 ) shifts the optimal thickness towards thinner cells .
results shown in fig . [ fig : fig5](b ) are calculated including parasitic losses @xmath40 .
we estimate that this is approximately the level of parasitic losses in the hit cells discussed above .
the trends are similar to those calculated assuming @xmath36 , yet the maximum achievable efficiency is proportionally decreased .
moreover , for a given efficiency level , the optimal thickness is increased . the results of fig .
[ fig : fig5 ] allow us to estimate the optimum thickness and the required material quality to reach a given efficiency level .
for example , to surpass 26% efficiency , in the idealized case of no parasitic losses and complete ( lambertian ) light trapping , we need an electron diffusion length around @xmath41 . in the more realistic case , when parasitic losses are included
( say , @xmath40 ) , achieving efficiency above 26% requires a diffusion length higher than @xmath42 for an optimal thickness of @xmath43 .
( a ) efficiency as a function of the thickness and diffusion length @xmath38 of the minority carriers ( electrons ) in the base , calculated for the structures with a lambertian light trapping .
( b ) the same quantity but including factor related to parasitic losses @xmath40 .
we note that @xmath38 is related to srh recombination , originating from defects in the base .
the scale on the right shows the corresponding carrier lifetime , calculated assuming electron mobility @xmath44.,scaledwidth=70.0% ]
efficiency as a function of ( a ) front and ( b ) back surface recombination velocity ( srv ) , and of the absorber thickness . for the bulk transport losses ,
we assume only auger recombination.,scaledwidth=70.0% ] efficiency as a function of the front surface recombination velocity ( srv ) and the absorber thickness .
for the bulk transport losses , we assume auger and srh recombination .
the material quality ( i.e. , @xmath38 ) is reduced , so that the maximum achievable efficiency is 27% for @xmath45 ( a ) and 25% for @xmath46 ( b).,scaledwidth=70.0% ] let us now estimate the constraints on the efficiency imposed by surface recombination , which is another extrinsic loss mechanism related to defect states at the surface . in fig .
[ fig : fig6 ] we show the efficiency as a function of ( a ) front and ( b ) back surface recombination velocity ( srv ) , and of the absorber thickness . at first , for the bulk transport losses we assume only auger recombination .
it can be seen that approaching the efficiency limits requires srv to be less than a few cm / s , with solar cells being more sensitive to recombination at the rear interface rather than to recombination at the front . in the calculations
above , we assumed a perfect material quality . yet , in our previous work we have demonstrated that the importance of surface recombination depends on bulk recombination rate , i.e. , the higher is the material quality , the more important are losses at the surface @xcite . from now on we focus on front surface recombination , neglecting recombination at the back , which has anyway to be of the order of a few cm / s to target high efficiency .
in realistic devices , there is usually an oxide passivating layer at the back and point contacts are implemented . for this reason
, back srv can be significantly reduced . on the other hand
, front srv may be increased due to the texturing . in fig .
[ fig : fig7 ] we plot the efficiency as a function of the front surface recombination velocity ( srv ) and absorber thickness . for bulk transport losses ,
we assume auger and srh recombination .
we adjust the diffusion length @xmath38 of the minority carriers ( electrons ) in the base , so that the maximum achievable efficiency is reduced to a certain value .
this kind of analysis allows to answer the following question : having a given material quality , what is the maximum allowed value of the front srv to achieve the limiting efficiency ? in fig .
[ fig : fig7](a ) the maximum possible efficiency is around 27% , which is obtained for @xmath45 . with this material quality ,
the optimal absorber thickness is @xmath2 , and the maximum efficiency can be achieved for the thickness range between 20 and @xmath47 . in this case
, front srv should be below a few tens of cm / s . in fig .
[ fig : fig7](b ) we further reduce diffusion length to @xmath46 , so that the maximum possible efficiency is 25% .
the region of highest efficiency ( 25% or slightly above ) is clearly wider than in the previous plot .
also the conditions for surface recombination at the front interface are relaxed : in this case , front srv should be below a few hundreds of cm / s .
this analysis confirms that the impact of surface recombination decreases with decreasing material quality .
( a ) diffusion length of electrons in the base @xmath48 as a function of the maximum achievable efficiency . ( b )
corresponding ( optimal ) thickness of the absorbing layer.,scaledwidth=70.0% ] to summarize the results presented in this paper , in this section we estimate the requirements for approaching a given energy conversion efficiency . in this regard ,
we analyse both bulk material quality ( related to diffusion length of electrons in the base @xmath38 ) and quality of the surfaces ( related to surface recombination velocity ) . in fig .
[ fig : fig8](a ) we show the diffusion length of electrons in the base @xmath38 as a function of the maximum achievable efficiency , whereas fig .
[ fig : fig8](b ) indicates the corresponding ( optimal ) thickness of the absorbing layer .
these results are extracted from fig .
[ fig : fig5 ] , i.e. , for each @xmath38 we have extracted the maximum achievable efficiency and the corresponding cell thickness . in fig .
[ fig : fig8](a ) we can distinguish two trends . in the case of no parasitic losses ( black line ) , for the efficiency range below around 28% , @xmath48 and the corresponding maximum achievable efficiency scale linearly .
yet , above 28% , significant increase of @xmath38 gives only a minor , if any , improvement of efficiency .
( a ) front surface recombination velocity ( srv ) as a function of the maximum achievable efficiency . ( b )
corresponding optimal thickness of the absorbing layer .
we consider cells limited by auger recombination ( the black lines ) , as well as cells with srh recombination : @xmath49 with maximum achievable efficiency around 27% ( the red lines ) and @xmath50 with maximum achievable efficiency around 25% ( the blue lines ) .
the colour regions indicate 1% ( relative ) tolerance intervals for each material quality.,scaledwidth=70.0% ] the red line in fig .
[ fig : fig8](a ) shows the results calculated including parasitic losses , that is @xmath40 .
parasitic losses result in a shift of the red curve towards lower efficiencies , with the maximum efficiency of around 26% .
indeed , the measured record efficiency is slightly less than 26% , which identifies parasitic losses as a major roadblock for going beyond 26% .
we can also see that an optimal absorber thickness changes significantly in the considered efficiency range , as shown in fig .
[ fig : fig8](b ) . for small @xmath48 ,
the optimal thickness is of the order of a few tens of micrometers .
then , with increasing @xmath48 , the optimal thickness increases rapidly , reaching around @xmath16 for @xmath48 of the order of millimetres
. a similar analysis can be performed regarding surface recombination . in fig .
[ fig : fig9](a ) we show front surface recombination velocity ( srv ) as a function of the maximum achievable efficiency , and in fig . [ fig : fig9](b ) we show the corresponding optimal thickness of the absorbing layer .
these results have been extracted from fig .
[ fig : fig6 ] and [ fig : fig7 ] .
we consider cells limited by auger recombination ( the black lines ) , as well as cells with srh recombination : @xmath45 with maximum achievable efficiency around 27% ( the red line ) and @xmath46 with maximum achievable efficiency around 25% ( the blue lines ) .
in general , the cells are fairly insensitive to recombination at the front interface if srv is below a few hundreds of cm / s .
this is of the order of magnitude of the effective srv , currently measured for nanostructured solar cells @xcite .
moreover , increasing srv increases the optimal absorber thickness , as thinner cells are more sensitive to surface recombination . as pointed out previously , efficiency as a function of the absorber thickness exhibits a wide maximum .
for this reason , in fig .
[ fig : fig9](b ) we include tolerance intervals for each material quality .
the intervals are indicated by the colour regions , and show thickness range in which efficiency changes by no more than 1% relative .
for example , if we consider auger - limited cells ( the black lines ) , and assume front srv equal to @xmath51 , we can see that the efficiency changes by no more than 1% relative in the thickness range from 80 to @xmath1 .
we note that the calculations are performed up to @xmath1 ( which we consider a practical limit for fabrication ) .
the results presented above show that the optimal absorber thickness is determined by an interplay between bulk and surface losses .
nevertheless , for a reasonably high bulk and surface quality ( consistent with present - day fabrication techniques ) , an optimal thickness is always above @xmath52 .
therefore , the conclusion that decreasing the absorber thickness below a few tens of micrometers is counter - productive holds for a wide range of material parameters .
however , we emphasize that reducing parasitic losses is required in order to increase the energy conversion efficiency above 26% .
in conclusion , we presented an electro - optical framework to calculate the limiting efficiency of c - si solar cells . in our approach
, we take the analytical photogeneration rate ( assuming full or partial light trapping ) and numerically solve the drift - diffusion equations to obtain the cell performance .
this gives a realistic description of carrier dynamics in the device and allows easily introducing intrinsic and extrinsic loss mechanisms .
comparison with present - day hit structures suggests that silicon solar cells are limited by ( in decreasing order of significance ) @xmath11 , ff , and @xmath3 .
thus , improving @xmath11 ( via light trapping and reducing parasitic losses ) and ff ( by improving the quality of contacts ) is a key to further boost the performance beyond the current efficiency record . in the first part of this contribution ,
we calculated the efficiency limits of c - si solar cells .
therefore , we focused on the cells limited by intrinsic auger recombination .
we have demonstrated that the limiting efficiency as a function of the absorber thickness exhibits a wide maximum : @xmath2 thick cell can be nearly as efficient as the solar cell with the optimal absorber thickness ( around @xmath16 ) .
it is therefore more practical to consider the _ optimal thickness range _ , rather than a single optimal thickness .
in this regard , we argued that decreasing the thickness below around @xmath2 is counter - productive , as it significantly reduces the maximum achievable efficiency .
when extrinsic srh recombination is considered , we notice that thicker cells are more sensitive to bulk losses , and therefore the conditions to reach the limiting efficiency for thinner cells are relaxed . yet , including surface recombination shows the opposite trend : thicker cells are less sensitive to surface recombination , as the surface - to - volume ratio is reduced .
nevertheless , for a reasonably high bulk and surface quality , it remains true that decreasing the thickness below a few tens of micrometers is counter - productive in terms of the efficiency .
the current efficiency record for single junction silicon solar cells is 25.6% @xcite .
based on the analysis presented in this paper , we believe that the efficiency above 26% is a reasonable next step , which is within the reach of present - day fabrication techniques . yet , this requires , in the first place , reduction of parasitic losses . surpassing an efficiency of around 27% puts severe constraints on surface recombination .
increasing the efficiency above the calculated limit for single junction c - si solar cells requires novel technologies , like silicon - perovskite tandem structures @xcite .
the efficiency limits of silicon solar cells are usually calculated using idealized diode equation @xcite : @xmath53 where @xmath54 is the photogenerated current density , @xmath55 is the cell thickness , and @xmath56 is the total recombination rate . in this appendix , we elaborate on the difference between the results obtained using eq .
( [ eq : diode_eq ] ) and numerical solution of drift - diffusion equations by means of finite - element approximation .
this allows us to investigate the accuracy of the results obtained using eq .
( [ eq : diode_eq ] ) in a wide absorber thickness range . in fig .
[ fig : fig_appendix ] we show ( a ) efficiency , ( b ) open - circuit voltage @xmath3 , ( c ) short - circuit current @xmath11 , and ( d ) fill factor ff as a function of the absorber thickness .
the red dashed lines refer to the results obtained using the idealized diode equation , whereas the black lines denote the results obtained by solving the drift - diffusion equations by means of finite - element approximation .
the material parameters and recombination models are the same in both approaches : we include auger recombination according to richter _ et al .
_ @xcite and bgn according to the model by schenk @xcite .
( a ) efficiency , ( b ) open - circuit voltage @xmath3 , ( c ) short - circuit current @xmath11 , and ( d ) fill factor ff as a function of absorber thickness .
the red dashed lines refer to the results obtained using the idealized diode equation , whereas the black lines refer to the denote the results obtained by solving the drift - diffusion equations by means of finite - element approximation . in ( a )
we show the comparison with the results digitalized from ref .
@xcite.,scaledwidth=100.0% ] by performing the calculations using eq .
( [ eq : diode_eq ] ) , we aim at reproducing the results reported in ref . @xcite . in fig . [
fig : fig_appendix](a ) we compare our calculations of efficiency with the results presented in the reference work .
the results obtained using eq .
( [ eq : diode_eq ] ) are very close to the results digitalized from ref .
we attribute small discrepancies to the fact that in our calculations we neglect free - carrier absorption and photon recycling : as discussed above , these two effects are likely to compensate each other , and indeed their impact on the cell performance is negligible . for this reason , we have also neglected these effects in the calculations presented in the main part of this paper .
finally , the calculations presented in this work are performed at temperature @xmath57 , whereas the results in the reference work are performed for @xmath58 .
the discrepancy between the results obtained using eq .
( [ eq : diode_eq ] ) and full numerical simulations increases with decreasing absorber thickness , as demonstrated in fig .
[ fig : fig_appendix ] . in the limiting case of very thick cells ,
both approaches give essentially the same results .
yet , the difference in efficiency for @xmath0 thick cell is close to 2% ( absolute value ) , i.e. , the ideal diode treatment gives higher efficiency .
since @xmath11 calculated using both approaches is the same , is the same in both cases also confirms that the mesh used in the finite - element calculations is dense enough to properly reproduce the photogeneration profile .
] we can trace the discrepancy back to the difference in @xmath3 and fill factor : @xmath3 calculated using eq .
( [ eq : diode_eq ] ) increases linearly with decreasing thickness , whereas @xmath3 obtained from the full numerical simulations tends to saturate .
moreover , ff calculated using eq .
( [ eq : diode_eq ] ) increases with decreasing absorber thickness , whereas ff obtained from the full numerical simulations shows the opposite trend ( small fluctuations may be due to numerical inaccuracy ) .
we can therefore conclude that the results obtained using the idealized diode equation become less accurate with decreasing absorber thickness , i.e. , in this approach the efficiency for thin cells is overestimated .
we attribute this inaccuracy to the assumptions regarding the treatment of the space - charge region ( scr ) in the idealized diode formalism : the idealized diode equation can be derived from the drift - diffusion equations assuming that there is no carrier generation nor recombination in scr @xcite .
it means that the currents in scr are constant . with the assumed doping concentrations ,
the width of scr is around @xmath59 .
therefore , if the cells are thick , scr is only a small part of the whole cell .
yet , in the case of thin cells , scr is a significant part of the whole cell , and the assumption that there is no carrier generation nor recombination in scr seriously disturbs the current distribution in the cell .
this work was supported by the eu through marie curie action fp7-people-2010-itn project no
. 264687 prophet .
the authors are grateful to angelo bozzola for carefully reading the manuscript and marco liscidini for many fruitful discussions .
41ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop `` , '' ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop `` , '' @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1002/pip.2385 [ ( ) , 10.1002/pip.2385 ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop `` , '' ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1002/pip.2606 [ ( ) , 10.1002/pip.2606 ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop ( ) @noop _ _ ( , ) | it is currently possible to fabricate crystalline silicon solar cells with the absorber thickness ranging from a few hundreds of micrometers ( conventional wafer - based cells ) to devices as thin as @xmath0 . in this work ,
we use a model single - junction solar cell to calculate the limits of energy conversion efficiency and estimate the optimal absorber thickness .
the limiting efficiency for cells in the thickness range between 40 and @xmath1 is very similar and close to 29% . in this regard , we argue that decreasing the thickness below around @xmath2 is counter - productive , as it significantly reduces the maximum achievable efficiency , even when optimal light trapping is implemented .
we analyse the roles of incomplete light trapping and extrinsic ( bulk and surface ) recombination mechanisms . for a reasonably high material quality ,
consistent with present - day fabrication techniques , the optimal thickness is always higher than a few tens of micrometers .
we identify incomplete light trapping and parasitic losses as a major roadblock in improving the efficiency upon the current record of 25.6% for silicon solar cells . finally , considering the main parameters that impact solar cell performance , we quantify the constraints and requirements for achieving a specified energy conversion efficiency , which is important for a proper design strategy of high efficiency silicon solar cells . |
the space density of peaks in a cosmological density field smoothed with a filter is sensitive to the shape of the linear power spectrum of mass fluctuations , even for filter sizes where the fluctuations are in the non - linear regime ( croft & gaztaaga 1998 ) . in de & croft ( 2007 ) ,
hereafter paper i we explored the sensitivity of the peak density to parameters related to the initial power spectrum , as well as redshift distortions and variations in the galaxy halo occupation distribution ( e.g. , berlind & weinberg 2002 ) . the theory of peaks in a gaussian density field was set out in detail by bardeen ( 1986 , hereafter bbks ) .
the relationship between the peak density and power spectrum in bbks can be used to explore such parameters as the spectral index and its dependence on scale ( kosowsky and turner , 1991 ) along with the neutrino mass . in this paper
we use data from the 2df galaxy redshift survey ( hereafter 2dfgrs , colless 2001 ) , to constrain cosmological parameters using the peak density . in a recent analysis of the 2dfgrs final data set cole ( 2005 ) employed a direct fourier method to compute the power spectrum .
these authors put constraints on several parameters using the directly measured power spectrum shape .
they assumed a primordial form for p(k ) with @xmath11 and @xmath12 along with a negligible neutrino mass .
this gave preferred values of @xmath13 and a baryon fraction of @xmath14 ( 1@xmath15 errors ) .
this analysis therefore implies a significantly lower mass density compared to the @xmath16 often taken as standard . in combination with cmb data from wmap ( spergel 2003 ) cole
find @xmath17 also on large scales some evidence was seen by cole of the baryon oscillations predicted by cdm models .
our present work is complimentary to this large scale linear theory analysis , with the peak density enabling constraints to be placed on the linear power spectrum and cosmological parameters from data on smaller scales .
this paper is the second in a series . in paper
i , we examined the number density of peaks in cosmological simulations and compared to results from linear peak theory ( bardeen 1986 , hereafter bbks ) over a range of density field smoothing filter scales .
this provided knowledge of the length scales for which agreement between theory and simulation can be expected .
the dark matter simulations used to compare to peak theory showed good agreement for filter scales between 3 - 30@xmath18 .
this was assuming that the mean interparticle separation was smaller than the filter scale ( for agreement at better than the @xmath19 level ) .
in addition to simple tests using simulations , we created galaxy catalogues using lists of dark matter halos and the halo occupation distribution formalism ( zheng 2005 . )
good correspondance was found between the peak density measured from the galaxies and dark matter for filter scales @xmath20 .
the peak density can be used to constrain cosmology through its dependence on the power spectrum of mass fluctuations , @xmath21 .
in particular we work with the asymptotic number density of maxima , i.e. the number density of peaks of all heights and which is found by bbks to be : @xmath22 where @xmath23 is defined to be the following ratio of moments of the power spectrum : @xmath24 here @xmath25 the scale dependence of @xmath21 is probed by smoothing the density field with gaussian filter with comoving radius @xmath26 , i.e. multiplying @xmath21 by @xmath27 .
croft & gaztaaga ( 1998 ) found that there is an simple relationship that holds to high accuracy between the local slope of the power spectrum @xmath1 at wavenumber @xmath2 and the filter scale @xmath28 .
the peak number density is therefore sensitive to the power spectrum slope ( but not its amplitude ) and the parameters which govern it can be tested .
paper i focussed on several such aspects of the power spectrum such as the neutrino mass , the possible rolling of the spectral index @xmath1 , and redshift distortions .
it was seen that upon increasing the neutrino mass the peak density decreases , as it does when the rolling of the spectral index is made more negative .
the effect of redshift distortions was also quantified , showing that they act to suppress the number density of peaks on small scales .
this information is useful for the present paper which deals with the peak density in redshift space measured from an observed galaxy catalogue .
our approach in this paper is to first test our recovery of the true peak density using mock catalogues derived from simulations that have had the 2df masks and selection function applied to them .
we then apply the same estimation techniques to the 2df data , using the recovery from mock data to estimate the reliability of the method .
the 2df peak density as a function of filter scale is then used to constrain cosmological parameters .
the plan for the paper is as follows : in section 2 we describe the simulated observations ( mock catalogues ) , including the detailed reasoning behind their use .
we give an overview of the 2dfgrs and in appendix a describe how the mock catalogues were generated .
the latter includes a description of the selection function , magnitude limits and geometrical limitations of the survey and how they were applied to n - body simulations to produce mock catalogues . in section 3
we detail tests of the process of finding peaks in the simulations and mock catalogues , including selection of the volume limits for a given filter size and how we deal with incomplete sky coverage . in section 4
we describe the calculation of the observed peak density from the 2df survey data using the completeness functions calibrated from mock catalogues .
we also describe various sources of uncertainty , and how they propagate into our evaluation of cosmological parameters . in section 5
we present the best fit values of @xmath3 , @xmath1 and d@xmath1/dln@xmath2 including error estimates , using markov chain monte carlo analysis . in section 5
we conclude and discuss possible future work .
the main aim of this paper is the extraction of accurate information about particular cosmological parameters , the neutrino mass , slope of the power spectrum and variation of spectral index with scale ( rolling . ) we can make good estimates of these quantities if we marginalize over parameters whose values are already well known .
this involves use of present day knowledge regarding certain other parameters such as the density of matter , cosmological constant , and baryon density , @xmath29,@xmath30 , @xmath31 .
we work with the space density of peaks in the smoothed galaxy density field . to recover the actual space density of peaks we need to address the issues related to the limitations and particular nature of the galaxy survey . in the 2df
survey data , limitations arise for several reasons .
for example certain galaxies may not make it into the survey because they are faint , but the magnitude limit varies over the survey s angular extent .
very bright objects affect the detection of those nearby , and so there are exclusion holes in the survey .
the survey geometry also has a complicated nature , making it different from the uniform contiguous volume best suited for finding and quantifying peaks . to be able to estimate the space density of peaks in the universe from the observations it is therefore important to be able to compute a completeness function for each sub volume of the survey region .
we do this by computing the ratio of the number of peaks in simulations with full sampling and no boundary effects to the peak density estimated from mock catalogues covering the same simulation volume .
the mock catalogues are simulated observations ( simulations combined with all the relevant observational constraints ) .
when using the mock catalogues , we make the assumption that the completeness function which we generate using them can be applied to the real universe .
this means that we assume that the simulations which are the basis of the mock catalogues are a close enough approximation to the statistical distribution of matter in the real universe that the completeness is valid .
we show in section 3.1 that in practice this is not an important restriction as the completeness is in fact unity ( i.e. no peaks are missed ) for most of the survey volume we actually use . in appendix
a we describe how we generate the mock catalogues . as they are created to have the same geometry and limitations as the 2dfgrs we first describe the observational dataset .
the 2dfgrs ( colless 2001 ) resulted in one of the largest catalogues of galaxy redshifts made so far , following a major spectroscopic survey which took advantage of the unique capabilities of the 2df facility at the anglo - australian observatory . the 2dfgrs obtained spectra for 245591 objects , mainly galaxies , brighter than a nominal extinction - corrected magnitude limit of @xmath32 .
reliable ( 2df quality flag @xmath33 ) redshifts were obtained for 221414 galaxies .
the galaxies cover an area of approximately 1500 square degrees selected from an extended version of the apm galaxy survey ( maddox 1996 ) in three regions : an ngp strip , an sgp strip and random fields scattered around the sgp strip . in figure [ 2df ] we show a projection of the galaxy distribution in the 2dfgrs in the equatorial plane as well as the positions of peaks found after smoothing the density field with a gaussian filter of radius @xmath34 .
( the peak finding will be described later ) .
different masks are used when interpreting the 2df data to characterize the completeness of the survey in various ways as a function of position on sky .
the magnitude limit mask gives the extinction - corrected magnitude limit of the survey at each position on sky .
the redshift completeness mask quantifies the fraction of galaxies above the magnitude limit with measured redshifts .
we use the recommended method ( colless 2001 ) to define this latter mask , it being specified by the complete set of 2 degree fields that were used to tile the survey region for spectroscopic observations . each point on the sky inside the survey boundary
is covered by at least one 2 degree field , but more often by several overlapping fields .
a sector is defined as the region delimited by a unique set of overlapping 2 degree fields .
this is the most natural way of partitioning the sky , as it takes into account the geometry imposed by the pattern of 2 degree fields and the way in which the galaxies were targeted for spectroscopic observations . within each sector with position angle @xmath35 , the redshift completeness @xmath36 , is the ratio of the number of galaxies for which redshifts have been obtained , @xmath37 , to the total number of objects contained in the parent catalogues , @xmath38 so that @xmath39 .
we note that when analyzing the data , the redshift completeness of a given sector , @xmath36 , should be clearly distinguished from the redshift completeness of a given field denoted , by @xmath40 , since multiple overlapping fields can contribute to a single sector .
the @xmath41-mask gives the dependence of the redshift completeness on apparent magnitude .
we describe the @xmath41-mask in more detail in the appendix a. we use these masks ( i.e. values of completeness for different patches of sky ) along with the incorporation of redshift limits , magnitude limits and a completeness cut - off to make our mock catalogues from simulations ( see appendix a ) .
in order to get a feel for how well peaks that are seen in the fully sampled dark matter density field can be recovered in the mock galaxy catalogues , we carry out a one to one comparison . to find peaks operationally ,
we first assign the galaxy ( or particle ) positions to a grid , which is then smoothed with a gaussian filter .
peaks are then associated with local maxima in this smoothed density field ( grid cells which have a higher density than the surrounding 26 cells ) .
the first step towards comparing peak locations in mocks and the fully sampled dark matter density field is to ensure that the same piece of the simulation is used in both the mock catalogues and the fully sampled case .
as mentioned before , our simulations have a box size of @xmath42 mpc and we replicate this box , making use of the periodic boundaries to get the desired survey volume for the mock catalogues .
we apply the 2df mask and selection function , resulting in a volume that matches the shape of the 2df redshift survey . when finding peaks from the mock catalogues it is important to assign weights to the galaxies , the appropriate ones being related to the inverse of the selection function . in order to find the weights , we generate _ random catalogues _ ,
in which points are initially randomly distributed .
we apply the 2df survey mask and selection function to these points , in the same way as with the mock galaxy catalogues .
we then place a grid on top of this distribution of points and count the number of points in each grid cell .
the ratio of the number of points in the grid cell to the expected number for a uniform distribution gives the inverse of the appropriate weight to use for that cell . in this way
, we compute a weight for each cell on the grid . in order to make sure the weights have an acceptably low contribution from poisson noise
we make the number of points in the random catalogues twenty times that in the mock galaxy catalogues .
we place the same three - dimensional grid ( of size @xmath43 cells ) onto the distribution of galaxies in the mock catalogues .
the box size was set equal to the distance @xmath44 as described earlier .
we assign the weighted galaxy density to the grid and smooth this field in fourier space with a gaussian filter .
we carry this out for several different smoothing filter radii , @xmath45 , each time locating the local maxima in the field . for our one - to - one comparison ,
we compare directly the positions of peaks found from the mock catalogues and from the fully sampled simulations . because of shot noise , we do nt expect peaks in the mocks to be found in exactly the same places . in figure [ matching_sim_mock ] , we plot projections of the ngp and sgp regions for one of the mock catalogues , with symbols showing where the peaks were found and other symbols showing where the true peaks lie ( measured from the full simulation . )
we can see there that the distrbutions of points are similar in general , but that there are differences , particularily at the far edges of the survey .
we quantify this in figure [ random_match ] where we show the fraction of peaks in the mock catalogues which have a true peak within one filter radius .
we can see that good peak detections , defined in this way occur approximately 50% of the time , roughly independent of the filter size . in the same plot we have carried out this comparison with peaks found from the random catalogues , to show
how likely it is that false peaks arise from poisson fluctuations .
we can see that this is @xmath46 times less likely , showing that we are truly extracting peak information from the mock catalogues .
the one - to - one test carried out in this section is meant to be illustrative of how well peaks can be identified in galaxy catalogues .
the information presented in figures [ matching_sim_mock ] and [ random_match ] is not used in our measurement of the peak density from observations .
the completeness corrections used for this purpose are described in the next section . in order to compare an observed peak density with theoretical models , we need to statistically correct for the effects of shot noise and of survey boundaries .
our one - to - one comparison above does not show the aggregate effect on the peak density of these effects , but only that approximately 50% of peaks are found within 1 filter radius of their true locations .
if we make the assumption that the 2df observations are affected in the same way by boundaries and shot noise as our mock catalogues , we can use our mock catalogues to compute an effective completeness ( which varies spatially ) .
this will enable us to translate an observed peak density into the density we would have measured with fully information and denser sampling .
our calculations of the effective completeness involve first locating the local maxima in the simulation with full sampling and in the mock catalogues , as described above . by examining the peak density as a function of distance from the edge to the survey we have found that boundaries do have a substantial effect ( for example , in figure [ 2df ]
it can be seen that peaks occur preferentially at the edges ) . rather than applying a large correction for this effect , we decide to not consider the peaks in the volume near the boundaries .
we choose to remove from our number density computation all peaks and volume closer to the side edges of each bin within @xmath47 in azimuthal distance and @xmath48 in declination distance on each side .
our approach is to only include regions of the simulation for which the completeness correction is of order unity ( see below . )
we compare the number of peaks from fully sampled simulations and mock catalogues to yield the completeness . to do this we divide space into radial bins ( distance measured from the origin ) of width @xmath49 for filter radii less than @xmath50 mpc and of width of @xmath51 for larger filter radii .
for each radial bin @xmath52 , a completeness @xmath53 given by @xmath54 @xmath55 and @xmath56 are respectively the number of peaks in the @xmath57 bin measured from the simulation and the mock catalogues .
the completeness correction for a particular filter size , @xmath45 ( denoted by @xmath58 ) is found by averaging over completeness values obtained from 20 mock catalogues . from an observed number of peaks , the corrected number of peaks in the @xmath57 bin
, @xmath59 can therefore be computed from @xmath60 where @xmath61 is observed number of peaks in the @xmath57 radial bin for the same filter radius . for most of the mock survey volume , the completeness we which we compute in this way is @xmath62 , indicating that we are statistically finding all peaks , with no need for any correction .
the completeness correction therefore has a small effect , and there is no significant difference in our results if we do not include it .
the completeness correction does however enable us to include a larger volume of the survey than without it , resulting in smaller error bars .
the completeness correction @xmath58 rises above 1.0 for large distances from the origin , as the number density of galaxies goes down .
for example , with a filter scale @xmath63 between a radial distance of 300 @xmath18 ( where @xmath64 ) and the edge of the survey volume used ( at @xmath65 ) @xmath58 averages approximately @xmath66 . in figure [ mock_sgp_ngp ]
we have plotted peak densities for different @xmath45 values for both ngp and sgp mock catalogues .
we have also shown with lines the average for mock catalogues .
the averaged corrected peak densities ( multiplied by @xmath67 ) are shown for both the ngp and sgp mocks .
we find that for all @xmath45 values they agree very well with the peak densities from the fully sampled simulations . in figure [ number_counts ]
we have shown the number of galaxies in 30 radial bins for @xmath68 mpc .
we have already seen that in the very nearby universe it was necessary to superimpose two mock catalogues in order to achieve the required number density .
we use equation [ correct ] to calculate the peak density in the observed universe from the 2dfgrs data . to compute the correction factors , we use 20 mock catalogues which have a comparable number density of objects to that in the survey .
these mocks are used to calculate the error bar on each data point corresponding to each @xmath45 . in figure
[ obs_peaks ] we show the peak densities for the sgp and ngp .
we also indicate the best - fit theoretical curve which corresponds to a model with @xmath69 , @xmath70 , @xmath71 ev ( total mass of three massive neutrino species in ev ) .
we descibe how the model fitting was done in section 5 below . in figure [ weight_obs ]
we show the weighted average of the peak density measured from the sgp and ngp . for illustrative purposes
, we also show the peak densities predicted by various models .
the observational weighted average of ngp and sgp data was calculated using : @xmath72 where @xmath73 and @xmath74 are respectively the number density of peaks in the sgp region of the survey and the fractional error on that value .
.[pktab ] the space density of peaks measured from the 2df redshift survey data , as a function of filter scale @xmath45 [ cols="^,^,^ " , ] in figure [ mcmc1d ] we show the 1 dimensional marginalized probability distributions for the 3 parameters @xmath1 , @xmath75 , @xmath3 .
the width of the likelihood distributions for @xmath1 , @xmath76 are large , indicating that the peak density constrainty will not be competitive with other measures of these parameters .
the @xmath3 probability distributions are relatively narrow , however .
the best fit values of the parameters in the marginalized distributions and their range of uncertainty are give in table [ params ] .
we can see that @xmath1 is consistent within the large 1 @xmath15 error bars with the wmap5 results ( dunkley 2009 ) , @xmath77 .
the 2 @xmath15 upper limit on the total mass of 3 neutrino species , @xmath3 is 2.48 ev .
this is in the same range as upper limits from a variety of other cosmological probes ( see elgaroy 2007 for a recent review . )
by measuring the space density of peaks in the smoothed 2df galaxy survey density field as a function of filter scale , we have constrained the values of the cosmological power spectrum parameters d@xmath1/dln@xmath2 , @xmath1 and @xmath78 .
we find values that are consistent within 1 @xmath15 of those measured from other cosmological observables ( e.g. , dunkley 2009 , seljak 2005 ) , and are in support of the standard @xmath79cdm scenario . in linear theory ,
the peak density is not affected by the amplitude of fluctuations . as simulation tests have shown
that this is also true in the non - linear regime , for filter scales as small as @xmath80 , this means that our constraints on cosmological parameters come entirely from the shape of the power spectrum . because of degeneracies in the shape of the power specrum with cosmological parameters it was necessary to assume prior values for some parameters ( @xmath81 and @xmath82 ) from the wmap ( spergel 2003 ) results in order to derive our own limits .
various sources of systematic errors are possible , when measuring peak statisics , for example those related to the fact that the peak density in a smoothed field is sensitive to boundary effects , which are difficult to correct for ( unlike the case of the correlation function , for example . ) in the present work , we have made conservative cuts , ignoring space close to the survey boundaries and validated these using mock catalogs to test both the overall recovery of the peak density and the recovery of individual peaks .
an additional source of error comes from the fact that redshift distortions tend to merge peaks together which are separate in real space , as was seen in paper i. we have dealt with this effect , by including a power spectrum suppression term due to the small scale random velocity dispersion ( see paper i for details and tests ) .
this is the one part of the analysis for which peak finding may be most sensitive to the relationship between mass and light .
our tests using the halo occupation distribution in paper i have show that this is unlikely to be a problem at the level of current observational uncertainties .
our constraints have come from an analysis of filter lengthscales between @xmath83 .
this approach is therefore complementary to much work in large scale structure which has been done by analyzing the galaxy power spectrum on somewhat larger length scales ( wavenumber @xmath84 , corresponding to wavelengths @xmath85 ) . on these larger scales ,
galaxy fluctuations are assumed to be linearly related to mass , and direct measurements of the galaxy power spectrum are used to constrain the mass power spectrum and hence cosmological parameters ( e.g. , tegmark 2004 ) .
recently , it was shown by sanchez & cole ( 2007 ) that shapes of the power spectra measured from the 2df and sdss galaxy surveys ( e.g. , cole 2005 , tegmark 2004 ) are not in agreement , due to the @xmath86-band selected sdss galaxies having a stronger scale - dependent bias .
this results in differences in the cosmological parameters inferred from the two surveys which are larger than the quoted measurement uncertainties . from the tests in cg97 and paper i
, we expect that our peak based measurement to be more robust both to differences in galaxy bias and non - linear evolution . in the future it will be useful to compute the peak number density from the sdss survey data , in order to check consistency
. the larger size of the sdss final dataset will enable the use of larger filter scales and reduce the error bars on parameters .
sd was supported in part by nsf grant ast-0707704 and department of energy award number de - fg02 - 07er41517 .
we thank the referee for useful comments which improved the paper .
bardeen j. , bond j.r .
, kaiser n. , szalay a.s . , 1986 , , 304 , 15 berlind , a. , & weinberg , d.h .
, 2002 , , 575 , 587 bond j.r .
, efstathiou g. , 1984 , , 285 , l45-l48 cole et al . , 2005 , mnras , 362 , 505c colless et al .
, 2001 , mnras , 328 , 1039c croft r.a.c .
, gaztaaga e. , 1998 , , 495 , 554c croft r.a .
c. , weinberg d.h . ,
bolte m. , burles s. , hernquist l. , katz n. , kirkman d. , tytler d. , 2002 , , 581 , 20c de , s. and croft r.a.c . , mnras , 2007 , 382 , 1591 dunkley , j. el al . , 2009 ,
apjs , 180,306 eisenstein d.j . , hu w. , 1999 , , 511 , 5 elgaroy , o. , 2007 , nuclear physics b , 168 , 51 erdodu et al . , 2004 , mnras , 352 , 939 kosowsky a. , turner m. , 1995 , phrvd , 52 , 1739 maddox , s. j. , efstathiou , g. & sutherland , w. j. , , 1996 , 283 , 1227 matsubara t. , szalay a.s . , pope a.c . , 2004 , , 606 , 1 norberg et al .
, 2001 , mnras , 328 , 64 peacock et al . , 2001 , nature , 410 , 169 sanchez , a. g. , & cole , s. , 2007 , submitted , arxiv:0708.1517 schechter , p. , 1976
, , 203 , 297 seljak , u. , et al . , 2005 , phys .
d , 71 , 103515 2005 , , 203 , 297 spergel , d. , et al . , 2003 , , 148 , 175s tegmark , m. , et al .
, 2004 , phys rev d. , 69 , 3501 verde et al . , 2003 , , 148 , 195v zheng et al . , 2005 , , 633 , 791z
in order to incorporate observational constraints into a simulation we need to first compute the the selection function .
we choose our luminosity function to have the form of a schechter ( 1976 ) function ( following erdodu 2004 ) and define where @xmath88 is the radial selection function at a distance @xmath86 which is obtained by integrating over the luminosity function @xmath89 .
@xmath90 is the minimum luminosity which could be observed at a distance @xmath86 .
@xmath91 is the lowest measured luminosity in the survey , which is taken as the luminosity @xmath90 at the comoving distance @xmath86 of @xmath92 mpc , equivalent to a lower redshift cutoff for the data sample we use . to choose magnitudes for the galaxies in our mock catalogues
, we numerically solve equation @xmath93 for @xmath94 using a random value of the selection function between 0 to 1 . using this value of the luminosity we estimate the apparent magnitude at point ( @xmath86,@xmath95,@xmath96 ) , @xmath95 being the azimuthal angle , @xmath96 being the declination and @xmath86 being the comoving radial distance .
there is a dependence on angular coordinates that comes from the fact that @xmath91 is estimated as the minimum luminosity at a fixed radial distance but its value varies with @xmath95 and @xmath96 .
once we have found the minimum luminosity at @xmath86 from the last integral equation , we use following equation to compute the apparent magnitude ( @xmath97 ) : where @xmath99 is the luminosity distance , @xmath100 is the redshift , and @xmath101 , equivalent to the absolute @xmath102 limit .
the extra factor of @xmath103 is due to the inclusion of the @xmath2 correction .
the mock catalogues are made by applying the mask and luminosity function mentioned above to the outputs of n - body simulations . to generate the simulation we use the same cosmology as described in paper i , an lcdm running spectral model with spectral index @xmath11 , @xmath104 , @xmath16 , @xmath105 , and without massive neutrinos .
we use the transfer function described in eisenstein and hu ( 1999 ) .
these simulations are identical to those used in paper i with respect to box size ( @xmath106 ) , mass fluctuation amplitude , and other relevant parameters ( particle number @xmath107 ) . in paper
i , we tested the effect on the space density of peaks of using a galaxy halo occupation distribution ( hod , e.g. , berlind & weinberg 2002 ) to create the galaxy density field rather than particles .
we found no measurable difference for filter scales @xmath108 and therefore in the present paper , in the interests of having the maximal space density of objects , we use particles as proxies for galaxies in our mock catalogues . in order to produce mock catalogues ,
we first apply the geometrical constraints ( angular coverage ) of the survey to our simulations .
there are galaxies in the 2df survey at redshifts as deep as @xmath109 . to cover the whole of this observed volume it is necessary to replicate the particle distribution of the simulation ( box size of @xmath110 mpc ) so that it can cover a sphere of radius @xmath111 mpc .
to do this , we translate all particles in the same direction with a comoving distance equal to the box size ( the simulations have periodic boundary conditions ) , repeating this in three mutually orthogonal directions to yield enough volume to inscribe the sphere in it .
we note that although the box has been replicated several times , the actual volume of a single box is similar to that of the survey , because of the wedge like geometry of the survey .
having carried out this replication , we compute the comoving distance from the observer for our lcdm cosmology , and compute the redshifts of all the particles , including the effect of peculiar velocities .
after we select the simulation particles which have the same angular coverage as in the real survey , we reject those which have a sector completeness , @xmath40 less than @xmath112 .
we assign random values to the selection function to calculate @xmath94 as described above .
this equation is then solved to obtain apparent magnitudes for each galaxy .
we apply magnitude limits ( described below ) to exclude faint objects from the mock catalogues .
one important point to note is that in the survey there is also a bright magnitude limit .
it is necessary to ignore very bright galaxies as they can affect observations of neighboring galaxies .
the presence of holes in the angular mask to exclude them also results in holes in our mock catalogues .
our chosen values for the bright and faint magnitude limits are @xmath113 and @xmath114 respectively .
galaxies falling outside this range are ignored . using the computed apparent magnitude we calculate the magnitude dependent incompleteness @xmath115 , which is given by where @xmath117 and @xmath41 ( different for each mask cell ) are parameters which were set by colless ( 2001 ) by comparing to a simple power law model for galaxy number counts ( see equations @xmath118 of colless 2001 . )
these values of @xmath41 were obtained from masks made publically available by the 2dfgrs collaboration for each mock galaxy we compare this completeness to a randomly generated number between 0 and 1 , rejecting galaxies when the number is greater than the completeness .
after this , the final mock catalogue is output . to be able to average over random fluctuations we make many catalogues from 5 simulations generated with different random seeds as well as by randomizing the observer positions in the simulations .
we use 4 different observer positions per simulation , making 20 different mock catalogues .
one problem we encounter when making the mocks is that for small distances from the observer the number density of particles in the simulations is less than the number density of galaxies in the survey . for this region , we clone particles to make up the difference .
our results are insensitive to whether this is done , as we restrict our analysis to smoothing filter radii much larger than the particle mean separation at this distance .
we describe this procedure in more detail below . before determining the peak density from the catalogues
it is necessary to choose the region which can be analysed for each given filter size .
this is because the space density of galaxies decreases with distance from the origin .
if the space density of galaxies is too low for a given filter size , the smoothed density field from the catalogue will not give a correct representation of the smoothed underlying density field .
in paper i , it was shown that the smoothing filter scale should be greater than the mean separation between galaxies to avoid serious underestimation of the peak density .
additionally , we have a finite sized ( @xmath43 cells ) grid which we use to locate peaks ( local maxima on the grid ) .
we make sure at all times that the grid cell size is also less than the mean separation between galaxies , in order to avoid missing peaks . if _
w _ is the cell width and @xmath119 is the mean intergalaxy distance , then we ensure that @xmath120 .
we have checked to see that this criterion is sufficient , and our results are not sensitive to this exact choice . for a given value of @xmath121
, we construct a volume limited catalogue , and compute the mean separation between galaxies . in figure [ mean_sep ]
we show the dependence of this average inter - galaxy separation @xmath119 ( @xmath122^{\frac{1}{3}}$ ] , where _ v _ is the volume in a radial bin centered on @xmath44 and @xmath123 is the number of galaxies ) with distance from the origin @xmath44 in both mock catalogues and the actual 2df survey .
as mentioned above , for each filter radius @xmath45 we can safely increase @xmath44 to the maximum value where @xmath124 . in order to determine this value , we averaged over the mock catalogues and fitted a fourth order polynomial in @xmath125 to yield a fitting function for the radial distance,@xmath126 . | we use the number density of peaks in the smoothed cosmological density field taken from the 2df galaxy redshift survey to constrain parameters related to the power spectrum of mass fluctuations , @xmath0 ( the spectral index ) , d@xmath1/dln@xmath2 ( rolling in the spectral index ) , and the neutrino mass , @xmath3 . in a companion paper
we use n - body simulations to study how the peak density responds to changes in the power spectrum , the presence of redshift distortions and the relationship between galaxies and dark matter halos . in the present paper
we make measurements of the peak density from 2df galaxy redshift survey data , for a range of smoothing filter scales from @xmath4 .
we use these measurements to constrain the cosmological parameters , finding @xmath5 , @xmath6ev , @xmath7n@xmath8k@xmath9 , at the 68 % confidence level , where @xmath3 is the total mass of three massive neutrinos . at 95% confidence
we find @xmath10 ev .
these measurements represent an alternative way to constrain cosmological parameters to the usual direct fits to the galaxy power spectrum , and are expected to be relatively insensitive to non - linear clustering evolution and galaxy biasing .
[ firstpage ] |
the scheme for quantum information processing described in the main text exploits a cooper - pair box ( cpb ) [ ] , whose hamiltonian can be written in the basis of the charge states @xmath124 ( with @xmath125 ) [ ] as @xmath126 where @xmath127 is the josephson energy , @xmath128 is the charging energy and @xmath129 is the dimensionless gate charge .
the numerical diagonalization of ( [ cpb2 ] ) with @xmath130 , @xmath131 and @xmath132 [ ] yields a highly anharmonic spectrum .
hence , we can safely truncate the hilbert space of the cpb to the lowest three levels shown in the central part of fig .
2(b ) of the main text . in this regime coherence times of the cpb
are considerably longer than the cz gating time [ ] and the spectrum is sufficiently anharmonic to obtain high fidelities in our simulations .
we describe here a simpler setup , which could be exploited for the first proof - of - principle experiments .
this is formed by a cpb and two distinguishable spin @xmath17 ensembles @xmath20 and @xmath23 within the same cavity .
we exploit the lowest three levels of the cpb : @xmath35 , @xmath133 and @xmath134 , with transition energies @xmath27 and @xmath28 ( @xmath38 is assumed ) .
the excitation energies of the spin ensembles are @xmath18 and @xmath21 .
two different harmonics of the resonator are taken into account , of frequency @xmath19 and @xmath22 .
@xmath19 is intermediate between @xmath18 and @xmath27 , while @xmath22 is intermediate between @xmath21 and @xmath28 . in this way
the cavity mode @xmath19 ( @xmath22 ) can be coupled both to the spin gap @xmath18 ( @xmath21 ) and to the superconducting gap @xmath27 ( @xmath28 ) , by tuning its frequency .
we recall here the definition of the logical 2-qubit states : @xmath135 we note that the cpb does not enter in the definition of the qubits : in the idle configuration it remains in its ground state @xmath35 .
single - qubit rotations are implemented in a way similar to that described in the text : rotations about the @xmath61 axis of the bloch sphere , @xmath59 , are realized by means of off - resonant pulses , while @xmath68 are performed by tuning @xmath136 to @xmath29 for the proper amount of time ( @xmath137 ) . by exploiting the hybrid encoding ( [ logical ] ) the cz gate
can be implemented as in the scalable setup . here ,
however , a single resonator is considered and thus photon - hopping does not occur . as stated in the text , cz is performed with a two - step rabi oscillation of the cpb between @xmath35 and @xmath134 accompanied by the absorption and emission of the two photons entering the definition of the two qubits .
this is done through the intermediate states : @xmath138 during the simulation of a cz in the simpler setup described for the first proof - of - principle experiments.,width=302 ] we now examine the effect of the three - step pulse sequence used to implement cz on the four logical states ( [ logical ] ) .
* let s start by considering the two qubits initially prepared in state @xmath82 .
first step : @xmath19 is tuned to @xmath27 by means of a @xmath83-pulse , which transfers the excitation from the photon @xmath20 to the intermediate level @xmath133 of the cpb , carrying the system to state @xmath139 .
second step : a @xmath140-pulse brings @xmath22 into resonance with @xmath28 , thus inducing a full rabi oscillation between states @xmath139 and @xmath141 .
third step : the repetition of the first pulse brings back the system to @xmath82 , with a phase @xmath83 acquired during the rabi flop in the second step .
the extra phase due to the absorption / emission processes occurring in the first and third steps can be set to zero by properly choosing the delay between the two corresponding @xmath83-pulses . *
if the system is initially in @xmath99 there is only the photon of frequency @xmath19 and the @xmath140-pulse ( second step ) leaves the state @xmath139 unaffected .
the two @xmath83-pulses ( first and third steps ) brings the state from @xmath99 to @xmath139 and back , but the phase acquired during this process is set to zero by the same delay fixed previously .
* if the system is in @xmath142 , the two @xmath83-pulses do not induce any transition .
in fact , in this state there is only a photon in mode @xmath23 which is never brought into resonance with the transition @xmath27 .
thus , also the rabi swap ( second step ) is ineffective , as the cpb is always in its ground state . *
finally , state @xmath100 is completely unaffected by the present sequence .
hence , the sequence of pulses described above implements a cz , i.e. , in matrix representation on basis ( [ logical ] ) : @xmath143 as photon - hopping is not required , the gating times are reduced with respect to the scalable setup , to about @xmath144 ( see the simulation reported in fig . [ proof ] ) .
following [ ] , we include the photon leakage by modeling the evolution of the system as due to an effective hamiltonian @xmath145 involving the complex cavity frequencies @xmath146 . with this substitution the photon part of the hamiltonian changes into @xmath147 \hat{a}_\gamma^{\dagger } \hat{a}_\gamma .\ ] ] here @xmath148 is the cavity loss rate , assumed in the 10 khz range [ ] , and @xmath136 are the real frequencies introduced in the text . fig . [ leakage ]
shows the absolute squared value of the components of the system wave - function as a function of time , during the simulation of a @xmath149 and of a cz gate .
no significant difference can be noted with respect to fig .
3(b ) of the text : the non - hermitian evolution due to @xmath145 leads only to few photons lost over thousand during our gating times . finally , it has been experimentally shown that pure dephasing of the cavity modes is negligible ( see , e.g. , [ ] where the measured value of the dephasing time @xmath150 approximately corresponds to twice the value of the photon decay time @xmath151 ) .
in the simulations reported in the text the cavity modes are modulated by means of step - like pulses .
we have checked that the same results can be achieved by linearly varying the resonator frequency on a timescale @xmath152 : this regime is well within the validity of the rotating - wave approximation and no generation of unwanted photons due to dynamical - casimir effect is observed [ ] .
a. wallraff , d.i .
schuster , a. blais , l. frunzio , r. s. huang , j. majer , s. kumar , s.m .
girvin , and r.j .
schoelkopf , nature ( london ) * 431 * , 162 ( 2004 ) ; h. paik , d. i. schuster , lev s. bishop , g. kirchmair , g. catelani , a. p. sears , b. r. johnson , m. j. reagor , l. frunzio , l. i. glazman , s. m. girvin , m. h. devoret , and r. j. schoelkopf , phys .
lett . * 107 * , 240501 ( 2011 ) .
j. majer , j.m .
chow , j.m .
gambetta , jens koch , b.r .
johnson , j.a .
schreier , l. frunzio , d.i .
schuster , a.a .
houck , a. wallraff , a. blais , m.h .
devoret , s.m .
girvin , and r.j .
schoelkopf , nature ( london ) * 449 * , 443 ( 2007 ) .
y. kubo , f.r .
ong , p. bertet , d. vion , v. jacques , d. zheng , a. drau , j .- f .
roch , a. auffeves , f. jelezko , j. wrachtrup , m.f .
barthe , p. bergonzo , and d. esteve , phys . rev .
lett . * 105 * , 140502 ( 2010 ) .
you , f. nori , phys . rev .
b * 68 * , 064509 ( 2003 ) ; a. blais , j. gambetta , a. wallraff , d.i .
schuster , s.m .
girvin , m.h .
devoret , and r.j .
schoelkopf , phys .
a * 75 * , 032329 ( 2007 ) .
f. w. strauch , p. r. johnson , a.j .
dragt , c.j .
lobb , j. r. anderson , f. c. wellstood , phy .
lett . * 91 * , 167005 ( 2003 ) .
a. palacios - laloy , f. nguyen , f. mallet , p. bertet , d. vion , d. esteve , j. low temp . phys . * 151 * , 1034 ( 2008 ) ; m. sandberg , f. persson , i.c .
hoi , c.m .
wilson , and p. delsing , phys ..
, * t137 * , 014018 ( 2009 ) .
houck , h.e .
tureci , and j. koch , nat .
physics * 8 * , 292 ( 2012 ) .
m. mariantoni , h. wang , t. yamamoto , m. neeley , r. c. bialczak , y. chen , m. lenander , e. lucero , a. d. oconnell , d. sank , m. weides , j. wenner , y. yin , j. zhao , a. n. korotkov , a. n. cleland , john m. martinis , science , * 334 * , 61 ( 2011 ) .
additional trivial phases can be straightforwardly eliminated by short @xmath153 operations as in the implementation of @xmath154 . in this way also the phase acquired during photon - hopping processes can be zeroed . by properly setting the delay between the two @xmath83-pulses corresponding to @xmath89 , the associated absorption / emission processes
yields a zero additional phase .
a. megrant , c. neill , r. barends , b. chiaro , y. chen , l. feigl , j. kelly , e. lucero , m. mariantoni , p.j.j .
omalley , d. sank , a. vainsencher , j. wenner , t. c. white , y. yin , j. zhao , c. j. palmstrom , j.m .
martinis , and a.n .
cleland , appl .
. lett . * 100 * , 113510 ( 2012 ) . | we introduce a scheme to perform quantum - information processing that is based on a hybrid spin - photon qubit encoding .
the proposed qubits consist of spin - ensembles coherently coupled to microwave photons in coplanar waveguide resonators .
the quantum gates are performed solely by shifting the resonance frequencies of the resonators on a ns timescale .
an additional cavity containing a cooper - pair box is exploited as an auxiliary degree of freedom to implement two - qubit gates .
the generality of the scheme allows its potential implementation with a wide class of spin systems .
3 cm a classical computer is made of a variety of physical components specialized for different tasks . in the same way
, a quantum computer will probably be a hybrid device exploiting the best characteristics of distinct physical systems . in this spirit
, much work has recently been done to achieve strong coupling of high - quality factor coplanar - waveguide resonators with superconducting qubits , such as cooper - pair boxes ( cpbs ) [ ] and transmons [ ] and/or spin ensembles ( ses ) [ ] .
superconducting qubits coupled to a microwave cavity field were proposed for quantum information processing ( qip ) [ ] , using classical fields [ ] or external voltages [ ] as a manipulation tool . during the last years several theoretical works have considered the possibility of joining the fast processing of superconducting qubits to the long coherence times of ses [ ] , which can be naturally exploited as quantum memories .
cavity photons can be used as a bus to transfer the quantum state from cpbs to spin ensembles , and to couple distant cpb qubits , leading to an effective interaction necessary to perform two - qubit gates [ ] .
recently , it was theoretically shown that a minimal architecture solely based on ses can be exploited for full qip [ ] , by employing a measurement - based scheme in which photons are still used as a quantum bus .
+ here we introduce a qualitatively different approach based on a hybrid spin - photon encoding of the qubits .
our scheme differs from previous ones because here the spin and photon degrees of freedom enter on an equal footing in the definition of the qubit .
this allows us to to perform all the manipulations simply by tuning the resonance frequencies of microstrip - line superconducting resonators ( fig . 1 ) .
experimentally , the resonator frequencies have already been shown to be variable on a ns timescale [ ] , and up to tenths of the fundamental - mode frequency @xmath0 [ ] .
besides , they can be assembled in a series of lumped elements , realizing large arrays of different geometries [ ] . such capability ,
along with the possibility of individually addressing the resonators , represents a prerequisite for scalability . as to the spins ,
the few requisites of the scheme can be fulfilled by a large variety of systems , ranging from diluted transition - metal or rare - earth ions to molecular nanomagnets [ ] . in our hybrid encoding , each physical qubit is represented by a resonator mode and a se .
we describe the quantum gates in the elementary unit of a scalable setup , namely two resonators containing different ses , where each qubit is encoded in the state of a distinct se - resonator pair .
these are connected by an interposed cavity containing a cpb , which is exploited as an auxiliary degree of freedom to implement two - qubits gates .
+ _ definition of the qubit
_ we consider a resonant cavity containing a single photon in a mode of frequency @xmath1 , and an ensemble of @xmath2 identical and non - interacting spins @xmath3 , initially prepared in the ground state @xmath4 . if the resonator mode is tuned to the gap of the two - level system , @xmath5 , after some time the se will collectively absorb the photon and evolve into the state @xmath6 .
transitions between @xmath7 and @xmath8 are described by @xmath9 and @xmath10 [ ] . in the low - excitation regime , the collective excitations of the se can be described as a harmonic oscillator and @xmath11 = 1 $ ] .
+ within the single - excitation subspace of the system formed by a resonator mode and a se , we introduce the hybrid encoding of the qubit @xmath12 : @xmath13 where @xmath14 is the photon occupation number of the cavity mode coupled with the se .
thus , the logical state of the qubit depends on whether the excitation is stored within the se or the quantized field of the resonator .
+ _ description of the scalable setup _ to achieve universal qip it is sufficient to perform a two - qubit quantum gate such as the controlled@xmath15 ( cz ) , and arbitrary single - qubit rotations around two non - parallel axes [ ] . hereafter , we describe our scheme for the quantum - gate implementation in the basic unit of a scalable setup , i.e. , a system of two qubits ( @xmath16 ) , encoded in the hybrid states of two distinguishable ses coupled to the modes of two different stripline resonators , as in fig .
2(a ) . in its general form
, the scheme can be implemented within any bipartite lattice of cavities [ ] . both ses consist of effective @xmath17 spins , but with different energy gaps : @xmath18 is coupled to the harmonic of frequency @xmath19 in cavity @xmath20 , while @xmath21 to the harmonic of frequency @xmath22 in cavity @xmath23 , as in the level scheme of fig .
2(b ) .
a third resonator @xmath24 , which is not used to encode qubits , is located in between the cavities @xmath20 and @xmath23 .
it contains a cpb which we exploit to implement cz .
as we show in appendix a , for reasonable values of the josephson and charge energies [ ] , the cpb is characterized by the anharmonic spectrum reported in the central part of fig .
2(b ) ( hereafter we consider only the three lowest levels ) .
in cavity b we consider two different harmonics , @xmath25 and @xmath26 , respectively close to the gaps @xmath27 and @xmath28 of the cpb .
, corresponding to transition frequency @xmath29 between single - spin states @xmath30 and @xmath31 ( @xmath32 ) .
the cpb - cavity couplings are indicated with capital letters .
dashed lines represent the frequencies of the cavity modes.,width=313 ] in the idle configuration , the cavity modes @xmath33 and @xmath34 are detuned and qubits encoded in different cavities evolve independently from one another .
hence , single qubit rotations can be implemented by varying the resonance frequency of the relevant resonator , while the cpb is left in its ground state and no photons are present in cavity @xmath24 .
the interaction between neighboring qubits is switched on when the two cavities @xmath20 and @xmath23 are brought into resonance with @xmath24 and photons can jump from cavities @xmath20 and @xmath23 to @xmath24 .
as explained below , cz gates can be obtained by exploiting a two - step rabi oscillation of the cpb between its ground state @xmath35 and the excited state @xmath36 ( fig .
2(b ) ) .
a three - level superconducting system has been exploited for performing cz gates in a different scheme in [ ] .
+ the total hamiltonian of the ses+cpb coupled to the quantized cavity fields reads : @xmath37 the first term describes the ses , @xmath20 and @xmath23 , as independent harmonic oscillators [ ] ( @xmath38 ) : @xmath39 the time - dependent photonic hamiltonian , which in our scheme is entirely responsible for the qubits manipulation , is : @xmath40 where @xmath41 . within the rotating - wave approximation ,
the spin - photon and cpb - photon couplings read : @xmath42 the coupling constants @xmath43 are enhanced by a factor @xmath44 with respect to their single - spin counterpart @xmath45 ( see , e.g. , ref . ) , while @xmath46 are referred to a single superconducting unit .
finally , the last term in eq . [ hamil ] describes the photon hopping induced by the evanescent coupling between the modes @xmath33 and @xmath47 of the two neighboring cavities [ ] : @xmath48 other hopping terms can be easily made negligible by engineering the two different cavities .
hereafter , we will use the interaction picture , with @xmath49 . + _
quantum gates _ in order to perform one- and two - qubit gates , we exploit the absorption ( emission ) of the photons entering the hybrid encoding ( eq . [ encoding ] ) .
these processes can be straightforwardly controlled by tuning the frequencies of the cavity modes by a quantity @xmath50 for suitable time intervals .
we will refer to such variations of the resonator frequencies as _ pulses_. in order to make the manipulation experimentally easier , we choose ( see fig .
2(b ) ) @xmath51 to be intermediate between @xmath52 and @xmath27 , while @xmath53 is close to @xmath54 and @xmath54 is close to @xmath28 . in the idle configuration @xmath55
, the cavity - mode frequencies are significantly detuned from the spin energy gaps ( @xmath56 ) and @xmath57 is ineffective .
in addition , the cavities @xmath20 and @xmath23 are far - detuned from @xmath24 ( @xmath58 ) and the effect of inter - cavity coupling is negligible . hence , single - qubit gates can be performed independently on each cavity .
+ in particular , a rotation @xmath59 of an arbitrary angle @xmath60 about the @xmath61 axis of the bloch sphere is performed by off - resonance pulses .
these induce a phase difference between the @xmath62 and @xmath63 states of the hybrid qubits ( eq . [ encoding ] ) , given by @xmath64 ( @xmath65 is the pulse duration ) . for simplicity
we assume step - like pulses , @xmath66 , so that a generic @xmath59 rotation is obtained by setting @xmath67 .
conversely , @xmath68 rotations are obtained by tuning the frequency of the cavity mode to match the corresponding energy gap of the se ( @xmath69 , with @xmath70 ) for the time @xmath71 .
the additional phase @xmath72 of the @xmath73 qubit state is straightforwardly corrected by an @xmath74 rotation . a simulation of this gate is reported in fig .
3(a ) in terms of the overlaps @xmath75 between the system state @xmath76 and the logical two - qubits states @xmath77 ( @xmath78 ) . + the implementation of two - qubit gates requires the coupling between the degrees of freedom that are used to encode the two qubits .
the cz gate is performed with a two - step rabi oscillation of the cpb between @xmath79 and @xmath80 accompanied by the absorption and emission of the two photons entering the definition of the two qubits . a multi - step pulse sequence involving the auxiliary states
@xmath81 is adopted .
we illustrate it considering the two qubits initially in @xmath82 .
the first step corresponds to the transfer of the photon of mode @xmath20 in cavity @xmath20 into mode @xmath24 in cavity @xmath24 , by means of a @xmath83-pulse that brings the two modes into resonance , @xmath84 .
simultaneously , the photon of mode @xmath23 is transferred into mode @xmath85 by varying @xmath86 .
this induces the transition @xmath87 . as a second step ,
the photon of frequency @xmath88 is absorbed by the cpb , after a @xmath83-pulse corresponding to @xmath89 , bringing the system into state @xmath90 .
then , a @xmath91-pulse corresponding to @xmath92 brings the mode @xmath93 into resonance with the @xmath94 transition of the cpb , thus inducing a complete rabi flopping between the states @xmath95 and @xmath96 . as a result
, a phase @xmath83 is added to @xmath95 .
finally , the repetition of the first two steps brings the state back to @xmath97 with an overall phase @xmath83 [ ] .
the time evolution of the two - qubit state @xmath76 induced by this pulse sequence is reported in fig .
3(b ) .
conversely , the other basis states do not acquire any phase and therefore this sequence implements the cz gate .
indeed , the basis states @xmath98 and @xmath99 contain only one photon and thus the full two - step rabi oscillation can not occur ( see appendix b ) , while the state @xmath100 is completely unaffected .
+ in a single - qubit rotation @xmath101 . dashed black lines represent @xmath102 .
negligible components are not shown .
( b ) time - dependence of the main components of @xmath103 in a cz gate .
auxiliary states @xmath104 , @xmath105 and @xmath106 are described in eq .
[ aux].,width=313 ] these simulations have been performed by assuming fundamental frequencies @xmath107 , @xmath108 and @xmath109 ghz [ ] , and using the first and second harmonic ( second and third harmonics ) for the @xmath20 and @xmath23 ( @xmath24 ) cavities . with this choice , photon hopping between modes other than those included in eq
. [ hopping ] is negligible . as a figure of merit for cz , we compute the _ fidelity - loss _ @xmath110 corresponding to the basis states @xmath111 : @xmath112 where @xmath113 is the final system state , when the qubits are initialized in @xmath114 .
we found @xmath115 for cz and less for the other gates demonstrating the accuracy of the operations .
analogous values of @xmath110 are obtained for initial states corresponding to linear superpositions of the basis states .
smaller values of @xmath110 can be obtained by using larger frequencies or larger detunings .
+ the frequency is varied up to @xmath116 for the fundamental harmonic , and proportionally for the others .
similar detunings have been experimentally shown in ref . .
we have assumed realistic values of the cpb - cavity @xmath117mhz and se - cavity coupling rates @xmath118mhz , corresponding to @xmath119 spins [ ] , and tunneling rate @xmath120mhz , which has already been shown experimentally [ ] .
larger values of @xmath121 and @xmath122 would reduce the gating times , but would also increase the fidelity - loss @xmath110 , unless compensated by larger detunings .
+ differently from alternative approaches ( see e.g. , [ ] ) , the main source of decoherence within our hybrid qubit encoding is determined by photon leakage due to cavity modes losses , since spin coherence times can be very long ( see below ) .
we have checked that cavity loss rates in the 10 khz range [ ] , i.e. photon lifetime of tens of microseconds , result only in few photons lost over thousand within the gates timescale .
we report full numerical simulations including these effects in appendix c , confirming the robustness of our scheme .
we notice that pure dephasing of cavity modes is negligible in such resonators [ ] . as far as the cpbs are concerned , in the present scheme they are always in the ground state , except during the implementation of cz gates . using a cpb as a non - linear element instead of a transmon or a phase qubit increases the anharmonicity of the system , while reducing the coherence times [ ] .
however , it is sufficient that the coherence time of the cpb is much longer than the cz gating time .
this can be easily achieved with present technology [ ] . at last
, the tuning of the resonator frequency could introduce extra dissipation to the system .
these effects have been analyzed experimentally ( see , e.g. , ref .
[ ] ) and found to be small .
they are mainly attributed to thermal noise and they could be further reduced with high frequency resonators , as the ones assumed in this work , and working at very low temperatures where q - factors are known to be larger and more stable .
+ _ physical implementation _ initialization and read - out can be achieved by exploiting the cpbs [ ] .
several classes of systems can be exploited as ses for the implementation of our scheme .
a simple possibility would be to exploit the different gyromagnetic factors of 3d and 4f ions , in order to obtain suitable level schemes in an applied magnetic field .
the embedding of these ions in a non - magnetic crystalline matrix allows to reduce harmful dipolar interactions , and to enhance the decoherence time up to the ms range [ ] .
another promising class of spin systems is represented by molecular nanomagnets ( mnms ) [ ] .
these spin clusters form crystals in which the molecules behave as identical and non - interacting magnetic units , and can be diluted by using non - magnetic analogues . besides
, mnms present degrees of freedom , such as spin chirality , that can be manipulated by the electric - field component of the cavity field [ ] and are substantially protected from the magnetic environment [ ] .
+ in summary , we have developed a scheme for quantum - information processing with spin ensembles in superconducting stripline resonators , exploiting a hybrid spin - photon encoding of the qubits .
our scheme is qualitatively different from previous ones because the spin and photon degrees of freedom enter the definition of the qubit on an equal footing . in this way
, the evolution can be induced simply by tuning the cavity frequency to the spin - energy gaps .
arbitrary single and two - qubit gates are implemented , over much shorter times than typical decoherence times of cavity - photons and spin - ensembles .
promising candidates for the spin degrees of freedom are diluted magnetic ions or molecular nanomagnets .
the application of this scheme to an @xmath123 array of cavities enables general quantum algorithms as well as quantum simulators [ ] to be implemented .
+ very useful discussions with a. auffves are gratefully acknowledged .
this work was financially supported by the italian firb project `` new challenges in molecular nanomagnetism : from spin dynamics to quantum - information processing '' and by fondazione cariparma . |
in recent years , eruptive objects with maximum luminosities intermediate between those of classical novae ( cne ) and supernovae ( sne ) have been discovered in increasing numbers , as surveys for galactic and extragalactic transients are made with greater depth and sky coverage by professional and amateur astronomers .
the term `` supernova impostors '' was introduced by van dyk et al.(2000 ) following the outburst of sn 1997bs .
sn 1997bs was classified as a sn iin event , `` n '' denoting narrow emission lines during outburst , in contrast to the high ejection velocities of typical sne .
however , it reached an absolute magnitude at maximum of only @xmath7 , more than 3 mag fainter than a typical core - collapse sn .
van dyk et al . argued that sn 1997bs was instead a `` superoutburst '' of a massive luminous blue variable ( lbv ) , analogous to those experienced by the galactic objects p cygni and @xmath8 carinae ( see humphreys , davidson , & smith 1999 ) .
the subject of intrinsically faint sne iin and @xmath8 car analogs has been summarized by van dyk ( 2005 ) and smith et al.(2008 ) .
more recently , sn 2008s , which had a sn iin - type spectrum but reached only @xmath9 , has been discussed by thompson et al .
( 2008 ) and smith et al .
prieto et al . (
2008 ) have identified the precursor of sn 2008s in pre - outburst mid - ir images obtained with the _
spitzer space telescope _ , demonstrating that the progenitor was a luminous star deeply embedded in circumstellar dust .
v838 monocerotis is the prototype of an apparently separate class of transients also lying in the cn - sn gap .
v838 mon erupted in 2002 , illuminating a spectacular light echo that provided a geometric distance ( sparks et al .
2008 ) implying an absolute magnitude at maximum of @xmath10 . unlike a cn ,
v838 mon became progressively redder during its outburst , eventually becoming enshrouded in circumstellar dust .
its outburst light curve ( bond et al .
2003 ) was very different from that of a sn or lbv eruption , showing a series of four maxima separated by about a month each .
the cause of this outburst remains uncertain , but a leading suggestion is that it was a stellar collision or merger ( e.g. , tylenda & soker 2006 , corradi & munari 2007 , and references therein ) .
possible extragalactic analogs of v838 mon
although this interpretation is controversial include the `` m31 red variable '' ( `` m31 rv '' ) which reached @xmath11 ( bond & siegel 2006 and references therein ) , and a luminous red transient in m85 ( rau et al . 2007 ;
kulkarni et al . 2007 ) , which attained @xmath12 .
m31 rv and the m85 transient appear to have occurred in old populations , suggesting that they did not arise from very massive progenitors . in this _
letter _ , we report observations of a luminous transient discovered in 2008 may in the nearby spiral galaxy ngc 300 .
we obtained extensive ground - based imaging , photometry , and spectroscopy , and we also investigated pre- and post - outburst space - based optical and infrared images of the transient .
this _ letter _ presents an overview of our results , along with a preliminary discussion of the nature of this remarkable object .
subsequent publications will give fuller details .
we designate the object as the `` ngc 300 optical transient , '' hereafter `` ngc 300 ot . ''
the outburst of ngc 300 ot was discovered by monard ( 2008 ) during his sn search program with the bronberg observatory 0.3-m telescope equipped with a ccd camera .
the ot was detected at broad - band magnitude 14.3 on a frame obtained on 2008 may 14 ( all dates in this paper are ut ) .
subsequent inspection of an image from 2008 april 24 , taken in morning twilight shortly after ngc 300 had emerged from behind the sun , showed that the ot was already rising , at @xmath216.3 mag .
it was fainter than 18 mag on 2008 february 8 , and on all previous bronberg observations .
ngc 300 is an sad spiral in the sculptor group , located just outside the local group .
the ot lies in a spiral arm with active star formation .
we adopt a distance modulus of @xmath13 ( @xmath14 mpc ) , based on cepheids ( gieren et al .
2005 ) and the red - giant tip ( rizzi et al . 2006 ) . at maximum
the ot was much fainter than a typical sn , but brighter than any cn . because of our interest in luminous red transients , we began a program of photometric and spectroscopic monitoring , using the 1.3- and 1.5-m telescopes at cerro tololo interamerican observatory operated by the smarts consortium .
the photometric monitoring in particular was intensive at first , because of our expectation that the ot might exhibit periodic spikes , like those during the outburst of v838 mon ( bond et al .
2003 ) .
we used the andicam optical / near - ir direct camera ( depoy et al .
2003 ) on the smarts 1.3-m with ccd and ir detectors , operated simultaneously , for the _ bvri _ and _ jhk _ photometry .
the _ bvri _
magnitudes of the ot were determined differentially with respect to a nearby comparison star , calibrated on photometric nights using landolt ( 1992 ) standard fields .
the _ jhk _
magnitudes are differential with respect to the same star , calibrated using its 2mass photometry . the j2000 position of the ot , based on astrometry of andicam frames calibrated against the usno - nomad catalog , is ra = 00:54:34.51 , dec = @xmath15:38:31.4 , with errors in each coordinate of about @xmath16@xmath17 .
figure 1 shows the _ bvrijhk _ light curve .
unfortunately , the rise to maximum was poorly covered , but it appears that it was more rapid than the subsequent slower decline .
the brightest @xmath5 magnitude we measured was 14.69 , on the first night of smarts observations ( 2008 may 15 ) , corresponding to @xmath18 to @xmath19 for the adopted distance and a reddening lying in the range @xmath20 to 0.4 ( see 3.1 ) .
following maximum light , the ot has declined smoothly in brightness at all wavelengths , while becoming steadily redder .
for example , @xmath21 evolved from 3.1 in mid - may to 7.4 at the beginning of 2008 december .
the rate of decline in the optical lessened around the beginning of 2008 september , but at this writing the brightness continues to decrease in both the optical and near - ir .
the @xmath22 magnitudes are not fading as rapidly as in the other filters , due to strong h@xmath23 emission .
we obtained low- and moderate - resolution optical spectroscopy of ngc 300 ot throughout its outburst , using the smarts 1.5-m cassegrain spectrograph .
our first low - resolution spectrum was observed on 2008 may 15 ( bond , walter , & velsquez 2008 ) and is shown in figure 2 ( top spectrum ; resolution 17.2 ) . this spectrum exhibits strong emission at h@xmath23 , h@xmath24 , the triplet at 854284988662 , and the unusual forbidden [ ] doublet at 72917323 .
the balmer lines are only slightly resolved at the velocity resolution ( @xmath25 ) of the spectrum .
the underlying continuum resembles a high - luminosity f - type supergiant , with h and k seen in absorption along with several weaker luminosity - sensitive lines such as the triplet at 7774 .
our photometry at maximum and shortly afterwards yields @xmath26 , implying that the inter- and circumstellar extinction could be as high as @xmath27 , since a mid- to late f - type supergiant normally has @xmath28 .
the mean foreground reddening of ngc 300 is @xmath29 ( gieren et al .
2005 ) .
this spectrum is clearly not that of a cn nor a typical sn . with strong and [ ] emission and an f - type absorption spectrum
, it resembles that of the type iin supernova impostor sn 2008s ( steele et al .
2008 ; smith et al .
it is also closely similar to the spectra of the well - studied post - red supergiant irc + 10420 ( jones et al .
1993 ; humphreys , davidson , & smith 2002 ) , variable a in m33 in its recent quiescent state ( humphreys et al . 2006 ) , and v838 mon near its maximum ( e.g. , wisniewski et al .
these f - type supergiant spectra with superposed narrow emission lines of relatively low excitation are indicative of eruptions that produce an optically thick wind or slowly expanding envelope .
as the ot declined , its spectrum became dominated by the h@xmath23 and emission .
one example , selected from our monitoring program , is the bottom spectrum shown in figure 2 , obtained on 2008 august 23 . by this date ,
the continuum had faded dramatically and reddened somewhat , but the strong emission remained .
we also obtained moderate - resolution ( 3.1 ) spectra regularly with the smarts 1.5-m spectrograph , and on three occasions with the magellan 6.5-m telescope and mage echellette spectrograph ( marshall et al . 2007 ;
resolution 1.5 at , 1.1 at h@xmath23 ) : 2008 july 6 , aug 30 , and sep 1 .
these spectra show that the triplet , h@xmath23 , and h@xmath24 have double - peaked emission lines , with the blue component being the stronger .
figure 3 shows the h@xmath23 and emission features from the magellan echellette spectra on 2008 july 6 ( top spectrum ) and the average of the august 30 and september 1 observations ( bottom spectrum ) , and illustrates the double structure of these lines .
double emission is usually attributed either to a bipolar outflow or a rotating disk , but in the case of an eruption , the double lines are most likely formed in a bipolar wind .
irc + 10420 shows very similar double - peaked emission and has strong independent evidence for a bipolar outflow ( davies et al .
2007 ; patel et al . 2008 ) .
the peak of the ot s primary blue component slowly increases in strength with time relative to the red component , supporting our conclusion that the outflow is bipolar , with the nearer , blue - shifted lobe expanding towards us .
the doppler velocities measured from both the hydrogen and emission lines in the echellette spectra indicate expansion velocities for the primary wind of @xmath270@xmath30 with respect to the star s systemic velocity of about @xmath31 .
there is evidence as well for additional emission components , appearing as shoulders or secondary bumps on the primary blue and red peaks at velocities of @xmath32@xmath33 .
a wind velocity of @xmath2@xmath34 is typical of the winds of f - type supergiants , and is somewhat lower than the winds associated with lbvs in eruption .
it is very similar to the expansion velocity of @xmath2@xmath35 of irc + 10420 ( jones et al .
1993 ; humphreys et al . 2002 ) .
by contrast , the [ ] lines , formed in a very low - density region , do not show a double - peaked structure ; they probably arise in more - distant and slower - moving material , likely formed before the current eruption .
we imaged the site of the ot twice with the _ hubble space telescope _ ( ) , in the director s discretionary ( dd ) program go-11553 ( pi : bond ) .
we used the wide field planetary camera 2 ( wfpc2 ) with the f450w and f814w filters ; observations were made on 2008 june 9 and september 1 . in addition , the archive contains two sets of deep observations made before the outburst , obtained with the wide field channel of the advanced camera for surveys ( acs ) and the f435w , f475w , f555w , f606w , and f814w filters .
the acs observations were made on 2002 december 25 ( go-9492 , pi : f. bresolin ) and 2006 november 810 ( go-10915 , pi : j. dalcanton ) .
we used these data to locate the ot precisely , search for a progenitor object , study the surrounding stellar population , and set limits on a light echo from the outburst .
in addition to the data , we were awarded dd time on the eso very large telescope ( vlt ) to obtain near - ir imaging of the ot , using the adaptive - optics naco camera ( lenzen et al .
2003 ; rousset et al .
the vlt / naco images were obtained on 2008 june 6 .
the wfpc2 exposure times for our dd observations ranged from 5 to 260 s , chosen so as not to saturate the image of the ot , but to show enough of the surrounding star field for precise registration with the pre - outburst acs frames .
the archival acs frames are considerably deeper , with total exposure times of 1080 , 2976 , 1080 , 3030 , and 4524 s in f435w , f475w , f555w , f606w , and f814w , respectively .
the vlt / naco @xmath36-band frames were four 600 s exposures , combined in the eso pipeline to produce a single image .
figure 4 shows a montage of post- and pre - outburst images .
the ot lies amid a rich field of resolved stars .
remarkably , however , there is no significant detection of a progenitor object at the ot s precise location .
the two nearest reliably detected sources lie about @xmath37 and @xmath38 from the site , well in excess of our registration error of about @xmath39 .
these two stars are red giants with f606w magnitudes of about 26.3 and 27.0 . in the deepest stacked pre - outburst images there are suggestions of local maxima at the ot site , but these are all consistent with noise . from the three deepest acs stacked images , we find the following limiting magnitudes ( vega - mag scale , 3@xmath40 upper limits ) for the ot progenitor : 28.3 ( f475w ) , 28.5 ( f606w ) , and 26.6 ( f814w ) .
similar limiting magnitudes were reported earlier , based on registering the 2006 acs images with ground - based magellan / clay 6.5-m images , by berger & soderberg ( 2008 ) , but with larger positional uncertainties .
the ot was thus very unlike a typical core - collapse sn , for which luminous progenitor stars are being found relatively easily in archival images at distances greater than that of ngc 300 ( e.g. , smartt et al .
2008 and references therein ) , nor was it similar to erupting lbvs , which even in quiescence have high optical luminosities ( e.g. , humphreys & davidson 1994 ) .
the acs images also provide deep cmds for the environment surrounding the ot .
these will be presented in detail in a subsequent publication , but we note that there is a rich population of young objects in the immediate vicinity of the ot .
for example , within @xmath41 ( @xmath223 pc ) of the ot site there are over a dozen blue main - sequence stars brighter than about f606w magnitude 26 , reaching as bright as @xmath42 ( corresponding to b1 stars of @xmath32@xmath43 ) .
the surrounding spiral - arm field is rich in blue clusters and associations .
we compared the post - outburst images of 2008 june 9 and september 1 in search of an expanding light echo , similar to the one surrounding v838 mon ( e.g. , bond 2007 , sparks et al .
2008 , and references therein ) .
although a light echo of the surface brightness and angular size of v838 mon s , moved to the distance of ngc 300 , would have been detectable at resolution , no echo is seen .
the very faint limits on the pre - outburst optical brightness of ngc 300 ot initially suggested to us that there was a similarity with v838 mon , whose progenitor was also optically inconspicuous ( e.g. , afar & bond 2007 , and references therein ) .
however , the very different light curve and spectroscopic behavior of the ot as its outburst proceeded casts doubt on this parallel .
moreover , a striking difference emerged when one of us ( prieto 2008 ) reported that the progenitor of the ot _ was _ detected as a bright mid - ir source in archival observations obtained with _ spitzer _ on 2003 november 21 ( pi : g. helou ) and 2007 december 28 ( pi : r. kennicutt ) .
the source was present in all four irac bands ( 3.6 , 4.5 , 5.8 , and 8 ) , and in the mips 24 band .
a black - body fit to the 3.624 fluxes gives the following results for the luminosity , effective temperature , and radius of the ot progenitor : @xmath44 , @xmath45 k , and @xmath46 au .
these are consistent with a heavily dust - enshrouded red supergiant of @xmath210@xmath6 ( e.g. , maeder & meynet 2000 ) .
there was no variability of the progenitor between 2003 and 2007 .
further detailed discussion of the _ spitzer _ observations is given by thompson et al.(2008 ) .
a picture thus emerges in which a luminous massive star , optically inconspicuous because it was deeply embedded in a dusty envelope , underwent a sudden outburst that cleared most of the surrounding dust . in many respects , ngc 300
ot has behaved remarkably similarly to the supernova impostor sn 2008s ( see 1 ) .
sn 2008s was also heavily enshrouded by dust prior to its outburst , with an intrinsic luminosity of @xmath47 based on its ir sed , implying an initial mass of @xmath2@xmath48 .
the energy released in the ot s eruption was relatively modest . integrating over the extinction - corrected visual light curve , beginning with the initial detection in 2008 april , and assuming that the bolometric correction is zero , we find that the total energy released was @xmath2@xmath49@xmath50 ergs .
this is similar to the energy radiated by sn 2008s , @xmath2@xmath51 ergs ( smith et al .
2008 ) , but considerably less than the @xmath2@xmath52 ergs emitted in @xmath8 car s great eruption , and of course the @xmath2@xmath53 ergs in true core - collapse sne . given the large amounts of pre - outburst obscuring dust , both ngc 300 ot and sn 2008s must have experienced significant mass loss as post - main - sequence stars , and were most likely red supergiants ( rsgs ) , post - rsgs , or conceivably lower - mass stars at the tip of the agb just before the outbursts .
thompson et al . (
2008 ) have suggested that these events may represent a new class of low - luminosity outbursts arising from stars in the mass range @xmath28@xmath54 . by comparison with multi - epoch _ spitzer _
irac observations in m33 , they have identified a group of rare , reddened sources that they call extreme - agb stars , which may be the progenitor class in a short - lived stage lasting only @xmath55@xmath56 yr immediately preceding the outbursts .
there are also many enshrouded oh / ir stars in our own galaxy , including agb - tip stars and red supergiants with very red mid - ir colors ; some may be as red as the progenitors of the ot and sn 2008s .
here we suggest that ngc 300 ot , and by analogy sn 2008s , were oh / ir stars before their eruptions . given both objects lack of obvious variability from _ spitzer _ observations several years apart , these ots were not mira variables just before outburst .
they had presumably left the region of rsgs and agb stars in the hr diagram , and were on a blue loop back to warmer temperatures .
the ngc 300 ot s bipolar wind at @xmath3275 km s@xmath57 also supports a warmer temperature for the progenitor . during this transition
, both stars reached a state in which they exceeded the eddington limit for their luminosities and masses and suddenly initiated outflows , bipolar in the case of ngc 300 ot . exactly why this should happen remains uncertain : the eruptions may have resulted from some type of as - yet unexplained failed supernova , a binary merger , or a photospheric eruption
. continued observations of the ngc 300 ot as it fades are especially important .
we currently observe the spectrum of its wind , but if it does not form another cocoon of dust , we may eventually observe the survivor , determine its true nature , and possibly gain some insight into the cause of its eruption . partial support for this research was provided by nasa through grant go-11553 from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
stsci s participation in the smarts consortium is supported by the stsci director s discretionary research fund .
we thank m. hernndez and j. velsquez for conducting the 1.5-m observations , j. espinoza , a. miranda , and j. vasquez for the 1.3-m observations , and michelle buxton and suzanne tourtellotte for scheduling the 1.3-m telescope .
g. pietrzyski kindly obtained one of the mage spectra .
we thank kris davidson , chris kochanek , kris stanek , and todd thompson for useful discussions . | a luminous optical transient ( ot ) that appeared in ngc 300 in early 2008 had a maximum brightness , @xmath0 to @xmath1 , intermediate between classical novae and supernovae .
we present ground - based photometric and spectroscopic monitoring and adaptive - optics imaging of the ot , as well as pre- and post - outburst space - based imaging with and _ spitzer_. the optical spectrum at maximum showed an f - type supergiant photosphere with superposed emission lines of hydrogen , , and [ ] , similar to the spectra of low - luminosity type iin `` supernova impostors '' like sn 2008s , as well as cool hypergiants like irc + 10420 .
the emission lines have a complex , double structure , indicating a bipolar outflow with velocities of @xmath2@xmath3 . the luminous energy released in the eruption was @xmath2@xmath4 ergs , most of it emitted in the first 2 months . by registering new images with deep archival frames , we have precisely located the ot site , and find no detectable optical progenitor brighter than broad - band @xmath5 magnitude 28.5 .
however , archival _ spitzer _ images reveal a bright , non - variable mid - ir pre - outburst source .
we conclude that the ngc 300 ot was a heavily dust - enshrouded luminous star , of @xmath210@xmath6 , which experienced an eruption that cleared the surrounding dust and initiated a bipolar wind .
the progenitor was likely an oh / ir source which had begun to evolve on a blue loop toward higher temperatures , but the precise cause of the outburst remains uncertain . |
let @xmath3 be a finite - dimensional affine algebraic group defined over a field @xmath4 of characteristic zero
. for any ( discrete ) group @xmath5 , the set of all representations of @xmath5 in @xmath3 has a natural structure of an affine @xmath4-scheme called the _ representation scheme _ @xmath6 .
the representation schemes and associated representation varieties play an important role in many areas of mathematics , most notably in representation theory and low - dimensional topology . in representation theory , the fundamental problem is to understand the structure of representations of @xmath5 in @xmath3 .
one can approach this problem geometrically by studying the natural ( adjoint ) action of the group @xmath3 on the variety @xmath6 .
when @xmath7 is algebraically closed and @xmath5 is finitely generated , the equivariant geometry of @xmath8 is closely related to the representation theory of @xmath5 : the equivalence classes of representations of @xmath5 in @xmath3 are in bijection with the @xmath3-orbits in @xmath8 , and the geometry of @xmath3-orbits determines the algebraic structure of representations . this relation has been extensively studied since the late 70s , and the representation varieties have become a standard tool in representation theory of groups ( see , for example , @xcite ) . in topology ,
one is usually interested in global algebro - geometric invariants of spaces defined in terms of representation varieties of fundamental groups .
for example , if @xmath9 is a knot in @xmath10 , many classical invariants of @xmath11 arise from its character variety @xmath12/\!/g \,$ ] , which is the ( categorical ) quotient of the representation variety of the fundamental group of the knot complement @xmath13 .
these invariants include , in particular , the alexander polynomial @xmath14 ( in the simplest case when @xmath15 , see , e.g. , @xcite ) , the so - called @xmath16-polynomial @xmath17 ( see @xcite ) , the casson invariant @xcite and the famous chern - simons invariant @xcite , all of which are defined for @xmath18 . in fact , for @xmath18 , the entire character variety , or rather its coordinate ring @xmath19\ , $ ] , has a purely topological interpretation as a kauffman bracket skein module of @xmath20 ( see @xcite ) . despite being useful tools , the representation varieties exhibit some intrinsic deficiencies .
first of all , these varieties are usually very singular , which makes it hard to understand their geometry .
thus , in representaion theory , one faces the problem of resolving singularities of @xmath8 . in topology
, the use of representation varieties is mostly limited to ( compact orientable ) surfaces , hyperbolic @xmath21-manifolds and knot complements in @xmath10 , all of which are known to be aspherical spaces .
the homotopy type of such a space is completely determined by the isomorphism type of its fundamental group , which makes representation varieties of these groups very strong and efficient invariants . for more general spaces
, however , one needs to take into account a higher homotopy information , and looking at representation varieties of fundamental groups ( or even , higher homotopy groups ) is not enough .
a natural way to remedy these problems is to replace the representation functor @xmath22 with its ( non - abelian ) derived functor @xmath23 much in the same way as one replaces non - exact ( additive ) functors in classical homological algebra ( such as ` @xmath24 ' and ` @xmath25 ' ) with corresponding derived functors ( ` @xmath26 ' and ` @xmath27 ' ) .
geometrically , passing from the representation scheme @xmath8 to the derived representation scheme @xmath28 amounts to desingularizing @xmath8 , while topologically , this brings in a new homology theory of spaces that captures a good deal of homotopy information and refines the classical representation varieties of fundamental groups in an interesting and non - trivial way .
to explain this idea in more precise terms , we recall that the representation scheme @xmath8 is defined as the functor on the category of commutative @xmath4-algebras : @xmath29 assigning to a @xmath4-algebra @xmath16 the set of families of representations of @xmath30 in @xmath3 parametrized by the @xmath4-scheme @xmath31 .
it is well known ( and easy to prove ) that the functor is representable , and we denote the corresponding commutative algebra by @xmath32 may be thought of as the coordinate ring of the affine @xmath4-scheme @xmath8 , and we sometimes write @xmath33 $ ] to emphasize this interpretation . ] .
varying @xmath5 ( while keeping @xmath34 fixed ) , we can now regard @xmath35 as a functor on the category of groups : @xmath36 which we call the _ representation functor _ in @xmath34 .
the functor extends naturally to the category @xmath37 of simplicial groups , taking values in the category @xmath38 of simplicial commutative algebras . both categories
@xmath37 and @xmath38 carry standard ( simplicial ) model structures , with weak equivalences being the weak homotopy equivalences of the underlying simplicial sets .
the functor @xmath39 is not homotopy invariant : in general , it does not preserve weak equivalences and hence does not descend to a functor between the homotopy categories @xmath40 and @xmath41 .
however , it is easy to check that @xmath42 does map weak equivalences between cofibrant objects in @xmath37 to weak equivalences in @xmath38 ( see lemma [ sgfun ] ) .
hence , by standard homotopical algebra , it has a ( total ) left derived functor @xmath43 we call the _ derived representation functor _ in @xmath3 .
heuristically , @xmath44 may be thought of as a
best possible " universal approximation to the representation functor at the level of homotopy categories . when applied to a simplicial group @xmath5 , the functor
is represented by a simplicial commutative algebra which ( abusing notation ) we denote by @xmath28 .
the homotopy groups of @xmath28 depend only on @xmath5 and @xmath3 , with @xmath45 $ ] being canonically isomorphic to @xmath46 .
in particular , if @xmath5 is a discrete simplicial group , then @xmath45 \cong \gamma_g $ ] . extending our terminology from @xcite , we will refer to @xmath47 $ ] as the _ representation homology _ of @xmath5 in @xmath3 and denote it @xmath48 .
we should mention that representation homology of associative and lie algebras was introduced and studied in @xcite .
the idea of deriving the representation functor was motivated by noncommutative geometry , where the representation functor plays an important role ( see @xcite and also @xcite ) . next , we recall that the model category @xmath37 of simplicial groups is quillen equivalent to the category of reduced simplicial sets , @xmath49 , which is , in turn , quillen equivalent to the category @xmath50 of pointed connected topological spaces .
these classical equivalences are given by two pairs of adjoint functors @xmath51 the construction of which will be briefly reviewed in section [ klg ] . here
, we only recall that @xmath52 is the kan loop group functor that assigns to a reduced simplicial set @xmath53 a semi - free simplicial group @xmath54 , which is a simplicial model of the based loop space @xmath55 ( see @xcite ) .
the functor @xmath52 preserves weak equivalences and hence descends to a functor between the homotopy categories of @xmath56 and @xmath37 . combining @xmath57 with the derived representation functor
, we define the _ representation homology _ of @xmath53 by @xmath58\ .\ ] ] the representation homology @xmath59 is a graded commutative algebra that depends only on the homotopy type of @xmath60 and hence is a homotopy invariant of the corresponding space @xmath61 . in degree zero , we have @xmath62\,$ ] , where @xmath63 is the fundamental group of @xmath60 . to avoid confusion , we emphasize that @xmath64 in general ; however , if @xmath60 is a @xmath65-space ( _ e.g. _ , @xmath66 for a discrete group @xmath5 ) , then indeed we have @xmath67 , so there is no ambiguity in our notation .
the goal of the present paper is threefold .
first , we give an elementary construction of representation homology in terms of classical ( abelian ) homological algebra .
our construction is analogous to pirashvili s construction of higher order hochschild homology , and it provides a natural interpretation of representation homology as functor homology .
this opens the way to efficient computations and places representation homology in one row with other classical invariants such as hochschild and cyclic homology .
second , we construct spectral sequences and natural maps relating representation homology to other homology theories associated with spaces ( including @xmath68 , the higher hochschild homology , the @xmath69-equivariant homology of the free loop space @xmath70 and the stable homology of the automorphism groups of f.g .
free groups @xmath71 ) .
third , we compute representation homology explicitly in a number of interesting cases , including the spheres @xmath72 , suspensions @xmath73 , closed surfaces of arbitrary genus and some classical 3-dimensional spaces , such as the link complements in @xmath74 and the lens spaces @xmath2 .
one of our main results is the comparison theorem that expresses the representation homology of a simply - connected topological space of finite rational type in terms of its quillen and sullivan models .
we now proceed with a summary of the main results of the paper .
we begin by recalling pirashvili s definition of higher hochschild homology @xcite .
let @xmath75 denote the full subcategory of finite sets with objects @xmath76 ( for @xmath77 ) and @xmath78 .
if @xmath79 is the category of @xmath4-vector spaces , any functor @xmath80 with values in @xmath79 extends naturally to a functor @xmath81 on the category of all sets ( by taking the left kan extension along the inclusion @xmath82 ) .
now , given @xmath80 and any simplicial set @xmath83 , we can define a simplicial @xmath4-vector space @xmath84 by composition @xmath85 following @xcite , we denote the homotopy groups of @xmath84 by @xmath86\,$ ] , where @xmath87 stands for the dold - kan normalization functor .
any commutative @xmath4-algebra @xmath16 defines a functor @xmath88 by assigning to an object @xmath89 the @xmath90-th tensor power @xmath91 and to a morphism @xmath92 the linear map @xmath93 , @xmath94 , with @xmath95 for @xmath96 .
the _ higher hochschild homology _ of @xmath60 with coefficients in @xmath16 is then defined by the formula @xmath97 . as observed in @xcite ,
if @xmath98 is the ( simplicial ) circle , then @xmath99 coincides with the classical hochschild homology @xmath100 of the commutative algebra @xmath16 .
the key fact behind this construction is that @xmath101 is a prop ( i.e. , a permutative category ) with monoidal structure @xmath102 , and the category of @xmath4-algebras over @xmath103 ( i.e. , the category of strong monoidal functors @xmath104 ) is equivalent ( via @xmath105 ) to the category of commutative @xmath4-algebras .
now , to define representation homology we replace @xmath101 by the category @xmath106 of finitely generated free groups with objects @xmath107 ( for @xmath108 ) and morphisms being the arbitrary group homomorphisms .
it is known that @xmath109 is also a prop with monoidal structure @xmath110 , and the category of @xmath4-algebras over @xmath109 is equivalent to the category of commutative hopf @xmath4-algebras . under this equivalence
, a commutative hopf algebra @xmath111 corresponds to the ( strong monoidal ) functor @xmath112 , @xmath113 , which actually takes its values in the category of commutative algebras . as in the case of sets ,
the functor @xmath114 extends naturally to the category of all ( based ) free groups @xmath115 , whose objects are the free groups @xmath116 given with a prescribed generating set @xmath117 and morphisms are the arbitrary group homomorphisms . mimicking the pirashvili construction , for a reduced simplicial set @xmath53 and a commutative hopf algebra @xmath118
, we can consider the composition of functors @xmath119 where @xmath120 is the kan loop group construction of @xmath60 and @xmath121 is the ( left ) kan extension of @xmath122 along the inclusion @xmath123 .
this defines a simplicial commutative algebra @xmath124 , whose homotopy groups we denote by @xmath125 \ .\ ] ] our main observation is that this definition is _ equivalent _ to our original definition of representation homology given in terms of the derived representation functor @xmath126 , see .
precisely ( _ cf .
_ proposition [ rephom0 ] ) , let @xmath34 be an affine group scheme over @xmath7 with coordinate ring @xmath127 .
then , for any @xmath53 , there is a natural isomorphism of graded commutative algebras @xmath128 thanks to proposition [ iprop1 ] , we may ( and will ) use the notation @xmath129 and @xmath130 interchangeably , without causing confusion .
although its proof is almost immediate , proposition [ iprop1 ] has a number of important implications .
first , we state the following theorem , which is the main result of section [ section_rephom ] ( see theorem [ rephomdtp2 ] ) . for any @xmath53
, there is a natural first quadrant spectral sequence @xmath131 converging to the representation homology of @xmath60 .
the spectral sequence relates the representation homology @xmath132 of a space @xmath60 to its pontryagin algebra @xmath133 . to describe the @xmath134-term of we recall that @xmath133 has a natural structure of a graded cocommutative hopf algebra with coproduct induced by the alexander - whitney diagonal and the product by the eilenberg - zilber map . for each @xmath135 , the assignment @xmath136_q\,$ ] , where @xmath137_q $ ] is the @xmath138-th graded component of the @xmath90-th tensor power of @xmath139 , defines a functor @xmath140 , which is the first argument of the `` tor '' in .
the `` tor '' itself is the ( abelian ) derived functor of the tensor product @xmath141 between covariant and contravariant @xmath142-valued functors over the ( small ) category @xmath106 .
the spectral sequence is a counterpart of pirashvili s fundamental spectral sequence for higher hochschild homology ( _ cf . _
* theorem 2.4 ) ) ; however , in the case of representation homology it takes a more geometric form .
theorem [ ithm1 ] has several interesting applications .
for example , if @xmath60 is a @xmath143-space , the spectral sequence degenerates giving an isomorphism ( _ cf .
_ corollary [ bgamma ] ) @xmath144 , \o(g))\ , \ ] ] where @xmath145 $ ] is the group algebra of @xmath5 equipped with the natural ( cocommutative ) hopf algebra structure .
the isomorphism shows that the representation homology of ( discrete ) groups has a natural ` tor ' interpretation , similar to the classical ( connes ) interpretation of the hochschild and cyclic homology ( see ( * ? ? ?
* chap . 6 ) ) .
furthermore , combined with a standard comparison theorem for spectral sequences , theorem [ ithm1 ] implies that the representation homology @xmath130 is stable under pontryagin equivalences inducing isomorphisms of pontryagin algebras @xmath146 . ] , and hence , if @xmath60 is simply - connected , then @xmath130 is actually a _ rational _ homotopy invariant of @xmath60 ( _ cf .
_ proposition [ rephomhtpinv ] ) .
next , in section [ sect5 ] , we show that representation homology is not only analogous to higher hochschild homology but actually related to it in a remarkably simple , geometric way .
the main result of this section reads ( _ cf .
_ theorem [ repvshh1 ] and theorem [ repvshh ] ) : let @xmath118 be a commutative hopf algebra .
@xmath147 for any simplicial set @xmath148 , there is a natural isomorphism @xmath149 where @xmath150 is a pointed simplicial set obtained from @xmath60 by adjoining functorially a basepoint @xmath151 , and @xmath152 is the @xmath153reduced@xmath154 suspension functor on the category of pointed simplicial sets .
@xmath155 for any pointed simplicial set @xmath156 , there is a natural isomorphism @xmath157 where @xmath158 is the pirashvili - hochschild homology of the commutative algebra @xmath118 with coefficents in @xmath7 viewed as an @xmath159-module via the hopf algebra counit @xmath160 .
the proof of part @xmath147 of theorem [ mt51 ] is based on milnor s classical @xmath161-construction @xcite that gives a simplicial group model for the space @xmath162 .
part @xmath155 follows from @xmath147 .
theorem [ mt51 ] has strong implications : in particular , it allows one to compute the representation homology of suspensions in a completely explicit way .
indeed , it is known that @xmath73 for any pointed connected space @xmath60 is _ rationally _ homotopy equivalent to a bouquet of spheres of dimension @xmath163 . since representation homology depends only on the rational homotopy type of a space , the isomorphism together with pirashvili s computations @xcite of higher hochschild homology of spheres , implies ( _ cf .
_ proposition [ rephomsusp ] ) for any pointed connected space @xmath60 of finite type , there is an isomorphism @xmath164 where @xmath165 is the linear dual of the lie algebra of @xmath3 , and @xmath166 is the reduced @xmath153singular@xmath154 homology of @xmath60 with constant coefficients in @xmath165 . by induction ,
proposition [ psusp ] implies @xmath167 \right)\,$ ] for all @xmath168 . in particular , for @xmath169 , we have @xmath170)\ , \
n \ge 2 \ .\ ] ] in section [ sect6 ] , we look at representation homology of simply - connected spaces .
we recall that , by a fundamental theorem of quillen @xcite , to any such space @xmath60 one can associate a connected ( chain ) dg lie algebra @xmath171 called a quillen model of @xmath60 .
the lie algebra @xmath171 encodes the rational homotopy type of @xmath60 and hence determines any homotopy invariant of @xmath172 over @xmath173 .
it is natural to ask how to express the representation homology of @xmath60 over @xmath173 in terms of @xmath171 .
the answer to this question is given by the following result ( _ cf .
_ theorem [ conj1 ] ) , which we refer to as the ` comparison theorem ' and which is technically the main theorem of the present paper .
let @xmath60 be a @xmath174-connected pointed topological space of finite rational type .
then , for any affine algebraic group @xmath3 defined over @xmath173 with lie algebra @xmath175 , there is an isomorphism of graded commutative @xmath173-algebras @xmath176 in theorem [ cthm ] , @xmath177 denotes the representation homology of the dg lie algebra @xmath171 with coefficients in the ( finite - dimensional ) lie algebra @xmath178 , as defined in our earlier paper @xcite . ] . using the results of @xcite ,
we can restate theorem [ cthm ] in terms of a sullivan model of @xmath60 . recall that a sullivan model of @xmath60 is a commutative cochain dg algebra @xmath179 related to @xmath171 via a quasi - isomorphism @xmath180 , where @xmath181 is the chevaley - eilenberg cochain complex of @xmath171 .
such a commutative dg algebra model always exists and provides an alternative way to describe the rational homotopy type of @xmath60 ( see @xcite ) .
now , for a fixed @xmath179 , we let @xmath182 denote the augmentation ideal of @xmath179 corresponding to the basepoint of @xmath60 , and define @xmath183 to be the current lie algebra of @xmath178 over @xmath182 . then , as consequence of ( * ? ? ?
* theorem 6.5 ) and our comparison theorem , we get ( _ cf . _
theorem [ c1sullivan ] ) @xmath147 for any @xmath174-connected pointed space @xmath60 of finite @xmath173-type , there is an isomorphism @xmath184 where the right - hand side is the classical @xmath153chevalley - eilenberg@xmath154 cohomology of the dg lie algebra @xmath185 equipped with homological grading .
@xmath155 if @xmath3 is a reductive affine algebraic group over @xmath4 , then @xmath186 where the right - hand side is the relative lie algebra cohomology of the pair @xmath187 .
part @xmath155 of theorem [ hrsullivan ] allows us to compute the @xmath3-invariant part of representation homology of some classical simply - connected spaces .
for example , we have ( _ cf . _ theorem [ rephomcpr ] and corollary [ rephomcprexpl ] ) let @xmath3 be a complex reductive lie group of rank @xmath188 .
let @xmath189 denote the exponents of the lie algebra of @xmath34 .
then , for any @xmath190 , there is an isomorphism of graded commutative algebras @xmath191 where the generator @xmath192 has homological degree @xmath193 , @xmath194 .
we deduce theorem [ strmconj ] from the so - called strong macdonald conjecture , a celebrated conjecture in representation theory proposed by feigin and hanlon in the early 80s and proved by fishel , grojnowski and teleman in 2008 ( see @xcite ) .
it is interesting to note that the isomorphism has a natural topological interpretation : it is given by a canonical algebra map which we call the _ drinfeld homomorphism _ : @xmath195 \,\to \ , \hr_\ast(x , g)^g\ .\ ] ] the domain of is the graded symmetric algebra of the direct sum of certain ( topologically defined ) eigenspaces of the @xmath196-equivariant homology of the free loop space @xmath197 ( see section [ hreh ] ) .
the drinfeld homomorphism exists for _ any _ simply - connected space @xmath60 , and it happens to be an isomorphism in the following cases : @xmath198 ( @xmath199 ) , @xmath200 ( @xmath201 ) and @xmath202 ( @xmath203 odd , @xmath204 ) .
it would be interesting to find more examples and give an abstract ( topological ) characterization of all such spaces . despite its simple statement ,
the proof of theorem [ cthm ] is fairly long and technical : it occupies most of section [ sect6 ] and relies heavily on results of quillen @xcite . for reader s convenience , we outline the main steps of the proof in section [ sect6.3.1 ] .
we close section [ sect6 ] with a conjectural generalization of theorem [ cthm ] to non - simply connected spaces ( see conjecture [ conj5 ] ) .
our conjecture is inspired by a recent work of buijs , flix , murillo and tanr who proposed a natural generalization of quillen models to non - simply connected spaces ( see @xcite ) . in section [ sec7 ] , we compute representation homology of some classical non - simply connected spaces .
our examples include closed surfaces of arbitrary genus ( both orientable and non - orientable ) as well as some three - dimensional spaces ( the link complements in @xmath74 and @xmath10 , the lens spaces @xmath2 and a general closed orientable @xmath21-manifold ) .
the representation homology of surfaces and link complements is expressed in terms of classical hochschild homology of @xmath205 and related commutative algebras .
for example , for the link complements in @xmath74 , we prove ( _ cf .
_ theorem [ hrlink ] ) let @xmath206 be a link in @xmath74 obtained as the alexander closure of a braid @xmath207 . then the representation homology of the complement of its @xmath153regular@xmath154 neighborhood in @xmath74 is given by @xmath208 the right - hand side of is the usual hochschild homology of the associative algebra @xmath209 with bimodule coefficients .
the bimodule @xmath210 is isomorphic to @xmath211 as a left module , while the right action of @xmath212 is twisted by an element @xmath213 viewed as an automorphism of @xmath214 via the artin representation of the braid group @xmath215 .
the representation homology of the classical lens space @xmath2 and , more generally , of an arbitrary closed orientable @xmath21-manifold @xmath216 is expressed in terms of its heegaard splitting by a differential ` tor ' functor ; this gives rise to an eilenberg - moore spectral sequence converging to @xmath217 ( see theorem [ lensspaces ] and theorem [ 3mnld ] for precise statements ) . we should mention that the key to our calculations in section [ sec7 ] is lemma [ hopushout ] which says that the derived representation functor preserves homotopy pushouts . since the representation functor is not ( left ) quillen ( see remark after lemma [ sgfun ] ) , this result is not immediate and requires a fairly technical proof which we defer to appendix [ scalgf ] . in section [ sec7 ]
, we also discuss a multiplicative version of the derived harish - chandra conjecture proposed in @xcite .
if @xmath3 is a connected reductive group with a maximal torus @xmath218 and @xmath219 is the associated weyl group , then for any space @xmath60 , there is a natural map @xmath220 which we call the _ derived harish - chandra homomorphism _ ( _ cf . _
* section 7 ) ) . in view of , by the classical chevalley restriction theorem @xcite , the map is an isomorphism for any odd - dimensional sphere @xmath221 .
we conjecture that is also an isomorphism for _ the product of two odd - dimensional spheres _ , @xmath222 @xmath223 , at least when @xmath3 is a classical group of type @xmath224,@xmath225 or @xmath226 @xmath227 . in the simply - connected case ( for @xmath228 ) , this conjecture was stated in @xcite in terms of representation homology of lie algebras , and some of its consequences and special cases were proved .
for the two - dimensional torus @xmath229 , we can reformulate it in the following explicit form ( see section [ s7.1.1 ] , conjecture [ conj2 ] ) : @xmath230^w\ , \ ] ] where @xmath231 is the lie algebra of @xmath232 ( i.e. , a cartan subalgebra of @xmath178 ) and @xmath233 is its linear dual . as for the drinfeld homomorphism , it would be interesting to find more examples of spaces , for which the map is an isomorphism , and/or give an abstract characterization of all such spaces . in the last section of the paper ,
we give another interpretation of representation homology as the hochschild - mitchell homology of a certain bifunctor on the category of finitely generated free groups @xmath234 .
such an interpretation is useful for several reasons .
first , it allows us to define representation _
cohomology _ in a natural way ( by simply replacing the hochschild - mitchell homology with the hochschild - mitchell cohomology of the same bifunctor ) .
second , it suggests that it is natural to extend the definition of representation ( co)homology by taking the hochschild - mitchell ( co)homology of @xmath234 with coefficients in an arbitrary bifunctor @xmath235 : i.e. , @xmath236 .
third and most important , it exhibits a close analogy with _ topological _
hochschild homology , which is known to be isomorphic to the hochschild - mitchell homology of the category @xmath237 of finitely generated free _
abelian _ groups ( see @xcite ) . motivated by this analogy
, we construct functorial trace maps @xmath238 relating homology of the automorphism groups of f.g .
free groups with appropriate coefficients to representation homology .
these maps are compatible with natural inclusions @xmath239 and hence have an stable limit as @xmath240 .
the corresponding stable map @xmath241 can be viewed a non - abelian analogue of the classical dennis trace relating topological hochschild homology to stable homology of general linear groups .
we conjecture that the map @xmath242 is actually an isomorphism , whenever @xmath243 is a _ polynomial _ bifunctor ( _ cf . _ conjecture [ conj3 ] ) .
this is a non - abelian analogue of a theorem of scorichenko @xcite .
the paper contains two appendices . in appendix
[ doldkan ] , we describe an abstract monoidal version of the classical dold - kan correspondence relating the category of ( non - negatively graded ) dg @xmath244-algebras and the category of simplicial @xmath244-algebras for an arbitrary @xmath4-linear operad @xmath244 . this version of the dold - kan correspondence is needed for our proof of comparison theorem in section [ sect6 ] .
the main result of appendix [ doldkan ] is theorem [ nnstar_equiv ] , which states that when @xmath4 is a field of characteristic @xmath245 , there is a quillen equivalence between the category of ( non - negatively graded ) dg @xmath244-algebras and the category of simplicial @xmath244-algebras .
various special cases of this theorem have appeared in the literature .
first of all , when @xmath244 is the lie operad , a slightly weaker version ( namely , a quillen equivalence between the category of _ positively graded _
dg lie algebras and reduced simplicial lie algebras ) was proved in ( * ? ? ?
* part i , theorem 4.6 ) . in _ loc .
_ quillen also outlines a proof for the commutative operad ( which controls commutative unital @xmath4-algebras ) under the same reducedness assumptions . for general ( nonreduced ) commutative algebras , the proof of the dold - kan correspondence
is given in ( * ? ? ?
* proposition a.1 ) .
the case of the associative operad is treated in greater generality ( for any commutative ring @xmath4 ) in @xcite , where the dg associative algebras and simplicial associative algebras are viewed as monoids in the ( symmetric ) monoidal model categories of chain complexes and simplicial @xmath4-modules , respectively . in this case
, the dold - kan correspondence follows from an abstract comparison theorem between monoids in different ( symmetric ) monoidal model categories .
the arguments that establish each of these special cases seem to apply only to the case in hand . to the best of our knowledge ,
a unified proof for any linear operad is missing in the literature .
our theorem [ nnstar_equiv ] fills in this gap . the key argument we use to prove theorem [ nnstar_equiv ]
is sketched in ( * ? ? ?
* remark 6.4.5 ) in the special case of the commutative operad .
theorem [ nnstar_equiv ] is crucial for the proof of our comparison theorem ( theorem [ cthm ] ) . while quillen s original result for reduced dg lie algebras is sufficient for this proof , the full strength of theorem [ nnstar_equiv ] is needed to prove proposition [ liereps ] , which is an interesting result on its own . in our second appendix ( appendix [ scalgf ] )
, we collect some technical facts about simplicial commutative algebras .
most of these facts should be well known to experts and go back to the classical works of andr @xcite and quillen @xcite .
however , the main results of this appendix proposition [ reedy_smooth_tensor ] and proposition [ reedy_smooth_colim ] did not seem to appear in the literature in this form and generality . these propositions are needed to prove lemma [ hopushout ] , which we use in the proofs of our main theorems in section [ section_rephom ] and section [ sect5 ] and also for explicit computations in section [ sec7 ] . while a weaker version of lemma [ hopushout ] ( that can be proven directly ) suffices for the results in section [ section_rephom ] and section [ sect5 ] , the full strength of this lemma is required for computations in section [ sec7 ] .
the paper is organized as follows . in section [ s2 ] , we introduce notation and recall some basic facts about simplicial sets and spaces . in section [ s3 ] , we review pirashvilli s construction of higher hochschild homology . in section [ section_rephom ] , we define representation homology in two equivalent ways ( via the derived representation functor and the pirashvilli - type construction ) and prove some basic properties . in section [ section_rephom ] , we establish the isomorphism between the representation homology of suspensions and higher hochschild homology . in section [ sect6 ] ,
we prove our main comparison theorem for simply - connected spaces and discuss its implications . in section [ sec7 ] , we explicitly compute representation homology for certain non - simply connected spaces .
finally , in section [ sec8 ] , we identify representation homology in terms of hochschild - mitchell homology and construct a non - abelian analogue of the dennis trace map relating representation homology to the stable homology of automorphism groups of finitely generated free groups .
we are very grateful to ayelet lindenstrauss and birgit richter for interesting questions and suggestions .
we also thank jim davis , siddhartha gadgil , martin kassabov , mike mandell and sasha patotski for discussions and comments .
the second author expresses his gratitude to the department of mathematics , indian institute of science , bengaluru for conducive working conditions during his visit in the summer of @xmath246 .
research of the first two authors was partially supported by the simons foundation collaboration grant 066274 - 00002b ( ` representation homology ' ) .
in this section , we introduce notation and recall some basic facts about simplicial sets .
standard references for this material are @xcite , @xcite and ( * ? ? ?
* chapter 8) .
let @xmath247 denote the simplicial category .
recall that the objects of @xmath247 are the finite ordered sets @xmath248\,:=\,\{0 , 1 , \ldots , n\}$ ] , @xmath249 , and the morphisms are the ( weakly ) order preserving maps @xmath248 \to [ m ] $ ] .
a _ simplicial object _ in a category @xmath250 is a contravariant functor from @xmath247 to @xmath250 : i.e. , @xmath251 the simplicial objects in @xmath250 form a category , with morphisms being the natural transformations of functors .
we denote this category by @xmath252 . if @xmath253 , we write @xmath254)$ ] .
the category @xmath247 is generated by two distinguished classes of morphisms @xmath255 and @xmath256 , whose images under @xmath257 are called the _
face _ and _ degeneracy maps _ of @xmath60 , respectively .
the map @xmath258 { \xrightarrow{}}[n]$ ] is the ( unique ) injection that does not contain `` @xmath259 '' in its image ; the corresponding face map is denoted by @xmath260 .
similarly , for @xmath249 , the map @xmath261 { \xrightarrow{}}[n]\,$ ] is the ( unique ) surjection in @xmath247 that takes value `` @xmath259 '' twice . the image of @xmath262 under @xmath60 is the degeneracy map @xmath263 .
the face and degeneracy maps of a simplicial object satisfy the following _ simplicial relations _ : @xmath264 thus , a simplicial object in @xmath265 is determined by a family @xmath266 of objects in @xmath250 together with morphisms @xmath267 and @xmath268 satisfying the relations .
the object @xmath269 is usually called the `` set '' of _ @xmath90-simplices _ of @xmath60 , and the @xmath245-simplices are usually called the _ vertices _ of @xmath60 .
we let @xmath270 denote the category of simplicial sets ( _ i.e. _ , simplicial objects in the category @xmath271 ) .
a simplicial set @xmath272 is called _ reduced _ if it has a single vertex , i.e. @xmath273 . the full subcategory of @xmath274 consisting of reduced simplicial sets will be denoted @xmath275 .
a simplicial set @xmath276 is called _ pointed _ if there are distinguished simplices @xmath277 , one in each degree , such that @xmath278 for all @xmath279 .
the sequence @xmath280 is called a basepoint of @xmath272 .
the category of pointed simplicial sets will be denoted @xmath281 .
note that @xmath282 can also be viewed as a full subcategory of @xmath283 as every reduced simplcial set has a canonical ( unique ) basepoint .
given @xmath284 , the set of _ nondegenerate _ @xmath90-simplices of @xmath60 is defined to be @xmath285 every element of @xmath286 can be uniquely expressed in terms of the nondegenerate elements of @xmath60 ( see ( * ? ? ? * lemma 11 ) for a precise statement ) .
in particular , a simplicial set can be defined by specifiying its nondegenerate simplices together with the restriction of each face map to the set of nondegenerate simplices .
we give a few basic examples of simplicial sets that will be used in this paper . to any object
@xmath287 one can associate a simplicial object @xmath288 , with @xmath289 and @xmath290 , @xmath291 being the identity map of @xmath16 for all @xmath292 .
this gives a fully faithful embedding @xmath293 .
the objects of @xmath252 arising this way are called _ discrete simplicial objects_. the @xmath90-dimensional geometric simplex is the topological space @xmath294 let @xmath295 denote the vertex of @xmath296 with @xmath297-th coordinate @xmath174 . for any morphism @xmath298 { \xrightarrow{}}[n]$ ] in @xmath247 , there is a ( unique ) linear map @xmath299 sending @xmath295 to @xmath300 , that restricts to a map of topological spaces @xmath301 . the collection
@xmath302 forms a _ cosimplicial space _ , _
i.e. _ , a ( covariant ) functor @xmath303 , where @xmath304 denotes the category of ( compactly generated weakly hausdorff ) topological spaces .
this functor is faithful : it gives a topological realization of the simplicial category , which was historically the first definition of @xmath305 .
let @xmath306 denote the yoneda embedding .
the functor @xmath307 assigns to @xmath308 $ ] a simplicial set @xmath309_{\ast } $ ] called the _
standard @xmath90-simplex_. explicitly , @xmath309_{\ast}$ ] is given by @xmath310_k\,:=\,\hom_{\delta}([k],[n])\,\cong\ , \{(n_0,\ldots , n_k)\,|\,0
\leq n_0 \leq \ldots \leq n_k \leq n\ } \,,\ ] ] where a function @xmath311 { \xrightarrow{}}[n]$ ] is identified with the sequence of its values @xmath312 . under this identification
, the nondegenerate simplices correspond to _ strictly _ increasing functions , and the face and degeneracy maps in @xmath309_{\ast}$ ] are given by @xmath313 by yoneda lemma , for any simplicial set @xmath60 , there is a natural bijection @xmath314_{\ast } , x)\,\cong\ , x_n\,,\ ] ] which shows that @xmath315_{\ast}$ ] ( co)represents the functor : @xmath316 .
the yoneda functor @xmath317 can be also regarded as a cosimplicial object in the category of simplicial sets . in particular , for any @xmath279 , there are @xmath318 coface maps @xmath319_{\ast } { \xrightarrow{}}\delta[n]_{\ast } $ ] , @xmath320 . using these maps , we define the _ boundary _ of @xmath309_{\ast}$ ] to be the simplicial subset @xmath321_{\ast}\,:=\ , \bigcup _ { 0 \leq i \leq
n } d^i(\delta[n-1]_{\ast } ) \,\subset \,\delta[n]_{\ast}\ , \ ] ] the _ simplicial @xmath90-sphere _ is then defined to be the corresponding quotient set @xmath322_{\ast}/\partial \delta[n]_{\ast}$ ] .
it is easy to see that the only nondegenerate simplices in @xmath323 are in degree @xmath245 and @xmath90 , with @xmath324 and @xmath325 , where @xmath117 is the image of the map @xmath326_n$ ] in @xmath327 .
note that @xmath328 for all @xmath297 .
thus , the simplicial structure of @xmath323 reflects the standard cw decomposition of the @xmath90-sphere @xmath329 with one cell in dimension @xmath245 and one cell in dimension @xmath90 .
the simplicial @xmath174-sphere @xmath330 is called the
_ simplicial circle_. by example [ sstd ] , we have @xmath331_k\,\,\cong\,\{(\underbrace{0,\ldots,0}_{\text{$i$ } } , \underbrace{1,\ldots,1}_{\text{$k+1-i$ } } ) \,|\ , i=0,1,\ldots , k+1\}$ ] and @xmath332_k\,=\,\{(0,\ldots 0),(1,\ldots,1)\}$ ] .
hence , @xmath330 is given explicitly by @xmath333 with @xmath334 corresponding to the basepoint @xmath335 .
there is an important functor @xmath336 assigning to each simplicial set @xmath272 a topological space @xmath61 called the _
geometric realization _ of @xmath60 .
explicitly , the space @xmath61 is defined by @xmath337 where each set @xmath286 is equipped with discrete topology and the equivalence relation is given by @xmath338 more formally ( see , e.g. , ( * ? ? ?
* section 1.3 ) ) , the functor @xmath336 can be defined as the ( left ) kan extension @xmath339 of the geometric simplex @xmath340 along the yoneda embedding @xmath341 .
it follows from this definition that @xmath342_*| \cong \delta^n $ ] for all @xmath108 , and in general , @xmath343 , where the colimit is taken over all morphisms of @xmath315_{\ast } { \xrightarrow{}}x\,,\,n \geq 0 $ ] .
the category @xmath344 has a standard model structure , where the weak equivalences are the morphisms inducing weak homotopy equivalences of the corresponding geometric realizations .
the cofibrations are levelwise injective maps and the fibrations are the kan fibrations ( see @xcite ) .
this structure gives a model structure on @xmath282 .
let @xmath345 be a pointed topological space .
the _ ( total ) singular complex _ of @xmath60 is a simplicial set @xmath346 defined by @xmath347 .
the _ eilenberg subcomplex _
@xmath348 of @xmath346 is @xmath349 if @xmath60 is connected , the natural inclusion @xmath350 is a weak equivalence of simplicial sets .
further , if we restrict @xmath351 to the category @xmath352 of connected pointed spaces , we get the pair of adjoint functors @xmath353 which induce mutually inverse equivalences of the homotopy categoriess : @xmath354 .
this equivalence justifies the following standard convention which we will follow throughout the paper
. * convention .
* we shall not notationally distinguish between a reduced simplicial set @xmath60 and its geometric realization @xmath355 .
nor shall we distinguish notationally between a topological space and a ( reduced ) simplicial model of that space .
recall that for an arbitrary associative @xmath4-algebra @xmath16 and an @xmath16-bimodule @xmath356 , the classical hochschild chain complex @xmath357 is defined by @xmath358 , @xmath249 , with differential @xmath359 given by the formula @xmath360 where @xmath361 and @xmath362 .
the homology of @xmath357 is denoted @xmath363 and called the _
hochschild homology of @xmath16 with coefficients in @xmath356_. for _ commutative _ algebras , there is a remarkable generalization of the hochschild complex , due to pirashvili @xcite , that depends on an arbitrary simplicial set @xmath272 and specializes for @xmath364 to the original classical construction . in this section ,
we review the definition of pirashvili s complex and give two natural interpretations of its homology groups in terms of loday construction and functor tensor products .
all results in this section can be found in ( or easily derived from ) pirashvili s papers @xcite .
let @xmath365 denote the category of finite pointed sets with objects @xmath308 = \{0,1 , \ldots , n\ } $ ] , @xmath366 , and morphisms @xmath367 \to [ m ] $ ] being arbitrary set maps such that @xmath368 .
let @xmath369 be a covariant functor .
we extend @xmath370 to the category @xmath371 of _ all _ pointed sets in a natural way , using the left kan extension along the inclusion @xmath372 .
we keep the notation @xmath370 for the extended functor : explicitly , @xmath373 is given by @xmath374)\,$ ] , where the colimit is taken over all pointed inclusions @xmath308
{ \,\hookrightarrow\,}x $ ] . given a pointed simplicial set @xmath375
, we define a simplicial @xmath4-vector space @xmath376 as the composition of functors @xmath377 we denote the homotopy groups of @xmath376 by @xmath378 and recall that @xmath379 $ ] , where @xmath380 is the dold - kan normalization functor ( see appendix [ sdk ] ) .
now , any commutative @xmath4-algebra @xmath16 and an @xmath16-module @xmath356 ( viewed as a symmetric bimodule ) give rise to a functor @xmath381 that assigns to the set @xmath308 $ ] the vector space @xmath382 and to a pointed map @xmath383 \to [ m ] $ ] , the action of @xmath384 on @xmath385 given by @xmath386 where @xmath387 for @xmath388 .
following @xcite , we denote this functor by @xmath389 , and for a pointed simplicial set @xmath390 , define @xmath391 thus , @xmath392 is the homology of the complex @xmath393 $ ] , which we call the _ pirashvili - hochschild complex _ of @xmath16 with coefficients in @xmath356 associated to @xmath60 .
let @xmath394 be the simplicial circle .
recall that the set of @xmath90-simplices @xmath395 can be identified with the set of monotone sequences of @xmath245 s and @xmath174 s of length @xmath396 modulo the identification @xmath397 ( see section [ sscirc ] ) . for a nonzero sequence @xmath398 ,
let @xmath399 denote the position of the first @xmath174 .
the map @xmath400 identifies @xmath401 with @xmath248 $ ] . under this identification
, the degeneracy map @xmath402 { \xrightarrow{}}[n+1 ] $ ] corresponds to the unique monotone injection skiping @xmath403 in its image and the face map @xmath404 { \xrightarrow{}}[n-1]$ ] is given by @xmath405 for @xmath406 , @xmath407 for @xmath408 , @xmath409 and @xmath410 for @xmath411 . from this description of @xmath412
, it is easy to see that the pirashvili complex @xmath413 for @xmath414 is precisely the classical hochschild complex @xmath415 . thus , @xmath416 for any commutative algebra @xmath16 and @xmath16-module @xmath356 . in a similar way , one can exlicitly describe the pirashvili complex @xmath417 for the @xmath90-dimensional simplicial sphere @xmath418 .
the corresponding homology groups @xmath419 are denoted @xmath420}_\ast(a , m ) $ ] and called in @xcite _ the hochschild homology of @xmath421 of order @xmath90_. in the present paper , we will mostly deal with two cases : @xmath422 and @xmath423 , where in the last case the module structure on @xmath7 comes from an augmentation @xmath424 . to simplify the notation
we will write @xmath425 for @xmath426 and regard @xmath427 as a functor on the category of ( pointed ) simplicial sets assuming @xmath16 to be fixed . we will refer to this functor as a _ higher hochschild homology _ of spaces .
we now give a useful interpretation of higher hochschild homology that exhibits the functor @xmath428 as a special case of the loday construction @xcite .
the starting point is the simple observation that the category @xmath429 of commutative algebras is a cocartesian monoidal category tensored over the category @xmath430 .
this means that @xmath430 acts on @xmath429 in the natural way , _
i.e. _ , there is a bifunctor @xmath431 where @xmath432 is defined to be the @xmath117-indexed coproduct of copies of @xmath16 in the category @xmath429 .
now , given a simplicial set @xmath60 , we may consider the composition of functors @xmath433 which , abusing notation , we keep denoting @xmath434 .
there is a natural isomorphism of functors @xmath435 .
this is immediate from the following commuative diagram @xmath436 \delta^{\mathrm{op } } & \rto^{\quad x \quad } & \set & \rto^{(\,\mbox{--}\ , \otimes a)\ } & \ve_k\\ & \rdto_{x } & \uto & \ruto_{\mathcal l(a , a ) } & \\ & & \set_{\ast } & & \\
\end{diagram}\ ] ] where the vertical arrow is the forgetful functor from @xmath437 to @xmath274 .
it follows from lemma [ defequiv ] that @xmath438 . as a consequence
, we have for any @xmath390 and @xmath439 , @xmath440 has the structure of a graded commutative @xmath4-algebra which is independent of the choice of a basepoint in @xmath60 . indeed , by definition , @xmath441 has the structure of a simplicial commutative @xmath4-algebra that depends only on the underlying simplicial set of @xmath60 .
there is another , more conceptual way to define higher hochschild homology , using homological algebra of functor categories over props .
recall that a prop is a permutative category @xmath442 whose set of objects is indexed by ( or identified with ) the natural numbers @xmath443 and whose monoidal structure @xmath444 is given by addition in @xmath443 ( see @xcite ) .
a @xmath4-algebra over a prop @xmath445 is a strict symmetric monoidal functor from @xmath445 to the tensor category @xmath79 . to define hochschild homology we take @xmath445 to be a category @xmath101 of finite sets with monoidal structure given by disjoint union .
more precisely , we let @xmath101 denote the full subcategory of @xmath446 whose objects are the sets @xmath447 for @xmath108 ( where , by convention , @xmath448 ) and morphisms are arbitrary set maps .
the monoidal structure on @xmath101 is given by @xmath449 .
it is well known and easy to prove ( see , e.g. , ( * ? ? ?
* section 2 ) ) that the category of @xmath4-algebras over @xmath450 is equivalent to the category @xmath451 , the equivalence being given by the functor @xmath452 $ ] .
we will write @xmath453 for the algebra over @xmath103 corresponding to the commutative algebra @xmath439 . now , let @xmath103-@xmath454 ( resp . ,
@xmath454-@xmath103 ) denote the category of all covariant ( resp .
, contravariant ) functors from @xmath103 to the category of vector spaces .
the notation suggests that one should think of the objects of @xmath103-@xmath454 and @xmath454-@xmath103 as left and right @xmath103-modules , respectively .
these categories are both abelian with enough projective and injective objects .
furthermore , they are related by a bifunctor @xmath455 that is right exact with respect to each argument , preserves sums and is left balanced ( see , e.g. , ( * ? ? ?
1.5 ) ) . explicitly , for a right @xmath103-module @xmath456 and a left @xmath103-module @xmath457 , @xmath458/r \,,\ ] ] where @xmath459 is the subspace spanned by the vectors of the form @xmath460 with @xmath461 and @xmath462 and @xmath463 running over all maps in @xmath464 .
next , we consider the functor @xmath465\ , \ ] ] where @xmath466 $ ] denotes the vector space generated by a set @xmath117 , and extend to the category simplicial sets in two steps .
first , we define a functor @xmath467 by taking the left kan extension of along the natural inclusion @xmath468 , and then we extend this degreewise to simplicial sets . abusing notation
, we will continue to denote the resulting functor by @xmath469 .
composing @xmath470 with the normalization functor @xmath471 assigns to every simplicial set @xmath272 a chain complex and hence an object in the derived category @xmath472 that we denote by @xmath473 .
now , recall that any commutative algebra @xmath16 defines an algebra @xmath474 over the prop @xmath101 that can be viewed as an object in @xmath475-@xmath476 . with this interpretation of @xmath16
, we have the following result . for any @xmath83 and @xmath439
, there is a natural isomorphism @xmath477\ ] ] although theorem [ dtp ] is not explicitly stated in @xcite , it can be deduced from results of this paper .
we do not give a proof of theorem [ dtp ] here as in the next section , we prove the analogous theorem for representation homology ( see theorem [ rephomdtp ] ) .
in this section , we define representation homology of a ( reduced ) simplicial set in a way parallel to pirashvili s construction of hochschild homology .
a new key ingredient is kan s construction @xcite of a simplicial group model of the loop space @xmath478 of a topological space @xmath60 .
we begin by reviewing this classical construction ( for details and proofs we refer to ( * ? ? ?
* chapter vi ) and ( * ? ? ?
* chapter v ) ) . let @xmath479 denote the category of simplicial groups .
it has a standard model structure , where the weak equivalences and fibrations of simplicial groups are the weak equivalences and fibrations of the underlying simplicial sets .
we note that , unlike @xmath344 , the model category @xmath479 is fibrant : by a classical theorem of moore , every simplicial group is a kan complex ( see ( * ? ? ?
* theorem 17.1 ) ) . a simplicial group @xmath480 is called _ semi - free _ if there is a sequence of subsets @xmath481 , one in each degree , such that @xmath482 is freely generated by @xmath483 , and the set @xmath484 is closed under degeneracies of @xmath5 , _
i.e. _ , @xmath485 for all @xmath486 and @xmath77 .
the subset @xmath487 is called the set of _ nondegenerate generators _ of @xmath30 of degree @xmath90 . one can show that every element in @xmath488 is nondegenerate ( when considered as an element of the underlying simplicial set ) , and a semi - free simplicial group is determined by specifying the sets of nondegenerate generators @xmath488 and the face elements of these generators .
semi - free simplicial groups are cofibrant objects in the model category @xmath479 .
the kan loop group construction provides an important class of semi - free simplicial groups that arise naturally from reduced simplicial sets . to be precise
, the kan construction defines a pair of adjoint functors : @xmath489 where @xmath52 is called the _ kan loop group functor _ ( or _ cobar construction _ ) and @xmath490 is the _ classifying simplicial complex _ ( or _ bar construction _ ) .
the functor @xmath52 preserves weak equivalences and cofibrations , while @xmath491 preserves weak equivalences and fibrations ( see ( * ? ? ?
* proposition v.6.3 ) ) .
hence , is a quillen pair , which is actually a quillen equivalence : i.e. , the functors @xmath52 and @xmath491 induce mutually inverse equivalences between the homotopy categories of @xmath49 and @xmath37 ( see ( * ? ? ?
* corollary v.6.4 ) ) . combining this with the classical quillen equivalence between topological spaces and simplicial sets ( see section [ qeq ] ) : @xmath492 we get an equivalence of the homotopy categories @xmath493 . for further use , we recall the explicit construction of the functor @xmath494 . given a reduced simplicial set @xmath495 , the set of @xmath90-simplices of @xmath496 is defined by @xmath497 where @xmath498 and the isomorphism is induced by the inclusion @xmath499 .
the degeneracy maps @xmath500 are induced by the degeneracy maps @xmath501 of the simplicial set @xmath60 , and the face maps @xmath502 are given by @xmath503 conversely , given a simplicial group @xmath504 , the simplicial set @xmath505 is defined by @xmath506 and @xmath507 for @xmath249 .
the degeneracy and face maps of @xmath505 are given explicitly in @xcite .
we note that when restricted to _
discrete _ simplicial groups , the functor @xmath491 coincides with the usual nerve construction , _
i.e. _ , @xmath508 for any discrete group @xmath509 .
the kan loop group @xmath510 of any reduced simplicial set @xmath60 is semi - free .
more precisely , for each @xmath511 , the composite map @xmath512 is injective when restricted to the subset @xmath513 , and the image @xmath514 forms the set of nondegenerate generators @xmath515 in degree @xmath516 of the semi - free basis @xmath517 of @xmath518 .
the following fundamental theorem clarifies the meaning of the kan loop group construction . for any reduced simplicial set @xmath60
, there is a weak homotopy equivalence @xmath519 where @xmath520 is the @xmath153moore@xmath154 based loop space of @xmath355 .
a detailed proof of theorem [ kanstheorem ] can be found in @xcite ( see _ loc .
_ , theorem 26.6 ) .
its significance becomes clear from the following considerations .
given any path - connected cw complex @xmath307 one can choose a pointed connected simplicial set @xmath521 such that @xmath522 . if @xmath272 is the path - connected component of @xmath523 containing the basepoint , then @xmath276 is a reduced simplicial set such that @xmath524 because @xmath525 is connected .
hence , applying the kan loop group construction to @xmath60 , we get @xmath526 .
thus , @xmath527 is a semi - free simplicial group model of the based loop space of @xmath525 . in this way
, the based loop space of any path - connected cw complex admits a simplicial group model .
we now introduce several ( equivalent ) definitions of the representation homology of a reduced simplicial set with coefficients in an algebraic group @xmath3 .
we begin with the construction of derived representation schemes of simplicial groups , following the approach of @xcite ( see @xcite for details ) .
let @xmath3 be an affine algebraic group scheme over @xmath4 . by definition
, @xmath3 is given by the representable functor from the category of commutative @xmath4-algebras to the category of groups : @xmath528 the algebra @xmath205 that represents is called the coordinate ring of @xmath34 : it is a commutative hopf algebra with coproduct @xmath529 , @xmath530 , corresponding to the multiplication in @xmath34 .
the functor has a left adjoint functor which we denote by @xmath531 we call the _ representation functor _ in @xmath34 , and think of the commutative algebra @xmath35 assigned to a discrete group @xmath5 as the coordinate ring @xmath532 $ ] of the affine scheme @xmath8 parametrizing the representations of @xmath5 in @xmath34 .
it is easy to see that the algebra @xmath35 has the following explicit presentation @xmath533 \otimes_k \o(g)\,]/ \langle\!\langle r \rangle\!\rangle\ , \ ] ] where the relations @xmath459 are given by @xmath534 now , to construct the derived representation functor we embed the category of groups into the category @xmath535 of simplicial groups and extend the functor to @xmath535 in the natural way , assigning to a simplicial group @xmath536 the simplicial commutative algebra @xmath537 . we will keep the notation @xmath538 for this extended representation functor : @xmath539 the functor maps weak equivalences between cofibrant objects in @xmath37 to weak equivalences in @xmath540 , and hence has a total left derived functor @xmath541 suppose that @xmath542 is a weak equivalence between cofibrant simplicial groups .
since @xmath543 is a fibrant model category , @xmath5 and @xmath544 are both fibrant - cofibrant objects . by whitehead
s theorem , the map @xmath384 has then a homotopy inverse @xmath545 , such that @xmath546 and @xmath547 .
now , any homotopy between fibrant - cofibrant objects can be realized using a good cylinder object in @xmath37 . since @xmath543 is a simplicial model category
, there is a natural choice of good cylinder objects for @xmath30 and @xmath544 : namely , @xmath548 \rightarrow \gamma\,$ ] , and similarly for @xmath549 . for such cylinder objects , the simplicial homotopies ( see ( * ? ? ? * def .
5.1 ) ) can be defined by explicit combinatorial relations which are obviously preserved by the functor @xmath550 .
thus , we conclude that the morphism @xmath551 is a homotopy inverse of @xmath552 in @xmath540 and hence @xmath553 and @xmath554 are mutually inverse isomorphisms in @xmath555 .
the existence of the derived functor follows now from ( * ? ? ?
despite the fact that the representation functor is a left adjoint functor , it is _ not _ left quillen . indeed , by definition , any left quillen functor preserves cofibrations ; in particular , maps cofibrant objects to cofibrant . to see that this is not the case for take @xmath556 , the multiplicative group , and apply to the free group on one generator @xmath557 , viewed as a discrete simplicial group in @xmath479 .
the result is @xmath558 $ ] , which is _ not _ a cofibrant simplicial algebra in @xmath559 .
thus , the simplicial adjunction @xmath560 is _ not _ a quillen pair , so the result of lemma [ sgfun ] is not automatic . for a fixed simplicial group @xmath561
, we define @xmath562 and call @xmath563 the _ derived representation scheme _ of @xmath5 in the algebraic group @xmath3 . by definition , @xmath563 is a simplicial commutative algebra viewed as an object of @xmath564 .
following @xcite , we refer to the corresponding homotopy groups as the _ representation homology _ of @xmath509 in @xmath565 : @xmath566 : = \h_\ast(n[\drep_{g}(\gamma)])\ ] ] it is easy to check that the functor commutes with @xmath567 , so that for any @xmath561 , there is a natural isomorphism in @xmath451 : @xmath568\,\cong\,[\pi_0(\gamma)]_{g}\ ] ] hence , for an ordinary group @xmath569 ( viewed as a discrete simplicial group ) , we have @xmath570 \cong \gamma_{g } = \o[\rep_g(\gamma ) ] \ .\ ] ] this justifies our notation and terminology for @xmath563 .
we record one useful property of the derived representation functor that will play a role in computations in section [ sec7 ] .
the functor @xmath571 preserves homotopy pushouts .
the proof of lemma [ hopushout ] requires some technical results about simplicial commutative algebras ; we therefore defer it to appendix [ scalgf ] ( see proposition [ reedy_smooth_colim ] and corollary [ corhpo ] ) .
we now define the representation homology of a ( reduced ) simplicial set by mimicking pirashvili s definition of higher hochshild homology .
our starting point is the known fact ( see , e.g. , ( * ? ? ?
* sect . 5 ) ) that the category of commutative hopf algebras over a field @xmath4 is equivalent to the category of @xmath4-algebras of the prop of finitely generated free groups .
to be precise , let @xmath106 denote the full subcategory of @xmath572 whose objects the free groups based on the sets @xmath573 for @xmath108 .
we denote such groups by @xmath574 ( where , by convention , @xmath575 is the identity group ) and write @xmath576 for the corresponding group algebras over @xmath4 .
the category @xmath106 is a prop , with monoidal product @xmath444 being the free product of groups , so that @xmath577 .
a commutative hopf algebra @xmath578 over @xmath4 defines the ( strong monoidal ) covariant functor @xmath579 , @xmath580 , which we denote by @xmath581 .
the assignment @xmath582 gives an equivalence between the category of commutative hopf algebras over @xmath4 and the category of @xmath4-algebras over the prop @xmath106 .
dually , the category of _ cocommutative _ hopf algebras is equivalent to the category @xmath4-algebras over the opposite prop @xmath583 .
now , observe that for any commutative hopf algebra @xmath111 , the functor @xmath584 takes values in the category of commutative algebras , _
i.e. _ , it can be viewed as a functor @xmath585 .
we extend this last functor to the category @xmath115 of all free groups by taking the left kan extension along the inclusion @xmath123 . to be precise ,
let @xmath115 denote the category of _ based _ free groups whose objects are pairs @xmath586 , where @xmath587 is a free group with a specified generating set @xmath117 , and morphisms are arbitrary group homomorphisms @xmath588 ( not necessarily , preserving the generating sets ) .
we have the natural inclusion functor @xmath589 that takes @xmath590 to @xmath591 .
the kan extension of @xmath581 along @xmath592 then defines a functor @xmath593 that assigns to the free group @xmath594 on a set @xmath117 the commutative algebra @xmath595 ( _ cf . _ ) .
we continue to denote this functor by @xmath581 .
let @xmath276 be a reduced simplicial set ( or equivalently , a connected pointed cgwh topological space ) .
recall that the kan loop group construction gives a functor @xmath596 that takes @xmath308 \in \delta^{\mathrm{op } } $ ] to the free group @xmath597 based on the set @xmath598 .
now , given a commutative hopf algebra @xmath578 , we consider the composition of functors @xmath599 which defines a simplicial commutative algebra @xmath600 .
the _ representation homology _ of @xmath60 in @xmath601 is defined by @xmath602\,=\ , \h_{\ast}[n(\underline{\mathcal h}(\lgr{x}))]\,\text{.}\ ] ] if @xmath603 is an affine group scheme with coordinate ring @xmath604 , we also write @xmath605 for @xmath606 .
clearly , a morphism @xmath607 of reduced simplicial sets induces a map of graded commutative algebras @xmath608 .
thus , representation homology defines a covariant functor @xmath609 .
the following proposition justifies the above definition of representation homology . for any @xmath53
, there is a natural isomorphism of graded commutative algebras @xmath610\,\text{.}\ ] ] in particular , @xmath611 , where @xmath612 is the fundamental group of @xmath172 .
if @xmath613 , we have natural isomorphisms @xmath614 for any set @xmath117 .
this implies that @xmath615 in @xmath616 . on the other hand , by proposition [ kanprop ] , the simplicial group
@xmath527 is semi - free , and hence a cofibrant object in @xmath479 .
this implies that @xmath617 in @xmath618 .
thus , @xmath619 is represented by the simplical commutative algebra @xmath620 , so @xmath621 \cong \pi_{\ast}[\underline{\mathcal h}(\lgr{x})]\,$ ] .
the isomorphism for @xmath622 is obtained by composing with and the natural isomorphism of groups @xmath623 .
let @xmath30 be a discrete group , and let @xmath624 be the classifying space ( _ i.e. _ , the simplicial nerve ) of @xmath509 . as a simple application of proposition [ rephom0 ] , we get @xmath625\,$ ]
. in particular , @xmath626 .
the kan adjunction gives the canonical cofibrant resolution @xmath627 in @xmath479 .
since @xmath5 is discrete , we have @xmath628 , and the result follows from proposition [ rephom0 ] . for any @xmath629
, there is a natural isomorphism @xmath630 recall that the wedge sum is a ( categorical ) coproduct in @xmath275 . since @xmath494 is a left adjoint functor , we have @xmath631 . by lemma [ hopushout ]
, it follows that @xmath632 .
the desired result is now immediate from knneth s theorem and proposition [ rephom0 ] .
now , we introduce the functor categories @xmath633-@xmath454 and @xmath454-@xmath633 , whose objects are all covariant ( resp . , contravariant ) functors from @xmath106 to the category of vector spaces .
we regard these objects as left and right modules over @xmath106 , respectively .
both categories are abelian with sufficiently many projective and injective objects .
there is a natural bifunctor @xmath634 which is right exact with respect to each argument , preserves sums and is left balanced in the sense of @xcite .
explicitly , this bifunctor can be defined by formula with @xmath101 replaced by @xmath106 . since @xmath635 is left balanced , the derived functors with respect to each argument are naturally isomorphic , and we denote their common value by @xmath636 .
note that for any left @xmath633-module @xmath637 , the functor @xmath638-@xmath639 is left adjoint to the functor @xmath640-@xmath109 , where @xmath641 is the right @xmath633-module @xmath642 for any vector space @xmath643 .
similarly , for any right @xmath633-module @xmath644 , the functor @xmath645 is left adjoint to the functor @xmath646-@xmath454 .
hence , both functors @xmath647 and @xmath648 commute with colimits . to state our first theorem
we need some notation .
first , we recall that if @xmath509 is any group , @xmath649 $ ] is a cocommutative hopf algebra : thus , @xmath649 $ ] defines a right @xmath633-module in @xmath454-@xmath633 .
now , if @xmath60 is a reduced simplicial set , @xmath650 $ ] defines a simplicial right @xmath633-module in @xmath651-@xmath633 . applying the normalization functor @xmath652 to this simplicial module
, we get a chain complex of @xmath633-modules and hence an object in the derived category @xmath653 . abusing notation
, we will denote this object by @xmath654 } ) $ ] .
for any @xmath53 and any commutative hopf algebra @xmath601 , there is a natural isomorphism of graded commutative algebras @xmath655 } ) \otimes^{\l}_{\ffgr } \underline{\mathcal h}]\ .\ ] ] to prove theorem [ rephomdtp ] we need a simple lemma .
recall that for @xmath656 , we denote by @xmath576 the group algebra of the free group based on the set @xmath657 . regarding it as a cocommutative hopf algebra
, we get a right @xmath633-module which ( to simplify the notation ) we also denote by @xmath658 . for each @xmath108 ,
the @xmath633-module @xmath659 is a projective object in @xmath660 . for a fixed @xmath108 , let @xmath661 $ ] denote the standard right @xmath633-module associated to the object @xmath662
this is a projective object in @xmath660 . indeed , by yoneda lemma
, there is a natural isomorphism @xmath663 for any @xmath664 .
the sequence of @xmath633-modules @xmath665 is exact in @xmath660 if and only if the sequence of @xmath4-vector spaces @xmath666 is exact for all @xmath108 .
it follows that @xmath667 is an exact functor , and hence @xmath668 is a projective object in @xmath669 . on the other hand , for any @xmath670 , we have @xmath671 \,\cong\ , k[{\langle n \rangle}^{\times m } ] \,\cong\,[k{\langle n \rangle}]^{\otimes m } \,=\,{k{\langle n \rangle}}({\langle m \rangle})\ , \ ] ] which shows that @xmath672 as right @xmath633-modules .
this finishes the proof of the lemma . by lemma [ gammangrp ] , for any @xmath108
, @xmath673 is a projective right @xmath633-module such that @xmath674 .
since colimits of projective modules are flat and commute with left kan extensions , this implies that @xmath675 is a _ flat _
right @xmath633-module and @xmath676 for any set @xmath117 . extending the last isomorphism levelwise to simplicial sets , we get an isomorphism of simplicial vector spaces @xmath677 \otimes_{\ffgr } \underline{\mathcal h } \,\cong\ , \underline{\mathcal h}(\lgr{x } )
further , since each @xmath678}$ ] is a flat right @xmath633-module , the normalized chain complex @xmath654 } ) $ ] is a complex of flat @xmath633-modules , hence we have a natural isomorphism in the derived category @xmath653 : @xmath679 } ) \otimes^{\l}_{\ffgr } \underline{\mathcal h}\,\cong\ , \underline{n}({k[\lgr{x } ] } ) \otimes_{\ffgr } \underline{\mathcal h } \,\cong\ , n(\underline{\mathcal h}(\lgr{x}))\,\text{.}\ ] ] at the homology level , this induces the desired isomorphism of theorem [ rephomdtp ] . for our next theorem
we recall that the singular chain complex @xmath680 of the ( moore ) loop space @xmath681 of a pointed topological space @xmath60 has a natural structure of a cocommutative dg hopf algebra .
the coproduct on @xmath682 is induced by the alexander - whitney diagonal , while the product comes from the structure of a topological monoid on @xmath683 via the eilenberg - zilber map ( see , e.g. , ( * ? ? ?
* section 26 ) ) . thus , the homology @xmath133 of @xmath478 is a graded cocommutative hopf algebra called the _
pontryagin algebra _ of @xmath60 .
now , any graded cocommutative hopf algebra @xmath684 defines a graded right @xmath633-module @xmath685 ( _ i.e. _ , a contravariant functor from @xmath106 to the category of graded vector spaces ) . for @xmath135 ,
we let @xmath686 denote the graded component of @xmath687 of degree @xmath138 ; thus , @xmath140 is a right @xmath633-module that assigns @xmath688_q $ ] , the @xmath138-th graded component of the graded vector space @xmath689 .
note that the @xmath633-module @xmath686 depends on _ all _ graded components of the hopf algebra @xmath690 , and not solely on @xmath691 . with this notation , we can now state our second theorem , which is an analogue of ( * ? ? ?
* theorem 2.4 ) for representation homology .
there is a natural first quadrant spectral sequence @xmath692 converging to the representation homology of @xmath60 .
recall from the proof of theorem [ rephomdtp ] that @xmath693 } ) $ ] is a non - negatively graded chain complex of flat right @xmath633-modules .
hence , for any left @xmath694-module @xmath581 , there is a standard ` hypertor ' spectral sequence ( see , e.g. , ( * ? ? ?
* application 5.7.8 ) ) : @xmath695})],\,\underline{\mathcal h } ) \ \underset{p}{\longrightarrow}\ \h_{p+q}\,[\underline{n}({k[\lgr{x } ] } ) \otimes_{\ffgr } \underline{\mathcal h}]\ .\ ] ] by theorem [ rephomdtp ] , the limit of this spectral sequence is isomorphic to @xmath696 . to prove the theorem we need only to show that @xmath697 } ) ] \cong \underline{\h}_{\,\ast}(\omega x;k ) \,$ ] as graded right @xmath633-modules . by kan
s theorem [ kanstheorem ] , @xmath698 is a topological group , which is weakly equivalent , as an @xmath699-space , to the based loop space @xmath478 .
this implies , in particular , that @xmath700)]\cong \h_{\ast}(\omega x;k ) $ ] as graded hopf algebras , and hence @xmath701)]\cong \underline{\h}_{\,\ast}(\omega x;k)\ , $ ] as graded @xmath633-modules .
note that @xmath702)\,$ ] stands here for the normalized chain complex of the simplicial hopf algebra @xmath650 $ ] , while @xmath703})\,$ ] in the above spectral sequence denotes the normalized chain complex of the simplicial @xmath633-module @xmath704 $ ] .
we need to check that @xmath705 } ) ] \cong \underline{\h}_{\,\ast } [ n({k[\lgr{x } ] } ) ] \,$ ] as graded @xmath633-modules .
now , the simplicial @xmath633-module @xmath706}$ ] assigns to @xmath707 the simplicial vector space @xmath708^{\otimes m } = \ { k[\lgr{x}_n]^{\otimes m } \}_{n \ge 0 } $ ] . by the eilenberg - zilber theorem ,
the normalized chain complex of this simplicial vector space is homotopy equivalent to @xmath709)^{\otimes m}$ ] , while , by kunneth s formula , the homology of @xmath702)^{\otimes m}\,$ ] is naturally isomorphic to @xmath710)]^{\otimes m}$ ] .
this shows that @xmath711}))({\langle m \rangle})\,\cong\,\h_{\ast } [ n({k[\lgr{x}]})]^{\otimes m } $ ] for any @xmath712 , completing the proof of the theorem .
theorem [ rephomdtp2 ] has several interesting implications .
first , we consider one important special case when the spectral sequence collapses at @xmath713-term .
let @xmath5 be a discrete group .
then , for any affine algebraic group @xmath34 , there is a natural isomorphism @xmath714 , \o(g))\ .\ ] ] in particular , @xmath715 \otimes_{\ffgr } \o(g)\,$ ] .
the classifying space @xmath716 is an eilenberg - maclane space of type @xmath65 .
its loop space @xmath478 is homotopy equivalent to @xmath5 , where @xmath5 is considered as a discrete topological space .
hence , @xmath717 for all @xmath718 , while @xmath719 $ ] as a hopf algebra .
thus , for @xmath716 , the spectral sequence collapses on the @xmath720-axis , giving the required isomorphism . combining the isomorphisms of corollary [ compwithdrep ] and corollary [ bgamma ]
, we can express the representation homology of @xmath5 ( originally defined as a non - abelian derived functor ) in terms of classical abelian homological algebra : @xmath721 \cong { { \rm{tor}}}_{\ast}^{\ffgr}(k[\gamma],\o(g ) ) \ .\ ] ] in degree 0 , we have a natural isomorphism expressing the coordinate ring of the representation variety @xmath8 as a functor tensor product : @xmath722 \cong k[\gamma ] \otimes_{\ffgr } \o(g)\ .\ ] ] this last isomorphism was found in @xcite , and it was one of the starting points for the present paper .
the result of theorem [ rephomdtp2 ] holds for any ( not necessarily , monoidal ) left @xmath633-module .
in particular , if we take a _ reductive _ affine algebraic group @xmath3 and define a left @xmath633-module @xmath723 by the formula @xmath724^g
= \o(g \times \stackrel{n}{\ldots } \times g)^g\,$ ] , then , for any @xmath725 , we obtain a homology spectral sequence @xmath726 converging to the @xmath3-invariant part of representation homology of @xmath60 .
the proof of corollary [ bgamma ] shows that , for @xmath716 , the spectral sequence collapses on the @xmath720-axis , giving an isomorphism @xmath727 , \o(g)^g)\ .\ ] ] in degree @xmath245 , we therefore have @xmath728^g \cong k[\gamma ] \otimes_{\ffgr } \o(g)^g $ ] . to state futher consequences of theorem [ rephomdtp2 ] we introduce some terminology .
we will say that a map @xmath729 of pointed topological spaces is a _ pontryagin equivalence _
( over @xmath4 ) if it induces an isomorphism @xmath730 of pontryagin algebras ( or equivalently , a quasi - isomorphism @xmath731 of dg hopf algebras ) .
the next result is obtained by applying to a standard comparison theorem for homology spectral sequences ( see ( * ? ? ?
* theorem 5.1.12 ) ) .
if @xmath729 is a pontryagin equivalence , the induced map on representation homology @xmath732 is an isomorphism for any hopf algebra @xmath111 .
we remark that corollary [ rephomhlinv ] does not say that an _ arbitrary _ isomorphism of hopf algebras @xmath730 gives an isomorphism @xmath733 .
( indeed , an abstract isomorphism of pontryagin algebras need not even induce a map on representation homology . ) still , corollary [ bgamma ] shows that if both @xmath60 and @xmath525 are aspherical spaces , then any isomorphism of pontryagin algebras induces an isomorphism on representation homology .
next , we recall that the singular chain complex @xmath734 of any space @xmath60 is naturally a dg coalgebra with comultiplication defined by the alexander - whitney diagonal . moreover , if @xmath60 is path - connected , there is a quasi - isomorphism of dg coalgebras ( see ( * ? ? ? * theorem 6.3 ) ) @xmath735\ , \ ] ] where @xmath736 is the classical bar construction .
since @xmath737 preserves quasi - isomorphisms , any pontryagin equivalence @xmath738 of path - connected spaces is necessarily a homology equivalence , _
i.e. _ , it induces an isomorphism on singular homology @xmath739 .
the converse is not always true unless @xmath272 and @xmath307 are simply - connected . in the latter case
, we have the following well - known result ( _ cf . _
* part i , prop .
let @xmath740 be a map of simply - connected pointed topological spaces .
the following conditions are equivalent : 1 .
@xmath384 is a rational homology equivalence : i.e. @xmath741 ; 2 .
@xmath384 is a rational pontryagin equivalence : i.e. @xmath742 ; 3 .
@xmath384 is a rational homotopy equivalence : i.e. @xmath743 .
the equivalence @xmath744 follows a classical theorem of adams @xcite that asserts that , for any simply - connected space @xmath60 , there is a quasi - isomorphism of dg algebras : @xmath745\,$ ] , where @xmath746 is the cobar construction . to prove that @xmath747 we first recall that , for any simply - connected @xmath60 , the @xmath173-vector space @xmath748 carries a natural bracket ( called the whitehead product ) making it a graded lie algebra is called the _ homotopy lie algebra _ of @xmath60 . ] .
thus , a map @xmath729 is a rational homotopy equivalence if and only if it induces an isomorphism of lie algebras @xmath749 .
then , a classical theorem of milnor and moore ( see ( * ? ? ? * theorem 21.5 ) ) implies that the hurewicz homomorphism @xmath750 induces an isomorphism of graded hopf algebras @xmath751 , where @xmath752 is the universal enveloping algebra of @xmath753 .
this yields the equivalence @xmath747 .
we say that a map @xmath754 of simply - connected spaces is a _ rational homotopy equivalence _ if the equivalent conditions of lemma [ equivl ] hold .
a rational homotopy equivalence induces an isomorphism on representation homology .
thus , @xmath755 depends only on the rational homotopy type of @xmath60 . by lemma [ equivl]@xmath756
, a rational homotopy equivalence @xmath757 induces an isomorphism @xmath758 . since @xmath759 , we have @xmath760 , and the universal coefficient theorem implies that @xmath761 .
the claim then follows from corollary [ rephomhlinv ] .
next , we look at higher connected spaces . recall that a space @xmath60 is said to be _
@xmath90-connected _ if @xmath60 is path - connected and its first @xmath762 homotopy groups vanish , _
i.e. _ @xmath763 for @xmath764 .
let @xmath272 be an @xmath90-connected space for some @xmath765 , and let @xmath766 .
then @xmath767 0 & \mbox{\rm for } & 1 \leq q < n\\*[1ex ] \h_{q+1}(x ; \g^ * ) & \mbox{\rm for } & n \leq q < 2n \end{array } \right.\ ] ] where @xmath768 is the lie algebra of @xmath3 and @xmath165 is its @xmath4-linear dual .
if a space @xmath60 is @xmath90-connected , its homotopy lie algebra @xmath769 is @xmath90-reduced , _
i.e. _ @xmath770 for @xmath771 .
since @xmath772 and @xmath760 , we have @xmath773 , @xmath774 for @xmath775 , and @xmath776 where the last isomorphism is a consequence of the rational hurewicz theorem ( see , e.g. , @xcite ) .
now , recall that for a fixed @xmath777 , the right @xmath633-module @xmath778 is defined as the functor @xmath779 , @xmath780_q $ ] .
it follows from this definition that @xmath781 0 & \mbox{\rm for } & 1 \leq q \leq n-1\\*[1ex ] \lin_k^ * \otimes \h_{q+1}(x ; k ) & \mbox{\rm for } & n \leq q \leq 2n-1 \end{array } \right.\ ] ] where @xmath782 is the linearization functor : @xmath783 and @xmath784 denotes its composition with linear duality .
thus , for @xmath60 @xmath90-connected , the @xmath785-terms of the spectral sequence can be identified as @xmath786 by lemma [ gammangrp ] , the right @xmath633-module @xmath787 is projective .
hence , @xmath788 for @xmath789 . on the other hand , by proposition [ tors ]
( see section [ pfconj1 ] below ) , @xmath790 , while @xmath791 for @xmath792 .
hence , for @xmath793 , we have @xmath794 the vanishing of @xmath795 for all @xmath796 in the range @xmath797 shows that the spectral sequence collapses on the @xmath138-axis for these values of @xmath138 .
thus , we have @xmath798 for @xmath797 . by and , these are the desired isomorphisms . for an arbitrary ( pointed connected ) topological space @xmath60 , the 5-term exact sequence associated to the spectral sequence reads @xmath799,\ , \underline{\mathcal h } ) \to \underline{\h}_{\,1}(\omega x ; k ) \otimes_{\ffgr } \underline{\mathcal h } \to \hr_1(x , \mathcal h ) \to { { \rm{tor}}}_{1}^{\ffgr}(k[\pi_1(x)],\ , \underline{\mathcal h } ) \to 0\ ] ] if the fundamental group @xmath800 is free , then by lemma [ gammangrp ] , the corresponding right @xmath633-module is flat , and hence in that case , we have an isomorphism @xmath801 it would be interesting to give an explicit interpretation of @xmath802 for an arbitrary @xmath60 similar to in the case of simply - connected spaces . in this section ,
we give another construction of representation homology that does not use the kan equivalence .
our starting point is a general categorical principle which asserts that any left adjoint functor on the category of simplicial sets with values in a ( cocomplete ) category @xmath250 arises from a cosimplicial object in @xmath250 .
specifically , a cosimplicial object @xmath803 gives rise to the simplicial adjunction @xmath804 where @xmath805 denotes the left kan extension of @xmath806 along the yoneda embedding @xmath807 , and @xmath808 is the functor assigning to @xmath809 the simplicial set @xmath810 ) , a)\}_{n \ge 0 } $ ] .
this gives an equivalence between the category @xmath811 of cosimplicial objects in @xmath250 and the category of simplicial adjunctions with values in @xmath812 ( see , e.g. , ( * ? ? ?
3.1.5 ) ) .
the fundamental example is the cosimplicial space @xmath813 defined by the geometric simplices @xmath814 . under the above equivalence , it corresponds to the classical adjunction between simplicial sets and topological spaces : @xmath815 where @xmath816 is the geometric realization functor defined in section [ geomreal ] . in general , it is therefore natural to think of functors of the form @xmath817 as _ realization functors _ of simplicial sets in @xmath812
. now , let us take an affine algebraic group @xmath34 and consider its classifying space @xmath818 .
this is naturally a simplicial affine @xmath4-scheme , and its coordinate ring @xmath819 is a cosimplicial commutative @xmath4-algebra .
the next proposition shows that the ` realization ' functor corresponding to @xmath820 is just the classical representation functor @xmath550 ( see section [ secrephom ] ) . for any @xmath53
, there is a natural isomorphism of commutative algebras @xmath821 \ ] ] in particular , if @xmath509 is a discrete group , then @xmath822 . for any @xmath823
, we have canonical isomorphisms @xmath824 here , the third and fourth isomorphisms follow from the fact that @xmath825 is a _ discrete _ simplicial group , so that @xmath826 and @xmath827 .
the desired proposition follows now from yoneda lemma .
proposition [ tensoroverdelta ] suggests the possibility of defining representation homology of spaces in terms of the ( non - abelian ) derived tensor product @xmath828 .
we briefly outline this construction and refer the reader to @xcite for more details and proofs .
we will need some technical results from abstract homotopy theory for which we will refer to @xcite and @xcite .
from now on , we let @xmath829 denote the category of simplicial commutative @xmath4-algebras .
this is a simplicial model category , tensored and cotensored over @xmath274 .
given two functors @xmath830 and @xmath831 , we recall that the functor tensor product @xmath832 is defined by the coequilizer ( _ cf . _
* ( 4.1.1 ) ) ) @xmath833 \to [ m ] } x([m ] ) \otimes f([n ] ) \
\underset{f_*}{\overset{f^*}{\rightrightarrows}}\ \coprod_{[n ] \in \delta } x([n ] ) \otimes f([n ] ) \right)\ .\ ] ] the assignment @xmath834 defines a bifunctor @xmath835 , which we would like to replace by a homotopy invariant derived functor . to this end
, we equip the categories @xmath836 and @xmath837 with reedy model structures , where the weak equivalences are given by the pointwise weak equivalences of diagrams in the underlying model categories @xmath274 and @xmath838 .
relative to the reedy product model structure on @xmath839 and the natural ( simplicial ) model structure on @xmath840 , the functor tensor product @xmath841 then becomes a quillen bifunctor ( see ( * ? ? ?
* theorem 19.7.2 and corollary 19.7.4 ) ) .
this means , in particular , that @xmath841 is a left quillen functor with respect to each of its arguments , provided the other argument is chosen to be a reedy cofibrant object .
moreover , there is a well - defined total left derived functor @xmath842 now , it is known ( see ( * ? ? ?
* corollary 15.8.8 ) ) that the category @xmath843 is reedy cofibrant , _
i.e. _ every simplicial object in @xmath274 is cofibrant with respect to the reedy model structure on @xmath843 .
hence , to compute the derived tensor product @xmath844 it suffices only to replace the diagram @xmath370 by its reedy cofibrant model in @xmath845 .
if @xmath831 is a pointwise cofibrant diagram ( with respect to the natural model structure ) in @xmath846 , then @xmath847 has a _
canonical _ reedy cofibrant model in @xmath845 given by its bar construction ( see ( * ? ? ?
4.2 ) ) . for any @xmath53
, there is a natural isomorphism in @xmath848 : @xmath849 the cosimplicial commutative algebra @xmath850 is given ( in cosimplicial degree @xmath203 ) by the commutative algebra @xmath851 this cosimplicial commutative algebra has a canonical resolution by a cosimplicial simplicial commutative algebra , denoted @xmath852 , which is given ( in cosimplicial degree @xmath203 ) by the simplicial commutative algebra @xmath853 ) \ , \otimes \,(\co(g ) \otimes \delta[m-1 ] ) \,\otimes\ , \ldots \ , \otimes \ , ( \mathcal{o}(g ) \otimes \delta[0])\ ] ] we refer the reader to @xcite for a precise definition of @xmath854 .
one can show that there is an isomorphism of simplicial commutative algebras @xmath855 although the cosimplicial simplicial commutative algebra @xmath852 is not reedy cofibrant , using , one can show that it is reedy smooth in the sense of appendix [ scalgf ] , definition [ reedy_smooth ] .
theorem [ tenrep ] is therefore a consequence of proposition [ reedy_smooth_tensor ] .
* remark 1 .
* formula may be thought of as an analogue of the loday construction for higher hochschild homology ( see section [ phhfromfin ] ) .
indeed , the loday construction can be interpreted as follows .
given a commutative algebra @xmath856 , consider the functor @xmath857 , defining a cosimplicial object in @xmath846 .
then , it follows from lemma [ defequiv ] that @xmath858 $ ] for any smooth commutative algebra @xmath16 .
the results of the next section suggest that , for any space @xmath60 , there is a homotopy equivalence @xmath859 * remark 2 . * instead of resolving @xmath860 by a cosimplicial simplicial commutative algebra
, one could resolve it by a cosimplicial commutative dg algebra @xmath861 .
the resulting tensor product @xmath862 will then correspond to the derived representation scheme under the quillen equivalence @xmath863 ( see theorem [ nnstar_equiv ] ) . as in the case of simplicial commutative algebras
, it suffices to require the resolution @xmath864 to be reedy smooth in @xmath865 , meaning that all the latching maps are required to be smooth extensions of commutative dg algebras in an appropriate sense ( see definition [ reedy_smooth ] ) .
the simplicial dg scheme @xmath60 corresponding to a reedy smooth cosimplicial commutative dg algebra @xmath866 is called _ injective _ in @xcite .
thus , the proof of theorem [ tenrep ] shows that the homotopy type of the derived representation scheme coincides with the homotopy type of the derived space @xmath867 of local systems defined in @xcite .
in section [ rhs ] , we defined representation homology by analogy with hochschild homology , using kan s simplicial loop group construction . in this section , we establish a direct relation between these two homology theories using another classical construction in simplicial homotopy theory due to j. milnor @xcite .
we begin by recalling a standard simplicial model for a ( reduced ) suspension @xmath868 of a space @xmath60 .
the _ suspension functor _ on pointed simplicial sets is defined by @xmath869 where @xmath870 is the reduced cone over @xmath60 . for a pointed simplicial set @xmath871 ,
the set of @xmath90-simplices in @xmath872 is given by @xmath873 with all @xmath874 being identified to @xmath151 .
the face and degeneracy maps in @xmath872 are defined by @xmath875 ( d^x_{i - m}(x ) , m ) & \mbox { if $ \ m \leq i \leq n$ } \end{array } \right .
\\*[2ex ] s_j\,:\ , c(x)_n { \xrightarrow{}}c(x)_{n+1}\ , , & & ( x , m)\mapsto \left\ { \begin{array}{ll } ( x , m+1 ) & \mbox { if $ \ 0 \leq j < m$}\\*[1ex ] ( s^x_{j - m}(x ) , m ) & \mbox { if $ \ m \leq j \leq n$ } \end{array } \right . \\\end{aligned}\ ] ] where @xmath876 for all @xmath877 . the embedding @xmath878 is given by @xmath879 , and @xmath880 is defined to be the corresponding quotient set .
note that , unlike @xmath881 , the simplicial set @xmath868 is reduced , since @xmath882 in @xmath73 for all @xmath883 ( in particular , we have @xmath884 ) .
now , for any pointed simplicial set @xmath60 , there is a homotopy equivalence @xmath885 , where @xmath886 is reduced suspension of the geometric realization of @xmath60 in the usual topological sense .
the next two theorems constitute the main result of this section . for any commutative hopf algebra @xmath601 and any _ pointed _
simplicial set @xmath60 , there is a natural isomorphism of graded commutative algebras @xmath887 to state the next theorem , we recall that there is a natural way to make an arbitrary simplicial set pointed by adding to it a disjoint basepoint .
to be precise , the forgetful functor @xmath888 has a left adjoint @xmath889 obtained by extending to simplicial sets the obvious functor @xmath890 on the category of sets .
explicitly , if @xmath891 is a simplicial set , then @xmath892 for all @xmath90 , and the face and degeneracy maps of @xmath893 are the ( unique ) basepoint - preserving extensions of the corresponding maps of @xmath60 .
being a left adjoint , the functor @xmath894 commutes with colimits ; in particular , we have @xmath895 where @xmath896 is the space obtained from @xmath355 by adjoining a basepoint . for any commutative hopf algebra @xmath601 and any simplicial set @xmath60
, there is an isomorphism of graded commutative algebras @xmath897 the proofs of theorem [ repvshh1 ] and theorem [ repvshh ] are based on a classical simplicial group model of the spaces @xmath898 , which we now briefly review . for a pointed simplicial set @xmath899
, we define @xmath900 .
then , by kan s theorem [ kanstheorem ] , there is a homotopy equivalence of spaces @xmath901 the following observation is due to j. milnor @xcite ( see also ( * ? ? ?
* theorem v.6.15 ) ) . for any @xmath902
, @xmath903 is a semi - free simplicial group generated by the simplicial set @xmath9 with basepoint identified with @xmath904 , i.e. @xmath905 the face and degeneracy maps are induced by the face and degeneracy maps of @xmath11 . by definition of the reduced suspension , we have @xmath906 for all @xmath907 and @xmath908 for all @xmath909 .
hence , @xmath910 , with @xmath911 being the basepoint .
it follows that @xmath912 to calculate the face and degeneracy maps , we recall from section [ klg ] that @xmath913 and @xmath914 for @xmath915 .
similarly , @xmath916 for all @xmath917 .
this proves the desired lemma .
recall that , for a commutative hopf algebra @xmath601 , we denote by @xmath918 the functor @xmath919 on the category of based free groups obtained from the @xmath633-module @xmath920 by taking its left kan extension along the inclusion @xmath921 ( see section [ rhs ] ) .
there is an isomorphism of functors from @xmath344 to @xmath922 : @xmath923 where @xmath601 in the right - hand side is regarded as a commutative @xmath4-algebra . by lemma [ miln ] , for any simplicial set @xmath924
, there are natural isomorphisms of groups @xmath925_n\,\cong\ , { \langle x_n \rangle}\,$ ] , @xmath926 , with structure maps on @xmath927 being compatible with those of @xmath60 . by applying the functor @xmath578
, we thus get isomorphisms of simplicial commutative algebras @xmath928_\ast ) \,\cong\ , \underline{\mathcal h}[{\langle x_\ast \rangle } ] \,\cong\ , x_\ast \otimes \mathcal h\ , \ ] ] which are obviously functorial in @xmath60 .
this proves the proposition .
theorem [ repvshh ] is an immediate consequence of the above proposition . to prove theorem [ repvshh1 ]
, we first note that , although the unreduced cone on a space @xmath60 coincides with the reduced cone on @xmath893 , the corresponding suspensions differ . instead , for any pointed space @xmath60
, there is a homotopy equivalence ( see @xcite ) @xmath929 from this we can deduce the following for a pointed topological space @xmath60 , there is a natural isomorphism @xmath930 applying corollary [ wedgeprod ] to , we have @xmath931 .
now , since @xmath932 , theorem [ repvshh ] implies @xmath933 , where @xmath111 is concentrated in degree @xmath245 .
it follows that @xmath934 as desired .
lemma [ unpsusp ] shows that @xmath935 . combining this last isomorphism with that of theorem [ repvshh ]
, we now conclude @xmath936 this proves theorem [ repvshh1 ] .
we conclude this section with a few simple examples illustrating the use of theorems [ repvshh1 ] and [ repvshh ] .
more examples will be given in the next two sections .
in what follows , @xmath3 denotes an arbitrary affine algebraic group and @xmath937 stands for its lie algebra .
the representation homology of the circle @xmath938 is given by @xmath939 and @xmath940 for @xmath941 .
this follows , for example , from lemma [ gammangrp ] and corollary [ bgamma ] ( since @xmath942 ) .
now , for higher dimensional spheres , we have @xmath943)\,$ ] for all @xmath944 .
note that @xmath945 for all @xmath946 . by theorem [ repvshh1 ]
, we conclude @xmath947 ) \otimes_{\mathcal o(g ) } k \
, \cong \ , { { \rm{sym}}}_k(\g^{\ast}[n-1])\,\text{,}\\\ ] ] where the second isomorphism follows from ( * ? ? ?
* section 5.5 ) .
we now generalize the previous example to arbitrary suspensions .
let @xmath868 be the suspension of a pointed connected space @xmath60 of finite type .
then @xmath948 where @xmath949 stands for the reduced @xmath153singular@xmath154 homology of @xmath60 with constant coefficients in @xmath165 . consequently , by induction , @xmath950\right)\ , \quad \forall\,n\ge 1 \ .\ ] ] it is known ( see ( * ? ? ? * theorem 24.5 ) ) that @xmath73 is rationally homotopy equivalent to a bouquet of spheres : @xmath951 , where each @xmath952 have dimension @xmath953 .
by proposition [ rephomhtpinv ] , it thus suffices to compute @xmath954 for @xmath955 .
note that the reduced homology @xmath956 of @xmath117 is isomorphic to @xmath957 with trivial coproduct , where @xmath958 is a basis element of homological degree @xmath959 .
now , by corollary [ wedgeprod ] and proposition [ rephomsph ] , we have @xmath960 ) \\ & \cong & { { \rm{sym}}}_k \biggl(\bigoplus_{i \in i } \g^{\ast}[n_i -1 ] \biggr ) \ , \cong \ , { { \rm{sym}}}_k\biggl(\bigoplus_{n \geq 2 } \g^{\ast } \otimes \h_n(\sigma x;k)[n-1]\biggr)\\ & \cong & \ , { { \rm{sym}}}_k\biggl(\bigoplus_{n \geq 1 } \g^{\ast } \otimes \h_n(x;k)[n]\biggl)\,\cong\ , { { \rm{sym}}}_k \left[\g^\ast \otimes { \overline{\mathrm{h}}}_\ast(x;k)\right]\ , \cong \ , { { \rm{sym}}}_k \ , { \overline{\mathrm{h}}}_\ast(x ; \g^ * ) \ , \end{aligned}\ ] ] where the last isomorphism is a consequence of the universal coefficient theorem .
proposition [ rephomsusp ] actually holds at the chain level : namely , the derived representation schemes of the suspensions @xmath880 of connected spaces are formal , _ i.e. _ dg algebra models representing these derived schemes are quasi - isomorphic to their homology . as a consequence of theorem [ repvshh ] ,
lemma [ unpsusp ] and proposition [ rephomsusp ] , we have the following general formula for the higher hochschild homology with coefficients in commutative hopf algebras . for any pointed
connected topological space @xmath60 of finite type , @xmath961\,\text{.}\ ] ] we remark that the isomorphism is a refinement of pirashvili s generalization of the classical hkr theorem which ( in our notation ) asserts that @xmath962 for any smooth commutative algebra @xmath16 ( _ cf . _
* theorem 4.6 ) ) .
since both sides of make sense with @xmath963 replaced by any smooth commutative algebra @xmath16 , it is natural ask when @xmath964\quad \mbox{for any connected space $ x\,$?}\ ] ] note that , by proposition [ pirhhsusp ] , holds for any polynomial algebra @xmath965 $ ] , @xmath966 .
in recent years , there have been a number of interesting _ topological _ generalizations of higher hochschild homology , such as factorization homology ( see , e.g. , @xcite ) and higher topological hochschild homology ( see @xcite ) .
it seems natural to expect that representation homology admits similar topological refinements , and the relation between higher hochschild homology and representation homology established in this section extends to this more general framework .
if @xmath60 is a simply - connected topoloical space of finite type , the rational homotopy type can be described by a differential graded lie algebra @xmath967 called a lie model of @xmath60 . in this section ,
we address the natural question : what is the relation between the representation homology of @xmath60 and that of the lie model @xmath967 ?
our main theorem , which we call comparison theorem , asserts that the two homology theories are isomorphic .
we begin by reviewing the definition of derived representation schemes of lie algebras and associated character maps , which we call the drinfeld traces . for details and proofs ,
we refer the reader to @xcite .
let @xmath178 be a fixed finite - dimensional lie algebra over @xmath4 .
given an ( arbitrary ) lie algebra @xmath968 , we are interested in classifying the representations of @xmath969 in @xmath178 . the corresponding moduli scheme @xmath970 is defined by its functor of points : @xmath971 assigning to a commutative @xmath4-algebra @xmath16 the set of families of representations of @xmath969 in @xmath178 parametrized by the @xmath4-scheme @xmath31 . the functor @xmath970 is represented by a commutative algebra @xmath972 , which has the following canonical presentation ( _ cf . _
6.1 ) ) : @xmath973 \otimes \xi^ * \,\rangle\!\rangle}\ , \ ] ] where @xmath165 is the vector space dual to @xmath178 and @xmath974 is the linear map @xmath975 dual to the lie bracket on @xmath178 .
the universal representation @xmath976 is given by the natural map @xmath977 \otimes \xi_i\ , \ ] ] where @xmath978 and @xmath979 are dual bases in @xmath178 and @xmath165 .
the algebra @xmath980 has a canonical augmentation @xmath981 induced by the zero map @xmath982 .
thus the assignment @xmath983 defines a functor with values in the category of augmented commutative algebras : @xmath984 we call the _ representation functor _ in @xmath178 .
geometrically , one can think of @xmath985 as a coordinate ring @xmath986 $ ] of the based affine scheme @xmath970 , with the basepoint corresponding to the trivial representation .
next , let @xmath3 be an affine algebraic group over @xmath4 associated with the lie algebra @xmath175 .
observe that for any @xmath968 , @xmath34 acts naturally on @xmath980 by automorphisms : this action is functorial in @xmath969 .
we write @xmath987 for the subfunctor of @xmath988 defined by taking the @xmath3-invariants : @xmath989 the algebra @xmath990 represents the affine quotient scheme @xmath991 parametrizing the closed orbits of @xmath34 in @xmath992 .
the functor can be extended to the category of dg lie algebras @xmath993 it is shown in @xcite that , for a fixed @xmath994 , the corresponding commutative dg algebra @xmath980 represents an affine dg scheme parametrizing the dg lie representations of @xmath969 in @xmath178 .
now , although the functor is not homotopy invariant ( it does not preserve quasi - isomorphisms and hence does not descend to @xmath995 ) , it is a left quillen functor and hence has a well - behaved left derived functor ( see ( * ? ? ?
* theorem 6.4 ) ) @xmath996 for a fixed dg lie algebra @xmath994 , we then define @xmath997 . in a similar fashion
, we construct the derived functor of the invariant functor @xmath998 : @xmath999 and define @xmath1000 .
note that , abusing notation , we write @xmath1001 and @xmath1002 for the commutative dg algebras in @xmath1003 instead of the eponymous derived schemes .
to compute @xmath1004 and @xmath1005 we use the cofibrant resolutions in @xmath1006 , which , in practice , are given by semi - free dg lie algebras .
thus , we have isomorphisms @xmath1007 where @xmath1008 is a(ny ) cofibrant resolution of @xmath1009 in @xmath1006 . finally , for any dg lie algebra @xmath969 , we define the _ representation homology _ of @xmath969 in @xmath178 by @xmath1010 \ .\ ] ] this is a graded commutative @xmath4-algebra , which depends on @xmath969 and @xmath178 ( but not on the choice of resolution of @xmath1011 ) .
if @xmath968 is an ordinary lie algebra , then there is an isomorphism of commutative algebras @xmath1012 which justifies the name ` derived representation scheme ' for @xmath1013 .
in addition , if @xmath3 is reductive , we have @xmath1014\ , \cong \ , \hr_{\ast}(\mathfrak{a},\g)^{g } .\ ] ] thus , the homology of @xmath1002 can be viewed as the invariant part of the representation homology of @xmath969 . our next goal is to describe certain natural maps with values in representation homology .
these maps can be viewed as ( derived ) characters of finite - dimensional lie representations . from now on
, we assume that @xmath178 is a reductive lie algebra over @xmath4 .
we let @xmath1015 denote the space of invariant polynomials on @xmath178 , and for each @xmath1016 , we write @xmath1017 for the subspace of homogeneous polynomials of degree @xmath1018 .
consider the symmetric ad - invariant @xmath4-multilinear forms on @xmath1019 of degree @xmath1020 .
every such form is induced from the universal one : @xmath1021 , which takes its values in the space @xmath1022 of coinvariants of the adjoint representation of @xmath969 in @xmath1023 .
the assignment @xmath1024 defines a ( non - additive ) functor on the category of lie algebras that extends in a natural way to the category of dg lie algebras : @xmath1025\ .\ ] ] it can be shown ( see ( * ? ? ?
* thm . 7.1 ) ) that the functor @xmath1026 has a ( left ) derived functor @xmath1027 which takes its values in the derived category @xmath1028 of @xmath4-complexes .
we write @xmath1029 for the homology of @xmath1030 and call it the _ lie - hodge homology _ of @xmath969 .
this terminology is justified by the fact that there is a natural direct sum decomposition ( see ( * ? ? ?
7.2 ) and ( * ? ? ?
1 ) ) @xmath1031 that is a koszul dual of the classical hodge decomposition of cyclic homology of commutative algebras . now , observe that , for any commutative algebra @xmath1032 , there is a natural symmetric invariant @xmath1033-linear form @xmath1034 on the current lie algebra @xmath1035 .
hence , by the universal property of @xmath1036 , we have a canonical map @xmath1037 { \xrightarrow{}}\lambda^{(d)}(\mathfrak{a } ) \otimes b \,\text{.}\ ] ] applying @xmath1036 to the universal representation and composing with , we define @xmath1038 & \rto & \lambda^{(d)}(\g ) \otimes \mathfrak{a}_{\g } \end{diagram}\,\text{.}\ ] ] on the other hand , for the lie algebra @xmath178 , we have a canonical ( nondegenerate ) pairing @xmath1039 induced by the linear pairing between @xmath165 and @xmath175 . replacing the lie algebra @xmath969 in by its cofibrant resolution @xmath1040 and using ,
we define the morphism of complexes @xmath1041 for a fixed polynomial @xmath1042 , this morphism induces a map on homology @xmath1043 , which we call the _ drinfeld trace _ associated does depend on the choice of @xmath1044 but we suppress this in our notation . ] to @xmath1044 .
it is easy to check that the image of is contained in the invariant subalgebra @xmath1045 of @xmath1046 , hence the drinfeld trace is actually a map @xmath1047 in this section for simplicity , we assume that @xmath1048 to use directly results from quillen s rational homotopy paper @xcite .
as explained in remark [ rem636 ] , the results of this section extend to an arbitrary field of characteristic @xmath245 by a universal coefficient argument .
[ [ section ] ] let @xmath60 be a 1-connected topological space of finite rational type .
recall ( _ cf . _
@xcite ) that one can associate to @xmath272 a commutative cochain dg algebra @xmath1049 , called a _ sullivan model _ of @xmath60 , and a connected ( chain ) dg lie algebra @xmath1050 , called a _ quillen model _ of @xmath60 .
each of these algebras is uniquely determined up to homotopy and each encodes the rational homotopy type of @xmath60 .
the relation between them is given by a dg algebra quasi - isomorphism @xmath1051 where @xmath1052 is the chevalley - eilenberg cochain complex of @xmath1050 .
the homology of @xmath1053 is the homotopy lie algebra @xmath1054 , while the cohomology of @xmath1055 is the rational cohomology algebra @xmath1056 of @xmath60 . among quillen models of @xmath60
, there is a _
minimal _ one given by a semi - free dg lie algebra @xmath1057 generated by a graded @xmath173-vector space @xmath643 with differential @xmath1033 satisfying @xmath1058\,$ ] . such a minimal model is determined uniquely up to ( noncanonical ) isomorphism . in particular , @xmath1059 $ ] ( see @xcite , p. 326 ) .
now , given an algebraic group @xmath3 , one can associate to a 1-connected space @xmath60 two representation homologies : the representation homology @xmath1060 of @xmath272 with coefficients in @xmath3 ( in sense of section [ rephomsect ] ) and the representation homology @xmath1061 of a lie model @xmath1050 of @xmath60 with coefficients in the lie algebra of @xmath3 ( in sense of section [ drep ] ) .
the following theorem , which is one of the main results of the paper , shows that the two constructions agree . for any affine algebraic group @xmath3 with lie algebra @xmath175
, there is an isomorphism of graded commutative @xmath173-algebras @xmath1062 theorem [ conj1 ] can be restated in terms of the sullivan model of @xmath60 .
recall that for any commutative ( dg ) algebra @xmath1063 , the tensor product @xmath1064 has a natural dg lie algebra structure with lie bracket defined by @xmath1065 \,=\,[\xi\,,\,\eta ] \otimes ab$ ] .
in particular , if @xmath60 is a pointed @xmath174-connected topological space of finite rational type , its sullivan model @xmath1066 is an augmented commutative dg algebra , and we can consider the dg lie algebras @xmath1067 and @xmath1068 both of which are cohomologically graded .
we write @xmath1069 and @xmath1070 for the classical ( absolute and relative ) chevalley - eilenberg cohomologies of the lie algebras @xmath1068 and @xmath1067 equipped with _
homological _ grading .
the following is a consequence of theorem [ conj1 ] and one of the main results of @xcite ( see _ loc .
_ , theorem 6.5 ) . for any affine algebraic group @xmath3 with lie algebra @xmath175
, there is an isomorphism of graded commutative algebras , @xmath1071 moreover , if @xmath3 is reductive , there is an isomorphism of graded commutative algebras @xmath1072 since the sullivan model of @xmath60 is uniquely determined up - to homotopy , it suffices to prove the desired theorem for a particular choice of sullivan model of @xmath60 .
let @xmath1073 be the minimal quillen model of @xmath60 .
then , @xmath1074 is connected , i.e , concentrated in positive homological degree and finite dimensional in each homological degree .
hence , @xmath1075 is @xmath1076-connected ( i.e , its coaugmentation coideal is concentrated in degrees @xmath1077 ) and finite dimensional in each homological degree .
the graded @xmath173-linear dual of @xmath1078 is @xmath1079 , which is a sullivan model of @xmath60 .
moreover , @xmath1078 is koszul dual to @xmath1074 , there is a quillen equivalence @xmath1080 between the category of @xmath1081 of ( coaugmented , conilpotent ) dg lie coalgebras and the category @xmath1006 of dg lie algebras ( see ( * ? ? ? * theorem 3.1 and 3.2 ) ) .
thus , there is a quasi - isomorphism of dg lie algebras @xmath1082 , which shows that @xmath1078 is koszul dual to @xmath1074 . ] .
it follows from theorem [ conj1 ] and ( * ? ? ?
* theorem 6.5 ( b ) ) ( also see _ loc .
_ , theorem 6.3 and the subsequent remark ) that @xmath1083 if , moreover , @xmath3 is reductive , we have @xmath1084 the first isomorphism above follows from the fact that all ( quasi-)isomorphisms in the proof of theorem [ conj1 ] are @xmath3-equivariant .
indeed , every @xmath3-action involved is induced by the @xmath3-action on the left @xmath1085-module @xmath1086 coming from the conjugation action of @xmath3 on itself .
this finishes the proof of the theorem . before proving theorem [ conj1 ] , we record one useful consequence that gives an explicit dg algebra model for the representation homology of @xmath60 in terms of the minimal quillen model of @xmath60 .
let @xmath1087 be the minimal quillen model of @xmath60 .
then , @xmath1088 is a canonical dg @xmath173-algebra whose homology is isomorphic to @xmath1089 .
thus , as graded algebras , @xmath1090\,,\ ] ] where the differential @xmath1091 is given on generators by @xmath1092 where @xmath1093 is the universal representation .
since @xmath1094 is a semi - free ( hence , cofibrant ) dg lie algebra , @xmath1095 $ ] .
the first assertion is then immediate from theorem [ conj1 ] .
the algebra isomorphism @xmath1096 follows easily from formula .
the formula for the differential @xmath1097 can follows easily from the fact that the universal representation @xmath1098 is a _ differential graded _ lie algebra homomorphism .
* example 1 .
* recall ( see example 5 , ( * ? ? ?
24 ) ) that the minimal lie model for the complex projective space @xmath1099 is given by the free lie algebra @xmath1100 generated by @xmath1101 , where the degree of @xmath1102 is @xmath1103 , and the differential is defined on generators by @xmath1104 , @xmath1105 $ ] for all @xmath1106 . by corollary [ cancomp ] , we have @xmath1107 \,\cong\ , \h_\ast \big [ { { \rm{sym}}}\big ( \bigoplus_{i=1}^r \g^{\ast } \cdot v_i \big ) , \partial \big]\,,\ ] ] where @xmath1108 denotes a copy of @xmath1109 in degree @xmath1103 indexed by @xmath1102 and where the differential @xmath1033 is given on generators by @xmath1110 here , the cobracket on @xmath1111 is given by @xmath1112 in sweedler notation .
* example 2 .
* as another application of corollary [ cancomp ] , we can easily recover the result of proposition [ nconnect ] .
indeed , let @xmath272 be an @xmath90-connected space for some @xmath1113 .
consider the minimal quillen model @xmath1114 of @xmath60 .
then @xmath1115 for all @xmath1116 .
by the rational hurewicz theorem , @xmath1117 for all @xmath1118 .
hence , @xmath1119 for @xmath1120 .
then the ( nonzero ) elements of @xmath1121 $ ] must have homological degree @xmath1122 , and therefore , by minimality of @xmath1123 , @xmath1124 for @xmath1125 .
the differential @xmath1097 on @xmath1126 then vanishes on chains of degree @xmath1127 , and the isomorphisms are immediate from corollary [ cancomp ] .
the proof of theorem [ conj1 ] is based on several technical results . as observed in section [ proof3 ] ,
@xmath1128 $ ] for any semi - free simplicial lie model @xmath1129 of @xmath60 .
now , @xmath1130 , where @xmath1131 is the simplicial right @xmath1085-module associated with the simplicial cocommutative hopf algebra @xmath1132 .
our first step is to prove theorem [ dtpliemodel ] , which states that @xmath1089 is isomorphic to the homology of the derived tensor product @xmath1133 .
recall that quillen s rational homotopy theory gives a zig - zag of maps @xmath436 \q\lgr{x } & \rto & \widehat{\q}\lgr{x } & \rto^g & \widehat{r } & \lto & r \end{diagram}\,,\ ] ] where @xmath1134 stands for completion with respect to the canonical augmentation . here
, the map @xmath1135 ( which is by no means unique ) is a weak equivalence in the model category of simplicial complete cocommutative hopf algebras ( @xmath1136 s ) .
the first and third arrows in the above zig - zag induce isomorphisms on all homotopy groups ( see ( * ? ? ?
* sec.3 , part i ) ) . in section [ proof1 ] ,
we use a relatively straightforward extension of the arguments of _ loc .
_ to verify ( in a series of propositions ) that the above zig - zag gives a zig - zag of the associated simplicial right @xmath1085-modules where every arrow induces isomorphisms on all homotopy groups .
the only subtlety here is that the notion of weak - equivalence in @xmath1136 is _ a priori _ different from that of a map inducing an isomorphism on all homotopy groups ( see ( * ? ? ?
* sec.4 , part ii ) ) .
this makes it necessary to argue that the map on simplicial right @xmath1085-modules induced by @xmath1135 indeed induces isomorphisms on all homotopy groups .
the proof of theorem [ dtpliemodel ] follows easily from the verifications in section [ proof1 ] as well as theorem [ rephomdtp ] .
starting from theorem [ dtpliemodel ] , we proceed to argue in section [ proof3 ] that @xmath1137 as graded vector spaces .
the crucial ingredient is proposition [ tors ] ( stated and proven in section [ proof2 ] ) , which may well be of independent interest .
this proposition states that if @xmath643 is a vector space , then @xmath1138 for @xmath1139 , where @xmath1140 is the right @xmath1085-module corresponding to the cocommutative hopf algebra @xmath1141 , the universal enveloping algebra of the free lie algebra generated by @xmath643 .
surprisingly , the proof of proposition [ tors ] relies on topological arguments : it uses theorem [ dtpliemodel ] as well as earlier computations of the representation homology of wedges of spheres ( proposition [ rephomsusp ] ) . we do not currently know a completely algebraic proof of this result .
proposition [ tors ] also implies ( see proposition [ liereps ] ) another result that could be of independent interest : for any lie algebra @xmath1011 , there is a natural isomorphism @xmath1142 finally , in section [ proof4 ] , we show that the isomorphism @xmath1137 of graded vector spaces constructed in section [ proof3 ] is indeed an isomorphism of graded commutative algebras .
we do this by exhibiting for any @xmath1143 a morphism of simplicial commutative algebras inducing the isomorphism @xmath1144 $ ] for @xmath1145 .
to show this , we first note that the canonical filtration ( by powers of the augmentation ideal ) on @xmath459 induces a filtration on the right @xmath1085-module @xmath1131
. then we use a generic connectivity argument due to curtis ( * ? ? ?
4 ) to show that @xmath1146 for @xmath1147 ( proposition [ quotientr ] ) . this allows us to replace @xmath1131 with @xmath1148 when computing homologies in degree @xmath1149 of @xmath1133 ( i.e , @xmath1150 for @xmath1145 ) .
again as a consequence of proposition [ tors ] , the @xmath90-simplices of @xmath1151 are right @xmath1085-modules whose higher tor s with @xmath1086 vanish .
it follows from these facts that the composite map @xmath436 \underline{\q\lgr{x } } \otimes_{\mathfrak{g } } \co(g ) & \rto & \underline{\widehat{\q}\lgr{x } } \otimes_{\mathfrak{g } } \co(g ) & \rto & \underline{\widehat{r } } \otimes_{\mathfrak{g } } \co(g ) & \rto & \underline{r}/f^r\underline{r } \otimes_{\mathfrak{g } } \co(g ) \end{diagram}\ ] ] induces the isomorphism @xmath1152 $ ] for @xmath1145 ( on @xmath1153 s ) .
it is not difficult to check that the maps above are morphisms of simplicial commutative algebras .
this concludes our argument .
[ [ section-1 ] ] by theorem [ nnstar_equiv ] there are quillen equivalences refining the dold - kan correspondence @xmath1154 where @xmath1155 denotes the category of simplicial objects in a category @xmath1156 . by proposition [ nstar_semifree_prop ] ,
applying the functor @xmath1157 to a semi - free quillen model of @xmath60 gives a reduced semi - free simplicial lie model of @xmath60 .
let @xmath1158 be a reduced semi - free simplicial lie model of @xmath60 .
consider the simplicial cocommutative hopf algebra @xmath1159 as well as the simplicial complete cocommutative hopf algebra @xmath1160 ( where the completion is with respect to the canonical augmentation ) .
these correspond to the right @xmath1085-modules @xmath1131 and @xmath1161 , which assign to @xmath1162 the the tensor product @xmath1163 and the completed tensor product @xmath1164 respectively .
@xmath1165 $ ] .
the proof of theorem [ dtpliemodel ] relies on several propositions from rational homotopy theory .
these propositions , in turn , are based on the following lemma .
let @xmath643 be a filtered reduced simplicial vector space .
let @xmath1166 denote the completion of @xmath643 with respect to the given filtration .
more generally , for any @xmath1167 , one has the simplicial vector spaces @xmath1168 and @xmath1169 , where @xmath1170 denotes the completed tensor product @xmath1171 .
let @xmath1172 denote the image in @xmath1170 of the symmetrization idempotent @xmath1173 .
let @xmath1174 . recall that a @xmath1175-equivalence ( see @xcite ) is a morphism inducing isomorphisms on all homotopy groups .
suppose that for each @xmath1176 , @xmath1177 for @xmath1178 sufficiently large .
then , @xmath1179 for each @xmath1176 , @xmath1180 for @xmath1178 sufficiently large .
+ @xmath1181 the map @xmath1182 is a @xmath1175-equivalences for all @xmath203 .
+ @xmath1183 the map @xmath1184 is a @xmath1175-equivalence . by a long exact sequence of homotopy groups ( lesh )
argument , the natural map @xmath1185 is an isomorphism for @xmath1178 sufficiently large .
thus , the inverse system @xmath1186 is eventually constant .
thus , @xmath1187 .
it follows from ( * ? ? ?
* part i , prop .
3.8 ) that @xmath1188 for @xmath1178 sufficiently large .
since @xmath1189 , we see that @xmath1190 for @xmath1178 sufficiently large . again by a lesh argument , @xmath1180 for @xmath1178 sufficiently large .
this proves @xmath1179 .
moreover , by the eilenberg - zilber and knneth theorems , @xmath1191 $ ] for @xmath1178 sufficiently large ( since the same is true for @xmath1192 ) .
it follows that the inverse system @xmath1193\}$ ] is eventually constant . arguing as for the case when @xmath1192 , we see that @xmath1194 $ ] for @xmath1178 sufficiently large .
this proves that the natural map @xmath1182 induces an isomorphism on @xmath1195 for any fixed @xmath138 .
this proves @xmath1181 .
since the map @xmath1196 is @xmath1197-equivariant and since @xmath1198 and @xmath1172 are the images of the symmetrization idempotent @xmath1199 acting on @xmath1200 and @xmath1170 respectively , the natural map @xmath1201 is a @xmath1175-equivalence .
thus , the map @xmath1202 is a @xmath1175-equivalence . since @xmath643 is reduced and by @xmath1181 , @xmath1203 for @xmath1204 ( by the eilenberg - zilber and knneth theorems ) .
it follows that @xmath1205 for @xmath1147 . applying @xmath1181 to @xmath1206 with filtration given by @xmath1207
, we see that the map @xmath1208 is a @xmath1175-equivalence .
this proves @xmath1183 .
the canonical map of @xmath1085-modules @xmath1209 is a @xmath1175-equivalence .
it needs to be shown that the map @xmath1210 is a @xmath1175-equivalence . by
* part i , thm .
3.7 ) , for any fixed @xmath138 , @xmath1211 vanishes for @xmath1178 sufficiently large .
lemma [ completions ] @xmath1181 then implies that the map @xmath1210 is a @xmath1175-equivalence , as desired .
recall that @xmath527 denotes the kan loop group functor applied to a reduced simplicial / cellular model of @xmath60 .
then , @xmath1212 is a simplicial cocommutative hopf algebra equipped with a canonical augmentation .
the completion @xmath1213 of @xmath1212 with respect to its canonical augmentation is a simplicial complete cocommutative hopf algebra ( @xmath1136 ) .
@xmath1212 as well as @xmath1213 correspond to simplicial right @xmath1085-modules , which we denote by @xmath1214 and @xmath1215 respectively .
the map @xmath1216 is a @xmath1175-equivalence .
we need to show that for each @xmath203 , the map @xmath1217 is a @xmath1175-equivalence . by lemma [ completions ]
@xmath1181 , this follows one we verify that for any fixed @xmath138 , @xmath1218 for @xmath1178 sufficiently large .
this is immediate from ( * ? ? ?
* thm . 4.72 ) .
we recall that the category @xmath1219 of reduced @xmath1136 s is a model category , whose cofibrant objects are retracts of semi - free @xmath1136 s .
the definition of semi - free @xmath1136 is the obvious extension to the simplicial setting of the definition of a free complete cocommutative hopf algebra : the free comple cocommutative hopf algebra generated by a vector space @xmath643 is @xmath1220 , where @xmath643 is primitive .
we now apply quillen s rational homotopy theory : in @xcite , quillen proves several equivalences of homotopy categories ( see _ loc .
211 , fig .
2 ) from which it follows that there is an isomorphism in @xmath1221 @xmath1222 . by theorem 4.7 of _ loc .
, there is a morphism @xmath1223 that is a simplicial homotopy equivalence .
denote the corresponding map of right @xmath1085-modules by @xmath1224 .
@xmath1225 is a @xmath1175-equivalence . by ( * ?
* part i , theorem 3.7 ) and lemma [ completions ] , the completion map @xmath1226 is a @xmath1175-equivalence .
similarly , it can be shown that the map @xmath1227 is a @xmath1175 equivalence . to prove the desired lemma
, we need to show that @xmath1228 is a @xmath1175 equivalence for each @xmath203 .
since the diagram @xmath436 \widehat{\q}\lgr{x}^{\otimes m } & \rto & \widehat{\q}\lgr{x}^{\widehat{\otimes } m}\\ \dto_{g^{\otimes m } } & & \dto_{g^{\widehat{\otimes } m}}\\ \widehat{r}^{\otimes m } & \rto & \widehat{r}^{\widehat{\otimes } m}\\ \end{diagram}\ ] ] commutes , it suffices ( by the eilenberg - zilber and knneth theorems ) to show that @xmath1135 is a @xmath1175-equivalence .
let @xmath1229 denote the functor of primitive elements . by ( * ?
* appendix a , cor .
2.16 ) , there is an isomorphism of simplicial vector spaces @xmath1230 .
for the same reason , @xmath1213 is isomorphic to @xmath1231 as simplicial vector spaces . since @xmath1232 is a canonical retract of @xmath1233 , @xmath1234 for @xmath1178 large enough ( since the same holds for @xmath1233 ) and for the same reason , @xmath1235 for @xmath1178 sufficiently large . by lemma [ completions ] @xmath1183 ,
the horizontal arrows in the commutative diagram below are @xmath1175-equivalences . @xmath436 { { \rm{sym}}}(\mathcal p \widehat{\q}\lgr{x } ) & \rto & \widehat{{{\rm{sym}}}}(\mathcal p \widehat{\q}\lgr{x } ) \ , \cong\ , \widehat{\q}\lgr{x}\\ \dto_{{{\rm{sym}}}(\mathcal p g ) } & & \dto_{g}\\ { { \rm{sym}}}(\mathcal p \widehat{r } ) & \rto & \widehat{{{\rm{sym}}}}(\mathcal p \widehat{r } ) \,\cong\ , \widehat{r}\\ \end{diagram}\ ] ] by ( * ? ? ?
* part ii , theorem 4.7 ) , @xmath1236 is a @xmath1175-equivalence .
thus , the left vertical arrow in the above diagram is a @xmath1175-equivalence .
it follows that @xmath1135 is a @xmath1175-equivalence , as desired .
( proof of theorem [ dtpliemodel ] ) by proposition [ comp1 ] , proposition [ comp2 ] and proposition [ comp3 ] , there is a diagram of complexes of right @xmath1085-modules in which each arrow is a quasi - isomorphism @xmath1237 hence , there is an isomorphism in the derived category @xmath1238 of complexes of @xmath173-vector spaces @xmath1239 the desired result now follows from theorem [ rephomdtp ] .
[ [ section-2 ] ] let @xmath643 be a @xmath173-vector space and let @xmath1140 denote the right @xmath1085-module corresponding to the cocommutative hopf algebra @xmath1240 .
the module @xmath1140 has a weight grading induced by the weight grading on @xmath1241 in which @xmath643 has weight @xmath174 .
let @xmath1242 denote the component of @xmath1140 of weight @xmath138 .
for example , @xmath1243 is the @xmath1085-module defined by @xmath1244 , see .
the following proposition may be of independent interest .
let @xmath3 be an affine algebraic group over @xmath173 , with lie algebra @xmath175 .
then @xmath1245 0 & \mbox{\rm if}\ i > 0 \end{array } \right.\ ] ] in particular , @xmath1246 = 0\,$ ] for all @xmath1247 and @xmath1248 .
the following lemma ( see ( * ? ? ?
4.3 ) ) is essential for the proof of proposition [ tors ] . for the benefit of the reader
, we outline a proof of this lemma that is different from that given in _
cit._. for any @xmath1249 , there is a natural isomorphism of commutative algebras @xmath1250 where @xmath1251 is the representation algebra defined in .
let @xmath1252 . from the left @xmath1085-module @xmath1086
, one can form the right @xmath1085-module @xmath1253 , which assigns @xmath1254 to @xmath1162 . since @xmath1255 is a commutative @xmath4-algebra and since @xmath1086 is a strictly monoidal left @xmath1085-module
, @xmath1256 acquires the structure of a lax monoidal right @xmath1085-module .
this structure is given by the maps @xmath436 \hom_k(\co(g)^{\otimes m } , b ) \otimes \hom_k(\co(g)^{\otimes n } , b ) & \rto^{\mu_b \circ ( \mbox{--}\,\otimes\,\mbox{-- } ) } & \hom_k(\co(g)^{\otimes ( m+n ) } , b ) \end{diagram}\,,\ ] ] where @xmath1257 is the product on @xmath1255 . by the standard @xmath1258 adjunction
, there is a natural isomorphism of @xmath4-vector spaces @xmath1259 it is a routine verification to check that under this isomorphism , the commutative @xmath4-algebra homomorphisms from @xmath1260 to @xmath1255 correspond to the right @xmath1085-module homomorphisms from @xmath1261 to @xmath1253 that respect the ( lax ) monoidal structure . since @xmath1086 is a coalgebra and @xmath1255 is an algebra , @xmath1262 has an algebra structure ( with product given by convolution ) .
another routine verification shows that the set of right @xmath1085-module homomorphisms from @xmath1261 to @xmath1253 that respect the ( lax ) monoidal structure is in ( natural ) bijection with the set of @xmath4-algebra homomorphisms @xmath1263 from @xmath1261 to @xmath1262 that satisfy the following additional conditions : @xmath1264 for all @xmath1265 and @xmath1266 . here
, @xmath1267 and @xmath117 stand for the counit and antipode of @xmath1261 respectively the coproduct in @xmath1261 is given by @xmath1268 in sweedler notation .
it is not difficult to verify that the third condition above follows from the first two . as shown in (
* example 3.4 ) , the algebra homomorphisms from @xmath1261 to @xmath1262 satisfying the above conditions are in natural bijection with lie algebra homomorphisms from @xmath1011 to @xmath1269 .
indeed , @xmath1263 satisfies these conditions for all @xmath1270 in @xmath1261 iff it satisfies these conditions for @xmath1271 . for @xmath1272 ,
these conditions are equivalent to the assertion that @xmath1273 is a @xmath4-linear derivation on @xmath1086 with respect to the homomorphism @xmath1274 , where @xmath1275 denotes the canonical augmentation on @xmath1086 .
such derivations are indeed in bijection with elements of @xmath1276 .
we thus , have a natural bijection @xmath1277 the desired lemma now follows from the yoneda lemma . since @xmath1278 , the required isomorphism for @xmath1279 follows from lemma [ ualphatensorog ] .
to prove the vanishing of @xmath1280 for @xmath1281 , we assign @xmath643 ( homological ) degree @xmath1076 .
then @xmath1140 is a graded right @xmath1085-module , whose component in degree @xmath1282 is @xmath1242 .
thus , @xmath1283 \,\cong\ , \bigoplus_{2q+i = n } { { \rm{tor}}}^i_{\mathfrak{g}}[(\underline{tv})_q , \co(g)]\,\text{.}\ ] ] the desired proposition will follow once we show that @xmath1284 \,\cong\ , { { \rm{sym}}}(\g^\ast \otimes v ) \,\text{.}\ ] ] equip @xmath1285 ( see formula [ nstar_semifree_formula ] in appendix [ doldkan ] ) with the simplicial cocommutative hopf algebra structure given by its identification with @xmath1286 .
this gives @xmath1287 the structure of a simplicial right @xmath1085-module ( which we denote by @xmath1288 ) .
this module assigns to the free group @xmath1162 the simplicial vector space @xmath1289 .
note that since @xmath643 has degree @xmath1076 , @xmath1290 is a semi - free simplicial lie model for the space @xmath60 given by a wedge of @xmath1291 @xmath21-spheres . by theorem [ dtpliemodel ] and proposition [ rephomsusp ] ,
@xmath1292 \,\cong\ , \hr_\ast(x , g ) \,\cong\ , { { \rm{sym}}}(\g^{\ast } \otimes v ) \,\text{.}\ ] ] the desired proposition is therefore , a consequence of lemma [ dgtosimp ] below .
the unit of the adjunction @xmath1293 induces a quasi - isomorphism of right @xmath1085-modules @xmath1294 . the morphism @xmath1294 is defined by the family of maps @xmath1295 where the last arrow is the eilenberg - zilber map ( which is well - defined for @xmath1296 because of the associativity of the eilenberg - zilber map for @xmath1297 ) . that this is a quasi - isomorphism follows from the knneth theorem and the fact that @xmath1293 is a quasi - isomorphism .
it follows from the associativity of the eilenberg - zilber map that the maps @xmath1298 indeed assemble into a morphism of right @xmath1085-modules .
[ [ section-3 ] ] we now show that @xmath1299 as graded vector spaces . without loss of generality , we may assume that @xmath967 is a semi - free dg lie model of @xmath60 .
since the representation functor @xmath1300 is left adjoint , it commutes with @xmath1157 , i.e , there is a commutative diagram of functors @xmath436 \dgl_{\q } & \rto^{n^{\ast } } & \mathtt{slie}_\q\\ \dto^{(\,\mbox{--}\,)_\g } & & \dto_{(\,\mbox{--}\,)_\g}\\ \cdga_\q & \rto^{n^{\ast } } & \mathtt{scalg}_\q\\ \end{diagram}\ ] ] thus , @xmath1128 $ ] , where @xmath1301 . by lemma [ ualphatensorog ] , @xmath1302 , where @xmath1303 .
thus , @xmath1304\,\cong\,\h_\ast[\mathrm{c}(\underline{r } ) \otimes_{\mathfrak{g } } \co(g ) ] $ ] , where @xmath1305 stands for associated chain complex .
since , @xmath1129 is a semi - free simplicial lie model of @xmath60 , the right @xmath1085-module of @xmath90-simplices in the simplicial right @xmath1085-module @xmath1131 is of the form @xmath1140 for some vector space @xmath643 .
it follows from proposition [ tors ] that @xmath1306 is a complex of right @xmath1085-modules whose higher tor s with @xmath1086 vanish .
thus , the map @xmath1307 is a quasi - isomorphism .
note that there is a quasi - isomorphism of complexes of right @xmath1085-modules @xmath1308 .
that @xmath1299 as graded vector spaces follows from theorem [ dtpliemodel ] , by which @xmath1309 $ ] .
as yet another consequence of proposition [ tors ] , we have the following result which may be of independent interest . for any lie algebra @xmath1310
, there is a natural isomorphism @xmath1311 as argued above , it suffices to show that @xmath1312 $ ] for any semi - free simplicial resolution @xmath1313 of @xmath1011 . by lemma [ ualphatensorog ] , @xmath1314 . since @xmath1313 is semi - free , the right @xmath1085-module of @xmath90-simplices in the simplicial right @xmath1085-module @xmath1315 is of the form @xmath1140 for some vector space @xmath643 .
it follows from proposition [ tors ] that the map @xmath1316 is a quasi - isomorphism .
the desired proposition then follows once we establish that @xmath1315 is a simplicial resolution of @xmath1317 . for this
, we need to check that for any @xmath203 , @xmath1318 resolves @xmath1319 .
this follows from the eilenberg - zilber and knneth theorems .
[ [ section-4 ] ] to complete the proof of theorem [ conj1 ] , it remains to show that @xmath1299 as graded @xmath173-algebras .
for this , given any @xmath1320 , we shall produce a morphism of simplicial commutative algebras that induces the isomorphism @xmath1321 $ ] for @xmath1322 .
recall that @xmath1303 is a semi - free simplicial associative algebra filtered by powers of its augmentation ideal .
this filtration induces a filtration on the simplicial right @xmath1085-module @xmath1131 : if the algebra of @xmath90-simplices of @xmath459 is @xmath1241 for some vector space @xmath643 , then the right @xmath1085-module of @xmath90-simplices of @xmath1323 is @xmath1324 .
the following connectivity result holds for the filtered right @xmath1085-module @xmath1131 . for @xmath1147 , we have @xmath1146 .
it needs to be shown that for all @xmath1162 , @xmath1325 for @xmath1147 . for @xmath1326 ,
this is obvious . for @xmath1192 ,
this is ( * ? ? ?
* part i , thm .
3.7 ) . for arbitrary @xmath203
, we generalize the argument in _ loc . cit._.
the functor @xmath1327 takes @xmath245 to @xmath245 and commutes with direct limits . by (
* remark 4.10 ) , the arguments in section 4 of _ loc .
_ proving lemma ( 2.5 ) therein apply to this functor as well .
it therefore , suffices to verify the desired proposition for @xmath1328 , where @xmath1329 is the free simplicial lie algebra generated by @xmath1330 , where @xmath11 is a finite wedge sum of simplicial circles .
note that in this case , @xmath1331 , and @xmath643 is a _ connected _ simplicial vector space . in this case ,
@xmath1332 . that @xmath1195 of each summand vanishes for @xmath1322 follows from the eilenberg - zilber and knneth theorems .
this proves the desired proposition . for @xmath1178 sufficiently large , all arrows in the following commutative diagram
induce isomorphisms on the homology groups
@xmath1333\,,\,i \leq q$ ] .
@xmath436 \mathrm{c}(\underline{r } ) \otimes^{\l}_{\mathfrak{g } } \co(g ) & \rto & \mathrm{c}(\underline{r}/f^r\underline{r } ) \otimes^{\l}_{\mathfrak{g } } \co(g)\\ \dto & & \dto \\
\mathrm{c}(\underline{r } ) \otimes_{\mathfrak{g } } \co(g ) & \rto & \mathrm{c}(\underline{r}/f^r\underline{r } ) \otimes_{\mathfrak{g } } \co(g)\\ \end{diagram}\ ] ] both @xmath1306 and @xmath1334 are complexes of right @xmath1085-modules whose higher tors with @xmath1086 vanish by proposition [ tors ] .
it follows that the vertical arrows induce isomorphisms on all homology groups .
it therefore , suffices to show that the horizontal arrow on top of the above diagram induces isomorphisms on @xmath1333\,,\,i \leq q$ ] for @xmath1178 sufficiently large .
consider the good truncation @xmath1335 ( see ( * ? ? ?
1.2.7 ) ) of a chain complex @xmath1305 of right @xmath1085-modules .
the exact sequence @xmath1336 of complexes of right @xmath1085-modules gives a distinguished triangle in @xmath1238 for any right @xmath1085-module @xmath380 .
@xmath1337\ ] ] it is easy to see that @xmath1338 for @xmath1339 .
the long exact sequence of homologies associated with the above distinguished triangle then implies that @xmath1340 by proposition [ quotientr ] , the map @xmath1341 is a quasi - isomorphism for @xmath1147 .
thus , the map @xmath1342 { \xrightarrow{}}\h_\ast[\tau_{<q+1 } \mathrm{c}(\underline{r}/f^r\underline{r } ) \otimes^l_\mathfrak{g } \co(g)]$ ] is an isomorphism of graded @xmath173-vector spaces .
the desired proposition now follows from .
note that the filtration on @xmath1233 induces a filtration on the right @xmath1085-module @xmath1161 .
clearly , @xmath1343 .
the following corollary is immediate from proposition [ comp1 ] and proposition [ goingdown ] . for @xmath1178
sufficiently large , all arrows in the following commutative diagram induce isomorphisms on the homology groups @xmath1333\,,\,i \leq q$ ] .
@xmath436 \mathrm{c}(\underline{r } ) \otimes^{\l}_{\mathfrak{g } } \co(g ) & \rto & \mathrm{c}(\underline{\widehat{r}})\otimes^{\l}_{\mathfrak{g } } \co(g)\\ \dto & \ldto & \\ \mathrm{c}(\underline{r}/f^r\underline{r } ) \otimes^{\l}_{\mathfrak{g } } \co(g ) & & \\ \end{diagram}\ ] ] recall that there is a weak equivalence between cofibrant objects in @xmath1219 @xmath1223 inducing a map of simplicial right @xmath1085-modules @xmath1225 ( see proposition [ comp3 ] ) .
consider the following commutative diagram , where the second arrow on the top and bottom rows is induced by @xmath1225 .
@xmath1344 \mathrm{c}(\underline{\q\lgr{x } } ) \otimes^{\l}_{\mathfrak{g } } \co(g ) & \rto & \mathrm{c}(\underline{\widehat{\q}\lgr{x } } ) \otimes^{\l}_{\mathfrak{g } } \co(g ) & \rto & \mathrm{c}(\underline{\widehat{r } } ) \otimes^{\l}_{\mathfrak{g } } \co(g ) & & \\
\dto & & \dto & & \dto & \rdto & \\ \mathrm{c}[\underline{\q\lgr{x } } \otimes_{\mathfrak{g } } \co(g ) ] & \rto & \mathrm{c}[\underline{\widehat{\q}\lgr{x } } \otimes_{\mathfrak{g } } \co(g ) ] & \rto & \mathrm{c } [ \underline{\widehat{r } } \otimes_{\mathfrak{g } } \co(g ) ] & \rto & \mathrm{c}[\underline{r}/f^r\underline{r } \otimes_{\mathfrak{g } } \co(g)]\\ \end{diagram}\ ] ] by proposition [ rephomdtp ] the left vertical arrow in induces isomorphisms on all homologies .
the two arrows on the top row of induce isomorphisms on all homologies by propositions [ comp2 ] and [ comp3 ] respectively .
the diagonal arrow induces isomorphisms on @xmath1333\,,\,i \leq q$ ] for @xmath1178 sufficiently large by proposition [ goingdown ] and corollary [ gdcor ] .
an isomorphism @xmath1345\,,\,i\leq q$ ] is thus induced on homologies ( for sufficiently large @xmath1178 ) by the composition of the maps on the bottom row of . that the composition of maps in the bottom row is a map of dg commutative algebras follows from the fact that each of the maps @xmath1346 is a morphism of simplicial commutative algebras .
indeed , this last fact follows from ( * ? ? ?
3.4 ) and the facts that @xmath1086 is a lax - monoidal left @xmath1085-module , the @xmath90-simplices of the right @xmath1085-modules @xmath1347 and @xmath1151 are lax - monoidal for each @xmath90 , and the morphisms @xmath1348 are natural transformations of lax - monoidal functors on @xmath90-simplices for each @xmath90 .
this completes the proof of theorem [ conj1 ] .
the results of this section go through with @xmath173 replaced by any field @xmath4 of characteristic @xmath245 .
indeed , the proofs of propositions [ comp1 ] and [ comp2 ] work for any such field @xmath4 . for proposition [ comp3 ] ,
we work with a semi - free simplicial lie model @xmath1129 of @xmath60 over @xmath173 . the corresponding lie model over @xmath4 is @xmath1349 .
the corresponding @xmath1136 over @xmath4 is @xmath1350 .
the @xmath1175-equivalence of @xmath1136 s ( over @xmath173 ) @xmath1351 extends to a @xmath1175-equivalence of @xmath1136 s over @xmath4 @xmath1352 .
this proves proposition [ comp3 ] over @xmath4 .
theorem [ dtpliemodel ] , proposition [ tors ] , theorem [ conj1 ] and proposition [ liereps ] can then be proven over @xmath4 as done above ( over @xmath173 ) . in this section
, we assume that @xmath1353 , and @xmath3 is a complex _ reductive _ group of finite rank @xmath188 .
we denote by @xmath175 the lie algebra of @xmath3 .
recall that for any homogeneous invariant polynomial @xmath1354 , one has the associated drinfeld trace map ( see formula ) @xmath1355 the drinfeld trace maps associated with a set @xmath1356 of homogeneous generators of @xmath1357 assemble to a homomorphism of graded commutative algebras @xmath1358 where @xmath1359 are the fundamental degrees of the lie algebra @xmath178 .
if @xmath1360 is a quillen lie model of a simply connected space @xmath60 , we refer to the map as a _
drinfeld homomorphism _ for @xmath60 . with this terminology , we can now state the main result of this section . for @xmath1361 @xmath1362 ,
the drinfeld homomorphism is an isomorphism .
as shown in example 1 , there is an isomorphism of graded commutative algebras + @xmath1363 $ ] , where @xmath1364 is the minimal lie model of @xmath1365 .
further , all ( quasi)-isomorphisms in the proof of theorem [ conj1 ] are @xmath3-equivariant .
indeed , the @xmath3-action on all complexes involved in the proof of theorem [ conj1 ] is induced by the @xmath3-action on the left @xmath1085-module @xmath1086 coming from the conjugation action of @xmath3 on itself . since @xmath3 is reductive , it follows that @xmath1366 $ ] ( as graded vector spaces ) . on the other hand , @xmath1364 is koszul dual to the graded linear dual of the _ cohomologically graded _
commutative algebra @xmath1367/(u^{r+1})$ ] , where @xmath1368 has cohomological degree @xmath1076 . by (
* thm . 6.5
( b ) ) , @xmath1369\,\cong\ , \h^{-\ast}[\g[u]/(u^{r+1}),\g;\c]$ ] .
a related cohomology is computed in ( * ? ? ?
a. ) , which states that if @xmath1370 has cohomological degree @xmath245 , then @xmath1371/(z^{r+1}),\g;\c]$ ] is the free exterior algebra on @xmath1372 generators of cohomological degree @xmath1373 and @xmath1370-weights @xmath1374 , where @xmath203 runs over the exponents @xmath1375 of @xmath175 . here .
the ( relative ) lie algebra cohomology being considered in ` continuous ' lie algebra cohomology .
in fact , setting @xmath1011 to be the semi - free dg lie algebra koszul dual to @xmath1376/(z^{r+1})$ ] , the results of _ loc .
_ imply that the drinfeld trace @xmath1377 is an isomorphism for @xmath1011 .
in fact , for @xmath1378 , @xmath1379 is a vector space in homological degree @xmath1380 ( where @xmath1381 ) having a basis of @xmath1178 elements with @xmath1370-weights @xmath1374 .
the claim then follows from the observation that the drinfeld trace for @xmath1364 and the drinfeld trace for @xmath1011 are identical as morphisms of @xmath1382-graded commutative algebras , though they differ as morphisms of @xmath1383-graded commutative algebras .
this argument is identical to that deducing ( * ? ? ?
* thm 9.1 ) from the corresponding result in @xcite .
it follows from ( * ? ? ?
3.2 and prop .
a.3 ) that under the identification @xmath1369\,\cong\ , \h^{-\ast}[\g[u]/(u^{r+1}),\g;\c]$ ] mentioned in the above proof , the drinfeld trace map is identified with the map ( * ? ? ?
* equation ( 3.1 ) ) given by teleman with an explicit formula ( see _ loc .
_ , equation ( 2.2 ) ) .
the latter map is implicitly described in the work of feigin @xcite . as a consequence of theorem [ rephomcpr ] , we get the following explicit description of ( the invariant part of ) representation homology of complex projective spaces .
let @xmath3 be a complex reductive lie group of rank @xmath188 .
let @xmath189 denote the exponents of the lie algebra of @xmath34 .
then , for @xmath190 , there is an isomorphism of graded commutative algebras @xmath1384 where the generator @xmath192 has homological degree @xmath193 . in the limit
, we have @xmath1385 note that , under the isomorphism of @xmath1382-graded vector spaces @xmath1386 , a basis element of homological degree @xmath1380 and @xmath1370-weight @xmath1387 corresponds to a basis element of homological degree @xmath1388 - 2m-1 $ ] and @xmath1368-weight @xmath1387 ( since @xmath1368 has homological degree @xmath1389 .
this shows that @xmath1390 has @xmath1178 basis elements of homological degree @xmath1391 for each exponent @xmath203 of @xmath175 .
hence , the first statement follows from theorem [ rephomcpr ] .
the second statement follows from the first , if we notice that the map @xmath1392 induced by any map of spaces @xmath1393 vanishes for @xmath1394 for degree reasons .
if @xmath1360 is a lie model of a simply - connected space @xmath60 of finite rational type , the lie - hodge decomposition has a natural topological interpretation described in @xcite .
this interpretation is based on a well - known theorem of jones @xcite that identifies the ( reduced ) cyclic homology @xmath1395 of the universal enveloping algebra of @xmath1396 with the ( reduced ) @xmath196-equivariant homology of the free loop space @xmath197 of @xmath60 @xmath1397 recall that @xmath1398 is the space of all continuous maps @xmath1399 equipped with compact open topology , and as such , it carries a natural circle action ( induced by the action of @xmath196 on itself ) . using the finite self - covering maps @xmath1400 ,
we may define , for all @xmath249 , the graded linear endomorphisms on the @xmath196-equivariant homology of @xmath1401 : @xmath1402 which are called the _
frobenius operations_. under the jones isomorphism , the @xmath1033-th hodge summand @xmath1403 of @xmath1404 corresponds to the common eigenspace @xmath1405 of frobenius operations @xmath1406 with the eigenvalues @xmath1407 for all @xmath1408 ( see ( * ? ? ?
* theorem 1.2 ) ) . on the other hand , by our theorem [ conj1 ] ,
the representation homology of the lie model @xmath1050 can be identified with the representation homology of @xmath60 .
thus , for any simply - connected space , the drinfeld homomorphism can be rewritten in purely topological terms @xmath1409 \,\to \ , \hr_\ast(x , g)^g\ , \ ] ] which exhibits an interesting relation between representation homology and @xmath1410-equivariant homology of simply - connected spaces .
note that the domain and the target of the map make sense for an arbitrary connected space @xmath60 .
it would be interesting to see if there is a direct construction of that does not pass through the above identifications and works for all ( not necessarily simply - connected ) spaces .
now , for @xmath1411 , the result of theorem [ rephomcpr ] reads @xmath1412 \ .\ ] ] modulo theorem [ conj1 ] , the isomorphism is equivalent to the strong macdonald conjecture . as we mentioned in the introduction , this famous conjecture proposed by feigin and hanlon in the early 80s was recently proved in @xcite by an algebraic _ tour de force_. the above reformulation suggests that there might exist a more conceptual argument using topological means . in a series of recent papers @xcite , buijs , flix , murillo and tanr associated a free dg lie algebra model @xmath1413 to _ any _ finite simplicial complex @xmath60 .
unlike quillen models , the dg lie algebras @xmath1414 are assumed , in general , to be not connected but _
complete _ with respect to the canonical decreasing filtrarion @xmath1415 defined by @xmath1416 and @xmath1417 $ ] .
the @xmath245-simplices of @xmath60 correspond to the degree @xmath1418 generators of @xmath1419 that satisfy the maurer - cartan equation , while the @xmath90-simplices of @xmath60 correspond to generators in degree @xmath1420 . for any connected , finite simplicial complex @xmath60 ,
the dg lie algebra @xmath1414 itself is acyclic ( i.e. , @xmath1421 ) .
the topological information about @xmath60 is contained in a dg lie algebra @xmath1422 obtained from @xmath1414 by twisting its differential by maurer - cartan elements corresponding to the vertices of @xmath60 , i.e. @xmath1423\,$ ] , where @xmath1424 denotes ( the maurer - cartan element corresponding to ) a vertex of @xmath60 .
now , the main result of @xcite ( see _ loc .
_ , theorem a ) says that , if @xmath60 is simply - connected , then @xmath1425 is quasi - isomorphic to a quillen model of @xmath60 .
this motivates the following conjectural generalization of our theorem [ conj1 ] .
let @xmath1426 be a complete free dg lie algebra model associated to a reduced simplicial set @xmath60 .
let @xmath1427 $ ] be the twisted differential on @xmath1414 corresponding to the ( unique ) basepoint of @xmath60 .
note that @xmath1428 $ ] has a canonical augmentation @xmath1267 corresponding to the trivial ( zero ) representation .
let @xmath1429 $ ] denote the adic completion of @xmath1430 $ ] with respect to the augmentation ideal of @xmath1267 .
similarly , @xmath1431 has a canonical augmentation corresponding to the trivial ( identity ) representation of @xmath1432 .
let @xmath1433 denote the corresponding completion of @xmath1089 .
@xmath1434 \,\text{.}\ ] ] note that conjecture [ conj5 ] holds for @xmath60 simply - connected : indeed , in this case , @xmath1435 is quasi - isomorphic to a quillen model @xmath967 of @xmath60 and @xmath1428\,\cong\,\q$ ] .
thus , the right - hand side of the conjectured isomorphism is @xmath1436 .
similarly , @xmath1437 , which implies that @xmath1438 .
thus , conjecture [ conj5 ] is equivalent to theorem [ conj1 ] for simply - connected spaces .
in this section , using standard topological decompositions , we compute representation homology of some classical non - simply connected spaces .
our examples include closed surfaces ( both orientable and non - orientable ) as well as some three - dimensional spaces ( link complements in @xmath74 , lens spaces and general closed orientable @xmath21-manifolds ) .
the representation homology of surfaces and link complements is given in terms of classical hochschild homology of @xmath205 ( or @xmath1439 for some @xmath1440 ) with twisted coefficients .
the representation homology of a closed 3-manifold @xmath216 is expressed in terms of a differential ` tor ' , which gives rise to an ( eilenberg - moore ) spectral sequence converging to @xmath1441 .
as a cell complex , the @xmath1076-torus @xmath229 can be constructed as the homotopy cofibre ( the mapping cone ) of the map @xmath1442 , where the subscripts on the circles indicate the generators of the respective fundamental groups , and the map itself is specified , up to homotopy , by its effect on these generators : @xmath1443 : = a b a^{-1 } b^{-1 } \ .\ ] ] thus @xmath1444\,$ ] , where the homotopy colimit is taken in the category @xmath1445 of connected pointed spaces .
applying to this the kan loop group functor @xmath1446 ( more precisely , the composition and @xmath1447 preserve homotopy colimits , and hence so does their composition .
] of @xmath1446 with the eilenberg subcomplex functor , see section [ qeq ] ) , we get a simplicial group model for @xmath1448 : @xmath1449\ .\ ] ] here @xmath1450 and @xmath1451 are the free groups on the generators @xmath1452 and @xmath1453 respectively ; the map @xmath1454 is given by , and the homotopy colimit is taken in the category @xmath37 of of simplicial groups .
now , by lemma [ hopushout ] , the derived representation functor @xmath23 preserves homotopy pushouts for any algebraic group @xmath3 .
hence , it follows from that @xmath1455 \cong { \rm hocolim } [ \,k \leftarrow \o(g ) \xrightarrow{\alpha _ * } \o(g\times g ) \ , ] \ , \ ] ] where the homotopy colimit is taken in @xmath1456 , and the map @xmath1457 is induced by ( explicitly , @xmath1458 ) $ ] for @xmath1459 ) . since @xmath1460\ , \cong\ , \o(g\times g ) \otimes^{\l}_{\o(g ) } k\ , \ ] ] by proposition [ rephom0 ] , we conclude that @xmath1461 where @xmath1462 is viewed as a ( right ) @xmath1463-module via the algebra map @xmath1464 . by standard homological algebra ( see ( * ? ?
* theorem 2.1 , p. 185 ) ) , we can identify the tor - groups in as the classical hochschild homology of @xmath205 with coefficients in the bimodule @xmath1465 , where the right @xmath205-module structure is given via the map @xmath1464 and the left module structure via the augmentation map @xmath1466 : @xmath1467 alternatively , for classical ( matrix ) groups @xmath34 , we can give an explicit ` small ' dg algebra model for the representation homology @xmath1468 .
specifically , let @xmath1469 denote the maximal ( augmentation ) ideal of @xmath1470 corresponding to the identity element @xmath1471 .
assume that @xmath1472 is generated by a _
regular _ sequence of elements @xmath1473 in @xmath205 , so that @xmath1474 .
consider the free module @xmath1475 and define the @xmath1476-module map @xmath1477 by @xmath1478 .
then , associated to @xmath1479 is the ( global ) koszul complex @xmath1480 with differential @xmath1481 since @xmath1472 is generated by a regular sequence , the canonical projection @xmath1482 is a quasi - isomorphism of complexes , and therefore @xmath1483 is a free resolution of @xmath7 over @xmath205 .
it follows from that @xmath1484\ , \ ] ] where @xmath1485 is a commutative dg algebra with differential @xmath1486 .
in particular , @xmath1487 for all @xmath1488 .
we conclude this example with a conjectural description of the @xmath3-invariant part of representation homology @xmath1489
. our conjecture can be viewed as a multiplicative analogue of the derived harish - chandra conjecture proposed in @xcite .
assume that @xmath3 is a connected reductive algebraic group of rank @xmath1490 defined over an algebraically closed field @xmath1491 of characteristic zero .
let @xmath218 be a cartan subgroup ( i.e. , a maximal torus ) in @xmath1492 , and let @xmath1493 be the corresponding weyl group .
note that , since @xmath1494 is commutative , the map @xmath1495 associated to @xmath1494 factors through the augmentation @xmath1496 .
hence , by , we have canonical isomorphisms @xmath1497 where @xmath1498 is the augmentation ideal for @xmath1494 and @xmath1499 is the lie algebra of @xmath1494 ( i.e. , a cartan subalgebra of @xmath178 ) .
now , by functoriality , the natural inclusion @xmath1500 induces a map of simplicial commutative algebras @xmath1501^g \,\to \,\drep_{t}[{{\mathbb g}}({{\mathbb t}}^2)]^w\ , \ ] ] which is ( a multiplicative analogue of ) the derived harish - chandra homomorphism constructed in @xcite
. then , the multiplicative version of the derived harish - chandra conjecture states we illustrate conjecture [ conj2 ] for @xmath1502 . since @xmath1503_{1 \leq i , j \leq
n } \,$ ] , the elements @xmath1504 form a regular sequence in @xmath1505 generating the maximal ideal @xmath1472 , so we have a canonical commutative dg algebra representing @xmath1506 : @xmath1507_{1 \leq i , j \leq n}\ .\ ] ] here the variables @xmath1508 and @xmath1509 have homological degree @xmath245 , @xmath1510 have homological degree @xmath174 , and @xmath1511 and @xmath1512 denote the determinants of the generic matrices @xmath1513 and @xmath1514 . the differential on @xmath1515 can be written in matrix terms as @xmath1516 where @xmath1517 and @xmath1518 is the identity @xmath1519-matrix . the harish - chandra homomorphism @xmath1520^{s_n}\ ] ]
is given explicitly ( on generators ) by the following map @xmath1521 and the derived harish - chandra conjecture asserts that @xmath1522 induces an isomorphism ( _ cf .
_ ) @xmath1523^{s_n}\ , \ ] ] where @xmath1524 have homological degree @xmath1525 and the symmetric group @xmath1526 acts diagonally by permuting the variables .
note that , in the case of @xmath1527 , unlike for other algebraic groups , conjecture [ conj2 ] follows from the derived harish - chandra conjecture for the corresponding lie algebra @xmath1528 stated in @xcite .
this is because the harish - chandra map @xmath1522 can be obtained by formally localizing the derived harish - chandra map for the lie algebra @xmath1528 ( _ cf . _
4 ) ) . in particular
, the evidence collected in @xcite for @xmath1528 also supports conjecture [ conj2 ] for @xmath1529 .
we list some of this evidence here . 1 .
conjecture [ conj2 ] holds for @xmath1530 and @xmath1531 .
this follows from theorem 4.1 and theorem 4.2(ii ) of @xcite .
2 . for all @xmath1532 ,
the map is degreewise surjective .
this follows from ( * ? ? ?
* theorem 4.2(i ) ) .
3 . for all @xmath1532 , @xmath1533 for @xmath1534 . this follows from ( * ? ? ?
* theorem 27 ) .
4 . for any @xmath34 as in conjecture [ conj2 ] ,
the map is an isomorphism in homological degree _
i.e. @xmath1535 .
this follows from a theorem of thaddeus @xcite ( see also @xcite ) .
finally , we remark that , for @xmath1536 , @xmath1537 and @xmath1538 , the harish - chandra map is known to be an isomorphism in homological degree @xmath245 : @xmath1539 for all tori @xmath1540 , @xmath1541 ( see @xcite ) .
however , by results of ( * ? ? ?
5.2 ) , the above isomorphism does not extend to higher homological degrees when @xmath1542 . in other words ,
the derived harish - chandra homomorphism @xmath1543 is not a week equivalence for higher dimensional tori @xmath1540 , @xmath1542 .
the above computation of representation homology of the @xmath1076-torus naturally generalizes to riemann surfaces of an arbitrary genus . to be precise ,
let @xmath1544 denote a closed connected orientable surface of genus @xmath1545 . as a 2-dimensional cell complex
, @xmath1544 can be described as the homotopy cofibre of the map @xmath1546 defined by @xmath1547\,[a_2 , b_2]\,\ldots \,[a_g , b_g ] \ , \ ] ] where @xmath1548 denote the @xmath1549- and @xmath1550-cycles on @xmath1544 generating the fundamental group @xmath1551 .
this gives the simplicial group model @xmath1552\,$ ] of @xmath1553 , which , in turn , implies @xmath1554 \cong { \rm hocolim } \,[\,k \leftarrow \o(g ) \xrightarrow{\alpha^g _ * } \o(g^{2g})\,]\ , \cong\ , \o(g^{2 g } ) \otimes^{\l}_{\o(g ) } k\ , \ ] ] where the map @xmath1555 is defined by @xmath1556\,[x_2 , y_2]\,\ldots \,[x_g , y_g])\ , \quad f \in \o(g)\ .\ ] ] by proposition [ rephom0 ] , we conclude @xmath1557 where @xmath1558 is the bimodule with left and and right @xmath1463-module structure given by the maps @xmath1559 and @xmath1560 , respectively . in case when @xmath1561 is generated by a regular sequence , we can also express the representation homology of @xmath1544 as the homology of the commutative dg algebra @xmath1562 , where @xmath1483 is the global koszul complex constructed in section [ s7.1.1 ] : @xmath1563\ .\ ] ] as in the torus case , for a reductive group @xmath3 with a cartan subgroup @xmath1494 , there is an algebra map induced by the derived harish - chandra homomorpism @xmath1564 : @xmath1565^w\ ] ] where @xmath219 operates diagonally on the target .
however , for @xmath1566 , this map seems far from being an isomorphism even for classical groups @xmath34 as in conjecture [ conj2 ] .
let @xmath1567 denote a closed connected non - orientable surface of genus @xmath1568 .
for example , @xmath1569 and @xmath1570 is the klein bottle .
it is known that @xmath1567 is homotopy equivalent to the homotopy cofibre of the map @xmath1571 : @xmath1572 where @xmath1573 denote generators of the corresponding fundamental groups .
this gives the following simplicial group model for @xmath1574 : @xmath1575\ , \ ] ] which , in turn , gives the presentation @xmath1576 \cong { \rm hocolim } \,[\,k \leftarrow \o(g ) \xrightarrow{\beta^g _ * } \o(g^g)\,]\ , \cong\ , \o(g^g ) \otimes^{\l}_{\o(g ) } k\ , \ ] ] where @xmath1577 is defined by @xmath1578 for @xmath1579 .
thus , @xmath1580 by a link @xmath206 in @xmath74 we mean a smooth ( oriented ) embedding of the disjoint union @xmath1581 of ( a finite number of ) copies of @xmath1582 into @xmath74 .
the link complement @xmath1583 is then defined to be the complement of an ( open ) tubular neighborhood of the image of @xmath1129 in @xmath74 . to describe a simplicial group model for @xmath60 we recall two classical facts from geometric topology ( _ cf .
_ @xcite ) .
first , by a well - known theorem of j. w. alexander , every link @xmath1129 in @xmath1584 can be obtained geometrically as the closure of a braid @xmath213 in @xmath74 ( we write @xmath1585 to indicate this relation ) .
second , for each @xmath1586 , the braids on @xmath90 strands in @xmath1584 form a group @xmath215 ( the artin braid group ) , which admits a faithful representation by automorphisms of the free group @xmath71 ( the artin representation ) .
specifically , the group @xmath215 is generated by @xmath1587 elements ( `` flips '' ) @xmath1588 subject to the relations @xmath1589 and in terms of these generators , the artin representation @xmath1590 is given by @xmath1591 to simplify the notation we will identify @xmath215 with its image in @xmath1592 under . the next proposition can be viewed as a refinement of a classical theorem of artin and birman ( * ? ? ?
* theorem 2.2 ) describing the fundamental group of the link complement @xmath1593 in terms of the artin representation ( see remark below ) .
let @xmath1585 be a link in @xmath74 given by the closure of a braid @xmath207 .
then @xmath1594\ , \ ] ] where @xmath213 acts on @xmath71 via the artin representation .
note that the homotopy pushout in coincides with the homotopy coequalizer @xmath1595^{{{\rm{id } } } } \ar@<-0.4ex>[r]_{\beta } & { { \mathbb f}}_n } \ ! ] \,$ ] of the two endomorphisms @xmath1596 and @xmath213 of @xmath71 .
hence , implies @xmath1597 \cong \text{coeq } [ \!\xymatrix { { { \mathbb f}}_n \ar@<0.4ex>[r]^{{{\rm{id } } } } \ar@<-0.4ex>[r]_{\beta } & { { \mathbb f}}_n } \ ! ] \cong \langle x_1 , \ , \ldots\,,\ , x_n\ |\ \beta(x_1 ) = x_1,\ ,
\ldots,\ , \beta(x_n ) = x_n \rangle\ , \ ] ] which is the artin - birman presentation of the link group @xmath1598 . the proof is based on a simple van kampen type argument ( _ cf .
_ @xcite ) .
let s place the @xmath90-braid @xmath213 in a regular position in the region @xmath1599 in @xmath74 , so that its starting points @xmath1600 and end points @xmath1601 are located on the @xmath1370-axis with coordinates @xmath1602 .
the link @xmath1129 is the closure of @xmath1603 obtained by joining the points @xmath1604 to @xmath1605 ( @xmath1606 ) by simple arcs in the region @xmath1607 , as shown in the picture @xmath1608{mayer_vietoris_argument.png}\ ] ] now , let @xmath1583 denote the complement of @xmath1129 . define @xmath1609 with a ( common ) basepoint @xmath335 in @xmath1610 .
it is easy to see that @xmath1611 is homeomorphic to the cylinder over @xmath1612 , which is , in turn , homotopic to @xmath1613 , where @xmath1614 is a two - dimensional disk in ( the @xmath1615-plane ) @xmath1616 encompassing the points @xmath1617 .
similarly , we have @xmath1618 \simeq { { \mathbb d}}^2\!\setminus\!\{q_1 ,
\ldots , q_n\ } \,$ ] , and @xmath1619 under these identifications , the natural inclusions @xmath1620 can be identified with @xmath1621 where the map @xmath1622 is determined ( uniquely up to homotopy ) by the braid @xmath213 and the map @xmath1623 is determined by the trivial braid connecting the points @xmath1624 and @xmath1605 .
thus , we can represent @xmath60 in @xmath1625 as the homotopy pushout of the diagram .
next , recall that @xmath483 can be identified with the mapping class group of @xmath1626 comprising ( the isotopy classes of ) orientation - preserving homeomorphisms that fix pointwise the boundary of @xmath1614 . as a mapping class group , @xmath215 acts naturally on the fundamental group @xmath1627 and the latter can be identified with the free group @xmath71 on generators @xmath1628 represented by small loops in @xmath1629 around the points @xmath1604 .
it is well - known ( see @xcite ) that the action of @xmath483 on @xmath71 arising from this construction is precisely the artin representation .
now , using the map @xmath1623 we identify @xmath1630 with @xmath1613 in and apply the loop group functor to this diagram of spaces . as a result
, we get the equivalence , which completes the proof of the proposition .
to state our main theorem we introduce some notation .
first , observe that , for any algebraic group @xmath3 , the artin representation @xmath1631 induces naturally a braid group action @xmath1632 $ ] , which we denote by @xmath1633 . on the standard generators ,
this action is defined by @xmath1634 now , for a braid @xmath207 , we let @xmath1635 denote the @xmath212-bimodule whose underlying vector space is @xmath211 , the left action of @xmath209 is given by multiplication , while the right action is twisted by the automorphism @xmath1636 .
let @xmath1637 be a link in @xmath74 given by the closure of a braid @xmath207 .
then @xmath1638 \cong \o(g^n ) \otimes_{\o(g^{2n})}^{\l } \o(g^n)_{\beta}\ .\ ] ] consequently , @xmath1639 by proposition [ prop1 ] and lemma [ hopushout ] , we have @xmath1640 & \cong & { \rm hocolim}\,[\,\o(g^n ) \xleftarrow{\,(\beta_*,\,{{\rm{id}}})\ , } \o(g^n ) \otimes_k \o(g^n ) \xrightarrow{\,({{\rm{id}}},\,{{\rm{id}}})\ , } \o(g^n)\,]\\ & \cong & { \rm hocolim}\,[\,\o(g^n ) \xleftarrow{\,(\beta_*,\,{{\rm{id}}})\ , } \o(g^{2n } ) \xrightarrow{\,({{\rm{id}}},\,{{\rm{id}}})\ , } \o(g^n)\,]\\ & \cong & \o(g^n ) \otimes_{\o(g^{2n})}^{\l } \o(g^n)_\beta \ .\end{aligned}\ ] ] this completes the proof of the theorem .
theorem [ hrlink ] exhibits an interesting analogy between the representation homology of link complements in @xmath74 and their legendrian contact homology in the sense of l. ng ( see @xcite ) .
this analogy is explained in the recent paper @xcite , where a new algebraic construction of link contact homology is given .
roughly speaking , in terminology of @xcite , @xmath1641 \,$ ] represents the algebraic ` homotopy closure ' of the braid @xmath207 in the category of simplicial commutative algebras , while the legendrian contact homology of @xmath1593 can be computed from a certain dg category @xmath1642 that represents the homotopy braid closure of @xmath207 in the category of ( small pointed ) dg @xmath4-categories .
note that theorem [ hrlink ] computes the representation homology of the topological _ space _ @xmath1643 , not of the _ link group _ @xmath1644 , which is the fundamental group of @xmath1593 .
even when @xmath206 is a knot in @xmath74 ( i.e. , a link with one component ) , the representation homologies @xmath1645 and @xmath1646 differ , because @xmath1593 is not a @xmath1647-space . ] . in knot theory ,
one is usually interested in representation varieties of the knot group @xmath1644 , so it is important to understand the relation between @xmath1645 and @xmath1646 . a natural way to approach
this problem is to consider @xmath1648 as a link in @xmath10 by adding to @xmath74 one point at infinity .
if @xmath1649 is a knot , by papakyriakopoulos sphere theorem , the complement @xmath1650 is an aspherical space , and @xmath1651 .
hence , for any knot @xmath1129 , @xmath1652 , so it suffices to clarify the relation between @xmath1645 and @xmath1653 . to this end , we observe that the natural inclusion @xmath1654 fits into the cofibration sequence @xmath1655 , so that @xmath1656\ , \ ] ] where @xmath1657 is chosen in such a way that it encloses @xmath206 in @xmath74 .
applying the kan functor to , we get @xmath1658\ .\ ] ] to describe the induced map @xmath1659 , we note that @xmath1660\,$ ] , hence @xmath1661\ .\ ] ] now , if we identify @xmath1662 as in proposition [ prop1 ] , then @xmath1659 is determined by the morphism of diagrams @xmath1663 1 & \lto & { { \mathbb f}}_1 & \rto & 1 \\ \dto & & \dto _ { } & & \dto \\ { { \mathbb f}}_n & \lto^{\ ( \beta , { { \rm{id}}})\ } & { { \mathbb f}}_n \amalg { { \mathbb f}}_n & \rto^{\ ( { { \rm{id } } } , { { \rm{id}}})\ } & { { \mathbb f}}_n \end{diagram}\ ] ] where the map in the middle is given ( on free generators ) by @xmath1664 . note that the left square in commutes because the product @xmath1665 stays fixed under the artin representation for any @xmath207 . the map @xmath1666 induces a map of simplicial commutative algebras @xmath1667 \to
\drep_g[{{\mathbb g}}(\r^3 \!\setminus\ !
l)]\ , \ ] ] which ( to simplify the notation ) we denote by the same symbol . by theorem [ hrlink ] , @xmath1638 \,\cong \,\o(g^n ) \otimes_{\o(g^{2n})}^{\l } \o(g^n)_{\beta}\ .\ ] ] on the other hand , by , @xmath1668 \,\cong\ , k \otimes_{\o(g)}^{\l } k \ , \cong \ , \lambda_k({{\mathfrak m}}/{{\mathfrak m}}^2 ) \cong \lambda_k(\g^*)\ , \ ] ] where the exterior algebra @xmath1669 ) $ ] is regarded as a dg algebra with trivial differential . with these identifications ,
the map is induced by the algebra homomorphism @xmath1670 now , we can regard @xmath1671 as a dg module over the dg algebra @xmath1672 . as a consequence of and theorem [ hrlink ] , we have then the following let @xmath1637 be a link in @xmath10 given by the closure of a braid @xmath207 .
then @xmath1673 \,\cong\ , k \otimes^{\l } _ { \lambda(\g^ * ) } [ \o(g^n ) \otimes_{\o(g^{2n})}^{\l } \o(g^n)_{\beta}]\ .\ ] ] consequenlty , there is a natural spectral sequence @xmath1674 converging to the representation homology of @xmath1675 .
recall that , for coprime integers @xmath720 and @xmath138 , the lens space @xmath2 of type @xmath1676 is defined as the quotient @xmath1677 of the @xmath21-sphere @xmath1678 viewed as the unit sphere in @xmath1679 modulo the ( free ) action of the cyclic group @xmath1680 given by @xmath1681 .
this definition shows that @xmath2 is a compact connected @xmath21-manifold , whose universal cover is @xmath10 and the fundamental group is @xmath1680 .
special cases include @xmath1682 , @xmath1683 and @xmath1684 .
to compute the representation homology of @xmath1685 we will use a well - known topological construction of these spaces in terms of dehn surgery in @xmath10 ( see , e.g. , ( * ? ? ?
recall that if @xmath1686 is a knot in @xmath10 and @xmath1687 are two integer numbers , the @xmath1688 dehn surgery on @xmath9 is a @xmath21-dimensional space obtained by removing from @xmath10 the interior @xmath1689 of a regular tubular neighborhood @xmath1690 , which is a @xmath21-dimensional solid torus @xmath1691 , and then gluing @xmath1692 back to @xmath1693 in such a way that the meridional curve of @xmath1692 is identified with a @xmath1676-curve on the boundary of @xmath1694 . for the trivial knot @xmath1686 , it is easy to see that the @xmath1695 dehn surgery on @xmath9 gives precisely the lens space @xmath2 . in this case , the knot complement @xmath1693 is homeomorphic to the solid torus @xmath1692 , so the space @xmath2 can be obtained by gluing together two solid tori along their boundary . to describe this in more precise terms , we consider the solid torus @xmath1692 as a subset in @xmath1679 : @xmath1696 we identify @xmath1697 as the boundary of @xmath1692 in @xmath1698 and denote by @xmath1699 the natural inclusion . now , for the given pair @xmath1700 of coprime numbers , we choose @xmath1701 , so that @xmath1702 , and define the ` gluing ' map @xmath1703 by factors as @xmath1704 , where @xmath1705 is a homeomorphism of the torus defined by formula that depends on the matrix @xmath1706 .
the assignment @xmath1707 induces an isomorphism from @xmath1708 to the mapping class group of @xmath1448 .
formula shows that @xmath1709 can be obtained as a join of two circles , @xmath1710 , twisted by @xmath1711 . ]
@xmath1712 then the @xmath1695 dehn surgery construction of @xmath1713 can be described as the pushout in @xmath1445 : @xmath1714\ .\ ] ] since @xmath1715 is a cofibration in @xmath1445 , we can replace the colimit in by a homotopy colimit and then replace the diagram of solid tori by a homotopy equivalent diagram of circles : @xmath1716\ .\ ] ] in this diagram , the map @xmath1717 is given by the canonical projection @xmath1718 and @xmath1719 is the composition @xmath1720 defined by @xmath1721 .
now , applying the kan loop group functor , we get a simplicial group model for @xmath2 : @xmath1722 \,\cong \ , { \rm hocolim}\,[\,{{\mathbb f}}_1 \xleftarrow{\pi } { { \mathbb g}}({{\mathbb t}}^2 ) \xrightarrow{\gamma } { { \mathbb f}}_1\,]\ .\ ] ] recall ( see ) that @xmath1723 is given in @xmath37 by the homotopy cofibre of the commutator map @xmath1724 , @xmath1725 \,$ ] , where @xmath1726 and @xmath1550 are the generators of @xmath1451 corresponding to the meridian and longitude in @xmath1448 . in terms of these generators , the maps @xmath1727 and @xmath1728 in
are induced by @xmath1729 where @xmath1730 is a generator of @xmath1450 . now , assume that @xmath34 admits a global koszul resolution @xmath1731 described in section [ s7.1 ] .
then , we have am explicit dg algebra model for @xmath1732 given by @xmath1733 . applying to the derived representation functor
, we get @xmath1734 \,\cong\ , { \rm hocolim}\,[\,\o(g ) \xleftarrow{\pi _ * } { { \mathcal a}}_\ast({{\mathbb t}}^2 , g ) \xrightarrow{\gamma _ * } \o(g)\,]\ , \ ] ] the maps @xmath1735 and @xmath1736 in are determined by ; on the degree @xmath245 component of the dg algebra @xmath1737 , they are given by @xmath1738 @xmath1739 using @xmath1735 and @xmath1736 , we can make @xmath205 into ( left and right ) dg modules over the dg algebra @xmath1737 , which we denote by @xmath1740 and @xmath1741 respectively . with this notation , we have the following result that completes our calculation .
the representation homology of a @xmath21-dimensional lens space @xmath1742 is given by @xmath1743 where @xmath1744 denotes the differential tor taken over the dg algebra @xmath1745 .
in particular , there is an eilenberg - moore homology spectral sequence @xmath1746 converging to the representation homology of @xmath2 .
the above construction of lens spaces generalizes to arbitrary closed @xmath21-manifolds .
specifically , it is well known ( see , e.g. , @xcite ) that every closed connected orientable @xmath21-manifold @xmath216 admits a heegaard decomposition @xmath1747 which can be written as @xmath1748\ , \ ] ] where @xmath1749 is a handlebody of genus @xmath1750 , @xmath1715 is the natural inclusion identifying @xmath1751 and @xmath1728 is a gluing map defined as the composition @xmath1752 , where @xmath1711 is an ( orientation - preserving ) diffeomorphism of @xmath1753 representing an element in the mapping class group @xmath1754 . in particular , for @xmath1755 , the heegaard diagram becomes ; in fact , the lens spaces can be characterized as ( closed ) @xmath21-manifolds that admit heegaard decompositions of genus @xmath1525 .
since @xmath1749 is homotopy equivalent as a cell complex to the bouquet of @xmath1135 circles @xmath1756 , we can represent the homotopy type of @xmath216 by @xmath1757\ .\ ] ] this gives the simplicial group model @xmath1758\,$ ] , and hence @xmath1759\,\cong\ , { \rm hocolim}\,[\,\o(g^g)\,\xleftarrow{\pi _ * } \,{{\mathcal a}}_\ast(\sigma_g , g ) \,\xrightarrow{\gamma_*}\ , \o(g^g)\,]\,\cong\ , \o(g^g ) \,\otimes^{\l}_{{{\mathcal a}}_\ast}\ , \o(g^g)\ , \ ] ] where @xmath1760 is an explicit dg algebra model for the representation homology @xmath1761 ( see section [ s7.1.2 ] ) . as a result
, we have the following generalization of theorem [ lensspaces ] to @xmath21-manifolds of higher genus .
let @xmath216 a closed connected orientable @xmath21-manifold .
assume that @xmath216 has a heegaard decomposition determined by an element @xmath1762 in the mapping class group of @xmath1544 .
then the representation homology of @xmath216 is given by @xmath1763 where @xmath1744 is the differential tor taken over the dg algebra @xmath1764 . in particular , there is an eilenberg - moore homology spectral sequence @xmath1765 converging to the representation homology of @xmath356 .
in this section , we identify representation homology as the hochschild - mitchell ( hyper)homology of a certain bifunctor on the category of finitely generated free groups @xmath234 with values in chain complexes of @xmath4-vector spaces .
let @xmath1766 be a small category . by a @xmath838-_bimodule _ ,
we mean a bifunctor @xmath1767 , which is contravariant in the first argument and covariant in the second .
we write @xmath1768 for the category of @xmath838-bimodules . for any @xmath1769
, one can define the ( hochschild - mitchell ) homology @xmath1770 and cohomology @xmath1771 of @xmath846 with coefficients in @xmath243 . for a precise definition and basic properties of these classical ( co)homology theories
we refer to @xcite ( a good summary can also be found in ( * ? ? ?
* appendix c ) ) . here
, we only recall that @xmath1772 and @xmath1773 are functors ( covariant and contravariant , respectively ) on the category of @xmath838-bimodules , such that @xmath1774 and @xmath1775 are universal @xmath1776-sequences , with @xmath1777 and @xmath1778 being canonically isomorphic to the coend @xmath1779 and the end @xmath1780 of the bifunctor @xmath243 .
moreover , the ( co)homology theories @xmath1781 and @xmath1782 have good functorial properties with respect to the first argument : in particular , any functor @xmath1783 between small categories induces a natural map on homology @xmath1784 , where @xmath1785 is the restriction functor on bimodules defined by @xmath1786 . to express representation homology in terms of hochschild - mitchell homology , we need to slightly extend the above classical setting
specifically , we will consider chain complexes of @xmath838-bimodules , which are simply bifunctors @xmath1787 with values in the category of chain complexes of @xmath4-vector spaces , and define @xmath1788 and @xmath1789 to be the hochschild - mitchell _ hyperhomology _ and the hochschild - mitchell _ hypercohomology _ of @xmath1790 , respectively .
now , given two chain complexes of right and left @xmath838-modules , say @xmath1791 and @xmath1792 , we define the chain complex of @xmath838-bimodules @xmath1793 by assigning to @xmath1794 the tensor product @xmath1795 of the corresponding chain complexes . with this notation , we have for any @xmath53 and any commutative hopf algebra @xmath111 , there is a natural isomorphism @xmath1796 ) \boxtimes_k { \underline{\mathcal h}}\,)\ .\ ] ] for any small category @xmath846 and any @xmath838-modules @xmath1797 and @xmath1798 with values in @xmath1799 , where @xmath7 is a commutative ring , there is a natural ( grothendieck ) spectral sequence ( see , e.g. , ( * ? ? ?
* ( c.10.1 ) ) ) : @xmath1800)\ , \longrightarrow\ , \h_{p+q}[m\otimes^{\l}_{\cc } n]\ .\ ] ] when @xmath7 is a field , this spectral sequence degenerates giving an isomorphism @xmath1801\,$ ] . in our situation , we have @xmath1802 ) \boxtimes_k { \underline{\mathcal h}}\ , ) \,\cong\ , \h_\ast[{\underline{n}}(k[\lgr{x } ] ) \otimes^{\l}_{\ffgr } { \underline{\mathcal h}}]\ , \ ] ] which in composition with the isomorphism of theorem [ dtp ] gives
. lemma [ hrhh ] motivates the following definition .
the _ representation cohomology _ of @xmath272 in @xmath118 is defined by @xmath1803 ) \boxtimes_k { \underline{\mathcal h}}\,)\ ] ] more generally , for any @xmath633-bimodule @xmath1787 , we define the _ representation homology _ and the _ representation cohomology _ of @xmath1790 by @xmath1804 in the case when @xmath243 is an ordinary @xmath633-bimodule ( with values in @xmath79 ) , this definition says that the representation ( co)homology of @xmath1790 is just the classical hochschild - mitchell ( co)homology of @xmath243 . for an arbtirary ( associative unital ) ring @xmath1805 , denote by @xmath1806 the full subcategory of @xmath459-@xmath1807 whose
objects are the free modules @xmath1808 , @xmath1809 . for any @xmath459-bimodule @xmath380 , consider the bifunctor @xmath1810 defined by @xmath1811 .
then , a theorem of pirashvili and waldhausen @xcite asserts that the hochschild - mitchell homology @xmath1812 is naturally isomorphic to the _ topological hochschild homology _
@xmath1813 of the ring @xmath1805 with coefficients in the bimodule @xmath87 .
it is therefore natural to _ define _ the topological hochschild homology of @xmath459 with coefficients in an arbitrary bifunctor @xmath1814 by ( _ cf . _
* chap . 13 ) ) @xmath1815 for @xmath1816 , the category @xmath1817 is equivalent to the category @xmath1818 of finitely generated free abelian groups , which ( as our notation suggests ) is the abelianization of the category @xmath106 .
the abelianization functor @xmath1819 induces a natural map @xmath1820 for any @xmath1821-bimodule @xmath1822 , and conversely , for any @xmath633-bimodule @xmath1823 , associated to the functor @xmath1454 , there is an andr - type spectral sequence ( see ( * ? ? ?
* theorem 1.20 ) ) : @xmath1824 converging to the representation homology of @xmath1790 .
thus , representation homology may be viewed as a non - abelian analogue of topological hochschild homology , and it is natural to ask for ` non - abelian ' analogues of various constructions known for topological hochschild homology . in the next section ,
we outline one such construction which may be thought of as a non - abelian version of the dennis trace map .
recall and @xcite for an extension to arbitrary @xmath1825-bimodules . ] that the classical dennis trace maps the stable homology of the general linear groups of a ring @xmath459 to topological hochschild homology of @xmath1826 : @xmath1827 where @xmath1032 is an arbitrary bimodule over @xmath1806 .
we generalize this map to the non - abelian setting .
let @xmath1828 denote the automorphism group of the free group on generators @xmath1628 .
we will regard @xmath1829 as the automorphism group @xmath1830 of the object @xmath1831 in the category @xmath106 .
there are obvious inclusions @xmath1832 defined by @xmath1833 , where @xmath1834 for @xmath1835 and @xmath1836 .
we set @xmath1837 .
now , consider an arbitrary bimodule @xmath243 on the category @xmath106 , i.e. a bifunctor @xmath1838 .
for each @xmath1839 , let @xmath1840 and define the linear maps @xmath1841 where @xmath1842 and @xmath1843 are induced by the natural inclusion @xmath1844 and the natural projection @xmath1845 , respectively .
put @xmath1846 where the inductive limit is taken with respect to .
next , observe that each @xmath1847 carries a natural @xmath1848-module structure : namely , @xmath1849,@xmath1850 , where @xmath1851 and @xmath1852 .
moreover , for all @xmath1853 , there is a commutative diagram @xmath436 d_n & \rto^{g^\ast \ , g_\ast } & d_n\\ \dto^{p_n^{\ast } \ , ( i_n)_\ast } & & \dto_{p_n^{\ast } \ , ( i_n)_\ast}\\ d_{n+1 } & \rto^{\tilde{g}^\ast \ , \tilde{g}_\ast } & d_{n+1}\\ \end{diagram}\ .\ ] ] as a consequence , the @xmath4-vector space @xmath1854 carries a natural ( inductive ) @xmath1855-module structure .
thus , we can consider the homology groups @xmath1856 for all @xmath279 and @xmath1857 .
since homology commutes with direct limits , we can identify @xmath1858 next , we construct natural maps relating @xmath1859 and @xmath1860 to representation homology @xmath1861 .
we proceed as in @xcite , but with one important modification . instead of using the mac lane isomorphism , which identifies group homology with classical hochschild homology in the case of @xmath1862 and which is not available in our non - abelian setting
, we will use the baues - wirsching isomorphism , which identifies group homology with baues - wirsching homology @xcite . specifically , regarding each group @xmath1863 as a category with a single object , we consider the natural functors @xmath1864 mapping the single object of @xmath1829 to @xmath1865 .
then , by ( * ? ? ?
* remark 1.10 ) , we have natural isomorphisms @xmath1866 where @xmath1867 denotes the baues - wirsching factorization category of a small category @xmath1868 ( see @xcite ) .
similarly , by ( * ? ? ?
* proposition 1.13 ) , there are natural isomorphisms @xmath1869 now , observe that , for each @xmath108 and any @xmath1085-bimodule @xmath1790 , there is a canonical map @xmath1870 induced by embedding each @xmath1871 into the coend diagram of the bifunctor @xmath1790 . since @xmath1872 is an exact functor from the category of left @xmath1873-modules to the category of left @xmath1874-modules , the sequence @xmath1875 forms a covariant left @xmath1776-functor . since @xmath1876 is a universal @xmath1776-functor , the map @xmath1877 extends to canonical linear maps @xmath1878 as in @xcite , it is easy to check that the maps are compatible when passing from @xmath90 to @xmath396
. hence , we can stabilize by passing to the inductive limit as @xmath1879 . with identification ,
the resulting stable map reads @xmath1880 this is a non - abelian analogue of the classical dennis trace map . as in the classical case
, it is natural to ask : when is an isomorphism ? motivated by a theorem of scorichenko @xcite ( see also @xcite ) , we propose a conjectural answer : we conclude this section a few remarks related to conjecture [ conj3 ] .
* a famous theorem of galatius @xcite asserts that natural maps from the symmetric group @xmath1526 to @xmath1829 ( defined by permuting the generators ) induce isomorphisms @xmath1881 this implies that @xmath1882 for all @xmath941 , where @xmath16 is any constant @xmath4-module provided @xmath4 has characteristic @xmath245 ( which we always assume in this paper ) .
conjecture [ conj3 ] implies ( * ? ? ?
* theorem 1 ) , which says that @xmath1883 for all @xmath941 , when @xmath243 is a polynomial bifunctor , _ constant _ with respect to its contravariant argument . indeed , for such bifunctors
, we have @xmath1884 for @xmath941 because @xmath106 has a terminal object .
the direct analogue of conjecture [ conj3 ] is false in the abelian case . indeed ,
if @xmath1032 is a constant bifunctor on @xmath1806 , then @xmath1885 vanishes in positive degrees ( since @xmath1806 has terminal object ) , but @xmath1886 may be highly nontrivial ( see @xcite ) .
the correct version of conjecture [ conj3 ] proved in @xcite replaces the stable group homology with waldhausen s stable @xmath11-theory . in the non - abelian case ,
one case also state a version of conjecture [ conj3 ] for the stable @xmath11-theory of automorphism groups @xmath1829 instead of group homology ; however , we expect that in this case , the two theories are actually isomorphic .
we briefly outline an argument behind this expectation .
let @xmath1887 denote the commutator subgroup of @xmath1888 .
it is known that @xmath1887 is a perfect normal subgroup , hence we can form the ` plus construction ' @xmath1889 let @xmath1890 denote the homotopy fiber of the map @xmath1891 .
we have a canonical group homomorphism @xmath1892 that equips any @xmath1888-module with a @xmath1893-action .
in particular , the @xmath1888-module @xmath1894 arising from a @xmath694-bimodule @xmath243 may be viewed as a @xmath1893-module , and hence defines a local system on @xmath1890 .
the _ stable @xmath11-theory _
@xmath1895 is then defined to be @xmath1896 , the homology of @xmath1890 with coefficients in the local system @xmath1894 .
now , consider the serre spectral sequence associated to the homotopy fibration @xmath1897 : @xmath1898 if @xmath1888 acts _ trivially _ on @xmath1899 ( as it happens in the classical case , see @xcite ) , then , since @xmath1900 is a homology equivalence for trivial coefficients , the above spectral sequence becomes @xmath1901 however , by galatius theorem @xcite , we know that @xmath1902 for @xmath1903 for any constant coefficients over @xmath4
. hence , the above spectral sequence must collapse on the @xmath720-axis , giving a desired isomorphism @xmath1904 .
the dold - kan correspondence is a classical result that establishes an equivalence between the category @xmath1905 of non - negatively graded chain complexes in an abelian category @xmath1906 and the category @xmath1907 of simplicial objects in @xmath1906 . in this appendix
, we will describe a monoidal enrichment of this correspondence relating the category of ( nonnegatively graded ) dg @xmath244-algebras to the category of simplicial @xmath244-algebras for an arbitrary @xmath4-linear operad @xmath244 . for simplicity
, we will fix a commutative ring @xmath4 with unit , and consider only the abelian category @xmath1908 . to any simplicial @xmath4-module @xmath1909 we can associate the chain complex @xmath1910\ ] ] with @xmath1911 for @xmath108 and the differential @xmath1091 given by @xmath1912 .
the assignment @xmath1913 defines a functor @xmath1914 from the category of simplicial @xmath4-modules to the category of connective chain complexes of @xmath4-modules .
the functor @xmath380 is called the _
normalization functor_. a classical theorem due to dold and kan ( see ( * ? ? ?
* theorem 8.4.1 ) ) asserts that @xmath87 is an equivalence of categories . for any simplicial @xmath4-module @xmath1915 ,
the homology groups of the chain complex @xmath1916 are naturally isomorphic to the homotopy groups @xmath1917 of the geometric realization of @xmath60 ( see ( * ? ? ?
* theorem 22.1 ) ) .
this justifies the notation @xmath1918 $ ] , which we use throughout the paper .
the inverse @xmath1919 of the normalization functor is defined as follows . for any chain complex @xmath1920 , the degree @xmath90 part of the simplicial @xmath4-module @xmath1921
is given by @xmath1922 \twoheadrightarrow [ r ] } \ , v_r\ ] ] we think of @xmath1921 as adjoining to @xmath643 the degeneracies of all elements in @xmath643 .
we write an element @xmath1923 in the summand corresponding to @xmath1924 as @xmath1925 .
when @xmath1926 , we simply write this as @xmath1270 , or @xmath1927 if we want to emphasize that we consider @xmath1270 to be an element in @xmath1921 rather than @xmath643 . as suggested by the notation , this determines the degeneracy maps in @xmath1921 : namely , @xmath1928 ( recall from section [ simpobj ] that @xmath1929^{\ast}$ ] ) .
the face maps in @xmath1921 are determined by the requirement that @xmath1930 for all @xmath915 , and the canonical map @xmath1931 , \qquad \qquad \quad x \ , \mapsto \ , \eta(x ) = x\ ] ] commutes with differentials , _
i.e. _ @xmath1932 . since all elements of @xmath1921 other than @xmath1927 are sums of degenerations of @xmath1927
, specifying the face maps on these elements determines all the face maps in @xmath1921 .
this defines a simplicial @xmath4-module @xmath1921 and hence the functor @xmath1919 ( see @xcite for more details ) .
it is easy to check that this functor is indeed the inverse of the normalization functor @xmath380 .
there is an alternative way to define the normalization functor . for each simplicial @xmath4-module @xmath1933
, we can take the chain complex @xmath1934 defined by @xmath1935 then one can show ( see @xcite ) that the canonical map @xmath1936 of chain complexes given by the composition @xmath1937 is an isomorphism . notice that the inverse of the normalization functor has an important feature : the collection of @xmath4-modules @xmath1938 , as well as the degeneracy maps between them , depends only on the graded module @xmath643 and not on its differential .
in other words , defines a functor @xmath1939 from the category @xmath1940 of graded @xmath4-modules to the category @xmath1941 of @xmath1942-systems of @xmath4-modules . similarly , gives a functor @xmath1943 .
this will play a role in our construction of the monoidal dold - kan correspondence in the next section .
it is a classical fact that the dold - kan normalization functor @xmath1944 can be endowed with a symmetric lax monoidal structure . to describe it ,
we first introduce some notations . given two simplicial modules @xmath1945 over a commutative ring @xmath4 , we denote by @xmath1946 the result of applying the tensor product levelwise , _
i.e. , _ @xmath1947 .
then , there is a _ quasi - isomorphism _ of chain complexes @xmath1948 called the _ eilenberg - zilber shuffle map _ , which is natural ( in @xmath60 and @xmath525 ) , symmetric , associative and unital in the obvious sense ( see , e.g. , @xcite for details ) .
this shuffle map allows one to transfer algebraic structures from a simplicial module @xmath16 to its normalization @xmath1949 .
for instance , if @xmath16 is a simplicial associative algebra , then @xmath1949 is a dg algebra ; if @xmath16 is a simplicial commutative algebra , then @xmath1949 is a commutative dg algebra , _
etc_. in general , for any @xmath4-linear operad @xmath244 , one can consider the category @xmath1950 of simplicial @xmath244-algebras as well as the category @xmath1951 of non - negatively graded dg @xmath244-algebras .
if @xmath1952 is a simplicial @xmath244-algebra , then each @xmath90-ary operation @xmath1953 gives a map @xmath1954 one can then use the eilenberg - zilber shuffle maps to construct the maps @xmath1955{{{\rm sh } } } \ , } \
n ( \ , a \ , { \ , \bar{\otimes } \,}\ , \stackrel{n}{\ldots } \ , { \ , \bar{\otimes } \,}\ , a \ , ) { \ , \xrightarrow [ ] { n(\alpha_a(\mu ) ) } \ , } n(a)\ ] ] which form the structure maps for a dg @xmath244-algebra on @xmath1949 .
this defines a functor @xmath1956 in the special case when @xmath244 is the lie operad , this last functor has already appeared in @xcite .
quillen showed that it has a left adjoint in that case .
his proof generalizes directly to an arbitrary operad .
the functor has a left adjoint @xmath1957 .
as in @xcite , for any @xmath1958 , we define @xmath1959 as the following ( degreewise ) coequalizer of simplicial @xmath244-algebras
@xmath1960 ^ -{\alpha_\ast } \ar@<-0.5ex>[r]_-{{{\rm sh}}_\ast } & t_{\cp}(n^{-1}(a ) ) } } \right]\ ] ] where @xmath1961 and @xmath1962 are induced by @xmath1963 and the eilenberg - zilber maps @xmath1964 respectively .
notice that the proof shows that the underlying @xmath1965-system of @xmath244-algebra of @xmath1959 depends only on the graded algebra structure of @xmath16 ( see the discussion at the end of the previous subsection ) .
this observation will allow us to describe the simplicial @xmath244-algebra @xmath1959 in the case when @xmath16 is semi - free .
we first consider the commutative diagrams of functors @xmath1966 ^ -{{\rm forget } } \ar[d]_{n } \ar@{}[rd]|{(1 ) } & { { \tt smod}}_k \ar[d]^{n } \\ { \mathtt{dga}}(\cp ) \ar[r]^-{{\rm forget } } & { { \tt ch}_{\geq 0}}(k ) } } \qquad \qquad \vcenter { \xymatrix { \salg(\cp ) \ar@{}[rd]|{(2 ) } & { { \tt smod}}_k \ar[l]_-{t_{\cp } } \\ { \mathtt{dga}}(\cp ) \ar[u]^{n^ * } & { { \tt ch}_{\geq 0}}(k ) \ar[u]_{n^{-1 } } \ar[l]_-{t_{\cp } } } } \ ] ] where we denote by @xmath1967 the free algebra functors in both simplicial and dg contexts .
the square @xmath1968 obviously commutes up to isomorphism of functors .
the square @xmath756 is obtained by replacing every functor on the square @xmath1968 by its left adjoint .
therefore , it also commutes up to isomorphism of functors .
the commutativity ( up to isomorphism ) of the square @xmath756 can be written as @xmath1969 in other words , @xmath1970 of a free dg @xmath244-algebra is free .
the same is true for semi - free algebras .
recall that a dg @xmath244-algebra is said to be _ semi - free _ if its underlying graded algebra is free over a degreewise free graded @xmath4-module @xmath643 . similarly , a simplicial @xmath244-algebra @xmath16 is said to be _
_ semi - free__-algebra is called semi - free if there is a collection of subsets @xmath1971 , called a _ basis _ ,
that is closed under degeneracies and that @xmath1972 for each @xmath90 .
it is clear that our definition implies this . to see the converse ,
notice that any basis element that is not the degeneracy of any other basis element is in fact non - degenerate in the underlying simplicial set of @xmath16 .
let @xmath643 be the graded @xmath4-module with a basis given by these non - degenerate basis elements .
then an application of ( * ? ? ?
* lemma i.2.11 ) shows that @xmath1973 as a @xmath1965-system of @xmath244-algebras . ]
if its underlying @xmath1965-system of @xmath244-algebras is of the form @xmath1973 for a degreewise free graded @xmath4-module @xmath643 .
the above discussion leads to the following [ nstar_semifree_prop ] the functor @xmath1957 sends semi - free dg @xmath244-algebras to semi - free simplicial @xmath244-algebras .
we have seen that @xmath1970 sends free algebras to free algebras .
since the underlying @xmath1965-system of @xmath1959 depends only on the graded algebra structure of @xmath16 , the result follows .
next , we consider the adjunction map @xmath1974 in the case when @xmath1975 is semi - free over a graded complex @xmath643 . to describe this map ,
we first give a different interpretation of the eilenberg - zilber shuffle map .
namely , we view it a collection of maps that connect two symmetric monoidal structures on the category @xmath1976 of chain complexes on @xmath4 . we will use the quotient " form of the normalization functor .
thus , we consider the equivalence of categories @xmath1977 .
one can use this equivalence to transport the symmetric monoidal structure @xmath1978 on @xmath1979 to a symmetric monoidal structure @xmath1980 on @xmath1976 .
precisely , we define @xmath1981 for @xmath1982 .
then the eilenberg - zilber shuffle maps can be written as @xmath1983 now , suppose that a dg @xmath244-algebra @xmath1958 is semi - free over a graded @xmath4-module @xmath643 , _
i.e. , _ @xmath1984
then by as well as the discussion that follows , the dg @xmath244-algebra @xmath1985 has a similar description @xmath1986 moreover , the adjunction map @xmath1987 is given by @xmath1988 this description of the adjunction map will be useful in the next subsection when we compare the model structures on simplicial @xmath244-algebras and dg @xmath244-algebras . by (
* section ii.4 , theorem 4 ) , there is a model structure on the category @xmath1950 of simplicial @xmath244-algebras where a map @xmath1989 is a weak equivalence ( resp . ,
fibration ) if and only if the map of the underlying simplicial sets is a weak equivalence ( resp . ,
fibration ) .
moreover , it is shown in @xcite that if @xmath4 is a field of characteristic @xmath245 , then the category @xmath1951 of dg @xmath244-algebras also has a model structure in which a map @xmath1989 is a weak equivalence ( resp .
, fibration ) if and only if the map of the underlying ( connective ) chain complexes is a weak equivalence ( resp . , fibration ) . from now on
, we assume that @xmath4 is a field of characteristic @xmath245 , and the categories @xmath1950 and @xmath1951 are equipped with the model structures described above .
then the normalization functor @xmath1990 preserves fibrations and weak equivalences .
therefore , the adjunction @xmath1991 & \salg(\cp ) \ , : \ , n \ar@<0.5ex>[l ] } \ ] ] is a quillen pair .
in fact , we have the following theorem , which is the main result of this appendix . [ nnstar_equiv ]
the quillen pair is a quillen equivalence .
it suffices to show that , for any semi - free dg @xmath244-algebra @xmath1975 , unit of the adjunction is a weak equivalence . composing this adjunction map with the isomorphism @xmath1992
, we can consider the map @xmath1993 which depends only on the underlying graded @xmath244-algebra structure of @xmath16 . if @xmath16 is free , _
i.e. , _ when the differential on @xmath1975 is induced by the differential on a chain complex @xmath643 , then for each @xmath249 , the map @xmath1994 is a quasi - isomorphism because it is induced by the eilenberg - zilber shuffle map ( which is always a quasi - isomorphism ) .
since @xmath4 is a field of characteristic @xmath245 , the same is true when we pass to @xmath1995-coinvariants @xmath1996 this shows that the map is a quasi - isomorphism in the case when @xmath16 is free . in the general case , when @xmath1975 is semi - free over a graded @xmath4-module @xmath643 , choose a homogeneous basis of @xmath643 , and assign a weight grading @xmath1997 for each such basis element @xmath1270 .
this induces a grading on @xmath16 , where an element @xmath1998 has weight grading @xmath1999 .
the underlying @xmath1965-system of @xmath1959 is given by @xmath2000 .
therefore , its elements in degree @xmath203 are sums of elements of the form @xmath2001 where @xmath2002 \twoheadrightarrow [ r_i]$ ] and @xmath2003 .
assign the weight grading @xmath1999 to this element .
then it is clear that all the degeneracy maps preserve this weight grading .
this induces a weight grading in the normalization @xmath2004 .
moreover , the map preserves this grading . in general , the differentials on both sides of do not preserve the grading .
however , if we let @xmath2005 be the filtration on @xmath16 induced by the weight grading , then we can always choose the weight grading on the homogeneous basis of @xmath643 inductively so that @xmath2006 . moreover , if we let @xmath2007 be the filtration on @xmath2004 induced by the weight grading on @xmath2004 , then we have @xmath2008 for all @xmath90 .
this can be shown , for example , by considering the filtration @xmath2009 on @xmath1959 induced by the weight grading and showing that the face maps preserve the filtration , _
@xmath2010 for all face maps @xmath290 .
therefore , both the domain and target of the map of chain complexes admit filtrations by subcomplexes , such that the map preserves these subcomplexes .
since these filtrations are induced by gradings , the graded @xmath4-modules associated to these filtrations can be canonically identified with the original graded @xmath4-modules , _
i.e. , _ we have @xmath2011 as graded @xmath4-modules .
in fact , a direct inspection of both sides shows that these isomorphisms are actually isomorphisms of chain complexes , where @xmath643 has zero differentials .
hence , the induced map @xmath2012 on the associated graded chain complexes coincides with the adjunction map @xmath2013 for the free algebra @xmath2014 with zero differential .
this map is a quasi - isomorphism by our argument in the free case .
since the filtrations @xmath2015 and @xmath2016 are bounded below and exhaustive , the map induces an isomorphism on homology by the eilenberg - moore comparison theorem ( * ? ? ?
5.5.11 ) .
as mentioned above , in the case when @xmath244 is the lie operad , theorem [ nnstar_equiv ] goes back to quillen . if @xmath244 is the associative operad ( and @xmath4 is _ any _ field ) , theorem [ nnstar_equiv ] is proved ( by a different method ) in @xcite .
fix a field @xmath4 of characteristic @xmath245 . by theorem [ nnstar_equiv ] ,
there is a quillen equivalence between the model category @xmath2017 of simplicial commutative algebras and the model category @xmath2018 of connective commutative differential graded algebras : @xmath2019 & { \mathtt{scomm\,alg}}_k \ , : \ , n \ar@<0.5ex>[l ] } \ .\ ] ] [ finite_type_equiv ] for an object @xmath2020\in \ho({\mathtt{scomm\,alg}}_k)$ ] , the following conditions are equivalent : + @xmath1968 @xmath2021 is a finitely generated @xmath4-algebra , and each @xmath2022 is finitely generated as a module over @xmath2021
. + @xmath756 the homotopy type @xmath2023 ) \in \ho(\cdga_k^+)$ ] has a representative @xmath2024 such that @xmath1255 is semi - free dg algebra over @xmath4 with finitely many generators in each homological degree .
+ @xmath2025 the homotopy type @xmath2020 $ ] has a representative @xmath2026 in @xmath2017 such that each @xmath2027 is finitely generated over @xmath4 . for any @xmath2020 \,\in\,\ho({\mathtt{scomm\,alg}}_k)$ ] , we can define an augmentation of @xmath2020 $ ] to be either a map @xmath2020 { \xrightarrow{}}[k]$ ] in @xmath2028 , or a map @xmath2029 in @xmath2017 , or a map @xmath2030 in @xmath429 ( which are clearly equivalent ) .
we will denote an augmentation as @xmath2031 , thinking of it as a @xmath4-point on the affine scheme @xmath2032 .
any augmentation @xmath2031 allows one to linearize @xmath16 .
namely , let @xmath2033 be the kernel of the map @xmath2034 , then we define @xmath2035 .
this gives the linearization functor from the category of augmented simplicial commutative algebra to the category of simplicial @xmath4-modules @xmath2036 the functor @xmath2037 has a right adjoint , given by assigning the simplicial commutative algebra @xmath2038 to a simplicial @xmath4-module @xmath356 .
this forms a quillen adjunction , and allows one to define the derived functor @xmath2039 which is the classical cotangent complex @xmath2040 ( see @xcite ) .
given a map @xmath2041 { \xrightarrow{}}[b]$ ] in @xmath2028 , any augmentation @xmath2042 { \xrightarrow{}}[k]$ ] of @xmath2043 $ ] determines an augmentation @xmath2044 { \ , \xrightarrow[]{f } \ , } [ b ] { \ , \xrightarrow[]{\varepsilon_x } \ , } [ k]$ ] of @xmath2020 $ ] .
the map @xmath384 is said to be _ tale _ at @xmath1270 if the induced map @xmath2045 is an isomorphism in @xmath2046 .
let @xmath2047 be the algebraic closure of @xmath4 . then
a map @xmath2041 { \xrightarrow{}}[b]$ ] in @xmath2028 is called _ tale _ if its base change @xmath2048 { \xrightarrow{}}[b_{\bar{k}}]$ ] to the algebraic closure @xmath2047 is tale for every augmentation @xmath2049 { \xrightarrow{}}[\bar{k}]$ ] .
the proposition below roughly says that a simplicial commutative algebra of quasi - finite type is determined by its underlying scheme as well as its derived cotangent spaces .
when stated in terms of differential graded commutative algebras , this result follows either from ( * ? ? ?
* proposition 1.3 ) or ( * ? ? ?
* corollary 2.7 ) .
the following version for simplicial commutative algebras then follows from theorem [ nnstar_equiv ] .
[ behrend_prop ] let @xmath2041 { \xrightarrow{}}[b ] $ ] be a morphism in @xmath2028 between two objects of quasi - finite type .
suppose @xmath384 is tale and induces an isomorphism in @xmath2050{\cong } \ , } \pi_0(b)$ ] .
then @xmath384 is an isomorphism in @xmath2028 .
homotopical arguments often require imposing the cofibrancy condition on objects and morphisms between simplicial commutative algebras .
as observed in @xcite , the above proposition allows one to replace the cofibrancy condition by certain smoothness conditions . to this end , we introduce the following replacement of cofibrations .
we say that a map @xmath2051 of simplicial commutative algebras is a _
smooth extension _ if it can be written as an ( infinite ) composition @xmath2052 such that each map @xmath2053 is a pushout of the form @xmath2054 \otimes c_n \ar[r ] \ar[d ] & \delta[n ] \otimes c_n \ar[d ] \\
sk_{n-1}(a ) \ar[r ] & sk_n(a ) } \ ] ] where @xmath2055 is some ( formally ) smooth commutative algebra in @xmath429 .
then , we say that a simplicial commutative algebra @xmath16 is _ smooth _ if the structure map @xmath2056 is a smooth extension .
notice that smooth extensions are preserved under compositions and pushouts .
moreover , if we focus on cotangent complexes , then smooth objects and smooth extensions between objects share the same crucial properties as cofibrant objects and cofibrations between objects .
this is the content of the following proposition , for a detailed proof of which we refer to @xcite .
[ smooth_obj_lin ] @xmath1968 suppose that @xmath16 is a smooth simplicial commutative algebra , then for any augmentation @xmath2057 , the canonical map @xmath2058 is an isomorphism in @xmath2059 .
+ @xmath756 suppose that @xmath2051 is a smooth extension of simplicial commutative algebras , then for any augmentation @xmath2060 , the induced map @xmath2061 on the linearizations is a cofibration in @xmath2062 .
recall that , if @xmath2067 is a reedy cofibrant diagram in any model category , then @xmath806 is objectwise cofibrant .
a similar statement is true for reedy smooth diagrams @xmath2064 .
namely , one can write each @xmath2068 as a ( possibly transfinite in each degree , and all the morphism sets in @xmath2063 are finite .
this condition is satisfied in all the examples of reedy categories that we consider in this paper . ] ) composition of smooth extensions , and therefore it is itself a ( direct limit of ) smooth simplicial commutative algebras .
thus , one can apply proposition [ smooth_obj_lin ] to deduce the following [ reedy_smooth_lin ] suppose that @xmath2064 is a reedy smooth diagram , then for each @xmath2065 and each @xmath2047-valued augmentation @xmath2069 , we have + @xmath1968 the canonical map @xmath2070 is an isomorphism in @xmath2059 .
+ @xmath756 the latching map @xmath2066 induces a cofibration on the linearization @xmath2071 in @xmath2062 .
we will be primarily concerned with the case when the reedy category @xmath2063 is the simplicial category @xmath247 . in this case , one can tensor any diagram @xmath2072 with a simplicial set @xmath2073 to obtain @xmath2074 .
the latching maps for @xmath16 can be written in terms of these tensor products .
in fact , by ( * ? ? ?
* proposition 16.3.8 ) , the latching map for @xmath16 at @xmath248\in \delta$ ] can be identified with @xmath2075 \otimes_{\delta } a \,\longrightarrow \ , \delta[n ] \otimes_{\delta } a \ , \cong \ , a(\alpha)\ ] ] for each fixed @xmath60 , the functor @xmath2076 has a total left derived functor .
the corresponding derived tensor product @xmath2077 is represented by the tensor prodcut @xmath2078 for any reedy cofibrant replacement @xmath2079 of @xmath16 .
recall that a simplicial set @xmath60 is _ finite _ if it has finitely many nondegenerate elements .
the following lemma asserts that tensoring over a finite simplicial set preserves quasi - finiteness .
it can be proved by induction on the filtration of @xmath60 by its skeletons @xmath2080 ( see @xcite for details ) .
[ tensor_finite_type ] let @xmath60 be a finite simplicial set .
suppose that @xmath2081 \mapsto \{a^n_m\}_{m \geq 0}$ ] is a cosimplicial simplicial commutative algebra such that each @xmath2082 is finitely generated , then both the tensor product @xmath2083 and the derived tensor product @xmath2084 are of quasi - finite type .
[ reedy_smooth_tensor ]
let @xmath60 be a finite simplicial set .
suppose that @xmath2072 is a reedy smooth cosimplicial simplicial commutative algebras , then the canonical map @xmath2085 is an isomorphism in @xmath2028 .
the @xmath2086 part of both @xmath2084 and @xmath2083 are given by @xmath2087 .
therefore , by proposition [ behrend_prop ] and lemma [ tensor_finite_type ] , it suffices to show that @xmath720 is tale . to this end , consider the category of simplicies @xmath2088 .
being a left kan extension , the tensor product @xmath2083 can be written as the colimit of the diagram @xmath2089{a } \ , } { \mathtt{scomm\,alg}}_k$ ] : @xmath2090,c ) \in
\delta \downarrow x}{\colim } \ , a^{n } \ .\ ] ] now , choose any reedy cofibrant resolution @xmath2091 , where @xmath2092 , and consider the induced map @xmath2093,c )
\in \delta \downarrow x}{\colim } \ , q^{n } \ , { \ , \xrightarrow[]{p } \ , } \ , \underset{([n],c ) \in
\delta \downarrow x}{\colim } \ , a^{n } \ , \cong \ , x \otimes_{\delta } a \ .\ ] ] for each @xmath2094,c )
\in \delta \downarrow x$ ] , any @xmath2047-valued augmentation @xmath2095 induces an augmentation ( still denoted as @xmath2031 ) of the simplicial commutative algebra @xmath2096 .
this determines a map @xmath2097 of @xmath2098-diagrams of augmented simplicial commutative algebras . by ( the proof of ) (
* proposition 16.3.12 ) , both of @xmath2099 and @xmath2100 are still reedy smooth when considered as @xmath2098-diagrams of simplicial commutative algebras .
therefore , by lemma [ reedy_smooth_lin](2 ) , both of @xmath2101 and @xmath2102 are reedy cofibrant as @xmath2098-diagrams of simplicial @xmath4-modules .
moreover , by lemma [ reedy_smooth_lin](1 ) , the linearization map @xmath2103 is a degreewise quasi - isomorphism .
therefore , @xmath2104 induces a quasi - isomorphism on colimits . in other words , for any augmentation @xmath2095 , we have a quasi - isomorphism @xmath2105 since @xmath60 is finite , @xmath2083 is a smooth simplicial commutative algebra .
therefore , by lemma [ reedy_smooth_lin](1 ) , the linearization @xmath2106 appearing as the target of the map @xmath2107 represents its derived linearization . by cofibrancy of @xmath2079 and
hence @xmath2078 , the same is true for the domain of this map .
this shows that @xmath2108 is tale , which completes the proof .
the above argument can be modified to show that , in some cases , to compute homotopy colimit , it suffices to resolve a diagram by a reedy smooth diagram . to this end
, we adopt the terminology of @xcite , and say that a category @xmath2063 is _ very small _ if 1 .
@xmath2063 has a finite set of objects and a finite set of morphisms , and 2 .
there exists an integer @xmath380 such that , if @xmath2109{f_1 } \ , } a_1 { \ , \xrightarrow[]{f_2 } \ , } \ldots { \ , \xrightarrow[]{f_n } \ , } a_n \,$ ] is a string of composable morphisms of @xmath2063 with @xmath2110 , then some @xmath2111 is the identity morphism in @xmath2063 . for any object @xmath2065 in such a category , define its degree as the largest integer @xmath90 such that there exists a string @xmath2109{f_1 } \ , } a_1 { \ , \xrightarrow[]{f_2 } \ , } \ldots { \ , \xrightarrow[]{f_n } \ , } a_n = \alpha \,$ ] of composable morphisms , with none of the morphisms @xmath2111 being the identity .
then , every non - identity morphism raises degree .
therefore , the category @xmath2063 is a reedy category , with @xmath2112 , and @xmath2113 consisting only of the identity morphisms .
[ reedy_smooth_colim ] let @xmath2063 be a very small category , and let @xmath1044 be an @xmath2063-shaped diagram of simplicial commutative algebras .
suppose that @xmath1044 is reedy smooth , then the canonical map @xmath2114 is an isomorphism in @xmath2028 .
first , we show that both @xmath2115 and the ordinary colimit @xmath2116 are of quasi - finite type .
the fact that @xmath2116 is of quasi - finite type follows from condition ( 3 ) of proposition [ finite_type_equiv ] . to see that @xmath2115 is of quasi - finite type ,
we first construct inductively a cofibrant resolution @xmath2079 of @xmath1044 such that each @xmath2117 is degreewise finitely generated .
then , @xmath2118 is again of quasi - finite type by condition ( 3 ) of proposition [ finite_type_equiv ] .
this allows one to apply proposition [ behrend_prop ] to the map @xmath2119 .
any augmentation @xmath2120 determine a compatible system of augmentations @xmath2121 , which can be further pulled back to a compatible system of augmentations @xmath2122 .
taking linearizations gives a map @xmath2123 the assumption that @xmath2124 is reedy smooth implies ( see lemma [ reedy_smooth_lin ] ) that this map is a termwise quasi - equivalence map of reedy cofibrant @xmath2063-shaped diagram in @xmath2062 .
therefore , @xmath384 induces a quasi - isomorphism on colimits : i.e. , the map @xmath2125 is a quasi - isomorphism . by smoothness of @xmath2116 and cofibrancy of @xmath2126 ,
this map represents a map on derived linearizations as well .
this shows that @xmath2127 is tale , finishing the proof .
suppose we are given a diagram @xmath2128 $ ] of simplicial groups , where @xmath509 is semi - free , and the two arrows are semi - free extensions .
let @xmath1044 be the diagram @xmath2129 $ ] obtained from @xmath2128 $ ] by applying the representation functor .
then we have to show that the canonical map @xmath2114 is an isomorphism in @xmath2028 .
since the weak equivalences in @xmath2017 are preserved under filtered colimits , it suffices to assume that the simplicial groups @xmath2130 and @xmath2131 are of finite type ( _ i.e. , _ they are semi - free with finitely many nondegenerate generators ) . while @xmath1044 is not , in general , reedy cofibrant , it is reedy smooth for any algebraic group @xmath3 .
the result therefore follows from proposition [ reedy_smooth_colim ] .
j. f. adams , _ on the cobar construction _ , proc .
usa * 42 * ( 1956 ) , 409412 .
m. andr , _ homologie des algbres commutatives _ , grund .
, springer verlag , 1974 .
d. ayala and j. francis , _
factorization homology of topological manifolds _ , j. topology * 8 * ( 2015 ) , 10451084
. h. j. baues and g. wirsching , _ cohomology of small categories _ ,
j. pure appl .
algebra * 38 * ( 1985 ) , no . 2 - 3 , 187211 .
k. behrend , _ differential graded schemes i : perfect resolving algebras _ ; arxiv : math/0212225 yu . berest , g. khachatryan and a. ramadoss , _ derived representation schemes and cyclic homology _ , adv . math . * 245 * ( 2013 ) , 625689 .
berest and a. ramadoss , _ stable representation homology and koszul duality _ , j. reine angew
* 715 * ( 2016 ) , 143187 .
berest , g. felder , a. patotski , a. c. ramadoss and t. willwacher , _ representation homology , lie algebra cohomology and the derived harish - chandra homomorphism _ , j. eur
( 2017 ) , to appear ; ` arxiv:1410.0043 ` .
berest , g. felder , a. patotski , a. c. ramadoss and t. willwacher , _ chern - simons forms and higher character maps of lie representations _
, internat .
res . notices * 2017 * , no .
1 , 158212 .
berest , g. felder and a. ramadoss , _ derived representation schemes and noncommutative geometry _ , contemp . math .
* 607 * ( 2014 ) , 113162 .
berest , a. eshmatov and w .- k .
yeung , _ perverse sheaves and knot contact homology _ , c. r. math . acad .
paris , ser .
i ( 2017 ) , http://dx.doi.org/10.1016/j.crma.2017.02.007 .
berest , a. ramadoss and y. zhang , _ dual hodge decompositions and derived poisson brackets _ , selecta math .
( 2017 ) , doi:10.1007/s00029 - 017 - 0309 - 7 .
j. s. birman , _ braids , links , and mapping class groups _ , princeton univ .
press , princeton , n.j .
m. brun , g. carlsson and b. dundas , _ covering homology _ , adv . math .
* 225 * ( 2010 ) , 31663213 .
u. buijs , y. flix , a. murillo and d. tanr , _ lie models of simplicial sets and representability of the quillen functor _ , ` arxiv:1508.01442 ` u. buijs , y. flix , a. murillo and d. tanr , _ the homotopy theory of complete differential graded lie algebras _ , ` arxiv:1601.05331 ` u. buijs , y. flix , a. murillo and d. tanr , _ maurer - cartan elements in the lie models of finite simplicial complexes _
, ` arxiv:1606.08794 ` . h. cartan and s. eilenberg , _ homological algebra _ , princeton university press , princeton , n.j . ,
i. ciocan - fontanine and m. kapranov , _ derived quot schemes _ , ann .
ens * 34 * ( 2001 ) , 403440 .
i. ciocan - fontanine and m. kapranov , _ derived hilbert schemes _ , j. amer .
soc . * 15 * ( 2002 ) , no .
4 , 787815 . c. chevalley , _ the betti numbers of the exceptional simple lie groups _ , proc . intern .
congress of math . ,
cambridge , mass . , 1950 ,
ii , amer . math .
soc . , providence , r. i. , 1952 , pp .
d. cooper , m. culler , h. gillet , d. long , p. shalen , _ plane curves associated to character varieties of 3-manifolds _ , invent . math . * 118 * ( 1994 ) , 4784 .
e. b. curtis , _ some relations between homotopy and homology _ , ann . math . * 82 * ( 1965 ) , 386413 .
a. djament and c. vespa , _
sur lhomologie des groupes dautomorphismes des groupes libres coefficients polynomiaux _ , comment .
* 90 * ( 2015 ) , 3358 . v.
g. drinfeld , _ on quasitriangular quasi - hopf algebras and on a group that is closely connected with @xmath2132 _ , leningrad math . j. , * 2 * ( 1991 ) , 829860 . w. dwyer and j. spalinski , _ homotopy theories and model categories _ in _ handbook of algebraic topology _ ,
elsevier , 1995 , pp .
. b. feigin , _ on the cohomology of the lie algebra of vector fields and of the current algebra _
sovietca * 7 * ( 1988 ) , 4962 .
y. felix , s. halperin and j .- c .
thomas , _ rational homotopy theory _ , graduate texts in mathematics * 205 * , springer verlag , new york , 2001 .
y. felix , s. halperin and j .- c .
thomas , _ differential graded algebras in topology _ in _ handbook of algebraic topology _ , elsevier , 1995 , pp . 829865 .
s. fishel , i. grojnowski and c. teleman , _ the strong macdonald conjectures and hodge theory on the loop grassmannian _ ,
ann . math .
( 2 ) * 168 * ( 2008 ) , 175220 .
v. franjou , e. m. friedlander , t. pirashvili and l. schwartz , _ rational representations , the steenrod algebra and functor homology _ , panoramas et synthses * 16 * , soc .
france , paris , 2003 .
b. fresse , _ homotopy of operads and grothendieck - teichmller groups _ , vol .
ii . to be published as vol 217.2 of _ mathematical surveys and monographs _ ,
s. galatius , _ stable homology of automorphism groups of free groups _ , ann . of math .
( 2 ) * 173 * ( 2011 ) , no . 2 , 705 - 768 .
i. galvez - carrilo , f. neumann and a. tonks , _ andre spectral sequences for baues - wirsching cohomology of categories _ , j. pure appl .
algebra * 216 * ( 2012 ) , no .
11 , 25492561 .
i. glvez - carrilo , f. neumann and a. tonks , _ thomason cohomology of categories _ , j. pure appl . alg .
* 217 * ( 2013 ) , 21632179 .
s. i. gelfand and yu .
manin , _ methods of homological algebra _ , springer verlag , berlin , 2003 .
e. getzler and m. kapranov , _ cyclic operads and cyclic homology _ , geometry , topology and physics , 167 - 201 , conf .
lecture notes geom .
topology , iv , intl . press ,
cambridge ma 1995 .
g. ginot , _ notes on factorization algebras , factorization homology and applications _ , in _ mathematical aspects of quantum field theories _ , math . phys .
springer , cham , 2015 , pp .
g. ginot , t. tradler and m. zeinalian , _ higher hochschild homology , topological chiral homology and factorization algebras _ , comm .
* 326 * ( 2014 ) , 635686 .
v. ginzburg , _ lectures on noncommutative geometry _ , ` arxiv : math/0506603 ` . r. goerss and r. jardine , _ simplicial homotopy theory _ , modern birkhuser classics .
birkhuser verlag , basel , 2009 .
m. hartl , t. pirashvili and c. vespa , _ polynomial functors from algebras over a set - operad and nonlinear mackey functors _ , int .
res . not .
imrn * 2015 * , no . 6 , 14611551 .
k. hess and a. tonks , _ the loop group and the cobar construction _ , proc .
* 138 * ( 2010 ) , 18611876 .
v. hinich , _ homological algebra of homotopy algebras _ , comm .
algebra * 25 * ( 1997 ) , 32913323 .
v. hinich , _ dg coalgebras as formal stacks _ , j. pure appl .
algebra * 162 * ( 2001 ) , 209250 .
p. s. hirschhorn , _ model categories and their localizations _ , mathematical surveys and monographs * 99 * , ams , providence , ri , 2003
. j. hollender and r. m. vogt , _ modules of topological spaces , applications to homotopy limits and @xmath1887 structures _ , arch
* 59 * ( 1992 ) , 115129 .
m. hovey , _ model categories _ , mathematical surveys and monographs * 63 * , amer . math .
soc . , providence , ri , 1999 .
j. huebschmann , _ extended moduli spaces , the kan construction , and lattice gauge theory _ ,
topology * 38*(1999 ) , 555596 .
m. jibladze and t. pirashvili , _ cohomology of algebraic theories _ , j. algebra * 137 * ( 1991 ) , no .
2 , 253296 .
j. d. s. jones , _
cyclic homology and equivariant homology _ , invent . math . * 87 * ( 1987 ) , 403423 .
d. m. kan , _ on homotopy theory and c.s.s . groups _ , ann
* 68 * ( 1958 ) , 3853 .
d. m. kan , _ on c.s.s . complexes _
, amer . j. math .
* 78 * ( 1957 ) , 449476 .
m. kapranov , _ injective resolutions of @xmath2133 and derived moduli spaces of local systems _ , j. pure appl . alg .
* 155 * ( 2001 ) , 167179 .
m. kassabov and a. patotski , _
character varieties as a tensor product _ , ` arxiv:1610.02694 ` b. keller , _
@xmath16-infinity algebras , modules and functor categories _ , contemp . math .
* 406 * ( 2006 ) , 6793 .
p. kirk and e. klassen , _ chern - simons invariants of @xmath21-manifolds and representation spaces of knot groups _ , math .
* 287 * ( 1990 ) , 343367 .
p. kirk and e. klassen , _ chern - simons invariants of @xmath21-manifolds decomposed along tori and the circle bundle over the representation space of @xmath2134 _ , comm .
* 153 * ( 1993 ) , 521557 .
s. klaus and m. kreck , _ a quick proof of the rational hurewicz theorem and a computation of the rational homotopy groups of spheres _ , math .
cambridge philos .
soc . * 136 * ( 2004 ) , 617623 .
m. kontsevich and a. rosenberg , _ noncommutative smooth spaces _ in _ the gelfand mathematical seminars 1996 - 1999 _ , birkhuser boston , boston , 2000 , pp .
lin , _ a knot invariant via representation spaces _
, j. differential geom .
* 35 * ( 1992 ) , 337357 .
loday , _ oprations sur lhomologie cyclique des algbres commutatives _ , invent . math .
* 96 * ( 1989 ) , 205230 .
cyclic homology _ ,
wiss . * 301 * , 2nd edition , springer - verlag , berlin , 1998 .
loday , _ homotopical syzygies _ ,
contemp . math .
* 265 * ( 2000 ) , 99127 .
a. lubotsky and a. magid , _ varieties of representations of finitely generated groups _ , mem .
* 58 * ( 1985 ) , no .
336 , 117 pp .
s. maclane , _ categorical algebra _ , bull .
* 71 * ( 1965 ) , 40106 .
m. majewski , _ rational homotopical models and uniqueness _ , mem .
* 143 * ( 2000 ) , no .
682 , 149 pp .
j. p. may , _ simplicial objects in algebraic topology _ , chicago lectures in mathematics , university of chicago press , chicago , il , 1992 .
c. mcmullen , _ the alexander polynomial of a 3-manifold and the thurston norm on cohomology _ , ann .
sci . cole norm .
* 35 * ( 2002 ) , 153171 .
j. milnor , _ on the construction @xmath2135 $ ] _ in _ algebraic topology _ , j. f. adams ( ed . ) , lecture notes of the london math . soc . * 4 * ( 1972 ) , 119136 g. massuyeau and v. turaev , _ brackets in representation algebras of hopf algebras _
, ` arxiv:1508.07566 ` .
b. mitchell , _ rings with several objects _ , adv .
* 8 * ( 1972 ) , 1161 .
l. ng , _ knot and braid invariants from contact homology i _ ,
geom . topol .
* 9 * ( 2005 ) , 247297 .
l. ng , _ knot and braid invariants from contact homology ii _ ,
geom . topol .
* 9 * ( 2005 ) , 16031637 .
t. pirashvili , _ hodge decomposition for higher order hochschild homology _ , ann .
sci . cole norm .
( 4 ) * 33 * ( 2000 ) , 151179 .
t. pirashvili , _ on the prop corresponding to bialgebras _ , cahiers de topologie et gom
. diff cat .
* 43 * ( 2002 ) , 221239 .
t. pirashvili and f. waldhausen , _ mac lane homology and topological hochschild homology _ , j. pure appl .
* 82 * ( 1992 ) , 8198 .
j. przytycki and a. sikora , _ on skein algebras and @xmath2136-character varieties _ , topology * 39 * ( 2000 ) , 115148 .
d. quillen , _ rational homotopy theory _ ,
ann . math . *
90 * ( 1969 ) , 205295 .
d. quillen , _ homotopical algebra _ , lecture notes in math .
* 43 * , springer - verlag , berlin , 1967 .
d. quillen , _ one the ( co-)homology of commutative rings _ , 1970 applications of categorical algebra ( proc .
pure math .
, vol xvii , new york , 1968 ) , 65 - 87 .
soc . , providence , ri .
e. riehl , _ categorical homotopy theory
_ , new mathematical monographs * 24 * , cambridge university press , cambridge , 2014 .
d. rolfsen , _ knots and links _ , mathematics lecture series * 7 * , publish or perish , inc . , berkeley , calif . , 1976 .
j. rotman , _ an introduction to homological algebra _ , 2nd edition , universitext , springer , new york , 2009 .
m. scharlemann , _ heegaard splittings of compact @xmath21-manifolds _ , in _ handbook of geometric topology _ , north holland , 2002 , pp . 921953 .
s. schwede and b. shipley , _ equivalences of monoidal model categories _ , algebr .
geom . topol .
* 3 * ( 2003 ) , 287334 .
a. scorichenko , _ stable @xmath11-theory and functor homology _ ,
thesis , northwestern university , 2000 .
a. s. sikora , _
character varieties of abelian groups _ , math . z. * 277 * ( 2014 ) , 241256 . c. teleman , _ some hodge theory from lie algebras _ in _ motives , polylogarithms and hodge theory_. int .
press lect .
ii , somerville , ma , 2002 , pp .
m. thaddeus , _ mirror symmetry , langlands duality , and commuting elements of lie groups _
, internat .
notices * 22 * ( 2001 ) , 11691193 .
b. ton and g. vezzosi , _ algbres simpliciales @xmath0-quivariantes , thorie de de rham et thormes hkr multiplicatifs .
_ compos . math .
* 147 * ( 2011 ) , no . 6 ,
19792000 . v. turchin and th .
willwacher , _ hochschild - pirashvili homology on suspensions and representations of @xmath2137 _ , preprint ` arxiv:1507.08483 ` .
wai - kit yeung , _ derived representation schemes of simplicial groups _ ,
thesis , in preperation . c. a. weibel , _ an introduction to homological algebra _
, cambridge studies in advanced mathematics * 38 * , cambridge university press , 1995 . | in this paper , we study representation homology of topological spaces , that is a natural homological extension of representation varieties of fundamental groups . we give an elementary construction of representation homology in terms of classical ( abelian ) homological algebra .
our construction is parallel to the loday - pirashvili construction of higher hochschild homology @xcite ; in fact , we establish a direct geometric relation between the two theories by proving that the representation homology of the ( reduced ) suspension of a ( pointed connected ) space is isomorphic to its higher hochschild homology .
we also construct some natural maps and spectral sequences relating representation homology to other standard homology theories associated with spaces ( such as pontryagin algebras , @xmath0-equivariant homology of the free loop space and stable homology of automorphism groups of f.g .
free groups ) .
we compute representation homology explicitly ( in terms of known invariants ) in a number of interesting cases , including spheres , suspensions , complex projective spaces , riemann surfaces and some 3-dimensional manifolds ( such as link complements in @xmath1 and the lens spaces @xmath2 ) .
one of our main results , which we call the comparison theorem , expresses the representation homology of a simply - connected topological space of finite rational type in terms of its quillen and sullivan models . |
in public health studies , clustered or longitudinal binary responses may be collected on a group of individuals where only a subset of these individuals are susceptible to having a positive response .
for example , questionnaires may ask teenagers who are dating to answer a series of questions about dating violence . as in the next generation health study , a larger proportion of all zero responses are observed than would occur by chance ; presumably many individuals who are not dating filled in all zeros on the questionnaire ( also known as `` structural zeros '' ) .
while there may be alternative reasons for structural zeros , for example , participants giving socially desirable responses , we believe this accounts for only a small fraction of zero inflation .
interest is in making inference about the correlated binary responses for those who are susceptible ( i.e. , inference about dating violence among individuals who were dating ) .
there is an extensive literature on zero - inflated poisson and binomial models [ @xcite ; @xcite ] that provide early references , along with more recent work on zero - inflated ordinal data [ @xcite ] and zero - inflated sum score data with randomized responses [ @xcite ] .
@xcite reviewed various statistical models incorporating zero inflation in both discrete and continuous outcomes for cross - sectional data .
@xcite discussed cross - sectional binary regression with zero inflation , and proved the model identifiability when at least one covariate is continuous . @xcite
first considered longitudinal or clustered data with zero - inflated binomial or poisson outcomes .
they incorporated a random effect structure to model the within - subject correlation and proposed an em algorithm to estimate the parameters .
@xcite extended the work of @xcite by proposing a generalized estimation equation ( gee ) approach to model several zero - inflated distributions in a longitudinal setting .
@xcite presented a hurdle model with random effects for repeated measures of zero - inflated count data .
there has been no work , however , on zero - inflated clustered binary data . a component of the next generation health study examines the prevalence and correlates of dating violence among 2787 tenth - grade students , following them over seven years .
dating violence is common among adolescents , may impact adolescent expectations regarding adult intimate relationships [ @xcite ] , and has been found to be associated with increased risk of depression and engagement in high - risk behaviors [ @xcite and @xcite ] .
thus , dating violence among adolescents merits interest from both developmental and public health perspectives [ @xcite ] .
investigators involved in the next study are primarily interested in identifying the risk factors associated with dating violence .
@xcite found a relationship between high - risk behaviors ( i.e. , depressive symptoms , alcohol use , smoking and drug use ) , gender and the prevalence of dating violence victimization .
a total of 10 questions were asked about dating violence .
five of the questions were on dating violence victimization : did your partner ( 1 ) insult you in front of others , ( 2 ) swear at you , ( 3 ) threaten you , ( 4 ) push or shove you , or ( 5 ) throw anything that could hurt you ; the other five were similar questions on perpetration : did you ( 1 ) insult your partner in front of others , ( 2 ) swear at your partner , ( 3 ) threaten your partner , ( 4 ) push or shove your partner , or ( 5 ) throw anything that could hurt your partner ? as illustrated in figure
[ fig1 ] , the distribution of the number of `` yes '' responses is clumped at zero .
when we fit the frequencies with a zero - inflated binomial distribution , the zero - inflation probability is estimated to be about 58% .
the binomial distribution yields a poor fit to the frequencies for two reasons .
first , the prevalence of `` yes '' responses is unequal across different questions ; second , the responses from the same subject are correlated .
but this only serves as an intuitive visualization of zero inflation .
one can argue that the clump of zeros might be due to the high correlation of the binary responses within the same subject ; and , therefore , we also fit the generalized linear mixed model ( glmm ) and plot the fitted frequencies in figure [ fig1 ] .
glmm attempts to fit the spike at 0 , and hence tends to overestimate the within - subject correlation . in this paper , we hope to explore whether zero inflation exists while allowing for the cluster effects .
we propose maximum - likelihood ( ml ) and gee approaches to simultaneously account for the zero inflation and clustering in the multiple binary responses .
the major difference between our work and the previous work is that @xcite , @xcite , and @xcite all considered the zero inflation at the observation level , while in our paper the zero inflation is at the subject level ( meaning that with a structural zero , all the binary responses from a subject are zero ) . for our dating violence example , subjects have all zero responses because they are not susceptible to the condition ( e.g. , in a relationship ) .
the proposed methods are evaluated and compared in simulation studies .
we then reexamine the relationships between high - risk behaviors and dating violence among teenagers using the proposed analysis strategy accounting for zero inflation .
in section [ sec2 ] we present both maximum - likelihood and gee approaches for parameter estimation .
section [ sec3 ] discusses the identifiability of the proposed model and proposes a likelihood ratio test for zero inflation .
simulation study results are presented in section [ sec4 ] .
the next dating violence data is analyzed in section [ sec5 ] , and a discussion follows in section [ sec6 ] .
let @xmath0 be the multivariate binary outcome for subject @xmath1 ( @xmath2 ) , and @xmath3 be the corresponding matrix of covariates .
let @xmath4 be the latent class , so that @xmath5 always takes the value of @xmath6 ( structural zero ) if @xmath7 , and @xmath5 follows a multivariate binary distribution with density @xmath8 if @xmath9 , where @xmath10 is a vector of parameters .
we suppress the subscript @xmath1 when there is no confusion .
let @xmath11 be the prevalence of the latent class 1 . in our example
, @xmath9 indicates that subject @xmath1 is susceptible to the possibility of dating violence ( i.e. , potential of answering the dating violence questions in a positive fashion ) , while @xmath7 indicates that the subject is not susceptible .
if both @xmath12 and @xmath13 are observed , the individual contribution to the full data likelihood is @xmath14 the observed likelihood of @xmath12 is then given by @xmath15 here we assume that the zero - inflation probability @xmath16 is the same across all the subjects in the sample .
this could easily be extended to allow @xmath16 to depend on covariates , for example , with a logistic regression model .
we use a generalized linear mixed effects model ( glmm ) to describe the multivariate distribution , @xmath17 : @xmath18 where @xmath19 , @xmath20 is the vector of random effects following the multivariate normal distribution @xmath21 , @xmath22 is the design matrix of the random effects , and @xmath23 is the known link function .
the parameter vector @xmath10 consists of the parameter of interest @xmath24 and the nuisance parameters in the variance component @xmath25 .
assume @xmath26 s are mutually independent given @xmath27 and @xmath20 , and let @xmath28 be the probability density function of @xmath20 .
then the likelihood for subject @xmath1 becomes @xmath29 the integral with respect to the random effects can be approximated by gaussian
hermite quadrature as @xmath30 where @xmath31 is the @xmath32th quadrature grid point and @xmath33 is the associated weight [ @xcite ] .
the parameter estimation for @xmath16 and @xmath10 can be found by maximizing the log - likelihood for all @xmath34 subjects , @xmath35 .
the variance estimation is calculated from the inverse of the observed information : @xmath36 and can be implemented by the ` optim ` function in ` r ` [ @xcite ] .
likelihood - based inference makes full distributional assumptions on @xmath37 .
when these assumptions are correct , the estimator gains efficiency ; otherwise , classical inference has poor statistical properties .
we explore the estimating equations approach [ @xcite ] that only specifies a structure for the conditional mean @xmath38 .
suppose @xmath39 where @xmath23 is the known link function and @xmath40 is the regression coefficients of interest .
unconditional on @xmath4 , the `` marginal '' mean of @xmath5 is given by @xmath41 the estimating equations can then be written as @xmath42 where @xmath43 and @xmath44 is the working covariance matrix for @xmath5 [ @xcite ] .
we can decompose @xmath44 as @xmath45 with @xmath46 being the diagonal matrix of the variance of @xmath26 [ which is @xmath47 and @xmath48 being the working correlation matrix specified by some nuisance parameter @xmath49 .
if the mean model ( [ eq0-mean ] ) is correct , the estimating equations ( [ eq1-gee1 ] ) are always consistent regardless of the working correlation , and choosing an approximately correct working correlation generally leads to improved efficiency . in the context of zero - inflated regression , we propose two ways to specify the working correlation : marginal and conditional specification .
the marginal correlation directly makes assumptions on @xmath48 , which is similar to the standard gee : the _ marginal independent correlation _ assumes @xmath50 , the @xmath51-dimensional identity matrix ; the _ marginal exchangeable correlation _ assumes that @xmath52 , where @xmath53 is the @xmath54 square matrix of ones .
we refer to these two different approaches as gee - mi and gee - me , respectively .
the conditional correlation exploits the zero - inflated structure and utilizes the conditional covariance , @xmath55 , to derive the unconditional covariance @xmath56 .
a similar idea was first used by @xcite to derive their gee estimator for observation - level zero inflation . by the law of total covariance , for @xmath57 , @xmath58 where @xmath59 is the @xmath60 element of @xmath61 , and @xmath62 is the @xmath63th element of @xmath64 .
the variance of @xmath26 is given by @xmath65 .
conditional independence correlation _ assumes that @xmath66 , so the working correlation is @xmath67 with the @xmath60 element as @xmath68 the _ conditional exchangeable correlation _ assumes that @xmath69 that is , a correlation of @xmath70 between any @xmath26 and @xmath71 _ given _ @xmath72 .
therefore , the @xmath60 element of the working correlation @xmath73 is @xmath74 we refer to these conditional gee approaches as gee - ci and gee - ce , respectively .
similar to the ordinary gee , an unstructured working correlation can be assumed that allows for distinct correlations for each pair of outcomes . with the unstructured gee , the marginal and conditional specification of working correlation
are equivalent , that is , @xmath75 we refer to this approach as gee - un . with each of the five forms of working correlation matrices
, we could solve ( [ eq1-gee1 ] ) using the newton raphson method to obtain the corresponding parameter estimates @xmath76 . with the exchangeable or unstructured correlation structure
, we iteratively update @xmath70 from its moment estimator and @xmath40 from equation ( [ eq1-gee1 ] ) [ @xcite ] . according to the standard theory of gee , the variance of the estimated @xmath76 has the usual sandwich form @xmath77 , where @xmath78 we note that the regression parameters in the glmm and gee are not directly comparable as they have different interpretations .
the former is interpreted as the `` subject - specific effect '' conditional on a subject @xmath1 , while the latter is the `` population - averaged effect '' or `` marginal effect '' [ @xcite ] .
thus , glmm and gee are not compatible for nonidentity link functions . in other words , if the glmm is true , the marginal expectation by integrating out the random effects @xmath20 may not preserve the linear additive form of the covariates .
however , for binary regression with a probit link and random intercept , glmm and gee are compatible .
we adopt a probit random effects model for both the simulations and example analysis .
let @xmath79 and @xmath80 be the c.d.f . and p.d.f .
of the standard normal distribution .
consider the generalized linear mixed effects model with a probit link and a random intercept only , @xmath81 by integrating out @xmath82 , the marginal probability of @xmath26 is computed as follows : @xmath83 while glmm estimates @xmath84 , gee estimates @xmath85 .
the latter is a probit regression model as well , with the regression coefficients , @xmath86 .
this allows us to compare the performance of glmm and gee by comparing the marginal effects of the covariates , which is our interest in the dating violence analysis of the next study .
in general , zero - inflated models are mixtures of two parametric parts , a point mass at zero ( equivalently , a binary distribution with @xmath87 ) and a parametric distribution for the nonstructural zero part .
typically , zero - inflated models are identified by observing a larger number of zeros than would be consistent with the parametric model .
for example , with poisson or binomial outcomes , one can observe excessive proportion of zeros with a histogram .
for a single binary outcome , zero inflation can not be distinguished from rare events , unless covariate dependence is introduced . when there is a continuous covariate @xmath88 , zero inflation is identified because of the linear effect of @xmath88 on the binary response through a known link function .
@xcite proved a weaker sufficient condition for identifiability when covariates are all categorical : to identify a two component mixture of logistic regressions with a binary response , the covariate vector needs to take at least 7 distinct values .
@xcite also used the same argument to prove the identifiability of zero - inflated ordinal regression .
single binary outcome can be seen as a special case of our proposed model with @xmath89 and @xmath90 .
as more information is available with @xmath91 , our model is also identified under follmann and lambert s condition .
@xcite proved the model identifiability for the zero - inflated binary regression with at least one continuous covariate . using a similar technique
, we can prove our model identifiability .
for gee with a probit link , consider @xmath92 and @xmath93 to be two parameter vectors that yield the same conditional mean @xmath94 , that is , @xmath95 equivalently , @xmath96 .
suppose the @xmath97th component of @xmath98 ( i.e. , @xmath99 ) is continuous , then we can take the partial derivative with respect to @xmath99 , which yields @xmath100 taking the partial derivative on both sides of ( [ eq2-iden ] ) with respect to @xmath99 , and with some algebra , it follows that @xmath101 , and hence @xmath102 . from ( [ eq1-iden ] ) , we further get @xmath103 .
this proves the identifiability gee - ci and gee - mi . in gee - un
, the association parameters are indeed obtained from a moment estimator of the correlation between @xmath26 and @xmath71 .
since the mean model is identified , the variance and correlation are also identified . for exchangeable working correlation ,
the association parameter is the `` average '' correlation between all @xmath26 and @xmath71 pairs with @xmath57 , which is identifiable as well .
we now prove the identifiability of the random effects model ( [ eq0-glmm ] ) with a probit link .
with normally distributed random effects , the mean of @xmath26 could be marginalized as @xmath104 , where @xmath25 is the variance covariance matrix of random effects @xmath20 .
we further assume that a continuous covariate is contained in @xmath98 but not in @xmath105 .
then the same argument of ( [ eq2-iden ] ) still applies by denoting @xmath106 as @xmath40 , which proves the identifiability of the regression coefficients up to a scale .
now it suffices to prove the identifiability of @xmath25 .
denote @xmath107 as the correlation coefficient of @xmath26 and @xmath71 ( @xmath57 ) given @xmath72 .
note that from section [ sec2.2 ] , we have @xmath108 since @xmath40 is identifiable , @xmath109 is also identifiable .
therefore , if two parameter vectors @xmath110 and @xmath111 lead to the same @xmath112 and @xmath113 , @xmath107 must be the same . furthermore , the regular glmm is identifiable , suggesting that the correlation structure @xmath107 conditional on @xmath72 is uniquely defined by @xmath25 .
hence , we prove @xmath114 , and , consequently , the identifiability of the ml estimator is established .
we also note that when @xmath90 and no covariates are available , the repeated binary counts could be collapsed into a binomial distribution .
the problem then reduces to the zero - inflated binomial model , which is clearly identifiable . in the presence of the random effects ,
collapsing the binary counts leads to an over - dispersed binomial distribution .
@xcite discussed the zero - inflated beta - binomial model , where the over - dispersion is controlled by a beta distributed random intercept .
our model assumes that the over - dispersion comes from a normal distributed random intercept .
another way to view the proposed model is a mixture of random effect distributions . recall that @xmath26 follows a bernoulli distribution with probability @xmath115 , given by @xmath116 instead of introducing the latent class @xmath4
, we assume that @xmath20 is a mixture of normal distribution and a point mass at @xmath117 : @xmath118 when @xmath119 , the probability @xmath115 is always 0 for @xmath120 , so @xmath5 is the structural zero .
it is easy to show that the likelihood is exactly the same as the proposed model . in practice , one may wish to test for the existence of zero inflation , which can be performed under the parametric model framework .
the likelihood ratio statistic is given by @xmath121 where @xmath122 is the maximized log - likelihood for the zero - inflated model , and @xmath123 is the maximized log - likelihood for the ordinary glmm .
as the null hypothesis ( @xmath87 ) is on the boundary of the parameter space , the asymptotic null distribution of @xmath124 is a mixture of @xmath125 and point mass at 0 , with equal mixture probabilities [ @xcite ] .
theoretically , we could also construct a score test statistic similar to the test proposed by @xcite for zero inflation in a poisson distribution .
however , for our problem , the likelihood function involves intractable integrals , making the score and information matrix both difficult to evaluate .
so in our application , we apply the likelihood ratio test .
motivated by the next study , the data generation for the simulation studies mimics the real example . to evaluate the statistical properties of the above methods , simulation studies of the true model and a misspecified model
were run with two different levels of within - cluster correlation
. a sample size of @xmath126 with a cluster size of @xmath127 questions is considered .
the simulations were repeated 5000 times to compare the performance of the naive estimator ( glmm , where the zero - inflation is ignored ) , the maximum - likelihood ( ml ) estimator and the five gee estimators ( gee - mi , gee - ci , gee - me , gee - ce and gee - un ) .
we calculated the average ( mean ) and standard deviation ( sd ) of the estimated parameters , average of the estimated standard errors ( se ) and 95% ci coverage rates ( cover ) based on the wald intervals to evaluate the robustness and efficiency of the gee and the maximum - likelihood approaches .
twenty gaussian
hermite quadrature points were used for computing the glmm and ml estimators .
we also tried 10 and 40 quadrature points as well as the adaptive quadrature with 250 simulated data sets . in our simulations , the results are very similar for differing number of quadrature points .
our experience for generalized linear mixed models with the logit link function is that gaussian quadrature works very well , and in most situations agq is not needed . in terms of numerical efficiency
, we found that the computation time for agq is about 1020 times longer than the fixed quadrature .
@@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + ml & @xmath128 & 0.700 & 0.700 & 0.014 & 0.014 & 0.949 + & @xmath129 & 0.500 & 0.499 & 0.040 & 0.039 & 0.953 + & @xmath130 & 0.000 & 0.000 & 0.039 & 0.040 & 0.959 + & @xmath131 & 0.894 & 0.895 & 0.027 & 0.027 & 0.949 + & @xmath132 & -0.447 & -0.448 & 0.051 & 0.051 & 0.951 + & @xmath133 & -0.358 & -0.358 & 0.050 & 0.051 & 0.956 + & @xmath134 & 0.179 & 0.179 & 0.051 & 0.050 & 0.949 + & @xmath135 & 0.358 & 0.358 & 0.050 & 0.051 & 0.948 + glmm & @xmath129 & 0.500 & 1.352 & 0.047 & 0.046 & 0.000 + & @xmath130 & 0.000 & -0.444 & 0.029 & 0.030 & 0.000 + & @xmath131 & 0.894 & 0.570 & 0.028 & 0.025 & 0.000 + & @xmath132 & -0.447 & -0.301 & 0.034 & 0.034 & 0.013 + & @xmath133 & -0.358 & -0.239 & 0.033 & 0.034 & 0.060 + & @xmath134 & 0.179 & 0.114 & 0.032 & 0.032 & 0.489 + & @xmath135 & 0.358 & 0.224 & 0.031 & 0.032 & 0.015 + the estimated parameters for ml and glmm methods were marginalized , as we described in section [ sec2.3 ] . in the following sections , we evaluate the performance of the maximum likelihood and gee under a correctly specified and a misspecified model .
additional simulation results are reported in the supplementary material [ @xcite ] , including ( a ) the performance of the proposed model with a smaller sample size ( @xmath136 ) ; ( b ) sensitivity of assuming a constant zero - inflation probability when the probability is affected by covariates ; ( c ) performance of zero - inflated beta - binomial model .
we generated a continuous subject - level covariate @xmath137 from a standard normal distribution and categorical covariate @xmath138 to denote the questions for each subject .
the zero - inflation indicator @xmath4 was generated from @xmath139 with @xmath140 .
the outcome @xmath26 was generated from a probit random effects model : @xmath141 where @xmath142 is the indicator function and @xmath82 is the random intercept following a normal distribution @xmath143 .
we fixed the regression parameters @xmath144 .
the variance component @xmath145 was taken to be @xmath146 and @xmath147 , respectively , to reflect weak ( pearson correlation of about 0.1 ) and strong ( pearson correlation of about 0.45 ) within - cluster correlations .
the simulation results are shown in tables [ tab1 ] and [ tab2 ] , where the true regression parameters are the marginal covariate effects given by @xmath148 .
@@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - mi & @xmath128 & 0.700 & 0.704 & 0.041 & 0.040 & 0.951 + & @xmath130 & 0.000 & -0.002 & 0.096 & 0.096 & 0.951 + & @xmath131 & 0.894 & 0.897 & 0.061 & 0.061 & 0.947 + & @xmath132 & -0.447 & -0.448 & 0.060 & 0.060 & 0.950 + & @xmath133 & -0.358 & -0.358 & 0.056 & 0.057 & 0.952 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.054 & 0.952 + & @xmath135 & 0.358 & 0.358 & 0.059 & 0.059 & 0.946 + gee - ci & @xmath128 & 0.700 & 0.701 & 0.029 & 0.029 & 0.948 + & @xmath130 & 0.000 & 0.001 & 0.069 & 0.070 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.043 & 0.953 + & @xmath132 & -0.447 & -0.449 & 0.055 & 0.056 & 0.953 + & @xmath133 & -0.358 & -0.359 & 0.053 & 0.054 & 0.958 + & @xmath134 & 0.179 & 0.179 & 0.052 & 0.052 & 0.951 + & @xmath135 & 0.358 & 0.360 & 0.056 & 0.056 & 0.946 + gee - me & @xmath128 & 0.700 & 0.701 & 0.032 & 0.032 & 0.953 + & @xmath130 & 0.000 & 0.001 & 0.077 & 0.078 & 0.953 + & @xmath131 & 0.894 & 0.898 & 0.049 & 0.049 & 0.949 + & @xmath132 & -0.447 & -0.449 & 0.058 & 0.058 & 0.955 + & @xmath133 & -0.358 & -0.359 & 0.055 & 0.056 & 0.956 + & @xmath134 & 0.179 & 0.180 & 0.054 & 0.054 & 0.952 + & @xmath135 & 0.358 & 0.359 & 0.057 & 0.058 & 0.950 + gee - ce & @xmath128 & 0.700 & 0.701 & 0.029 & 0.029 & 0.950 + & @xmath130 & 0.000 & 0.001 & 0.069 & 0.070 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.043 & 0.953 + & @xmath132 & -0.447 & -0.449 & 0.055 & 0.056 & 0.954 + & @xmath133 & -0.358 & -0.359 & 0.052 & 0.054 & 0.958 + & @xmath134 & 0.179 & 0.179 & 0.052 & 0.052 & 0.951 + & @xmath135 & 0.358 & 0.360 & 0.056 & 0.056 & 0.946 + gee - un & @xmath128 & 0.700 & 0.702 & 0.036 & 0.036 & 0.952 + & @xmath130 & 0.000 & 0.000 & 0.085 & 0.085 & 0.952 + & @xmath131 & 0.894 & 0.897 & 0.053 & 0.053 & 0.948 + & @xmath132 & -0.447 & -0.448 & 0.058 & 0.059 & 0.954 + & @xmath133 & -0.358 & -0.358 & 0.055 & 0.056 & 0.954 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.054 & 0.952 + & @xmath135 & 0.358 & 0.359 & 0.058 & 0.058 & 0.948 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover
* + ml & @xmath128 & 0.700 & 0.701 & 0.024 & 0.024 & 0.951 + & @xmath129 & 1.500 & 1.503 & 0.088 & 0.089 & 0.955 + & @xmath130 & 0.000 & -0.001 & 0.053 & 0.053 & 0.950 + & @xmath131 & 0.555 & 0.555 & 0.033 & 0.033 & 0.949 + & @xmath132 & -0.277 & -0.278 & 0.037 & 0.038 & 0.952 + & @xmath133 & -0.222 & -0.222 & 0.037 & 0.037 & 0.949 + & @xmath134 & 0.111 & 0.111 & 0.036 & 0.036 & 0.954 + & @xmath135 & 0.222 & 0.222 & 0.037 & 0.037 & 0.951 + glmm & @xmath129 & 1.500 & 2.248 & 0.071 & 0.073 & 0.000 + & @xmath130 & 0.000 & -0.419 & 0.032 & 0.029 & 0.000 + & @xmath131 & 0.555 & 0.384 & 0.029 & 0.026 & 0.000 + & @xmath132 & -0.277 & -0.204 & 0.027 & 0.027 & 0.230 + & @xmath133 & -0.222 & -0.163 & 0.027 & 0.027 & 0.387 + & @xmath134 & 0.111 & 0.079 & 0.025 & 0.026 & 0.763 + & @xmath135 & 0.222 & 0.155 & 0.025 & 0.026 & 0.271 + gee - mi & @xmath128 & 0.700 & 0.712 & 0.083 & 0.088 & 0.949 + & @xmath130 & 0.000 & -0.002 & 0.161 & 0.170 & 0.966 + & @xmath131 & 0.555 & 0.560 & 0.072 & 0.075 & 0.962 + & @xmath132 & -0.277 & -0.279 & 0.050 & 0.050 & 0.949 + & @xmath133 & -0.222 & -0.223 & 0.046 & 0.046 & 0.950 + & @xmath134 & 0.111 & 0.111 & 0.040 & 0.040 & 0.954 + & @xmath135 & 0.222 & 0.223 & 0.047 & 0.049 & 0.953 + gee - ci & @xmath128 & 0.700 & 0.709 & 0.064 & 0.064 & 0.954 + & @xmath130 & 0.000 & -0.006 & 0.121 & 0.123 & 0.969 + & @xmath131 & 0.555 & 0.556 & 0.056 & 0.057 & 0.958 + & @xmath132 & -0.277 & -0.278 & 0.045 & 0.045 & 0.951 + & @xmath133 & -0.222 & -0.222 & 0.042 & 0.042 & 0.946 + & @xmath134 & 0.111 & 0.111 & 0.038 & 0.039 & 0.952 + & @xmath135 & 0.222 & 0.223 & 0.044 & 0.045 & 0.951 + gee - me & @xmath128 & 0.700 & 0.706 & 0.058 & 0.057 & 0.947 + & @xmath130 & 0.000 & -0.001 & 0.115 & 0.114 & 0.954 + & @xmath131 & 0.555 & 0.557 & 0.053 & 0.053 & 0.952 + & @xmath132 & -0.277 & -0.279 & 0.045 & 0.044 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.042 & 0.042 & 0.947 + & @xmath134 & 0.111 & 0.111 & 0.039 & 0.039 & 0.955 + & @xmath135 & 0.222 & 0.223 & 0.043 & 0.044 & 0.953 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - ce & @xmath128 & 0.700 & 0.707 & 0.056 & 0.056 & 0.951 + & @xmath130 & 0.000 & -0.003 & 0.111 & 0.111 & 0.956 + & @xmath131 & 0.555 & 0.556 & 0.052 & 0.051 & 0.951 + & @xmath132 & -0.277 & -0.278 & 0.043 & 0.043 & 0.952 + & @xmath133 & -0.222 & -0.222 & 0.041 & 0.041 & 0.950 + & @xmath134 & 0.111 & 0.111 & 0.038 & 0.038 & 0.955 + & @xmath135 & 0.222 & 0.223 & 0.043 & 0.043 & 0.954 + gee - un & @xmath128 & 0.700 & 0.709 & 0.068 & 0.068 & 0.950 + & @xmath130 & 0.000 & -0.003 & 0.132 & 0.134 & 0.959 + & @xmath131 & 0.555 & 0.558 & 0.060 & 0.060 & 0.952 + & @xmath132 & -0.277 & -0.278 & 0.046 & 0.046 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.044 & 0.043 & 0.948 + & @xmath134 & 0.111 & 0.111 & 0.039 & 0.039 & 0.951 + & @xmath135 & 0.222 & 0.223 & 0.045 & 0.046 & 0.952 + both the ml and the five gee methods perform reasonably well , in terms of small bias and good ci coverage rate .
glmm is seriously biased with poor ci coverage .
it can be seen that the ml method is the most efficient , as it makes use of the full distributional assumption of the observed data . on the contrary ,
gee only relies on the first moments of the outcome . in estimating @xmath16 , the zero - inflated probability , the ses of the gee approaches are more than twice as large as the se of the ml method .
the ses for other parameters are also significantly smaller for the ml method . of the five gee methods
, we found that gee - ce is the most efficient with the smallest se , while gee - mi is the least efficient . by exploiting the correlation structure induced by the zero - inflation process ,
the conditional independence and exchangeable working correlation both gain a substantial amount of efficiency , compared to their marginal counterparts .
this result is consistent with the simulation results in @xcite .
the ses for gee - ce and gee - ci are quite close , implying that adding working dependence to the outcome given @xmath9 would not help much as long as the dependence due to zero inflation is accounted for
. we did observe a bigger improvement of gee - ce versus gee - ci for the strong correlation case .
but the improvement of gee - ci versus gee - mi is even larger .
therefore , we recommend that it is more important to make use of the zero - inflation structure in the gee estimators . although gee - un has the most flexible form of working correlation , it is not as efficient as gee - ci or gee - ce , probably due to estimating a larger number of nuisance parameters .
we found that gees may occasionally not have a solution or have a boundary solution ( @xmath149 ) in about 12% of the simulations with @xmath150 .
our experience is that nonconvergence or boundary solutions occur more often when the covariate effects are weaker , the within - cluster correlation is stronger , the true zero - inflation probability is closer to 1 , or the model is more severely misspecified .
we consider a misspecified model where only the first three questions are correlated and the last two are independent .
the data generation for @xmath26 with @xmath151 was the same as in section 3.1 , but @xmath26 for @xmath152 was generated as follows : @xmath153 the random intercept @xmath82 was not added to the last two questions , but a factor of @xmath154 was divided to the coefficients to keep the marginalized regression coefficients the same . in this case , the ml estimator is from a misspecified model since the random intercept model imposes correlation among all the questions .
for gee , only the working correlation is misspecified , while the first moment of @xmath26 is still correct .
the simulation results are presented in tables [ tab3 ] and [ tab4 ] .
@@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + ml & @xmath128 & 0.700 & 0.706 & 0.013 & 0.013 & 0.925 + & @xmath129 & 0.500 & 0.273 & 0.047 & 0.047 & 0.000 + & @xmath130 & 0.000 & -0.008 & 0.038 & 0.039 & 0.953 + & @xmath131 & 0.894 & 0.899 & 0.024 & 0.024 & 0.948 + & @xmath132 & -0.447 & -0.446 & 0.051 & 0.053 & 0.963 + & @xmath133 & -0.358 & -0.357 & 0.049 & 0.053 & 0.964 + & @xmath134 & 0.179 & 0.177 & 0.053 & 0.052 & 0.949 + & @xmath135 & 0.358 & 0.355 & 0.053 & 0.053 & 0.949 + glmm & @xmath129 & 0.500 & 1.176 & 0.042 & 0.040 & 0.000 + & @xmath130 & 0.000 & -0.441 & 0.029 & 0.030 & 0.000 + & @xmath131 & 0.894 & 0.570 & 0.028 & 0.024 & 0.000 + & @xmath132 & -0.447 & -0.303 & 0.035 & 0.036 & 0.019 + & @xmath133 & -0.358 & -0.241 & 0.034 & 0.035 & 0.080 + & @xmath134 & 0.179 & 0.114 & 0.034 & 0.034 & 0.519 + & @xmath135 & 0.358 & 0.223 & 0.033 & 0.034 & 0.021 + gee - mi & @xmath128 & 0.700 & 0.704 & 0.041 & 0.040 & 0.951 + & @xmath130 & 0.000 & -0.002 & 0.095 & 0.095 & 0.951 + & @xmath131 & 0.894 & 0.896 & 0.059 & 0.060 & 0.949 + & @xmath132 & -0.447 & -0.447 & 0.059 & 0.060 & 0.953 + & @xmath133 & -0.358 & -0.357 & 0.056 & 0.057 & 0.953 + & @xmath134 & 0.179 & 0.178 & 0.057 & 0.057 & 0.950 + & @xmath135 & 0.358 & 0.358 & 0.061 & 0.062 & 0.948 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - ci & @xmath128 & 0.700 & 0.701 & 0.030 & 0.030 & 0.950 + & @xmath130 & 0.000 & 0.000 & 0.070 & 0.071 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.044 & 0.951 + & @xmath132 & -0.447 & -0.449 & 0.055 & 0.056 & 0.956 + & @xmath133 & -0.358 & -0.358 & 0.053 & 0.054 & 0.957 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.055 & 0.949 + & @xmath135 & 0.358 & 0.360 & 0.059 & 0.059 & 0.951 + gee - me & @xmath128 & 0.700 & 0.702 & 0.034 & 0.033 & 0.951 + & @xmath130 & 0.000 & 0.000 & 0.079 & 0.080 & 0.952 + & @xmath131 & 0.894 & 0.897 & 0.049 & 0.050 & 0.949 + & @xmath132 & -0.447 & -0.448 & 0.058 & 0.058 & 0.955 + & @xmath133 & -0.358 & -0.358 & 0.055 & 0.056 & 0.954 + & @xmath134 & 0.179 & 0.179 & 0.057 & 0.057 & 0.952 + & @xmath135 & 0.358 & 0.359 & 0.061 & 0.061 & 0.951 + gee - ce & @xmath128 & 0.700 & 0.701 & 0.030 & 0.030 & 0.951 + & @xmath130 & 0.000 & 0.000 & 0.070 & 0.071 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.044 & 0.950 + & @xmath132 & -0.447 & -0.448 & 0.055 & 0.056 & 0.955 + & @xmath133 & -0.358 & -0.358 & 0.053 & 0.054 & 0.956 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.055 & 0.950 + & @xmath135 & 0.358 & 0.359 & 0.059 & 0.059 & 0.952 + gee - un & @xmath128 & 0.700 & 0.703 & 0.037 & 0.036 & 0.952 + & @xmath130 & 0.000 & -0.001 & 0.086 & 0.087 & 0.951 + & @xmath131 & 0.894 & 0.897 & 0.053 & 0.054 & 0.948 + & @xmath132 & -0.447 & -0.448 & 0.058 & 0.059 & 0.953 + & @xmath133 & -0.358 & -0.358 & 0.055 & 0.056 & 0.954 + & @xmath134 & 0.179 & 0.179 & 0.057 & 0.057 & 0.950 + & @xmath135 & 0.358 & 0.359 & 0.061 & 0.062 & 0.950 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + ml & @xmath128 & 0.700 & 0.730 & 0.015 & 0.015 & 0.472 + & @xmath129 & 1.500 & 0.624 & 0.037 & 0.038 & 0.000 + & @xmath130 & 0.000 & -0.046 & 0.039 & 0.039 & 0.776 + & @xmath131 & 0.555 & 0.561 & 0.024 & 0.024 & 0.941 + & @xmath132 & -0.277 & -0.273 & 0.036 & 0.046 & 0.986 + & @xmath133 & -0.222 & -0.218 & 0.036 & 0.045 & 0.987 + & @xmath134 & 0.111 & 0.103 & 0.047 & 0.044 & 0.933 + & @xmath135 & 0.222 & 0.208 & 0.046 & 0.045 & 0.934 + glmm & @xmath129 & 1.500 & 1.201 & 0.038 & 0.039 & 0.000 + & @xmath130 & 0.000 & -0.416 & 0.029 & 0.030 & 0.000 + & @xmath131 & 0.555 & 0.378 & 0.024 & 0.022 & 0.000 + & @xmath132 & -0.277 & -0.206 & 0.027 & 0.034 & 0.425 + & @xmath133 & -0.222 & -0.164 & 0.027 & 0.034 & 0.613 + & @xmath134 & 0.111 & 0.079 & 0.034 & 0.033 & 0.816 + & @xmath135 & 0.222 & 0.153 & 0.033 & 0.032 & 0.432 + gee - mi & @xmath128 & 0.700 & 0.712 & 0.080 & 0.082 & 0.948 + & @xmath130 & 0.000 & -0.005 & 0.153 & 0.159 & 0.963 + & @xmath131 & 0.555 & 0.558 & 0.066 & 0.067 & 0.954 + & @xmath132 & -0.277 & -0.278 & 0.048 & 0.049 & 0.946 + & @xmath133 & -0.222 & -0.222 & 0.045 & 0.045 & 0.949 + & @xmath134 & 0.111 & 0.111 & 0.054 & 0.054 & 0.951 + & @xmath135 & 0.222 & 0.223 & 0.058 & 0.06 & 0.954 + gee - ci & @xmath128 & 0.700 & 0.707 & 0.066 & 0.066 & 0.944 + & @xmath130 & 0.000 & -0.001 & 0.129 & 0.130 & 0.966 + & @xmath131 & 0.555 & 0.558 & 0.058 & 0.058 & 0.960 + & @xmath132 & -0.277 & -0.279 & 0.046 & 0.046 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.043 & 0.043 & 0.950 + & @xmath134 & 0.111 & 0.112 & 0.053 & 0.052 & 0.950 + & @xmath135 & 0.222 & 0.224 & 0.057 & 0.058 & 0.951 + gee - me & @xmath128 & 0.700 & 0.709 & 0.069 & 0.068 & 0.948 + & @xmath130 & 0.000 & -0.004 & 0.132 & 0.133 & 0.951 + & @xmath131 & 0.555 & 0.557 & 0.060 & 0.060 & 0.952 + & @xmath132 & -0.277 & -0.278 & 0.046 & 0.046 & 0.946 + & @xmath133 & -0.222 & -0.222 & 0.043 & 0.043 & 0.946 + & @xmath134 & 0.111 & 0.112 & 0.054 & 0.054 & 0.950 + & @xmath135 & 0.222 & 0.224 & 0.058 & 0.059 & 0.953 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - ce & @xmath128 & 0.700 & 0.709 & 0.067 & 0.066 & 0.946 + & @xmath130 & 0.000 & -0.004 & 0.129 & 0.130 & 0.959 + & @xmath131 & 0.555 & 0.556 & 0.058 & 0.058 & 0.954 + & @xmath132 & -0.277 & -0.279 & 0.045 & 0.046 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.043 & 0.043 & 0.948 + & @xmath134 & 0.111 & 0.111 & 0.053 & 0.052 & 0.948 + & @xmath135 & 0.222 & 0.224 & 0.057 & 0.058 & 0.951 + gee - un & @xmath128 & 0.700 & 0.711 & 0.073 & 0.073 & 0.951 + & @xmath130 & 0.000 & -0.005 & 0.140 & 0.142 & 0.954 + & @xmath131 & 0.555 & 0.557 & 0.061 & 0.061 & 0.952 + & @xmath132 & -0.277 & -0.278 & 0.047 & 0.047 & 0.947 + & @xmath133 & -0.222 & -0.222 & 0.044 & 0.044 & 0.948 + & @xmath134 & 0.111 & 0.111 & 0.053 & 0.053 & 0.950 + & @xmath135 & 0.222 & 0.223 & 0.058 & 0.059 & 0.953 + with @xmath155 , the ml approach is almost unbiased for estimating @xmath16 as well as the regression coefficients . when @xmath129 increases to 1.5 , the ml estimator becomes slightly biased with poor ci coverage , especially for @xmath16 and @xmath130 .
the estimation of other parameters appears to be robust to the model misspecification , except that the ses for @xmath132 and @xmath133 overestimate the true variability . on the other hand ,
the five gee methods all perform quite well , in terms of little bias and close - to - nominal coverage rates .
similar to the previous simulation study , we observed that about 1% of the gee simulations did not converge for @xmath150 .
although the maximum - likelihood approach is biased , its standard error is much smaller than the gee approaches .
for example , the ml estimator for @xmath16 in table [ tab4 ] has a se only a quarter as large as that of the gee - ci and gee - ce estimators . as a result of the variance - bias trade - off
, the mean squared error for the ml estimator is still smaller than gee .
if the interest is in estimation , one can still argue that the ml performs better ; but if the interest is in hypothesis testing , gee methods are preferred , as they are more robust and preserve the correct type i error rate . from the above two sets of simulation studies , we would generally recommend the ml estimator in practice because of its high efficiency .
the correlation structure of the outcome is critical in identifying the zero - inflation process .
therefore , a full parametric assumption for the correlation can lead to good efficiency in the estimation .
however , if this parametric assumption does not hold , the ml estimator could have poor ci coverage rates . in order to perform hypothesis testing
, we would prefer the gee approaches , which only rely on the correct mean model and are not sensitive to the working correlation assumption . among the five gee approaches ,
the gee - ci and gee - ce are the most favorable , because they are more efficient by exploiting the dependence structure induced by the zero - inflation process . in practice ,
the gee - ci and gee - ce estimators may be computed in conjunction with the ml estimator as a sensitivity analysis .
in this section we fit our proposed ml and gee models to the dating violence example , together with the naive glmm model .
a total of 2787 students were enrolled in the study , among which 664 left all the dating violence questions blank .
these 664 subjects are either not in a relationship , and thus skipped these questions , or they did not respond at all to the whole survey . in the remaining 2123 subjects ,
39 were excluded because they only answered part of the dating violence questions , and 61 were excluded because they have missing data in the covariates .
the final analysis sample was @xmath156 .
the clustered outcomes of interest ( @xmath26 ) are the ten questions of dating violence , including five victimization and five perpetration questions .
we can see from figure [ fig1 ] that the frequency histogram of `` yes '' responses shows a huge spike at 0 .
it seems likely that some students who answered all the questions with `` no '' were not in a relationship , that is , a zero inflation of the outcome .
define the latent variable @xmath9 if the subject is in a relationship and 0 otherwise .
we included gender ( @xmath157 ) , depressive symptoms ( @xmath158 ) , family relationship ( @xmath159 ) and family influence ( @xmath160 ) as the predictors of @xmath26 given @xmath9 .
the ds score comes from the questionnaire of depressive symptoms and is on the continuous scale ranging from 1 to 5 , with the larger score indicating worse depressive symptoms .
the fr ( ranging from 0 to 10 ) measures the participant s satisfaction with the relationship in his / her family , with 10 being a very good relationship .
the fi score ( ranging from 1 to 7 ) is the family influence on the participants not verbally or physically abusing their romantic partner , with a higher score being greater influence .
we adjust for question number as a factor and question type ( victimization vs. perpetration ) , in order to account for different prevalence of yes responses .
the interactions between question type and other covariates ( @xmath161 , @xmath162 , @xmath163 and @xmath164 ) are also included .
the summary statistics of these variables are described in table [ tab5 ] . @@ld2.6@ * variable * & + gender : female & 56.6% + ds score & 2.0 ( 1.0 ) + family relationship & 7.4 ( 2.3 ) + family influence & 5.7 ( 1.8 ) + question 1v insult you & 18.5% + question 1p insult your boyfriend / girlfriend & 16.8% + question 2v swear at you & 31.3% + question 2p swear at your boyfriend / girlfriend & 26.1% + question 3v threaten you & 7.2% + question 3p threaten your boyfriend / girlfriend & 5.6% + question 4v push you & 13.5% + question 4p push your boyfriend / girlfriend & 9.9% + question 5v throw object at you & 4.5% + question 5p throw object at your boyfriend / girlfriend & 3.8% + denote @xmath27 to be the design matrix including all the covariates and interaction terms mentioned above .
we fit the probit random effects model @xmath165 where @xmath166 is the random intercept .
this model has the same marginal mean as the marginal probit regression model : @xmath167 as @xmath168 for @xmath169 . for comparative purposes , we report the marginal regression coefficients @xmath40 for all the analyses .
= @@ld2.10d2.10d2.10d2.10d2.10d2.10d2.10@ * parameter * & & & & & & & + ( intercept ) & -0.192 ( 0.176 ) & -0.746 ( 0.177 ) & -0.055 ( 0.254 ) & -0.159 ( 0.225 ) & -0.497 ( 0.222 ) & -0.341 ( 0.221 ) & -0.201 ( 0.232 ) + question 2 & 0.528 ( 0.037 ) & 0.374 ( 0.024 ) & 0.578 ( 0.071 ) & 0.508 ( 0.057 ) & 0.475 ( 0.058 ) & 0.488 ( 0.057 ) & 0.529 ( 0.061 ) + question 3 & -0.754 ( 0.043 ) & -0.577 ( 0.031 ) & -0.814 ( 0.069 ) & -0.750 ( 0.061 ) & -0.716 ( 0.063 ) & -0.729 ( 0.061 ) & -0.768 ( 0.063 ) + question 4 & -0.332 ( 0.035 ) & -0.250 ( 0.026 ) & -0.360 ( 0.053 ) & -0.335 ( 0.048 ) & -0.315 ( 0.047 ) & -0.324 ( 0.047 ) & -0.339 ( 0.049 ) + question 5 & -1.003 ( 0.050 ) & -0.773 ( 0.036 ) & -1.069 ( 0.083 ) & -0.996 ( 0.073 ) & -0.949 ( 0.077 ) & -0.969 ( 0.074 ) & -1.013 ( 0.075 ) + question type & 0.222 ( 0.122 ) & 0.158 ( 0.089 ) & 0.264 ( 0.173 ) & 0.234 ( 0.150 ) & 0.215 ( 0.146 ) & 0.223 ( 0.145 ) & 0.215 ( 0.155 ) + ds score & 0.192 ( 0.036 ) & 0.178 ( 0.035 ) & 0.250 ( 0.043 ) & 0.220 ( 0.038 ) & 0.226 ( 0.037 ) & 0.219 ( 0.037 ) & 0.230 ( 0.039 ) + gender & 0.132 ( 0.067 ) & 0.153 ( 0.054 ) & 0.171 ( 0.076 ) & 0.118 ( 0.068 ) & 0.186 ( 0.067 ) & 0.160 ( 0.066 ) & 0.164 ( 0.070 ) + family relationship & -0.037 ( 0.014 ) & -0.035 ( 0.012 ) & -0.042 ( 0.019 ) & -0.039 ( 0.017 ) & -0.041 ( 0.015 ) & -0.041 ( 0.016 ) & -0.041 ( 0.017 ) + family influence & -0.114 ( 0.017 ) & -0.087 ( 0.015 ) & -0.137 ( 0.021 ) & -0.135 ( 0.018 ) & -0.095 ( 0.017 ) & -0.112 ( 0.017 ) & -0.123 ( 0.018 ) + question type@xmath170ds score & 0.037 ( 0.025 ) & 0.026 ( 0.018 ) & 0.034 ( 0.032 ) & 0.037 ( 0.029 ) & 0.028 ( 0.026 ) & 0.034 ( 0.028 ) & 0.035 ( 0.029 ) + question type@xmath170gender & -0.409 ( 0.051 ) & -0.300 ( 0.037 ) & -0.446 ( 0.066 ) & -0.392 ( 0.057 ) & -0.373 ( 0.056 ) & -0.379 ( 0.056 ) & -0.399 ( 0.059 ) + question type@xmath170family & -0.003 ( 0.010 ) & -0.002 ( 0.008 ) & -0.009 ( 0.014 ) & -0.007 ( 0.012 ) & -0.006 ( 0.012 ) & -0.006 ( 0.012 ) & -0.005 ( 0.013 ) + relationship + question type@xmath170family & 0.025 ( 0.012 ) & 0.019 ( 0.009 ) & 0.031 ( 0.015 ) & 0.025 ( 0.013 ) & 0.026 ( 0.013 ) & 0.024 ( 0.013 ) & 0.028 ( 0.013 ) + influence + @xmath171 & 0.571 ( 0.031 ) & 1 & 0.523 ( 0.062 ) & 0.626 ( 0.075 ) & 0.690 ( 0.104 ) & 0.663 ( 0.089 ) & 0.587 ( 0.070 ) + @xmath129 & 1.033 ( 0.065 ) & 1.627 ( 0.055 ) & & & & & + the results of the ml , glmm and gee estimations are listed in table [ tab6 ] . from the ml estimation
, we can see that all four subject - level covariates are significant : the probability of dating violence perpetration was higher for females , those who are more depressed , those who have a worse relationship in their family , and those who are less influenced by their guardians .
the interaction terms between question type and gender , and question type and family influence were both significant , suggesting ( a ) boys are more likely to be the victims of dating violence , and ( b ) the family influence has a slightly higher impact on dating violence perpetration than victimization .
the finding regarding greater male dating violence victimization in this age is in line with previous studies [ @xcite ; @xcite ] .
the impacts of depression score and family relationship are similar regardless of question type .
the directions of association are expected and consistent with some of the findings in @xcite .
the zero - inflation probability is estimated to be 0.571 , that is , we expect about 43% of the sample ( about 860 subjects ) to be structural zeros .
we suspect that a majority of them were not in a relationship , but answered all the dating violence questions with `` no .
'' however , there may be alternative reasons for the zero inflation , for example , some kids may give socially desirable answers in the survey and hence underreport dating violence .
however , we believe that this only accounts for a small fraction of the structural zeros .
the likelihood ratio test statistic for zero inflation is @xmath172 ( @[email protected] ) .
the parameter estimations by gee are generally close to ml , but the standard errors are larger . the naive glmm method estimated smaller covariate effects , which could be biased due to ignoring the zero - inflated nature of the data .
as pointed out by a referee , the zero - inflation problem could be avoided by including a filter question of asking whether the subject had a relationship or not . the filter question was not included because the study investigators felt that it was an unreliable question to ask .
relationships between teenagers today can not easily be characterized , and the investigators felt that explicitly asking this question may limit important responses about violence [ @xcite ] .
there are other cases where the susceptible population can not be ascertained accurately .
for example , in drug abuse studies , specifically asking whether individuals are drug abusers might not be a question that would result in a reliable response .
but we may be interested in whether the abusers seek particular types of treatment .
additionally , this approach may be relevant to questions regarding immigration status , where there may be legal implications ( perceived or real ) in answering , and in questions on mental health status , where the respondent may have a reduced ability to accurately report their status .
in public health research , excessive zero responses may occur if the population which is susceptible to respond is not carefully screened or is unknown .
the resulting zero inflation may have an effect on the results obtained by conventional methods of analysis . in the next generation health study , investigators were interested in identifying predictors of dating violence in teenagers . examining these regression relationships are of interest for those individuals who are in a relationship ( i.e. , the susceptible condition ) .
many more individuals completed this study component than investigators thought would be in a relationship at this age .
this led to what appears to be zero - inflated clustered binary data .
we developed both ml and gee approaches for analyzing such data . through simulations and analysis of the real data example
, we found that the ml approach is substantially more efficient than the gee approaches .
however , under moderate model misspecification , the ml approach may result in biased inference .
it is recommended that , as a sensitivity analysis , both ml and gee approaches be applied in applications . in the gee approach ,
we treat the regression parameters and nuisance working correlation as orthogonal , that is , a gee1 approach with the parameters in the working correlation estimated by a moment estimator .
potential efficiency gain could be achieved using an improved version of gee1 [ @xcite ] or gee2 [ prentice and zhao ( @xcite ) ; liang , zeger and qaqish ( @xcite ) ] , by establishing another set of estimating equations on the second moment .
however , gee2 requires a correct variance structure with working third and fourth moment model , which is hard to verify with the presence of zero inflation , and results in bias under second , third and fourth moment misspecification .
the previous work of @xcite adopted prentice and zhao s gee2 that maintains the parameter orthogonality in their second moment estimating equations , and they argued that only making a first moment assumption may lead to parameter nonidentifiability .
this is true in their case , where the zero - inflation probability is at the observation level , so @xmath174 and @xmath62 might be confounded in @xmath175 .
however , in the case of subject - level zero inflation , we have shown the model identifiability of the gee1 estimators . in principle , zero - inflated models can not be identified nonparametrically ; parametric assumptions for the nonzero part play a fundamental role in model estimation .
for example , zero inflation in poisson and binomial data can be determined by the lack of fit in the zero cells of these respective distributions . in this paper , we assume that the nonzero distribution is given by a generalized linear mixed model with normal random effects .
the ml approach exploits the correlation structure in order to distinguish structural zeros and random zeros .
intrinsically , gee only uses the mean structure of the binary data in order to estimate regression parameters and , unlike ml , does not use the entire distribution for estimation . in our application , there were very few missing data
. however , in many studies with sensitive psychological or behavioral questions , there may be informative missingness .
an advantage of the ml approach is that it can more easily be extended to account for informative missing responses [ see @xcite , e.g. ] .
the proposed methodology implicitly assumes that the subjects answer the questions truthfully .
if this assumption does not hold , the parameter estimation is likely to be biased .
we could formulate a likelihood approach if we had good prior information about the distribution of false negative occurrence across the questions .
this would require a verification subsample corresponding to the survey questions ( maybe obtained through interviews of parents and friends ) on a fraction of teenagers . the zero - inflation probability in our model is assumed to be constant , but it is straightforward to include covariates in both ml and gee approaches . in addition , our focus is on a cross - sectional inference , that is , analyzing the dating violence data at one point in time ( 11th grade ) .
understanding behavior change from adolescence over time is interesting but also challenging . in the longitudinal
setting , the zero - inflation probability is time - dependent that can probably be modeled by a latent markov process .
we will leave it for future exploration .
we greatly appreciate and enjoyed reading the very insightful comments and suggestions by the associate editor and two anonymous referees .
their comments not only remarkably improved the quality of the paper , but also motivated our deeper thoughts of possible extensions of the methodology in the future . | the next generation health study investigates the dating violence of adolescents using a survey questionnaire .
each student is asked to affirm or deny multiple instances of violence in his / her dating relationship .
there is , however , evidence suggesting that students not in a relationship responded to the survey , resulting in excessive zeros in the responses .
this paper proposes likelihood - based and estimating equation approaches to analyze the zero - inflated clustered binary response data .
we adopt a mixed model method to account for the cluster effect , and the model parameters are estimated using a maximum - likelihood ( ml ) approach that requires a gaussian hermite quadrature ( ghq ) approximation for implementation .
since an incorrect assumption on the random effects distribution may bias the results , we construct generalized estimating equations ( gee ) that do not require the correct specification of within - cluster correlation . in a series of simulation studies ,
we examine the performance of ml and gee methods in terms of their bias , efficiency and robustness .
we illustrate the importance of properly accounting for this zero inflation by reanalyzing the next data where this issue has previously been ignored . , , |
random heterogeneous multiphase materials or media are ubiquitous .
examples include composites , porous media , biological materials as well as cosmological structures , and their macroscopic properties are of great interest @xcite . in the first part of this series of two papers @xcite ( henceforth referred to as paper i ) , we proposed a theoretical formalism to model and categorize heterogeneous materials via two - point correlation functions @xmath1 , which can be interpreted as the probability of finding two points separated by the the dsiplacement vector @xmath2 in one of the phases @xcite .
in particular , we introduced the idea of the two - point correlation function space and its basis functions .
in general , @xmath0 of a medium can be expressed by a map @xmath3 on the associated basis functions , which is composed of convex - combination and product operations .
we also suggested a set of basis functions by examining certain known realizable analytical two - point correlation functions .
moreover , we introduced an efficient isotropy - preserving @xmath0-sampling algorithm , namely the _ lattice - point _
algorithm , but left the details of the algorithm for another paper . here
we will provide algorithmic details of the lattice - point methodology and consider several nontrivial applications to illustrate the practical utility of our theoretical formalism .
in particular , we will apply the lattice - point algorithm to generate three - dimensional ( 3d ) digitized realizations of a fontainebleau sandstone and a boron carbide / aluminum composite from two - dimensional ( 2d ) tomographic images of their slices through the materials . to justify whether the reconstructions are successful , one should also measure other statistical descriptors of the media @xcite . here
we compute the two - point cluster function @xmath4 @xcite ( see definition and discussions in sec .
iv ) , which contains nontrivial `` connectedness '' information of the phases of interest . by comparing @xmath4 of the target and reconstructed media ,
we demonstrate that @xmath0 is indeed sufficient to capture the salient structural features in these cases .
we also study the reconstruction of a binary laser - speckle pattern in two dimensions , and find that the algorithm fails to reproduce the target pattern accurately .
we conclude that in general reconstructions using @xmath0 only work well for heterogeneous materials with single - scale structures ( those having only one characteristic length scale ) .
however , two - point information via @xmath0 is not sufficient to accurately model multi - scale media ( those composed of structural elements associated with multiple characteristic length scales ) .
moreover , we will show how the proper convex combinations of basis functions enable one to obtain two - point correlation functions with desired properties and thus enable one to generate a variety of structures with controllable morphological features .
note that here we will mainly focus on modeling heterogeneous materials in three dimensions , which complements the two - dimensional examples that we considered in paper i. yeong and torquato formulated the ( re)construction problem as an energy - minimization problem using simulated annealing @xcite .
this has become a very popular ( re)construction technique @xcite . in this method , a nonnegative objective function @xmath5 , called the energy , " is defined as the sum of squared differences between the target and sampled correlation functions . in our case , the energy is given by @xmath6 ^ 2,\ ] ] where @xmath7 and @xmath8 is the target and sampled two - point correlation function , respectively .
an important issue in the ( re)constructions based on the method of simulated annealing is the choice of the _ energy threshold _ @xmath9 , i.e. , the error tolerance of discrepancies between the correlation functions of generated medium and the imposed ones . when the energy of the ( re)constructed medium is below @xmath9 , the ( re)construction process is terminated .
the energy threshold is a key indicator of how accurately the medium is ( re)constructed . in our previous work , @xmath9 was always chosen to be a very small number , e.g. , @xmath10 ; however , no quantitative analysis on why such a value should be chosen was given . in this paper
, we will discuss in detail the significance of the energy threshold for both the orthogonal @xmath0-sampling algorithm ( only sampling @xmath0 along convenient orthogonal directions ) @xcite and the lattice - point algorithm . in the lattice - point procedure , we consider the digitized medium ( pixel system ) to be a lattice - gas " system , in which the black pixels behave like nonoverlapping `` particles '' moving from one lattice site to another .
the two - point correlation function of the medium is obtained by binning all the distances between the black pixels and dividing the number in each bin by the total number of distances between all lattice sites within that bin . to generate a trial configuration , a randomly selected `` particle '' ( black pixel )
is given a random displacement subjected to the nonoverlapping constraint .
only the distances between the moved `` particle '' and all the others need to be recomputed in order to obtain @xmath0 of the trial configuration . in this way
, all directions in the digitized medium are effectively sampled ; moreover , the complexity of the algorithm is linear in @xmath11 , i.e. , the number of nonoverlapping particles . for heterogeneous materials containing well - defined inclusions
, we will show that the computational speed of the ( re)construction process can be increased by an algorithm modification using surface optimization , which is essentially a biased pixel - selection procedure .
this algorithm modification is based on the fact that in the later stages of ( re)constructions , further refinements of the configurations are achieved by the moves of black pixels on the surfaces of formed pixel - clusters .
thus , only the `` surface pixels '' are selected and given random moves to generate trial configurations , i.e. , the surfaces are optimized .
the physical analog of this process is solidification , in the later stage of which the re - arrangements of solutes only occur on the surfaces of nucleated particles .
we will see in the following that one can obtain a much lower error ( discrepancies between the correlation functions of the ( re)constructed medium and the imposed ones ) when surface optimization is properly applied .
this improvement enables one to quantitatively study the non - uniqueness problem of the ( re)constructions @xcite .
the rest of the paper is organized as follows : in sec .
ii , we describe the lattice - point algorithm and the algorithm modification using surface optimization in great detail . the choice of different pixel - lattices is also discussed . in sec .
iii , we discuss the significance of the energy threshold for both the orthogonal @xmath0-sampling method and the lattice - point algorithm . in sec .
iv , we suggest a general form of the map @xmath3 and apply the theoretical formalism to model a fontainebleau sandstone , a boron carbide / aluminum composite and a binary laser - speckle pattern .
we also show how to construct materials with structural properties of interest by manipulating the parameters in the basis functions and choosing proper combination coefficients . in sec .
v , we make concluding remarks .
in paper i , we derived the exact algebraic equations for the ( re)construction problem and showed that the equations have an infinite number of solutions and can not be solved rigorously . in principle
, the yeong - torquato scheme enables one to obtain one of the solutions efficiently .
the most time - consuming steps of the scheme are the samplings of the two - point correlation function at every trial configuration .
an efficient @xmath0-sampling method would dramatically speed up the ( re)construction process .
furthermore , an isotropy - preserving algorithm is required in ( re)constructions whenever a radial two - point correlation function @xmath8 ( @xmath12 ) is employed for the case of statistically homogeneous and isotropic media .
the lattice - point algorithm is designed to sample the digitized representation of a statistically homogeneous and isotropic medium in all possible directions efficiently . for simplicity
, we will illustrate the idea in 2d .
implementation of the algorithm in three dimensions is a straightforward extension . instead of considering the digitized medium as a collection of black and white pixels
, one can think of the medium as a lattice - gas system : the black pixels are the `` gas molecules '' and the white pixels are unoccupied lattice sites , as shown in fig .
the `` gas molecules '' are free to move from one lattice site to another , subject to the impenetrability condition , i.e. , each lattice site can only be occupied by one `` gas molecule '' .
thus , the black pixels behave like hard particles and the volume fraction of black phase is conserved during the evolution of the system
. @xmath13{fig1a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig1b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] for a statistically homogeneous and isotropic particle system ( e.g. , an equilibrium hard - sphere system or an equilibrium lattice - gas system in @xmath14-dimensional euclidean space @xmath15 ) , the most basic statistical descriptor of the spatial correlations of the particles is the pair correlation function @xmath16 @xcite .
the quantity @xmath17 is proportional to the conditional probability of finding the center of a particle in a @xmath14-dimensional spherical shell of volume @xmath18 , given that there is another particle at the origin . here
@xmath19 is the number density of the system and @xmath20 is the surface area of a @xmath14-dimensional sphere of radius @xmath21 , which is given by @xmath22 where @xmath23 is the euler - gamma function .
hence , for a finite system , integrating @xmath24 over the volume yields @xmath25 , i.e. , all the particles except the one at the origin .
alternatively , @xmath17 is the average number of particles at a radial distance between @xmath21 and @xmath26 from a reference particle .
practically , @xmath16 can be obtained from simulations by generating a histogram for the number of particles @xmath27 contained in a concentric shell of finite thickness ( `` bin '' width ) @xmath28 at radial distance @xmath21 from a arbitrarily chosen reference particle .
the two - point correlation function @xmath8 of a statistically homogeneous and isotropic medium can be interpreted as the probability that both of ends of a randomly oriented line segment with length @xmath21 fall into the phase of interest , say , the `` black '' phase . by comparison
, we can see that the method of computing @xmath16 of an isotropic particle system implies a natural way of obtaining @xmath8 of a statistically isotropic digitized medium , which efficiently uses all possible vector - information in the medium . for each black pixel ( or `` gas molecule '' ) @xmath29 , the distances between pixel @xmath29 and all the other pixels @xmath30 are computed and binned to generate a histogram for the number of black pixels separated from each other by distance @xmath21 .
another histogram for the number of lattice sites separated from a reference site by distance @xmath21 is also generated by computing and binning all possible site - separation distances .
suppose the histograms for black pixels and lattice sites are stored in the array @xmath31 $ ] and @xmath32 $ ] , respectively .
it is easy to show that the two - point correlation function can be obtained from @xmath33 = bn[r]/sn[r],\ ] ] ( see fig .
[ fig2 ] ) . in other words , to compute @xmath0 we first calculate the fraction of occupied lattice sites separated by distance @xmath21 ( @xmath34 is an integer ) from a reference black pixel .
then we average this fraction over all black pixels ( by choosing every black pixel as reference pixel once ) to obtain @xmath8 .
this procedure is consistent with the geometrical probability interpretation of @xmath0 . for the digitized media
, a natural bin width could be the characteristic size of the pixel of the lattice , e.g. , the edge length of a square pixel if square lattice is used . @xmath13{fig2a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig2b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] at each step in the simulated annealing process , a trial configuration is generated by moving a randomly selected black pixel to an unoccupied lattice site .
configuration array _ can be used to speed up this process . in our 2d implementation ,
the configuration array is a 2d array with the entries being 1 and 0 , corresponding to the occupied and unoccupied lattice sites , respectively .
when a trial move is made , first we need to test whether it violates the impenetrability condition by checking whether the underlying lattice site is occupied or not .
this process just requires constant access time to an entry of the configuration array .
note that several trial moves can be attempted before a trial configuration is found for a high - density system . after a trial configuration
is generated , the old position information of the selected pixel is stored in the array designated @xmath35 . the next step is to recompute the two - point correlation function for the trial configuration . to do this efficiently , a _ distance matrix _ is set up when the system is initialized and updated if a trial configuration is accepted .
note that this matrix is symmetric .
for each trial configuration , only the position of a randomly selected black pixel is changed , e.g. , the @xmath36 pixel .
thus , only the @xmath36 row and column of the distance matrix need to be updated , which requires @xmath11 operations ( @xmath11 is the total number of black pixels in the system ) .
the old entries in the @xmath36 row ( or column ) of the distance matrix are stored in the array @xmath37 .
it is also unnecessary to recompute the entire array @xmath31 $ ] . recall that @xmath31 $ ] contains the binned number of black pixels separated from each other by distances between @xmath21 and @xmath38 ( @xmath28 is the bin width )
once the @xmath36 pixel is selected , the distances between pixel @xmath39 and all the other pixels are binned and stored in the array @xmath40 $ ] . after the trial configuration
is generated , the new distances between pixel @xmath39 and all the other pixels are binned and stored in the array @xmath41 $ ] .
the array @xmath31 $ ] is updated as follows for each @xmath21 : @xmath42 - bno[r ] + bnn[r ] \rightarrow bn[r].\ ] ] @xmath43 $ ] is then recomputed using eq .
( [ eq2 ] ) and the trial configuration is accepted with the probability given by eq .
( 57 ) in paper i. if the trial configuration is rejected , all information of the old configuration can be restored easily using the arrays @xmath35 , @xmath37 , @xmath40 $ ] and @xmath41 $ ] . for example , @xmath31 $ ] of the old configuration can be restored by @xmath44 - bnn[r ] + bno[r ] \rightarrow bn[r].\ ] ] the above procedures are repeated until the energy of the ( re)constructed medium is below the energy threshold [ see eq .
( [ eq201 ] ) in sec .
iii ] or the total number of evolutions reaches the prescribed limit value . in the ( re)construction of media composed of well - defined `` particles '' or large clusters , we find out that in the later stages of simulated annealing , the random pixel - selection process is very inefficient .
many trial moves are rejected because a majority of selected pixels are inside the formed `` particles '' or clusters and it is energetically unfavorable to move them outside .
this requires the use of a biased pixel - selection process , i.e. , surface optimization .
the idea of surface optimization is analogous to the physical process of solidification : when the nuclei of proper sizes have been formed , they capture more solutes from surrounding solution to further decrease the total free energy . in the simulated annealing process ,
once the `` nuclei '' are formed , the random pixel - selection process is replaced by a biased pixel - selection process , i.e. , only those in the surrounding `` solution '' or on the surface of a `` nucleus '' are selected to be moved to or along the surface of randomly selected `` nuclei '' .
it is natural and easy to incorporate the surface optimization with the lattice - point algorithm because the black pixels are already considered as `` molecules '' .
each black pixel is assigned a `` free energy '' , which is the minus of the number of the nearest neighbors of that pixel . if a pixel is inside a `` nuclei '' , it has the largest number of nearest neighbors and the lowest `` free energy '' , which equals @xmath45 if a square lattice is used and @xmath46 if a triangular lattice is used .
if the pixel is on the surface of a `` nucleus '' or in the surrounding `` solution '' , its `` free energy '' will be relatively higher .
the highest free energy is @xmath47 , which means the pixel is separated from all the others .
thus , the black pixels are grouped into two subsets : the _ low - energy _
subset that contains pixels with lowest `` free energy '' and the _ high - energy _
subset that contains the other pixels .
when the trial move is made , only the pixels in the high - energy subset are selected to bias the move . if the trial move is accepted , the free energy of the moved pixel and its neighbor pixels are recomputed ; the subsets are updated .
@xmath48{fig3.eps } \end{array}$ ] numerical experiments show that applying surface optimization properly will further decrease the final energy of the ( re)constructed medium by a factor of @xmath49 ( see fig .
[ fig3 ] ) , if the same cooling schedule is used .
thus , surface optimization enables one to efficiently produce more accurate structures associated with the imposed correlation functions .
moreover , by setting a lower energy threshold , one can address the non - uniqueness issue quantitatively .
progress on this topic will be reported in our future publications . a 2d digitized heterogeneous material ( medium )
can be represented as a 2d array and the real morphology of the material also depends on the lattice on which the system of pixels is built @xcite . here , we mainly focus on two commonly used lattices in the literatures , namely , the square lattice and the triangular lattice ( see fig . [ fig4 ] ) .
@xmath13{fig4a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig4b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] implementation of the algorithms discussed above on a square lattice is easier than that on a triangular lattice because the former has orthogonal lattice vectors .
however , the pixels of a square lattice ( squares ) only have @xmath50-fold symmetry while those of a triangular lattice ( hexagons ) have @xmath51-fold symmetry , as shown in fig . [ fig4 ] .
we find that for the media with long - range correlations or large - scale structures , using a square lattice usually introduces undesirable anisotropy in the ( re)constructed systems , as shown in fig .
[ fig5 ] ; while for the media without long - range order or large - scale structures , both lattices work well .
@xmath13{fig5a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig5b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] it is worth pointing out that the triangular lattice is superior to the square lattice in the ( re)construction of _ anisotropic _ materials . in that case
, one can not count on lattice - point algorithm , which only uses radial averaged structural information .
several optimization directions need to be specified and processed separately .
the triangular lattice has six intrinsic directions due to the higher symmetry of its pixel shape while the square lattice only has four ( see fig .
[ fig4 ] ) .
methods using more optimization directions have been proposed by torquato and coworkers @xcite , i.e. , along the @xmath52- and @xmath53-degree directions in the square lattice ; but one needs to pay the cost of complexity of implementation in these cases .
an important issue that has not been emphasized in our previous work on the ( re)construction algorithms is the choice of the _ energy threshold _ @xmath9 , i.e. , the error tolerance of discrepancies between the statistical properties of the generated structure and the imposed ones . recall that in our case , the _ energy _ is defined as the sum of squared differences between target @xmath7 and @xmath8 of the constructed medium [ see eq .
( [ eq201 ] ) ] and the energy threshold @xmath9 is a prescribed value of @xmath5 such that when @xmath54 , the ( re)construction is terminated . @xmath9
indicates how accurately the medium is ( re)constructed . a smaller @xmath9 means the two - point correlation function of the generated medium matches the imposed one better . in the following , we will estimate the change of the energy of a digitized medium caused by a perturbation of its original structure ( i.e. , displacing a randomly selected black pixel ) .
this change of energy can be considered as the energy difference between the target and the constructed medium .
thus , by imposing a particular value of the energy difference ( i.e. , @xmath9 ) , we can in turn estimate the number of perturbed black pixels . in this way
, we quantitatively relate the energy threshold to the question of how well the medium is ( re)constructed . in general , the energy difference depends on how @xmath0 is sampled , the linear size of the system @xmath55 and the number of black pixels @xmath11 ( volume fraction of black phase @xmath56 ) , which will be discussed accordingly .
first , we consider the energy threshold of the lattice - point algorithm
. the two - point correlation function of a digitized medium is computed from eq .
( [ eq2 ] ) . if the generated structure perfectly matches the target structure , the energy given by eq .
( [ eq201 ] ) should be exactly 0 .
suppose the perfect ( re)constructed structure is perturbed by moving one of its black pixels to an unoccupied lattice site ; then @xmath8 is different from @xmath7 and @xmath5 becomes a small positive number .
note that we assume the perturbed structure does not have the same @xmath0 as the original structure , i.e. , there is no structural degeneracy .
let @xmath57 denote the smallest positive value of @xmath5 . from eq .
( [ eq201 ] ) , we see that the difference between @xmath0 and @xmath58 of every bin contributes to @xmath5 . from eq .
( [ eq2 ] ) and the fact that the last bin ( farthest away from the reference center ) contains the largest number of pairs of lattice sites , we have @xmath59\right ) - \hat{s}_2\left(\left[\frac{n}{2}\right]\right)}\right ] ^2 = \frac{4}{\omega_n^2},\ ] ] where @xmath60 $ ] is the integer part of @xmath61 and @xmath62 is the number of elements of the integer set @xmath63 , which is defined by @xmath64 \le m , n \le \left [ \frac{n}{2}\right ] , \left ( \left[\frac{n}{2}\right]-1\right)^2 \le m^2+n^2 \le \left(\left [ \frac{n}{2}\right]\right)^2 \right\}.\ ] ] in other words , eq . ( [ eq801 ] ) estimates @xmath57 using the change of @xmath5 from 0 caused by removing ( adding ) one pair of black pixels from ( to ) the last bin ( the pixels bounded in pairs due to the symmetry of the distance matrix ) .
also note that the pair of pixels are originally in ( or moved to ) a `` bin '' for the pair distance larger than @xmath60 $ ] , which we do not take into account for computation of @xmath0 . for a particular value of @xmath9
, the maximum number of _ misplaced pairs _ of black pixels can be estimated by @xmath65 the ratio of the number of misplaced pairs of black pixels over the total number of pairs of black pixels is given by @xmath66 instead of specifying @xmath9 , one can also specify the ratio @xmath67 ; and the threshold can be computed from eq .
( [ eq804 ] ) : @xmath68 for example , consider a system composed of @xmath69 pixels ( @xmath70 ) and @xmath71 .
the total number of lattice sites is @xmath72 and the total number of black pixels is @xmath73 .
@xmath62 can be obtained numerically , which is @xmath74 . from eq .
( [ eq801 ] ) , we have @xmath75 .
suppose we choose the threshold @xmath76 , from eq .
( [ eq803 ] ) , we have the number of misplaced pairs of black pixels @xmath77 , which seems to be a large number . however , when considering the total number of pairs in the system , we have @xmath78 ; from eq .
( [ eq804 ] ) , the ratio is given by @xmath79 which means only one out of a half million pairs is put in the wrong bin .
thus , the medium is ( re)constructed to a very high accuracy . in our simulations ,
we choose the threshold @xmath76 . for the orthogonal @xmath0-sampling algorithm ,
@xmath8 is sampled line ( column ) by line ( column ) by moving a line ( column ) segment of length @xmath21 one pixel distance each time and counting the times that both ends of the segment are black pixels .
this number is then divided by the total number of times that one moves the line ( column ) segment ( total number of pixels on the line ( column ) ) to obtain @xmath0 of that line ( column ) .
finally , the @xmath0 sampled from different lines and columns of the digitized medium are averaged to compute the @xmath0 of the whole medium .
similarly , consider the perfect ( re)constructed structure is perturbed by moving one of its black pixels to an unoccupied lattice site and there is no structural degeneracy , @xmath57 is given by @xmath80 where @xmath55 is the linear size of the system .
for a particular @xmath9 , the maximum number of _ misplaced black pixels _ is given by @xmath81 thus , the ratio of misplaced black pixels over the total number of black pixels can be obtained by @xmath82 if stead , @xmath67 is specified , the required energy threshold @xmath9 can be obtained from eq .
( [ eq903 ] ) : @xmath83 consider the same system used in the previous section ( @xmath70 , @xmath71 ) . if we require only 10 black pixels are misplaced , thus , the ratio @xmath84 . from eq .
( [ eq904 ] ) , we have @xmath85 in other words , if we choose the threshold @xmath76 , only 10 black pixels in the system are misplaced . the medium is ( re)constructed to a high accuracy .
it is worth noting that in the above discussions , we assume that the perturbed structure does not have the same @xmath0 of the original structure , i.e. , there is no structural degeneracy . in general , however , structural degeneracy does exist ( i.e. , media with the same @xmath0 but different @xmath86 , @xmath87 , ... ) .
results concerning these non - uniqueness issues will be reported in our future publications .
in this section , we illustrate the practical utility of our theoretical formalism in both two and three dimensions . in particular , we will consider two kinds of applications .
firstly , given a set of basis functions ( may not be complete ) , one can express the scaled autocovariance functions ( two - point correlation functions ) of a statistically homogeneous and isotropic medium in terms of the specified basis functions with certain accuracy .
realizations of the materials can be generated using proper ( re)construction procedures and subsequent analysis can be performed on the images to obtain effective macroscopic properties of interest ; see , e.g. , ref .
secondly , the map @xmath3 ( see the discussion below ) enables one to construct candidates of realizable two - point correlation functions with properties of interest , which in turn enables one to design and investigate materials with desired structural characteristics .
the ( re)construction of realizations of 3d medium from the information obtained from a 2d micrograph or image is of great value in practice @xcite .
therefore , we will apply the lattice - point algorithm to generate three - dimensional digitized realizations of a fontainebleau sandstone and a boron carbide / aluminum composite from two - dimensional tomographic images of their slices through the materials . in a successful reconstruction ,
other deemed crucial structural characteristics besides @xmath0 obtained from the reconstructed medium should also agree closely with that of the target medium .
consequently , in order to judge quantitatively how well the reconstructions are , we will measure and compare another important morphological descriptor , i.e. , the two - point cluster function @xmath88 , defined to be the probability of finding two points at @xmath89 and @xmath90 in the same cluster of the phase of interest @xcite .
for statistically homogeneous and isotropic media , @xmath4 only depends on the relative scaler distances between the points , i.e. , @xmath91 .
note that @xmath4 contains nontrivial topological `` connectedness '' information .
when large clusters are present in the medium , @xmath4 becomes a long - ranged function and its integral will diverge if the phase of interest percolates .
the measurement of @xmath4 for a 3d material sample can not be made from a 2d cross - section of the material , since it is an intrinsically 3d microstructural function @xcite .
to sample @xmath4 from a digitized medium , we associate each pixel with a cluster - index that indicates to which cluster the pixel belongs , and only bin the distances of pixel - pairs in the same cluster
. then the number of pair distances in each bin is normalized by the total number of distances between the lattice sites within that bin , which is similar to the procedure of sampling @xmath0 discussed in sec .
we will also study the reconstruction of a binary laser - speckle pattern in two dimensions and show that the algorithm can not reproduce the pattern accurately .
moreover , we introduce and discuss a classification of heterogeneous materials into _ multi - scale media _ and _ single - scale media_. a multi - scale medium ( msm ) is the one in which there are multiple characteristic length scales associated with different structural elements .
examples of msm include fractal patterns and hierarchical laminate composites @xcite .
a single - scale medium ( ssm ) is the one composed of structural elements associated with only one characteristic length scale , such as a fontainebleau sandstone and a boron carbide / aluminum composite .
we conclude that in general reconstructions using @xmath0 only work well for heterogeneous materials with single - scale structures .
however , two - point information via @xmath0 is not sufficient to capture the key structural features of multi - scale media . before presenting the aforementioned applications
, we will first consider a general form of @xmath3 , which includes all possible convex combinations of the basis functions . in paper
i , we introduced the idea of expressing the two - point correlation function of a statistically homogeneous and isotropic medium through a selected set of bases of the two - point correlation function space . in practice , it is convenient to use the scaled autocovariance functions that are equivalent to the two - point correlation functions , which are defined as follows : @xmath92 suppose @xmath93 is a set of bases of scaled autocovariance functions and @xmath3 is an map on @xmath93 composed of convex combinations and products of @xmath94 ( @xmath95 ) , thus @xmath96 = \wp[f_1(r),f_2(r), ... ,f_m(r)],\ ] ] is also a realizable scaled autocovariance function .
note that the choice of basis functions is not unique . in general
, one can consider that @xmath97 $ ] takes the form @xmath98 = \sum\limits_i \alpha_i f_i(r ) + \sum\limits_{i , j } \beta_{ij } f_i(r)f_j(r ) + \cdots,\ ] ] where the coefficients satisfy the condition @xmath99 then one can use standard regression methods to obtain the set of coefficients such that @xmath100 $ ] is the best approximation of the target function pointwisely .
the basis functions we consider here include debye random medium function @xmath101 , the family of polynomial functions @xmath102 , damped oscillating function @xmath103 , overlapping - sphere function @xmath104 and symmetric - cell material function @xmath105 , which are discussed in paper i. @xmath106{fig6a.eps}\\ \mbox{\bf ( a ) } \\\\\\
\includegraphics[height=5.5cm , keepaspectratio]{fig6b.eps}\\ \mbox{\bf ( b ) } \end{array}$ ] first , we investigate structural properties of a fontainebleau sandstone from a two - dimensional tomographic image of a slice through the material sample .
sandstone is an important porous medium in geo - physical and petroleum applications and has been the focus of many studies @xcite . a microstructural image of a slice of a fontainebleau sandstone is shown in fig .
[ fig6](a ) , in which the black areas are solid phases ( phase 1 ) and the white areas are void phases ( phase 2 ) .
the two - point correlation function of the _ void _ phase @xmath107 is shown in fig . [
fig6](b ) .
@xmath108{fig7a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig7b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] the scaled autocovariance function of the void phase in fontainebleau sandstone @xmath109 can be approximated by the convex combination of @xmath101 and @xmath103 as follows : @xmath110 where @xmath111 , @xmath112 are the combination coefficients and @xmath113 @xmath114 where @xmath115 , @xmath116 are the effective correlation length ; @xmath117 is the oscillating frequency and @xmath118 is the phase angle .
the two - point correlation function @xmath107 is approximated by @xmath119 where @xmath120 , @xmath121 are volume fractions of the solid and void phase , respectively ; @xmath122 is the discrepancy between the sampled two - point correlation function and the basis - function approximation .
the average of the absolute values of discrepancies @xmath123 , which indicates how well the sampled two - point correlation function is approximated by the convex combination , is defined as : @xmath124 where @xmath125 is the sample - length .
note that although the general form of @xmath3 given by eq .
( [ eq7 ] ) would work well if a complete set of basis functions is given , in the present case ,
our practical approach is enough .
@xmath126{fig8.eps}\\ \end{array}$ ] the 3d reconstruction of the fontainebleau sandstone from @xmath107 obtained from the digitized image of a 2d slice [ fig .
[ fig6](a ) ] is shown in fig .
visually , the reconstruction provides a good rendition of the true sandstone microstructure , as can be seen by comparing the 2d images .
as pointed out earlier , to ascertain whether the reconstruction is quantitatively successful , we also measure @xmath4 of the target and generated media .
since it is an intrinsically 3d microstructural function , we only compute and compare the two - point cluster functions of the 2d slices .
the measured @xmath127 for the void phase of the target and the reconstructed slices of the fontainebleau sandstone are shown in fig .
[ fig71 ] .
the figure reveals that although @xmath127 of the generated medium is slightly below that of the target medium , the discrepancies are acceptable ( e.g. , the largest discrepancy @xmath128 ) .
thus , we consider the reconstruction is successful .
also note that the close agreement of the two - point cluster function indicates that @xmath127 is largely determined by @xmath129 of the medium and suggests that @xmath127 might be expressed as a functional of @xmath129 .
@xmath106{fig9a.eps}\\ \mbox{\bf ( a ) } \\\\\\
\includegraphics[height=5.5cm , keepaspectratio]{fig9b.eps}\\ \mbox{\bf ( b ) } \end{array}$ ] @xmath108{fig10a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig10b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] @xmath126{fig11.eps}\\ \end{array}$ ] a 2d digitized image of a boron carbide / aluminum ( @xmath130/@xmath131 ) inter - penetrating composite and the sampled two - point correlation function of the _ aluminum _ phase ( white phase ) @xmath132 @xcite are shown in fig . [ fig8 ] .
as we can see from fig .
[ fig8](b ) , @xmath132 is essentially an exponentially decreasing function without any significant short - range correlation .
thus , @xmath132 can be approximated by @xmath133 where @xmath134 , @xmath135 are combination coefficients ; @xmath136 and @xmath137 are volume fractions of the boron carbide ( black ) phase and the aluminum ( white ) phase , respectively . @xmath101 and @xmath103
are given by eq .
( [ eq101 ] ) and eq .
( [ eq102 ] ) , respectively ; the parameters used are @xmath115 , @xmath138 and @xmath139 , @xmath118 .
the average of the absolute values of discrepancies @xmath140 is given by @xmath141 where @xmath125 is the sample - length .
the 3d reconstruction of the boron carbide / aluminum composite from @xmath132 obtained from the digitized image of a 2d slice [ fig . [ fig8](a ) ] is shown in fig .
[ fig9 ] . as in the previous section ,
to verify that the reconstruction is quantitatively successful , we compute and compare the two - point cluster function @xmath142 for the aluminum phase of the target and reconstructed 2d slices of the composite . as shown in fig .
[ fig91 ] , the reconstruction also slightly underestimates @xmath142 with acceptable discrepancies ( e.g. , the largest discrepancy @xmath143 ) .
the medium can be considered as successfully reconstructed .
also note that @xmath142 is long - ranged , indicating large clusters of the aluminum phase present in the media .
although in the above two examples our theoretical formalism works well , there are situations where the microstructural information contained in @xmath8 alone is not sufficient to accurately reconstruct a heterogeneous material .
one such example is the multi - scale structure of a binary laser - speckle pattern ( fig .
[ fig10 ] ) .
the figure reveals that there are three structural elements : `` particles '' , `` stripes '' and a background `` noise '' ( individual black pixels dispersed throughout the white phase ) .
thus , there are three characteristic length scales in the medium associated with these structural elements .
@xmath106{fig12a.eps}\\ \mbox{\bf ( a ) } \\\\\\
\includegraphics[height=5.5cm , keepaspectratio]{fig12b.eps}\\ \mbox{\bf ( b ) } \end{array}$ ] the reconstruction of the speckle pattern is shown in fig .
[ fig11 ] . comparing fig .
[ fig11 ] with fig .
[ fig10 ] , we can see that instead of reproducing all the structural elements in the target medium , the ( re)construction program seems to mix them up to generate a single - scale structure that has the same ( or to a very high accuracy ) two - point correlation function as the target medium .
we note that this is a numerical example of structural degeneracy of @xmath8 .
@xmath106{fig13.eps}\\ \end{array}$ ] the laser - speckle pattern we considered is also an example of the _ multi - scale media _ (
the separation of length scales in msm results in the inefficiency of the bulk - based structure characteristics ( e.g. , @xmath144-point correlation functions ) , since usually they can only pick up structural information associated with the largest length scale . for example , consider the dislocations in a crystalline solid as the phase of interest ; it is clear that the two - point correlation function of the `` dislocation '' phase is identically zero since the `` dislocation '' phase has no measure compared with the bulk `` crystal '' phase ( i.e. , the volume of dislocations is zero compared with that of the bulk crystal ) . in this extreme example , the ratio @xmath145 of the two characteristic length scales associated with the `` dislocation '' phase @xmath146 and the `` crystal '' phase @xmath147 is zero , i.e. , @xmath148 . in digitized media ,
the ratio of length scales is always positive due to the discrete nature of the system ; however when the ratio is significantly different from unity , the two - point correlation function alone is not able to capture the key structural features . on the other hand , extensive experience with successful reconstructions of single - scale media ( ssm ) from @xmath0
shows that the two - point correlation functions are indeed sufficient to determine the structures of ssm to a high accuracy .
thus , we conclude that in general reconstructions using @xmath0 work well for heterogeneous materials with single - scale structures , while the microstructural information contained in @xmath0 is not sufficient to accurately model multi - scale media .
note that more morphological information ( e.g. , the lineal - path function @xcite and the two - point cluster function , etc . ) can be used to model msm more accurately . however , even the cluster - type functions containing connectedness information of the media such as @xmath4 can not completely characterize msm statistically . from eq .
( [ eq4 ] ) , one can construct candidates of realizable two - point correlation functions using the basis functions .
given a set of basis functions with diverse and interesting properties , one would be able to construct a two - point correlation function that exhibits all the useful properties of the basis functions to some extend and generate an `` optimal '' structure that realizes all the desired structural features .
thus , the theoretical formalism enables one to design materials with _ structural _ properties of interest .
given an accurate structure - property relation , one could even design materials with _
physical _ properties of interest by manipulating their two - point correlation functions .
for example , adams , gao and kalidindi recently developed a methodology to obtain finite approximations of the second - order properties closure in single phase polycrystalline materials , from which the second - order microstructure design can proceed @xcite .
as pointed out in paper i , we know very little about the basis function set @xmath93 at this stage .
our choice of @xmath93 is based on the criteria that @xmath94 s should have simple analytical forms and they present typical features of certain known two - point correlation functions .
important features exhibited by most realizable two - point correlation functions are monotonically deceasing or damped oscillating , which corresponds to materials without or with significant short - range order , respectively .
another feature of two - point correlation functions that may affect the structures of the corresponding materials is the smoothness of the function , which is not emphasized in our previous work . in the following
, we will see through several examples that the properties of basis functions can be observed in the generated structures ; also the media with desired structural properties can be obtained by manipulating the combination coefficients and the parameters ( e.g. , effective correlation length ) in the basis functions . in the paper
i , we already investigated hypothetical correlation functions combining the exponentially decreasing and damped - oscillating features . in this section
, we provide an example of non - smooth correlation functions ( with discontinuous derivatives ) , i.e. , the polynomial function of order two @xmath149 , which is defined as @xmath150 where the parameter @xmath151 is the effective correlation length .
other correlation functions having this non - smooth feature include the functions of known constructions ( e.g. , @xmath104 , @xmath105 , etc . ) and other polynomial functions @xcite . a typical medium associated with this type of functions is composed of dispersions of fully penetrable `` particles '' .
the monotonically decaying part of the functions determines the size and shape of the `` particles '' and the penetrability is consistent with the flat `` tail '' of the functions , which implies no spatial correlation between the `` particles '' .
when these functions are use as dominant basis functions , one can expect the generated media also contain well - defined overlapping `` particles '' ; thus , surface optimization could be applied in the ( re)constructions .
however , when functions like debye random medium function are dominant in the convex combination , this algorithm modification should be used with care because the media could contain `` clusters '' of all sizes and shapes , which would significantly reduce the efficiency of the modified algorithm using surface optimization .
besides @xmath152 , we also use debye random medium function @xmath101 and damped - oscillating function @xmath103 as the basis functions . consider the simplest form of @xmath3 for these three basis functions , i.e.
, @xmath153 where @xmath154 ( @xmath155 ) satisfies @xmath156 and @xmath157 ; and @xmath101 , @xmath103 and @xmath152 are given by eq .
( [ eq101 ] ) , eq .
( [ eq102 ] ) and eq .
( [ eq14 ] ) , respectively .
the characteristic length scales of the generated structures are determined by the parameters in the basis functions ; and the ratios of the characteristic lengths are chosen to be close to unity , i.e. , the hypothetical media belong to ssm .
@xmath126{fig14.eps}\\ \end{array}$ ] @xmath108{fig15a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig15b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] suppose we choose the combination coefficients @xmath158 ; and choose the following values for the parameters in the basis functions : @xmath159 ; @xmath160 , @xmath161 , @xmath118 ; @xmath162 .
the combined autocovariance function @xmath163 is shown in fig .
[ fig12 ] and the constructed structure with volume fraction of black phase @xmath71 is shown in fig .
[ fig120 ] . from the figures
, we can see the effects of each basis function on the generated structure .
the short - range correlations in the structure are determined by @xmath152 , which allows spatially uncorrelated `` particles '' with diameter @xmath151 to form .
these `` particles '' would form clusters of different sizes as the volume fraction of the black phase increases .
the middle - range correlations in the structure are dominated by the oscillation part in @xmath163 , which is the contribution of @xmath103 . as we have shown in paper i ,
the parameter @xmath164 is manifested as a characteristic repulsion among different structure elements ( appearing in different forms at different volume fraction ) with diameter of order @xmath164 ; while @xmath165 controls the overall exponential damping , and thus , the effective range of the repulsion .
the large - scale correlation is dominated by the long `` tail '' of @xmath101 , which allows clusters of all shapes and sizes to form . in fig .
[ fig120 ] , we can clearly identify the structure elements associated with each basis function .
for example , the stripe - like structures are associated with the oscillating feature of @xmath163 , the width of which is approximately @xmath164 .
several `` particles '' with diameter @xmath166 can be found dispersed in the white phase , which correspond to the contribution of @xmath152 .
note that most `` particles '' form clusters together with the `` stripes '' at this relatively high volume fraction @xmath71 .
finally , the spatial distribution of large - scale clusters is determined by the `` tail '' of @xmath101 .
@xmath126{fig16.eps}\\ \end{array}$ ] @xmath108{fig17a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig17b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] suppose now we would like to generate a similar structure as the one in fig .
[ fig120 ] but with larger `` particles '' and clusters .
to `` grow '' the `` particles '' , we need to increase the parameter @xmath151 in @xmath152 , which controls the effective diameter of the `` particles '' .
to form larger clusters , we need to reduce the `` repulsion '' between the structural elements ; or equivalently , to suppress the oscillation introduced by @xmath103 .
this can be done by either increasing the parameter @xmath165 in @xmath103 or decreasing the coefficient @xmath167 .
we first adopt the former method .
thus , to construct a new structure with the required properties , we use the following parameters : @xmath159 ; @xmath168 , @xmath169 , @xmath118 ; @xmath170 ; and use the same @xmath154 @xmath171 as the last example . the combined autocovariance function @xmath163 is shown in fig .
[ fig13 ] and the constructed structure with volume fraction of black phase @xmath71 is shown in fig .
[ fig130 ] . from the figures
, we can see the generated medium indeed contains larger `` particles '' and clusters ; besides , there are still stripe - like structures associated with the repulsion part of @xmath163 .
@xmath126{fig18.eps}\\ \end{array}$ ] @xmath108{fig19a.eps } & \includegraphics[height=5cm , keepaspectratio]{fig19b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] now we would like to further increase the size of clusters in the constructed medium while keeping all the parameters in the basis functions unchanged .
this time we need to adjust the combination coefficients to obtain the required structure . to generate larger clusters , the repulsion effect need to be further suppressed
; thus , we need to reduce @xmath167 . since the sum of the coefficients must equal 1 , we also need to increase the other two @xmath154 ( @xmath172 ) .
note that further increasing @xmath173 makes @xmath152 dominant in the combination ; however , as we discussed before , @xmath152 corresponds to spatial uncorrelated `` particles '' and the size of clusters formed by these `` particles '' is determined by the volume fraction the `` particle '' phase , which can not be changed here .
so we choose to increase @xmath174 of @xmath101 based on the fact that the large correlation length and long `` tail '' of @xmath101 significantly stimulate the forming of large clusters .
thus , in the construction we use the following combination coefficients : @xmath175 ; and use the same parameters in the basis functions as the last example .
the combined autocovariance function @xmath163 is shown in fig .
[ fig14 ] and the constructed structure with volume fraction of black phase @xmath71 is shown in fig .
[ fig140 ] . from the above examples , we see that even a simple convex combination of the basis functions ( the simplest form of map @xmath3 ) can provide variety of candidates of realizable two - point correlation functions .
we can also obtain desired structures by manipulating the combination coefficients and the parameters in the basis functions .
thus , we argue that in general given a complete set of basis functions , eq .
( [ eq7 ] ) enables one to construct candidates of realizable functions with desired properties and to design materials with structural properties of interest .
in this paper , we described in great detail the lattice - point algorithm , which has been shown to be both efficient and isotropy - preserving for ( re)constructing statistically homogeneous and isotropic media .
the digitized medium ( pixel system ) can be considered as a lattice - gas system , in which the black pixels behave like nonoverlapping `` particles '' moving from one lattice site to another .
the two - point correlation function of the medium is sampled by binning all distances between black pixels and dividing the number in each bin by the total number of distances between all lattice sites in that bin . the energy threshold indicating to what accuracy the medium is ( re)constructed has been discussed in detail for the first time . this quantity is directly related to the issue of non - uniqueness of ( re)constructions .
we have also described an algorithm modification using surface optimization to further speed up the ( re)construction process and discussed its implementation based on our lattice - gas version of the digitized media .
numerical experiments have shown that by applying the surface optimization algorithm properly ( i.e. , in the ( re)constructions of media composed well - defined `` particles '' ) , the final energy can be reduced by a factor of @xmath49 if the same cooling schedule is used . the choice of different pixel - lattices has also been discussed .
we have applied the theoretical formalism proposed in paper i to model several examples of real materials . in particular
, we used the lattice - point algorithm to generate 3d realizations of a fontainebleau sandstone and a boron carbide / aluminum composite from 2d images of their slices through the materials .
the two - point cluster functions in the reconstructions in these two different cases matched the corresponding ones in the original materials , thus demonstrating the efficiency and accuracy of the ( re)construction algorithm in these cases .
we also studied the reconstruction of a binary laser - speckle pattern in 2d , in which the algorithm fails to reproduce the target pattern accurately .
we conclude that in general reconstructions using @xmath0 only work well for heterogeneous materials with single - scale structures .
however , two - point information via @xmath0 is not sufficient to accurately model multi - scale media . in paper
i , we pointed out that given a complete set of basis functions , one can obtain the basis - function approximations of the two - point correlation functions for the materials of interest to any accuracy . here
we only used a simple form of eq .
( [ eq7 ] ) due to our limited knowledge of the basis functions set .
we also constructed realizations of materials with desired structural characteristics by manipulating the combination coefficients and the parameters in the basis functions .
. ( [ eq7 ] ) enables one to construct candidates of realizable functions with desired properties and to design materials with structural properties of interest .
we are now developing efficient ( re)construction procedures that take into account additional microstructural information that characterize the media .
for example , using our lattice - gas version of the digitized media , the two - point cluster functions @xmath4 can be directly incorporated into the yeong - torquato scheme @xcite , which enables one to obtain more accurate ( re)constructions of the target media . also , a quantitative structure - property relation is necessary for the further applications of the theoretical formalism .
finite element analysis on the generated media will be performed to establish the relation . such work will be reported in our future publications . | in the first part of this series of two papers , we proposed a theoretical formalism that enables one to model and categorize heterogeneous materials ( media ) via two - point correlation functions @xmath0 and introduced an efficient heterogeneous - medium ( re)construction algorithm called the `` lattice - point '' algorithm . here
we discuss the algorithmic details of the lattice - point procedure and an algorithm modification using surface optimization to further speed up the ( re)construction process .
the importance of the error tolerance , which indicates to what accuracy the media are ( re)constructed , is also emphasized and discussed .
we apply the algorithm to generate three - dimensional digitized realizations of a fontainebleau sandstone and a boron carbide / aluminum composite from the two - dimensional tomographic images of their slices through the materials . to ascertain whether the information contained in @xmath0 is sufficient to capture the salient structural features , we compute the two - point cluster functions of the media , which are superior signatures of the microstructure because they incorporate topological connectedness information .
we also study the reconstruction of a binary laser - speckle pattern in two dimensions , in which the algorithm fails to reproduce the pattern accurately .
we conclude that in general reconstructions using @xmath0 only work well for heterogeneous materials with single - scale structures
. however , two - point information via @xmath0 is not sufficient to accurately model multi - scale random media .
moreover , we construct realizations of hypothetical materials with desired structural characteristics obtained by manipulating their two - point correlation functions . |
flattened , rotating circumstellar disks reprocess light from newly born stars in t tauri and herbig ae / be systems @xcite .
disks can modify the spectral energy distribution ( sed ) of the systems , and reveal themselves in resolved images at various wavelengths @xcite . in a disk ,
dust grains only exist outwards the sublimation radius @xmath0 , where dust reach sublimation temperature @xmath1 , leaving a dust - free hole at the center .
protoplanetary disks are usually optically thick to stellar radiation ( with exceptions such as transitional disks , @xcite ) . as a result ,
most part of the disk receives direct starlight only at the surface , under a small grazing angle determined by the shape of the surface @xcite . on the other hand ,
the entire wall of the rim at @xmath0 is in full sight of the star , heated by intense stellar radiation . as a result ,
the rim is vertically puffed up under the hydrostatic equilibrium condition as pressure balances gravity @xcite . as shown by @xcite and @xcite , a puffed inner rim blocks starlight from reaching to the disk behind , thereby creating a shadow in the outer disk .
depending on the disk properties , the outer disk may re - emerge from the shadow at a certain distance , or remain shadowed throughout .
the puffed up rim and the self - shadowing effect are more prominent in herbig ae / be stars than in t tauri stars , as dust sublimation occurs closer to the less luminous and cooler t tauri stars , and the covering angle of the rim is smaller @xcite .
self - shadowing by a disk rim is a key component in explaining observations of herbig ae / be stars .
in particular , the difference between group i and ii objects @xcite has been interpreted as that group i disks are flared while group ii disks are flat and ( completely ) self - shadowed by a puffed up rim @xcite .
also , a puffed up rim has been involved to explain the light curves of ux orionis stars @xcite , dust emission features in disks @xcite , and cooling of the surface layer in disks @xcite .
the change of rim height has been proposed to account for the variability of disk infrared excess @xcite as well as the overall surface brightness of the disk in scattered light imaging @xcite . moreover ,
an evolutionary sequence has been proposed by @xcite and @xcite : as the shadow cast by the rim extends further as the amount of small grains decreases due to grain growth and settling , eventually an initially flared group i object evolves into a flat self - shadowed group ii object at late stages .
recently , thanks to a few high angular resolution imaging instruments that have been mounted on 8-meter class ground based mirrors in the past few years , near - infrared ( nir ) imaging of polarized scattered light from disks have found structures that may be produced by self - shadowing in disks .
@xcite suggested that the two nulls seen on the ring in the @xmath2 band polarized intensity ( pi ) image of the hd 142527 system are shadows cast by an misaligned inner disk ( see also @xcite , @xcite , @xcite , @xcite , and @xcite ) .
@xcite found that the two spiral arms in the mwc 758 system may have shadows in its @xmath2 band pi images ( see also @xcite ) .
in addition , various rings and arcs and gaps in disks have also been suggested to be manifestations of self - shadowing ( e.g. hd 144432 , @xcite ; hd 169142 , @xcite ; hd 163296 , @xcite ; tw hydrae , @xcite ; hd 169142 , @xcite ) . in these cases ,
the gap region has been interpreted as being shadowed by puffed up inner disk structures , with the outer gap edge presumed to be the edge of the shadow .
on the theory side , the effect of shadowing have been explored by many in the past through radiative transfer modeling , with the shadow cast by either a puffed up inner rim or hydro features produced by planets ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . however , most works focused on the sed or the thermal emission of the disk
. the effects of a puffed up rim in high angular resolution scattered light imaging have not been throughly examined . in particular , whether observed rings and arcs in disks can be the outer edge of shadows remain unanswered . in this paper ,
the properties of the shadow cast by a puffed up inner rim in nir polarized light imaging are studied through radiative transfer simulations using the @xcite code .
disk vertical structures are self - consistently calculated assuming hydro static equilibrium ( hseq ) .
synthetic images of disks with either a naturally puffed up rim or a parametrized rim are produced , and their properties are measured and compared with observations .
the structure of the paper is as following .
the models and mcrt simulations are introduced in section [ sec : models ] ; the results are presented in section [ sec : results ] , followed by a summary and discussion in section [ sec : summary ] .
mcrt simulations are carried out using the ( * ? ? ?
* see also @xcite ) code , which has been used to model protoplanetary disks in the past ( e.g. @xcite ) .
the simulation setup is largely adopted from @xcite , and is briefly summarized here .
the central star has @xmath3 , @xmath4 , and @xmath5k , typical for a herbig ae / be system .
the disk is axisymmetric , and is constructed in spherical coordinates @xmath6 ( radial ) and @xmath7 ( polar , midplane is at @xmath8 and poles are at @xmath9 ) .
the grid has 500 cells in @xmath6 direction with logarithmic spacing and 200 cells in @xmath7 direction with power law spacing to over sample the inner disk and the disk midplane .
the inner edge of the disk is at dust sublimation radius @xmath100.5 au , where dust temperature reaches 1600 k. the outer edge of the disk is at 200 au .
the surface density @xmath11 follows a power law with radius as @xmath12 where @xmath13 is the normalization at @xmath14 .
the @xmath15 power law has been suggested by both detailed disk structure calculation ( e.g. * ? ? ?
* ) and mm disk observations ( e.g. * ? ? ?
* ; * ? ? ?
the total mass of the disk is assumed to be @xmath16
. a small accretion rate of @xmath17/yr is assumed to represent a typical accretion state , and also to provide nominal heating to the shadowed region to prevent the development of a known ripple instability @xcite .
all simulations are run with 1 billion photon packages .
synthetic @xmath2 band ( 1.6 @xmath18 ) images for short , which has the unit [ mjy arcsec@xmath19 , or [ ergs s@xmath20 @xmath21 hz@xmath20 arcsec@xmath19 .
this quantity is sometimes referred to as `` spectral radiance '' . ] and sed are produced for each model .
all models are assumed to be at 140 pc away .
full - resolution mcrt images are convolved by a circular gaussian kernel with a full width half max ( fwhm ) of @xmath22 to achieve an angular resolution comparable to observations by subaru , vlt , and gemini ( the fwhm of an airy disk is @xmath23 for a primary mirror with a diameter @xmath24 m at @xmath25 ) . in simulations ,
photons from the central star are absorbed and reemitted or scattered by the dust in the surrounding disk .
the temperature in each grid cell is calculated based on the radiative equilibrium algorithm described in @xcite .
the anisotropic scattering phase function is approximated using the henyey - greenstein function .
polarization is calculated assuming a rayleigh - like phase function for the linear polarization @xcite .
the vertical density structure of the disk is solved through iterations assuming self - consistent hseq in the vertical direction @xmath26 , as gas pressure balances gravity : @xmath27 where @xmath28 is pressure and @xmath29 is density .
simulations are iterated on the temperature @xmath30 calculated from the radiative equilibrium solution of the radiation transfer equation at least 10 times to achieve convergence on the vertical structures @xcite . in this
study the stellar radiation is the main energy source of the disk ( heating from accretion has little effect except in the umbra region right behind the rim ) , and the gravity from the central star is the sole gravitational source . note that the vertical structure of the disk ( as well as the puffed up rim ) may be affected by other mechanisms , such as the gravitational perturbation from an orbiting companion ( particularly if on an inclined orbit ) , or turbulences in the disk .
the naturally puffed up rims based on the hseq condition in this study do not account for these additional mechanisms .
there are two types of dust grains in the models , the `` small '' grains and the `` big '' grains .
the gas - to - dust - mass ratio ( including both the small and big grains ) is assumed to be 100:1 ( therefore the total mass of the dust grains @xmath31 is @xmath32 ) .
the small grains represent primordial dust particles .
their properties are assumed to be the standard interstellar medium ( ism ) grains as in @xcite .
these grains contain silicate , graphite , and amorphous carbon , and their size distribution is a smooth power law in the range of 0.02 - 0.25 @xmath18 followed by a sharp cut off beyond 0.25 @xmath18 .
the small grains are well mixed with gas as the two are dynamically well coupled - sized grains may still have some degree of settling , depending on the strength of the turbulences in disks that stir up dust grains vertically @xcite . due to a lack of good understanding of the nature and strength of the turbulences in protoplanetary disk , settling of these sub-@xmath18-sized small grains
is not considered in this work . ] .
the large grains represent an advanced stage of grain growth and coagulations .
model 1 in @xcite is adopted for their properties .
they are made of amorphous carbon and astronomical silicates , and have a power - law size distribution as @xmath33 , an exponential cutoff beyond @xmath34 , and a maxim size of 1 mm .
the big grains tend to settle to the disk midplane , as they usually dynamically decouple from the gas and are no longer held in the vertical direction by the gas pressure @xcite .
the vertical distribution of the big grains is assumed to be a gaussian profile with a small scale height @xmath35 ( i.e. their vertical structure is parametrized and not evolved in the hseq iterations ) .
therefor the big grains form a thin layer at the disk midplane .
the details of the properties and spatial distribution of the big grains have little effect on the scattered light images and sed presented below , as long as they are well embedded at the disk midplane .
the optical properties of both grains can be found in figure 2 in @xcite .
two sets of models , series a and b , are constructed and listed in table [ tab : models ] .
model base is the fiducial model , which has no big grains .
series a is constructed by fixing @xmath31 while decreasing @xmath36 and increasing @xmath37 , similar to the ml series in @xcite and bl series in @xcite .
this scenario ( from model base to a4 ) is to mimic an evolutionary sequence : the amount of small grains decreases with time as grain growth and coagulation gradually convert them into bigger particles that settles down at the disk mid - plane .
the inner rim in base to a4 is naturally puffed up under the hseq condition ( only the small grains ) . as will be shown in the next section , while series a does gradually turn from a flared disk ( base ) into a self - shadowed disk ( a4 ) ,
the boundary between the shadowed inner disk and the flared outer disk is not sharp .
this is because the rim does not have a sharp edge in the vertical density distribution , and the radial optical depth of the rim only gradually decreases with height . to maximize the sharpness of the rim edge and produce features with the highest contrast possible in scattered light imaging , series b
is created . in b1 to b6 , an artificial opaque rim is placed at @xmath38 to @xmath39 , with a parametrized height @xmath40 expressed in the angle subtended by half of the rim to the star ( i.e. the vertical structure of the artificial rim is fixed and not evolved under the hseq condition ) .
the opaque rim is constructed by filling the region with small grains of a uniform density @xmath41 , so that the radial optical depth @xmath42 of the rim is greater than @xmath43 at wavelengths of interests , therefore completely blocks the starlight in the solid angle subtended by the rim .
in other words , in the polar direction @xmath44 inside the rim , then @xmath45 outside the rim . this artificial vertically constant density structure
is only designed in order to create the sharpest possible shadows .
it is not meant to have a physical motivation .
both the naturally puffed up rim and the parametrized artificial rim are illustrated in figure [ fig : verticalprofiles ] .
this section presents the main modeling results , first the structure of the disks ( section [ sec : structure ] ) and then observational aspects of the models ( section [ sec : modelobservations ] ) .
the density and temperature structure for models base and b2 are shown in figure [ fig : verticalprofiles ] in spherical coordinates , with the disk surface over - plotted , defined as where the radial optical depth @xmath42 at 0.4 @xmath18 at each polar angle @xmath7 from the stellar surface ( the stellar flux peaks aroundt 0.4 @xmath18 ) reaches 1 . in spherical coordinates a shadowed region has a flat surface ( constant @xmath7 with increasing @xmath6 ) , while a flared region has a rising surface ( increasing @xmath7 with increasing @xmath6 ) .
the disk surface traces out the boundary between the hot disk atmosphere and the cool interior , as most starlight intercepted by the disk is deposited near the surface . in the model base ,
a puffed up inner rim is clearly visible in the density structure as a vertically extended structure close to @xmath0 .
the rim casts a shadow behind from @xmath0 to @xmath46 au , where the surface comes out of the shadow and starts rising .
however , as discussed in section [ sec : models ] , the puffed up rim has a blurry edge . as a result ,
the edge of the shadow is not sharp , and the disk surface smoothly transits from flat to flared around 3 au . in the model b2
, a parametrized artificial rim is placed at @xmath0 with @xmath47 .
the sharp rim edge results in a sharp turning point on the @xmath48 surface at @xmath49 au , where the surface suddenly rises . inside the shadow region ,
the disk does not have a uniform structure , and can be separated into two regions .
an umbra exists immediately behind the rim ( boundary indicated by the dashed line ) , within which the entire surface of the star is not visible . in the umbra ,
practically the only heating sources are radiation diffusion and accretion .
this region becomes sufficiently cold to collapse to the midplane . on the other hand ,
the region in between the umbra and the @xmath48 surface receives additional heating from the following ways : 1 .
once outside the umbra and moving towards @xmath50 , the disk receives direct radiation from a bigger and bigger polar region on the star .
2 . inward heat diffusion from @xmath50 heats up the adjacent region , smoothing out the transition and preventing a sudden jump in @xmath51 @xcite .
3 . indirect illumination and heating from the halo @xcite .
in particular , closer to @xmath50 , secondary ( multiple ) scattering from the disk atmosphere in the flared region becomes more important .
nir flux from the opposite side of the inner rim .
the strength of additional heating sources 1 - 3 increases with radius , puffs up the disk moderately starting at @xmath522 au . eventually the @xmath48 surface comes out of the shadow at 13 au and disk stays flared outward .
the @xmath48 surface for all models are plotted in figure [ fig : surface ] . as shown in @xcite and @xcite , from base to a4
the disk gradually evolves from flared to flat ( completely self - shadowed by its own puffed up rim ) , as the amount of small grains decreases .
the shadowed - to - flared - disk transition occurs more and more smoothly , and @xmath50 can not be clearly defined . on the other hand ,
the surface transition is sharp in the b series , as the surface suddenly turns upward at [ @xmath50,@xmath40 ] ( marked by dots in the right panel and labeled in table [ tab : models ] .
the entire disk in b6 is inside the shadow so @xmath50 is not defined . ) .
the transition point rises and moves outward as @xmath40 increases , and the surface always converges to model base once out of the shadow , indicating that the flared part of the disk in all models are similar .
the midplane temperature of the disk @xmath51 is shown in figure [ fig : midplanet ] for the b series .
there are two bumps on the curves .
the first bump occurs when the disk comes out of the umbra , and the second bump occurs at the shadow edge @xmath50 . in the umbra
all models have the same @xmath51 profile ( best traced out by model b6 ) .
the location of the first bump moves outward as @xmath40 increases and the umbra extends .
after the first bump the shadow reduces the midplane temperature , consistent with @xcite , and models converge again to another profile in cases that the two bumps are not too close to each other ( b3 to b6 ) . finally , all models converge to the base model at @xmath54 , echoing the convergence of the @xmath48 surface in figure [ fig : surface ] , and demonstrating the similarity of the models in the flared region .
the @xmath48 surface officially comes out of the shadow at @xmath50 and starts to be directly irradiated by the bulk part of the stellar surface ( @xmath55latitude@xmath56 ) .
however , the transition to the `` flared profile '' occurs much earlier than @xmath50 .
for example , b4 starts to deviate from its `` shadow profile '' ( traced out by b6 ) at @xmath5210 au , and gradually converges towards its `` flared profile '' ( traced out by base ) until the eventual merger at @xmath57 au .
this echoes the smooth transition in temperature and density structures seen in figure [ fig : verticalprofiles ] .
vertical density structure of the disk is determined by the temperature structure under the hseq condition .
therefore , at radius of a fraction of @xmath50 , the disk already starts to deviate from its thin `` shadow state '' and become modestly puffed up , until eventually smoothly joins the flared disk at @xmath50 .
this is illustrated in figure [ fig : verticalprofiles ] for b2 , in which case the thin `` shadow state '' ends at @xmath522 au , and in between 2 au and @xmath49 au the disk is modestly puffed up , and beyond 13 au the disk becomes fully flared . in summary ,
the disk in series a evolves from a mostly flared disk to a self - shadowed flat disk , but the shadowed - to - flared - disk transition is smooth , due to the lack of a sharp rim edge . in the extreme cases , with parametrized sharp rims in series b , the transition on the @xmath48 surface becomes sharp , but
the transition in the underlying disk density and temperature structure is still smooth and occurs over a radius range .
sed for all models are shown in figure [ fig : sed ] .
series a mimics the evolutionary sequence from group i to ii , as shown in @xcite .
the 1 - 100 @xmath18 infrared excess drops as the total amount of small grains decreases and the disk becomes more and more self - shadowed , meanwhile the far infrared and mm emissions beyond 100 @xmath18 rise due to more big grains . for series
b , the nir excess from 1 - 10 @xmath18 comes from the artificial rim , which increases as the rim grows higher from b1 to b6 .
meanwhile the 10 - 100 @xmath18 excess decreases as the shadow edge expands and a bigger disk region falls into the shadow .
this is similar to the `` seesaw '' phenomenon in @xcite , which has been explained by puffed up rims with varying heights . beyond 100 @xmath18
the emissions mainly come from the outer part of the disk , which remains flared and has a similar density and temperature structure in most models thus the emission is little changed . scattered light images trace structures at the disk surface , set by the dust distribution and intrinsically determined by the temperature structure under the hseq condition . while formally inside @xmath50 , the shadowed region may still receives and scatters nir photons from ( 1 ) the polar region of the star ( in the region immediately below the radial @xmath48 surface ) ; ( 2 ) the bulk part of the star ( high in the disk atmosphere ) ; ( 3 ) the opposite side of the rim ; and ( 4 ) secondary scattering photons from the disk halo ( in particular at radius close to @xmath50 ) . if the rim is naturally puffed up , stellar photons may also be scattered by the tenuous rim edge into the shadowed region
therefore , it is expected that the shadowed region to be not completely dark , and the surface brightness should converge to the flared disk when getting closer to @xmath50 .
@xmath2 band polarized intensity images for model b2 and b4 are shown in figure [ fig : images ] . in b2 , a marginal ring / gap structure ( i.e. a region with pi@xmath58 ) at @xmath50 can be seen at face - on viewing angle , inside which the shadowed region has low surface brightness . if we define `` ring / gap contrast '' as the ratio of the azimuthally averaged surface brightness at the peak of the ring to the bottom of the gap , the contrast is 1.33 for the raw image and 1.03 for the convolved image . at a 45@xmath59 inclination angle ,
the ring becomes elliptical , and due to the forward / backward scattering effect the front side is brighter than the back side @xcite . in b4 ,
the shadow edge is not distinctive in all panels . at face - on viewing angle , the azimuthally
averaged surface brightness decreases with radius throughout the disk ( see figure [ fig : image - rp ] ) .
the radial profile of face on @xmath2 band pi images for all models are shown in figure [ fig : image - rp ] .
for the a models , converting small grains into vertically settled big grains gradually turns a bright flared disk into a faint flat one .
the brightness of the disk drops by 1 - 2 orders of magnitude .
however this process happens in a rather smooth manner that the surface brightness of the entire disk drops all together , while the outer disk drops faster than the inner region .
the dependence of pi on @xmath6 changes from pi@xmath60 in the base model to pi@xmath61 in a4 .
the surface brightness distribution remains smooth and relatively featureless throughout in all cases .
in particular , no ring / gap structure exists at the edge of the shadowed region , in either the original images or the @xmath62 scaled images . for the b models ,
the shadowed - to - flared - disk transition is clearly visible on the radial profiles . in general
, the surface brightness radial profile can be described by a 3-stage broken power law : ( 1 ) well inside the shadow different models have a similar radial profile as pi@xmath63 ( best seen in b3 to b6 as their shadows extend far enough ) ; ( 2 ) in the flared region all models converge to pi@xmath60 , the same as the base model ; and ( 3 ) a transition region with pi @xmath64 ( constant pi ) connects the two .
this is consistent with the trend seen earlier in the disk density , temperature , and surface structures , that b models have essentially the same structure both in the inner shadowed region and the outer flared region , with a wide transition region in between . and
similar to the @xmath51 profile , the transition occurs over a radius range and not suddenly . for b2 to b5 in b1
is too close to the star and beyond the capability of current nir imaging facilities . ] , radial profiles smoothly deviate from the @xmath65 profile around @xmath66 , and gradually converge to the @xmath60 profile around @xmath50 .
b3 has a marginal `` gap / ring '' structure similar to b2 , with very weak ring / gap contrast ratios ( listed in table [ tab : models ] ) .
pi never increases with radius throughout the disk in b4 to b6 , and a ring / gap structure is not present around the shadow edge .
the difference between b2/b3 and b4/b5 is probably caused by the fact that the transition region is wider in b4/b5 due to their large shadow size , so the transition occurs even more smoothly .
the effects of puffed up inner rims in protoplanetary disks in both temperature and density structure and nir scattered light imaging are studied .
two types of rims are explored : natural rims that are self - consistently puffed up under the hydrostatic equilibrium condition ( series a ) , and artificial opaque rims with parametrized heights and sharp edges in the vertical direction ( series b ) .
the main results from the models are : 1 .
an inner rim naturally puffed up under the hseq condition creates a shadow behind , within which the disk surface is flat and beyond which it is flared ( the a models ) . decreasing the amount of small grains gradually turn the entire disk into a self - shadowed and flat disk , confirming previous results such as in @xcite ( figure [ fig : surface ] ) .
2 . for naturally puffed up rims ,
the shadowed - to - flared - disk transition occurs smoothly , and no sharp / distinct turning point on the disk surface exists .
this is mainly due to the fact that hseq rims do not have a sharp vertical edge , and their `` radial optical depth '' gently decreases with height . as a result ,
the scattered light images have a smooth / featureless surface brightness distribution . from base to a4 , the surface brightness of the disk drops by 1 - 2 orders of magnitude , and the radial dependence changes from a shallower @xmath67 to a steeper @xmath68 ( figure [ fig : image - rp ] ) .
3 . the most extreme rims , which are parametrized artificial rims with a razor sharp vertical edge in the b series ,
do create a sharp turning point on the radial @xmath48 surface ( figure [ fig : surface ] ) .
this point formally defines the edge of the shadow @xmath50 , beyond which the surface is flared and sees the bulk part of the star .
both the inner shadowed region and the flared region are similar in models with different rim heights .
on the other hand , the shadowed - to - flared - disk transition does not occur abruptly in the underlying disk density and temperature structures .
the midplane temperature starts to deviate from its `` shadow profile '' at a fraction of @xmath50 ( figure [ fig : midplanet ] ) , and eventually converges to its `` flared profile '' around @xmath50 .
the density structure of the disk follows the change in @xmath51 under the hseq condition , thereby exhibits a broad transition region as well ( figure [ fig : verticalprofiles ] ) .
this is mainly caused by partial illumination by the polar region of the star , multiple scattering , and radiation diffusion within the disk .
4 . even the parametrized razor sharp rims in the b models can not produce prominent rings and arcs in scattered light images as what have been directly imaged recently . on the other hand ,
self - shadowing by the rim does have a clear effect on the radial profile of the scattered light .
the surface brightness radial profiles of the b models can be described by a 3-stage broken power law : two steep profiles in the inner shadowed region and the flared region , connected by a shallow profile in a broad transition region ( figure [ fig : image - rp ] ) .
the enhancement around the shadow edge however is usually not strong enough to make a prominent ring / gap structure .
b2 and b3 show a very marginal ring / gap structure around @xmath50 ( figure [ fig : images ] ) , while other models remain largely smooth and featureless throughout in both the original and @xmath62 scaled images .
remarks on the implications of the models and observations : 1 .
a faint disk in nir direct imaging observations with a relatively steep surface brightness radial profile ( i.e. @xmath69 ) is consistent with small grains being heavily depleted and the entire disk is self - shadowed by the inner rim .
the surface brightness threshold depends on the stellar luminosity . for a @xmath43 k , 2.5 @xmath70 , and 2 @xmath71 central star ,
the threshold is about 1/@xmath72 mjy arcsec@xmath73 .
such kind of faint disks may have already been found in observations ( e.g. hd 139614 , subaru direct imaging , m. fukagawa , private communication , hashimoto et al . in prep . ) .
2 . even with the sharpest possible rims as the parametrized rims in b models , the shadow edge fails to manifest itself as prominent rings / arcs in scattered light images .
the highest ring / gap contrast ratio in full resolution images is only 1.33 , and practically no ring / gap structure is found in convolved images with an angular resolution of 0.05@xmath74 .
note that the convolved images in this study have the ideal chance to review potential disk structures , as ( 1 ) they are convolved by a gaussian psf , which smears the intrinsic full resolution images less than a realistic psf that has a similar kernel but more extended wings ; ( 2 ) no observational noises and instrument effects are included , which are expected to further weaken the detectability of features in images .
3 . realistic rims in a models can only produce structures even smoother than b models in scattered light images .
radial structures with high contrast recently observed in nir direct imaging surveys , such as the arcs in the hd 163296 system @xcite , the gap in the hd 169142 system @xcite , the spiral arms in the sao 206462 system @xcite and the mwc 758 system @xcite , and various rings and arcs in transitional disks ( e.g. * ? ? ?
* ; * ? ? ?
* ) are unlikely to be the shadow edges of puffed up rims .
it is more likely that they are created by local density or scale height structures @xcite , such as gaps ( e.g. * ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) and spiral arms ( e.g. dong et al .
2015b ) produced by planets in the density distribution .
4 . though unable to create arcs / rings and gaps in disks , a shadow cast by a puffed up rim may still have an effect on the surface brightness radial distribution of scattered light images . with a vertically sharp rim ,
scattered light images show a 3-stage broken power law : 2 steep radial profiles in the inner and other disk connected by a shallow transition region around the shadow edge .
such systems may have already been observed : the @xmath2 band scattered light image of tw hydrae has a shallow @xmath75 transition region located at @xmath7620 - 40 au , in between 2 steep @xmath65 and @xmath77 profiles in the inner and outer disk @xcite .
the shallow region may actually be a gap , which can be either confirmed or ruled out by future observations with better image quality and s / n ratio . if the shadow region is not a gap , then comparing with the b models this may suggest that the disk is self - shadowed , and the shadow edge is around @xmath78 au ( note that tw hydrae has an @xmath79 au inner hole at the center , @xcite , therefore perhaps the shadow is created by a puffed up cavity wall at @xmath80 au instead of a rim at @xmath81 au ) .
however a self - consistent structure puffed by the hseq condition is unlikely to account for the observation , as the shadowed - to - flared - disk transition is too smooth .
the needed puffed up structure has to have a somewhat sharp cutoff , perhaps sculpted by other mechanisms in the system .
the realization of such a structure with a physical origin can be the subject of future studies .
i thank misato fukagawa , jun hashimoto , and barbara whitney for useful discussions and help on the paper , and the referee cornelis dullemond for a helpful referee report . this work is supported by nasa through hubble fellowship grant hst - hf-51320.01-a awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa , under contract nas 5 - 26555 .
this research made use of the savio cluster at uc berkeley , and the lawrencium cluster at the lawrence berkeley national laboratory ( supported by the director , office of science , office of basic energy sciences , of the u.s .
department of energy under contract no .
de - ac02 - 05ch11231 )
. cccccccc model & parametrized / natural rim & @xmath82 & @xmath40 ( @xmath59 ) & @xmath50 ( au ) & ring / gap contrast + base & natural & 1 & n / a & n / a & n / a + a1 & natural & @xmath83 & n / a & n / a & n / a + a2 & natural & @xmath84 & n / a & n / a & n / a + a3 & natural & @xmath85 & n / a & n / a & n / a + a4 & natural & @xmath86 & n / a & n / a & n / a + b1 & parametrized & 1 & 10 & 5 & n / a + b2 & parametrized & 1 & 12.5 & 13 & 1.33/1.03 + b3 & parametrized & 1 & 15 & 30 & 1.10/1.02 + b4 & parametrized & 1 & 17.5 & 52 & n / a + b5 & parametrized & 1 & 22.5 & 147 & n / a + b6 & parametrized & 1 & 27.5 & n / a & n / a + | we explore whether protoplanetary disks with self - shadowing from puffed up inner rims exhibit observable features in scattered light images .
we use both self - consistent hydrostatic equilibrium calculations and parameterized models to produce the vertically puffed up inner rims .
we find that , in general , the transition between the shadowed and flared regions occurs in a smooth manner over a broad radius range , and no sudden jump exists at the outer edge of the shadow in either the disk temperature or density structures . as a result ,
a puffed up rim can not create sharp ring / arc / spiral - arm - like features in the outer disk as have been detected in recent direct nir imaging of disks . on the other hand ,
if the puffed up rim has a sharp edge in the vertical direction , the shadowing effect can produce a distinct 3-stage broken power law in the radial intensity profile of the scattered light , with 2 steep surface brightness radial profiles in the inner and outer disk joined by a shallow transition region around the shadow edge .
these types of scattered light profiles may have already been observed , such as in the recent subaru direct imaging of the tw hydrae system . |
linear programming ( lp ) decoding , as an approximation to maximum - likelihood ( ml ) decoding , was proposed by feldman _ et al .
many observations suggest similarities between the performance of lp and iterative message - passing methods , e.g. in @xcite .
for example , we know that the existence of low - weight pseudo - codewords degrades the performance of both methods ( @xcite , @xcite ) . therefore , it is reasonable to make use of the simpler geometrical structure of lp decoding to make predictions on the performance of message - passing algorithms . on the other hand , there are differences which prevent us from making an explicit connection between these two approaches . for instance , given an ldpc code , adding additional parity checks that are satisfied by all the codewords can only improve lp decoding , while with message - passing algorithms , these parity checks may have a negative effect by introducing short cycles in the tanner graph .
this property of lp decoding allows improvements by tightening the relaxation .
another characteristic of lp decoding ( the _ ml certificate property _ ) is that its failure to find the ml codeword is detectable .
more specifically , the decoder always gives either the ml codeword , or a nonintegral pseudo - codeword as the solution .
these two properties motivate the use of an adaptive approach in lp decoding which can be summarized as follows : given a set of constraints that describe a code , start the lp decoding with a few of them , and sequentially and adaptively add more of the constraints to the problem until either the ml codeword is found or no further `` useful '' constraint exists .
the goal of this paper is to explore the potential of this idea for lp decoding .
we also prove some properties for the lp relaxation of ml decoding which can be useful for performance analysis of lp and/or iterative decoding algorithms .
we show that by putting lp in an adaptive setting , we can obtain the same performance as if a huge number of constraints were added to the relaxation from the beginning .
in particular , we have observed that while the number of constraints per check node required for convergence is exponential in the check node degrees for lp decoding , the adaptive method generally converges with a number of constraints which is a ( small ) constant , independent of degree distributions .
this property makes it feasible to apply lp decoding to higher - density codes .
the rest of this paper is organized as follows . in section
ii , we review feldman s lp decoding . in section iii , we introduce and analyze an adaptive algorithm to solve the original lp problem more efficiently . in section iv , we study how adaptively adding additional constraints can improve the performance .
finally , section v concludes the paper . due to space limitations ,
some of the proofs of the results are omitted .
consider a binary linear code @xmath0 of length @xmath1 .
if a codeword @xmath2 is transmitted through a binary - input memoryless channel , the ml codeword given the received vector @xmath3 is the solution to the optimization problem @xmath4 where @xmath5 is the vector of log - likelihood ratios defined as @xmath6 as an approximation to ml decoding , feldman _
proposed a relaxed version of this problem by first considering the convex hull of the local codewords defined by each row of the parity - check matrix , and then intersecting them to obtain what is called the _
fundamental polytope _ by koetter _
this polytope has a number of integral and nonintegral vertices , but the integral vertices exactly correspond to the codewords of @xmath0 .
therefore , whenever lp gives an integral solution , it is guaranteed to be the ml codeword . in feldman s relaxation of the decoding problem ,
the following is done for each row @xmath7 of the parity - check matrix .
suppose that the @xmath8th check node has the neighborhood set @xmath9 , i.e. @xmath10 contains the indices of the variable nodes that are directly connected to this check node .
then , add the following constraints to the problem : @xmath11 throughout the paper , we refer to the constraints of this form as _ parity - check constraints_. in addition , for any element @xmath12 of the optimization variable , @xmath13 , the constraint that @xmath14 is also added .
as any odd - sized subset @xmath15 of the neighborhood @xmath10 of each check node introduces a unique parity - check constraint , there are @xmath16 constraints corresponding to each check node of degree @xmath17 .
therefore , the total number of constraints and hence , the complexity of the problem , is exponential in terms of the maximum check node degree , @xmath18 .
this becomes more significant in a high density code where @xmath18 increases with the code length , @xmath1 . in this section , we show that lp relaxation of linear codes has some properties that allow us to solve the optimization by using a much smaller number of constraints . given a constraint of the form @xmath19 and a vector @xmath20 , we call ( [ general const ] ) an _ active constraint _ at @xmath21 if @xmath22 and a _ violated constraint _ or , equivalently , a _ cut _ at @xmath21 if @xmath23 considering a constraint that generates a cut @xmath24 at point @xmath13 , we can immediately make the following observations : @xmath25 the following theorem reveals a special property of the constraints of the lp decoding problem . [ one cut per check ] at any given point @xmath26^n$ ] , at most one of the constraints introduced by each check node can be a cut .
( proof omitted . ) having a linear @xmath27 code with @xmath28 parity checks , a natural question is how we can find all the cuts defined by the lp relaxation at any given point @xmath29 . for any check node and an odd subset @xmath15 of its neighborhood that introduces a cut , we know from ( [ obsv2 ] ) that the members of @xmath15 are the variable nodes with the largest values among the neighbors of the check node .
therefore , sorting the elements of @xmath13 before searching for a cut can simplify the procedure .
consider a check node @xmath8 . without loss of generality , assume that variable nodes @xmath30 , are the neighbors of this check node , and they are sorted with respect to their values such that @xmath31 .
the following algorithm is an efficient way to find the cut generated by this check node at @xmath32 , if it exists .
[ find cuts ] _ step 1 : set @xmath33 , @xmath34 and @xmath35 .
_ _ step 2 : check the constraint ( [ constraints ] ) .
if it is violated , we have found the cut . exit .
_ _ step 3 : set @xmath36 . if @xmath37 , move @xmath38 and @xmath39 ( the two largest members of @xmath40 ) from @xmath40 to @xmath15 _ _ step 4 : if @xmath37 and ( [ obsv1 ] ) is satisfied , go to step 2 ; otherwise , the check node does not provide a cut at @xmath41 . _
if redundant calculations are avoided , this algorithm can find the cut generated by the check node , if it exists , in @xmath42 time , where @xmath43 is the degree of the check node .
repeating the procedure for each check node , and considering @xmath44 complexity for sorting @xmath13 , the time required to find all the cuts at point @xmath13 becomes @xmath45 .
the simplex lp algorithm starts from a vertex of the problem polytope and visits different vertices of the polytope by traveling through the edges until it finds the optimum vertex .
the time required to find the solution is approximately proportional to the number of vertices that have been visited , and this , in turn , is determined by the number and properties of the constraints in the problem .
hence , if we eliminate some of the intermediate vertices and only keep those which are close to the optimum point , we can reduce the complexity of the algorithm . to implement this idea in the adaptive lp decoding scheme , we run the lp solver with a minimal number of constraints to ensure boundedness of the solution , and depending on the lp solution , we add only `` the useful constraints '' that cut the current solution from the feasible region .
this procedure is repeated until no further cut exists . to start the procedure ,
we need at least @xmath1 constraints so that the problem has a vertex that can become the solution of the first round . using the condition that @xmath14
, we add one side of these inequalities for each @xmath46 , depending on whether increasing @xmath12 increases or decreases the objective function . in other words , for each @xmath47 , we initially have the constraint @xmath48 the optimum ( and only ) vertex of this initial problem corresponds to the result of ( uncoded ) hard - decision based on the received vector .
now we proceed with the following algorithm : [ adaptive lp ] _ step 1 : setup the initial problem according to ( [ initial constraints ] ) .
_ _ step 2 : run the lp solver .
_ _ step 3 : search for all the cuts for the current solution .
_ _ step 4 : if one or more cuts are found , add them to the problem constraints and go to step 2 . _ if at any iteration of algorithm [ adaptive lp ] no cut is found , the current solution is the the solution of the lp decoder with all the relaxation constraints given in section ii .
the claim follows from the fact that if at any stage no cut is found , the current solution is in the fundamental polytope .
[ n iterations ] the adaptive algorithm ( algorithm [ adaptive lp ] ) converges with at most @xmath1 iterations .
the final solution is a vertex @xmath49 of the problem space determined by the initial constraints along with those added by the adaptive algorithm .
therefore , we can find @xmath1 such constraints , @xmath50 whose corresponding hyperplanes uniquely determine this vertex .
this means that if we set up an lp problem with only those @xmath1 constraints , the optimal point will be @xmath49 .
now , consider the @xmath51th intermediate solution , @xmath52 , that is cut off at the end of the @xmath51th iteration .
at least one of the constraints , @xmath53 , should be violated by @xmath52 , otherwise since @xmath52 has a lower cost than @xmath49 , @xmath52 would be the solution of lp with these @xmath1 constraints .
but we know that the cuts added at the @xmath51th iteration are all the possible constraints that are violated at @xmath52 .
consequently , at least one of the cuts added at each iteration should be among @xmath54 ; hence , the number of iterations is at most @xmath1 .
this theorem applies to any general lp problem where there is a fixed set of constraints , and at each iteration we add all the cuts generated by this set of constraints .
the adaptive algorithm has at most @xmath55 constraints at the final iteration
. follows from theorem [ one cut per check ] and theorem [ n iterations ] . for high - density codes of fixed rate ,
this bound guarantees convergence with @xmath56 constraints , whereas the standard lp and the polytope given in @xcite for high - density codes respectively require exponential and @xmath57 constraints . to observe the complexity reduction due to the adaptive approach for lp decoding ,
we have performed simulations over random regular ldpc codes of various lengths , degrees , and rates on the awgn channel .
all the experiments were performed with the low snr value of @xmath58 db , since in the high snr regime the recieved vector is likely to be close to a codeword , in which case the algorithm converges fast , rather than demonstrating its worst - case behavior . in the first scenario , we studied the effect of changing the check node degree @xmath17 from @xmath59 to @xmath60 while keeping the code length at @xmath61 and the rate at @xmath62 .
the simulation was performed over 400 blocks for each value of @xmath17 .
the maximum ( average ) number of iterations required to converge started from around @xmath63 ( @xmath64 ) for @xmath65 , and decreased monotonically down to @xmath66 ( @xmath67 ) for @xmath68 .
the average and maximum numbers of parity - check constraints in the final iteration of the algorithm are plotted in fig .
[ const vs d_c ] .
it is observed that both the average and the maximum values are almost constant , and remain below @xmath69 for all the values of @xmath17 . for comparison , the total number of constraints required for the standard ( non - adaptive ) lp decoding problem , which is equal to @xmath16 is also included in this figure .
the decrease in the number of required constraints translates to a large gain for the adaptive algorithm in terms of the running time .
this gain increases exponentially with @xmath17 , so that with the lp solver that we have used in our work ( glpk @xcite ) , the adaptive algorithm always converges several thousand times faster that standard lp for @xmath70 .
, for fixed length @xmath61 and rate @xmath62.,width=3 ] [ const vs d_c ] in the second case , we studied random ( 3,6 ) codes of lengths @xmath71 to @xmath72 .
for all values of @xmath1 , the maximum ( average ) number of required iterations remained between @xmath73 and @xmath74 ( @xmath75 and @xmath76 ) .
the average and maximum numbers of parity - check constraints in the final iteration are plotted versus @xmath1 in fig .
[ const vs n ] .
we observe that the number of constraints is generally between @xmath77 and @xmath78 . , for fixed rate @xmath62 and check node degree @xmath79.,width=3 ] [ const vs n ] in the third experiment
, we investigated the effect of the rate of the code on the performance of the algorithm .
[ const vs m ] shows the average and maximum numbers of parity - check constraints in the final iteration where the block length is @xmath80 and the number of parity checks , @xmath81 , increases from @xmath82 to @xmath83 .
the variable node degree is fixed at @xmath84 .
we see that the average and maximum numbers of constraints are respectively in the ranges @xmath85 to @xmath86 and @xmath87 to @xmath88 for most values of @xmath81 .
the relatively large drop of the average number for @xmath89 with respect to the linear curve can be explained by the fact that at this value of @xmath81 the rate of failure of lp decoding was less than @xmath90 at @xmath58 db , whereas for all the other values of @xmath81 , this rate was close to @xmath91 . since the success of lp decoding generally indicates proximity of the received vector to a codeword , we expect the number of parity checks required to converge to be small in such a case , which decreases the average number of constraints . , for @xmath80 and @xmath84.,width=3 ]
[ const vs m ] based on these simulation results , we observe that in practice the algorithm performs much faster than is guaranteed by theorem [ n iterations ] .
these observations suggest the following conjecture .
[ conjecture on complexity ] for a random parity - check code of length @xmath1 with @xmath92 parity checks and arbitrary degree distributions , as @xmath1 increases , the adaptive lp decoding algorithm converges with probability arbitrarily close to @xmath91 in at most @xmath93 iterations and with at most @xmath94 final parity - check constraints per check node , where @xmath93 and @xmath94 are constants independent of the length , rate and degree distribution of the code .
[ int positions in pcw ] if for a given code of length @xmath1 , the adaptive algorithm converges with at most @xmath95 final parity - check constraints , then each pseudo - codeword of this lp relaxation should have at least @xmath96 integer elements . to see this ,
note that each pseudo - codeword corresponds to the intersection of at least @xmath1 active constraints .
if the problem has at most @xmath97 parity - check constraints , then at least @xmath96 constraints of the form @xmath98 or @xmath99 should be active at each pseudo - codeword , which means that at least @xmath96 positions of the pseudo - codeword are integer - valued .
the complexity reduction obtained by adaptive lp decoding inspires the use of cutting - plane techniques to improve the error rate performance of the algorithm . specifically ,
when lp with all the original constraints gives a nonintegral solution , we try to cut the current solution , while keeping all the possible integral solutions ( codewords ) feasible . in the decoding problem
, the new cuts can be chosen from a pool of constraints describing a relaxation of the maximum - likelihood problem which is tighter than the fundamental polytope . in this sense ,
the cutting - plane technique is equivalent to the adaptive lp decoding of the previous section , with the difference that there are more constraints to choose from .
the effectiveness of this method depends on how closely the new relaxation approximates the ml decoding problem , and how efficiently we can search for those constraints that introduce cuts .
_ @xcite have mentioned some ways to tighten the relaxation of the ml decoding , including adding redundant parity checks ( rpc ) , and using lift - and - project methods .
( for more on lift - and - project , see @xcite and references therein . )
gomory s algorithm @xcite is also one of the most well - known techniques for general integer optimization problems , although it suffers from slow convergence .
each of these methods can be applied adaptively in the context of cutting - plane techniques .
the simple structure of rpcs makes them an interesting choice for generating cuts .
there are examples where even the relaxation obtained by adding all the possible rpc constraints does not guarantee convergence to a codeword . in other words
, it is possible that we obtain a nonintegral solution for which there is no rpc cut , although the general case is still not well - studied . also , finding efficient methods to search for rpc cuts for a given nonintegral solution is an open issue . on the other hand ,
as observed in simulation results , rpc cuts are generally strong , and a reasonable number of them makes the resulting lp relaxation tight enough to converge to an integer - valued solution . in this work ,
we focus on cutting - plane algorithms that use rpc cuts .
a redundant parity check is obtained by modulo-2 addition of some of the rows of the parity - check matrix , and this new check introduces a number of constraints that may include a cut .
there is an exponential number of rpcs that can be made this way , and in general , most of them do not introduce cuts . hence , we need to find the cuts efficiently by exploiting the particular structure of the decoding problem .
in particular , we observe that cycles in the graph have an important role in determining whether an rpc generates a cut .
this property is explained by theorem [ rpc cycles ] .
[ minimal subset ] given a current solution , @xmath13 , the subset @xmath100 of check node indices is called a _ cut - generating collection _ if the rpc made by modulo-2 addition of the parity checks corresponding to @xmath101 introduces a cut .
[ rpc cycles ] let @xmath102 be a collection of check node indices in the tanner graph of the code , and let @xmath103 be the subgraph made up of these check nodes , the variable nodes directly connected to them , and all the edges that connect them .
then , @xmath101 can be a cut - generating collection only if @xmath103 contains a cycle that only passes through variable nodes whose corresponding current values are fractional .
( proof omitted . )
this result motivates the following algorithm to search for cuts .
[ search for rpc cuts ] _ step 1 : having a solution @xmath13 , prune the tanner graph by removing all the variable nodes with integer values .
_ _ step 2 : starting from an arbitrary check node , randomly walk through the pruned graph until a cycle is found .
_ _ step 3 : create an rpc by combining the rows of the parity - check matrix corresponding to the check nodes in the cycle .
_ _ step 4 : if this rpc does not introduce a cut , go to step 2 . _ by exploiting some of the properties of the problem as described above , the search for each cut becomes much faster than the naive search ; however , this procedure is still the complexity bottle - neck of the decoding , and it becomes prohibitively complex for codes of length on the order of several hundred bits and above . on the other hand , experiments demonstrate that the number of cuts required to converge to the ml codeword ( if the convergence is possible ) does not grow very rapidly with length .
this motivates more study of the properties of rpc cuts and efficient schemes to find them .
.,width=3 ] [ rpc wer ] to demonstrate the performance improvement achieved by using the rpc cutting - plane technique , we present simulation results for a random regular @xmath104 ldpc code of length @xmath105 .
the adaptive nature of the algorithm allows us to smoothly trade complexity for performance by changing the number of trials in the search for each rpc cut , i.e. the number of iterations in algorithm [ search for rpc cuts ] , as well as the total number of calls of the lp solver . in this experiment
, we fix the total number of iterations of algorithm [ search for rpc cuts ] to @xmath106 , and declare decoding failure if no cut is found after @xmath106 trials .
the word error rate ( wer ) of the algorithm is plotted versus snr in fig .
[ rpc wer ] for different values of @xmath106 . for comparison ,
the wer of pure lp decoding , i.e. with no rpc cut , and a lower bound on the wer of the ml decoder have been included , as well . in order to obtain this lower bound , we count the number of times that the cutting - plane lp algorithm with a large value of @xmath106 converges to a codeword other than the transmitted codeword , and divide that by the number of blocks . due to the ml certificate property of lp decoding , we know that ml decoding would fail in those cases , as well . on the other hand ,
ml decoding may also fail in some of the cases where lp decoding does not converge to an integral solution .
therefore , this estimate gives a lower bound on the wer of ml decoding .
the numerical results suggest that the cutting - plane lp decoding with rpc cuts can significantly outperform the pure lp decoding , at the cost of increased complexity .
in this work , we studied the potential for improving lp decoding , both in complexity and error correction capability , by using an adaptive approach .
the key idea was to use the fact that we can always recognize the failure of lp decoding to find the ml codeword , a property that message - passing algorithms only have in specific cases such as the erasure channel .
this feature allows us to add only constraints that are `` useful '' , depending on the current status of the algorithm . in the algorithm proposed in section iii ,
the complexity is significantly reduced and becomes independent of the degree distributions , making it possible to apply lp decoding to parity - check codes of arbitrary densities . however , since general purpose lp solvers are used at each iteration , the complexity in terms of the block length is still super - linear , as opposed to linear as in the message - passing algorithms . an interesting question is whether we can design special lp solvers for decoding of ldpc codes that can take advantage of the sparsity of the constraints and other properties of the problem to converge in linear time .
section iv serves as a first step to explore the application of cutting - plane techniques in lp decoding .
a desirable feature of this approach is that by changing the parameters of the algorithm we can smoothly trade complexity for performance .
in contrast , if we want to get the same performance gains by tightening the relaxation in a non - adaptive setting , the required complexity increases much faster .
we showed that redundant parity checks provide strong cuts , even though they may not guarantee ml performance .
a major open problem is to find efficient ways to search for these cuts by exploiting their properties , and to determine specific classes of codes for which rpc cuts are more effective .
furthermore , the effectiveness of cuts generated by other techniques , such as lift - and - project cuts and gomory cuts , as well as specially designed cuts , needs further study . 1 j. feldman , m. j. wainwright , and d. karger , `` using linear programming to decode binary linear codes , '' _ ieee trans .
inform . theory _
954 - 972 , mar .
o. vontobel and r. koetter , `` on the relationship between linear programming decoding and min - sum algorithm decoding , '' _ proc .
intl symp . on inform .
theory and applications _ , pp .
991 - 996 , parma , italy , oct . 2004 .
o. vontobel and r. koetter , `` graph - cover decoding and finite - length analysis of message - passing iterative decoding of ldpc codes , '' _ sumbitted to ieee trans .
theory_. `` gnu linear programming kit , '' available at ` http://www.gnu.org/software/glpk ` .
m. laurent , `` a comparison of the sherali - adams , lovsz - schrijver and lasserre relaxations for 0 - 1 programming , '' _ mathematics of operations research , _ vol .
470 - 496 , 2003 .
r. gomory , `` an algorithm for integer solutions to linear programs , '' _ princeton ibm mathematical research report _
, nov . 1958 . | detectability of failures of linear programming ( lp ) decoding and its potential for improvement by adding new constraints motivate the use of an adaptive approach in selecting the constraints for the lp problem . in this paper
, we make a first step in studying this method , and show that it can significantly reduce the complexity of the problem , which was originally exponential in the maximum check - node degree .
we further show that adaptively adding new constraints , e.g. by combining parity checks , can provide large gains in the performance . |
the smallest molecular cores observed to date have at least @xmath06 orders of magnitude greater angular momentum per unit mass than the sun , suggesting that they would eventually greatly exceed the breakup velocity if no angular momentum was lost during the evolutionary process .
it is widely believed that most if not all low - mass stars form with circumstellar accretion disks ; accreting solar - like pre - main sequence ( pms ) stars have rotational velocities 1/5th to 1/10th of breakup speeds despite the angular momentum transported to the star via accretion from its circumstellar disk .
these fundamental observations lead to the conclusion that an angular momentum regulation mechanism must be at work during the accretion phase .
stellar winds alone as conventionally invoked can not explain such angular momentum loss unless they operate completely differently than expected ( see , e.g. , sills et al .
2000 and references therein ) ; these young solar - like pms stars are completely convective , so winds must slow the rotation of the whole star ( not just the outer layers ) , and the timescale for winds to significantly slow a star s rotation is longer than the evolutionary timescales involved here .
all recent thinking focuses on angular momentum transfer mechanisms intimately related to the accretion process itself .
these include ( a ) transfer of stellar angular momentum to a spreading accretion disk ( _ e.g. _ knigl 1991 , knigl & pudritz 2000 ) , and ( b ) transfer of angular momentum to an accretion - driven wind , launched either at the disk / magnetosphere boundary ( _ e.g. _ shu et al .
2000 ) or elsewhere in the star - accretion disk system ( matt & pudritz 2005 ) .
all of the above disk - related regulation mechanisms posit that an accretion disk is present , and that the central star is ` locked ' ( or nearly so ) to a constant angular velocity fixed by the keplerian speed at the radius at which the stellar magnetosphere and disk are linked ( co - rotation radius ) even as the pms stars contract . for more than 10 years (
including , e.g. , edwards et al . 1993 ) , observers have attempted to test the hypothesis that disk - related processes account for angular momentum regulation among solar - like pms stars .
the test is seemingly simple : compare the observed rotation periods or projected equatorial rotational velocities of pms stars surrounded by disks ( and thus , by hypothesis , regulated ) with those that lack disks and are presumed free to spin up in response to contraction toward the main sequence absent a regulating disk .
the results of these tests have , however , been mixed ; for example , rebull ( 2001 ) and makidon et al .
( 2004 ) find no clear correlation between disk excess and rotation rate in orion and ngc 2264 , respectively , but herbst et al .
( 2002 ) and lamm et al .
( 2004 ) find such a correlation in these same clusters .
moreover , herbst et al.(2002 ) find a weak correlation between the rotation rate and the _ size _ of the disk excess , which is not found by , e.g. , stassun et al .
( 1999 ) or rebull ( 2001 ) .
two key factors complicate this apparently simple test and may lead to ambiguous results : ( a ) the inaccuracies inherent in near - ir excesses , and ( b ) the influence of sample size .
near - ir excesses ( typically @xmath3-band ) , although the most widely available , are not sufficiently robust , owing to the presence of inner disk holes and projection effects ( see , e.g. , hillenbrand et al . 1998 ) and/or contrast between the disk and the photosphere and/or disk geometry ( e.g. , the size of the inner disk wall ) .
these young stars are variable , and use of near - ir indicators is often also compromised by the vexing problem of non - simultaneity in the photometric observations ( optical through infrared ) typically used to derive @xmath3-band excesses . a recent paper by rebull et al .
( 2004 ) identifies the second key factor as the sample size .
rotation periods among young pms stars surrounded by disks span a factor of 20 .
the broad period distribution among young pms stars means that extracting a ` signal ' of period changes between samples of stars surrounded by disks and those that lack disks requires a large sample before it is possible to distinguish ` locked ' from ` freely spinning ' objects .
the fact that near - ir excesses ` miss ' @xmath030% of stars which longer wavelength observation reveal to have clear disk signatures complicates separation of disked and non - disked samples a potentially fatal difficulty for small samples ( @xmath4 several hundred ) stars .
we now are in a position to overcome both of these problems . over the past decade ,
more than 1700 rotation periods have been measured for solar - like pms stars , @xmath0900 in orion alone . moreover ,
the advent of the spitzer space telescope ( werner et al .
2004 ) provides a powerful tool for robust identification of stars surrounded by disks , via observation of the irac channel 1 ( 3.6 @xmath1 ) @xmath5 irac channel 4 ( 8 @xmath1 ) color index an index both highly diagnostic of emission arising in circumstellar accretion disks , and little influenced by inclination effects , inner disk holes or ( because of the method of data acquisition ) non - simultaneity of photometric observations . in this contribution , we discuss the period distributions for a sample of @xmath0500 stars drawn from the orion region ( @xmath01 myr ) for stars with and without circumstellar disks as diagnosed by [ 3.6 @xmath1 ] @xmath5 [ 8 @xmath1 ] colors ( hereafter denoted @xmath7-[8]$ ] ) . to anticipate our conclusions , we find that only a small fraction of rapidly rotating stars exhibit robust disk signatures , while the fraction of disks surrounding slowly rotating stars is much greater ( with a high degree of statistical significance ) .
we find in these spitzer data the clearest indication to date that stars with longer periods are more likely than those with short periods to have ir excesses suggestive of disks .
megeath et al . ( 2006 , in prep ) surveyed the orion molecular clouds ( @xmath01 myr ) with the irac instrument ( 3.6 , 4.5 , 5.8 , & 8 @xmath1 ; fazio et al . 2004 ) as part of a larger joint irac and mips gto investigation .
the complete irac survey covers 6.8 square degrees ( 4.4 square degrees in orion a and 2.4 square degrees in orion b ) , including the region near the trapezium ( see figure [ fig : where ] ) but also continuing along the molecular cloud to the southeast .
the `` traditional '' orion nebula cluster ( onc , @xmath00.5 myr ) is a @xmath8 region ( see fig . [
fig : where ] ) , including stars with periods found by herbst et al .
( 2000 , 2002 ) , stassun et al .
( 1999 ) , and carpenter et al .
the flanking fields ( ff ; @xmath01 myr ; rebull 2001 and rebull et al .
2000 ) are four fields to the n , s , e , and w of the onc , each of which is @xmath00.75@xmath9 on a side ( see fig .
[ fig : where ] ) . stars in the ff are slightly older than those in the onc ( ramirez et al .
periods for stars here come from rebull ( 2001 ) , stassun et al .
( 1999 ) and carpenter et al .
the irac survey follows the molecular cloud , even through the onc and ff region , so it includes most of field 2 , to the north of the onc , and all of field 4 , to the south of the onc .
less than a third of fields 1 and 3 are included in the irac survey .
see megeath et al .
( 2006 , in prep ) for more extensive details on data acquisition and reduction for the orion data . in summary , toward each region , four 12-second high - dynamic - range ( hdr ) frames were obtained .
the individual basic calibrated data ( bcd ) frames ( ver .
s11 ) were mosaicked together using the mopex software .
the sources were identified with a custom source finder written in idl and then measured with the idlphot package .
an aperture of 2.4@xmath10 and a sky radius of 2.4@xmath10 to 7.2@xmath10 were used .
the adopted zero points in units of dn / sec / pixel are 19.66 , 18.94 , 16.88 and 17.39 for a 10 pixel aperture radius in irac channels 1 through 4 respectively .
aperture corrections of 1.213,1.234,1.379 and 1.584 were applied to the data .
table [ tab : data ] contains irac fluxes for those stars with periods detected in both irac-1 and irac-4 ( see below ) . for stars in the region covered by irac observations ,
there are 854 stars with known periods .
the catalog of stars with periods were merged with a catalog of irac sources identified by an automated source finder developed for this program ( megeath et al .
2006 ) . in irac-1 ( 3.6 @xmath1 ) , 90% of those stars are detected , 96% are detected in irac-2 ( 4.5 @xmath1 ) , 81% are detected in irac-3 ( 5.8 @xmath1 ) , and 59% are detected in irac-4 ( 8 @xmath1 ) .
( we note for completeness that the mips 24 @xmath1counterpart to this gto survey retrieves only 13% of the stars with known periods ; for this reason , the mips data are not discussed here . )
because the measurement of periods necessarily implies optical ( or nir ) detection of the stellar photosphere , by examining only stars with measured periods , we have already effectively selected just the less - embedded stars that are brighter in irac wavelengths .
the entire irac survey completeness limit in this region ranges from 3 to 7 magnitudes deeper than that of the periodic sample , so without reference to backgrounds , all stars with periods should be detectable in the irac survey .
however , in reality the nebular background and source confusion increases closer to the trapezium .
the majority of periodic stars are detected , but those periodic stars missing irac detections come from the region closest to the trapezium , making the detection and identification of even some bright sources less reliable . the nebular background increases with wavelength , so a larger fraction of stars are missing 8 @xmath1 detections than 3.6 @xmath1 detections . since the onc is slightly younger than the surrounding regions ( ramirez et al .
2004 ) , this effectively means that the sample of stars with periods and detections at the longer wavelengths will be slightly skewed to older stars .
color - color plots like those appearing in megeath et al .
( 2006 , in prep ) for this sample show clear populations of disk candidates and non - disk candidates ; see , e.g. , allen et al.(2004 ) for more discussion on interpretation of irac colors .
we considered several different possible disk indicators to use here , but the separation between the disk candidates and non - disked candidates was most apparent when using a larger difference in wavelength . despite having fewer stars to consider when using the 8 @xmath1 detections , the conclusions of this analysis do not change significantly when
other disk indicators are considered .
therefore , we will use @xmath7-[8]$ ] measurements to identify disks , where we have set ( based on histograms of excesses ; see below ) the boundary between disk and non - disk candidates at @xmath7-[8]$]=1 magnitude . requiring both 3.6 and 8 @xmath1 detections
reduces the 854 available periods down to 464 stars , where the stars are preferentially lost from the core of the onc ( where the nebular emission and source confusion are both high ) .
these 464 stars appear in table [ tab : data ] .
none of our conclusions change significantly when considering other disk indicators involving at least one irac band ( such as the longer lever arm accorded by @xmath11 $ ] , or the additional stars provided by @xmath7-[5.8]$ ] ) . in the past
, it has been assumed that the distribution of measured periods is representative of the larger distribution of rotation rates .
samples of stars with and without measured periods have similar @xmath12 distributions ( e.g. , rhode et al.2001 ) .
however , recently , the sample of stars with measured periods has been shown to be significantly brighter in x - rays than the sample of stars without measured periods ( e.g. , rebull et al .
2006 in press ) , suggesting that there may be real differences in the two samples .
we looked for any difference in irac properties between those stars with measured periods and those stars without measured periods . in order to assemble a fair comparison , we only considered objects ( either with measured periods or without ) also detected in @xmath13 ( values taken from the literature ) and those located in a clear `` locus '' above the zero - age main sequence ( zams ) in the @xmath13/@xmath14color - magnitude diagram ( see rebull et al .
2000 for much more discussion , including cmds ) .
figure [ fig : all14_histo ] shows normalized histograms of the @xmath7-[8]$ ] colors for stars with measured periods and those otherwise similar stars that are likely cluster members but without measured periods . there is a clear separation of the disk candidates above @xmath7-[8]$]=1 and the non - disk candidates below that level
this demonstrates vividly the advantage of using mir disk indicators ; this effect is not found so prominently using other ( e.g. , nir ) disk indicators , which indeed has complicated earlier searches for disks .
we note that the separation of the disk candidates at @xmath7-[8]$]=1 is slightly clearer for the stars with periods than the stars without measured periods ; we suspect that this is because the sample of stars without measured periods are more likely to have non - cluster member interlopers .
we have normalized the histograms in the figure because of the vastly different numbers of stars ( 1014 without periods and 387 with periods ) . a kolmogorov - smirnoff ( k - s ) test on these ( un - normalized ) distributions reveal that they are statistically indistinguishable , and we conclude that the irac disk properties of the stars with and without measured periods are statistically indistinguishable .
figure [ fig : p14 ] plots @xmath7-[8]$ ] vs. period for all stars throughout the orion region .
there is still a clear separation of the disk candidates above @xmath7-[8]$]=1 and the non - disk candidates below that level . by examining the cumulative distribution function of all @xmath7-[8]$ ] colors
, we find an inflection in the curve near @xmath7-[8]$]=1 , verifying what we see by eye in the distribution of colors in fig .
[ fig : p14 ] ( or fig .
[ fig : all14_histo ] ) .
it is similarly already apparent that the distribution of @xmath7-[8]$ ] is different for shorter than it is for longer periods . by examining the cumulative distribution function of periods for all stars with
@xmath7-[8]>1 $ ] , there is an inflection point near @xmath151.8 d ( log @xmath150.25 ) .
figure [ fig : p14_histos ] shows the histograms of @xmath7-[8]$ ] , distinguishing @xmath161.8 d ( log @xmath160.25 ) , and @xmath171.8 d. clearly , a long - period star is more likely to have a disk than a short - period star , but there are also substantial numbers of long - period stars with little or no excess .
these latter stars may have just recently cleared their disks and have not yet spun up in response to contraction on their way to the zams .
figure [ fig : cumudist_14 ] shows the cumulative distributions of [ 3.6]@xmath5[8 ] for the samples with @xmath17 and @xmath41.8 d portrayed in fig . [ fig : p14 ] .
( the inflection point mentioned above near @xmath7-[8]=1 $ ] is readily apparent . )
a k - s test on these distributions reveal that they are statistically substantially different ( @xmath41 chance in @xmath18 that they are drawn from the same population ) .
the distributions of @xmath19 for samples with [ 3.6]@xmath5[8]@xmath20 and @xmath41 are also statistically different .
there are 11 fast - rotating stars that appear to have disks . of these 11 stars , one ( jw 783 )
is a likely double in the irac images , and one ( par 2097 ) falls in a particularly bright region of the nebula ( and therefore is likely to have more inaccurate 8 @xmath1 fluxes than typical ) .
the remaining 9 stars probably have reasonable measurements at 3.6 and 8 @xmath1 , despite many of them having some nebular background at 8 @xmath1 .
perhaps these fast rotating stars with disks have just recently experienced an accretion event , or they are descendents of ex ori - type stars .
these 9 stars are good targets for followup studies , and they are par 1409 , par 1799 , par 1938 , par 2123 , jw 104 , jw 879 ( a likely non - member based on proper motions ) , numbers 1053 and 3115 from hillenbrand ( 1997 ) , and number 864 from rebull ( 2001 ) .
we expected that there might be differences in this relation between @xmath19 and ir excess as a function of location within the cluster ( onc vs. ff ) .
there are differences between the onc and the ff , but the differences are all in the relative disk fractions .
the disk fraction is much higher in the onc , and although this could be a result of the fact that the onc is slightly younger than the ff ( ramirez et al .
2004 ) , this is likely also a result of the bias against faint 8 @xmath1detections of stars without bright infrared excesses in the onc imposed by the weaker photospheric emission and substantially higher background seen at longer wavelengths . as a function of physical location , there is no distinguishable difference of the clear separation of disk candidates from non - disk candidates , or of the tendency for fast rotators to lack disks .
we expected that there might be an effect of mass or age .
we derived masses and ages via placement in the @xmath21 color - magnitude diagram and via dantona and mazzitelli ( 1994,1998 ; dam ) models , as in rebull ( 2001 ) , with conversion from the theoretical to the observational plane using transformations found in hillenbrand ( 1997 ) .
( we note that of course masses derived in this fashion are model - dependent .
we could have investigated mass effects via the use of spectral types , but types only exist for 50% of our sample . in lieu of cutting the sample down by half ,
we have opted instead to proceed using masses so derived . )
previous investigators ( e.g. , herbst et al .
2002 ) report different period distributions for stars above and below a mass division of 0.25 m@xmath2 , also using dam models .
therefore , we also investigated separating the sample at this boundary , corresponding roughly to a spectral type of m2.5-m3 , or a mass of @xmath00.35 m@xmath2 using siess et al .
( 2000 ) models ( the z=0.02 model with no overshooting , using conversions from kenyon & hartmann 1995 ) , or roughly @xmath00.5 m@xmath2 using baraffe et al .
( 1998 ) models ( z=0 , he mass fraction=0.275 , mixing length=1.0 , conversions as presented therein ) .
as seen in figure [ fig : mass ] , stars with @xmath220.25 m@xmath2have a higher disk fraction , and the separation between the disk and non - disk candidates , while still readily apparent , is less sharp ; e.g. , there are fractionally more lower - mass stars with @xmath7-[8]$ ] values near 1 .
there is more scatter in the lower - mass stars ; the errors in periods are comparably small for all masses , but the errors in irac color are likely to be greater for the fainter stars .
the fastest rotators still tend to lack disks .
a two - dimensional k - s test indicates that there is @xmath41 chance in @xmath23 that these distributions were drawn from the same population .
there is a well - defined clump of higher - mass stars with ir excesses between log @xmath150.6 and 1 , which does not appear with such clarity in the lower - mass stars .
mean _ period for the distribution of stars with disks is identical within the scatter for both mass ranges ( log @[email protected] for the more massive stars and [email protected] for the less massive ) .
the _ modal _ ( or most likely ) period of these distributions of stars with disks are slightly different ; for the more massive stars , the mode of the distribution is log @xmath19=0.9 ( with 33% of the sample having log @[email protected] ) and 0.6 for the less massive stars ( this distribution is broader , so although the most frequent value is 0.6 , only 15% of the sample has log @[email protected] ) .
the higher mass `` clump '' suggests consistency with taurus ( edwards et al .
1993 ) in the sense that the taurus sample was dominated by k7-m0 stars , as is the left side of fig .
[ fig : mass ] .
we can speculate on many possible explanations for this .
it may be that disk locking is more efficient in the more massive stars , or that the parameters ( @xmath25 , @xmath26 ) relevant for disk locking found in these systems are more similar across stars in this mass range , such that the locking occurs at similar radii and periods for a larger fraction of the sample in the higher mass stars . as a function of age within this sample
, there is no distinguishable difference in the clear separation of disk candidates from non - disk candidates , or in the tendency for fast rotators to lack disks .
since younger stars may have a higher disk fraction , there is the expectation that there may also be some subtle second - order effects from these influences , but it is still true , regardless of the subsample , that faster rotators are less likely to have disks .
similar results can be derived from the independently obtained distributions of @xmath12 in this region ; see figure [ fig : vsini ] .
the relation is not nearly as clear here , due primarily to the fundamental resolution limits imposed by spectroscopy at both the fast - rotating ( due to smearing out of the absorption lines in these late - type stars ) and slow - rotating ( instrumental resolution is usually 10 - 12 km s@xmath27 ) ends of the distribution .
nonetheless , stars that are more rapidly rotating are still less likely to have disks .
the fact that this effect is also seen in the independent @xmath12sample is powerful confirmation ; of the 296 stars with measured @xmath12 and irac fluxes , only 154 of them also have measured periods .
basic conservation of angular momentum during the star formation process requires that the angular momentum of the central object must be regulated .
recent empirical studies ( rebull et al.2004 , herbst et al .
2005 ) have found that young low - mass pms objects must be draining angular momentum into their surroundings , but it has not been clear , until now , that the disks were the culprit .
the sample discussed here provides two significant advantages above and beyond previous studies : ( 1 ) the disk indicator we use here is robust against many of the problems plaguing shorter - wavelength disk indicators , and ( 2 ) the sample size is large , nearly 500 stars .
we have finally resolved two of the most important issues complicating earlier similar investigations .
we find a clear correlation between rotation rate and circumstellar disks . among the slower rotators ( stars with periods @xmath20 1.8d ) , the period distributions for stars with and without disks ( [ 3.6]@xmath5[8]@xmath20 and @xmath41 ) are statistically indistinguishable .
edwards et al .
( 1993 ) made a similar observation ( with many fewer objects ) in their early discussion of disk - locking , and speculated that the long - period non - disked objects have been released recently ( within a few hundred thousand years ) .
if this speculation is correct , then examination of the period distributions among older star - forming regions should reveal a significant separation in the period distributions betwen disked and non - disked stars .
the evolutionary timescales here may be important , as suggested by hartmann ( 2002 ) ; the timescales for angular momentum loss may be comparable to the evolutionary and/or disk clearing timescales .
especially if these stars have recently cleared their disks , it is possible that they are still behaving as disk - locked ; we could have disk _ braking _ without disk _
locking_. if this is the case , then the timescales for disk clearing are indeed comparable to the evolutionary timescales here . another possible explanation , as long et al .
( 2005 ) have suggested , is that combinations of parameters ( @xmath25 , @xmath26 ) could result in disk locking at different rotation rates for otherwise similar stars .
finally , it could be important to consider if these stars undergo intermittent periods of spinup and/or rapid accretion ( e.g. , lovelace et al .
1995 , li 1996 , kuker et al .
2003 , long et al.2005 , matt & pudritz 2005b ) , not just in terms of whether we have caught some stars in the act of spinning up , but also specifically the implications for disk luminosity in the irac bands during periods of high accretion rate . given the observed correlation , disks are the most obvious solution , but it is possible that there are other explanations . if the rapid rotators are more likely to be in close binary systems , then this correlation could be explicable in terms of disk dissipation by the companion and/or source confusion in determining the irac colors .
followup high - resolution spectroscopy of the fast rotators looking for indications of binarity will be required to resolve this issue .
the remaining issues to pursue in similar investigations include sampling ( a ) a broader range of environmental settings , e.g. , regions further from ( or lacking entirely ) o and b stars , and ( b ) a broader range of ages .
if the characteristics ( e.g. , ionization fraction , @xmath25 , lifetime ) of the disks around young stars are intimately tied to the net time spent near o and b stars , the distribution found in orion may be significantly different from those found in other clusters lacking high - mass stars . if the observations here in orion are representative of the characteristics of other young low - mass systems , we expect that the difference in period distributions of stars with and without mir excesses will increase with time , as suggested by the monte carlo models discussed in rebull et al .
( 2004 ) . in older clusters
, there should be a lower fraction of slowly rotating stars lacking disks , since those stars that may be `` just released '' here should have spun up in older clusters . in order to constrain the timescale for disk locking ( and understand whether or not there is an effect with cluster age ) by comparing orion to other clusters ,
ideally , we want to minimize errors by comparing irac excess distributions in other clusters , not those of other disk indicators .
additional clusters with substantial numbers of measured periods and spitzer observations are ngc 2264 ( @xmath03 myr ) , ic 348 ( @xmath03 myr ) and taurus - auriga ( @xmath01 myr ) . l.m.r . wishes to acknowledge many helpful conversations with s.e .
strom and s. c. wolff .
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 .
this research has made extensive use of nasa s astrophysics data system abstract service .
the research described in this paper was partially carried out at the jet propulsion laboratory , california institute of technology , under a contract with the national aeronautics and space administration .
allen , l. , et al.2004 , apjs , 154 , 363 baraffe , i. , et al.1998 , a&a , 337 , 403 carpenter , j. m. , hillenbrand , l.a . , & skrutskie , m. , f. 2001 , aj , 121 , 3160 dantona , f. , & mazzitelli , i. 1994 , , 90 , 467 ( dam ) dantona , f. , & mazzitelli , i. 1998 , http://www.mporzio.astro.it/@xmath0dantona/prems.html ( dam ) fazio , g. , et al .
, 2004 , apjs , 154 , 10 hartmann , l. , 2002 , apjl , 566 , 29 herbst , w. , rhode , k. l. , hillenbrand , l. a. , & curran , g. 2000 , , 119 , 261 herbst , w. , et al .
2002 , , 396 , 513 herbst , w. , bailer - jones , c. a. l. , & mundt , r. 2001 , apj , 554 , 197 herbig g. h. , & bell k. r. , 1988 , lick observatory bull .
1111 hillenbrand , l. , 1997 , aj , 113 , 1733 hillenbrand , l. , 1998 , aj , 116 , 1816 jones , b. f. , & walker , m. f. , 1988 , , 95 , 1755 kenyon , s. , & hartmann , l. , 1995 , apjs , 101 , 117 knigl , a. 1991 , , 370 , 39 knigl , a. & pudritz , r. 2000 , in protostars and planets iv , ed .
v. mannings , a. p. boss & s. s.russell ( tucson : university of arizona press ) , 759 edwards , s. , et al .
1993 , aj , 106 , 372 kuker , m. , et al .
2003 , , 589 , 397 li , j. , 1996 , , 456 , 696 long , m. , et al .
2005 , , 634 , 1214 lovelace , r. , et al .
1995 , mnras , 275 , 244 matt , s. , & pudritz , r. , 2005 , apjl , 632 , 135 matt , s. , & pudritz , r. , 2005 , mnras , 356 , 167 megeath , t. , et al . , 2006 , in prep parenago , p. p. , 1954
, trudy sternberg astron.inst . , 25 , 1 ramirez , s. , rebull , l. , stauffer , j. , strom , s. , hillenbrand , l. , hearty , t. , kopan , e. , pravdo , s. , makidon , r. , jones , b. , 2004 , aj , 128 , 787 rebull , l. , hillenbrand , l. a. , strom , s. e. , duncan , d. k. , patten , b. m. , pavlovsky , c. m. , makidon , r. b. , & adams , m. t. 2000 , , 119 , 3026 rebull , l. , et al . , 2001 ,
aj , 121 , 1676 rebull , l. , wolff , s. , strom , s , & makidon , r. b. 2002b , aj , 124 , 546 rebull , l. , wolff , s. , & strom , s. e. , 2004 , aj , 127 , 1029 rhode , k. , herbst , w. , & mathieu , r. 2001 , , 122 , 3258 siess , l. , dufour , e. , & forestini , m.2000 , , 358 , 593 , and @xmath4http://www - laog.obs.ujf - grenoble.fr / activites / starevol / evol.html@xmath20 ( sdf ) sills , a. , pinsonneault , m. , terndrup , d. , 2000 , , 534 , 335 shu , f. , najita , j. , shang , h. , & li , z .- y .
2000 , in protostars and planets iv , ed . v. mannings , a. p. boss & s. s. russell ( tucson : university of arizona press ) , p. 789 stassun , k. g. , mathieu , r. d. , mazeh , t. , & vrba , f. j. 1999 , , 117 , 2941 werner , m. , et al . , 2004 , apjs , 154 , 1 | early observations of t tauri stars suggested that stars with evidence of circumstellar accretion disks rotated slower than stars without such evidence , but more recent results are not as clear . near - ir circumstellar disk indicators , though the most widely available , are subject to uncertainties that can result from inner disk holes and/or the system inclination .
mid - infrared observations are less sensitive to such effects , but until now , these observations have been difficult to obtain .
the spitzer space telescope now easily enables mid - infrared measurements of large samples of pms stars covering a broad mass range in nearby star - forming regions .
megeath and collaborators surveyed the orion molecular clouds ( @xmath01 myr ) with the irac instrument ( 3.6 , 4.5 , 5.8 , 8 @xmath1 ) as part of a joint irac and mips gto program .
we examine the relationship between rotation and spitzer mid - ir fluxes for @xmath0900 stars in orion for stars between 3 and 0.1 m@xmath2 .
we find in these spitzer data the clearest indication to date that stars with longer periods are more likely than those with short periods to have ir excesses suggestive of disks . |
soft x - ray transients ( sxts ) are a subset of low - mass x - ray binaries ( lmxb ) that display episodic , dramatic x - ray and optical outbursts , which usually last for several months . in the interim
the sxts are in a state of quiescence during which the optical emission is dominated by the luminosity of the faint companion star @xcite . in quiescence the optical lightcurves exhibit the classical double - humped ellipsoidal modulation , which is due to the differing aspects that the tidally distorted secondary star presents to the observer throughout its orbit ( e.g. see @xcite ; @xcite ; @xcite and @xcite ) .
the lmxb sxt prototype , a062000 ( = v616mon ) was discovered in 1975 by the ariel-5 satellite @xcite during an x - ray outburst whose peak flux made it the brightest nonsolar source ever seen .
the optical counterpart was subsequently identified @xcite and after the system had faded back to its quiescent level , a062000 was found to be a 7.8hr binary system containing a k - type secondary star and a non - stellar continuum source attributed to an accretion disc ( @xcite ; @xcite ) .
a radial velocity study led to the discovery of the first black hole primary in an x - ray transient @xcite .
thereafter , many uv / optical / infrared studies of a062000 in quiescence have been undertaken ( @xcite ; @xcite ; shahbaz , naylor & charles 1994 ; mcclintock , horne & remillard 1995 ; @xcite ; @xcite ; gelino , harrison & orosz 2001 ) , that have enabled the system parameters to be determined .
the k3k4v secondary star has a radial velocity semi - amplitude of 433@xmath13@xmath3which gives a mass function of [email protected]@xmath4@xcite , the inclination is @xmath5=41@xmath13@xmath6 ( @xcite ; @xcite ) , the binary mass ratio @xmath7 is(@xmath8/@xmath9)=0.067 ( @xcite ; where @xmath9 and @xmath8 are the masses of the compact object and secondary star respectively ) and @[email protected]@xmath4(@xcite ; @xcite ) .
the high quiescent x - ray luminosity of a062000 compared to other black hole binaries indicates that it is not completely dormant @xcite , but that there is still some continuing accretion , although at a very low rate .
superposed on the ellipsoidal modulation of the secondary star are many rapid flares on timescales of tens of minutes or less ( @xcite ; @xcite ; @xcite ) , which appear to be a common feature in quiescent black hole and neutron star x - ray transients @xcite .
the results obtained from timing analyses alone have not proven conclusive and the origin of the variability still remains uncertain .
the most likely explanations are magnetic reconnection events in the disc , optical emission from an advective region , reprocessed x - ray variability and flickering from the accretion stream impact point @xcite .
a spectrophotometric study of the flares in v404cyg showed that the continuum and h@xmath0 emission line were at times correlated @xcite .
the kinematics of the h@xmath0 flare events suggest that the whole line profile participates in the flare , and can be explained if the optical flare is induced by variable photoionisation from the x - ray source .
here we report on our time - resolved optical spectrophotometric study of a062000 in quiescence .
we study the balmer line and continuum short - term variations and examine the spectrum of the variability .
we infer the physical parameters and origin of the short - term flare .
finally the low resolution spectra provides us with the kinematic resolution to study the emission line variations using the method of doppler tomography .
lll telescope & date & ut range + vlt & 2003 january 7 & 01:2308:27 + wht & 2001 december 1/2 & 22:5606:37 + wht & 2001 december 2/3 & 21:5006:59 + wht & 2001 december 4 & 00:5404:15 +
time resolved spectroscopic observations of a062000 were obtained on jan 7 2003 using eso s very large telescope ( vlt ) unit telescope 1 equipped with the focal reducer low dispersion spectrograph fors1 .
a log of the observations are given in table1 .
the data were taken with the 600v grism with an order - sorting filter ( gc435 + 31 ) centred at 6270 and integration times of 120s . to maximise spectrophotometric accuracy
a wide slit was used , and as a result the spectral resolution was determined by the seeing ( @xmath100.8for the entire night ) .
we obtained 32 spectra with a slit width of 2.5and and 96 spectra with a slit width of 2.0 .
the wavelength coverage and dispersion depends on the slit width , and we measured from the arc lines coverages and dispersions of 4933 - 7329at 1.17 per pixel for the 2.5slit and 4452 - 6794at 1.15 per pixel for the 2.0slit .
the corresponding effective resolutions set by the seeing were about 4.7 and 4.6 for the 2.5and 2.0wide slits , respectively .
although the atmospheric dispersion corrector was used , in order to minimise atmospheric dispersion along the slit when observations were taken at high airmass , the slit was aligned to a position angle of 131.7degrees east of north .
the crowded field of view of a062000 is such that we were able to center a non - variable comparison star ( 38north 42west of a062000 ) on the slit for slit light - loss corrections . finally , since the observing conditions were excellent , the spectrophotometric flux standard hz2 @xcite was observed .
the data were debiased and corrected for the pixel - to - pixel variations using the bias frames and tungsten lamp flat - fields respectively , taken as part of the daytime calibrations .
the sky was subtracted by fitting second - order polynomials in the spatial direction to the sky regions on either side of the object .
the spectra of a062000 and the comparison star were then optimally extracted @xcite with arc spectra extracted at the same location on the detector as the targets . the wavelength scale was determined through a fourth - order polynomial fit to more than 25 arc lines giving a root - mean square error of @xmath11 for all the spectra .
some residual flexure is to be expected , and the additional use of a wide slit means that there is some uncertainty due to the position of the centre of light of the target .
the latter term means that it is impossible to use the night sky lines to correct for this , so the comparison star s h@xmath0 absorption line was used .
the relative velocity shift of each comparison star spectrum with respect to the first spectrum was computed using the method of cross - correlation .
the velocity shifts were then applied to a062000 and the comparison star spectra .
finally the velocity shift of the first comparison star , computed using the position of the h@xmath0 absorption line , was added to the a062000 velocities .
we corrected the spectra for the instrumental response and determined the slit - losses to obtain absolute flux calibrated spectra .
a third order polynomial fit to the continuum of the spectrophotometric standard was used to remove the large - scale variation of the instrumental response and the slit - loss corrections were calculated using the comparison star spectra .
finally we use the colour excess of @[email protected] @xcite and the extinction law of @xcite to correct the spectra of a062000 for interstellar reddening assuming @xmath13=3.1 .
this is the starting point for the analysis of the spectra presented hereafter .
spectrophotometric observations of a062000 were also obtained with the william herschel telescope ( wht ) on 2001 december 14 using the isis dual - beam spectrograph in single red arm mode to maximise throughput .
the observations used the tek4 ccd and r158r grating yielding a dispersion of 2.9 per unbinned pixel .
all observations were either unbinned or we used @xmath14 binning .
the exposure time used was 120sec .
a wide slit of 4arcsec rotated to include a comparison star ( which was not the same as the comparison star used at the vlt ) was used to ensure spectrophotometric accuracy , so the actual spectral resolution was set by the seeing . this was 1.92.4arcsec on the first night , 1.01.7arcsec on the second night , and about 1.0arcsec on the third night , corresponding to effective resolutions of 1520 , 814 , and about 8 respectively .
a log of the observations are given in table1 .
all images were debiased and flat - fielded using standard iraf techniques .
one - dimensional spectra were then extracted of both a062000 and an on - slit comparison star using the optimal extraction method of @xcite .
variations in transparency and wavelength dependent slit - losses were corrected by fitting a smooth function to the ratio of the individual comparison spectra to their average .
wavelength calibration was done by interpolating from a series of cune+cuar arc spectra obtained at intervals of approximately an hour during each night . as the comparison star did not have strong enough absorption lines , further correction for flexure
was not performed . since the velocity resolution is low anyway ( 350900kms@xmath15 ) we decided to discard the detailed velocity information and use the h@xmath0 emission in a062000 to calculate wavelength corrections , and simply work with integrated line fluxes .
all three nights were non - photometric , so no attempt was made to apply absolute flux calibration and all fluxes are expressed relative units .
simultaneous photometry was obtained with the site2 ccd camera on the jacobus kapteyn telescope ( jkt ) .
an @xmath16 band filter was used with 60s exposures .
standard iraf data reduction techniques were used to process the images , and differential photometry was performed using the optimal photometry algorithm of @xcite . .
[ fig : flarepos ]
since the use of a wide slit adds uncertainty to the wavelength calibration , we only use the vlt to determine the radial velocity curve , as no appropriate correction could be performed for the wht data .
we took the vlt data and used the cross correlation method of @xcite to determine radial velocities of absorption lines pertaining to the secondary star . prior to cross - correlation the spectra were interpolated onto a constant velocity scale ( 60 @xmath3pixel@xmath15 ) and normalised by fitting a spline function to the continuum .
only regions of the spectrum devoid of emission lines ( 50005820 and 59006500 ) were used in the analysis .
the balmer emission lines ( h@xmath2 and h@xmath0 ) , hei ( 5875 ) and interstellar naid lines were masked out so that they did not affect the cross - correlation process .
we use the comparison star as a template , since its spectrum contains features of a late - type k - star . the resulting radial velocity curve
was then fitted with a sinusoid with the orbital period fixed at @xmath17=0.323016d .
we determine the radial velocity semi - amplitude @xmath18 to be [email protected]@xmath3 , the systemic velocity @xmath19=10@xmath12@xmath3and the time at orbital phase 0.0 , @xmath20 to be [email protected] ; phase 0.0 corresponds to the inferior conjunction of the secondary star .
the reduced @xmath21 of the fit was 2.7 and the uncertainties quoted are 1-@xmath22 and have been rescaled so that the reduced @xmath21 of the fit is 1 .
it should be noted that since we used the position of the comparison star s h@xmath0 absorption line to determine the velocity zero point of the instrumental flexure , the radial velocities are relative to the comparison star .
furthermore , given the uncertainties in wavelength calibration due to the large slit width , the value for @xmath19 should not be taken at face value . the radial velocity curve and sinusoidal
fits are shown in figure[fig : rvcurve ] .
the scatter in the radial velocity curve is caused by the uncertain wavelength calibration , due to the use of a wide slit .
we also used a template star spectrum obtained by shifting all the spectra of a062000 into the rest frame of the secondary star , using the radial velocity solution derived by @xcite .
the results obtained are the same as those obtained using the comparison star as a template .
the amplitude we obtain for @xmath18 is less than that obtained by @xcite .
this is because our coverage of the minimum in the radial velocity curve at phase 0.25 is not complete and so the derived value for @xmath18 is biased . by fitting the radial velocity curve with @xmath18 fixed at 432@xmath3 @xcite we obtain @[email protected] .
it should be noted that the @xmath20 we obtain is different by 0.43 phase to that quoted by @xcite , even though the nominal definition of orbital phase 0.0 adopted is the same .
this may have arisen from the potential ambiguity of defining phase 0.0 photometrically , as the minima at phase 0.0 and 0.5 can be almost the same depth and hence can be confused .
no such ambiguity exists with our spectroscopic ephemerides .
in figures[fig : vlt_lcurves ] and [ fig : bcurves ] we present the vlt continuum and emission line lightcurves as a function of orbital phase .
the continuum lightcurve was obtained by summing the flux across the 50005820 and 59006500regions .
the continuum lightcurve exhibits rapid short - term variations as seen previously by @xcite , @xcite and @xcite . to determine the flare lightcurve we subtract a fit to the lower envelope of the lightcurve .
we use an iterative rejection scheme to fit the lower envelope , similar to @xcite and @xcite . we reject points more than 2-@xmath23above the fit , then refit , repeating the procedure until no new points are rejected .
most noticeable are the brief flares at orbital phase 1.15 and 1.2 which have a rise time of @xmath2415min and amplitudes of 3 and 12 percent respectively .
a larger flare is also seen at phase 0.6 which has a rise time of @xmath2430min and an amplitude of @xmath2420 percent . in figure[fig : flarepos ] we have marked the beginning and end of the strongest most significant flare events and also the periods when flares are not seen . to extract the balmer line fluxes ,
the continuum background was removed by fitting and subtracting a smooth polynomial fit to the continuum regions around the lines before the line flux was integrated .
the balmer line lightcurves show significant variations , scatters of 3.8 and 7.6 percent for the h@xmath0 and h@xmath2 lightcurves respectively .
the h@xmath2 and h@xmath0 lightcurves are strongly correlated .
figure[fig : bcurves ] shows the h@xmath0 and h@xmath2 emission line wings and core determined by integrating the line flux .
the blue , core , and red wing fluxes were calculated by integrating the emission line over the velocity ranges 2500 to 500@xmath3 , -500 to 500@xmath3and 500 to 2000 @xmath3respectively .
one can clearly see that the h@xmath0 and h@xmath2 emission line wings and core are correlated with each other .
the balmer line lightcurves show some correlations with the flare lightcurve ( figure[fig : vlt_lcurves ] ) .
this is clearest for the flare event at phase 1.15 .
there is no noticeable participation from h@xmath0 , although the correlation is quite strong in h@xmath2 .
however , it could be that there is an underlying source of variable contamination in the h@xmath0 lightcurves such as from the disc ( see section[doppler ] ) , that prevents a complete correlation with the continuum lightcurve .
the fractional variability in h@xmath2 is a factor of @xmath242 larger than in h@xmath0 ( see figure[fig : vlt_lcurves ] ) . in v404cyg
the correlation between h@xmath0 and the continuum is much stronger .
also in v404cyg the h@xmath0 emission line varies considerably @xcite by a factor of @xmath242 more than in a062000 .
however , it could be that the optical depth in a062000 is larger than that in v404cyg , so the emission line flares are generally weaker , and h@xmath2 is more pronounced .
also , the persistent lines could be stronger relative to the flaring component in a062000 .
the h@xmath0/h@xmath2 ratio varies by 8 percent around a mean value of 2.7 , and is inversely correlated with the line fluxes , consistent with there being a higher optical depth in h@xmath0 than in h@xmath2 .
the ratio h@xmath0/h@xmath2 is consistent with case b recombination ( 2.85 for t=10@xmath25k @xcite ) and so indicates that the balmer emission lines are optically thin . for the flare event at phase 1.15 the balmer decrement drops from 3.0 to 2.6 during the flare event indicating that either the flare arises from more optically thick regions or most likely that the temperature increases during the flare ( from @xmath245000k to @xmath2430,000 k ; @xcite ) .
the amplitude of the line flare is larger than in the continuum by a factor of @xmath243 .
one might expect that is because the continuum lightcurve is diluted by light from the secondary star ( see section[low ] ) .
after this correction the amplitude of the flares in the continuum compared to the line are similar . from the data presented here for a062000 , some , but not all , flare events that appear in the continuum lightcurve are also present in the emission line , similar to what is observed in the continuum and h@xmath0 lightcurves of v404cyg @xcite . the wht lightcurves ( figure[fig : wht_lcurves ] )
exhibit no pronounced flares similar to those seen in the vlt data ( figure[fig : vlt_lcurves ] ) .
the continuum lightcurves are relatively smooth , with possibly some weak flares present .
these are not detectable in the h@xmath0 lightcurves , however , and the h@xmath2 data , where present , are too noisy to tell .
it may be that the wht observations represent a less active epoch ; such differences in flare behaviour between epochs do seem to occur in quiescent sxts , and can be very marked ( hynes et al . in preparation ) .
there have been several photometric studies of a062000 in quiescence which show a distorted double - humped modulation ( @xcite ; @xcite ; @xcite ; @xcite and @xcite ) .
the maxima and minima in the lightcurves vary in height and depth and on occasions the maxima have reversed ( see figure1 of haswell 1996 ) . furthermore , it has been reported that the average optical magnitude varies on a timescale of a few hundred days .
possible explanations for these variations are star - spots on the secondary star ( @xcite ; @xcite ) or a persistent superhump @xcite . to obtain a v - band lightcurve of a062000 from our spectra , we integrated the dereddened spectra with the johnson @xmath26-band filter response .
the resulting lightcurve is typical ; we see the classic distorted double - humped modulation and the maxima have different heights .
our lightcurve resembles the optical lightcurves obtained by @xcite , with the brighter maxima occurring at phase 0.25 .
the uncertainty in the reddening adds a 7 percent uncertainty to the the mean flux level.we use the x - ray binary model described in @xcite with the parameters @xmath7=0.067 , @xmath5=41@xmath6 , @xmath9=11@xmath4 a distance of @xmath27=1164pc and a secondary star with an effective temperature of t@xmath28=4600k ( @xcite ; @xcite ) , to determine the secondary star s ellipsoidal modulation .
the excess lightcurve is then obtained by subtracting the secondary star s ellipsoidal modulation from the underlying continuum of the de - reddened lightcurve ( see figure[fig : shump ] ) . to interpret the observed lightcurve we require the secondary star s ellipsoidal modulation plus the steady accretion disc s light with excess continuum emission ( i.e. superhump light ) between orbital phase 0.00 and 0.50 .
the mean level of the excess lightcurve i.e. the photometric veiling , suggests that the disc contributes 20 percent to the observed flux .
this is consistent with the disc fraction estimated using the secondary star s absorption lines ( in section[low ] ) , given that the mean level of the ellipsoidal lightcurve is at most only accurate to 20 percent due to the uncertainties in the distance .
although the exact amplitude and flux of the excess lightcurve depends on the model parameters that predict the secondary star s ellipsoidal modulation , the result that there is excess light near phase 0.2 remains firm . to determine the colour and possible origin of the excess light we compared the data taken around the maximum of the inferred superhump at orbital phase @xmath290.2 ( see figure[fig : shump ] ) to the data taken one orbital cycle earlier @xmath30)[email protected] . since the secondary star s ellipsoidal modulation contribution will be the same in both cases , any difference in the spectral shape
must be due to a superhump at @xmath31 , or a starspot depressing the ellipsoidal maximum at phase @xmath32 .
the continuum shapes of the spectra taken at @xmath31 and @xmath32 are not signifcantly different .
therefore , our data do not allow us to say if this excess light arises from star - spots or the superhump .
there have been many studies of a062000 where the fractional contributions of the light from the secondary star and accretion disc have been estimated . by subtracting the k secondary star spectrum from the broad - band data of a062000obtained ten months after the 1975 may outburst , @xcite found the total v - band light due to the disc at 5500 to be @xmath33=43@xmath16 percent .
he also determined a power - law form for the disc s light ( @xmath34 ) with an index of -2 .
similarly , using the same analysis , @xcite obtained @xmath33=40@xmath110 percent and @[email protected] for the disc s light in 1985 .
further observations by @xcite showed that the disc s fractional contribution at 5225 is roughly constant at @xmath3542 percent during 19851990 . using high resolution spectroscopy ,
@xcite find that the disc contributes 6@xmath13 percent at h@xmath0 and 17@xmath13 percent at h@xmath2 in 1991/1992 .
as pointed out by @xcite the difference in the disc s fractional contribution determined by @xcite and @xcite is probably due to the choice of the template star .
we model the average quiet - state spectrum of a062000 in terms of a template secondary star and non - variable accretion disc spectrum .
we use a k4 spectrum ( hd154712 ) taken from the gunn - stryker atlas @xcite to represent the secondary star and a power - law spectrum of the form @xmath36 to represent the light from the accretion disc .
since the spectral resolution of the a062000 spectrum is higher than the template stars in the gunn - stryker atlas , the spectrum of a062000 was rebinned to the spectral dispersion of the template star ( 10pixel@xmath15 ) .
figure[fig : qspectrum ] shows the results of a spectral synthesis of the continuum emission .
we find that the accretion disc s light has a power - law index of [email protected] and contributes 37@xmath113 percent to the observed continuum light at 5500 . from the residual of the fit
it can be seen that a power - law model for the disc predicts more flux short - ward of 5000 , so a power - law function does not best describe the disc s light .
if we model the spectrum of the disc with a blackbody function , we obtain a significantly better fit ( at the 99 percent level ) , where the disc contributes 58@xmath116 percent to the continuum flux at 5500 and the disc has a blackbody temperature of 4600@xmath1100k and a radius of [email protected]@xmath37 . using a k3 star in order to match the blue end of the spectrum does not give a better fit ( at the 99.99 percent confidence level . )
all the uncertainties quoted are 1-@xmath22 and have been rescaled so that the @xmath38 of the fit is 1 .
although our determination of the fractional contribution of the accretion disc s light to that observed is consistent with the findings of @xcite , @xcite and @xcite , the form of power - law description of the disc s light is not .
however , it should be noted that the non - variable accretion disc light is a composite of light from the steady - state accretion disc plus the excess light from star - spots or a late - superhump , which most probably have very different spectral shapes . for the case where the excess light is due to a late - superhump , the late - superhump modulation may have its origin in the outer regions of the disc in the changing stream - disc interaction ( rolfe , haswell & pattterson 2001 ) , the spectrum of the late - superhump is expected to be different compared to the steady - state disc . although the tidally heated stream - disc impact region will be hotter than the rest of the outer disc , it is not obvious how much hotter this will be compared to the inner regions of the disc , where viscous stresses due to differential rotation are much higher .
thus it is difficult to estimate the spectrum of the late - superhump without detailed computations .
therefore , it is no surprise that the disc s light and spectrum are observed to change with time , as it just reflects the variable behaviour of the accretion disc .
assuming that the light produced by the flares is simply added to the quiescent spectrum which contains the secondary star and light from the accretion disc ( assumed to be non - variable ) , the difference between flare - state spectra and the quiet - state spectra yields an estimate for the spectrum of the flares .
the actual portions of the lightcurves used to determine the flare - state and quiet - state spectra are marked in figure[fig : flarepos ] and were selected after subtracting the secondary star s ellipsoidal modulation .
figure[fig : fspectrum ] shows the resulting spectrum , which has been binned for clarity .
we compared the average flare spectrum taken during the beginning and end of the night .
they were found to be the same to within 1.5 percent . before the subtraction
we shift the spectra into the rest frame of the secondary star , so that the features from the secondary are removed cleanly .
the r - band mag varies on a timescale of a few hundred days with an amplitude of 0.3 mag @xcite .
however , since this timescale is much longer than the orbital period it is safe to assume that the lightcurve of the superhump of excess light does not change shape over the length of the orbital period .
the flare spectrum has a relatively flat continuum suggesting that a high temperature model is needed to fit the data .
we attempt to fit the flare spectrum with different models .
a fit using a blackbody ( @xmath39=3.93 ) gives a temperature of 5900@xmath1200k and a radius [email protected]@xmath37 ( 90 percent confidence ) .
a power - law fit ( @xmath39=3.85 ) of the form @xmath34 has an index of [email protected] , or @xmath40 ( 90 percent confidence ) .
a more realistic model is a continuum emission spectrum of an lte slab of hydrogen . using the synthetic photometry synphot package ( iraf / stsdas ) to compute the lte models we fit the continuum regions to estimate the temperature , radius and baryon column density of the slab .
we find that the @xmath21 surface has a broad minimum at @xmath39 of @xmath244 .
the requirement that the flare region is smaller than the area of the accretion disc ( 0.5@xmath41 ; @xcite ) adds a weak constraint , only ruling out temperatures less than @xmath245000k .
we constrain the temperature and equivalent radius of the optically thin lte slab to lie in the range 1000014200k and 0.0320.044 @xmath37 ( 99 percent confidence ) for a baryon column density in the range 10@xmath4210@xmath43 nucleons@xmath44 ( see figure[fig : chi ] ) .
the emission covers 0.050.08 percent ( 99 percent confidence ) of the accretion disc s projected surface area ( @xmath7=0.067 , @xmath5=41@xmath45 and @xmath9=11@xmath4 ; @xcite and @xcite ) .
tight constraints on the parameters of interest can not be placed because of the correlations between the temperature and column density ; a high temperature and low column density model gives the same @xmath21 as a low temperature and high column density model .
although this can only be done by using the balmer emission line fluxes , for a062000 this is very difficult as it is clear that the balmer emission lines are contaminated by the emission from the bright - spot ( see section[doppler ] ) .
aeaqr is an unusual cataclysmic variable in that it exhibits flaring behaviour which can be described in the framework of a magnetic propeller that throws out gas out of the binary system @xcite .
the flares are thought to arise from collisions between high - density regions in the material expelled from the system after interaction with the rapidly rotating magnetosphere of the white dwarf @xcite .
the spectrum of the flares in aeaqr can be described by a optically thin gas with a temperature of 800012,000k @xcite .
the spectrum of the flares seen in a062000 are described by optically thin gas with a similar temperature , it seems unlikely that the same mechanism producing the flares in aeaqr operates in the sxts .
[ although the temperature of the large flares in v404cyg have been estimated to be @xmath248000k , the uncertainties should be noted @xcite . ]
doppler tomography is a powerful tool in the study of accretion disc kinematics and emission properties @xcite .
this analysis method is now a widespread procedure in the study of emission - line profiles in both cataclysmic variables and low - mass x - ray binaries , providing a quantitative mapping of optically thin emission regions in the velocity space . when intrinsic doppler broadening by bulk flow is present , the tomograms provide a concise and convenient form of displaying phase - resolved line profiles .
however , there are several assumptions inherrant in doppler tomography that should be noted ; it is assumed that the emission is fixed in the co - rotating frame , the emission does not vary over the binary orbit and the motion is in the orbital plane . if these assumptions are not valid then doppler tomography can produce artefacts @xcite .
the balmer line profiles show the doubled peaked structure typical of an accretion disc , and periodic changes in the strength of the two emission peaks which are attributed to s - wave emission components ( see figure[fig : dopmaps ] ) . the trailed spectra show two s - wave components . prior to constructing the doppler images , the spectra were re - binned onto a constant velocity scale ( 60 @xmath3pixel@xmath15 ) and then continuum subtracted using a spline fit to the continuum regions close to the balmer lines .
when computing the doppler maps , we use the systemic velocity determined from the secondary star s absorption radial velocity curve @xcite .
the h@xmath0 and h@xmath2 doppler images show a broad ring background which is the emission from the accretion disc extending to high velocities .
the images also show regions of entended enhanced emission at the position of the stream / disc impact point and extended emission from the opposite side of the disc . in figure[fig :
dopmaps ] we show paths for the gas stream and the kepler velocity of the disc along the gas stream for @xmath7=0.067 and @xmath18=433@xmath3
. the bright - spot does not lie on either predicted path for the gas stream , but occurs half - way between them as is observed in ugem @xcite .
as explained by @xcite the most probable explanation for this is that we are seeing gas around the bright - spot after it has passed through the shock at the edge of the accretion disc . assuming that the velocity after the shock is a combination of the velocity of the gas stream and accretion disc
, we can estimate the radius of the post - shock emission , which most probably corresponds to the accretion disc radius . from the h@xmath0 and h@xmath2 maps
we deduce a radius of @[email protected]@xmath41 ( where @xmath41is the distance the the inner lagrangian point ) , similar to the values determined previously , @xmath470.5@xmath41 @xcite .
the h@xmath0 data suggest an anti - correlation with the continuum lightcurves , in both the wht and vlt datasets . such an effect
would , of course , be expected in the equivalent width , but not in the fluxes .
given that the effect repeats between the two independent datasets , this is unlikely an artifact , and probably rather indicates a real modulation in the h@xmath0 flux .
such an effect can be explained if it originates from a more optically thick region for which we see a varying projected area . in this case , a double - humped modulation , analogous to ellipsoidal modulations , is expected .
the most likely region to be responsible for this is the accretion stream impact and/or overflow , since the h@xmath0 and h@xmath2 doppler tomograms show that this region is much more prominent in h@xmath2 relative to the disc than in h@xmath0 , as expected if the h@xmath0 emission arises from more optically thick regions than the disc .
furthermore , we would expect emission to be out of phase with the ellipsoidal modulation .
the phase of maximum light would depend on the disc radius and how extended the stream impact region is , but up to a shift of 0.2 or so in phase relative to the ellipsoidal modulations is not implausible .
superhumps are optical modulations that arise when an accretion disc expands to the resonance radius , becomes elliptical and is forced to precess @xcite .
it is believed that enhanced viscous dissipation in the eccentric disc due to the tidal stress exerted by the secondary star is the source of light observed in the superhump lightcurves .
such an eccentric disc which gives rise to asymmetric emission line profiles , has been successfully used to explain the asymmetric line profiles in oycar @xcite , zcha ( @xcite , @xcite ) and amcvn @xcite .
the doppler maps of a062000 presented here show significant extended emission , strongest in h@xmath0 , with similar velocities to the stream / disc impact point , but from the opposite side of the disc between phase 0.25 and 0.50 ( see figure[fig : dopmaps ] ) .
this crescent - like structure is reminiscent of the feature seen in the doppler maps of xtej2123 - 058 in quiescence @xcite .
also , the orbital average h@xmath0 profiles of xtej1118 + 480 near quiescence show significant velocity variations and can be interpreted as evidence for an eccentric precessing disc @xcite .
thus the most likely explanation for the crescent structure seen in the doppler maps of a062000 presented here , is that it is due to an elliptical accretion disc .
however , it should be noted that such structure is not seen in the doppler maps determined by @xcite .
it is interesting to see what the doppler map of an eccentric disc would look like .
@xcite have attempted to compute such a map and find that the doppler map of an eccentric disc is shifted with respect to the system s centre of mass with asymmetric enhancements in intensity around the edge of wthe disc .
more recently , @xcite have performed detailed 2-d smoothed particle hydrodymanical ( sph ) simulations of an eccentric precessing disc in a binary system .
the synthetic trailed spectra show two distinct non - sinusoidal s - wave features which evolve with the disc precession phase which are produced by the gas stream and the disc gas at the stream - disc impact shock .
these emission regions are not fixed in the co - rotating frame and vary with time over the binary orbit . to show the effects of using data that violate the assumptions inherrant in dopper tomography , the authors produce maximum entropy doppler maps of the sph trailed spectra .
they find that an artefact of doppler tomography is that the stream impact emission is reconstructed in velocity space as a broad emission region between the gas stream ballistic path and the velocity of a circular keplerian disc along the ballistic stream trajectory i.e. the stream s `` kepler shadow '' ( see figure9 of @xcite ) . in both the h@xmath0 and
h@xmath2 trailed spectra of a062000 , characteristic rapidly turning red - blue - red non - sinusoidal s - wave features between orbital phases 0.2 and 0.6 are seen , and look very similar to the simulated sph trailed spectra of a precessing disc .
the reconstructed data between phase 0.8 and 1.0 indicate that the real data are not well reproduced by sum of sinusoids that vary on the orbital period .
if there is an eccentric precessing disc , as there must be if there is a superhump , the emission regions are not fixed in the co - rotating frame , it is not surprising that doppler tomography fails .
furthermore , the broad emission near the stream / disc impact region can be interpreted as the artefact of doppler tomography , which blurs the emission produced by the gas stream and the disc gas at the stream - disc impact region .
thus qualitatively our data suggests the presence of a precessing disc in a062000 .
given the mass ratio of a062000 , the 3:1 resonance is the dominant tidal instability and lies at @xmath48=0.66@xmath41 .
although we derive a disc radius less than this value ( see section[bspot ] ) , it should be noted that we actually determine the post - shock radius of the gas stream as it impacts the accretion disc , which most probably lies close to but not on the edge of the accretion disc .
furthermore , @xcite , @xcite and @xcite show that the effective radius of the disc as measured by the stream - disc impact region changes dramatically with superhump phase .
thus the method of doppler tomography will estimate a mean value for the disc s radius .
therefore , it is possible that the radius of the accretion disc extends up to the tidal radius and so the asymmetric brightness distribution in the accretion disc we observed can be explained by the presence of a tidally distorted accretion disc .
@xcite and @xcite have proposed an accretion flow model to explain the observations of quiescent black hole x - ray transients .
the accretion disc has two components , an inner hot advection - dominated accretion flow ( adaf ) that extends from the black hole horizon to a transition radius and a thin accretion disc that extends from the transition disc to the edge of the accretion disc .
interactions between the hot inner adaf and the cool , outer thin disc , at or near the transition radius could be a source of quasi - periodic variability .
the adaf flow requires electrons in the gas to cool via synchrotron , bremsstrahlung , and inverse compton processes which predict the form of the spectrum from radio to hard x - rays .
it has been suggested that the flares observed in a062000 and other quiescent x - ray transients arise from the transition radius ( @xcite ; @xcite ) .
indeed , the 0.78mhz feature detected in v404cyg has been used to determine the transition radius , assuming that the periodicity represents the keplerian period at the transition between the thin and advective disc regions @xcite , and a possible low frequency break in the power spectrum of a062000 may have a related origin @xcite . in the original adaf models the optical flux is produced by synchrotron emission . as pointed out in @xcite , it is difficult to explain the flare spectrum in terms of optically thin synchrotron emission , unless the electrons follow a much steeper power - law electron energy distribution compared to solar flares ; solar and stellar flares have a frequency spectrum with a power - law index of @xmath0=0.5 and an electron energy distribution with a power - law index of @xmath242
. however , more recently the adaf models have been questioned , and substantial modifications proposed .
@xcite emphasised that the bernoulli constant of the gas is positive and hence outflows are possible ( as also noted by @xcite ) . in the adiabatic inflow outflow solution ( adios )
model of @xcite most of the accretion energy that is released near the black hole is used to drive a wind from the surface of the accretion disc .
most of the gas that falls onto the outer edge of the accretion disc is carried by this wind away from the black hole , with the result that the hole s accretion rate is much smaller than the disc s accretion rate .
advective flows are also expected to be convectively unstable , as also remarked by @xcite . in these models the central accretion rate
can also be suppressed . for either of these cases ,
the effect is to shift the source of cooling outwards , and the optical synchrotron emission is reduced ( @xcite ; @xcite ) . in these more realistic cases
@xmath49 flares directly from the inner flow are less likely .
optical emission might still arise from the inner edge of the outer ( thin ) disc , or from reprocessing of x - ray variability .
it was argued by @xcite that since large ( @xmath24 few d ) flares in v404cyg involve enhancements of both blue and red wings of h@xmath0 , they must involve the whole disc .
while these observations indicate participation by a wide range of azimuths , they leave open the possibility that only the inner disc is involved . using simultaneous multicolour photometry @xcite determined the colour of similar large flares in v404cyg .
although the flare parameters determined are complicated by uncertainties in the interstellar reddening , the flare temperature was estimated to be @xmath248000k .
flares on timescales similar to those present in a062000 ( i.e. tens of minutes ) were also observed , but no physical parameters could be determined given the large uncertainties in the colour .
the large ( @xmath24 few hrs ) flares were observed to arise from regions that cover at least 3 percent of the disc s surface area . @xcite
have suggested that the large ( @xmath24 few hrs ) flares in v404cyg are produced in regions further out or further above the disc from a corona than the rapid ( @xmath240.5hr ) and more rapid ( @xmath24min ) flares .
if the flares in v404cyg and a062000 have the same origin , then it is interesting to note that the large flares ( @xmath24 few hrs ) observed in v404cyg cover a larger surface area of the disc compared to the rapid ( tens of mins ) flares observed in a062000 , which is consistent with the idea that the large flares arise from regions that extend further out into the disc compared to the rapid flares . although the flare temperatures derived suggest that large flares are cooler than rapid flares , the uncertainties are large and so no meaningful conclusion can be drawn at this stage .
accurate physical parameters for the flares can only be obtained by resolving the balmer jump and paschen continuum . for the data presented here on a062000 , there is evidence that some of the flare events in the continuum lightcurve are correlated with the balmer emission line lightcurves , similar to what is observed in v404cyg @xcite .
the value for the balmer decrement suggests that the persistent flux is optically thin and the decrease of the balmer decrement during the flares suggests a significant temperature increase .
we find that the optically thin spectrum of the flare , which lasts tens of minutes , has a temperature of @xmath2412000k and covers 0.08 percent of the disc s projected surface area ( see section[flare ] )
. in many high inclination systems the continuum light from the bright - spot produces a single hump in the lightcurve , because the bright - spot is obscured by the disc when it is on the side of the disc facing away from the observer .
since a062000 is at an intermediate inclination angle ( 41@xmath6 ; @xcite and @xcite ) where no strong obscuration effects are predicted , it is not surprising that the continuum lightcurve does not show obvious evidence of a bright - spot . from the doppler maps of a062000 ( see section[bspot ] ) , it is clear that the bright - spot is present in the balmer emission lines .
if the bright spot were present in the continuum then we would expect strong correlations between the continuum and emission line lightcurves .
indeed we see some evidence for such correlations in the lightcurves ( see figures[fig : vlt_lcurves ] and [ fig : wht_lcurves ] ) .
it is interesting to note that the flare temperature and area we obtain ( see section[flare ] ) is consistent with the bright - spot temperature and radius determined by @xcite .
therefore , with the data presented here we can not rule out the possibility that the flares arise from the bright - spot .
although , this hypothesis can only be tested by observing high inclination x - ray transients , where one would expect the orbital distribution of the flares and lightcurves to be indicative , it should be noted that the kinematics of the flares in v404cyg do rule out at least some of the flares coming from the bright spot .
we have presented a time - resolved spectrophotometric quiescent study of the optical variability in a062000 .
we observe the well known flare events which and find that some of the events appear in both the continuum and the emission line lightcurves .
the balmer line flux and variations suggest that that the persistent emission is optically thin and the drop in the balmer decrement for the flare event at phase 1.15 , suggests that either a significant increase in temperature occurred during the flare or that the flare is more optically thick than the continuum .
the data suggests that there are two hi emitting regions , the accretion disc and the accretion stream / disc region , with different balmer decrements .
the orbital modulation of h@xmath0 with the continuum suggests that the steeper decrement is most likely associated with the stream / disc impact region .
we find that flare spectrum can be described by an optically thin gas with a temperature in the range 1000014000k that covers 0.050.08 percent ( 90 percent confidence ) of the accretion disc s surface . given these parameters ,
the possibility that the flares arise from the bright - spot can not be ruled out .
finally , doppler images of the h@xmath0 and h@xmath2 emission show enhanced emission at the stream / disc impacts region as well as extended structure from the opposite side of the disc .
the trailed spectra show characteristic s - wave features that can be interpreted in the context of an eccentric accretion disc .
ts acknowledges support from the spanish ministry of science and technology under project aya200203570 .
rih is supported by nasa through hubble fellowship grant # hf-01150.01-a awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa , under contract nas 5 - 26555 . based on observations made with eso telescopes at the paranal observatory under programme 70.d-0766 and also with the william herschel telescope operated on the island of la palma by the issac newton group in the spanish observatorio del roque de los muchachos of the instituto
de astrofsica de canarias .
avni y. , bahcall , j.n .
, 1975 , apj , 197 , 675 ball g.h . , narayan r. , quataert e. , 2001 , apj , 552 , 221 blandford r.d . , begelman m.c , 1999 , mnras , 303 , l1 boley f. , wolfson r. , 1975 , iau circ.2819 casares j. , dubus g. , shahbaz t. , zurita c. , charles p.a . , 2002 , mnras , 329 , 29 elvis m. , grifiths c.g .
, turner m.j.l .
, page c. , 1975 , iau circ.2184 froning c.s . ,
robinson e.l . , 2001 ,
aj , 121 , 2212 foulkes s.b . , haswell c.a . , murray j.r
, rolfe d.j . , 2004 , mnras , 349 , 1179 gelino d.m . , harrison t.e .
, orosz j.e . , 2001 ,
aj , 122 , 2668 gunn j.e . ,
. , 1983 , apjs , 52 , 121 haswell c.a . ,
1992 , ph.d thesis , university of texas .
haswell c.a . , robinson e.l . ,
horne k. , stiening , r.f .
, abbott t.m.c . , 1993 ,
apj , 411 , 802 haswell c.a . , 1996 ,
in compact stars in binaries , iau symposium 165 , 351 hessman f.v .
, mantel k. , barwig h. , schoembs r , 1992 , a&a , 263 , 147 honey w.b .
, charles p.a . , whitehurst r. , barrett p.b . , smale a.p . , 1988 , mnras , 231 , 1 horne k. , 1986 , pasp , 98 , 609 hynes r.i . , zurita c. , haswell c.a . , casares j. , charles p.a . , pavlenko e.p .
, shugarov s.yu . ,
lott d.a . , 2002 , mnras , 330 , 1009 hynes r.i . , charles p.a . , casares j. , haswell c.a . , zurita c. , shahbaz t. , 2003 , mnras , 340 , 447 khruzina t.s . , cherepashchuk a.m. , 1995 , arep , 39 , 178 leibowitz e.m .
, hemar s. , orio m. , 1998 , mnras , 300 , 463 mcclintock j.e . ,
petro i.d . , remillard r. , ricker g.r . , 1983 , apj , 266 , l27 mcclintock j.e . , horne k. , remillard r. , 1995 , apj , 442 , 358 mcclintock j.e . , remillard r. , 2000 , apj , 531 , 956 marsh t.r . , horne k. , 1988 , mnras , 235 , 26 marsh t.r . , horne k. , schlegel e.m . , honeycutt r.k . ,
kaitchuck r.h .
, 1990 , apj , 364 , 637 marsh t.r .
, robinson e.vl , wood j.h . , 1994 , mnras , 266 , 137 narayan r. , yi i. , 1994 , apj , 428 , l13 narayan r. , barret d. , mcclintock j.e . , 1997 , apj , 482 , 448 narayan r. , mcclintock j.e . ,
yi i. , 1996 , apj , 457 , 821 naylor t. , 1998 , mnras , 296 , 339 osterbrock d.e . , 1987 , astrophysics of gaseous nebulae , freeman , san francisco oke j.b . , 1974 , apjss , 27 , 21 oke j.b . , 1977 , apj , 217 , 181 patterson j. , halpern j. , shambrook a. , 1993 , apj , 419 , 803 pearson k.j .
, horne k. , skidmore w. , 2003 , mnras , 338 , 1067 quataert e. , narayan r. , 1999 , apj , 520 , 298 rolfe d.j .
, haswell c.a . ,
patterson j. , 2001 , mnras , 324 , 529 seaton m.j . , 1979 , mnras , 187 , 73 shahbaz t. , naylor t. , charles p.a . , 1994 , mnras , 268 , 756 shahbaz , t. , bandyopadhyay r.m . , charles , p.a . , 1999 , a&a , 346 , 82 shahbaz t. , zurita c. , casares j. , dubus g. , charles p.a . , wagner r. , ryan e. , 2003a , apj , 585 , 443 shahbaz t. , dhillon v.s .
, marsh t.r . , zurita c. , hynes r.h .
, charles p.a .
, casares j. , 2003b , 346 , 1116 smith d. , 1999 , ph.d .
thesis , university of cambridge tjemkes s.a . , van paradijs j. , zuiderwijk e.j . , 1986 , a&a , 154 , 77 tonry j. , davis m. , 1979 , aj , 84 , 1511 van paradijs j. , mcclintock j.e .
, 1995 , in x - ray binaries , eds .
lewin , j. van paradijs , e.p.j .
van den heuvel , cup , p. 58
vogt n. , 1982 , apj , 252 , 653 welsh w.f . , horne k. , oke j.b . 1993 , apj , 406 , 229 whitehurst r. , king a.r . , 1991 , mnras , 249 , 25 wilson r.e . , devinney e.j .
, 1971 , apj , 166 , 605 wu c .- c .
, panek r.j .
, holm a.v . , schmitz m. , swank j.h . , 1983 , pasp , 95 , 391 wynn g.a , king a.r . ,
horne k. , 1997 , mnras , 286 , 436 zurita c. et al . , 2002 , mnras , 333 , 791 | we present a time - resolved spectrophotometric study of the optical variability in the quiescent soft x - ray transient a062000 . superimposed on the double - humped continuum lightcurve
are the well known flare events which last tens of minutes
. some of the flare events that appear in the continuum lightcurve are also present in the emission line lightcurves . from the balmer line flux and variations , we find that the persistent emission is optically thin . during the flare event at phase 1.15 the balmer decrement dropped suggesting either a significant increase in temperature or that the flares are more optically thick than the continuum .
the data suggests that there are two hi emitting regions , the accretion disc and the accretion stream / disc region , with different balmer decrements .
the orbital modulation of h@xmath0 with the continuum suggests that the steeper decrement is most likely associated with the stream / disc impact region . by isolating the flare s spectrum
we find that it has a frequency power - law index of [email protected] ( 90 percent confidence ) .
the flare spectrum can also be described by an optically thin gas with a temperature in the range 1000014000k that covers 0.050.08 percent ( 90 percent confidence ) of the accretion disc s surface . given these parameters ,
the possibility that the flares arise from the bright - spot can not be ruled out .
we construct doppler images of the h@xmath0 and h@xmath2 emission lines .
apart from showing enhanced blurred emission at the region where the stream impacts the accretion disc , the maps also show significant extended structure from the opposite side of the disc .
the trailed spectra show characteristic s - wave features that can be interpreted in the context of an eccentric accretion disc .
accretion , accretion discs binaries : close stars : individual : a062000 |
semiconductor leds and their counterpart solar cells are expected to play a major role in general lightning and renewable energy production .
the development of organic light emitting diodes ( oled ) has proven to be successful and they are on the market , notwithstanding the disadvantages of a poor carrier injection and transport properties compared to inorganic semiconductors . an idea to further improve such devices
is the use of an hybrid system that would combine the best properties of both materials together and potentially open new possibilities for optoelectronic devices@xcite . in the strong coupling regime ,
novel hybrid quasiparticles are formed , with properties diverse from those of the individual excitons leading to new uses in particular those requiring large nonlinearities as for optical switching . in the weak coupling regime ,
in which wannier and frenkel excitons maintain their individuality , a suitable use of each component would also bring new features@xcite and could overcome the drawback presented earlier .
for instance , carriers are electrically injected , transported and bound into excitons in the inorganic semiconductor subsystem , this latter being coupled to an organic light emitting subsystem via a frster energy transfer process@xcite .
this internal non - radiative energy transfer had been proposed and has been recently demonstrated to be efficient enough ( see ref . for a review of relevant work ) .
following early theoretical predictions@xcite , the energy transfer process has been observed from a quantum well to a quantum dot overlayer@xcite , and from a quantum well to an organic overlayer@xcite .
a way to further improve the transfer between both components is the use of a plasmonic resonance to enhance the electric field in the system and take advantage of the metallic cathode which is often mandatory in electrical devices .
indeed the electrodes are often responsible of parasitic and strong waveguiding , a considerable energy is trapped in the surface plasmon mode then the recovery of some of this energy becomes then of interest@xcite we deal here with such hybrid systems structured in a planar geometry whereby a wannier exciton in an inorganic semiconductor quantum well plays the role of the donor and a frenkel exciton in a crystalline organic overlayer that of the acceptor .
we study the role of an intermediate metallic layer that will carry the excitation through plasmons from the inorganic part to the organic .
moreover , we take into account retardation in the model and the possible quenching effects of the organic emission due to the metal layer . in section [ model ] , we describe the theory of frster energy transfer in a planar hybrid nanostructure taking into due account plasmonic excitations and retardation . in section [ diel ] ,
we use parameter values of typical hybrid structures to calculate the transfer rate for various configurations .
we discuss the role of the plasmonic enhancement of the transfer and the additional loss channels introduced by the metal .
we consider here the planar architecture shown in fig.[schema ] .
the donor subsystem consists of a semiconductor quantum well of thickness @xmath0 sandwiched between two semiconductor barriers of thickness @xmath1 . for simplicity , we take the same background dielectric constant @xmath2 for the well and the barriers and we assume the barriers to be infinitely high ( i.e. , the wannier exciton is fully contained in the well region ) . at the bottom ( @xmath3 ) lies a transparent glass substrate with dielectric constant @xmath4 . at the top
lies a metallic layer described by his complex dielectric constant @xmath5 , a spacer layer with constant @xmath6 and then ( @xmath7 ) lies the acceptor subsystem consisting of a crystalline organic medium with complex dielectric constant @xmath8 and thickness @xmath9 followed by air .
the bottom glass and the air parts are supposed to be semi - infinite .
the quantities @xmath4 , @xmath2 , @xmath10 and @xmath11 include only the contribution of higher resonances ( with respect to the exciton energies under consideration ) and we consider them to be real .
the quantity @xmath12 is the total dielectric function of the organic material , including in particular the resonant absorption due to the frenkel excitons , and thus it is a complex valued tensor @xcite , which however is assumed here isotropic for simplicity .
the complex dielectric constant @xmath13 also accounts for the metal losses .
here we extend the theoretical framework for the frster energy transfer rate in a hybrid nanostructure developed in references , and discussed in detail in reference , to take into account retardation effects as well as plasmonic resonances .
the former approach is equivalent to the usual frster theory based on the dipole - dipole interaction @xcite and leads to a macroscopic , semiclassical description of the energy transfer .
its main advantage is that the acceptor subsystem enters the calculation only via its complex ( frequency dependent ) dielectric tensor , a quantity directly accessible by experiment . in short
, the transfer rate is obtained from the joule losses suffered in the organic medium by the electric field induced by the quantum well exciton , through the metal .
this is appropriate to the weak coupling regime between donor and acceptor and occurs when the linewidth of the acceptor resonance is large compared to the interaction coupling donor and acceptor , leading to an irreversible energy transfer .
moreover , as the linewidth of 2d excitons in inorganic quantum wells are typically at least one order of magnitude smaller than the linewidth of the organic material excitons , we employ for simplicity a delta function approximation for the emission spectrum of the donor subsystem . as a matter of fact
, the presence of the wannier exciton of energy @xmath14 gives rise to a source term in the quantum well corresponding to the exciton polarization oscillating at frequency @xmath15 which can be written as : @xmath16 where @xmath17 is the matrix element of the electric dipole moment between the bloch functions of the conduction and valence band extrema and @xmath18 is the envelope function describing the bound electron - hole pair ( @xmath19 , @xmath20 being the electron and hole coordinates ) . in our simple model of a 2d wannier - mott exciton
, the polarization is given by the product of the @xmath21-wave function of the 2d relative motion of the electron - hole pair with the lowest subband envelope functions of electron and hole ( which are equal in the approximation of infinitely deep well ) and with the plane - wave function of the center - of - mass motion .
this description is only appropriate to coherent large radius excitons with a well defined center of mass wavevector , as usually occur in high quality inorganic semiconductor quantum wells .
taking into account their corresponding normalization we can write : @xmath22 where @xmath23 is the in - plane normalization area , @xmath24 the in - plane wave vector of the center - of - mass motion , @xmath25 the in - plane component of @xmath26 and @xmath27 is the 2d exciton bohr radius . then , differently from previous work ( ref . , ) , we look for the electric field @xmath28 resulting from this polarization calculated from maxwell s equations with the use of transfer matrix theory @xcite .
the dielectric tensor @xmath29 is piecewise constant corresponding to each different layer , see fig.[schema ] .
we choose it to be complex in the organic layer and the metal to described the absorption and losses in these material , and real in the inorganic part , as the background absorption is supposed to be higher in energy , and for the rest of the system .
the appropriate boundary conditions at the interfaces are the continuity of the tangential electric field and of the normal electric displacement between each layer .
knowing the electric field , we can calculate the transfer rate ( inverse transfer time ) of the excitation into the organic medium : @xmath30 this expression is equivalent to applying the fermi golden rule to the decay of one excited state in the quantum well into the excited states of the organic molecules in the linear regime approximation ( for a derivation of eq.[taugene ] in terms of a microscopic inelastic scattering rate of wannier excitons due to resonant two - level molecules in the organic medium see ref . ) . in eq.[taugene ] , and elsewhere below , the acceptor dielectric permittivity is evaluated at the frequency @xmath15 of the delta - like emission spectrum of the donor .
the power @xmath31 dissipated in the organic layer is given by @xmath32 .
such energy transfer mechanism , in the non - radiative approximation and without considering metallic layer , have been shown to be fast enough to efficiently quench the wannier exciton luminescence and to turn on the organic molecule light emission @xcite .
due to the in - plane translational symmetry of the source ( we are dealing with free excitons in the well ) , we consider the polarization for a given in - plane wave vector . in this case
three modes of different symmetry can be identified : longitudinal ( l ) where @xmath17 is along the in - plane wave vector , perpendicular ( z ) where the dipole moment is oriented along the z - axis , and transverse ( t ) for which the polarization is orthogonal to the two first cases .
for each one , the exciton polarization ( [ polarization ] ) gives rise to an electric field in the quantum well and then in all the structure .
to calculate this latter , we determine for all three cases the source term to be included in the transfer matrix theory , as detailed below . with @xmath33
we can write for the te mode : @xmath34 \\
b_z(z ) & = \frac{ck}{\omega } \left [ a_+^s(z_0 ) e^{iq(z - z_0 ) } + a_-^s(z_0 ) e^{-iq(z - z_0 ) } \right ] \end{aligned}\ ] ] and the tm mode : @xmath35 \\
e_z(z ) & = \frac{ck}{\varepsilon \omega } \left [ a_+^p(z_0 ) e^{iq(z - z_0 ) } + a_-^p(z_0 ) e^{-iq(z - z_0 ) } \right ] \end{aligned}\ ] ] with @xmath36 and @xmath37 the polarization .
the amplitude @xmath38 from each side of a dielectric slab of thickness d are linked by @xmath39 with @xmath40 at an interface between @xmath41 et @xmath42 we have @xmath43 with @xmath44 considering a slab that include a source of field , as the quantum well considered here , a particular solution of maxwell equations in the active domain leads to a source term @xmath23 for the amplitude .
the relation of the amplitude at both ends is @xmath45 where @xmath23 is related to the electric field at the border due to the intrinsic polarization distribution ( [ pol ] ) and @xmath46 is the matrix element for the whole slab , constituted by the matrix that we have previously seen , for a slab of thickness @xmath47 and dielectric permittivity @xmath42 sandwiched between two medias of permittivity @xmath41 and @xmath48 : @xmath49 .
@xmath23 is given by the continuity relations of the fields at each interface .
for @xmath36 and @xmath37 polarizations , writing @xmath50 and @xmath51 we obtain : @xmath52 \label{sources } \\
\nonumber s^p & = \frac{1}{2 } \bigg [ \left ( \begin{array}{c } \frac{\omega \varepsilon_2}{ck } e_z(z_- ) - \frac{\omega \varepsilon_1}{ck } e_l(z_- ) \\ \frac{\omega \varepsilon_2}{ck } e_z(z_- ) + \frac{\omega \varepsilon_1}{ck } e_l(z_- ) \end{array } \right ) \\ & -m_{\varepsilon_1 | \varepsilon_2}^p m_{d } \left ( \begin{array}{c } \frac{\omega \varepsilon_2}{ck } e_z(z_+ ) - \frac{\omega \varepsilon_1}{ck } e_l(z_+ ) \\ \frac{\omega \varepsilon_2}{ck } e_z(z_+ ) + \frac{\omega \varepsilon_1}{ck } e_l(z_+ ) \end{array } \right ) \bigg ] \label{sourcep}\end{aligned}\ ] ] where the fields @xmath53 , @xmath54 , @xmath55 are the electric fields at the border of the active domain .
so starting from the polarization term ( [ polarization ] ) , we solve maxwell equations to find a particular solution in the quantum well domain .
the field at the border will be used to find the general solution in the whole structure , giving a source term for the transfer matrix theory ( ( [ sources ] ) and ( [ sourcep ] ) ) .
choosing a symmetric particular solution according to the well width , the electric field is : @xmath56 with @xmath57 and @xmath58 .
that lead to the matrix elements ( [ sources ] ) and ( [ sourcep ] ) to be use : @xmath59 where the first row correspond to the traveling waves from the bottom to the top in fig.[schema ] , and the second row to the waves traveling in the opposite direction .
from here it is straightforward to determine the electric field in the whole structure and then the transfer rate to the organic media ( [ taugene ] ) .
we also determine the losses in the process which compete with the energy transfer described by @xmath60 .
the main channel will be the dissipation in the metal @xmath61 : @xmath62 and the direct radiative emissions @xmath63 ( as the final interest is the light emitted from the organic part ) given by the flux of the poynting vector in the @xmath64direction in any section of the final layers , glass or air , as we consider them semi - infinite here : @xmath65 we also consider a non - radiative channel of dissipation @xmath66 intrinsic to the well , independent of the @xmath67vector , that will be estimated @xcite and is of the order a few hundreds of picoseconds .
we will discuss below the dependency with the @xmath67vector , however in most cases of experimental interest , the transfer rate for an exciton distribution ( rather than that for one specific exciton ) should be considered via averaging over the initial state of the donor subsystem .
we assume , in particular , that all directions for the center of mass wave vector of the exciton in the well are equiprobable ( i.e. , the exciton distribution has cylindrical symmetry ) . for a thermalized population of quantum well excitons ,
each one having a well defined center of mass wavevector , we also average over the energy @xmath68 , with @xmath69 the exciton mass , according to the boltzmann distribution , assuming that the wavevector dependent energy transfer itself is not fast enough to modify such exciton distribution . in the case of 2d excitons ,
the wave - vectors amplitude distribution is given by : @xmath70 where @xmath71 .
we also neglect here for simplicity the effects on the energy transfer efficiency of the dynamic ionization - recombination equilibrium between excitons and electron - hole unbound pairs which has recently been considered for gaas based quantum wells @xcite and that should be less important in ii - vi based quantum wells having a higher exciton binding energy .
after the excitation is transfered to the organic part , we have to consider the emission of this latter to evaluate the quenching due to the metal layer .
we use the same approach as previously except a few details : as the bohr radius of the frenkel exciton is much smaller than the typical size of the layer , of the order of one molecule , we consider the emission from an assembly of point dipoles . the transfer by a point dipole at a given position @xmath72 inside the organic layer
can be described using the preceding theoretical framework considering a sum over all @xmath73-vectors for an infinitely fine layer of emitters situated at the same position @xmath72 . for each value of the @xmath73-vector
the layer is defined by @xmath74 and @xmath75 : @xmath76 which are the limit of ( [ sources ] ) and ( [ sourcep ] ) when the width of the well is going to zero , with @xmath77 replaced by @xmath78 for the point molecule , and @xmath79 replaced by @xmath80 to take into account the local field effect .
we neglect the absorption due to the quantum well as the emission time for the organic is shorter than the transfer time to the quantum well , plus stockes shift ( weak coupling regime @xcite ) .
from there , after integration , we obtain the electric field in the whole structure for a given molecule at position @xmath72 and then the radiative emission from the top @xmath81 , and the losses : the dissipation in the metal @xmath82 and the emission at the bottom in the glass side @xmath83 .
we integrate on all positions @xmath72 inside the organic to get the transfer rate from the assembly of molecules .
the initial excitation of each molecule is proportional to the energy transfered from the inorganic quantum well that reach its position .
the detail of the spatial repartition of the energy in the organic before the reemission of that latter is strongly dependent of the material and is out of the scope of this article @xcite .
we consider here the two limiting cases of a weak and strong diffusion of the frenkel exciton .
if we can neglect the diffusion then the excitation is proportional to the square of the electric field at the point of the molecule ( the integrand of eq.([taugene ] ) ) , on the contrary for strong diffusion the energy spread in the whole layer and the excitation is independent of the position and proportional to @xmath60 ( [ taugene ] ) .
the plasmonic resonance will increase the transfer rate to the organic but also induce losses , that have to be evaluated so we detail the efficiency of the transfer @xmath84 and the efficiency of the organic emission @xmath85 .
this latter is different for weak and strong diffusion hypothesis .
the two processes being independent , the internal efficiency @xmath86 will be the product of the two quantities .
the total efficiency of the device , that take into account the out - coupling of the light and the injection process , will not be studied here .
@xmath87 the explicit analytic form of the obtained results is much more cumbersome than in previous theoretical work @xcite , due to the more complex geometry including the metal layer and the retardation effects taken into account here . in the following we will then only discuss the results numerically .
in fig.[compa ] we show the results of ref .
( dashed lines ) together with the transfer rate obtained with the full calculation including retardation effects described above ( plain lines ) , for the different possible polarizations of the exciton at the energy emission of the znse well @xmath88 and thickness of the organic layer of @xmath89 . in the large @xmath73 values , where retardation effects can be neglected
, we see that both model coincide . for small @xmath73 values corresponding to the radiative part
the behavior is largely affected .
the @xmath90-polarized exciton transfer rate show a plateau the value of which depends on the thickness of the organic layer , as the propagating modes penetrate all the layer .
the @xmath91-polarization transfer rate shows a similar behavior in the non - radiative regime , but goes to zero when going to normal incidence ( @xmath92 ) this is a simple geometric effects as in this case there is no electric field at the interface ( see eq.([ewz ] ) ) .
we also notice a transfer from @xmath93-polarized exciton , mostly in the radiative range .
this latter is null if we take only into account the non - radiative part , as in this case it does not give rise to any charge distribution nor electric field . in between these two regimes , the position and intensity of the peak
is slightly modified and depends on the absorber thickness .
: plain lines are the exciton s transfer rate taking into account retardation effects for @xmath90-polarization ( blue ) , @xmath91-polarization ( red ) , and @xmath93-polarization ( green ) .
dashed lines are the exciton s transfer rate without retardation for @xmath90-polarization ( blue ) and @xmath91-polarization ( red ) .
the geometry here consist of a znte quantum well sandwiched between two layers of acceptor as in ref . .
the vertical lines are @xmath94 and @xmath95 where @xmath4 is the relative permittivity of the glass substrate.,width=302 ] we turn now to the complete geometry of fig.[schema ] including the metallic layer . for illustrative purposes
, we will focus on the effects of the plasmons and thus only change the overlayer configuration keeping all the rest of the heterostructure fixed .
we take into account a simplified description of the organic with only one exciton and we discuss only the @xmath90-polarization state in the quantum well .
the @xmath91-polarization presents the same behavior in the non - radiative part of the spectrum , however the transfer is killed when we approach orthogonal incidence ( @xmath73=0 ) , and the transverse polarization has a low transfer , only noticeable in the intermediate region with no effect of the plasmon . for the donor subsystem , we take a znte semiconductor quantum well for which @xmath96 , @xmath97 where @xmath27 is taken to be @xmath98 @xcite ( note that in a quantum well this value is half the one of the bulk ) and @xmath99 @xcite where @xmath100 is the electron mass . for the structure , we take @xmath101 , @xmath102 , @xmath103 , @xmath104 , @xmath105 , while the glass substrate and the air top layer are semi - infinite in the z - direction@xcite .
the organic medium is tetracene which absorption match the znte emission@xcite .
we assume that even close to the interface the power transferred to the organic medium does not lead to any saturation effect .
thus , the organic medium can be considered to be in the ground state and can be described by its linear dielectric response . in simple cases , it is possible to use an effective model @xcite , typically near the exciton resonances , where the optical response is modeled by a real background constant and several lorentz transitions @xmath106 , described by their energy @xmath107 , coupling amplitude @xmath108 and damping @xmath109 : @xmath110 @xmath111 is the background constant .
the values of the parameters are obtained by fitting experimental data . for clarity in the discussion about the plasmonic effects
, we restrict ourselves to the lowest excitation for which @xmath112 , @xmath113 , @xmath114 @xcite .
for the metallic layer we use silver with a permittivity described by a combination of drude and critical points models , as described in ref . , that cover the energy range @xmath115 .
[ cols= " < , > " , ] the second part of the process is the radiative emission from the organic component , which can be negatively affected by the proximity with the metal layer .
we consider each molecule as a single emitter .
the radiative transfer in the open part of the device is depending on the emitter position due to interference effects ( see fig.[onemol ] ) with an initial excitation density that depends on the position and on the exciton diffusion in the sample .
after integration we find a radiative decay for the frenkel exciton of about 40ns , which is in the range of the expected values for tetracene without considering any superradiance effect@xcite .
the spacing between the organic and the metal is a key parameter as it has an opposite effect on the transfer from the well and the radiative emission . at @xmath116 ,
the total efficiency is given fig.[qq ] and is less than one percent , 0.25% in the case of a strong diffusion .
that means that the quenching due to the presence of the metal is significant . as a function of the spacer thickness , for an exciton at @xmath116 with strong diffusion ( dashed line ) and weak diffusion ( plain line).,width=245 ]
we present a model for energy transfer in an hybrid device , taking into account both radiative and non - radiative part .
we study the role of a plasmonic resonance due to the presence of an intermediate metallic electrode .
we demonstrate that this latter is responsible of an effective increase of the internal energy transfer , up to a factor 5 to 44 at 300k according to the considered geometry , and more than a factor 200 for particular @xmath73-vectors .
the plasmonic resonance leads also to a large quenching of the emission by the organic , which reduce significantly the final internal efficiency . a way to improve
it further is to consider a patterned metal layer where most of the field will be directed in the dielectric , non absorbing holes in the metal , still taking advantage of the plasmonic enhancement .
the author would like to thank for valuable discussions and for hosting this research , and acknowledges fruitful discussions with and . | we consider the resonant energy transfer from a two - dimensional wannier exciton ( donor ) to a frenkel exciton of a molecular crystal overlayer ( acceptor ) when the active medias are separated by a metallic layer , possibly an electrode .
we characterize the effect of the surface plasmon on this process . using realistic values of material parameters
, we show that it is possible to change the transfer rate within typically a factor of 5 ( up to 44 according to geometrical configuration ) .
we then take into account the quenching of the organic luminescence due to the proximity with the metal .
this latter is significant and affect negatively the total internal efficiency that we discuss for different geometries . |
since the first observation of bose - einstein condensates ( becs ) in dilute weakly interacting gases of bosons in 1995 , great efforts both experimental and theoretical have been addressed on this subject . thus nowadays ,
thanks to the impressive progress of experimental techniques , becs represent a source of inspiration for new challenging problems in quantum physics . a notable aspect inherent in the dynamics of coupled becs
is that the nonlinear character of their equations of motion entails a particularly rich phenomenology where nonlinear effects including chaos can be studied in a quantum environment . among the latter , in recent experiments
have been observed phenomena such as super - fluidity @xcite , josephson tunneling @xcite and atom optics @xcite .
if the recent past has been prolific of experiments on becs arranged into arrays with huge number ( of the order of several hundreds ) of condensates , at the present , the experiments on chains with a few interacting becs ( of the order of several unities ) has received a limited attention @xcite-@xcite despite the increasing amount of theoretical work devoted to such systems . over the last few years ,
small_-chain models has been focused in particular on the two - well system ( dimer ) @xcite-@xcite and the three - well system ( trimer ) @xcite , @xcite while their many - well generalization has been studied in @xcite-@xcite .
the theoretical prediction of macroscopic effects such as coherent atomic tunneling @xcite , population self - trapping @xcite , and josephson junction effect @xcite , have been confirmed by recent observations in the experimental realization of the dimer @xcite-@xcite .
these results stimulate the theoretic study of the bec small chains in order to predict new macroscopic effects that , in addition to furnishing a deeper insight of stability properties and an increased control of systems , open the possibility of designing significant experiments . in this respect ,
the recent experimental technique discussed in @xcite and successfully applied to realize a two - well system provides an extremely effective tool for engineering small arrays .
the superposition of a parabolic confining potential to a linear optical lattice allows one to select two wells ( dimer ) with possibly different depths by changing the parabola position and amplitude .
likewise , a simple , suitable change of the latter should enable one to select @xmath0-well arrays , with @xmath0 arbitrary , whose ( local ) potentials feature a parabolic profile .
the trimer case is thus very promising in that , in addition to represent a practicable experimental objective , it exhibits behaviors typical of more complex arrays .
in fact , while the simplest bec array ( the dimer ) features an integrable dynamics @xcite , in the trimer case the apparently harmless addition of a further coupled condensate is sufficient to make the system nonintegrable . as a consequence ,
the trimer displays strong dynamical instabilities in extended regions of the phase space @xcite and a whole new class of behaviors including chaos . in general ,
unstable behaviors have been found in condensates subject to kicked pulses @xcite or to an anisotropic trapping potential @xcite and within soliton dynamics @xcite .
concerning the trimer , a conspicuous number of recent papers has addressed various aspects of its dynamics .
these include , within a mean - field approach , the occurrence of a transistor - like behavior @xcite , and the self - trapping @xcite in a triple - well asymmetric trap of repulsive bosons , the ring - trimer dynamics with attractive interaction in the presence of a single on - site defect @xcite , and the chaos control through an external laser pulse @xcite .
recently , the dynamical instability has been studied in terms of quasimomentum modes for a @xmath1-well semiclassical bose - hubbard model including a linear potential @xcite .
also , more related to purely quantum properties , second - quantized models of the trimer have been considered to study the properties of quantum states relevant to interwell particle exchange @xcite and entanglement @xcite , the semiclassical quantization of the bogoliubov spectrum @xcite , the structural changes of trimer eigenstates in terms of the tunneling parameter @xcite , and the occurrence of vortex - like states @xcite . in the present paper
we study the classical dynamics of an asymmetric open trimer ( aot ) made of three coupled becs arranged into an open chain geometry , where both the interwell tunneling amplitude @xmath2 and the relative depth @xmath3 of the central well , are regarded as adjustable parameters .
our objective is to evidence several macroscopic effects that can take place in the aot dynamics to provide a useful guide for future experiments . in section [ s : dyn ]
we introduce the semiclassical equations of the trimer mean - field dynamics . in section [
s : fp ] , after determining the solutions relevant to the fixed points of motion equations , i. e. the proper modes of trimer dynamics , we focus mostly on establishing , via standard procedures , their stability character for experimentally significant values of effective parameters @xmath4 and @xmath5 , where @xmath6 and @xmath7 are the effective interatomic scattering and the total boson number , respectively .
in section [ s : sd ] we review the ( three ) _ stability diagrams _
( one for each class of fixed points ) that summarize characters and properties of fixed points .
these supply an exhaustive , operationally valuable , account of the system stability properties and allow to determine the stable / unstable charater of the system evolution for initial conditions situated , in the phase space , close to a given fixed point . both the derivation and the use of such diagrams , rapidly sketched in @xcite , are discussed here ( and in the relevant appendices ) with some detail together with various new applications . by using the _ stability diagrams _ as a map for detecting the critical behaviors of trimer , in section [ s : macroeff ] we find parameter - tuned macroscopic effects ( both in the dimeric and nondimeric regime ) and evidence some features that may prove experimentally relevant .
such effects are shown to arise when crossing either the boundaries of regions with different stability characters or the boundaries of forbidden regions or , finally , by exploiting the coalescence / bifurcation mechanism characterizing certain fixed points . in section [ s : cdwi ] we consider the special case of @xmath8-like states relevant to the fixed points of the regime charaterized by a central depleted well .
we investigate the unexpected relation between the average population oscillations and the initial phase difference of the lateral condensates , suggesting a possible experiment .
finally , section [ s : conc ] is devoted to make some concluding remarks .
the quantum hamiltonian of the aot on a linear chain reads @xcite @xmath9\ ] ] where site boson operators @xmath10 satisfy standard commutators @xmath11 = \delta_{ik}$ ] , @xmath12 , @xmath13 is the energy offset of the wells , @xmath3 is the central - well relative depth , @xmath2 is the hopping amplitude and @xmath14 ( @xmath15 the scattering length , @xmath16 the boson mass ) is the on - site boson interaction .
the total boson number @xmath17 , commutes with hamiltonian ( [ e : ham3 ] ) .
the exact form of parameters @xmath6 and @xmath2 is calculated in @xcite showing how such quantities depend on @xmath16 , on the laser wavelength @xmath18 , and on the potential depth @xmath19 of potential @xmath20 determining the optical lattice .
the 3-well linear system , obtained by superposing the optical potential to parabolic potential @xmath21 confining bosons @xcite , @xcite , is expected to have lateral wells with depth @xmath13 smaller then the central - well depth @xmath22 due to the presence of the parabolic potential .
the dynamics of the aot with strong tunnelling amplitudes ( @xmath23 ) and large average numbers of atoms per well can be described by the mean - field form of ( [ e : ham3 ] ) .
the method developed in @xcite combines the coherent - state picture of quantum systems with the time - dependent variational approach @xcite providing the semiclassical version of quantum model ( [ e : ham3 ] ) in which the dynamical canonical variables are the expectation values @xmath24 , @xmath25 . the so - called trial macroscopic wave function @xmath26 is standardly assumed to be @xmath27 , where @xmath28 and @xmath29 are the glauber coherent states such that @xmath30 for each @xmath31 .
then , the resulting mean - field hamiltonian reads @xmath32 - w |z_2|^2-\frac{_t}{^2}[\ , z_2^ * ( z_1+z_3 ) + { \rm c.c . } ] \ , , \ ] ] which , through the poisson brackets @xmath33 , yields the dynamical equations @xmath34 the validity of such a semiclassical picture is confirmed experimentally @xcite , @xcite for the time scale ( hundreds of milliseconds ) of the macroscopic oscillations in a two - well system .
the effect of approximating the system state @xmath26 through a product of localized ( coherent ) states @xmath35 appears on a much longer time scale @xcite when @xmath26 must be expressed as a superposition of fock states @xmath36 ( @xmath37 ) to include ( quantum ) delocalization effects .
an early analysis of equations ( [ e : dyneq ] ) for @xmath38 is reported in @xcite , @xcite .
it is easy matter to check that system ( [ e : dyneq ] ) embodies the conservation of boson total number @xmath39 , namely @xmath40 .
constant @xmath7 can be exploited to introduce new rescaled variables @xmath41 whose evolution is governed by @xmath42 where @xmath43 and the prime denotes derivation with respect to the rescaled time variable @xmath44 .
this shows that the two significant parameters of the trimer dynamics are @xmath45 and @xmath46 ( namely , at fixed @xmath7 , @xmath47 and @xmath48 ) , parameter @xmath13 being incorporated in the time dependent factor @xmath49 of new variables @xmath50 .
furthermore , it is easy to verify that the condition @xmath51 is valid for @xmath52 ( or , equivalently , that @xmath53 holds for every @xmath54 ) if , initially , @xmath55 .
such a condition identifies an integrable subregime where the lateral condensates have the same population and phase ( see , e. g. , @xcite , @xcite and @xcite ) which in the following we shall refer to as _ dimeric regime_. we can get insight about the behaviour of the system in different dynamical regimes through the study of special exact solutions along with their stability character . with the latter we mean the simplest ( single mode ) periodic orbits in the phase space of the form @xmath56 . substituting the latter in equations ( [ e : dyneq2 ] ) provides the equations for variables @xmath57 .
such equations can be equivalently obtained from @xmath58 with @xmath59 , where the chemical potential @xmath16 , a lagrange multiplier , allows one to include the conserved quantity @xmath60 derived from the total boson number @xmath61 . owing to the condition @xmath62 we name fixed - point equations the equations determining @xmath63 , @xmath64 .
these have the form @xmath65 where @xmath66 and the fixed - point coordinates @xmath67 are normalized to unity : @xmath68 . in equation
[ e : fpsys ] we have exploited the system symmetry , inherited from equations [ e : dyneq2 ] , under a global phase shift @xmath69 to restrict coordinates @xmath67 ( which are in principle complex numbers ) to the real axis .
let us summarize the solutions of system ( [ e : fpsys ] ) .
_ central depleted well_. this parameter - independent solution is obtained when the central well is strictly depleted , that is @xmath70 , whereas the lateral ones have the same population and opposite phases . due to the constraint on the number
it must necessarily be @xmath71 , @xmath72 . in summary @xmath73
we shall refer to this solution as _ central depleted well _
( cdw ) . since configurations where either one or both of the lateral condensates have zero population are forbidden by equations ( [ e : fpsys ] ) , we are left with the following two possibilities .
_ dimeric solutions_. similar to ( [ e : dyneq ] ) and ( [ e : dyneq2 ] ) , equations ( [ e : fpsys ] ) possess dimeric solutions characterized by @xmath74 and @xmath75 .
these solutions can be expressed via the polar representation @xmath76 introducing this parametrization in ( [ e : fpsys ] ) one finds that the tangent @xmath77 of the angular coordinate @xmath78 identifying each fixed point is given by the real roots of @xmath79 with @xmath80 , @xmath81 . in the following
we shall refer to this class of solutions as _ dimeric fixed points _
( dfp ) whose derivation is discussed in [ s : dr ] . _ non - dimeric solutions_. the remaining single - mode solutions of system ( [ e : fpsys ] ) , characterized by @xmath82 , @xmath83 , will be referred to as _ non - dimeric fixed points _ ( nfp ) . in this case
it is appropriate to switch to the alternative set of coordinates @xmath84 , @xmath85 , and @xmath86 before setting @xmath87 in this case the radial coordinate @xmath88 reads @xmath89 and depends on the hamiltonian parameter @xmath46 and on parameter @xmath90 . by introducing this parametrization in system ( [ e : fpsys ] ) one finds that @xmath90 is given by the real roots of @xmath91 where @xmath92 , @xmath93 .
since the physical solutions of system ( [ e : fpsys ] ) have to be real it is clear that the roots which make @xmath94 or @xmath95 imaginary , must be discarded .
the next section contains a qualitative discussion of the three stability diagrams describing the entire set of solutions of system ( [ e : fpsys ] ) for any choice of @xmath46 and @xmath45 .
a detailed derivation of such fixed points is given in [ s : nr ] .
the stability diagrams introduced in @xcite allow one to know both number and stability character of the fixed points of trimer dynamics for any choice of significant parameters @xmath46 and @xmath45 .
in the previous section we have shown that vector @xmath96 describing a fixed point can be parametrized by the angular coordinate @xmath78 of the two - dimensional polar representations ( [ dfp ] ) , and ( [ ndfp_p ] ) and ( [ ndfp_ra ] ) for the dfp and the nfp , respectively . to determine parameter @xmath78 ( and thus the vector @xmath97 relevant to a given fixed point ) we must find the roots of the fourth - degree polynomial @xmath98 in the unknown quantity @xmath99 .
this allows one to determine @xmath100 in the dimeric case with @xmath101 ( see ( [ dfp_c ] ) ) , and @xmath102 in the non - dimeric case with @xmath103 ( see ( [ ndfp_pol ] ) ) . in the latter case vector @xmath104
is obtained by using the constraint @xmath105 .
therefore , for a given value of parameters @xmath46 and @xmath45 , the number of roots of ( [ genpol ] ) establishes how many dfp and nfp there exist .
this information is summarized in figures [ fig : dn ] and [ fig : nn ] .
the stability character of the fixed points solving equations ( [ e : fpsys ] ) is depicted in the stability diagrams of figures [ fig : ds ] and [ fig : ns ] for the dfp and the nfp case , respectively . in such figures
any point in the plane @xmath106 is implicitly associated to a definite value of parameter @xmath45 , owing to equation ( [ genpol ] ) .
this can be seen by rewriting ( [ genpol ] ) as @xmath107 = - \frac{\left[a_\nu \tan^2(\theta)+b_\nu\right]\tan(\theta)}{\tan^4(\theta)+c } \
, , \ ] ] which , for a given value @xmath108 , defines a curve @xmath109 $ ] in the plane @xmath78-@xmath46 consisting of two or more non - connected branches , both in the dimeric ( @xmath101 ) and the non - dimeric case ( @xmath103 ) . at the operational level , in both figures , for a given choice @xmath110 and @xmath111 ,
configuration parameter @xmath78 can be obtained by intersecting the curves @xmath109 $ ] , for the dimeric and non - dimeric case , and the straight line @xmath112 . once
( triplet @xmath46 , @xmath45 , @xmath78 of ) a fixed point has been identified , its stable / unstable character can be evinced from figures [ fig : ds ] or [ fig : ns ] thanks to the colour corresponding to a given value of @xmath78 and @xmath46 ( see the relevant figure captions ) .
the linear stability analysis determining the character of fixed points ( and , in practice , the whole structure of such phase diagrams ) is discussed in [ s : asd ] .
as an example , in figures [ fig : dn ] and [ fig : nn ] are reported with different dashing styles , five straight horizontal lines corresponding to several values of the @xmath45 .
the five curves drawn in figures [ fig : ds ] and [ fig : ns ] , refer to the straight lines drawn with the same dashing styles in figures [ fig : dn ] and [ fig : nn ] , for the dimeric and non - dimeric case , respectively .
notice that in general these curves consist of two or more non - connected branches . in the dimeric case
there are always two unbounded branches featuring an asymptote at @xmath113 or at @xmath114 . for @xmath115
$ ] a further , roughly parabolic , bounded branch appears .
this means that for any choice of the parameters @xmath46 and @xmath45 there are always at least two dfps , and that , for suitable choices of these parameters , two more fixed points appear . in the non - dimeric case ,
figure [ fig : ns ] , the curves @xmath116 $ ] consist of either one or two bounded parabola - like branches , hence featuring zero , two or four intersection with a constant-@xmath46 straight line .
however , for certain values of the parameters , not all of these intersections represent fixed points , i.e. real solutions of system ( [ e : fpsys ] ) . indeed , as it is discussed in [ s :
nr ] , some of the real roots of polynomial @xmath117 more precisely the ones lying in the patterned regions of figure [ fig : ns] result in complex - valued solutions of system ( [ e : fpsys ] ) and therefore must be discarded .
in particular , the patterned ( forbidden ) regions of figure [ fig : ns ] can be shown to correspond to the white region in the @xmath46-@xmath45 plane of figure [ fig : nn ] where no solution is permitted .
figures [ fig : ds ] and [ fig : ns ] also supply important information concerning fixed points and their characteristic energies . referring to the different colours used in such figures ,
a fixed point is a ( stable ) minimum , a stable saddle , an unstable saddle or a ( stable ) maximum of the energy function depending on whether the region it lies in is white , light grey , medium grey or dark grey .
as we mentioned above , other than the dfps and the nfps , a further fixed point , the cdw configuration , is present for any value of @xmath46 and @xmath45 .
the parameter independence of the relevant configuration @xmath118 , makes the use of a polar representation unnecessary , and allows one to draw its stability diagram in the parameter plane @xmath46-@xmath45 in a direct way .
this is illustrated in figure [ fig : cdw ] .
a careful look at stability diagrams [ fig : ds ] and [ fig : ns ] , in addition to provide , at first glance , number and properties of the fixed points of aot dynamics for any choice of @xmath46 and @xmath45 , allows one to single out at least three ways to produce macroscopic , and hence experimentally observable , effects by means of a simple change of model parameters .
_ crossing boundaries of regions with a different character . _ for suitable choices of @xmath45 , curves @xmath116 $ ] may cross the boundary dividing regions with different stability properties .
thus , by slightly and adiabatically changing the hamiltonian parameters ( e.g. the optical potential amplitude ) the system can modify significantly its behavior .
this is the case of ( one of ) the asymptotic branches of the dimeric curves , which , for any @xmath119 , crosses the boundary between the maxima and the unstable saddles region ( see , e. g. , the branch in figure [ fig : dn ] corresponding to @xmath120 in figure [ fig : nn ] ) . likewise , in the non - dimeric case , one of the parabola - like branches of curves @xmath116 $ ] always crosses the interesting unstable `` isthmus '' splitting the large stable ( light - grey ) region in figure [ fig : ns ] , provided that @xmath121 . _
coalescence_/_bifurcation of fixed points .
_ at the `` tips '' of the parabola - like branches of curves @xmath116 $ ] , a _ coalescence_/_bifurcation _ _ effect _ of pairs of fixed points takes place . hence , for some values @xmath122 of the hopping parameter it is possible to find a value @xmath123 with two coincident solutions of fixed - point equations located at the parabolic - branch tip . by varying @xmath45 ,
the branch tip moves away from @xmath122 , and the coincident solutions either split ( i.e. bifurcate ) or disappear accordingly .
let us emphasize that in the non - dimeric case ( figure [ fig : ns ] ) the coalescence phenomena always involve fixed points with different stability characters , whereas , in the dimeric case ( figure [ fig : ds ] ) only the coalescence of two fixed points of the same kind may happen , provided that @xmath124 ( see next section ) . _ crossing forbidden - region boundaries .
_ significant changes in the phase - space structure of the system are expected as well where the non - dimeric curves @xmath116 $ ] cross the boundaries of forbidden zones . indeed , in this case , a slight change in the parameters may cause the sudden appearance / disappearance of a fixed point . as to the cdw fixed points , which according to the hessian signature is always a saddle , figure [ fig : cdw ] clearly shows that it may display an either stable or unstable character , depending on the choice of parameters @xmath46 and @xmath45 based on these three mechanisms , we discuss the results of several numeric simulations evidencing how macroscopic effects are induced in the aot dynamics by slight changes in parameters @xmath46 and @xmath45 ( similar effects on the stability of the many - well system with @xmath125 , of vortexlike modes in a 1d array and of localized modes in a 2d array have been studied in @xcite , @xcite and @xcite , respectively ) .
these entail that trajectories based on the same initial configuration may exhibit an either regular or chaotic behavior .
the character of the aot dynamics , can be conveniently investigated by constructing appropriate poincar sections within the four - dimensional _ reduced _ phase - space spanned by the ( _ reduced _ ) set of canonical variables @xmath126 .
this reduction is obtained via the canonical transformation @xmath127 @xmath128 , where @xmath129 once the condition @xmath130 is taken into account and one notices that the @xmath131 evolution is nonautonomous ( see @xcite for details on the reduced dynamics ) .
we have numerically integrated the trimer equations of motion in the reduced coordinates , by a first - order bilateral symplectic algorithm .
moreover , we have qualitatively investigated the degree of regularity of the ensuing trajectories by means of poincar sections .
the latter are obtained by projecting the points belonging to a given submanifold , characterized by a given value of one of the four reduced variables , onto the plane spanned by two of the three remaining variables .
all of the poincar sections reported in the following refer to a set of trajectories based on an array of starting points surrounding a significant configuration of the system , usually a fixed point , of reduced coordinates @xmath132 . in every instance
the points belonging to the submanifold @xmath133 have been projected onto the @xmath134-@xmath135 plane .
the first macroscopic effect we discuss here has been briefly discussed in @xcite and concerns the dimeric configuration @xmath136 , with @xmath137 bosons in each lateral well and the remaining @xmath138 in the central one which is represented by the angular variable , @xmath139 .
one can show that the latter is a ( dimeric ) fixed point provided that @xmath140 .
furthermore , such a fixed point is either a maximum or an unstable saddle depending on whether @xmath45 is smaller or larger than the critical value @xmath141 .
this is clearly visible in figure [ fig : dexp ] , where two different parameter choices @xmath142 and @xmath143 , shown with the letters @xmath0 and @xmath144 , are considered .
the crossing of the boundary separating the stable and unstable regions has evident consequences .
the poincar sections ( reported in @xcite ) of trajectories , based on a set of points closely surrounding the fixed point under concern @xmath145 , mirror the dramatic change in the dynamical behaviour involved in such a slight parameter change . while choice @xmath142 results in the neat and regular poincar sections characterizing a stable fixed point , with choice @xmath143 sections densely fill an area considerably larger than the previous example , despite the relevant trajectories are based on the very same points @xcite .
an instability outbreak similar to that discussed above is found in the neighborhood of the `` instability isthmus '' of the non - dimeric stability diagram [ fig : ns ] .
indeed , the possibly present upper branch of the non - dimeric @xmath46 curve always crosses such unstable region .
this is the case of four of the five curves plotted in figure [ fig : ns ] and of the curve with @xmath146 plotted in figure [ fig : gexp ] .
the inset of figure [ fig : gexp ] is a magnification showing fixed points @xmath147 , @xmath148 and @xmath149 , located in the proximity of the instability `` isthmus '' for three slightly different values of @xmath46 ( @xmath150 , @xmath151 and @xmath152 ) .
these fixed points correspond to ( slightly ) different values of coordinate @xmath78 , thus the corresponding configurations are quite similar to each other . in these three cases
more than @xmath153 of the total population is almost equally shared by the central and one of the lateral condensates .
the phases of these condensates are parallel to each other and opposite to that of the remaining , almost empty , condensate .
this kind of configuration corresponds to the neighborhood of the reduced - phase - space point with @xmath154 .
figure [ fig : nexp ] shows the poincar sections based on a set of points closely surrounding it , for the three choices of @xmath46 listed above , @xmath155 , @xmath156 and @xmath157 . the first and last choice corresponding to trajectories based in the neighborhood of a stable fixed point are both characterized by extremely regular sections where the populations `` flicker '' about their initial configuration . in the remaining case
, the fixed point belongs to the `` unstable isthmus '' ( see figure [ fig : gexp ] ) , and the consequent irregular trajectories produce a quite different pattern .
also in the present case , a slight change in the parameters destroys the steady , flickering populations .
once again the poincar sections are indistinguishable from one another and densely fill an area considerably larger than in the previous cases . [
cols="^,^,^ " , ]
the last case considers several unexpected and interesting features of cdw fixed points .
remarkably , these fixed points are always present independently from the value of the parameters @xmath46 and @xmath45 , and have a simple configuration where the entire boson population is equally shared between the lateral condensates , being the phases opposite to each other . figure [
fig : cdw ] shows that the cdw are saddle points and are expected to exhibit a stable character in extended regions of the parameter plane .
this fact , is widely confirmed by our numeric simulations where , for initial configurations sufficiently close to the cdw fixed point , we observe a regular dynamics exhibiting ( possibly quite complex ) periodic oscillations of the three condensate populations . in particular , the lateral condensates cyclically exchange small amounts of bosons , so that their population imbalance oscillates . whereas the central condensate , whose population oscillates with a considerably smaller amplitude and a frequency twice than that of the lateral condensates , participates in the dynamics acting as a bridge between the lateral ones .
figure [ fig : cdwsu ] shows the evolution of a configuration quite close to the cdw for two parameter choices belonging to different stability regions of figure [ fig : cdw ] .
as expected , in the stable case ( panels * a * and * b * ) , the amplitude of the population oscillations increases with increasing displacement of the initial configuration from the cdw .
of course , in general , such displacement can not be increased indefinitely : the initial configuration would eventually leave the `` domain of influence '' of the cdw and the consequent trajectory would be determined by the stability character of a different fixed point .
the same considerations apply to configurations obtained from the cdw by setting the phase difference between the lateral condensates to a value @xmath158 and maintaining the total boson population equally shared by the lateral condensates . in this case
the oscillation amplitude is a function of @xmath159 with a minimum at @xmath160 , where it vanishes .
our numeric simulations confirm these predictions .
in fact , for sufficiently small @xmath46 s , regular oscillatory dynamics is observed over the entire range of @xmath161 .
this is clearly shown in figure [ fig : cdwpops ] , displaying the population oscillations pertaining to initial configurations identified by several values of @xmath161 .
a further important feature of the cdw fixed points is highlighted by figure [ fig : xpamp ] showing that for small values of @xmath46 , the amplitude @xmath162 of the oscillation of the population @xmath163 exhibits a quite simple dependence on the phase difference @xmath161 of the relevant initial configuration .
more precisely @xmath162 linearly depends on @xmath164 .
a rigorous discussion of such dependence is beyond the scope of the present article .
however , some insight in this sense can be gained by adopting an approach similar to that developed in @xcite in the study of the stability of the dimeric integrable subregime .
interestingly , the same linear behavior as in figure [ fig : xpamp ] is obtained by relaxing the requirement that the entire boson population is equally shared by the two lateral condensates .
this is clearly shown in figure [ fig : xpav ] , where we plotted the average oscillation amplitude relevant to initial configurations characterized by a certain phase difference @xmath161 and condensate populations @xmath165 slightly displaced from that of the cdw fixed point .
the plots displayed in figures [ fig : xpamp ] and [ fig : xpav ] suggest the possibility of measuring the phase difference between two equally populated condensates .
let us stressing the fact that , in order to profitably exploit this quantum effect to measure the difference of phase between two condensates , we need of a nondestructive real - time probing of the condensates such as the technique employed to image dimer - like condensates confined by magnetic traps @xcite .
the present paper has been inspired by the wish of gaining a deeper insight both on the complexity of three - well dynamics ( which , as noted above , is the paradigm of the behavior one can expect in more complex becs arrays ) and on the possibility of detecting macroscopic effects amenable to experimental observation .
recently , it has been proposed that the superposition of multiple laser beams could allow to create unusual optical lattices in a controlled manner @xcite . in this respect , we observe that a suitable choice of laser wavelengths could generate a superlattice whose elementary cell is characterized by a three - well potential such as the one addressed in the present paper .
the further minimization of the tunneling between subsequent trimeric cells should provide an alternative experimental realization of the system under concern . a nice example in
this sense is the trimerized kagom optical lattice @xcite providing ( multiple copies of ) the trapping potential typical of a symmetric closed trimer . in the present paper we focus on the rich phenomenology offered by the trimer and on the possibility of triggering macroscopic effects through a suitable choice of the adjustable parameters , relegating as much as possible to the appendices the technical details necessarily involved in a systematic analysis .
the macroscopic effects we evidence in this paper are obtained through a careful analysis of the stability diagrams introduced in @xcite , whose operational value is reviewed in section [ s : sd ] .
their derivation is thoroughly discussed in [ s : asd ] .
such diagrams describe the trimer phase space in terms of the stability character of dynamical fixed points for any choice of the significant parameters .
this wealth of information enriches the recent literature about the dynamical stability of mesoscopic bec arrays @xcite-@xcite , @xcite-@xcite , @xcite and supplies an essential tool in designing experimental realizations of the trimer dynamics .
also , at the theoretical level , it provides a sound basis for gaining a deeper insight into the quantum counterpart of classical dynamical instability , which is far from being fully understood .
the use of stability diagrams leads to identify in section [ s : macroeff ] three different mechanisms able to modify in a substantial way the trimer dynamics .
these are exploited in sections [ s : priming ] , [ s : losscoher ] and [ s : instsupp ] to prime instability in breather - like regular oscillations , loss of coherence in anti - breather states , and suppression of instability due to fixed - point disappearance , respectively .
macroscopic dynamical oscillations are shown to be triggered by means of slight parameter changes thus providing a stimulating basis for planning actual experiments in the scheme described in @xcite-@xcite .
the analysis of the stability diagrams in section [ s : cdwi ] also evidences that a simple configuration exists allowing to relate in a direct manner the initial phase difference between two condensates to the amplitude of the ensuing population oscillations .
this suggests a rather simple scheme aimed at testing experimentally the limits of validity of the mean - field picture , where each condensate is endowed with a macroscopic phase .
the study of the fixed points inherent in the classical picture reported in sections [ s : dyn ] and [ s : sd ] proves useful also in unraveling the complex structure of the energy spectrum of the quantum trimer , in that it allows to recognize a ( either exact or approximate ) mapping with a simpler dimeric system @xcite .
the mapping is shown to persist also at the quantum level , thus extending the results found in @xcite to the case of the quantum trimer .
these results provide a useful starting point in investigating the quantum counterpart @xcite of classical dynamical instability .
a deeper insight into the relation between the classical and quantum picture is certainly useful for the approaches to this issue based on the dynamical algebra method @xcite or on the measure of quantum - state entanglement @xcite .
the linear stability character of a fixed point @xmath166 can be determined by studying the eigenvalues of the matrix associated to the ( linearized ) equations governing the dynamics of the ( small ) displacements @xmath167 from the fixed point itself , where @xmath168 .
the hamiltonian , up to the second order in the small displacements @xmath169 , reads @xmath170 where @xmath171 and @xmath172 .
as it is well known , linear instability occurs when at least one of such ( in principle complex ) eigenvalues features a positive real part .
the linearized equations obtained from hamiltonian ( [ e : hpq ] ) through poisson s brackets @xmath173 ( these stem from the original ones @xmath174 ) have the form @xmath175 in accordance with the expected time reversal symmetry of the linear dynamics , the six ( complex ) eigenvalues of matrix @xmath176 come in pairs featuring the same modulus and opposite signs .
more precisely each pair of eigenvalues of @xmath176 consists in the signed square roots of an eigenvalue of the three - by - three matrix @xmath177 .
indeed , due to the diagonal block structure of @xmath176 , the relevant eigenvalue equation is @xmath178 , where in writing the characteristic polynomial of @xmath177 we took into account the fact that , similar to @xmath179 , such a matrix has a zero eigenvalue .
this feature is the consequence of the fact that both @xmath180 and @xmath181 are constrained by the condition @xmath182 . according to @xmath183
the fixed point is ( linearly ) stable provided that the three conditions @xmath184 are simultaneously met . indeed in this case
the two non - zero eigenvalues of @xmath177 are real and non - positive , so that the four non - zero eigenvalues of @xmath176 namely their signed square roots are purely imaginary .
conversely , whenever an eigenvalue of @xmath185 is complex or real positive , one of its square roots has necessarily a positive real part , giving rise to ( linear ) instability .
the following subsections describe in some detail the construction of the stability diagrams for the dfp , the nfp and the cdw . as illustrated in section [ s : dyn ] , the dimeric fixed points ( dfp ) pertain to the integrable dimeric subregime @xcite of the dynamics described by system ( [ e : dyneq ] ) , which is characterized by @xmath186 .
in this situation the fixed point configuration @xmath187 is described by @xmath188 and @xmath189 ( see eqs .
( [ dfp ] ) ) , where the tangent @xmath190 must be a real root of polynomial @xmath191 ( see formula ( [ dfp_c ] ) ) with @xmath80 , @xmath81 . before proceeding to a more detailed discussion of the dfp case , we remark that the previous polar representation of @xmath57 s allows one to readily relate variable @xmath78 to the fixed point @xmath187 .
particularly , @xmath192 shows how @xmath193 corresponds to a situation where the central condensate is almost depleted , so that the total population is ( equally ) shared by the lateral condensates .
conversely , when @xmath194 the lateral condensates are almost empty and the total population is concentrated in the central one .
notice also that the sign of @xmath78 is similarly strictly related to the phase differences between the central condensate and the lateral ones .
more precisely such phase differences are 0 or @xmath195 depending on whether @xmath78 is positive or negative . as we discussed above , solving the fixed point equations for the dfp amounts to finding the real roots of the fourth - degree polynomial @xmath196 ( henceforth denoted by @xmath197 ) .
four real solutions are found from @xmath198 for @xmath199^{{3}/{2}}\ ] ] and two real solutions elsewhere in the parameter plane , as it is illustrated in figure [ fig : dn ] . in section [ s :
sd ] , we showed how equation @xmath198 suggests that , for any choice of parameters @xmath200 , @xmath201 , the dfp can be found in the @xmath106 plane from the intersection of the parametric curve ( see [ e : gtauc ] ) @xmath202 with the vertical straight line @xmath112 .
the curve @xmath203 features two asymptotes at @xmath204 ( @xmath205 ) .
this is in agreement with the fact that polynomial @xmath206 has no less than two real solutions . in particular
it is easy to verify that when @xmath207 the values @xmath208 are solutions of @xmath198 .
hence , in this case , the two unbounded branches coalesce with their asymptotes , thus providing @xmath46-independent solutions corresponding to the configurations @xmath209 . more generally , when @xmath210 the branches of the curve @xmath203 , and hence the roots of @xmath198 , are characterized by a reflection symmetry with respect to the nearest asymptote ( see figure [ fig : ds ] ) .
this can be seen by noticing that @xmath211 = \frac{2\,\bar\nu\mp3 \sin(2\vartheta)}{\sin(2\vartheta ) } \frac{\cos(2\vartheta)}{2\sqrt 2}\ , , \
, \vartheta \in \left [ -\frac{\pi}{4},\frac{\pi}{4 } \right ] \ ] ] where the auxiliary angle @xmath212 describes the positive ( negative ) solutions @xmath213 ( @xmath214 ) .
one easily verifies that @xmath215 = \tau_d \bigl [ \tan(\pm { \pi}/{4}-\vartheta ) ; { 1}/{2}-\bar\nu \bigr ] $ ] .
_ local character of the dfp_. this information is obtained by exploiting the constraint @xmath216 and the fixed point equations ( [ e : fpsys ] ) .
after expressing the quantities appearing in matrices @xmath217 and @xmath179 ( see ( [ e : mqp ] ) ) as @xmath218 and @xmath219 , the analysis of the signature of both matrix @xmath217 and @xmath179 depending on parameter @xmath220 leads to the following conclusions . for any choice of the parameters @xmath46 and @xmath45 ,
a dfp is a minimum if the corresponding angular coordinate , @xmath78 , is positive .
if , conversely , @xmath221 the fixed point is either a maximum or a saddle point depending on whether @xmath46 is greater or less than @xmath222 where @xmath223 $ ] with @xmath224 $ ] and @xmath225 @xmath226 $ ] .
combining such results with the linear stability analysis based on equation @xmath183 ( defined after formula ( [ e : m11 ] ) ) confirms that both maxima and minima are , as expected , stable fixed points , whereas all the saddles of the dimeric regime are unstable .
the situation for the dfp is summarized in the corresponding _ stability diagram _ displayed in figure [ fig : ds ] .
the differently shaded region pertain to fixed points with a different stability character . since in this case @xmath82 , the polar representation of nondimeric fixed points ( nfp ) is not as simple as that of the dfp case .
still , one can take advantage from the conservation of the total boson number .
some algebraic manipulations of the fixed - point equations ( see _ non - dimeric solutions _ in section [ s : fp ] ) suggest to switch to the new coordinates @xmath227 @xmath228 , @xmath85 and @xmath86 , obeying the constraint @xmath229 , before setting @xmath230 and @xmath231 . this way @xmath232 and radial coordinate ( [ ndfp_ra ] )
are shown to depend on @xmath78 and @xmath46 as @xmath233 with @xmath234 , and the fixed points equations are equivalent to the equation for the roots of polynomial @xmath235 , ( henceforth denoted by @xmath236 ) , where @xmath92 , @xmath93 ( see [ ndfp_pol ] ) . recalling that we are ultimately looking for real solutions of system ( [ e : fpsys ] ) it is clear that the roots which make @xmath94 or @xmath237 imaginary must be discarded . before analysing how nfp are identified
, it is important to remark that in the nfp case the relation among the auxiliary variable @xmath78 and the configuration of the fixed point is not as transparent as in the dfp case , basically due to the @xmath78 dependence of the radial variable ( [ e : rg ] ) .
nevertheless , within the nfp stability diagram some special regions can be still recognized .
notice indeed that , according to the previous definitions of @xmath238 , @xmath239 and @xmath240 as well as to ( [ e : rg ] ) , the ( forbidden ) limit situations @xmath241 and @xmath242 correspond to the dimeric case @xmath243 and to the cdw case @xmath244 , @xmath70 , respectively . as we shall discuss below , these limit situations are actually the boundaries of two forbidden regions in the nfp diagram ( see ( [ e : forb ] ) and discussion ) . due to the restrictions on @xmath88 and @xmath240 , the diagram of figure [ fig : nn ] , giving the number of fixed points relevant to system ( [ e : fpsys ] ) on varying parameters @xmath46 and @xmath45 , is significantly richer than the corresponding diagram for the number of real roots of polynomial @xmath236 .
indeed the latter consists of four regions , delimited by the solid lines appearing in figure [ fig : nn ] .
two of these regions correspond to situations where @xmath245 features only two real roots , whereas within the remaining regions the roots are either four or none .
despite they are real , some of these roots can make either @xmath94 or @xmath246 complex .
hence some new regions appear within the ones delimited by the solid curves , where the number of solutions must be decreased by either one or two units .
as discussed in section [ s : sd ] , the features of diagram ( [ fig : nn ] ) can be better understood by resorting to the @xmath106 plane , where , similar to the dfp case , for a given choice @xmath112 , @xmath247 , the real roots of polynomial @xmath236 correspond to the intersections of curve @xmath248/(3{\alpha}^4 + 1)\ , , \quad \nu = \tilde\nu \ , , \ ] ] with the straight line @xmath112 . in the half plane @xmath249
such curve consists in general of two bounded , roughly bell - shaped , branches .
more precisely they are defined within the @xmath78 intervals @xmath250 $ ] and @xmath251 $ ] , respectively , where @xmath252 @xmath253 .
when @xmath254 ( @xmath255 ) the former ( latter ) branch , which from now on will be referred to as the negative ( positive ) branch , disappears .
these features readily explain the four regions in the diagram describing the number of real roots of @xmath236 .
hence , provided @xmath256 is sufficiently small , there are either four or two solutions depending on whether @xmath257 or @xmath258 .
indeed , for any straight line at a given value of @xmath46 , there exist for each branch a value of @xmath45 such that the branch and the straight line are tangent . denoting such values @xmath259 , where the superscript labels the branch , it is easy to check that @xmath260 is always a growing function of @xmath46
. hence two solutions relevant to the negative ( positive ) branch are always present for @xmath261 [ @xmath262 .
notice further that the presence of a region of parameters where four solutions are found implies that there exists some @xmath263 such that @xmath264 .
then it is clear that @xmath236 features four real roots in the region @xmath265 , no real roots in the region @xmath266 and two real roots elsewhere .
the explicit expressions of functions @xmath267 and @xmath268 for @xmath269 and @xmath270 , respectively , ( solid lines in figure [ fig : ns ] ) can be easily described in a parametric way by means of two curves @xmath271 , @xmath272 : the curve @xmath273 intersects @xmath274 at the tips of its bell - shaped branches while @xmath275 is such that @xmath276 at @xmath277 and @xmath278=\tau_t({\alpha})$ ] .
the requirement that @xmath88 , @xmath279 determines two forbidden regions in the stability diagram:@xmath280 and @xmath281 , where @xmath282 ( see figure [ fig : ns ] ) .
the information about the forbidden regions can be visualized in parallel in the @xmath283 plane by drawing the curves @xmath284 and @xmath285 describing the intersections between @xmath286 and the borders of the forbidden regions , @xmath287 and @xmath288 , respectively ( dotted curves of figure [ fig : nn ] ) . according to what we said so far ,
these curves enclose two regions where one of the real roots of polynomial @xmath236 must be excluded from the set of solutions of system ( [ e : fpsys ] ) : @xmath289 and @xmath290 .
similarly two of these real roots must be discarded in the regions @xmath291 and @xmath292 .
notice also that @xmath293 [ @xmath294 is expectedly tangent to @xmath268 [ @xmath295 at @xmath296 , [ @xmath296 ] .
_ local character of the nfp_. the study of the signature of both matrix @xmath217 and @xmath179 defined in ( [ e : mqp ] ) reveals that nfps are either maxima or saddle points depending on whether they lie within the region @xmath298 or elsewhere in the allowed zone . in parallel , the linear stability analysis based on polynomial @xmath299 ( see the discussion relevant to ( [ e : st_cond ] ) ) discloses the ( linear ) stability of most of the saddle points lying in the positive @xmath78 quadrant , namely the ones belonging to the light grey region of figure [ fig : ns ] .
indeed , within the allowed region @xmath300 , the stability condition ( [ e : st_cond ] ) fails to apply in the region @xmath301 , where @xmath302 , and in a peculiarly shaped region of the positive @xmath78 quadrant , where @xmath303 .
since @xmath304 is positive everywhere outside these regions , it turns out that linear stability characterizes not only , as expected , the maxima , but also most of the saddle points at positive @xmath78 .
as discussed in section [ s : dyn ] , there is only one parameter - independent cdw fixed point : @xmath305 , which , according to the signature of matrices @xmath217 and @xmath179 ( see [ e : mqp ] ) , is always a saddle point .
indeed , no matter for the sign of @xmath306 , the quadratic form has always an undefined signature . recalling that @xmath307 , the coefficients involved stability condition ( [ e : st_cond ] ) have the quite simple form @xmath308/{\tau^2}$ ] and @xmath309/{\tau^2}$ ] yielding @xmath310/\tau^4 $ ] .
hence , introducing the functions @xmath311}\ ] ] the cdw saddle is actually stable in the parameter regions @xmath312 , @xmath313 and @xmath314 , and unstable elsewhere .
notice that the former zone can be also described as @xmath315 , where @xmath316 is strictly related to the boundary of one of the forbidden zones of the nfp stability diagram .
actually , as we already noticed above , the points lying on such ( forbidden ) boundary @xmath317 have a cdw form , since @xmath318 there ( see ( [ e : forb ] ) and the defining formulas for @xmath319 , @xmath320 in section [ s : nr ] ) .
the stability diagram for the cdw fixed point is displayed in figure [ fig : cdw ] .
notice that , unlike figures [ fig : ds ] and [ fig : ns ] , it is in the plane of the parameters , @xmath46 and @xmath45 .
this is possible because , as we already remarked , there is always a single , parameter - independent cdw configuration .
99 oelkers n , and links j 2007 _ phys . rev .
b _ * 75 * 115119 wu b , and niu q 2001 _ phys . rev .
a _ * 64 * , 061603(r ) paraoanu g s 2003 _ phys .
a _ * 67 * , 023607 kalosakas g , rasmussen m , and bishop a r 2002 _ phys . rev
. lett . _ * 89 * , 030402 raghavan s , smerzi a , fantoni s , and shenoy s 1999 _ phys .
rev . a _ * 59 * , 620 aubry s , flach s , kladko k , and olbrich e 1996 _ phys .
lett . _ * 76 * , 1607 franzosi r , and penna v 2003 _ phys .
e _ * 67 * , 046227 stickney j a , anderson d z , and zozulya a a 2007 _ phys . rev . a _ * 75 * 013608 liu b , fu l - b 2007 _ phys . rev .
a _ * 75 * , 033601 pando c l and doedel e j 2007 _ phys .
e _ * 75 * , 016213 + pando c l and doedel e j 2005 _ phys .
e _ * 71 * 056201 chong g , hai w , and xie q 2005 _ phys .
e _ * 71 * , 016202 kolovsky a r , korsch h j and graefe e m _ arxiv _ : 0901.4719v1 @xmath321quant - ph@xmath322 $ ] mossmann s and jung c 2006 _ phys .
a _ * 74 * 033601 kolovsky a 2007 _ phys .
lett . _ * 99 * , 020401 hiller m , kottos t , and geisel t 2006 _ phys .
a _ * 73 * , 061604 lee c , tristram j. alexander , and kivshar y s 2006 _ phys .
* 97 * , 180408 | we study the dynamical stability of the macroscopic quantum oscillations characterizing a system of three coupled bose - einstein condensates arranged into an open - chain geometry .
the boson interaction , the hopping amplitude and the central - well relative depth are regarded as adjustable parameters . after deriving the _ stability diagrams _ of the system , we identify three mechanisms to realize the transition from an unstable to stable behavior and analyze specific configurations that , by suitably tuning the model parameters , give rise to macroscopic effects which are expected to be accessible to experimental observation . also , we pinpoint a system regime that realizes a josephson - junction - like effect . in this regime
the system configuration do not depend on the model interaction parameters , and the population oscillation amplitude is related to the condensate - phase difference .
this fact makes possible estimating the latter quantity , since the measure of the oscillating amplitudes is experimentally accessible . |
recently meyer et al . @xcite carried out high - precision measurements of the total cross sections near threshold for the reaction @xmath1 these measurements were confirmed by bondar et al . @xcite .
the early theoretical calculations @xcite underestimate these s - wave @xmath2 production cross sections by a factor of @xmath35 .
the basic features of these early calculations may be summarized as follows .
the pion production reactions are assumed to be described by the single nucleon process ( the born term ) , fig.1(a ) , and the @xmath4-wave pion rescattering process , fig.1(b ) .
the @xmath5-@xmath6 vertex for the born term is assumed to be given by the pseudovector interaction hamiltonian @xmath7 where @xmath8 is the axial coupling constant , and @xmath9 = 93 mev is the pion decay constant .
the first term represents @xmath10-wave pion - nucleon coupling , while the second term accounts for the nucleon recoil effect and makes @xmath11 galilean - invariant " . for @xmath4-wave pion production
only the second term contributes . since this second term is smaller than the first term by a factor of @xmath12 , the contribution of the born term to @xmath4-wave pion production
is intrinsically suppressed , and as a consequence the process becomes sensitive to two - body contributions , fig.1(b ) .
the s - wave rescattering vertex in fig.1(b ) is commonly calculated using the phenomenological hamiltonian @xcite @xmath13 the two coupling constants @xmath14 and
@xmath15 in eq.([h1 ] ) were determined from the @xmath16 and @xmath17 pion nucleon scattering lengths @xmath18 and @xmath19 as [ eq:4 ] @xmath20
@xmath21 the current algebra prediction @xcite for the scattering lengths , @xmath22 , implies that only chiral symmetry breaking terms will give a non - vanishing value of the coupling constant @xmath14 in eq.([h1 ] ) .
therefore @xmath14 is expected to be very small .
indeed , the empirical values @xmath23 and @xmath24 obtained by hhler _ et al._@xcite lead to @xmath25 and @xmath26 .
so the contribution of the @xmath14 term in eq.([h1 ] ) is significantly suppressed .
meanwhile , although @xmath15 is much larger than @xmath14 , the isospin structure of the @xmath15 term is such that it can not contribute to the @xmath2 production from two protons at the rescattering vertex in fig.1(b ) .
thus , the use of the phenomenological hamiltonians , eqs.([h0 ] ) and ( [ h1 ] ) , to calculate the born term and the rescattering terms illustrated in figs.1(a ) and ( b ) , gives significantly suppressed cross sections for the @xmath0 reaction near threshold .
therefore , theoretically calculated cross sections can be highly sensitive to any deviations from this conventional treatment .
these delicate features should be kept in mind in discussing the large discrepancy ( a factor of @xmath35 ) between the observed cross sections and the predictions of the earlier calculations .
a plausible mechanism to increase the theoretical cross section was suggested by lee and riska@xcite .
they proposed to supplement the contribution of the pion - exchange diagram , fig.1(b ) , with the contributions of the short - range axial - charge exchange operators which were directly related to heavy - meson exchanges in the nucleon - nucleon interactions @xcite . according to lee and riska , the shorter - range meson exchanges ( scalar and vector exchange contributions )
can enhance the cross section by a factor 35 .
subsequently , horowitz _ et al .
_ @xcite demonstrated , for the bonn meson exchange potential , a prominent role of the @xmath27 meson in enhancing the cross section , thereby basically confirming the conclusions of lee and riska .
the possible importance of heavy - meson exchanges may be inferred from the following simple argument .
consider fig.1(b ) in the center - of - mass ( cm ) system with the initial and final interactions turned off and with the exchanged particle allowed to be any particle ( not necessarily a pion ) . at threshold , @xmath28 , @xmath29 , @xmath30 , so that any exchanged particle must have @xmath31 mev and @xmath32 mev / c , which implies @xmath33 .
thus the rescattering process probes two - nucleon forces at distances @xmath3 0.5 fm corresponding to a typical effective exchanged mass @xmath34 mev .
its sensitivity to the intermediate - range @xmath6-@xmath6 forces indicates the possible importance of the two - body heavy meson axial exchange currents considered by lee and riska .
the particular kinematical situation we considered here shall be referred to as the _ typical threshold kinematics_. meanwhile , hernndez and oset @xcite considered the _ off - shell _ dependence of the @xmath35 @xmath4-wave isoscalar amplitude featuring in the rescattering process , fig.1(b ) .
they pointed out that the @xmath4-wave amplitude could be appreciably enhanced for off - shell kinematics pertinent to the rescattering process .
we have seen above that , for the _ typical threshold kinematics _ , the exchanged pion can indeed be far off - shell .
the actual kinematics of course may deviate from the _ typical threshold kinematics _ rather significantly due to energy - momentum exchanges between the two nucleons in the initial and final states , but the importance of the off - shell kinematics for the exchanged pion is likely to persist .
hernndez and oset examined two types of off - shell extrapolation : ( i ) the hamilton model for @xmath35 isoscalar amplitude based on @xmath27-exchange plus a short range piece@xcite , and ( ii ) an extrapolation based on the current algebra constraints . in either case the enhancement of the total cross section due to the rescattering process was estimated to be strong enough to reproduce the experimental data . a more detailed momentum - space calculation carried out by hanhart _
et al._@xcite supports the significant enhancement due to an off - shell effect in the rescattering process , although the enhancement is not large enough to explain the experimental data .
it should be emphasized that hanhart _
s calculation eliminates many of the kinematical approximations employed in the previous calculations .
given these developments based on the phenomenological lagrangians , we consider it important to examine the significance of these phenomenological lagrangians in chiral perturbation theory ( @xmath36pt ) @xcite which in general serves as a guiding principle for low - energy hadron dynamics .
in the present work we shall describe an attempt at relating the traditional phenomenological approaches to @xmath36pt .
the fact that @xmath36pt accounts for and improves the results of the current algebra also makes it a natural framework for studying threshold pion production .
furthermore , in this low - energy regime , it is natural to employ the heavy - fermion formalism ( hff)@xcite .
the hff has an additional advantage of allowing easy comparison with eqs.([h0 ] ) and ( [ h1 ] ) .
it should be mentioned , however , that the application of @xmath36pt to nuclei involves some subtlety . as emphasized by weinberg @xcite ,
naive chiral counting fails for a nucleus , which is a loosely bound many - body system .
this is because purely nucleonic intermediate states occurring in a nucleus can have very low excitation energies , which spoils the ordinary chiral counting . to avoid this difficulty
, one must first classify diagrams appearing in perturbation series into irreducible and reducible diagrams , according to whether or not a diagram is free from purely nucleonic intermediate states .
thus , in an irreducible diagram , every intermediate state contains at least one meson .
the @xmath36pt can be safely applied to the irreducible diagrams .
the contribution of all the irreducible diagrams ( up to a specified chiral order ) is then to be used as an effective operator acting on the nucleonic hilbert space .
this second step allows us to incorporate the contributions of the reducible diagrams
. we may refer to this two - step procedure as the _ nuclear chiral perturbation theory _ ( nuclear @xmath36pt ) .
this method was first applied by weinberg @xcite to chiral - perturbation - theoretical derivation of the nucleon - nucleon interactions and subsequently used by van kolck _ et al .
park , min and rho ( pmr ) @xcite applied the nuclear @xmath36pt to meson exchange currents in nuclei .
the success of the nuclear @xmath36pt in describing the exchange currents for the electromagnetic and weak interactions is well known @xcite .
the present paper is in the spirit of the work of pmr .
this article is organized as follows : in the next section we define our pion field and the chiral counting procedure . then in section
iii we present the two lowest order lagrangians , discuss their connection to the early works on this reaction and determine within certain approximations the numerical values of the effective pion rescattering vertex strength , @xmath37 . in section iv we briefly discuss the connection between the transition matrix for this reaction and the @xmath36pt calculated amplitude . in section v we present necessary loop corrections to the born term , and in section vi we calculate the cross section and discuss the various approximations and the uncertainties of the low energy constants in @xmath36pt . finally in section vii , after discussing some higher chiral order diagrams , we present our main conclusions . a work very similar in spirit to ours has recently been completed by cohen _ et al._@xcite .
the effective chiral lagrangian @xmath38 involves an su(2 ) matrix @xmath39 that is non - linearly related to the pion field and that has standard chiral transformation properties @xcite .
an example is @xcite @xmath40 ^ 2 } + i\bbox{\tau } \!\cdot\ !
\label{eq : uuu}\ ] ] in the meson sector , the sum of chiral - invariant monomials constructed from @xmath39 and its derivatives constitutes the chiral - symmetric part of @xmath38 .
furthermore , one can construct systematically the symmetry - breaking part of @xmath38 with the use of a mass matrix @xmath41 the chiral transformation of which is dictated by that of the quark mass term in the qcd lagrangian .
to each term appearing in @xmath38 one can assign a chiral order index @xmath42 defined by @xmath43 where @xmath44 is the summed power of the derivative and the pion mass involved in this term .
a low energy phenomenon is characterized by a generic pion momentum @xmath45 , which is small compared to the chiral scale @xmath46 1 gev .
it can be shown that the contribution of a term of chiral order @xmath42 carry a factor @xmath47 , where @xmath48 represents either @xmath45 or the pion mass @xmath49 .
this suggests the possibility of describing low - energy phenomena in terms of @xmath38 that contains only a manageably limited number of terms of low chiral order .
this is the basic idea of @xmath36pt .
the heavy fermion formalism ( hff)@xcite allows us to easily extend @xmath36pt to the meson - nucleon system . in hff , the ordinary dirac field @xmath50 describing the nucleon , is replaced by the heavy nucleon field @xmath51 and the accompanying small component field " @xmath52 through the transformation @xmath53 \label{n}\ ] ] with @xmath54 where the four - velocity @xmath55 is assumed to be almost static , i.e. , @xmath56 @xcite .
elimination of @xmath52 in favor of @xmath51 leads to expansion in @xmath57 . since @xmath58 1 gev @xmath59 , an expansion in @xmath57 may be treated like an expansion in @xmath60 .
@xmath38 in hff consists of chiral symmetric monomials constructed from @xmath39 , @xmath51 and their derivatives and of symmetry - breaking terms involving @xmath41 .
the chiral order @xmath42 in hff is defined by @xmath61 where @xmath44 is , as before , the summed power of the derivative and the pion mass , while @xmath62 is the number of nucleon fields involved in a given term . as before , a term in @xmath38 with chiral order @xmath42 can be shown to carry a factor @xmath63 . in what follows , @xmath42 stands for the chiral order defined in eq.([eq : nubar2 ] ) .
in addition to the chiral order index @xmath42 defined for each term in @xmath38 , we assign a chiral order index @xmath64 for each irreducible feynman diagram appearing in the chiral perturbation series for a multifermion system @xcite .
its definition is @xmath65 where @xmath66 is the number of nucleons in the feynman diagram , @xmath67 the number of loops , and @xmath68 the number of disconnected parts of the diagram .
the sum over @xmath69 runs over all the vertices in the feynman graph , and @xmath70 is the chiral order of each vertex .
one can show @xcite that an irreducible diagram of chiral order @xmath64 carries a factor @xmath71 . in the literature the term effective lagrangian " ( or effective hamiltonian " ) is often used to imply that that lagrangian ( or hamiltonian ) is only meant for calculating tree diagrams .
the hamiltonians given in eqs.([h0 ] ) and ( [ h1 ] ) are regarded as effective hamiltonians of this type .
we must note , however , that the effective lagrangian in @xmath36pt has a different meaning .
not only can @xmath38 be used beyond tree approximation but , in fact , a consistent chiral counting even demands inclusion of every loop diagram whose chiral order @xmath64 is lower than or equal to the chiral order of interest . as will be discussed below , for a consistent @xmath36pt treatment of the problem at hand ,
we therefore need to consider loop corrections
. however , since the inclusion of the loop corrections is rather technical , we find it useful to first concentrate on the tree - diagram contributions .
this simplification allows us to understand the basic aspects of the relation between the contributions from @xmath36pt and the phenomenological hamiltonians , eqs.([h0 ] ) and ( [ h1 ] ) .
therefore , in the next two sections ( iii and iv ) we limit our discussion to tree diagrams . a more elaborate treatment including loop corrections
will be described in section v.
in order to produce the one - body and two - body diagrams depicted in figs.1(a ) and 1(b ) , we minimally need ( see below ) terms with @xmath72 and @xmath73 in @xmath38 .
we therefore work with @xmath74 where @xmath75 represents terms of chiral order @xmath42 .
their explicit forms are @xcite [ eq:11 ] @xmath76 \\ & + \bar{n } ( i v \cdot d + g_a^ { } s \cdot u ) n \\ & - \frac12 \displaystyle \sum_a c_a ( \bar{n } \gamma_a n)^2 \label{eq17a}\end{aligned}\ ] ] @xmath77 in the above @xmath78 @xmath79 @xmath80 ) n,\ ] ] and @xmath81 is the covariant spin operator defined by @xmath82 .
\label{s}\ ] ] in @xmath83 above we have retained only terms of direct relevance for our discussion . the coupling constants @xmath84 and @xmath85 can be fixed from phenomenology @xcite .
they are related to the pion - nucleon @xmath27-term , @xmath86 ( @xmath87 average mass of the light quarks , @xmath88 ) , the axial polarizability @xmath89 and the isospin - even @xmath35 @xmath4-wave scattering length @xmath90 @xcite .
( the explicit expressions will be given below . )
it should be noted that in hff , a part of the term in @xmath83 with the coefficient @xmath91 , namely the @xmath92 piece , represents the @xmath4-wave @xmath5-@xmath6 scattering contribution , which in a traditional calculation is obtained from the crossed born - term . the four - fermi non - derivative contact terms in eq.([eq:11 ] ) were introduced by weinberg @xcite and further investigated in two- and three - nucleon systems by van kolck _
et al_. @xcite .
although these terms are important in the chiral perturbative derivation of the nucleon - nucleon interactions @xcite , they do not play a major role in the following discussion of the threshold @xmath0 reaction .
we therefore temporarily ignore these four - fermion terms and come back to a discussion of these terms in the last section .
the lagrangian ( [ lag ] ) leads to the pion - nucleon interaction hamiltonian @xmath93 where [ eq:17 ] @xmath94 n + \frac{1}{4f_\pi^2 } \bar{n } \bbox{\tau}\!\cdot\!\bbox{\pi } \!\times\!\dot{\bbox{\pi } }
n \label{eq : hint0}\ ] ] @xmath95 \bar{n } n \nonumber \\ & & \label{eq : hint1}\end{aligned}\ ] ] here @xmath96 represents the term of chiral order @xmath42 .
we now compare @xmath97 resulting from @xmath36pt , eq.([hint ] ) with the phenomenological effective hamiltonian @xmath98 , eqs.([h0 ] ) and ( [ h1 ] ) .
( the reader is reminded that the chiral index @xmath42 should not be confused with the suffix appearing in @xmath11 and @xmath99 . ) regarding the @xmath100 vertices , we note that the first term in @xmath101 and the first term in @xmath102 exactly correspond to the first and second terms , respectively , in @xmath11 .
thus the so - called galilean - invariance term naturally arises as a @xmath103 correction term in hff . as for the @xmath104 vertices
, we can associate the second term in @xmath105 to the @xmath15 term in @xmath99 , and second term in @xmath106 to the @xmath14 term in @xmath99 .
this suggests the following identifications : @xmath107 and @xmath108 \equiv \kappa(k , q ) \end{array } \label{eq : kappakq}\ ] ] in eq.([eq : kappakq ] ) , @xmath109 and @xmath110 stand for the four - momenta of the exchanged - and final pions , respectively , see fig.1(b ) . since , as already discussed , the @xmath15 term is not important for our purposes , we shall concentrate on the @xmath14 term .
the best available estimates of the coefficients @xmath111 ( i=1 - 3 ) can be found in refs.@xcite , which give [ eq:20 ] @xmath112 \\ & = -0.87 \pm 0.11 \,\mbox{gev}^{-1 } , \label{c1}\end{aligned}\ ] ] @xmath113 \\ & = -5.25 \pm 0.22 \,\mbox{gev}^{-1 } , \label{c2}\end{aligned}\ ] ] @xmath114 \\ & \hskip 4em + 2c_1 - c_3 + \frac{g_a^2}{8m_n } \\ & = 3.34 \pm 0.27 \,\mbox{gev}^{-1}. \label{c3}\end{aligned}\ ] ] the numerical results are based on the experimental values : @xmath115 mev@xcite , @xmath116@xcite , and @xmath117@xcite .
we shall show in section vi that the uncertainties in the numerical value for @xmath118 might be larger than quoted in eq.([c3 ] ) .
in fact , the terms in eqs.([c1])-([c3 ] ) proportional to the @xmath119 are @xmath120 corrections arising from finite terms of @xmath121 . however , since the present section is just an introduction to a later systematic treatment , this inconsistency in accuracy " will be ignored for the moment .
now , for _ on - shell _ low energy pion - nucleon scattering ,
_ i.e. _ , @xmath122 , we equate @xmath123 where @xmath124 from eq.([eq:20 ] ) we have @xmath125 -\frac{g_a^2}{8m_n } , \label{eq : cctilde1}\ ] ] which results in @xmath126 the above cited empirical value for @xmath127 leads to @xmath128 we now interpret these results in terms of @xmath14 of eq.([h1 ] ) .
conventionally , @xmath14 is determined from eq.([eq4a ] ) which is the first term in eq.([eq : kappa1 ] ) .
thus @xmath129 which gives @xmath130 or @xmath131 .
this is the
standard value " used in the literature @xcite . on the other hand , the r.h.s . of eq.([eq : onshell ] ) based on @xmath36pt gives from eq.([eq : kappa2 ] ) @xmath132 which is about twice as large as the conventional value .
this means the second term in eq.([eq : kappa1 ] ) is almost as large as the first term .
thus @xmath36pt leads to a substantial modification of the commonly used formula , eq.([eq4a ] ) or eq.([eqll ] ) .
this large higher chiral order " corrections due to @xmath121 [ the term proportional to @xmath119 in eq.([eq : kappa1 ] ) ] indicates that @xmath36pt does not converge very rapidly in this particular case .
this apparent lack of convergence is probably due to the fact that the first terms in expansion , the @xmath5-@xmath6 isoscalar scattering length @xmath127 , is exceptionally small . to develop further the connection between the traditional and the @xmath36pt approaches , we return to a discussion of eq.([eq : kappakq ] )
obviously , the constant @xmath14 can not be fully identified with @xmath133 which depends on the momenta @xmath134 and @xmath135 .
in fact , the momentum dependence of @xmath133 should play a significant role in describing the physical pion - nucleon elastic scattering process where @xmath136 , @xmath137 .
an additional crucial point in the present context is that , in the rescattering diagram fig.1(b ) , the exchanged pion can be far off - shell , and therefore the @xmath134 and @xmath135 dependence in @xmath133 may play an even more pronounced role . as an illustration , let us consider again the _ typical threshold kinematics _ discussed in the introduction : @xmath138 and @xmath139 . if we denote by @xmath37 the value of @xmath133 [ eq.([eq : kappakq ] ) ] corresponding to the _ typical threshold kinematics _ , we have @xmath140 \label{eq : kappath}\ ] ] the use of the central values for the coupling constants @xmath84 and @xmath85 leads to @xmath141 thus the strength of the @xmath4-wave pion - nucleon interaction here is much stronger than the on - shell cases , see eqs.([lambdaold ] ) and ( [ eq : lambda1c ] ) , and the sign of the off - shell coupling strength is _ opposite _ to the on - shell cases .
the first feature is qualitatively in line with the observation of hernndez and oset@xcite that the rescattering term should be larger than previously considered .
however , the sign of the typical off - shell coupling in our case [ eq.([kappath1 ] ) ] is opposite to the one used in ref.@xcite .
as will be discussed later , this flip of the sign drastically changes the pattern of interplay between the born and rescattering terms .
we must emphasize that the off - shell enhancement depends strongly on the values of @xmath84 and @xmath85 , which , as discussed in ref.@xcite , are not known very accurately .
it is therefore important to examine to what extent the existing large ambiguities in @xmath84 and @xmath85 affect the off - shell enhancement of the @xmath0 reaction .
we shall address this question in section vi .
as explained earlier , in the nuclear @xmath36pt we first use @xmath36pt to calculate the contributions of the irreducible diagrams .
let @xmath143 represent the contributions of all irreducible diagrams ( up to a specified chiral order @xmath64 ) for the @xmath0 process .
then we use @xmath143 as an effective transition operator in the hilbert space of nuclear wavefunctions .
consequently , the two - nucleon transition matrix element @xmath144 for the @xmath0 process is given by @xmath145 where @xmath146 ( @xmath147 ) is the initial ( final ) two - nucleon state distorted by the initial - state ( final - state ) interaction .
these distorted waves should be obtained by solving the schrdinger equation with nucleon - nucleon interactions generated by irreducible diagrams pertinent to nucleon - nucleon scattering , thereby incorporating an infinite number of reducible " ladder diagrams . in this section
we concentrate on the derivation of the transition operator @xmath143 , relegating the discussion of @xmath144 and eq.([eq : tmatrix ] ) to section vi .
we decompose @xmath143 as @xmath148 where @xmath149 represents the contribution from feynman diagrams of chiral order @xmath64 , as defined in eq.([eq : nununu ] ) .
the lowest value of @xmath64 occurs for the born term shown in fig.2(a ) . for @xmath4-wave @xmath5 production at threshold the @xmath150 vertex with @xmath151 , the first term in eq.([hint ] ) , can not contribute ; hence the lowest @xmath42 for @xmath150 vertex involving an external pion must be @xmath72 . the first term in eq.([eq : hint1 ] )
provides this vertex . according to eq.([eq : nununu ] ) , the chiral order of fig.2(a ) is given by @xmath152 as can be checked easily , there are no diagrams with @xmath153 since in the rescattering diagram , fig.2(b ) , the second term in eq.([eq : hint0 ] ) , which gives the @xmath154 vertex with @xmath151 , is not operative here due to the isospin selection rule .
the rescattering diagram in fig.2(b ) with the indicated value of @xmath42 at each vertex contributes to @xmath155 .
it should be noted that because of the @xmath156 term in the chiral counting expression , eq.([eq : nununu ] ) , exchange - current - type diagrams such as fig.2(b ) give higher values of @xmath64 . in this work
we truncate the calculation of the transition operator @xmath143 at @xmath157 thus , @xmath158 the above enumeration is , as briefly discussed in section iii , far from complete because loop diagrams and counter terms and finite terms from @xmath121 have been left out . in fig.3
we show the loop corrections to the born term [ fig.2(a ) ] .
the diagrams in fig.3 all have @xmath159 and hence are of the same chiral order as the leading order rescattering diagram , fig.2(b ) . as discussed earlier , for the @xmath0 reaction at threshold the contribution of the born term is numerically suppressed so that the rescattering diagram , which is formally of higher chiral order by two units of @xmath64 , plays an essential role .
this implies that a meaningful and consistent @xmath36pt calculation of this reaction must include the loop corrections to the leading - order born term .
however , we continue to postpone the discussion of loop corrections to the next section . the tree diagrams contributing to eq.([eq : calttrunc ] ) , figs.2(a ) and 2(b ) , are as follows .
the born term , fig.2(a ) , contributes to @xmath160 and the rescattering term , fig.2(b ) , contributes to @xmath161 .
these contributions are given , respectively , by [ eq:35 ] @xmath162 @xmath163 where @xmath164 and @xmath165 ( @xmath166 ) denote the initial and final momenta of the @xmath69-th proton , @xmath167 ; and @xmath168 is as defined in eq.([eq : kappakq ] ) .
we have emphasized above that the loop corrections to the born diagram , fig.2(a ) , which has chiral order @xmath169 , are of the same chiral order @xmath170 as the two - body pion rescattering process , fig.2(b ) .
these loop corrections therefore must be included in a consistent @xmath171 calculation . for our present purposes
it is not necessary to go into a general discussion of the renormalization of the parameters in @xmath172 .
instead we concentrate on an estimation of the size of the finite loop corrections to the specific tree level terms shown in fig.2 .
this will be done by applying standard feynman rules and using dimensional regularization @xcite .
specifically , we only need consider the loop corrections to the single @xmath173 vertex in the s - wave channel : @xmath174 \ ; ( v \cdot q ) \ ; \tau_i^0 \ ; \ ; { \cal v } \label{bornnew}\ ] ] where @xmath175 is the spin of the @xmath69-th proton and @xmath176 is the amplitude to be calculated . for the born term [ fig.2(a ) ] itself we have : @xmath177 given by eq.([tm1 ] ) .
the loop diagrams [ figs.3(a)-(f ) ] , which renormalize the s - wave born term , give the following contributions : [ loop ] @xmath178 @xmath179 @xmath180 @xmath181 + \frac { ( v p)^2 j_0 ( v p ) - ( v p')^2 j_0 ( v p ' ) } { v q } \right ) \label{loopd}\ ] ] here we have adopted the notations of ref.@xcite . thus @xmath182 where the divergence is included in @xmath183 . \label{l}\ ] ] in this expression @xmath184 denotes the dimensional regularization scale and @xmath185 .
furthermore , @xmath186 and @xmath187 in eqs .
( [ loop ] ) are defined by @xmath188 and @xmath189 \label{j2}\ ] ] the two contributions to @xmath176 , eqs.([loopc ] ) and ( [ loopd ] ) , originate from two different combinations of terms in eqs .
( [ eq:17 ] ) . to calculate eq.([loopc ] ) , the second term in eq.([eq : hint0 ] ) and the first term in eq.([eq : hint1 ] ) are used at the vertices , whereas eq.([loopd ] ) is calculated using the first term of eq.([eq : hint0 ] ) and the second term of eq.([eq : hint1 ] ) .
the standard renormalization consists in the following procedure : + ( 1 ) the loop contributions to @xmath176 are separated into a divergent part , which we take to be proportional to @xmath67 of eq.([l ] ) and which contains a pole at @xmath190 , and a finite part : @xmath191 ( 2 ) local counter terms , which are of the same chiral order as the loop diagrams , are added . in our case
these counter terms must come from the lagrangian @xmath192 .
@xmath193 to give two - nucleon diagrams with @xmath170 .
the unknown constants @xmath194 are then written as a sum of a finite and an infinite part @xmath195 the constants @xmath196 are determined by requiring that the infinite part of @xmath194 cancel the divergent part @xmath197 .
the remaining finite contributions which should be added to the born term via eq.([bornnew ] ) , are @xmath198 the amplitude @xmath199 contains energy - independent and energy - dependent parts , as can be seen in eq .
( [ loop ] ) . the energy - independent part can be absorbed in the renormalization of the following physical parameters : the pion wave function renormalization factor @xmath200 [ fig.3(e ) ] , the nucleon mass @xmath201 and the nucleon wave function renormalization factor @xmath202 [ ( fig.3(f ) ] , as well as the axial coupling constant @xmath8 [ figs.3(a),(b),(e),(f ) ] .
for the evaluation of the energy - dependent part we use the _ typical threshold kinematics _ : @xmath203 , @xmath204 , @xmath205 . putting these values into the corresponding terms in eq .
( [ loop ] ) , we obtain as the total contribution of the diagrams in fig.3 @xmath206 thus @xmath207 amounts to @xmath208 of the born term [ fig.2(a ) ] .
in addition we have finite contributions from the counterterms of @xmath192 , @xmath209 .
we note that only very few of the low energy constants in the counterterms @xmath210 are known @xcite .
some of the low energy constants in @xmath121 , @xmath211 have been estimated in ref.@xcite assuming @xmath212 resonance saturation .
the result indicates @xmath213 . for an estimate of the low energy constants
@xmath210 in @xmath192 it seems reasonable to assume that they are of the same order of magnitude as the @xmath211 in @xmath121 . to be conservative
let us assume @xmath214 ; then we expect @xmath215 .
it is clear that , if those coefficients were `` unreasonably large '' , the convergence of the whole chiral series would be destroyed .
altogether , after renormalization the total contributions from the loop terms are expected to amount to at most @xmath216 of the born term .
this is not a completely negligible contribution in the present context because , as will be discussed in the next section , there can be a significant cancelation between the born and the rescattering terms .
nevertheless , since our present treatment involves other larger uncertainties , we will neglect the renormalization of the born term and henceforth concentrate on the bare born term [ fig.2(a ) ] and the rescattering term [ fig.2(b ) ] .
we derived in section iv the effective transition operator @xmath143 arising from the tree diagrams and , in section v , we estimated the additional contributions due to the loop corrections and presented an argument for ignoring the loop corrections in this work .
these considerations lead to the the hff expression of @xmath143 up to order @xmath64 = 1 , given in eqs.([eq : calttrunc ] ) and ( [ eq:35 ] ) , and this @xmath143 is to be used in eq.([eq : tmatrix ] ) to obtain the two - nucleon transition matrix @xmath144
. a formally consistent " treatment of eq.([eq : tmatrix ] ) would consist in using for @xmath146 and @xmath147 two - nucleon wave functions generated by irreducible diagrams of order up to @xmath171 . a problem in this consistent
" @xmath36pt approach is that the intermediate two - nucleon propagators in fig.1 can be significantly off - mass - shell , which creates a difficulty in any @xmath36pt calculation .
another more practical problem is that , if we include the initial- and final- two - nucleon ( @xmath6-@xmath6 ) interactions in diagrams up to chiral order @xmath171 , these @xmath6-@xmath6 interactions are not realistic enough to reproduce the known @xmath6-@xmath6 observables .
a pragmatic remedy for these problems is to use a phenomenological @xmath6-@xmath6 potential to generate the distorted @xmath6-@xmath6 wavefunctions .
park , min and rho @xcite used this hybrid approach to study the exchange - current in the @xmath217 reaction and at least , for the low - momentum transfer process studied in ref.@xcite , the hybrid method is known to work extremely well .
apart from the above - mentioned problem , there is a delicate aspect in the derivation of an effective two - body operator from a given feynman diagram .
ordinarily , one works with @xmath218-space transition operators acting on @xmath218-representation wavefunctions , for the nuclear wavefunctions are commonly given in this representation . to this end ,
a feynman amplitude which is most conveniently given in momentum space , is fourier - transformed into the @xmath218-representation .
this method works best for low momentum transfer processes which have substantial transition amplitudes for on - shell initial and final plane - wave states .
however , the @xmath0 reaction at threshold does not belong to this category . for this reaction it is essential to recognize that the nucleon lines that appear as external lines in fig.2 are in fact internal lines in larger diagrams illustrated in fig.1 .
these internal lines can be far off - shell due to the initial- and final - state interactions . indeed without this
off - shell kinematics , the born term [ fig.2(a ) ] would not contribute at all ! in the conventional approach , however , one ignores this feature in deriving @xmath143 in coordinate representation .
for example , in fourier - transforming an operator of the type of @xmath219 , eq.([tp1 ] ) , even though @xmath164 and @xmath165 in eq.([tp1 ] ) in fact can be anything due to momentum transfers caused by the initial and final @xmath6-@xmath6 interactions , it is a common practice to keep the energy of the propagating pion fixed at the value determined by the asymptotic energies of the nucleons .
hanhart et al.@xcite made a critical study of the consequences of avoiding these kinematical approximations .
they worked directly with the two - nucleon wavefunctions in momentum representation . in the present work we do not attempt at detailed momentum - space calculations and simply use the conventional " fourier transform method .
because of this and a few other approximations adopted , the numerical work presented here is admittedly of exploratory nature .
nonetheless , as we shall show , our semiquantitative study of @xmath144 based on the chiral - theoretically motived transition operator @xmath143 provide some valuable insight into the dynamics of the threshold @xmath0 reaction .
let us denote the contribution of fig.2(b ) for plane - wave initial and final states by @xmath220 .
we first calculate this matrix element for the _ typical threshold kinematics _ described earlier ; for the meson variables , @xmath221 and @xmath222 with @xmath223 .
correspondingly , the coupling strength @xmath133 [ eq.([tp1 ] ) ] is taken to be @xmath224gev@xmath225 [ eq.([kappath1 ] ) ] .
subsequently , by liberating the momentum variables @xmath226 , @xmath227 , @xmath228 , and @xmath229 from the on - mass - shell conditions ( @xmath230 ) , we treat @xmath231 as a function of @xmath226 , @xmath227 , @xmath228 , and @xmath229 .
let @xmath232 stand for this function .
we still require momentum conservation at each vertex , which imposes the conditions @xmath233 , and @xmath234 .
@xmath232 can be easily fourier transformed to give @xmath235 in @xmath218 representation .
the simplified treatment described here , which is commonly used in the literature , shall be referred to as the _ fixed kinematics approximation_. now , in the _ fixed kinematics approximation _ , @xmath143 [ eqs.([eq : calttrunc ] ) , ( [ eq:35 ] ) ] is translated into differential operators acting on relative coordinate of the two - nucleon wavefunctions : [ eq:37 ] @xmath236 @xmath237 where the derivative operator with subscript @xmath238 is to act on the relative coordinate @xmath239 between two protons , and @xmath240 . the trivial isospin operator @xmath241 has been dropped . the yukawa function @xmath242 is defined with the _ effective _ mass @xmath243 .
we reemphasize that the simple yukawa form @xmath244 arises only when the _ fixed kinematics approximation _ just discussed is used .
from this point on , our calculation of @xmath144 follows exactly the traditional pattern described in the literature .
thus @xmath144 is evaluated by inserting the transition operators , @xmath245 and @xmath235 , eq.([eq:37 ] ) , between the initial and final nuclear states @xmath246 where @xmath10 and @xmath247 are the asymptotic relative three - momenta of the initial and final two - proton systems .
the wavefunctions are normalized as @xmath248 with @xmath249 being the @xmath6-@xmath6 scattering phase shifts . for simplicity , the coulomb interactions between the two protons is ignored .
( the coulomb force is known to reduce the cross section up to 30%.@xcite . )
the explicit expression for the transition amplitude at threshold is obtained as @xmath250 here , @xmath251 is the kinetic energy of the final state , and [ eq:40 ] @xmath252 @xmath253 the total cross section is obtained by multiplying the absolute square of the transition amplitude ( averaged over the initial spins and summed over the final spins ) with the appropriate phase space factor @xmath254 and the flux factor @xmath255 : @xmath256 where @xmath257 is the scattering length of the @xmath258 potential .
then the cross section can be simply expressed as @xmath259 where [ eq:44 ] @xmath260
@xmath261 under the approximation ( [ aprx ] ) , the energy dependence of the cross section is solely given by @xmath262 , which incorporates the phase space and the final state interaction effect ( in the watson approximation@xcite ) .
we have calculated the integrals @xmath263 and @xmath264 for representative nuclear potentials : the hamada - johnston ( hj ) potential@xcite , and the reid soft - core potential ( rsc)@xcite . the results are given in table i , and the corresponding cross sections are presented in table ii .
these results indicate that , for the nuclear potentials considered here , the value of @xmath265 is much too small to reproduce the experimental cross section .
if we define the discrepancy ratio @xmath266 by @xmath267 with @xmath268 taken from ref.@xcite , then @xmath269 ( @xmath270 ) for the hamada - johnston ( reid soft - core ) potential , and @xmath266 happens to be almost constant for the whole range of @xmath271 mev for which @xmath268 is known .
thus , although the off - shell behavior of the @xmath4-wave pion scattering amplitude derived from the chiral lagrangian does enhance the contribution of the rescattering process over the value reported in the literature , the sign change that occurs in @xmath272 as one goes from @xmath273 [ eq.([eq : onshell ] ) ] to @xmath37 [ eq.([eq : kappath ] ) ] results in a significant cancelation between the born term @xmath263 and the rescattering term @xmath274 , leading to the very small cross sections in table ii @xcite .
the drastic cancelation between @xmath263 and @xmath274 found here means that the calculated cross sections are highly sensitive to the various approximations used in our calculation and also to the precise values of the constants @xmath275 and @xmath85 of eq.([eq:20 ] ) .
we will discuss these two questions in the next two paragraphs .
we adopted the threshold kinematics approximation and neglected the energy - momentum dependence in eq.([eq : kappakq ] ) and treated the vertices in figs.1 and 2 as fixed numbers , _
i.e. _ , @xmath133 = @xmath37 = constant .
in addition , although the loop corrections of chiral order @xmath64 = 1 , shown in fig.3(a ) , automatically introduces energy - momentum dependent vertices , we ignored this feature .
the fact that the kinematics of the reaction eq.([eq : pppi ] ) requires highly off - shell vertices leads to the expectation that the vertex form factors can be very important and invalidate the _ threshold kinematics approximation _ leading to eq.([eq:37 ] ) . in this connection
we note that a momentum - space calculation @xcite , which is free from this approximation , indicates that even a negative value of @xmath14 could lead to the moderate enhancement of the cross section .
the strong cancelation between the born and rescattering terms also means that , even within the framework of the _ fixed kinematics approximation _ , the large errors that exist in the empirical value of @xmath127 and the @xmath275 and @xmath85 constants can influence the cross sections significantly . to assess this influence , we rewrite eq.([eq : kappath ] ) as @xmath276 the use of the experimental values for @xmath127 and @xmath277 quoted earlier leads to @xmath278 with this uncertainty taken into account , the ratio @xmath266 ranges from @xmath279 to @xmath280 for the hamada - johnston potential , and from @xmath281 to @xmath282 for the reid soft - core potential . to further examine the uncertainties in the @xmath83 constants we remark that the value of @xmath283 can be extracted from the known pion - nucleon effective range parameter @xmath284 .
the low energy pion - nucleon scattering amplitude is expanded as : @xmath285 where * q * is the pion momentum and @xmath286 @xcite .
if we use @xmath83 to calculate the s - wave pion - nucleon amplitude we find : @xmath287 and then eq.([eq : kappath ] ) leads to @xmath288 since @xmath85 is given directly by the experimental quantity @xmath89 [ eq.([c2 ] ) ] , we consider eq.([bplus ] ) as an alternative input to determine @xmath118 in terms of @xmath284 and @xmath85
. then eqs.([bplus]),([c2 ] ) and the experimental value of @xmath284 @xcite give @xmath289 we note that this value is larger than the one given in eq.([c3 ] ) , indicating that the determination of @xmath118 requires further studies .
with the new value of @xmath37 given in eq.([kappanew ] ) we find that the discrepancy ratio @xmath266 [ eq.([ratio ] ) ] can be as small as @xmath290 .
( in this case @xmath291 ; the exact cancelation between the born and the pion rescattering term occurs for @xmath292 . ) without attaching any significance to the detailed numbers above , we still learn the extreme sensitivity of @xmath293 to the input parameters and that , despite this high sensitivity , @xmath293 still falls far short of @xmath268 ( within the framework of the _ fixed kinematics approximation _ ) .
in this work we have used @xmath36pt to calculate the effective pion - exchange current contribution to the @xmath294 reaction at threshold . as stated repeatedly ,
our aim here is to carry out a systematic treatment of @xmath143 up to chiral order @xmath171 [ see eq.([eq : calttrunc ] ) ] .
however , in order to make contact with the expressions appearing in the literature @xcite , let us consider a very limited number of @xmath295 diagrams . to be specific , we consider a diagram in fig.2(b ) but with the @xmath296 ( @xmath10-wave ) @xmath100 vertex replaced with a @xmath72 ( @xmath4-wave ) vertex .
then , instead of eq.([tp1r ] ) , we will obtain @xmath297 , \label{textra}\ ] ] which is the two - body transition operator used in ref.@xcite .
thus , we do recover the usual phenomenological parameterization in @xmath36pt , but this is just one of many @xmath298 diagrams .
our systematic @xmath299 calculation excludes all @xmath298 diagrams .
we have also ignored the exchange current contributions from scalar and vector two - nucleon exchanges . following the @xmath36pt of refs .
@xcite the vector meson exchange is largely accounted for via the four - nucleon contact terms illustrated in fig.4(a ) . if we had retained the last two terms of eq.([eq:11 ] ) , the pion - nucleon interaction @xmath300 , eq.([eq : hint1 ] ) , would have had an additional piece @xmath301 the @xmath301 term of fig.4(a ) has a @xmath303 structure , which means it describes @xmath10-wave pion production and therefore does not contribute to the threshold @xmath0 reaction .
the @xmath4-wave pion production contact term , also belonging to the type of diagram illustrated in fig.4(a ) , enters as a @xmath304 recoil correction to @xmath301 and therefore is of chiral order @xmath64 = 2 .
formally , the chiral order @xmath295 diagrams have no place in the present calculation limited to @xmath171 .
however , in view of the great current interest in the possible large contribution of the heavy - meson exchange diagrams , we make a few remarks on the s - wave @xmath295 contact terms depicted in fig.4(a ) .
we note that the coordinate representation of this contact term contains @xmath305 .
meanwhile , in the threshold @xmath0 reaction the initial two - nucleon relative motion must be in @xmath10-wave ( because of parity ) and so its wavefunction vanishes at @xmath306 . thus , even in a chiral order @xmath295 calculation , the contact term fig.4(a ) corresponding to s - wave pion production will play no role . including meson loops corrections to these contact terms [ an example illustrated in fig.4(b ) ]
would smear out the @xmath307-function behavior , allowing them to have a finite contribution to the threshold @xmath0 reaction .
this involves , however , diagrams of even higher chiral order than @xmath295 .
thus , in order to include the strong effective isoscalar - vector repulsion of the @xmath6-@xmath6 forces ( @xmath308 exchange ) contained in the four - nucleon contact terms of weinberg s @xcite and van kolck s _ et al . _ s
@xcite @xmath36pt description , we have to go to chiral order @xmath309 .
meanwhile , one may picture the effective heavy mesons " as generated by multi - pion exchange diagrams like those illustrated in fig.5 .
these diagrams , which necessarily contain loops , represent a very limited class of @xmath310 diagrams .
for example , an important part of the effective scalar exchange between two nucleons involve intermediate @xmath5-@xmath5 @xmath4-wave interaction which requires at least two loop diagrams like fig.5(c ) .
thus , if we are to interpret the heavy - meson exchange diagrams of lee and riska@xcite in the framework of nuclear @xmath36pt , we must deal with terms with chiral order @xmath311 , which at present is beyond practical calculations . 1 .
using @xmath36pt in a systematic fashion we have shown that the contribution of the pion rescattering term can be much larger than obtained in the traditional phenomenological calculations .
this fact itself supports the suggestion of hernandez and oset @xcite that the off - shell @xmath4-wave pion - nucleon scattering should enhance the rescattering contribution significantly .
however , the sign of the enhanced rescattering vertex obtained in @xmath36pt is _ opposite _ to that used in ref.@xcite , at least for the _ typical threshold kinematics _ defined in the text .
this sign change in the coupling constant @xmath37 leads to a destructive interference between the born and rescattering terms instead of the constructive interference found in ref.@xcite .
the significant cancelation between these terms give rise to the very small cross section for the near - threshold @xmath0 reaction calculated in this work .
although our particular numerical results were obtained in what we call the _ fixed kinematics approximation _ , these results at least indicate that the large enhancement of @xmath293 obtained in ref.@xcite is open to more detailed examinations .
the _ fixed kinematics approximation _ ( which is commonly used in the literature ) should be avoided .
there are at least two reasons why this is not a good approximation for this reaction : ( i ) the initial- and final - state interactions play an essential role in the near - threshold @xmath0 reaction ; ( ii ) the theoretical cross section within the framework of the born plus rescattering terms is likely to depend on the delicate cancelation between these two terms . in a momentum space calculation @xcite , we can easily avoid the _ fixed kinematics approximation_. such a calculation will allow us to work with full off - shell kinematics , to incorporate the @xmath36pt form factors in the born term , and to reduce ambiguities in our calculation down to the level of uncertainties in the input parameters in @xmath36pt and the chiral counter terms . 3 .
several works @xcite indicate that the two - nucleon scalar ( sigma ) exchange can be very important .
we gave in the introduction a simple kinematical argument for its plausibility , and our dynamical calculation ( albeit of semiquantitative nature ) seems to indicate the necessity of the sigma exchange contribution in order to explain the observed cross sections for the threshold @xmath0 reaction .
it is of great interest to see to what extent an improved @xmath36pt calculation based on momentum - space representation helps sharpen the conclusion on the necessity of the sigma exchange diagram .
such a calculation is now in progress . if it is established that the heavy meson exchange diagrams play an essential role in the threshold @xmath0 reaction
, it seems that we must resort to a modified version of @xmath36pt , for a brute force extension of our treatment to @xmath312 seems extremely difficult .
an attempt to include vector meson degrees of freedoms explicitly can be found e.g. in ref.@xcite . a purely phenomenological approach as used in @xcite
may also be a useful alternative .
we are grateful to u. van kolck for the useful communication on ref .
one of us ( b .- y .
p ) is grateful for the hospitality of the nuclear theory group of the university of south carolina , where the main part of this work was done .
this work is supported in part by the national science foundation , grant no .
phys-9310124 . in practical calculations
we will choose the nucleon rest frame @xmath314 in which case eq.([n ] ) corresponds to the standard non - relativistic reduction of a spinor into upper and lower components and the covariant spin operator of eq.([s ] ) is simply given by @xmath315 . in eq.(11a )
the sum over @xmath316 runs over the possible combinations of @xmath317- and @xmath318- matrices : @xmath319 , @xmath320 , @xmath321 , and @xmath322 .
however , because of the fermi statistics ( fierz rearrangement ) , only two of the four coupling constants @xmath323 are independent .
we may remark en passant that this standard value " for @xmath14 determined from @xmath127 , is numerically close to the contribution of the crossed born term , which in hff is grouped with the @xmath118 term in @xmath83 as @xmath324 .
this remnant of the s - wave pion nucleon crossed born term in hff , appears in the second term in eq.([eq : kappa ] ) , and gives a value of @xmath127 , @xmath325 compatible with @xmath326 @xcite . | the usual theoretical treatments of the near - threshold @xmath0 reaction are based on various phenomenological lagrangians . in this work
we examine the relationship between these approaches and a systematic chiral perturbation method .
our chiral perturbation calculation indicates that the pion rescattering term should be significantly enhanced as compared with the traditional phenomenological treatment , and that this term should have substantial energy and momentum dependence .
an important consequence of this energy - momentum dependence is that , for a representative threshold kinematics and within the framework of our semiquantitative calculation , the rescattering term interferes destructively with the born - term in sharp contrast to the constructive interference obtained in the conventional treatment .
this destructive interference makes theoretical cross sections for @xmath0 much smaller than the experimental values , a feature that suggests the importance of the heavy - meson exchange contributions to explain the experimental data . |
the extensive air shower ( eas ) inverse approach to a problem of the primary energy spectra reconstruction in the region of @xmath0 pev energies has been an essential tool in the past decade @xcite .
basically , it follows from the high accuracies of recent experiments @xcite and the availability of the eas simulation code @xcite , which was developed in the framework of contemporary interaction models in order to compute the kernel functions of a corresponding integral equation set @xcite . at the same time
, the energy spectra of primary ( @xmath1 and @xmath2 ) nuclei obtained from the kascade experiment @xcite using the eas inverse approach disagree with the same data from the ongoing gamma experiment @xcite , where parameterization of the eas inverse problem is used .
+ below , a peculiarity of the eas inverse problem is investigated , and one of the possible reasons for the observed disagreements between the energy spectra in @xcite and @xcite is considered in the framework of the sibyll @xcite interaction model . the paper is organized as follows : in section 2 the eas inverse approach and the definition of the problem of uniqueness is described .
it is shown , that the abundance of primary nuclear species leads to pseudo solutions for unfolded primary energy spectra . the existence and significance of the pseudo solutions
are shown in section 4 .
the pseudo solutions for primary energy spectra were obtained on the basis of simulation of kascade @xcite shower spectra .
the eas simulation model is presented in section 3 . in section 5
the peculiarities of the pseudo solutions are discussed in comparison with the methodical errors of the kascade data .
the eas inverse problem is ill - posed by definition and the unfolding of the corresponding integral equations does not ensure the uniqueness of the solutions .
the regularized unfolding on the basis of _ a priori _ information on expected solutions ( smoothness , monotony and non - negativity ) in some cases can redefine the inverse problem @xcite and provide the appropriate solutions .
however , the expected singularities ( e.g. knees ) in the primary energy spectra at @xmath3 ev may erroneously be smoothed by regularization algorithms and vice versa , be imitated by the unavoidable oscillations @xcite of the solutions .
furthermore , the eas inverse problem implies evaluations of at least two or more unknown primary energy spectra from the integral equation set of fredholm kind @xcite .
these peculiarities have not been studied in detail and the problem of the uniqueness of solutions can limit the number of evaluated spectra .
+ let @xmath4 be the energy spectrum of a primary nucleus @xmath5 over the atmosphere , @xmath6 be the probability density function describing the transformation of @xmath5 and @xmath7 parameters of the primary nucleus to a measurable vector @xmath8 . then the eas inverse problem , i.e. the reconstruction of the energy spectra of @xmath9 primary nuclei on the basis of the detected spectra @xmath10 of eas parameters , is defined by the integral equation @xmath11 evidently , if @xmath12 are the solutions of eq .
( 1 ) , the functions @xmath13 should also be the solutions of ( 1 ) , provided equation @xmath14 is satisfied for the given measurement errors @xmath15 and for at least one of the combinations of the primary nuclei @xmath16 the number of combinations ( 3 ) stems from a possibility of the existence of a set of functions @xmath17 for each of the primary nuclei ( @xmath5 ) , which can independently satisfy eq .
+ for example , suppose that @xmath18 .
let us denote @xmath19 by @xmath20 and , for simplicity , set the right - hand side of eq .
( 2 ) to 0 . then
, following expression ( 3 ) , we find @xmath21 independent combinations of eq .
( 2 ) : @xmath22 for @xmath23 and @xmath24 , @xmath25 , @xmath26 , @xmath27 and @xmath28 with different @xmath29 functions .
the measurement errors @xmath30 on the right - hand side of these equations can both increase and decrease the domains of @xmath29 functions .
+ one may call the set of functions @xmath31 the pseudo functions with the corresponding pseudo solutions ( spectra ) @xmath13 .
the oscillating @xmath32 functions at @xmath33 are responsible for the first @xmath9 equations @xmath34 , @xmath35 , due to the positive - definite probability density function @xmath36 .
the pseudo solutions @xmath37 can be avoided by using iterative unfolding algorithms @xcite .
+ additional sources of the pseudo solutions originate from the mutually compensative effects at @xmath38 : @xmath39 inherent to eq .
( 2 ) for arbitrary groups of @xmath40 and @xmath41 primary nuclei .
since there are no limitations on the types of the pseudo functions ( except for @xmath42 ) that would follow from expression ( 4 ) , and the number of possible combinations ( 3 ) rapidly increases with the number of evaluated primary spectra ( @xmath9 ) , the problem of the uniqueness of solutions may be insoluble for @xmath43 .
moreover , the pseudo functions have to restrict the efficiency of unfolding energy spectra for @xmath44 , because the unification of @xmath45 primary nuclei spectra into @xmath46 nuclear species ( e.g. light and heavy ) inevitably increases the uncertainties of the kernel functions @xmath36 and thereby also increases the domains of the pseudo functions .
+ notice , that the pseudo solutions will always appear in the iterative unfolding algorithms if the initial iterative values are varied within large intervals . at the same time
, it is practically impossible to derive the pseudo functions from the unfolding of equations ( 1,2 ) due to a strong ill - posedness of the inverse problem .
however , for a given set of the measurement errors @xmath15 and the known kernel functions @xmath6 for @xmath35 primary nuclei , eq .
( 2 ) can be regularized by parametrization of the pseudo functions @xmath47 .
the unknown parameters @xmath48 can be derived from a numerical solution of parametric eq .
( 2 ) , and thereby one may also evaluate the parametrized pseudo functions @xmath31 .
+ below ( section 3 ) , an eas simulation model for computing the kernel function @xmath36 and replicating the kascade @xcite eas spectral errors @xmath15 is considered .
the primary energy spectra obtained in the kascade experiment were derived on the basis of the detected 2-dimensional eas size spectra @xmath49 and an iterative unfolding algorithm @xcite for @xmath50 primary nuclei @xcite . evidently , whether these solutions are unique or not depends on the significance of the arbitrary pseudo functions @xmath51 from eq .
+ we suppose that the convolution of the shower spectra @xmath52 at the observation level and corresponding measurement errors @xmath53 , @xmath54 @xcite are described by 2-dimensional log - normal distributions with parameters @xmath55 , @xmath56 , @xmath57 , @xmath58 and correlation coefficients @xmath59 between the shower size ( @xmath60 ) and the muon truncated size ( @xmath61 ) .
we tested this hypothesis by the @xmath62 goodness - of - fit test using the corsika(nkg ) eas simulation code @xcite for the sibyll2.1 @xcite interaction model , 4 kinds of primary nuclei ( @xmath63 ) , 5 energies ( @xmath64 pev ) and simulation samples for each of @xmath7 and @xmath5 : 5000 , 3000 , 2000 , 1500 , 1000 respectively in @xmath65 zenith angular interval .
the values of corresponding @xmath66 , ( @xmath67 , @xmath68 ) were distributed randomly in the interval @xmath69 for the measurement ranges of the kascade experiment ( @xmath70 and @xmath71 ) and the bin sizes @xmath72 .
+ notice , that the combined 2-dimensional log - normal distributions with parameters @xmath73 at @xmath74 , @xmath75 at @xmath76 , @xmath77 at @xmath78 and @xmath79 at @xmath80 , more precisely ( @xmath81 ) describe the shower spectra @xmath52 in the tail regions .
+ we performed an additional test of the log - normal fit of the @xmath82 spectra using multiple correlation analysis for the shower parameters simulated by the log - normal @xmath52 probability density functions and shower parameters obtained from the corsika eas simulations at power - law primary energy spectra ( @xmath83 ) and equivalent abundances of primary nuclei .
the corresponding correlation coefficients were equal to @xmath84 , @xmath85 , @xmath86 , @xmath87 , and were in close agreement for both methods of @xmath88 and @xmath89 generations .
we replicated the kascade 2-dimensional eas size spectrum @xmath90 ( and corresponding @xmath91 ) by picking out @xmath88 and @xmath92 randomly from the 2-dimensional shower spectra @xmath52 after randomly picking @xmath5 and @xmath7 parameters of a primary particle from the power - law energy spectra @xmath93 with a rigidity - dependent knee @xmath94 , the sharpness parameter @xmath95 and normalization of the all - particle spectrum @xmath96 .
the relative abundance of nuclei was arbitrarily chosen to be @xmath97 and @xmath98 for primary @xmath99 and @xmath2 nuclei respectively , which approximately conforms with the expected abundance from balloon and satellite data @xcite .
+ the mediate values of the parameters of the probability density function @xmath100 were estimated by the corresponding log - parabolic splines .
+ the total number of simulated eas events was set to @xmath101 in order to replicate the corresponding statistical errors @xmath102 of the kascade data .
on the basis of the obtained estimations of @xmath102 ( section 3 ) for the kascade experiment , we examined the uniqueness of unfolding ( 1 ) by @xmath62-the minimization : @xmath103 where @xmath104 represents the left - hand side of eq .
( 2 ) for 2 kinds of the empirical pseudo functions @xmath105 @xmath106 while @xmath107 , otherwise @xmath108 .
the unknown @xmath109 and @xmath110 parameters in expressions ( 7,8 ) were derived from @xmath62 minimization ( 6 ) .
the numbers of bins were @xmath111 and @xmath112 with the bin size @xmath113 .
+ in fact , the minimization of @xmath62 ( 6 ) for different representations ( 7,8 ) of the pseudo functions @xmath31 provides a solution of the corresponding parametric eq .
( 2 ) with a zero right - hand side . to avoid the trivial solutions @xmath114 and reveal the domains of the pseudo functions , the values of some of the parameters were arbitrarily fixed during the minimization of @xmath62 ( 6 ) .
the magnitudes of the fixed parameters were empirically determined via optimization of conditions @xmath115 and @xmath116 for the pseudo spectra with the fixed parameters .
+ the true primary energy spectra @xmath4 for @xmath117 nuclei ( 5 ) and the all - particle energy spectrum @xmath118 ( lines ) along with the corresponding distorted ( pseudo ) spectra @xmath13 ( symbols ) are presented in fig . 1 respectively . and the all - particle spectrum @xmath118 for @xmath117 nuclei ( lines ) and the corresponding pseudo solutions @xmath13 for the pseudo function ( 7 ) ( symbols).,width=264,height=264 ] the parameters of the pseudo functions ( 7 ) derived for @xmath119 ( @xmath120 ) are presented in table 1 .
+ .parameters @xmath121 ( tev@xmath122 ) and @xmath123 of the pseudo function ( 7 ) for different primary nuclei @xmath5 and @xmath124 tev . [ cols="<,^,^",options="header " , ] simulated showers.,width=264,height=264 ] since the measurement errors are negligibly small , the significance of the mutually compensative effects is well seen .
the singularity of the proton spectrum was approximately compensated by the @xmath125 and @xmath126 spectra .
this is due to both the large number ( @xmath127 ) of possible mutually compensative combinations ( 3 ) and the peculiarities of eas development in the atmosphere ( kernel functions @xmath36 , section 3 ) , which are expressed by the approximately log - linear dependences of the statistical parameters @xmath128 , @xmath129 , @xmath130 and @xmath131 of shower spectra @xmath36 on energy ( @xmath132 ) and nucleon number ( @xmath133 ) of primary nuclei @xcite .
the value of @xmath134 for a @xmath135 times smaller eas sample ( @xmath136 ) was equal to @xmath137 .
the results from figs . 13 show that the pseudo functions with mutually compensative effects exist and belong practically to all families - linear ( 7 ) , non - liner ( 8) and even singular ( 9 ) in a logarithmic scale .
+ the all - particle energy spectra in figs
. 13 are practically indifferent to the pseudo solutions of elemental spectra .
this fact directly follows from eq .
( 2 ) for pseudo solutions and is well confirmed by the identity of the gamma @xcite and kascade @xcite all - particle energy spectra in spite of disagreements of the elemental ( @xmath138 ) primary energy spectra ( see @xcite ) .
+ the @xmath62 minimization ( 6 ) uses mainly the nearest pseudo energy spectra with free parameters for compensation of the pseudo spectra with fixed parameters . and @xmath2 primary nuclei ( light shaded areas ) and corresponding `` methodical errors '' of the kascade unfolding spectra ( dark shaded areas ) taken from @xcite .
the solid and dotted lines resulted from pseudo functions ( 7 ) and the dashed lines stemmed from ( 8).,width=264,height=264 ] the significance of the pseudo functions @xmath51 in most cases exceeds the significance of the evaluated primary energy spectra @xmath4 and unfolding of ( 1 ) can not be effective for @xmath139 . + the unfolding of the primary energy spectra for @xmath50 will increase the number of possible combinations ( 3 ) of the pseudo solutions and the corresponding pseudo functions by a factor of two .
taking into account the large values of applied @xmath140 @xcite one may conclude that the contributions of the pseudo functions in the unfolded energy spectra of @xcite have to be dominant .
+ the `` methodical errors '' obtained in @xcite for @xmath50 define the uncertainties of the solutions intrinsic only to the given unfolding algorithms .
the existence and significance of the mutually compensative pseudo solutions follow from eqs .
( 1,2 ) and from the peculiarities of the shower spectra @xmath6 regardless of the unfolding algorithms .
+ comparison of the methodical errors @xmath141 for @xmath142 and @xmath143 from @xcite with corresponding errors @xmath144 due to the pseudo solutions from expressions ( 7,8 ) are shown in fig . 4 .
the magnitudes of the fixed parameters were empirically determined by maximizing @xmath145 ( left panel ) and @xmath146 ( right panel ) for a given goodness - of - fit test @xmath147 from @xcite .
+ it is seen that the methodical errors ( dark shaded areas ) from @xcite significantly underestimate the contribution of the pseudo solutions ( light shaded areas ) from expressions(7,8 ) .
moreover , the methodical errors from @xcite slightly depend on the primary energy ( or statistical errors ) , whereas the domains of the pseudo solutions strongly correlate with the statistical errors according to definition ( 2 ) .
the results show that the reconstruction of primary energy spectra using unfolding algorithms @xcite can not be effective and the disagreement between the kascade @xcite and gamma @xcite data is insignificant in comparison with the large domains of the mutually compensative pseudo solutions ( fig 4 ) of the unfolded spectra @xcite . + even though the oscillating pseudo solutions @xmath148 ( section 2 ) are possible to avoid using regularization algorithms @xcite , the mutually compensative effect ( 4 ) of the arbitrary pseudo functions @xmath149 intrinsic to the expression ( 2 ) is practically impossible to avoid at @xmath150 .
+ the uncertainties of solutions due to the mutually compensative pseudo functions can be obtained by varying the initial values of iterations within a wide range in the frameworks of a given unfolding algorithm .
+ to decrease the contributions of the mutually compensative pseudo solutions one may apply a parameterization of the integral equations ( 1 ) @xcite using _ a priori _
( expected from theories @xcite ) known primary energy spectra with a set of free spectral parameters .
this transforms the eas inverse problem into a set of equations with unknown spectral parameters , and thereby the eas inverse problem is transmuted into a test of the given primary energy spectra using detected eas data @xcite .
the reliability of the solutions can be determined by their stability depending on the number of spectral parameters , the agreement between the expected and detected eas data sets , and the conformity of the spectral parameters with theoretic predictions .
+ the all - particle energy spectra ( fig .
13 ) are practically indifferent toward the pseudo solutions for elemental spectra .
+ the obtained results depend slightly on the spectral representations of the shower spectra @xmath36 and the primary energy spectra @xmath4 .
i thank my colleagues from the gamma experiment for stimulating this work and the anonymous referee for suggestions which considerably improved the paper .
25 r. glasstetter et al . , proc .
26@xmath151 icrc , salt lake city , * 1 * ( 1999 ) 222 .
ter - antonyan , l.s .
haroyan , preprint hep - ex/0003006 ( 2000 ) .
h. ulrich et al . , proc .
27@xmath151 icrc , hamburg , * 1 * ( 2001 ) 97 .
ter - antonyan and p.l .
biermann , proc .
27@xmath151 icrc , hamburg , he054 ( 2001 ) 1051 ( astro - ph/0106091 ) .
samvel ter - antonyan and peter biermann , proc .
28@xmath151 icrc , tsukuba , he1 , ( 2003 ) 235 ( astro - ph/0302201 ) .
t. antoni et al . , astropart . phys . * 24 * ( 2005 ) 1 ( astro - ph/0505413 ) .
martirosov et al . , 29@xmath151 icrc , pune , he1.5 , * 8 * ( 2005 ) 9 ( astro - ph/0506588 ) . m. aglietta et al . , astropart . phys . * 10 * ( 1999 ) 1 .
glasmacher et al .
, astropart . phys .
* 10 * ( 1999 ) 291 .
t. antoni et al . , nucl .
@xmath152 meth . *
a513 * ( 2003 ) 490 .
ter - antonyan et al . , 29@xmath151 icrc , pune , he1.2 , * 6 * ( 2005 ) 105 ( astro - ph/0506588 ) .
a.p . garyka et al . , astropart .
( 2007 ) , doi : 10.1016/j.astropartphys .
2007.04.004 ( arxiv:0704.3200v1 [ astro - ph ] ) .
d. heck , j. knapp , j.n .
capdevielle , g. schatz , t. thouw , forschungszentrum karlsruhe report , fzka 6019 ( 1998 ) .
fletcher , t.k .
gaisser , p. lipari , t. stanev , phys.rev . *
d50 * ( 1994 ) 5710 .
r. gold , anl-6984 report , argonne ( 1964 ) .
b. wiebel - sooth , p.l . bierman and h. meyer ,
astrophys . 330
( 1998 ) 330 .
a.m. hillas , journal of physics * g31 * ( 2005 ) r95 .
hrandel , astropart . phys . * 21 * ( 2004 ) 241 .
t. stanev , p.l .
biermann , t.k .
gaisser , astron .
274 , ( 1993 ) 902 .
ter - antonyan , 28@xmath151 icrc , tsukuba , he2 ( 2003 ) 239 ( astro - ph/0303658 ) .
j.r.hrandel , astro - ph/0611387 ( 2006 ) . | the problem of the uniqueness of solutions during the evaluation of primary energy spectra in the knee region using an extensive air shower ( eas ) data set and the eas inverse approach is investigated .
it is shown that the unfolding of primary energy spectra in the knee region leads to mutually compensative pseudo solutions .
these solutions may be the reason for the observed disagreements in the elementary energy spectra of cosmic rays in the 1 - 100 pev energy range obtained from different experiments .
cosmic rays , primary energy spectra , extensive air shower , inverse problem .
96.40.pq , 96.40.de , 96.40.-z , 98.70.sa |
considerable resources are being devoted to advancing ingan light - emitting diodes ( leds ) , largely because of solid - state lighting . of particular interest is improving device efficiency @xcite , especially the mitigation of efficiency degradation with increasing excitation , i.e. , efficiency droop shen , hader2,dellaney .
to arrive at a solution , it is important to understand the principal mechanism of emission together with the effects of carrier leakage @xcite , auger recombination @xcite , junction heating @xcite , carrier and defect delocalizations chichibu , smowton , hader1 . equally important
is the change in efficiency with temperature , not only for application but also for better understanding of the physics contributing to efficiency droop .
this paper investigates led efficiency as functions of current density and lattice temperature .
the analysis uses a model that allows direct input of band - structure properties @xcite .
band - structure details are important , because underlying emission properties in wurtzite quantum - well ( qw ) structures are the changes in band energy dispersions , confinement energies and optical transition matrix elements with excitation .
these changes arise from screening of piezoelectric and spontaneous polarization fields @xcite .
the present approach differs substantially from the total - carrier - density rate - equation description used in most discussions involving ingan led efficiency . in terms of reproducing experimental efficiency versus injection current data , a particularly successful rate - equation model is the @xmath1 model @xcite .
the model s name derives from the phenomenological constants @xmath2 , @xmath3 and @xmath4 introduced to account for shockley - read - hall ( srh ) , radiative - recombination and auger - scattering carrier losses , respectively .
band - structure properties enter indirectly via these coefficients .
section 2 summarizes the model and presents the working equations .
section 3 demonstrates the application of the model by calculating internal quantum efficiency ( iqe ) as a function of injection current for different lattice temperatures .
following ref .
hader3 , the calculations are performed for a led with an in@xmath5ga@xmath6n qw active region .
the choices of input parameters are explained and comparison to recent experimental results is discussed .
this section also describes the contributions from auger carrier loss , plasma heating and carrier leakage .
then , the role of the band structure is illustrated by considering a led with an in@xmath7ga@xmath8n active region .
section 4 summarizes the paper .
the derivation of spontaneous emission from qw and barrier transitions starts with a hamiltonian adapted from quantum optics @xcite . with this hamiltonian ,
hiesenberg operator equations of motion are derived for the carrier populations and polarizations .
these equations contained operator products describing coupling to higher order correlations in light - matter interaction .
truncating at the hartree - fock level , as well as adding phenomenologically srh carrier loss and relaxation contributions from carrier - carrier and carrier - phonon scattering , lead to the working equations for the present investigation @xcite . for the qw population @xmath9 , @xmath10 \notag \\ & & -\gamma _ { c - p}\left [ n_{\sigma , n_{\sigma } , k_{\bot } } -f\left ( \varepsilon _ { \sigma , k_{\perp } } , \mu _ { \sigma } ^{l},t_{l}\right ) \right ] \notag \\ & & -\gamma
_ { ag}\left [ n_{\sigma , n_{\sigma } , k_{\bot } } -f\left ( \varepsilon _ { \sigma , k_{\perp } } , \mu _ { ag},t_{ag}\right ) \right ] \label{e1}\end{aligned}\]]where each qw state is denoted by the subscripts @xmath11 , representing charge ( @xmath12 or @xmath13 ) , subband index ( @xmath14 ) and in - plane momentum ( @xmath15 ) . in the above equation , @xmath2 is the srh coefficient , @xmath16 and @xmath17 are the effective carrier - carrier and carrier - phonon collision rates , respectively .
carrier loss by auger scattering is modeled with a term containing an effective auger scattering rate @xmath18 , where @xmath19 is the auger coefficient and @xmath20 is the total ( qw and barrier ) carrier density . in the summation , @xmath21 or @xmath22 ,
when @xmath12 or @xmath13 . for the barrier carrier population @xmath23 , @xmath24 \notag \\ & & -\gamma _ { c - p}\left [ n_{\sigma , k}^{b}-f\left ( \varepsilon _ { \sigma , k}^{b},\mu _ { \sigma } ^{l},t_{l}\right ) \right ] \notag \\ & & -\gamma _ { ag}\left [ n_{\sigma , k}^{b}-f\left ( \varepsilon _ { \sigma , k}^{b},\mu _ { ag},t_{ag}\right ) \right ] \label{e2}\end{aligned}\]]where each barrier state
is denoted by the subscripts @xmath25 , representing charge ( @xmath26 ) and 3-dimensional momentum ( @xmath27 ) . in the above equation
, @xmath28 is the current density , @xmath22 is the electron charge and @xmath29 .
the injected carrier distribution @xmath30 is a fermi - dirac function with chemical potential @xmath31 , temperature @xmath32 . in the above equations ,
@xmath33 and @xmath34 are the carrier energies in the qw and barrier , respectively . from the derivation ,
the spontaneous emission coefficients are given by @xmath35 @xmath36where @xmath37 and @xmath38 are the qw dipole matrix element and transition energy , @xmath39 and @xmath40 are the barrier dipole matrix element and transition energy , and @xmath41 is the host permittivity . for the asymptotic fermi - dirac distributions approached via carrier - carrier collisions @xmath42 and @xmath43 , the chemical potential @xmath44 and plasma temperature @xmath45 are determined by conservation of carrier density and energy . for @xmath46 and @xmath47 , which are reached via carrier - phonon collisions , the chemical potential @xmath48 is determined by conservation of carrier density and the lattice temperature @xmath49 is an input quantity . for @xmath50 and @xmath51 , which are reached via auger scattering , there is a loss of carrier density that eventually leads to a chemical potential @xmath52 of the intrinsic semiconductor , i.e. , approximately at mid gap , with its exact location depending on electron and hole dispersions .
@xmath53 is determined by noting that energy is conserved in an auger scattering event . when summed over all states , eqs .
( 1 ) plus ( 2 ) reduce to total carrier density rate equations used in the @xmath1 model , if we assume that the total emission contribution equals @xmath54 , where @xmath3 is an effective bimolecular radiative coefficient . in the summation over states ,
quantities such as the total carrier density and energy are computed by converting the sum over states to integrals , i.e. , @xmath55where @xmath56 and @xmath13 are the surface area and thickness of the active active region consisting of the qw and barrier layers .
the conversion to bulk ( 3-d ) density is via division by the qw layer width @xmath13 in the case of the qw and by the total barrier width @xmath57 in the case of the barrier .
further discussions involving implementation and comparison with results from quantum - kinetic calculations may be found in earlier reports chow1,chow2,waldmueller .
the dynamical solution gives the carrier densities in qw and barrier states .
band - structure information enters directly into eqs ( [ e1 ] ) and ( [ e2 ] ) via the dipole matrix elements @xmath58 , @xmath39 and carrier energies @xmath59 , @xmath60 . iterative solution of the @xmath61 and poisson equations @xcite gives the qw energies and dipole matrix elements .
for this calculation , the necessary bulk wurtzite material parameters are given in refs .
jenkins , wright , wei , ambacher .
when performing the band - structure calculation , quasiequilibrium condition is assumed to determine the qw and barrier bulk densities used in the solution of poisson equation .
this is an inconsistency that is acceptable provided the dynamical solution does not produce carrier distributions deviating too far from quasiequilibrium distributions .
application of the model involves first calculating the band structure for a range of carrier densities by solving the @xmath61 and poisson equations .
this gives the band - structure information needed for the solving the population dynamical equations .
to facilitate the numerics , carrier states are grouped into two categories : those belonging to the qws and those belonging to the barriers . with a high internal electric field , the distinction between qw and barrier states may be ambiguous . in this paper ,
the choice is made by calculating @xmath62dz@xmath63 , where integral is performed over the qws and @xmath64 is the wavefunction .
the states where the integral is greater than @xmath65 are grouped as qw states and the rest as barrier states . for the problem
being addressed , which are the excitation and temperature dependences of iqe , the distinction is only important because only qw transitions are affected by the quantum - confined stark effect ( qcse ) . for the barrier transitions ,
the dipole matrix element in the presence of an internal electric field is approximated by an average , where each transition is weighted according to the occupations of the participating states .
grouping the barrier states appreciably reduces numerical demand , which remains substantial because one is still keeping track of a large number of @xmath27-states . in the numerical solution of eqs .
( [ e1 ] ) and ( [ e2 ] ) , the band - structure quantities are updated at each time step according to the instantaneous carrier density .
when steady state is reached , iqe is determined from dividing the rate of carrier ( electron or hole ) loss via spontaneous emission by the rate of carrier injection , i.e. @xmath66
the above model was applied to compute the temperature dependence of iqe versus current density .
a motivation is that adding temperature dependence may produce further physical insight to efficiency droop , as well as provide more stringent testing of the model . in order to relate to experiments
expt1,expt2,hader3 , simulations were performed for two ledconfigurations .
the experimental devices had single - qw active regions , which circumvented complications arising from nonuniform population in multi - qw structures .
one device consisted a 2 nm in@xmath5ga@xmath6n qw between gan barriers @xcite , while the other device consisted a 3 nm in@xmath7ga@xmath8n qw between gan barriers @xcite .
the experiments @xcite show features of iqe dependences on excitation and temperature , that if understood , may shed light on led physics in general and the efficiency droop in particular .
they involve changes in shape of iqe versus current density curves with temperature and the relationships among these curves after the onset of droop . in the former , for lattice temperatures @xmath67 and above , both in@xmath5ga@xmath68n and in@xmath7ga@xmath8n leds exhibit iqe versus current density behavior showing the familiar efficiency droop , as described relatively well by the abc model .
interestingly , a second iqe bump appears at lower temperatures for the in@xmath5ga@xmath6n sample and becomes very pronounced at 100k .
the abc model is unable to describe this shape change , which does not occur in the in@xmath7ga@xmath8n sample at any temperature .
a goal of this study is to use our model to reproduce and understand the origin of the second bump and determine whether the different behavior between the two leds is intrinsic , i.e. , involving only differences in band structure .
additionally , the model will be used to examine the detailed relationship among iqe curves at high current densities . in the in@xmath5ga@xmath6n case
, the curves for different temperatures are clearly separated for the entire measured current density range . on the other hand , for in@xmath69ga@xmath8n , the iqe versus current curves for different temperatures either cross or merge with each other at high current densities .
previous modeling of these behaviors has lead to puzzling results , such as inferring a decrease in @xmath70 coefficient with increasing temperature , contrary to what is expected of auger scattering @xcite .
figure 1 shows results from the present model for the in@xmath5ga@xmath6n led .
the curves describe iqe versus current density at different lattice temperatures .
similar to experiment , there is a double bump excitation dependence in iqe .
the lower excitation bump decreases with increasing temperature and vanishes for @xmath71 .
input parameters that are assumed temperature dependent are the srh and auger coefficients , @xmath2 and @xmath70 , respectively and the carrier - phonon scattering rate @xmath72 they are adjusted to produce iqe behavior resembling those found in refs . @xcite and @xcite .
the srh coefficient is decreased with decreasing temperature to give the onset of emission as depicted in the figure .
the auger coefficient is adjusted to reproduce approximately the experimental maximum iqe values .
since auger carrier loss conserves energy , significant plasma heating may occur .
the role of carrier - phonon scattering is to relax the hot carrier distributions back to the lattice temperature .
both @xmath70 and @xmath73 affect the peak iqe .
however , the current density where iqe peaks is relatively sensitive to @xmath73 , which is chosen to produce an onset of droop in the tens of a/@xmath74 range .
figure 2 plots the values of @xmath2 , @xmath70 and @xmath73 versus temperature used in computing the curves in fig .
a decrease in @xmath2 with decreasing temperature is expected for defect related loss .
figures 2 ( b ) and 2 ( c ) show increases in auger coefficient and carrier - phonon scattering rate with increasing temperature , which are consistent with microscopic calculations .
the values of auger coefficient are within the range predicted for phonon - assisted auger scattering @xcite and smaller than most values obtained from experimental curve fitting with the abc model @xcite .
the values used for carrier - phonon scattering are higher than predicted by quantum kinetic calculations for typical iii - n structures .
however , it should be noted that those calculations are for intraband scattering among nearby states .
the present effective rates represent the relaxation of very energetic states populated by auger scattering to the qw ground state .
the energy differences are in the neighborhood of the bandgap energy ( @xmath75 ) . to give some insight into the origin of the double bump , fig .
3 ( a ) shows the spontaneous emission contributions from qw and barrier states versus current density for the @xmath76 curve in fig .
clearly , the spontaneous emission from barrier states dominates the emission from the qw at low current density , coinciding with the location of first iqe bump [ see left arrows in figs . 1 and 3 ( a ) ] . here ,
optical emission from barrier transitions occurs via the contribution @xmath77 , as soon as the product of electron and hole populations , @xmath78 becomes nonzero .
however , there may be no optical emission from qw transitions because of absence of localized qw states from the strong piezoelectric and spontaneous polarization fields ( quantum - confined stark effect ) .
there is a transition region where the screening of internal electric field results in the appearance of localized qw states . here
, the small energy separation between qw and barrier states gives comparable qw and barrier populations , especially for the holes .
however , the qw emission contribution @xmath79 can still be negligible , even though the product @xmath80 may be appreciable , because qcse spatially separates electrons and holes in the qws , resulting in very small dipole matrix elements for qw transitions . at high current density
[ see right arrows in fig . 1 and 3 ( a ) ]
, plasma screening restores the qw localized states and optical - dipole matrix element , leading to qw emission overtaking that from the barrier because of the advantage in 2-dimensional versus 3-dimensional carrier density of states .
the above explanation is supported by figs . 3 ( b ) and 3 ( c ) .
the solid curves are absolute square of envelope functions for electrons and holes at current densities indicated by arrows in fig .
each curve is displaced according to its energy for clarity .
envelope functions belonging to qw and barrier states are indicated by black and grey curves , respectively .
the former is clearly missing in fig .
3 ( b ) . the black dashed lines plot the confinement potentials and
they indicate a significantly stronger deformation in fig .
3 ( b ) compared to fig .
3 ( c ) due to the higher internal electrical fields in the lower current density situation . figure 4 plots the contributions to the @xmath76 iqe curve for various recombination and scattering processes .
the solid red curve is the sum of qw and barrier emission , with a slight change in slope at low current density indicating the transition from predominately barrier to predominately qw emission .
a second slope change at higher current density arises from the onset of auger carrier loss .
srh and leakage ( or failure to capture ) losses from the barrier is depicted by the blue dashed curve .
they are primarily responsible for the low iqe at low current densities .
the green dotted curve shows that defect ( srh ) loss from qw states are essentially negligible .
auger carrier loss accounts for the difference between the solid black curve and the sum of all the other curves . to illustrate the auger contribution , fig .
5 ( a ) shows the iqe versus current density for two auger coefficients : @xmath81 , which is used in fig . 1 , and @xmath82 , which is in the neighborhood of what is typically expected for materials with bandgap energy of approximately 2.7ev .
the dashed curve clearly shows that for @xmath70 less than @xmath83 , auger carrier loss has negligible effect on iqe .
the difference between the two curves indicates the auger contribution to the iqe curve in fig . 1 for @xmath76 .
during each auger event , an electron and a hole recombine , with the loss in energy transferred to a second electron or hole . in a wide bandgap material , such as ingan ,
the transferred energy is significant and can lead to highly - energetic carrier distributions , depending on the carrier - phonon scattering rate .
because of rapid carrier - carrier scattering , this change in carrier distributions may be described in terms of the plasma temperature @xmath45 .
figure 5 ( b ) shows the plasma temperature versus current density for the two auger coefficients .
the solid curve indicates significant rise in temperature for the @xmath76 curve in fig .
1 . the dashed curve deoicts the rise in temperature in the absence of auger loss , so that heating comes mainly from the relaxation of carriers from the barrier states , where they are injected , to the ground states , where emission occurs .
the role of band structure is investigated by performing simulations for the in@xmath7ga@xmath8n led , using the same parameter values as plotted in fig .
figure 6 shows the iqe versus current density for different temperatures .
clearly indicated is a missing or negligible second iqe bump .
band - structure calculations traced the reason to a smaller internal electric field with lower in concentration in the qw .
consequently , localized qw states exist for the entire temperature and current - density range considered , and their emission exceeds that of the barrier states at all current densities .
the absence of a transition from barrier to qw emission dominance gives rise to the lack of a double bump shape .
experimental evidence supporting the above argument may be found in expt2 .
there , the iqe versus current density curves shows single - bump behavior for @xmath49 between @xmath84 and @xmath67 , as in fig .
however , agreement is not complete for the simulations fail to exactly reproduce experimental observation at high current density , where the experiment shows crossing of the iqe curves for different temperatures . at high
excitation , the fitting parameters used in this study which influence led behavior are c and @xmath73 .
we had assumed that their values do not change considerably with a change in indium concentration from ref .
@xcite , we again consider a , c and @xmath73 as free - parameters .
figure 7 shows the iqe results for input parameters given in fig .
figure 7 ( a ) indicates better agreement with with ref . 14 .
however , agreement between theory and experiment for the in@xmath5ga@xmath85n led with the new set of parameters is degraded . on the other hand
the single- versus double - bump behavior remains consistent with experiment and appears robust to changes in the fitting parameters .
this paper uses an approach to modeling ingan leds that involves the self - consistent solution of band structure and carrier dynamics .
one motivation is systematic incorporation into a model , at a microscopic level and on equal footing , the effects of spontaneous emission , carrier capture and leakage , as well as nonequilibrium effects such as plasma heating .
another motivation is to provide direct input of band - structure properties , in particular , their carrier - density dependences arising from screening of piezoelectric and spontaneous polarization fields .
doing so allows comparison of different devices at the heterostructure - design level .
the end result is a non - equilibrium microscopic model that provides a more precise estimation of relative strengths of possible physical contributions compared to the commonly used @xmath1 model @xcite and is easier to implement than a first - principles , many - body approach @xcite .
the model reproduces the main experimental features of the observed droop behavior versus temperature relatively successfully .
shape changes of iqe versus current density with temperature are described with the optical emission from quantum - well and barrier transitions treated on equal footing .
the iqe droop is described by auger scattering and resulting plasma heating , with material parameters and their temperature dependences that are consistent with microscopic calculations . however , there are remaining discrepancies .
for example , while the experimental and theoretical maximum iqes can be made to roughly agree , the current densities where they occur are overestimated by the model .
several mechanisms may be responsible , e.g. , effects arising from doping profile , presence of carrier blocking layers and interface irregularities @xcite .
these effects are ignored in the present study , but may be readily incorporated into the model .
also excluded is a more detailed description of defect recombination , e.g. as proposed for describing the experimental data @xcite .
more serious may is the neglect of many - body effects , such as treated in refs @xcite and @xcite .
there is also room for refinement in treating the connection between the band - structure and population - dynamics aspects .
carrying out these improvements increases computational demands , and their implementation can be greatly facilitated by having more experimental data and better knowledge details on the experimental configurations .
this work is performed at sandia s solid - state lighting science center , an energy frontier research center ( efrc ) funded by the u. s. department of energy , office of science , office of basic energy sciences .
the author thanks helpful discussions with j. wierer , the hospitality of the technical university berlin and travel support provided by sfb787 . 99 m.r .
krames , o.b .
shchekin , r. mueller - mach , g.o .
mueller , l. zhou , g. harbers , and m. g. craford , status and future of high - power light - emitting diodes for solid - state lighting , j. display technology * 3 * , 160 - 175 ( 2007 ) .
j. hader , j.v .
moloney , b. pasenow , s.w .
koch , m. sabathil , n. linder , and s. lutgen , on the important of radiative and auger losses in gan - based quantum wells , appl .
92 * , 261103 - 261105 ( 2008 ) .
a. a. efremov , n. i. bochkareva , r. i. gorbunov , d. a. larinvovich , yu .
t. rebane , d. v. tarkhin , and yu .
g. shreter , effect of the joule heating on the quantum efficiency and choice of thermal conditions for high - power blue ingan / gan leds , semiconductors * 40 * , 605 ( 2006 ) .
s. f. chichibu , t. azuhata , m. sugiyama , t. kitamura , y. ishida , h. okumurac , h. nakanishi , t. sota , and t. mukai , optical and structural studies in ingan quantum well structure laser diodes , j. vac , sci .
b * 19 * , 2177 ( 2001 ) .
a. laubsch , m. sabathil , w. bergbauer , m. strassburg , h. lugauer , m. peter , s. lutgen , n. linder , k. strebubel , j. wagner , j. hader , j. v. moloney and s.w .
koch , on the origin of iqe-droop in ingan leds , phys .
status solidi c 6 , s913-s916 ( 2009 ) . s. choi , h. j. kim , s .- s .
kim , j. liu , j. kim , j .- h .
ryou , r. d. dupuis , a. m. fishcer , and f. a. ponce , improvement of peak quantum efficiency and efficiency droop in iii - nitride visible light - emitting diodes with an inaln electron - blocking layer , appl .
. lett . * 96 * , 221105 - 221107 ( 2010 ) .
1 . iqe versus current density for led with single in@xmath5ga@xmath86n quantum well and temperatures @xmath87 to @xmath88 at @xmath89 intervals .
the arrows for the @xmath76 iqe curve indicate the current densities for figs .
3(b ) and 3(c ) .
3 . ( a ) spontaneous emission contributions from qw and barrier states ( solid and dashed curves ,
respectively ) versus current density for @xmath90 curve in fig .
1 . solid curves in ( b ) and ( c ) are absolute square of envelope functions at zone center ( @xmath91 ) for electrons and holes at current densities indicated by arrows in fig .
the x - axis is along the growth direction .
each curve is displaced according to its energy for clarity .
envelope functions belonging to qw and barrier states are indicated by black and grey curves , respectively . the black dashed lines plot the confinement potentials .
( a ) iqe and ( b ) plasma temperature versus current density . in both plots ,
the lattice temperature is @xmath76 and auger coefficients are @xmath92 and @xmath83 ( solid and dashed curves , respectively ) .
iqe versus current density for led with single 3 nm in@xmath7ga@xmath93n qw and temperatures @xmath94 to @xmath88 at @xmath89 intervals .
the srh and auger coefficients , as well as the carrier - phonon scattering rates are same as those in fig . 1 for the in@xmath5ga@xmath6n structure . fig .
iqe versus current density for led with ( a ) 3 nm in@xmath7ga@xmath93n qw and ( b ) 2 nm in@xmath5ga@xmath6n qw . the latticetemperatures are @xmath94 to @xmath67 at @xmath89 intervals .
the srh and auger coefficients , as well as the carrier - phonon scattering rates are plotted in fig . | the changes in excitation dependence of efficiency with temperature is modeled for a wurtzite ingan light - emitting diode .
the model incorporates bandstructure changes with carrier density arising from screening of quantum - confined stark effect .
bandstructure is computed by solving poisson and k@xmath0p equations in the envelop approximation .
the information is used in a dynamical model for populations in momentum - resolved electron and hole states .
application of the approach shows the interplay of quantum - well and barrier emissions giving rise to shape changes in efficiency versus current density with changing temperature , as observed in some experiments .
ocis : ( 230.3670 ) light - emitting diodes ; ( 230.5590 ) quantum - well , -wire and -dot devices ; ( 250.5590 ) quantum - well , -wire and -dot devices . |
the mass spectrum of black holes ( bhs ) that form from the collapse of massive stars is highly uncertain .
an accurate dynamical mass estimate has been derived only for @xmath10 stellar bhs ( see table 2 of zel et al .
2010 for one of the most updated compilations , but see also lee , brown & wijers 2002 ; orosz 2003 ; narayan & mcclintock 2005 ) .
most of the derived bh masses are in the range @xmath11 , with an apparent absence ( in x - ray binaries ) of bhs with @xmath12 m@xmath2 , difficult to explain with observational biases ( zel et al .
2010 ) . in the milky way ( mw ) ,
the most massive bhs in x - ray binaries do not seem to significantly exceed @xmath13 m@xmath2 ( e.g. @xmath14 m@xmath2 for gs2023 + 338 , charles & coe 2006 ; @xmath15 m@xmath2 for grs 1915 + 105 , harlaftis & greiner 2004 ) , whereas a few bhs in nearby galaxies may have higher masses : m33 x-7 ( @xmath16 m@xmath2 , orosz et al . 2007 ) , ic 10 x-1 ( @xmath17 m@xmath2 , prestwich et al .
2007 ; silverman & filippenko 2008 ) and ngc 300 x-1 ( @xmath18 m@xmath2 , carpano et al .
2007 ; crowther et al . 2007 ; @xmath19 m@xmath2 , under reasonable assumptions for the inclination and for the mass of the companion , crowther et al . 2010 ) .
interestingly , these three relatively massive stellar bhs are hosted in regions with relatively low metallicity .
a metallicity @xmath20 z@xmath2 is estimated for the dwarf irregular galaxy ic 10 from an electron - temperature based calibration of spectra of hii regions ( @xmath21 , garnett 1990 ) , assuming z@xmath22 .
the metallicity of m 33 in proximity of x-7 ( i.e. at @xmath23 , where @xmath24 is the holmberg radius ) is @xmath25 z@xmath2 , and that of ngc 300 in the vicinity of x-1 ( i.e. at @xmath26 ) is @xmath27 z@xmath2 ( derived from the metallicity gradients of m 33 and of ngc 300 , respectively , provided by pilyugin , vlchez & contini 2004 ) . from a theoretical perspective
, our knowledge of the mass spectrum of stellar bhs is hampered by the uncertainties about two issues : ( i ) mass losses by stellar winds in massive stars ( during and especially after the main sequence , ms ) ; ( ii ) the hydrodynamics of core - collapse supernova ( sn ) explosions .
the rate of mass loss by stellar winds during the ms likely increases with the metallicity of the star ( @xmath28 , where @xmath29 , depending on the model , e.g. kudritzki , pauldrach & puls 1987 ; leitherer , robert & drissen 1992 ; maeder 1992 ; kudritzki & puls 2000 ; vink , de koter & lamers 2001 ; kudritzki 2002 ; belkus , van bever & vanbeveren 2007 ; pauldrach , vanbeveren & hoffmann 2012 ) . the behaviour of post - ms massive stars , such as luminous blue variable stars ( lbvs ) and wolf - rayet stars ( wrs ) , is much more uncertain ( e.g. vink & de koter 2005 ) . according to models of stellar evolution and sn explosion ( fryer 1999 ; fryer & kalogera 2001 ; woosley , heger & weaver 2002 ; heger et al .
2003a ) , a star with a final mass , of a star the mass bound to the star immediately before the collapse . ]
@xmath30 m@xmath2 can collapse quietly to a bh , after a weak ( if any ) sn explosion ( failed sn scenario ) . in the following
, we use the terms ` failed sn ' and ` direct collapse ' as synonymous , to indicate the silent collapse of a star to a bh after no or weak sn .
the actual value of the minimum final mass for a star to directly collapse into a bh is quite uncertain : our adopted fiducial value of 40 m@xmath2 is a conservative assumption ( fryer 1999 ) , and searches for sn progenitors provide some evidences for this theoretical scenario ( see smartt 2009 for a recent review ) . since
stellar winds are suppressed at low metallicity , metal - poor massive stars are more likely to have @xmath30 m@xmath2 than their metal - rich analogues . in the case of a failed sn , it is reasonable to expect that the mass of the remnant is comparable to the final mass of the progenitor star ( or at least more than half of it , see e.g. heger et al .
therefore , bhs may form , via this channel , with mass higher than in the case of sn explosion . according to belczynski
et al . ( 2010 ; hereafter b10 ) , the mass of a bh formed via direct collapse may be as high as @xmath31 m@xmath2 .
the models by b10 consider only the evolution of single , non - rotating stars .
rotation and binarity can affect the final mass of the remnant ( e.g. vanbeveren 2009 ; maeder & meynet 2010 , 2012 ) . in the following , we will refer to massive stellar bhs ( msbhs ) to indicate bhs with mass @xmath32 m@xmath2 formed via direct collapse .
the existence of msbhs in the nearby universe may be crucial for our understanding of x - ray sources .
the scenario of x - ray binaries powered by msbhs was recently proposed to explain a large fraction of the ultra - luminous x - ray sources ( ulxs , i.e. x - ray sources with luminosity , assumed isotropic , higher than @xmath33 erg s@xmath34 ) , without requiring excessive super - eddington factors or more exotic mechanisms ( e.g. mapelli , colpi & zampieri 2009 ; zampieri & roberts 2009 ; mapelli et al .
2010 , 2011a ) . in this paper ,
we present new @xmath0body simulations of dense intermediate - mass ( a few @xmath35 m@xmath2 ) young ( @xmath36 myr ) star clusters ( scs ) , including an accurate treatment of dynamics and updated recipes for metal - dependent stellar evolution , stellar winds and failed sne .
our aim is to study the formation and dynamical evolution of stellar bhs and msbhs in young scs with different metallicity . in the current paper ( which is the first of a series )
, we will focus on the effects of the dynamics of stellar bhs and msbhs on the population of x - ray sources . in the next papers of the series we will consider also other effects of bh dynamics ( e.g. the consequences for the population of gravitational wave sources ) .
we simulate scs , as most stars ( @xmath31 per cent ) are expected to form in scs ( e.g. lada & lada 2003 ) .
we restrict our analysis to intermediate - mass ( a few @xmath35 m@xmath2 ) young scs .
these form with a higher frequency than larger clusters in the nearby universe ( as the mass function of star clusters is @xmath37 , with @xmath38 , lada & lada 2003 ) , and are often sites of an intense x - ray activity : many bright high - mass x - ray binaries ( hmxbs ) and ulxs are associated with ob associations and with young intermediate - mass scs ( e.g. goad et al .
2002 ; zezas et al . 2002
; liu , bregman & seitzer 2004 ; soria et al . 2005 ; ramsey et al .
2006 ; terashima , inoue & wilson 2006 ; abolmasov et al . 2007 ; berghea 2009 ; swartz , tennant & soria 2009 ; tao et al . 2011 ; gris et al .
2011 , 2012 ) .
tens of intermediate - mass young scs have been discovered in the mw in the last few years ( e.g. bica et al .
2003 ; mercer et al . 2005 ; borissova et al . 2011 ; richards et al . 2012 ) .
the paper is organized as follows . in section 2 , we briefly review the previous work on this topic , to summarize the state - of - art and to highlight the differences with the present analysis . in section 3 ,
we describe the method adopted for the simulations . in section 4 , we present the results , focusing on the mass spectrum of bhs and on the effects of dynamics on accreting bhs in x - ray binaries . in section 5 ,
we summarize the most relevant results and discuss future challenges for @xmath0body simulations of scs .
most population synthesis codes study the formation of bhs from single stars and from stars in primordial binaries ( i.e. stars that are in the same binary since their formation ) .
this method has been widely used to investigate the population of x - ray binaries ( e.g. portegies zwart , verbunt & ergma 1997 ; hurley , tout & pols 2002 ; podsiadlowski , rappaport & pfahl 2002 ; podsiadlowski , rappaport & han 2003 ; belczynski et al .
2004a ; belczynski , sadowski & rasio 2004b ; rappaport , podsiadlowski & pfahl 2005 ; dray 2006 ; madhusudhan et al .
2006 , 2008 ; belczynski et al . 2008 ; linden et al .
2010 , hereafter l10 ) .
some studies include also recipes for the effects of metallicity on stellar evolution ( e.g. hurley et al .
2002 ; belczynski et al . 2004a ; dray 2006 ; belczynski et al . 2008 ; l10 ) . in particular ,
l10 adopt recipes for stellar winds and failed sne that are very similar to b10 , and find that msbhs are unlikely to power bright x - ray binaries .
this occurs because primordial binaries merge if they are sufficiently tight to enter a common envelope ( ce ) phase before the sn of the primary and if they are still in roche lobe overflow ( rlo ) at the end of the ce phase .
if they survive the ce phase , they may be tight enough to later enter a rlo phase when the secondary evolves , but typically their bhs are small , as a consequence of the mass lost during ce . in both cases ,
the binary does not evolve into a rlo hmxb with a msbh . on the other hand , l10 and the large majority of studies that take into account metallicity
do not include the effects of dynamics on the evolution of primordial binaries and on the formation of new binaries .
this is a strong limitation , as most stars ( @xmath31 per cent ) likely form in scs ( e.g. lada & lada 2003 ) , and most scs are collisional environments , that is sites where close encounters between stars and binaries ( three - body encounters ) are extremely frequent and have important consequences ( see , e.g. bonnell & kroupa 1998 , for a review dedicated to young scs ) .
binaries are important for close encounters , because they have a larger cross - section than single stars and because they have an energy reservoir ( their internal energy ) , which can be exchanged with single stars ( e.g. heggie 1975 ; heggie & hut 1993 ; davies 1995 ; colpi , mapelli & possenti 2003 ) .
close encounters with single stars statistically lead to the increase ( decrease ) of the binding energy of a hard ( soft ) binary , defined as a binary with binding energy higher ( lower ) than the average kinetic energy of a star in the sc ( heggie 1975 ) .
close encounters can even unbind binaries ( ionization ) .
dynamical exchanges are also possible , that is interactions where one of the former members of the binary is replaced by the single star ( for a description of the possible outcomes of a binary - single star interaction , see , e.g. sigurdsson & phinney 1993 ) .
finally , recoil velocities due to three - body encounters can cause the ejection of the star and/or of the binary from the parent cluster ( see section 4.1 of sigurdsson & phinney 1993 for a general definition of recoil velocity in three - body encounters ) .
recent @xmath0body simulations of dense young scs with msbh binaries , but without stellar - evolution recipes , showed that close encounters substantially affect the evolution of these binaries , inducing hardening , exchanges and even ejections from the parent cluster ( mapelli et al .
2011b ) .
blecha et al .
( 2006 , hereafter b06 ) is one of the few studies of massive bhs where a stellar evolution code ( although not accounting for different metallicities ) is combined with recipes for dynamics ( in part semi - analytical prescriptions and in part a @xmath0body integrator for three - body encounters ) .
b06 study the evolution of an intermediate - mass bh ( imbh ) born via runaway collapse ( portegies zwart & mcmillan 2002 ) at the centre of a 5@xmath39 m@xmath40 dense young sc .
the imbh mass in the simulations of b06 is in the range @xmath41 m@xmath2 , i.e. partially overlapping with our definition of msbhs .
b06 find that the dynamical effects are very important for the evolution of the imbh , as captures of stars and three - body encounters allow it to have a companion star for most of the sc lifetime .
the main differences between this paper and b06 are the following .
firstly , our treatment of dynamics is fully @xmath0body , rather than based on a hybrid code .
secondly , in our simulations msbhs form self - consistently , as a consequence of metallicity - dependent stellar evolution and dynamics , whereas b06 generate one imbh per cluster , following recipes from runaway - merger simulations .
thus , ` our ' msbhs are generally smaller than those by b06 , and do not form necessarily in the core .
thirdly , we included a metallicity - dependent treatment of stellar evolution and wind - mass losses .
at last , the total mass of each of the scs we simulate is a factor of @xmath42 smaller than in b06 simulations .
finally , monte carlo codes are suitable for the study of the largest collisional systems ( @xmath43 objects ) , such as globular clusters ( e.g. hnon 1971a , 1971b ; stodlkiewicz 1982 , 1986 ; giersz 1998 ; joshi et al .
2000 , 2001 ; fregeau et al . 2003 ; fregeau & rasio 2007 ) .
the number of objects in these systems makes prohibitive to run wide grids of @xmath0body models so far .
some monte carlo codes also include accurate recipes for metal - dependent stellar and binary evolution ( e.g. chatterjee et al .
2010 ; downing et al .
2010 , 2011 ; pattabiraman et al .
monte carlo codes were recently used to study the population of x - ray binaries in globular clusters ( e.g. ivanova et al .
2006 , 2008 ) , the evolution of possible sources of gravitational waves ( e.g. downing et al .
2010 , 2011 ) , and the formation of imbhs by runaway collapse ( e.g. grkan , freitag & rasio 2004 ) .
these studies do not include recipes for the formation of msbhs in the @xmath32 m@xmath40 range .
the simulations were done using the starlabstarlab/ ] public software environment ( see portegies zwart et al .
2001 ) , which allows to integrate the dynamical evolution of a sc , resolving binaries and three - body encounters .
starlab includes the seba code for stellar and binary evolution ( portegies zwart & verbunt 1996 ; nelemans et al .
2001 ) . in its original version ,
seba accounts only for a solar - metallicity environment .
we modified seba by including various effects of metallicity , as follows .
we added the metallicity dependence of stellar radius , temperature and luminosity , using the polynomial fitting formulas by hurley , pols & tout ( 2000 ) .
we changed the recipes for mass loss by winds for ms stars , by using the metal - dependent fitting formulas given by vink et al .
( 2001 ; see also b10 ) .
we added a very approximate treatment for lbv stars and for wr mass losses by stellar winds , according to the recipes by b10 .
in particular , we assume that a post - ms star becomes a lbv when its luminosity @xmath44 and radius @xmath45 satisfy the requirement that @xmath46 and @xmath47 ( humphreys & davidson 1994 ) .
the mass - loss rate by stellar winds for a lbv is then calculated as @xmath48 m@xmath2 yr@xmath34 , where @xmath49 ( chosen by b10 because it allows to reproduce the most massive known stellar bhs , see the introduction and zel et al .
2010 ) .
naked helium giants coming from stars with zero age ms ( zams ) mass @xmath50 m@xmath2 ( e.g. , van der hucht 1991 and references therein ) are labelled as wr stars in the new version of the code and undergo a mass - loss rate by stellar winds defined by @xmath51 m@xmath2 yr@xmath34 , where @xmath52 .
this formula was first used by b10 , and is a combination of the hamann & koesterke ( 1998 ) wind rate estimate ( taking into account wr wind clumping ) and vink & de koter ( 2005 ) wind @xmath53-dependence for wrs .
we assume that all stars with final mass @xmath54 m@xmath2 collapse quietly to a bh , without sn explosion ( fryer 1999 ; fryer & kalogera 2001 ) , and that all the bhs born in this way do not receive any natal kicks ( see fryer et al .
2012 , for a discussion of this assumption ) .
if @xmath54 m@xmath2 , the mass of the bh is equal to @xmath55 , where @xmath56 is the final mass of the carbon oxygen ( co ) content of the progenitor , while @xmath57 and @xmath58 are the residual mass of helium ( he ) and of hydrogen ( h ) , respectively .
@xmath59 is the fraction of he and h mass which collapses to the bh in the failed sn scenario .
the value of @xmath59 is uncertain , and can range between 0 and 1 ( depending whether a faint sn or completely no sn occurs ) .
we assume @xmath60 to match the maximum values of @xmath61 at low @xmath53 derived by b10 . if @xmath62 m@xmath2 , the sn takes place and we adopt the standard routine of starlab ( portegies zwart & verbunt 1996 ) , including a sn explosion and partial fallback of h , he and co , depending on the ratio between energy released by the sn and binding energy of each shell .
[ fig : fig1 ] shows the mass of the bh as a function of the zams mass for four different metallicities ( from 0.01 to 1 z@xmath2 , where we adopt z@xmath22 ) .
the values of @xmath61 shown in fig . [ fig : fig1 ] have been obtained from our code when binary evolution is switched off .
the main feature is the increase of the maximum allowed @xmath61 for decreasing metallicity .
it is interesting to note the abrupt step in @xmath61 for @xmath63 and @xmath64 m@xmath2 , in the case of @xmath3 z@xmath2 and @xmath65 z@xmath2 , respectively .
the step is produced by the minimum value of @xmath66 for which @xmath54 m@xmath2 and the direct collapse takes place , at a given metallicity .
the main differences between our treatment of the collapse and that by b10 are that ( i ) b10 do not use the total final mass @xmath67 to discriminate between direct collapse and ` standard ' sn , but the mass of the co core , assuming that all stars with final co mass @xmath68 m@xmath2 collapse directly to a bh ; ( ii ) b10 adopt slightly different recipes for fallback ( see portegies zwart & verbunt 1996 and belczynski et al .
2008 for details ) . because of these differences , in our simulations fallback is less efficient for zams masses in the @xmath69 m@xmath40 range .
furthermore , the maximum mass of bhs for @xmath70 is higher ( by a factor of 1.4 ) in our simulations with respect to b10 results .
our choice of using @xmath67 to discriminate between sn and direct collapse produces the abrupt step visible in fig .
[ fig : fig1 ] for low @xmath53 , which is much less pronounced in b10 .
our results are consistent with those of b10 in light of the uncertainty in the bh mass spectrum . in general , our stellar evolution recipes are less sophisticated than those adopted by b10 , but are the most accurate that can be presently implemented within the framework of a complete @xmath0body calculation , maintaining an acceptable computational time .
one of the simplifying approximations in our model ( as well as in b10 ) is that stars do not rotate .
rotation may have important effects both on the mass loss of the progenitor star , inducing rotational mixing ( e.g. , maeder & meynet 2010 , 2012 , and references therein ) , and on the formation of the bh , producing asymmetries in the collapse ( e.g. fryer & heger 2000 ; akiyama et al .
2003 ; fryer & warren 2004 ; ardeljan , bisnovatyi - kogan & moiseenko 2005 ; wheeler , akiyama & williams 2005 ; thompson , quataert & burrows 2005 ; woosley & bloom 2006 ) . on the other hand ,
how the effects of rotation affect the mass spectrum of bhs is debated and highly uncertain ( e.g. dessart , oconnor , & ott 2012 ; maeder & meynet 2012 ) . including the effects of rotation in our model
requires a significant revision of the adopted recipes for stellar evolution and wind mass losses , which goes beyond the aims of the current paper .
we will investigate the effects of rotation in a forthcoming paper .
for the evolution of binaries we maintain the recipes already present in the original version of seba ( portegies zwart & verbunt 1996 ) .
in particular , seba distinguishes among the evolution of detached binaries ( where both stars are smaller than their roche lobes ) , that of semi - detached binaries ( one of the components fills its roche lobe ) and that of contact binaries ( both stars fill their roche lobe ) . in the case of detached binaries ,
mass loss and accretion can occur via stellar winds ( according to the formulation by livio & warner 1984 ) . in the case of semi - detached binaries , either stable or unstable
rlo takes place , according to the formalism described in appendix c of portegies zwart & verbunt ( 1996 ) .
if the accretor is a bh , the maximum mass - accretion rate of the bh is constrained by the requirement that the luminosity does not exceed the eddington limit .
a further critical assumption about binary evolution is the ce efficiency ( we define as ce efficiency the product of the two degenerate parameters @xmath71 and @xmath72 , see equation 3 of podsiadlowski et al .
2003 for a standard definition ) .
for the runs presented in this paper , we adopt @xmath73 , which is a rather standard value and tends to favour the formation of bh binaries ( e.g. podsiadlowski et al .
test runs with different values of @xmath74 show that the choice of this parameter does not significantly affect the results for values @xmath75 , in agreement with podsiadlowski et al .
( 2003 ) and with l10 .
ll parameter & values + @xmath76 & 5 + @xmath77 & 5500 + @xmath78 [ pc ] & 0.4 + @xmath79 & 1.03 + imf & kroupa ( 2001 ) + @xmath80 [ m@xmath2 ] & 0.1 + @xmath82 [ m@xmath2 ] & 150 + @xmath83 & 0.1 + @xmath84 $ ] & 0.01 , 0.1 , 1.0 + in this paper , we focus on moderately dense scs , adopting a spherical king profile with central adimensional potential @xmath85 ( king 1966 ) .
each simulated sc is initially composed of @xmath86 stars , corresponding to a total mass @xmath87 m@xmath2 per sc .
the resulting core density , at the beginning of the simulation , is @xmath88 m@xmath2 pc@xmath89 .
the main parameters adopted for the initial conditions are reported in table 1 .
for the runs presented in this paper , we fix the primordial binary fraction to @xmath90 .
the single stars and the primary stars of each binary are generated according to a kroupa initial mass function ( imf , kroupa 2001 ) , with minimum and maximum mass equal to 0.1 and 150 m@xmath2 , respectively .
the masses of the secondaries ( @xmath91 ) are generated according to a uniform distribution between @xmath92 and @xmath93 ( where @xmath93 is the mass of the primary ) .
the initial semi - major axis @xmath94 of a binary is chosen from a distribution @xmath95 ( sigurdsson & phinney 1995 ; portegies zwart & verbunt 1996 ) , consistent with the observations of binary stars in the solar neighbourhood ( e.g. kraicheva et al .
1978 ; duquennoy & mayor 1991 ) .
we generate @xmath94 between r@xmath2 and @xmath96r@xmath2 , but discarding systems where the distance between the two stars at the pericentre is smaller than the sum of their radii ( portegies zwart , mcmillan & makino 2007 ) .
the maximum value of @xmath94 was chosen arbitrarily , but motivated by the need to include also a significant fraction of soft binaries erg , where @xmath97 is the gravitational constant .
for @xmath98 m@xmath2 this corresponds to @xmath99r@xmath40 ] .
the initial eccentricity @xmath101 of a binary is chosen from a thermal distribution @xmath102 , in the @xmath103 range ( heggie 1975 ) .
the simulated scs have half - mass relaxation time @xmath104 , where @xmath105 is the initial half - mass radius of the sc ( in our simulations @xmath106 pc ) .
thus , the core collapse time ( portegies zwart & mcmillan 2002 ) is @xmath107 .
we integrate the evolution of these scs for the first 100 myr , therefore studying a phase of the life of the cluster in which dynamical interactions are particularly intense .
we make three sets of runs corresponding to three different metallicities : 0.01 z@xmath2 , 0.1 z@xmath2 and 1 z@xmath2 . for each of these metallicities we simulate 100 different clusters , for a total of 300 scs .
the properties of the simulated scs ( total mass , number of stars , core density , core and half - mass radius ) are consistent with the properties of observed young intermediate - mass scs ( see e.g. the recent review by portegies , mcmillan & gieles 2010 ; see also hillenbrand & hartmann 1998 ; dias et al . 2002 ; portegies zwart 2004 ; pfalzner 2009 ; kuhn et al .
2012 ) .
finally , our simulations do not include recipes for the tidal field of the host galaxy .
accounting for the tidal field may increase the fraction of mass lost , and even transform the scs into unbound associations ( e.g. gieles & portegies zwart 2011 ) .
the effect of tidal fields will be added and discussed in forthcoming papers .
the simulations were run on the graphics processing unit ( gpu ) cluster ibm - plx at cineca .
the processors available on plx are six - cores intel westmere 2.40 ghz ( two per node ) , while the gpus are nvidia tesla m2070 and m2070q ( two per node ) .
each single job ran over one processor and two gpus and required 50 cpu hours on average .
starlab runs on gpus through the sapporo library ( gaburov , harfst & portegies zwart 2009 ) .
[ fig : fig2 ] shows the effects of binary evolution of the progenitor star on the final mass of the bh . in particular , the points in fig .
[ fig : fig2 ] show all the simulated bhs ( originated from stars in binaries ) whose mass differs by more than @xmath108 per cent from the bh mass calculated if the progenitor was a single star evolving in isolation . in the following
, we will simply refer to them as bhs with @xmath109 .
all the bhs with @xmath109 form from progenitors that were in a primordial binary and underwent a mass transfer ( mt ) phase before collapsing to a bh .
the effects of mt before the formation of the first bh can be dramatic , especially when the two stars undergo a ce phase before the first sn . depending on the binding energy of the envelope , a ce phase can end either with the ejection of the envelope or with the merger of the two stars ( see portegies zwart & verbunt 1996 for more details ) . if the envelope is ejected , the final bh mass will likely be smaller than expected for the evolution of an unperturbed progenitor ; whereas if the two stars merge , the final bh mass will be higher than expected for a single star . for simplicity , the code assumes that , if two stars merge during a ce phase , no mass is lost during the merger and that the merger remnant evolves as a ms star ( the merger product is a blue straggler star , if its mass is higher than the turn - off mass ) . for example , a ce phase followed by envelope ejection is responsible for the two bhs formed by stars with @xmath110 m@xmath2 and @xmath3 z@xmath2 , which have @xmath111 m@xmath2 , i.e. a factor of @xmath112 less than their analogues born from single stars or wider binaries .
interestingly , for @xmath113 five bhs form with mass @xmath114 m@xmath2 ( one of them with mass close to 40 m@xmath2 ) .
these five bhs all form from the merger of a primordial binary before the sn explosion of the primary star depends strongly on the treatment of the star ( which is a blue straggler star ) formed from the merger , and in particular on the mass lost by this star before the collapse .
this issue is delicate and deserves further study . ]
( see e.g. soria 2006 ) .
the merger is a consequence of the mt phase triggered by the evolution of the primary .
this can be a viable path to form msbhs even at solar metallicity ( see the recent paper by soria et al .
2012 for the case of an ulx powered by a bh with estimated mass @xmath115 m@xmath2 , in a @xmath116 environment ) .
alternatively , msbhs can form at @xmath117 even through the merger of two low - mass bhs ( e.g. belczynski et al .
2004b ) .
table 2 allows to understand the statistical importance of binary evolution on the final mass of bhs .
we define @xmath118 as the fraction of bhs with @xmath109 with respect to the total number of simulated bhs . from table 2 ,
@xmath119 for all the metallicities .
since the primordial binary fraction is @xmath90 , approximately all bhs that formed in primordial binaries have @xmath109 .
this depends on the chosen initial distribution of the semi - major axes and on the adopted recipes for ce , but also on the effects of dynamical encounters , as we will discuss in the next section .
this result indicates that the initial binary fraction is an essential ingredient to shape the mass distribution of bhs . the resulting mass distribution of bhs for the three considered metallicities is shown in fig .
[ fig : fig3 ] .
the low - mass tails of the three distributions do not differ significantly , showing a peak at @xmath120 m@xmath2 .
this is in fair agreement with the observational limits for mw bhs discussed by zel et al .
( 2010 ) , especially considering the uncertainties on fallback models .
the most significant difference among different metallicities appears at the high - mass tail : the most massive bhs at @xmath121 z@xmath2 have a cut - off at @xmath122 m@xmath2 , whereas a significant fraction of bhs form with @xmath123 m@xmath2 at @xmath65 z@xmath2 and with @xmath124 m@xmath2 at @xmath3 z@xmath2 .
llllllllllllll @xmath53 [ z@xmath125 & @xmath126 & @xmath127 & @xmath128 & @xmath129 & @xmath130 [ myr ] & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath118 & @xmath136 & @xmath137 + 0.01 & 9.28 & 0.28 & 0.20 & 1.2 & 37.5 & 0.069 & 0.036 & 0.054 & 0.017 & 2.7 & 0.11 & 0.13 & 0.24 + 0.1 & 8.80 & 0.33 & 0.25 & 1.3 & 31.5 & 0.064 & 0.039 & 0.043 & 0.018 & 2.5 & 0.09 & 0.13 & 0.16 + 1 & 8.50 & 0.31 & 0.25 & 1.4 & 30.0 & 0.042 & 0.020 & 0.036 & 0.019 & 1.7 & 0.11 & 0.006 & 0.0 + lllllllllll @xmath53 [ z@xmath125 & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 [ myr ] & @xmath143 & @xmath144 & @xmath145 & @xmath146 & @xmath147 + 0.01 & 1.18 & 0.86 & 0.75 & 1.5 & 52.2 & 0.203 & 0.195 & 0.102 & 0.102 & 4.4 + 0.1 & 1.12 & 0.87 & 0.75 & 1.8 & 44.6 & 0.134 & 0.125 & 0.054 & 0.054 & 4.5 + 1 & 0.05 & 0.80 & 0.00 & 2.5 & 41.1 & 0.000 & 0.000 & 0.000 & 0.000 & 0.0 + table 2 provides a striking evidence of the importance of dynamics on the evolution of bhs in binaries , although the statistics is still quite low .
the average number of bhs per cluster ( @xmath148 ) is consistent with the expectations for a kroupa imf and does not appreciably depend on the metallicity .
the fraction of bhs that became members of a binary at least once in the simulated time interval is @xmath149 , regardless of the metallicity .
this fraction is higher ( by a factor of @xmath1502 ) than the fraction we expect if all the bh binaries come from primordial binaries .
furthermore , the fraction of bhs that formed from single stars and then became members of a binary as a consequence of a dynamical exchange is very high ( @xmath151 ) . in particular , @xmath128 is very similar to @xmath127 , indicating that most of the bhs in binaries formed from single stars .
in addition , the average number of exchanges per bh binary ( @xmath129 ) is more than one during the simulated time interval .
therefore , dynamical exchanges dominate the life of binary bhs in the simulated young scs .
the average lifetime of bh binaries is @xmath152 myr in table 2 . in calculating @xmath130
, we assume that a binary survives even if one of its members exchanges , and dies only when it is completely ionized or when the simulation is stopped .
it is worth noting that bh@xmath153bh and bh@xmath153neutron star ( ns ) binaries ( i.e. binaries where both the primary and the secondary member are compact objects ) live longer than other bh binaries ( on average ) , because they are not perturbed by the stellar evolution of the secondary and are sufficiently massive to avoid ionization . in particular , the average lifetime of bh@xmath153bh binaries is 46 , 37 and 32 myr at @xmath3 , 0.1 and 1 z@xmath2 , respectively .
the average lifetime of bh@xmath153ns binaries is similar : 49 , 38 and 65 myr at @xmath3 , 0.1 and 1 z@xmath2 , respectively myr ) of bh@xmath153ns binaries at @xmath121 z@xmath2 is explained by statistical fluctuations : only four bh@xmath153ns binaries form at @xmath121 z@xmath2 in our simulations . ] .
we notice that the lifetime of bh@xmath153bh and bh@xmath153ns binaries is significantly longer at very low metallicity ( @xmath3 z@xmath2 ) . at this metallicity ,
the most massive msbhs ( @xmath154 m@xmath40 ) tend to produce very massive , hard and thus long - lived binaries .
we will focus on binaries composed of two compact objects in a dedicated paper .
the fraction of bhs that undergo wind accretion ( @xmath131 ) indicates that there is about one wind - accreting system every two young clusters in 100 myr .
interestingly , about half of these systems are consequences of dynamical exchanges ( see @xmath132 in table 2 ) .
this is true for all the considered metallicities ( for @xmath121 z@xmath2 the fraction of wind - accreting systems formed by exchanges is slightly lower than for lower metallicities , but this might be a fluctuation due to low statistics ) .
the fraction of bhs in rlo ( @xmath133 ) is lower than @xmath131 .
a large fraction of rlo systems are originated from a dynamical exchange ( in table 2 , @xmath155 ) .
most of the donor stars in rlo systems are post - ms stars ( @xmath31 per cent post - ms stars , versus @xmath156 per cent ms stars ) .
a non - negligible fraction of such post - ms stars in rlo systems are lbv and wr stars ( @xmath157 per cent of the total donor stars in rlo systems ) .
these results are in agreement with b06 .
we note that a non - negligible fraction of ms companions are blue straggler stars , i.e. stars rejuvenated by stellar mergers or by mt ( e.g. mapelli et al .
2004 , 2006 ) : they behave as ms stars , although they have mass higher than the turn - off mass . in table 2
, @xmath135 is defined as the average number of strong resonant interactions per binary ( i.e. interactions that lead to the formation of an unstable triple system and that change significantly the orbital period of the binary ) , calculated only for those binaries that will undergo rlo .
these interactions are mostly three - body encounters and , in a few cases , four - body encounters ( i.e. binary - binary encounters ) .
the number of strong interactions is quite high ( @xmath158 ) , confirming the importance of three - body encounters .
finally , the last two columns of table 2 show the number of msbhs normalized to the total number of bhs ( @xmath136 ) , and the number of rlo msbhs normalized to the total number of rlo bhs ( @xmath137 ) , respectively . @xmath136 and @xmath137 are similar , suggesting that the incidence of rlo systems among msbhs is comparable to the incidence of rlo systems among low - mass bhs .
we stress that the main results presented in this section , and especially the statistics of accreting systems , strongly depend on the assumed fraction of primordial binaries .
in fact , primordial binaries have at least two important effects on accreting systems .
firstly , a number of primordial binary systems are expected to be sufficiently tight to start rlo at early times , as a consequence of stellar evolution .
if there are no primordial binaries , the first rlo systems will appear later in the evolution of the sc , as an effect of dynamical interactions .
secondly , primordial binaries ( and especially hard primordial binaries ) represent an initial reservoir of binding energy .
this means that , if a bh becomes member of a hard primordial binary after an exchange , this binary might be sufficiently tight to start rlo immediately after the first exchange . on the contrary ,
non - primordial binaries that form from the encounter of three single stars are initially quite soft , and even the hardest among them become sufficiently tight to start rlo only after a lot of dynamical encounters ( e.g. hut et al .
these arguments are discussed quantitatively ( through a supplementary set of runs ) in appendix a. we also note that the results presented in this section strongly depend on the stellar evolution recipes and on the simplified model of failed sn adopted for our simulations . in appendix b , we highlight the differences with respect to a scenario where the formation of msbhs is strongly suppressed . in the forthcoming investigations , we will consider in more detail different models of stellar evolution . table 3 reports the same quantities as shown in the first 11 columns of table 2 , but calculated only for the msbhs ( rather than for all the bhs ) .
table 3 shows that dynamical interactions are crucial for msbhs , even more important than for low - mass stellar bhs .
first , the fraction of msbhs that became members of a binary at least once in the simulated time interval ( with respect to the total number of simulated msbhs ) is @xmath159 , i.e. a factor of @xmath160 higher than in the overall bh sample . at low metallicity ( @xmath3 and 0.1 z@xmath2 ) ,
most of the msbhs in binaries formed from single stars , as the fraction of msbhs that formed from single stars and then became members of a binary as a consequence of a dynamical exchange is @xmath161 . instead , for @xmath121 z@xmath2
, no msbhs in binaries form from single stars ( @xmath162 ) .
this is naturally explained by the fact that the only channel to form msbhs at @xmath121 z@xmath2 is the merger between two massive stars in a primordial binary ( see section 4.1 ) .
furthermore , the average number of exchanges per msbh binary is @xmath163 , higher than for the overall bh sample ( @xmath164 ) .
the lifetime of msbh binaries is slightly longer than that of the entire bh binary sample ( @xmath165 myr ) .
this is in fair agreement with b06 , finding that a bh with @xmath166 m@xmath2 , born from runaway collapse , spends about 25 to 60 per cent of the simulated time interval with a companion .
columns @xmath167 of table 3 provide a key result to understand the formation of x - ray binaries powered by msbhs .
the comparison between columns @xmath168 and @xmath169 indicates that the large majority of wind - accreting msbh binaries at low @xmath53 ( @xmath3 and 0.1 z@xmath2 ) were originated by a dynamical exchange ( @xmath170 ) .
the comparison between columns @xmath171 and @xmath172 indicates that all rlo msbh binaries at low metallicity ( @xmath3 and 0.1 z@xmath2 ) were originated by a dynamical exchange ( @xmath173 ) .
furthermore , all but one of the rlo msbh binaries in our simulations host msbhs that formed from single stars ( this is not shown in table 3 , but we checked it from our simulations ) .
the only msbh ( among rlo msbh binaries ) that did not form from a single star was originated from the merger of a primordial binary .
thus , even this msbh formed as a single object and acquired a stellar companion after dynamical exchange . the fraction of wind - accreting msbh binaries ( with respect to the total number of simulated msbhs ) is higher than the corresponding fraction for all bh binaries .
the fraction of rlo msbh binaries is similar to the corresponding fraction for all bh binaries , but the low statistics makes fluctuations very important ( we find 18 rlo msbh binaries over 300 simulations ) .
therefore , to study rlo msbh binaries in detail , it will be essential to run a much larger grid of simulations ( mapelli et al .
, in preparation ) .
we find no wind - accretion msbh binaries and no rlo msbh binaries for @xmath121 z@xmath2 , but this is likely an effect of the low statistics , as only five msbhs form in our simulations at @xmath121 z@xmath2 .
the results reported in table 3 are in fair agreement with the findings by l10 , based on population synthesis calculations .
in particular , l10 find that no or very few rlo systems powered by msbhs form from primordial binaries .
this is consistent with our finding that no rlo msbh binaries come from primordial binaries .
on the other hand , the code used by l10 does not account for dynamical interactions .
all rlo msbh binaries found in our simulations were originated from dynamical interactions , in the sense that the msbh became a member of the binary as a consequence of a dynamical exchange .
[ fig : fig4 ] , [ fig : fig5 ] and [ fig : fig6 ] show some important characteristics of mt bhs . in these figures ,
the filled circles are wind - accreting systems , while the crosses are rlo systems .
[ fig : fig4 ] shows the mass of the bh versus the mass of the companion star ( @xmath174 ) . at all considered metallicities , @xmath174 ranges from relatively low values ( @xmath175 m@xmath2 , mostly asymptotic giant branch stars ) to very high values ( @xmath176 m@xmath2 or more ) for both wind - accreting systems and rlo systems .
the masses of accreting bhs span all the possible bh masses for a given metallicity ( see fig . [
fig : fig3 ] , for comparison ) , even the higher masses . not only wind - accreting systems but also rlo systems can host msbhs ( @xmath177 m@xmath2 ) .
this is an important difference with respect to l10 results , and it is mainly the consequence of dynamical interactions ( which were not accounted by l10 , see the discussion in section 4.3 ) .
we note that for @xmath65 z@xmath40 there is even one rlo system powered by a @xmath178 m@xmath2 msbh .
this msbh is more massive than the maximum mass ( @xmath179 m@xmath2 ) that can be reached at this metallicity through the adopted models of stellar evolution .
in fact , such msbh comes from the merger between a @xmath180 m@xmath2 bh and a @xmath181 m@xmath2 ms companion .
[ fig : fig5 ] shows the mass of the bh versus the orbital period @xmath182 .
for all the considered metallicities , wind - accreting systems can form with periods as long as a few @xmath183 days , in agreement with previous studies ( e.g. l10 ) .
rlo systems have periods spanning from less than one day ( @xmath184 hours for one system at @xmath3 z@xmath40 ) up to @xmath185 yr .
we stress that the periods of rlo systems shown in fig .
[ fig : fig5 ] are the values of the period at the first roche lobe approach , as the code does not trace with accuracy the late stages of rlo and the time interval between snapshots is not sufficiently short to follow the evolution of all the systems .
for example , if a system evolves into tidal instability and the companion is a ms star , the code assumes that the system is undergoing merger and removes the binary from calculation ( portegies zwart & verbunt 1996 ) .
therefore , the plotted periods of rlo systems must be considered upper limits .
we note that the companion star is an evolved star , with a very large radius ( of the order of @xmath186 r@xmath40 ) , in most of the wide rlo systems ( @xmath187 yr ) .
[ fig : fig6 ] shows the mass of the bh versus the time elapsed since the beginning of the simulation .
this plot gives information about the duty cycle ( @xmath188 ) , defined as the lapse of time for which a binary is wind accreting and/or in rlo , divided by the total elapsed time in the simulation .
[ fig : fig6 ] indicates that most systems are in rlo for less than one snapshot ( corresponding to @xmath189 myr ) .
this is in fair agreement with the results by b06 , which find that rlo systems powered by imbhs are on average short - lived ( @xmath190 per cent of the simulated time interval ) .
wind accretion can last for a longer time ( a few myr ) .
relatively low - mass bhs ( @xmath191 m@xmath2 ) tend to start the rlo phase short after their formation .
these systems start rlo as a consequence of stellar evolution of the secondary and/or as an effect of the first sn kick ( see l10 ) . instead
, more massive bhs start rlo at later times ( @xmath192 myr after the beginning of the simulation ) .
this means that the most massive bhs were single or in relatively wide binaries and can start rlo only as a consequence of the hardening of the binary by three - body encounters and/or of a dynamical exchange .
therefore , dynamical interactions are essential to allow msbhs to power rlo x - ray binaries .
this difference is also important to understand which x - ray sources are associated with low - mass bhs and which x - ray sources might be powered by msbhs .
in fact , from fig . [
fig : fig6 ] we expect that the donor stars in x - ray binaries powered by msbhs are on average older than those in x - ray binaries powered by low - mass bhs
. in particular , most msbhs enter the rlo phase with companions that are @xmath193 myr old .
interestingly , most of the ulxs for which information about the stellar environment is available are associated with @xmath194 myr old stellar populations ( see e.g. soria et al . 2005 ; liu et al .
2007 ; gris et al . 2008 ; swartz et al . 2009 ; gris et al . 2011 ; voss et al . 2011 ) .
further hints about the importance of dynamics come from the variation of the orbital period .
[ fig : fig7 ] shows the final period @xmath195 of a rlo binary ( defined as the period at the beginning of rlo ) versus the initial period @xmath196 of the same binary ( defined as the period at the beginning of the simulation ) . for consistency , in fig .
[ fig : fig7 ] we show only primordial binaries that do not undergo dynamical exchanges before starting rlo . for most systems @xmath197 , as it was reasonable to expect .
the shrinking is due to the joint effect of stellar evolution ( e.g. a ce phase before the formation of the first bh forces the semi - major axis to shrink ) and of three - body encounters ( especially when the period changes by a factor of @xmath185 or more ) .
the large majority of systems have @xmath198 yr . systems with an initial period below this threshold merge before the formation of the first bh .
there are only two systems with @xmath199 yr , both for @xmath3 z@xmath2 .
these systems undergo at least one strong three - body encounter before the formation of the first bh .
the three - body encounters widen the semi - major axis of these binaries , allowing them to avoid merger and to survive till the formation of the first bh .
lllllll @xmath53 [ z@xmath125 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 + 0.01 & 0.50 & 0.23 & 0.03 & 0.10 & 0.006 + 0.1 & 0.41 & 0.19 & 0.04 & 0.09 & 0.009 + 1 & 0.46 & 0.24 & 0.001 & 0.08 & 0.007 + the possibility that a bh is ejected from the parent sc is relevant for various astrophysical issues .
for example , bright hmxbs and ulxs are often close to young scs and star forming regions , but displaced by @xmath205 pc with respect to their centre ( e.g. zezas et al .
2002 ; kaaret et al .
2004 ; berghea 2009 ; swartz et al .
2009 ; swartz 2010 ; rangelov , prestwich & chandar 2011 ; voss et al . 2011 ; poutanen et al .
this fact has been generally interpreted as the indication that bright x - ray sources are powered by runaway binaries , that is , by binaries that were ejected from the parent cluster because of a natal kick ( e.g. sepinsky , kalogera & belczynski 2005 ; zuo & li 2010 ) or because of a close encounter ( e.g. kaaret et al .
2004 ; berghea 2009 ; mapelli et al .
2011b ) . in our simulations
, bhs can be ejected both through sn explosion ( natal kick ) and through three - body encounters .
table 4 shows that @xmath206 per cent of simulated bhs are ejected from the sc , almost independently of the metallicity ( we classify a bh as ejected when its distance from the centre of mass of the sc is @xmath207 ) .
about half of the ejections are consequences of the natal kick , while the remaining half is due to three body encounters .
sn explosions can also unbind a binary system .
we estimate that @xmath208 primordial binaries per sc are ionized by a sn explosion leading to the formation of a bh ( we do not include nss in this estimate ) . at @xmath3 and 0.1 z@xmath2 , the fraction of ejected msbhs ( with respect to the total number of simulated bhs ) is @xmath209 .
since msbhs are @xmath210 per cent of all the bhs at these metallicities , this means that about @xmath211 per cent of all the simulated msbhs are ejected as a consequence of three - body encounters ( we recall that in our simulations we assume that msbhs receive no natal kick ) .
this percentage is moderately lower , but still in agreement with mapelli et al .
( 2011b ) , who find that @xmath179 per cent of all msbhs are ejected as a consequence of three - body encounters .
the difference between these two estimates can be explained by the fact that mapelli et al .
( 2011b ) generate the msbhs already in the initial conditions ( rather than letting them form later , through stellar evolution ) , and assume that all msbhs are members of primordial binaries .
thus , the available time for msbhs to undergo thee - body encounters is longer in the simulations by mapelli et al .
( 2011b ) .
this assumption increases the number of interactions involving msbh binaries , especially during the first stage of core collapse ( which occurs as early as @xmath212 myr in our simulated scs ) .
furthermore , the simulations in mapelli et al .
( 2011b ) do not include stellar evolution .
table 4 also shows that about 10 per cent of all the simulated bhs are ejected together with their companion star , regardless of the metallicity .
about one tenth of these ejected binaries enter rlo after leaving the cluster ( @xmath213 ) . since the total fraction of rlo systems is @xmath214 ( see table 2 ) , this means that about 20 per cent of all rlo systems enter the rlo phase after being ejected from the sc ( and in most of the cases the rlo is triggered by a dynamical interaction ) .
only one of these ejected rlo systems is powered by a msbh .
thus , most of the simulated rlo systems are inside the parent sc .
this result is apparently at odds with observations ( e.g. zezas et al .
2002 ; kaaret et al .
2004 ; berghea 2009 ; swartz et al . 2009 ; swartz 2010 ) , which indicate that a significant fraction of bright x - ray sources are offset from the parent cluster by more than 10 pc .
on the other hand , we stress that the simulations presented in this paper do not include the tidal field of the host galaxy : as our simulated scs have a relatively small initial mass , a large fraction of binaries can be stripped from the sc by external tidal forces .
finally , in the current paper we have assumed that the msbhs are born without natal kick .
this assumption depends crucially on the model of direct collapse and failed sn .
thus , the fraction of ejected msbhs reported in this paper must be regarded as a lower limit . in the previous sections
, we highlighted the differences between the behaviour of msbhs and that of low - mass stellar bhs , discussing the results of the @xmath0body simulations . in this section ,
we show how the results of the @xmath0body simulations can be intuitively understood in light of the interplay between stellar evolution and some of the basic properties of three - body encounters .
the three most relevant aspects from the physics of three - body encounters are the following .
\(i ) the probability of a single star ( or stellar remnant ) with mass @xmath215 to exchange into a binary ( @xmath216 ) is higher if @xmath217 or @xmath218 , where @xmath93 and @xmath91 are the masses of the two components of the binary ( e.g. hills & fullerton 1980 ; sigurdsson & phinney 1993 ) . in particular
, @xmath216 is very close to one , if the binary is hard ( so that it can not be easily ionized ) and if @xmath219 or @xmath220 ( hills & fullerton 1980 ) .
after the first @xmath10 myr of the cluster lifetime ( when the turn - off mass goes below @xmath221 m@xmath2 ) , single msbhs are among the most massive objects in the sc and are likely to replace the lowest - mass members of primordial binaries .
\(ii ) the cross section for three - body interactions is larger for more massive binaries ( see e.g. sigurdsson & phinney 1993 ; davies , benz & hills 1994 ; davies 2002 ; miller & hamilton 2002 ) . after the first @xmath10 myr of the cluster lifetime ,
msbh binaries become significantly more massive than the other binaries in the sc . thus , their rate of three - body encounters is higher than that of low - mass binaries .
\(iii ) hard binaries tend to harden as a consequence of three - body encounters , i.e. to reduce their semi - major axis @xmath94 ( e.g. heggie 1975 ; heggie & hut 1993 ; davies 1995 ; quinlan 1996 ; merritt 2001 ) .
the decrease of @xmath94 can start rlo into a bh binary .
a msbh binary in our simulated scs is hard if @xmath222 a.u.@xmath223 ( corresponding to a period @xmath224 yr@xmath223 ) .
therefore , most of the msbh binaries in our simulations are hard . combining stellar evolution with the above notions from the theory of three - body encounters , we can summarize the evolution of the simulated bhs as shown in fig .
[ fig : fig8 ] . in the following ,
we discuss the main differences between msbhs and low - mass bhs , relatively to the flow chart in fig .
[ fig : fig8 ] . case a : the bh forms from a single star .
if the bh is a msbh , then the exchange probability @xmath216 is very high already after @xmath185 myr since the sc formation , and the msbh becomes soon a binary member via dynamical exchange .
the resulting msbh binary undergoes efficiently three - body encounters and hardens rapidly .
if the bh has relatively low mass , its @xmath216 is low for most of the sc life : the low - mass bh can remain single for a much longer time , or even for the entire lifetime of the cluster .
furthermore , low - mass bhs are more easily ejected out of the sc .
case b : the bh forms from a primordial binary . depending on the initial semi - major axis of the binary and on the stellar evolution , the binary can merge or avoid merger before the formation of the bh . if the binary merges before the formation of the bh ( case b1 ) , then the bh forms as a single object and its evolution is the same as in the case a. if the binary does not merge before the formation of the bh ( case b2 ) , then there are at least three possibilities . case b2.1 : the sn kick of the primary unbinds the binary . in this case , the bh is substantially a single object and behaves as in case a. in our simulations , this case can occur only for low - mass bhs , as msbhs are assumed to form without natal kick .
case b2.2 : the binary remains bound and the stellar evolution of the secondary is faster than three - body encounters .
this occurs especially when the binary is very close and/or the mass of the secondary is similar to the mass of the primary . for binaries where the radius of the secondary star is close to the roche lobe at the time of formation of the bh ,
the binary enters a rlo phase because of the evolution of the secondary .
as already shown by l10 , msbhs can hardly form in such close binaries , because the first ce phase leads generally to the formation of small bhs . therefore , this case is more frequent for low - mass bhs .
case b2.3 : the binary remains bound and three - body encounters are more efficient than the stellar evolution of the secondary .
this occurs especially for relatively wide binaries ( where stellar evolution is not sufficient to drive rlo ) , and/or for low - mass secondary stars , which evolve much more slowly than the primary . in this case
, the evolution is very different depending on the mass of the bh and depending on whether the binary is hard or soft . in the following ,
we mention only three of the possible cases ( the most relevant for our simulations ) .
\(i ) if the binary is soft ( unlikely if a msbh is member of the binary ) , three - body encounters are expected to ionize it .
\(ii ) if the bh is a msbh and the binary is relatively hard , then three - body encounters harden the binary .
exchanges can occur , but are unlikely to remove the msbh from the binary , as it is more massive than most single stars .
the msbh binary survives and may undergo mt .
\(iii ) if the bh has a low mass ( lower than the companion mass ) and the binary is relatively hard , then three - body encounters harden the binary , but exchanges can remove the bh from the binary .
if it is expelled by an exchange , the bh becomes single again .
we stress that the flow chart in fig .
[ fig : fig8 ] is schematic and somehow simplistic , as it neglects some further effects that can take place ( e.g. the sn explosion of the companion star , the formation of a binary composed of two compact objects , the definitive ejection of the bh from the sc ) . despite this , fig .
[ fig : fig8 ] allows to understand why exchanges are so important to enhance the formation and evolution of msbh binaries ( see table 3 ) .
in fact , msbhs evolve predominantly from case a ( i.e. they form from single stars ) to case b2.3 , where exchanges and three - body encounters dominate the evolution of a hard and massive msbh binary .
it clarifies also why most low - mass bhs start rlo immediately after their formation ( see fig .
[ fig : fig6 ] ) , whereas msbhs enter rlo much later than their formation ( @xmath225 myr after the beginning of the simulation ) .
in fact , it is unlikely for msbhs to evolve through case b2.2 and to start rlo because of the stellar evolution of the secondary star in a primordial binary .
msbhs evolve mainly through case a and case b2.3 : they can efficiently acquire ( new ) companions because of dynamical exchanges , and their binaries harden because of three - body encounters . on the other hand ,
dynamical interactions occur on a longer timescale ( a few tens myr ) than massive star evolution . instead , low - mass bhs evolve mainly through case a , case b2.2 and case b2.3 . in case
a , low - mass bhs remain likely single objects , as the probability of an exchange into a binary is low .
case b2.2 can lead to mt on the timescale of massive star evolution .
case b2.3 can result in the ejection of the low - mass bh from the binary , as a consequence of dynamical exchanges .
therefore , most of the rlo systems powered by low - mass bhs evolve as described in case b2.2 , and only a few of them through case b2.3 . to describe the evolution of mass accretion and the emission properties of rlo binaries is beyond the scope of this paper . on the other hand , from the simulations described in the previous sections
, we can extract some basic hints about the evolution of rlo binaries .
the code we use limits the accretion rate so that the luminosity of the binary can not exceed the eddington limit for bhs .
we prefer not to change this assumption and not to introduce arbitrary rules for super - eddington accretion , as the dynamical code is not sufficiently accurate to distinguish between different accretion models . in a forthcoming paper , we will re - simulate the mt systems obtained from this study , without accounting for three - body encounters but with a more accurate recipe for the mt rate ( e.g. patruno & zampieri 2008 ) .
[ fig : fig9 ] shows the eddington luminosity ( @xmath226 ) for the simulated rlo systems .
there are important differences between the solar - metallicity simulations and the two sub - solar environments , although the statistics is quite low .
no rlo systems formed at @xmath227 have @xmath228 erg s@xmath34 ( corresponding to a bh mass @xmath229 m@xmath40 ) .
rlo systems formed at @xmath230 and @xmath231 reach @xmath232 erg s@xmath34 and @xmath233 erg s@xmath34 , respectively ( corresponding to bh mass @xmath234 and 58 m@xmath40 , respectively ) .
the mass spectrum of bhs born from the collapse of massive stars is very uncertain .
recent theoretical studies ( e.g. mapelli et al . 2009 ; zampieri & roberts 2009 ; b10 ) indicate that the mass of the bh depends on the metallicity of the progenitor star and that metal - poor massive stars can produce msbhs with a mass as high as @xmath235 m@xmath2 ( for @xmath231 ) .
observational results suggest that low - metallicity environments host more massive bhs than the mw ( e.g. prestwich et al .
2007 ) and highlight a possible anti - correlation between the population of bright x - ray sources ( especially ulxs ) and the metallicity of the host galaxy ( e.g. mapelli et al .
2010 , 2011a ) . however , it is still unclear whether theoretically predicted msbhs are efficient in powering x - ray sources .
studies based on population synthesis codes ( e.g. l10 ) , including metallicity effects but neglecting dynamical interactions , find that msbhs can hardly power bright x - ray sources , because of various effects connected with binary evolution . in this paper , we present preliminary simulations run with a @xmath0body plus stellar and binary evolution code , based on the public version of starlab .
these simulations include metallicity - dependent stellar evolution , binary evolution and an accurate treatment of three - body encounters .
we simulate dense young scs with three different metallicities : @xmath231 , @xmath230 and @xmath227 .
our simulations show that the dynamics is essential to map the population of bh binaries and that of x - ray binaries in young scs .
firstly , the rate of bh binaries formed through dynamical exchanges is very high for all the considered metallicities .
secondly , the hardening of the binary due to three - body encounters is an important mechanism to induce rlo in binaries hosting bhs .
the role of dynamics is particularly important for msbhs . in scs with @xmath3 and 0.1 z@xmath2 , about 75 per cent of simulated msbhs form from single stars and become members of binaries through dynamical exchanges in the first 100 myr of the sc life
this is a factor of @xmath6 more efficient than in the case of low - mass ( @xmath7 m@xmath2 ) stellar bhs .
we show that the vast majority of msbhs in rlo binaries originated from single stars and went through a dynamical exchange ( table 3 and section 4.3 ) .
this is consistent with the qualitative predictions from the theory of three - body encounters combined with stellar evolution ( section 4.6 ) .
in fact , the higher is the mass of the bh involved , the more important the effects of dynamics on x - ray binaries , as the rate of three - body encounters scales approximately with the total mass of the binary ( e.g. sigurdsson & phinney 1993 ) .
furthermore , the probability for a single bh to become a binary member through a dynamical exchange is higher if the mass of the single bh is higher than the mass of one of the two components of the binary ( e.g. hills & fullerton 1980 ) .
therefore , msbhs are even more influenced by dynamics than their low - mass analogues .
our results agree with the basic conclusion by l10 : it is very hard that msbhs power rlo x - ray binaries , if they form in primordial binaries that evolve unperturbed . on the other hand
, we find that the situation is completely different for msbhs that form in dynamically active environments ( such as dense young scs ) .
in these environments , msbhs efficiently power rlo x - ray binaries , as a consequence of dynamical exchanges and three - body encounters , which dramatically alter the orbital properties of primordial binaries .
our results are also in fair agreement with those by b06 , although there are some important differences , as b06 study imbhs born from the runaway collapse in denser scs , do not account for metallicity - dependent stellar evolution , and have a much simplified dynamical treatment .
both our paper and b06 find that bhs with mass @xmath236 m@xmath2 spend a large fraction of the simulated time with a companion star and that these massive bhs can power rlo binaries as a consequence of three - body encounters .
our preliminary results show that the study of msbhs is very promising , but there is a lot of work still to be done .
first , the sample of msbhs obtained from our simulations is statistically small ( 112 and 118 msbhs for the simulations with @xmath65 and @xmath237 z@xmath40 , respectively ) .
although our main results are physically well motivated and in agreement with previous studies about three - body encounters ( e.g. , hills & fullerton 1980 ; sigurdsson & phinney 1993 ) and population synthesis models ( e.g. l10 ) , a much larger sample is required to study in detail the properties of msbhs in rlo binaries . from the theoretical point of view
, different models of stellar evolution and especially of mass loss by stellar winds need to be investigated with the same approach . different environments ( not only dense young scs , but also globular clusters and open clusters ) need to be studied with larger statistics .
furthermore , in this paper we assume a primordial binary fraction @xmath90 .
higher binary fractions are not unrealistic ( e.g. delgado - donate et al . 2004 ; ivanova et al . 2005 ; sollima et al . 2010 ) and must be considered in future simulations .
the code described in this paper traces accurately the dynamics of close encounters , but adopts simplified recipes for stellar and binary evolution , as well as for mt .
the mt systems individuated by our simulations will be re - simulated with more accurate binary evolution codes , to trace how they evolve , which are their expected emission features and whether they can explain a fraction of the ulxs .
a fundamental question is where we can search for msbhs .
mw globular clusters can reach metallicities as low as @xmath231 .
thus , they may host msbhs , single or with low - mass companions
. metallicities of the order of @xmath238 are not infrequent in nearby irregular galaxies ( e.g. ic 10 ) , which are good candidates to further investigate the metallicity@xmath153bh mass connection .
we thank the anonymous referee for her / his critical reading of the manuscript , which improved it .
we made use of the public software package starlab ( version 4.4.4 ) and of the sapporo library ( gaburov , harfst & portegies zwart 2009 ) to run starlab on graphics processing units ( gpus ) .
we acknowledge all the developers of starlab , and especially its primary authors : piet hut , steve mcmillan , jun makino , and simon portegies zwart .
we thank the authors of sapporo , and in particular e. gaburov , s. harfst and s. portegies zwart .
we also thank paola marigo , alessia gualandris , antonella vallenari , rosanna sordo , laura greggio , mauro barbieri and monica colpi for stimulating discussions .
we acknowledge the cineca award n. hp10cxb7o8 and hp10c894x7 , 2011 for the availability of high performance computing resources and support .
mm , lz and er acknowledge financial support from inaf through grant prin-2011 - 1 .
lz acknowledges financial support from asi / inaf grant no . i/009/10/0 .
ab acknowledges financial support from miur through grant prin-2009 - 1 .
mm thanks the aspen center for physics , where part of this work was done .
abolmasov p. k. , swartz d. a. , fabrika s. , ghosh k. k. , sholukhova o. , tennant a. f. , 2007 , apj , 668 , 124 akiyama sh . , wheeler j. c. , meier d. l. , lichtenstadt i. , 2003 , apj , 584 , 954 ardeljan n. v. , bisnovatyi - kogan g. s. , moiseenko s. g. , 2005 , mnras , 359 , 333 belczynski k. , kalogera v. , zezas a. , fabbiano g. , 2004a , apj , 601 , l147 belczynski k. , sadowski a. , rasio f. a. , 2004b , apj , 611 , 1068 belczynski k. , kalogera v. , rasio f. a. , taam r. e. , zezas a. , bulik t. , maccarone th . j. , ivanova n. , 2008 , apjs , 174 , 223 belczynski k. , bulik t. , fryer c. l. , ruiter a. , valsecchi f. , vink j. s. , hurley j. r. , 2010 , apj , 714 , 1217 ( b10 ) belkus h. , van bever j. , vanbeveren d. 2007 , apj , 659 , 1576 berghea c. , 2009 , _ ultraluminous x - ray sources and their environment _ , phd thesis , http://students.cua.edu/79berghea/papers/dissertation.pdf bica e. , dutra c. m. , soares j. , barbuy b. , 2003 , a&a , 404 , 223 blecha l. , ivanova n. , kalogera v. , belczynski k. , fregeau j. , rasio f. , 2006 , apj , 642 , 427 ( b06 ) bonnell i. , kroupa p. , 1998
, to appear in the orion complex revisited , eds .
m.j.mccaughrean , a.burkert , asp conference series , arxiv : astro - ph/9802306 borissova j. , et al .
, 2011 , a&a , 532a , 131 carpano s. , pollock a. m. t. , prestwich a. , crowther p. , wilms j. , yungelson l. , ehle m. , 2007 , a&a , 466 , l17 charles p. a. , coe m. j. , 2006 , in : compact stellar x - ray sources . edited by walter lewin & michiel van der klis .
cambridge astrophysics series , no . 39 .
cambridge , uk : cambridge university press , 215 chatterjee s. , fregeau j. m. , umbreit s. , rasio f. a. , 2010 , apj , 719 , 915 colpi m. , mapelli m. , possenti a. , 2003 , apj , 599 , 1260 crowther p.
a. , carpano s. , hadfield l. j. , pollock a. m. t. , 2007 , a&a , 469 , l31 crowther p. a. , barnard r. , carpano s. , clark j. s. , dhillon v. s. , pollock a. m. t. , 2010 , mnras , 403 , l41 davies m. b. , benz w. , hills j. g. , 1994 , apj , 424 , 870 davies m. b. , 1995 , mnras , 276 , 887 davies m. b. , 2002 , in ` omega centauri , a unique window into astrophysics ' .
asp conference proceedings , vol . 265 .
edited by floor van leeuwen , joanne d. hughes , and giampaolo piotto .
san francisco , astronomical society of the pacific , p.215 delgado - donate e. j. , clarke c. j. , bate m. r. , hodgkin s. t. , 2004 , mnras , 351 , 617 dessart l. , oconnor e. , ott ch . , 2012 , apj , 754 , 76 dias w. s. , alessi b. s. , moitinho a. , lpine j. r. d. , 2002 , a&a , 389 , 871 downing j. m. b. , benacquista m. j. , giersz m. , spurzem r. , 2010 , mnras , 407 , 1946 downing j. m. b. , benacquista m. j. , giersz m. , spurzem r. , 2011 , mnras , 416 , 133 dray l. m. , 2006 , mnras , 370 , 2079 duquennoy a. , & mayor m. , 1991 , a&a , 248 , 485 fregeau j. m. , grkan m. a. , joshi k. j. , rasio f. a. , 2003 , apj , 593 , 772 fregeau j. m. , & rasio f. a. , 2007 , apj , 658 , 1047 fryer ch . l. , 1999 , apj , 522 , 413 fryer ch . l. , & heger a. , 2000 , apj , 541 , 1033 fryer ch . l. , kalogera v. , 2001 , apj , 554 , 548 fryer ch .
l. , & warren m. s. , 2004 , apj , 601 , 391 fryer ch .
l. , belczynski k. , wiktorowicz g. , dominik m. , kalogera v. , holz d. e. , 2012 , apj , 749 , 91 gaburov e. , harfst s. , portegies zwart s. , 2009 , new astronomy , 14 , 630 garnett d. r. , 1990 , apj , 363 , 142 gieles m. , portegies zwart s. f. , 2011 , mnras , 410 , l6 giersz m. , 1998 , mnras , 298 , 1239 goad m. r. , roberts t. p. , knigge c. , lira p. , 2002 ,
mnras , 335 , l67 gris f. , pakull m. w. , soria r. , motch c. , smith i. a. , ryder s. d. , bttcher m. , 2008 , a&a , 486 , 151 gris f. , kaaret p. , pakull m. w. , motch c. , 2011 , apj , 734 , 23 gris f. , kaaret p. , corbel s. , feng h. , cseh d. , tao l. , 2012 , apj , 745 , 123 grkan m. a. , freitag m. , rasio f. a. , 2004 , apj , 604 , 632 hamann w .- r .
, koesterke l. , 1998 , a&a , 335 , 1003 harlaftis e. t. , greiner j. , 2004 , a&a , 414l , 13 heger a. , fryer c.l . , woosley s.e . , langer n. , hartmann d.h . , 2003a , apj , 591 , 288 heger a. , woosley s.e . , fryer c.l .
, langer n. , 2003b , in from twilight to highlight : the physics of supernovae : proceedings of the eso / mpa / mpe workshop held at garching , germany , 29 - 31 juli 2002 , eso astrophysics symposia .
isbn 3 - 540 - 00483 - 1 . edited by w. hillebrandt and b. leibundgut .
springer - verlag , p. 3
heggie d. c. , 1975 , mnras , 173 , 729 heggie d. c. , hut p. , 1993 ,
apjs , 85 , 347 hnon m. h. , 1971a , ap&ss , 13 , 284 hnon m. h. , 1971b , ap&ss , 14 , 151 hillenbrand l. a. , hartmann l. w. , 1998 , apj , 492 , 540 hills j. g. , fullerton l. w. , 1980 , aj , 85 , 1281 humphreys r. m. , davidson k. , 1994 , astronomical society of the pacific , publications , 106 , 1025 hurley j. r. , pols o. r. , tout c. a. , 2000 , mnras , 315 , 543 hurley j. r. , tout c. a. , pols o. r. , 2002 , mnras , 329 , 897 hut p. , mcmillan s. , goodman j. , mateo m. , phinney e. s. , pryor c. , richer h. b. , verbunt f. , weinberg m. , 1992 , astronomical society of the pacific , publications , 104 , 981 ivanova n. , belczynski k. , fregeau j. m. , rasio f. , 2005 , mnras , 358 , 572 ivanova n. , heinke c. o. , rasio f. a. , taam r. e. , belczynski k. , fregeau j. , 2006 , mnras , 372 , 1043 ivanova n. , heinke c. o. , rasio f. a. , belczynski k. , fregeau j. , 2008 , mnras , 386 , 553 joshi k. j. , rasio f. a. , portegies zwart s. , 2000 , apj , 540 , 969 joshi k. j. , nave c. p. , rasio f. a. , 2001 , apj , 550 , 691 kaaret p. , alonso - herrero a. , gallagher j. s. , fabbiano g. , zezas a. , rieke m. j. , 2004 , mnras , 348l , 28 king i. r. , 1966 , aj , 71 , 64 kraicheva z. t. , popova e. i. , tutukov a. v. , iungelson l. r. , 1978 , astronomicheskii zhurnal , 55 , 1176 kroupa p. , 2001 ,
mnras , 322 , 231 kudritzki r. p. , pauldrach a. , puls j. , 1987 , a&a , 173 , 293 kudritzki r .-
p . , puls j. , 2000 , ara&a , 38 , 613 kudritzki r. p. , 2002 ,
apj , 577 , 389 kuhn m. a. , baddeley a. , feigelson e. d. , getman k. v. , broos p. s. , townsley l. k. , povich m. s. , naylor t. , king r. r. , busk h. a. , luhman k. l. , 2012 , arxiv:1208.3492 , to appear in the labyrinth of star formation , ( eds . ) d. stamatellos , s. goodwin , and d. ward - thompson , springer , in press lada ch . j. , lada e. a. , 2003 , ara&a , 41 , 57 lee c .- h . , brown g. e. , wijers r. a. m. j. , 2002 , apj , 575 , 996 leitherer c. , robert c. , drissen l. , 1992 , apj , 401 , 596 linden t. , kalogera v. , sepinsky j. f. , prestwich a. , zezas a. , gallagher j. , 2010 , apj , 725 , 1984 ( l10 ) liu j .- f . , bregman j. , miller j. , kaaret ph .
, 2007 , apj , 661 , 165 liu j .- f . , bregman j. , seitzer p. , 2004
, apj , 602 , 249 livio m. , warner b. , 1984 , the observatory , 104 , 152 madhusudhan n. , justham s. , nelson l. , paxton b. , pfahl e. , podsiadlowski ph . , rappaport s. , 2006 , apj , 640 , 918 madhusudhan n. , rappaport s. , podsiadlowski ph . , nelson l. , 2008 , apj , 688 , 1235 maeder a. , meynet g. , 2010 , new astronomy , 54 , 32 maeder a. , meynet g. , 2012 , reviews of modern physics , 84 , 25 maeder a. , 1992 , a&a , 264 , 105 ( m92 ) mapelli m. , sigurdsson s. , colpi m. , ferraro f. r. , possenti a. , rood r. t. , sills a. , beccari g. , 2004 , apj , 605l , 29 mapelli m. , sigurdsson s. , ferraro f. r. , colpi m. , possenti a. , lanzoni b. , 2006 , mnras , 373 , 361 mapelli m. , colpi m. , zampieri l. , 2009 , mnras , 395l , 71 mapelli m. , ripamonti e. , zampieri l. , colpi m. , bressan a. , 2010 , mnras , 408 , 234 mapelli m. , ripamonti e. , zampieri l. , colpi m. , 2011a , an , 332 , 414 mapelli m. , ripamonti e. , zampieri l. , colpi m. , 2011b , mnras , 416 , 1756 marks m. , kroupa p. , dabringhausen j. , pawlowski m. s. , 2012 , mnras , 422 , 2246 mercer e. p. , et al .
, 2005 , apj , 635 , 560 merritt d. , 2001 , apj , 556 , 245 miller m. c. , hamilton d. p. , 2002 , mnras , 330 , 232 narayan r. , mcclintock j. e. , 2005 , apj , 623 , 1017 nelemans g. , yungelson l. r. , portegies zwart s. f. , verbunt f. , 2001 , a&a , 365 , 491 orosz j. a. , 2003 , in van der hucht k. a. , herrero a. , esteban , c. , eds , proc . iau symp .
212 , _ a massive star odyssey : from main sequence to supernova _ , p. 365
orosz j. a. , et al .
, 2007 , nature , 449 , 872 zel f. , psaltis d. , narayan r. , mcclintock j. e. , 2010 , apj , 725 , 1918 patruno a. , zampieri l. , 2008 , mnras , 386 , 543 pattabiraman bh .
, umbreit s. , liao w .- k . , choudhary a. , kalogera v. , memik g. , rasio f. a. , 2012 , submitted to apj supplements , arxiv:1206.5878 pauldrach a. w. a. , vanbeveren d. , hoffmann t. l. 2012,a&a , 538 , 75 pfalzner s. , 2009 , a&a , 498 , l37 pilyugin l. s. , vlchez j. m. , contini t. , 2004 , a&a , 425 , 849 podsiadlowski ph . , rappaport s. , pfahl e. d. , 2002 , apj , 565 , 1107 podsiadlowski ph . , rappaport s. , han z. , 2003 , mnras , 341 , 385 portegies zwart s. f. , verbunt f. , 1996 , a&a , 309 , 179 portegies zwart s. f. , verbunt f. , ergma e. , 1997 , a&a , 321 , 207 portegies zwart s. f. , mcmillan s. l. w. , hut p. , makino j. , 2001 , mnras , 321 , 199 portegies zwart s. f. , mcmillan s. l. w. , 2002 , apj , 576 , 899 portegies zwart s. f. , 2004 , arxiv : astro - ph/0406550 , lecture note to appear in `` joint evolution of black holes and galaxies '' of the series in high energy physics , cosmology and gravitation .
iop publishing , bristol and philadelphia , 2005 , eds m. colpi , v.gorini , f.haardt and u.moschella portegies zwart s. f. , mcmillan s. l. w. , makino j. , 2007 , mnras , 374 , 95 portegies zwart s. f. , mcmillan s. l. w. , gieles m. , 2010 , ara&a , 48 , 431 poutanen j. , fabrika s. , valeev a. f. , sholukhova o. , greiner j. , 2012 , mnras , submitted , arxiv:1210.1210 prestwich a. h. , kilgard r. , crowther p. a. , carpano s. , pollock a. m. t. , zezas a. , saar s. h. , roberts t. p. , ward m. j. , 2007 , apj , 669 , l21 quinlan g. d. , 1996 , newastronomy , 1 , 35 ramsey c. j. , williams r. m. gruendl r. a. , chen c .- h .
r. , chu y .- h .
, wang q. d. , 2006 , apj , 641 , 241 rangelov b. , prestwich a. h. , chandar r. , 2011 , apj , 741 , 86 rappaport s. a. , podsiadlowski ph .
, pfahl e. , 2005 , mnras , 356 , 401 richards e. e. , lang c. c. , trombley ch . , figer d. f. , 2012 , aj , 144 , 89 sepinsky j. , kalogera v. , belczynski k. , 2005 , apj , 621l , 37 sigurdsson s. , phinney e. s. , 1993 , apj , 415 , 631 sigurdsson s. , phinney e. s. , 1995 , apjs , 99 , 609 silverman j. m. , filippenko a. v. , 2008 , apj , 678 , l17 smartt s. j. , 2009 , ara&a , 47 , 63 sollima a. , carballo - bello j. a. , beccari g. , ferraro f. r. , fusi pecci f. , lanzoni b. , 2010 , mnras , 401 , 577 soria r. , cropper m. , pakull m. , mushotzky r. , wu k. , 2005 , mnras , 356 , 12 soria r. , 2006 , in populations of high energy sources in galaxies , edited by meurs e. j. a. ; fabbiano g .. cambridge : cambridge university press , 473 soria r. , kuntz k. d. , winkler p. f. , blair w. p. , long k. s. , plucinsky p. p. , whitmore b. c. , 2012 , apj , 750 , 152 spitzer l. , 1987 dynamical evolution of globular clusters , princeton university press stodolkiewicz j. s. , 1982 , acta astronomica , 32 , 63 stodolkiewicz j. s. , 1986 , acta astronomica , 36 , 19 swartz d. a. , tennant a. f. , soria r. , 2009 , apj , 703 , 159 swartz d. a. , 2010 , galaxy wars : stellar populations and star formation in interacting galaxies asp conference series vol
. 423 . edited by smith b. , bastian n. , higdon s. j. u. , and higdon j. l .. san francisco : astronomical society of the pacific , p.277 tao l. , feng h. , gris f. , kaaret ph . , 2011 ,
apj , 737 , 81 terashima y. , inoue h. , wilson a. s. , 2006 , apj , 645 , 264 thompson t. a. , quataert e. , burrows a. , 2005 , apj , 620 , 861 vanbeveren d. , 2009 , new astronomy reviews , 53 , 27 van der hucht k. a. , 1991 , in ` wolf - rayet stars and interrelations with other massive stars in galaxies ' : proceedings of the 143rd symposium of the international astronomical union . edited by karel a. van der hucht and bambang hidayat . international astronomical union .
symposium no .
143 , kluwer academic publishers , dordrecht , p.19 vink j. s. , de koter a. , lamers h. j. g. l. m. , 2001 , a&a , 369 , 574 vink j. s. , de koter a. , 2005 , a&a , 442 , 587 voss r. , nielsen m. t. b. , nelemans g. , fraser m. , smartt s. j. , 2011 , mnras , 418 , l124 wheeler j. c. , akiyama s. , williams p. t. , 2005 , ap&ss , 298 , 3 woosley s. e. , heger a. , weaver t. a. , 2002 , reviews of modern physics , 74 , 1015 woosley s. e. , bloom j. s. , 2006 , ara&a , 44 , 507 zampieri l. , roberts t. , 2009 , mnras , 400 , 677 zezas a. , fabbiano g. , rots a. h. , murray s. s. , 2002 , apj , 577 , 710 zuo z .- y . , li x .- d . , 2010 , mnras , 405 , 2768
an initial binary fraction @xmath90 was adopted for all the simulations presented in the main text . in this appendix
, we show the results of a set of 100 supplementary runs without primordial binaries ( @xmath239 ) . while these simulations are quite unrealistic , as primordial binaries are expected to exist in scs , they are useful to illustrate the importance of primordial binaries .
all the other properties of the scs are the same as in the main text . in order to better see the effects on msbhs , we maximize their formation probability by taking @xmath3 z@xmath2 .
llllllllllll @xmath53 [ z@xmath125 & @xmath126 & @xmath127 & @xmath128 & @xmath129 & @xmath130 [ myr ] & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath136 & @xmath137 + 0.01 & 8.12 & 0.27 & 0.27 & 2.5 & 38.5 & 0.034 & 0.034 & 0.009 & 0.009 & 0.17 & 0.43 + table a1 shows some of the most relevant properties of the bh population in the runs with @xmath239 and @xmath3 z@xmath2 .
some interesting considerations can be drawn from the comparison of table a1 with table 2 .
first , the number of bhs per cluster is about 10 per cent lower than in the case with @xmath90 .
this is because a number of bhs form from the merger of relatively small stars ( 13 - 25 m@xmath40 ) in primordial binaries , if @xmath240 .
this is evident from fig .
[ fig : figa1 ] , where the initial mass spectrum of bhs in the runs with @xmath90 and with @xmath239 are compared .
no bhs with mass @xmath241 m@xmath2 form in the case with @xmath239 , whereas all the bhs with mass @xmath242 m@xmath2 formed in the runs with @xmath90 are remnants of merged progenitors .
we stress that the mass of bhs born from merged progenitors depends strongly on the assumptions about mass loss during the merger .
the fraction of bhs that become members of a binary at least once in the simulation ( @xmath127 ) in table a1 is almost the same as in table 2 . on the other hand , @xmath243 by construction in the case with @xmath239 .
thus , the number of exchanges that produce bh binaries is higher if @xmath239 .
similarly , the average number of exchanges per binary ( @xmath129 ) is almost double in the case of @xmath239 than in the case of @xmath90 .
these results indicate that exchanges involving bhs are more numerous if @xmath239 .
furthermore , the properties of accreting bh binaries are considerably affected by the assumed fraction of primordial binaries .
wind - accreting systems and especially rlo systems are strongly suppressed if @xmath239 .
this can be easily understood in light of the formation mechanism of binaries . if primordial binaries are allowed to form , then a number of primordial binary systems are expected to be sufficiently close to start rlo at early times , as a consequence of stellar evolution . in the absence of primordial binaries
, binaries can form only through the dynamical interaction of three single stars ( hereafter , three - body capture ) and through a tidal interaction between two single stars ( hereafter , tidal capture ) .
three - body capture binaries are generally wide and eccentric ( hut et al .
1992 ) , whereas tidal - capture binaries are very hard and generally merge ( e.g. , portegies zwart et al .
these two mechanisms are unlikely for most of the sc life , but are enhanced during core collapse ( spitzer 1987 ) . in our simulations
, we do not include any recipes for tidal - capture binaries .
thus we have only three - body capture binaries . only a small fraction ( @xmath244 ) of bh binaries
born from three - body capture survives ionization and becomes hard enough ( through three - body encounters ) to start rlo within 100 myr since the beginning of the simulation .
the right - hand panel of fig .
[ fig : figa2 ] confirms this consideration , by showing that all rlo systems switch on at relatively late times .
llllllllll @xmath53 [ z@xmath125 & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 [ myr ] & @xmath143 & @xmath144 & @xmath145 & @xmath146 + 0.01 & 1.36 & 0.85 & 0.85 & 2.6 & 51.8 & 0.110 & 0.110 & 0.022 & 0.022 + the last two columns of table a1 suggest another important consideration : if @xmath239 , msbhs are much more efficient than low - mass bhs in powering rlo systems .
in fact , the fraction of rlo systems powered by msbhs ( @xmath245 ) is about twice as high as the fraction of msbhs with respect to all bhs ( @xmath246 ) .
the bias towards high bh masses in rlo systems appears even more evident if we look at fig .
[ fig : figa2 ] : all the bhs powering rlo systems have masses @xmath247 m@xmath2 .
this is a strong confirmation that the more a bh is massive , the more its chances of entering a binary by dynamical exchange are high .
finally , table a2 shows some of the most relevant properties of msbhs in the runs with @xmath239 . from the comparison with table
3 , it is apparent that more msbhs can form in the absence of primordial binaries ( @xmath248 and 1.18 if @xmath239 and 0.1 , respectively ) . if @xmath240 , the mass loss following the ce phase in a primordial binary tends to produce smaller bh masses
finally , the fraction of rlo systems powered by msbhs is a factor of 10 smaller if @xmath239 than if @xmath90 . in summary ,
all the aspects discussed in this appendix indicate that primordial binaries have a crucial importance for the population of x - ray binaries in scs .
llllllllllll @xmath53 [ z@xmath125 & @xmath126 & @xmath127 & @xmath128 & @xmath129 & @xmath130 [ myr ] & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath136 & @xmath137 + 0.01 & 9.58 & 0.29 & 0.19 & 1.1 & 30.9 & 0.050 & 0.017 & 0.050 & 0.008 & 0.14 & 0.04 + llllllllll @xmath53 [ z@xmath125 & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 [ myr ] & @xmath143 & @xmath144 & @xmath145 & @xmath146 + 0.01 & 1.34 & 0.69 & 0.61 & 1.5 & 37.3 & 0.030 & 0.030 & 0.015 & 0.015 + in this paper , we follow the stellar evolution recipes described in section 3.1 and we assume that stars with @xmath54 m@xmath2 directly collapse to msbhs .
different stellar evolution recipes and different assumptions for the end of massive star life have important effects on the presented results .
a complete comparison between different stellar evolution recipes will be done in the next papers of the series . in this appendix ,
we just show a simplified , rather extreme case , in which the formation of msbhs is strongly suppressed . in particular , we assume that the maximum bh mass is @xmath249 m@xmath2 , even for @xmath3 z@xmath2 .
thus , in these runs the msbhs can have only @xmath250 m@xmath2 .
we performed 50 runs of sc evolution assuming @xmath249 m@xmath2 , @xmath3 z@xmath2 and leaving all the other properties of the scs as described in section 3 .
the main results for the bh population and for x - ray binaries powered by bhs are summarized in tables b1 and b2 .
table b1 shows that @xmath126 , @xmath127 , @xmath128 and @xmath129 are very similar to the case with @xmath3 z@xmath2 described in table 2 .
the most significant difference between table b1 and table 2 is the value of @xmath137 .
the fraction of msbhs that power rlo systems is much smaller in table b1 than in table 2 ( 0.04 versus 0.24 ) .
this confirms that msbhs in the mass range @xmath32 m@xmath2 are much more efficient in powering rlo systems than bhs with @xmath249 m@xmath2 .
the same conclusion can be derived from the comparison between table 3 ( for @xmath3 z@xmath2 ) and table b2 , where the statistics for msbhs is shown .
the msbhs in table b2 ( whose mass is @xmath251 m@xmath2 by construction ) are members of binaries and power rlo systems less often than the msbhs in table 3 . | we present @xmath0body simulations of intermediate - mass ( @xmath1 m@xmath2 ) young star clusters ( scs ) with three different metallicities ( @xmath3 , 0.1 and 1 z@xmath2 ) , including metal - dependent stellar evolution recipes and binary evolution . following recent theoretical models of wind mass loss and core collapse supernovae
, we assume that the mass of the stellar remnants depends on the metallicity of the progenitor stars .
in particular , massive metal - poor stars ( @xmath4 z@xmath2 ) are enabled to form massive stellar black holes ( msbhs , with mass @xmath5 m@xmath2 ) through direct collapse .
we find that three - body encounters , and especially dynamical exchanges , dominate the evolution of the msbhs formed in our simulations . in scs with @xmath3 and 0.1 z@xmath2 , about 75 per cent of simulated msbhs form from single stars and
become members of binaries through dynamical exchanges in the first 100 myr of the sc life .
this is a factor of @xmath6 more efficient than in the case of low - mass ( @xmath7 m@xmath2 ) stellar black holes .
a small but non - negligible fraction of msbhs power wind - accreting ( @xmath8 per cent ) and roche lobe overflow ( rlo , @xmath9 per cent ) binary systems .
the vast majority of msbh binaries that undergo wind accretion and/or rlo were born from dynamical exchange .
this result indicates that msbhs can power x - ray binaries in low - metallicity young scs , and is very promising to explain the association of many ultraluminous x - ray sources with low - metallicity and actively star forming environments .
black hole physics stars : binaries : general galaxies : star clusters : general x - rays : binaries methods : numerical stars : kinematics and dynamics . |
( color online ) schematic setup of trng based on the laser phase noise measurement using delayed self - homodyne method .
bs : beam splitter ; apd : silicon avalanche photodetector with the low ( high ) cutoff frequency of 50 khz ( 1 ghz ) ; adc : 8-bit binary analog - digital - converter working at @xmath4 mhz .
+ + fig .
( color online ) ( a ) the laser phase ( detection background ) noise of the laser field is observed at @xmath44 ma ( @xmath45 ma ) , while the threshold current is @xmath46 ma .
( b ) autocorrelation function of the beat signal vs time delay ( red line corresponds to autocorrelation of @xmath47 ) .
the sampling frequency is @xmath48 mhz ( @xmath49 ns ) .
+ + fig .
( color online ) ( a ) histogram of the laser intensity digitized by an 8-bit adc after reducing bias by exclusive - or post - processing , and after entropy amplification by bit extraction ( inset ) .
( b ) normalized entropy @xmath50 of the random bit sequence when the @xmath0-least - significant - bit ( @xmath0-lsb ) is adopted .
+ + fig .
( color online ) the results of the randomness tests of ( a ) diehard and ( b ) sts .
the diehard test is considered successful when all @xmath51 values satisfy @xmath52 .
the sts test is considered successful when all @xmath51 values are larger than 0.0001 and the passed portions are between @xmath53 .
ma ( @xmath45 ma ) , while the threshold current is @xmath46 ma .
( b ) autocorrelation function of the beat signal vs time delay ( red line corresponds to autocorrelation of @xmath47 ) .
the sampling frequency is @xmath48 mhz ( @xmath49 ns).,width=245 ] | we present a simple setup to implement truly random number generator based on the measurement of the laser phase noise . from the entropy point of view
, we estimate the number of truly random bits that can be extracted from the sampled byte . with a simple method of adopting the @xmath0-least - significant - bit
, we amplify the entropy of the original bit sequence and realize a truly random bit generation rate of @xmath1 mbps .
the application of random number generator ( rng ) covers wide areas of cryptography , statistical sampling , computer simulations , etc .
traditionally , rng is implemented using the algorithm - based pseudo - random number generator ( prng ) , while in most application areas , prngs are not enough .
so far , it is commonly accepted that pure quantum process is the only solution for truly random number generator ( trng ) .
recently , two kinds of quantum noise of laser , phase noise and relaxation oscillation noise , are chosen to generate truly random bit sequence with high generation rate @xcite .
phase noise , directly measured using self - delayed homodyne detection @xcite , can be used to generate random bits with good randomness , but the final random number generation rate is limited by the laser linewidth ; relaxation oscillation noise , amplified by chaotic behavior , can be used to achieve ultra - high random bit generation rate @xcite , but the random bits generated by chaotic laser involve more classical noise ( not truly random ) and so can not have good results in three - sigma randomness tests @xcite . to generate an original random bit sequence ,
either an oscilloscope @xcite or an analog - digital - converter ( adc ) @xcite is available for high speed multi - bit per measurement data recording , but due to data size limitation in one scan process , using oscilloscope will reduce the generation rate for long random bit sequence .
multi - bit per measurement method increases the random number generation rate but introduces additional correlations .
the proper number of bits extracted from the recorded data can be estimated by randomness evaluation . in this letter , we propose a simple method to evaluate the randomness from the entropy point of view , which gives the upper and lower estimations of the maximal bits that can be extracted from one measurement .
the laser phase noise ( gaussian random variable @xcite ) measurement based trng scheme is adopted , where a single - mode vertical cavity surface emitting laser ( vcsel ) is used and sampled by an adc .
after entropy amplification by bit extraction , we realize a @xmath1 mbps trng , which passes three standard statistical tests .
note that this bit extraction method is also applicable for other random sources .
the schematic setup is shown in fig .
[ setup ] ( see also @xcite ) and the delayed self - homodyne method is used to measure the phase noise of the vcsel .
the beat signal of @xmath2 and @xmath3 is detected by the avalanche photodetector ( apd ) and sampled at @xmath4 mhz by the 8-bit binary adc .
when the delay time is much longer than the coherence time of laser ( @xmath5 ) , the autocorrelation function of the laser electric field is eliminated @xcite , i.e. , @xmath6 , where @xmath7 and @xmath8 is the laser linewidth . in this case , the beat signal of these two mutually independent laser fields can be considered as a good candidate to implement trng . since the phase noise of the laser is inversely proportional to the laser power , while the classical noises are laser power independent @xcite , we choose working current just above the threshold current to ensure that the phase noise is dominant and the true randomness can thus be guaranteed . in experiment , a @xmath9 nm vcsel laser works at 1.4 ma ( threshold current 1.3 ma ) , and its linewidth is @xmath10 mhz ( @xmath11 ns ) . with delay time @xmath12 ns , the power spectral density of the beat signal is shown in fig .
[ spectrum ] ( a ) . in fig .
[ spectrum ] ( b ) , we calculate the autocorrelation function @xmath13 from the power density spectrum in fig .
[ spectrum ] ( a ) using wiener - khintchine theorem @xcite . without any post - processing , the autocorrelation of the laser source sampled by 8-bit adc at @xmath14 mhz ( time delay @xmath15 ns ) is smaller than @xmath16 . due to the dependence on the external parameters in the experiment ,
the sampled original data digitized by an 8-bit adc is biased . to reduce the bias
, we perform a simple exclusive - or post - processing between every two 8-bit original data , which is a common method to improve the randomness of rng @xcite . after the bit extraction , the histogram of the data
is shown in fig .
[ histogram ] ( a ) , well fitted with a gaussian profile .
since truly randomness should have a uniform distributed histogram with shannon entropy of 8 bits per byte , the randomness in this original data is limited by the correlation within the byte . to evaluate the randomness ,
we calculate shannon entropy using the experiment results ( histogram data ) : @xmath17 , where @xmath18 is the probability of the data in the @xmath19 digitized bin ( totally @xmath20 bins ) .
this can roughly be viewed as that this data can extract @xmath21 random bits out of every byte using appropriate , yet maybe complicated bit extraction post - processing . to assure that shannon entropy can evaluate the randomness of the source , we calculate the entropy of laser source according to its gaussian fitting curve ( fig .
[ histogram ] ( a ) ) using @xmath22 , where @xmath23 is the standard deviation of gaussian profile .
the result of the entropy of the laser source is @xmath24 bits per byte , and experiments show these two values of shannon entropy get closer when the size of the data is larger .
to avoid complicated post - processing , adopting @xmath0-least - significant - bit ( @xmath0-lsb ) out of each byte is an efficient method to amplify the randomness . but this method is not strong in eliminating the correlation , and extracting @xmath25-lsb according to shannon entropy can not pass the random tests .
therefore , we change the randomness indicator to min - entropy , @xmath26 , which uses only the maximal probability and represents the worst - case scenario @xcite . by simply adopting the last 6-bit in each byte according to the result , @xmath27 bits per byte , the histogram of the new random bit sequence shows good uniformity ( inset of fig .
[ histogram ] ( a ) ) and represents good randomness .
hence , we can consider shannon entropy and min - entropy , respectively , as the upper and lower estimations of the randomness of the laser source . for true randomness ,
shannon entropy and min - entropy have the same value and both equal @xmath28 bits per byte . in our experiments , after the bit extraction process , the values of these two entropies are amplified to @xmath29 and @xmath30 bits per byte , respectively . to further illustrate the entropy amplification , we calculate the normalized entropy @xmath31 of the random bit sequence before and after bit extraction ( fig . [ histogram ] ( b ) ) , where @xmath32 represents shannon entropy with the length of tested - bit sequences @xmath0 . since shannon entropy holds @xmath33 for true randomness , normalized entropy should be unity for any @xmath0
. it can be seen from fig .
[ histogram ] ( b ) that , after bit extraction , the entropy of the random bit sequence is amplified and gets much closer to unity . for randomness tests
, we continuously record a final random bit sequence of 1 gbit , and it passes all three standard random tests , i.e. , ent : a pseudorandom number sequence test program @xcite , diehard : a battery of tests of randomness @xcite , and nist - sts : the national institute of standards and technology - statistical test suite @xcite .
the ent results are : entropy @xmath34 bit per bit ( the optimum compression would reduce the bit file by @xmath35 ) .
@xmath36 distribution is @xmath37 ( randomly would exceed this value less than @xmath38 of the times ) .
arithmetic mean value of data bits is @xmath39 ( @xmath40 random ) .
monte carlo value for @xmath41 is @xmath42 .
serial correlation coefficient is @xmath43 .
the results of diehard and sts test are shown in fig .
[ tests ] ( a ) and ( b ) , respectively . in summary
, we implement a simple and compact trng based on the measurement of the laser phase noise , which passes all three standard statistical random tests . to amplify the entropy
, we estimate the maximal bits that can be extracted from each sampled byte by lsb - adoption , and realize a @xmath1 mbps truly random number generation rate .
this work is supported by the key project of national natural science foundation of china ( nsfc ) ( grant 60837004 ) and national hi - tech research and development ( 863 ) program .
we acknowledge the help from w. tang , w. wei , y. li , j. zhou and j. zhang .
99 h. guo , w. tang , y. liu , and w. wei , truly random number generation based on measurement of phase noise of a laser , " phys .
rev .
e * 81 * , 051137 ( 2010 ) ; arxiv : 0908.2893v3 [ quant - ph ] . b. qi , y. m. chi , h. k. lo , and l. qian , high - speed quantum random number generation by measuring phase noise of a single - mode laser , " opt
. lett . * 35 * , 312 ( 2010 ) ; arxiv : 0908.3351v2 [ quant - ph ] .
a. uchida , k. amano , m. inoue , k. hirano , s. naito , h. someya , i. oowada , t. kurashige , m. shiki , s. yoshimori , k. yoshimura , and p. davis , fast physical random bit generation with chaotic semiconductor lasers , " nat .
photon .
* 2 * , 728 ( 2008 ) .
i. reidler , y. aviad , m. rosenbluh , and i. kanter , ultrahigh - speed random number generation based on a chaotic semiconductor laser , " phys . rev .
lett . * 103 * , 024102 ( 2009 ) .
m. lax , classical noise . v. noise in self - sustained oscillators , " phys
. rev .
* 160 * , 290 ( 1967 ) . c. h. henry ,
theory of the linewidth of semiconductor lasers , " ieee j. quantum electron .
* 18 * , 259 ( 1982 ) .
k. vahala and a. yariv , occupation fluctuation noise : a fundamental source of linewidth broadening in semiconductor lasers , " appl .
. lett . * 43 * , 140 ( 1983 ) .
n. wiener , acta math .
* 55 * , 117 ( 1930 ) .
m. epstein , l. hars , r. krasinski , m. rosner , and h. zheng , design and implementation of a true random number generator based on digital circuit artifacts , " lect .
notes comput .
sci . *
2779 * , 152 ( 2003 ) .
m. wayne , and p. kwiat , low - bias high - speed quantum random number generator via shaped optical pulses , " opt .
express * 18 * , 9351 ( 2010 ) .
j. walker , ent : a pseudorandom number sequence test program , " http://www.fourmilab.ch / random/. g. marsaglia , diehard : a battery of tests of randomness , " http://www.stat.fsu.edu / pub / diehard/. nist , sts : the national institute of standards and technology - statistical test suite , " http://csrc.nist.gov/groups/st/toolkit/rng/stats_tests.html . |
although clasical novae occur rather frequently in our galaxy ( some tens of events per year ) , they are not expected to contribute greatly to the chemical enrichment of the galaxy because they eject relatively small masses of nuclear - processed material .
the rare cno nuclei @xmath2c , @xmath3n and @xmath4o are exceptions , and so is the light isotope @xmath1li , which is the topic of the present study ( hernanz & jos 2000 ) .
about 10% of the present galactic abundance of @xmath1li may have come from novae , which are one of the several sources ( mainly cosmic ray spallation and neutrino spallation in supernovae : cass , vangioni - flam & audouze 2001 ) that raise the @xmath1li abundance above the cosmologically - interesting value left by big bang nucleosynthesis .
nova nucleosynthesis of @xmath1li is also of interest in that the @xmath1li is produced as unstable @xmath1be by the reaction @xmath5he(@xmath6,@xmath0)@xmath1be , where @xmath5he is another big bang isotope whose subsequent evolution must be accounted for .
the @xmath7-decay of @xmath1be with half - life 53.28 d is accompanied in 10.5% of cases by emission of a @xmath0-ray of energy 478 kev that potentially provides a direct measurement of the nova contribution if it can be detected .
monitoring of the @xmath0-ray sky for novae is important because many events are never detected ( @xmath8 out of several dozen per year are discovered , often by amateur astronomers : harris et al .
1999 , 2000 , hereafter papers i and ii ) , raising the possibility that original discoveries of optically - undetected novae may be made by their @xmath0-ray emission alone .
this requires experiments with broad fields of view carried on space platforms over periods of the order of years .
data from four such experiments have been analyzed with this purpose in view : the gamma ray spectrometer on the _ solar maximum mission _
( _ smm _ , 19801989 : see e.g. leising et al .
1988 ) ; batse and comptel on the _ compton _ observatory ( 19912000 : hernanz et al . 2000 and iyudin et al .
1995 , 2001 ) ; and the transient gamma ray spectrometer ( tgrs ) on board _ wind _ , from which we will present results from 19951997 in this paper .
the properties of the tgrs detector , which we describe in 3 , are very suitable for monitoring line emission from novae .
the detector has very good energy resolution , which enables doppler - shifted lines from cosmic sources to be distinguished from background lines at the rest energy ; we previously exploited this property in searches for the 511 kev positron annihilation line , which is blueshifted by a few kev in novae ( paper i , paper ii ) . in this paper
we exploit another property of tgrs to search for the 478 kev line , namely the remarkable stability of the instrument s background @xmath0-ray line spectrum ( 4 ) .
our analysis relies to some extent on the expected properties of the @xmath0-ray line from the nova event .
the most important factor in the nucleosynthesis is the composition of the white dwarf upon which the explosion occurs , since material is expected to have entered the burning layer by diffusion upward and to have altered its composition substantially from what was accreted .
most common are expected to be degenerate co remnants of @xmath9 left by lower mass red giants .
more massive objects up to the limiting stable mass @xmath10 probably have one composition .
it is not clear where the transition in mass between co and one occurs , or whether indeed there is a range of overlap .
although the general white dwarf mass distribution peaks around @xmath11 , the distribution of white dwarf masses in novae is skewed towards higher masses , because the frequency with which nova outbursts recur is a steeply rising function of mass . the one remnants and the more massive co objects @xmath12 are therefore over - represented . a ratio of 2:1 for the two subclasses co : one is often assumed .
abundance analyses are not available for most novae .
properties may be deduced ( with great uncertainty ) from the correlations of white dwarf subclass with other observables .
it appears that explosions on co white dwarfs are less energetic .
the ejected masses are of order @xmath13 for both subclasses , but co novae have lower expansion velocities
@xmath14 km s@xmath15 ( warner 1995 ) .
one of the individual novae that we will consider , the one type cp cru , had a velocity of 2000 km s@xmath15 ( della valle 1996 ) .
theoretical models predict that co novae are the strongest sources of the 478 kev @xmath0-ray line ( gomz - gomar et al .
1998 , hernanz & jos 2000 ) since they may produce about @xmath16 of @xmath1be per event .
however , predictions of the @xmath1be mass fraction produced vary considerably ( see e.g. starrfield et al .
2000a , hix et al .
2000 , and hernanz & jos 2000 ) .
further uncertainty is added by the failure ( in general ) of models to reproduce the full @xmath13 of ejecta , falling short by an order of magnitude or more .
the ejecta velocities are also usually underestimated .
distance estimates for novae are highly uncertain .
an empirical relation exists between rate of decline and absolute visual magnitude : @xmath17 ( warner 1995 ) , where @xmath18 , the speed class , is the time taken for @xmath19 to increase by 2 from discovery .
following the explosion in and ejections by thermonuclear runaway , the remaining surface layers undergo hydrostatic h burning on a longer time - scale @xmath20 yr ( shore , starrfield , & sonneborn 1996 ) and a convective envelope developes . during this time
the bolometric luminosity is thought to remain constant at the eddington value @xmath21 ( warner 1995 ) , but the peak of the emission moves from the optical to the uv as expansion reveals hotter underlying layers .
this leads to a second method of estimating distance that has sometimes been used ( e.g. stickland et al .
1981 , evans et al . 1990 ) .
let it be assumed that at visual maximum almost all the flux is at visual wavelengths [ as stickland et al .
found for v1668 cyg ( 1978 ) ] .
the absolute visual magnitude is then known from the eddington formula @xmath22 erg s@xmath15 , where @xmath23 is the mean molecular weight and @xmath24 the white dwarf mass in @xmath25 ; the distance follows from a comparison of apparent and absolute visual magnitudes .
post - outburst behavior raises further questions about @xmath1be observability , which depends on the time - scale on which the expanding envelope becomes transparent to the @xmath0-ray line ; it obviously must be well within the 53 d half - life of @xmath1be if the line is to be seen.be in a low - velocity , late - time wind of which part is accelerated by a drag force due to the companion when a common envelope developes between primary and secondary ( livio et al .
it was applied to @xmath0-ray line emission in novae v838 her ( starrfield et al . 1992 , 1993 ) and v1974 cyg ( starrfield et al . 1993 ) .
models of this type will not become transparent to @xmath0-rays for @xmath26 d or two half - lives , severely depleting the detectable @xmath1be .
] we assume that this is the case , following gomz - gomar et al .
( 1998 ) , who predict @xmath0-ray transparency after 513 days , depending on the mass and composition of the model .
these models also predict large yields of @xmath1be and 478 kev line fluxes .
our measurements can therefore be regarded as an approach to testing these `` optimistic '' models . during 19951997 five novae
were discovered in the southern hemisphere .
they are listed in table 1 , with composition information where available , and distance estimates from the two empirical methods described above .
there are obvious sources of uncertainty in the distance measurements , notably the time of visual maximum ( before discovery ) and , in the case of the @xmath27 method , possible neglect of a bolometric correction and super - eddington velocities in some novae ( schwarz 1999 ) .
in addition , there is a single reliable distance measurement ( 3180 pc for cp cru from expansion parallax : downes & duerbeck 2000 ) .
relative to paper i ( see previous footnote ) there are a few changes in table 1 due to our improved knowledge of the times of visual maxima , thanks to diligent searches of the records of southern hemisphere amateurs ( f. bateson & a. jones , private communication ) .
in space - based @xmath0-ray astronomy the signal from any cosmic source is usually completely dominated by intense ambient background radiation arising from bombardment by the all - pervasive energetic cosmic ray particles , and by the ensuing secondary radiations such as @xmath7-decays .
the central problem that all experiments must face is therefore how to subtract away this huge background ( typically @xmath28 of count rate ) . in general
, the signal has to be modulated .
methods include pointing on and off source rapidly ( as with cgro/_osse _ , johnson et al .
1993 ) , modulation in time by a periodic occulter ( e.g. an opaque aperture stop , or the spinning spacecraft body like _ heao-3 _ : mahoney et al .
1982 ) , or even earth ( cgro/_batse _ : ling et al .
2000 ) , or modulation in space by a coded mask illuminating a segmented detector plane ( to be used by _
integral_/spi : vedrenne et al .
it is important to note that our work with tgrs used two distinct methods of background subtraction .
one is a mechanical occulter ( 5 ) , but the other is unique to tgrs unlike any other experiment tgrs can model its background lines and their temporal behavior accurately enough to perform a usable background subtraction ( 4 ) .
the two methods of background subtraction yield results for different `` samples '' of novae .
occultation measures the averaged collective @xmath1be emission from the general direction of the gc .
background modeling ought to detect sudden increases in an otherwise predictable background line time history , each the signature of an individual nova .
we can search the whole time history , not just the times of the known novae in table 1 , in the hope of detecting nearby undiscovered events .
the tgrs experiment is located on the south - facing surface of the rotating cylindrical _ wind _ body , which points permanently toward the southern ecliptic pole .
the spectrometer , a radiatively cooled 35 @xmath29 n - type ge crystal sensitive to energies between 20 kev8 mev , is kept at its operating temperature 85 k by a passive radiative cooler constructed mainly of be and mg .
apart from 30 mm of be / cu alloy that absorbs solar x - rays , it is unshielded , so that in normal ( non - burst ) operation background spectra are accumulated continuously from the entire southern ecliptic hemisphere . a 1 cm thick pb occulter attached to the _ wind _ body exploits the 3 s rotational period in order to modulate the signal from part of the sky .
the occulter subtends @xmath30 along the ecliptic plane , rather than exactly the plane of the ecliptic , which would be @xmath30 .
the @xmath31 offset was designed to optimize modulation of the gc . ]
as seen from the detector , the width of the band thereby occulted being @xmath32 fwhm at around 500 kev .
owens et al .
( 1995 ) reported the results of extensive monte carlo simulations of the instrument response as a function of energy and zenith angle .
the actual cross section 35 @xmath29 is customarily reduced to an `` effective '' area by factoring in the efficiency of detection at a given energy and angle .
further , the frequent cases where photon energy is only partly lost in the detector can not be used in the present analysis , since the counts appear in a smooth quasi - continuum at energies below the line energy , and no one count can be unambiguously associated with a weak line like that at 478 kev .
our analysis therefore only uses photons which undergo full energy loss ( where in any case the ge response as a function of energy loss is maximal ) , i.e. which appear in the channel(s ) at the nominal line energy and are referred to as the `` photopeak '' . in these terms , the photopeak effective area of the tgrs detector is @xmath33 @xmath29 at 478 kev and zenith angle @xmath34 .
it falls sharply with increasing energy but does not vary much with zenith angle ( owens et al .
the instrument response was operationally verified by on - orbit comparisons of @xmath0-ray burst spectra with those obtained by cgro/_batse _ ( seifert et al . 1998 ) and _ wind_/konus ( palmer et al .
1996 ) .
the sharp energy resolution of the detector is critical for our analysis .
scintillator detectors like _ smm_/grs and _ batse _ typically achieve @xmath35 kev resolution around 500 kev .
any line at 478 kev is therefore blended with the positron annihilation line at 511 kev , which is much stronger in both background and background - subtracted spectra .
the resolution of tgrs at this energy was nominally about 3 kev fwhm , which was achieved in the early months of the mission ( harris et al .
1998 ) ; thereafter resolution degraded due to accumulated damage from cosmic ray impacts .
the line profiles also became distorted , with marked tails on the low - energy wings .
we limited the data analysis to the period before fall 1997 , after which the deterioration became marked ( kurczynski et al .
the detector gain stability also deteriorated on this time - scale .
although the gain shift was not difficult to monitor using the positions of known narrow background lines , the effect on the analysis of occulted data was serious ( 5.1 ) .
the tgrs background spectra were binned in 1 kev energy channels over the entire 20 kev 8 mev range in 24 minute intervals .
the occulted spectra were accumulated in 128 angle bins covering each 3 s revolution of _
telemetry constraints did not allow all of these occulted data to be returned in finely - resolved energy bins , however .
data above 1 mev were discarded .
four energy windows were defined to be 20150 kev , 150400 kev , 4001000 kev , and 479543 kev , with energy binnings of 2 , 4 , 8 and 1 kev respectively .
the boundaries of the windows changed during the mission as a result of the gain shift .
the unusual orbits followed by _
wind _ since launch provide the key to the success of the tgrs background modeling .
they have all been highly elliptical ( acua et al . 1995 ) , and tgrs has thus spent virtually the whole mission in interplanetary space , avoiding major sources of @xmath0-ray background radiation such as earth albedo @xmath0-rays and energetic trapped charged particles .
the chief source of particle bombardment is the galactic cosmic ray ( gcr ) flux .
there have been very few interruptions in the data stream , generally caused by brief passages of the trapped radiation belts around perigee and by triggering of a special mode of data collection by solar flares and @xmath0-ray bursts .
the live time in this data set covers about 90% of the analysis period ( paper ii ) .
the resulting tgrs background lines may be strong , but they are stable enough for our purposes ( 4 below ) .
by contrast , experiments in low earth orbit are plagued by trapped radiation belt - induced radioactivity , generating background lines varying on a wide variety of @xmath7-decay time - scales , many of which never reach equilibrium , and can not be predicted on either long or short time - scales ( weidenspointner et al .
our analyses of the tgrs background and occulted data sets used quite distinct methodologies that were designed to measure different quantities . here
we focus on the background spectra . a background spectrum is composed of many narrow lines from radioactivity induced by gcr and secondary neutron impacts , superimposed on a continuum .
the large mass of be close to the tgrs detector in the radiative cooler and sunshield ( 3 ) gives rise to a strong line at exactly 478 kev from the interactions @xmath36be(@xmath37,@xmath38)@xmath1be and @xmath36be(@xmath39,@xmath40)@xmath1be ( naya et al .
we measured the strength of this line in successive spectra from the years 19951997 . any sudden increment in the strength
would correspond to the signal from a cosmic source of @xmath1be , and would be confirmed by a subsequent decline on a 53 d half - life .
this analysis can not detect a dc level of 478 kev line emission .
this method clearly relies on the underlying background 478 kev line being very stable , or at least varying in a simple and comprehensible way . as mentioned above , the predominant factor has been radioactivity induced by the gcr flux which varies slowly with the 11 yr solar activity cycle .
solar magnetic modulation is expected to cause the flux to peak around solar activity minimum , which occurred about halfway through our 19951997 analysis period .
we found evidence for this in the count rate for partial energy losses @xmath41 mev in the detector that is shown in fig .
1 ( top ) .
this is above the energy region characteristic of @xmath0-rays produced by nuclear interactions , so that the background must be due principally to gcr bombardment and the diffuse cosmic background , of which the former only is time - variable .
the day - to - day variability of the @xmath41 mev rate ( top full line ) shows a very weak @xmath42 d period that is caused by the _ wind _
orbit crossing the neutral current sheet in the interplanetary magnetic field ( which varies with the 28 d solar rotation ) .
when integrated over the 53 d time - scale characteristic of nova signals , the count rate shows a modest , smooth variation ( histogram ) that follows the trends expected for the gcr flux when subject to solar modulation .
this pattern of variability was confirmed by a large number of measurements of tgrs background line strengths performed for line identification purposes ( weidenspointner et al .
many prompt de - excitation lines were identified that arise from electromagnetic decays of excited nuclear states on submicrosecond time - scales , and it is clear that the intensities of these lines must trace the incident gcr flux effectively instantaneously .
these trends are also shown in fig . 1 ( full lines ) .
the background spectra were summed in 53 d intervals . in each interval
the energy range 460490 kev was fitted with a spectral model consisting of a power law and two lines .
there is a strong line at 472.2 kev due to @xmath43na from spallation of structural al , and the 477.6 kev @xmath1be line from the radiative cooler .
the @xmath43na line was slightly distorted by the low - energy tailing problem even before the fall of 1997 ( 3.2 ) .
it was therefore fitted by an asymmetric function @xmath44 for amplitudes @xmath45 , @xmath46 , line energy @xmath47 and width parameters @xmath48 and @xmath49 ( phillips & marlow 1976 ) .
the line at 478 kev was not fitted by this function , since other factors broaden it ( symmetrically ) to such an extent that the asymmetry is not significant .
one such factor is intrinsic ; this line is measured to be somewhat broader than the instrument resolution due to blending with a weak line of @xmath50co .
alternatively , folding its profile with a gaussian corresponding to 1000 km s@xmath15 ( typical for co novae : 2 ) has the same effect . a simple gaussian with this width was therefore included in the fit . a specimen of such a fit is shown in fig .
the amplitude in terms of count rate was corrected for efficiency using the photopeak effective area as a function of energy and zenith angle ( 3.2 ) .
the zenith angles of the five known 19951997 novae were readily obtained ; in the general case ( searching for a flux increase when no nova was known to be present ) we assumed a zenith angle of @xmath51 , which is the average for the estimated distribution of southern novae ( hatano et al .
1997 ) .
the time series of the line amplitudes fitted to the 478 kev line in the background spectra is shown in fig . 3 ( _ top _ ) .
we fitted this time series with a model by which @xmath1be is created in the instrument by a time - dependent gcr flux , and decays with a half - life 53.28 d. the various curves in figure 1 suggest that a simple parametrization of the gcr flux would be a linear increase between 1995 january and solar minimum in 1996 june , followed by an abrupt change of slope
. let the corresponding @xmath1be production rate at time @xmath52 be @xmath53 where @xmath54 and @xmath55 are constants and @xmath55 changes slope at solar minimum .
the time - dependent @xmath1be abundance is then @xmath56~ + ~\frac{c t}{\lambda}~+~n(t=0)~\exp{(-\lambda t)}\end{aligned}\ ] ] where @xmath57 d. we fitted the time series to this function ( which has 5 free parameters when account is taken of the discontinuity at solar minimum ) and searched for significant sudden flux increases over and above it .
this was done by fitting an exponentially decaying function with decay constant @xmath58 to each residual data point and the following ones .
the fit using eq .
( 5 ) is shown in fig . 3 ( _ top _ ) , and the residuals in fig . 3 ( _ bottom _ ) .
it is evident from fig .
3 ( _ top _ and _ bottom _ ) that the function ( 2 ) fitted the data very well , so the procedure of fitting the exponential decay could be applied without obvious systematic errors . the clear result in fig . 3 ( _ bottom _ )
is that no flux increase followed by exponential decline was found to appear in the 478 kev line , either at the times of known southern hemisphere novae or at any other time .
clearly no hitherto undiscovered novae were detected by @xmath0-ray line emission using this method .
the @xmath59 upper flux limits for this general case , and for the five known novae , are given in table 2 .
our analysis of the occulted data was designed to measure a quasi - steady , quasi - diffuse 478 kev @xmath0-ray line arising from the integrated emission of many novae that would have occurred towards the gc during the 53 d half - life of @xmath1be .
the principle was suggested by leising ( 1988 ) , and was put into practice by harris , leising & share ( 1991 ) .
our procedure was essentially identical to that which we previously employed to measure the diffuse positron annihilation spectrum from the gc ( harris et al .
1998 ) . when the count rate in an energy channel is plotted as a function of angle along the ecliptic , those angle bins corresponding to the direction of a source show dips .
the amplitude of the dip equals the source count rate at that energy , so that a spectrum of the source is built by taking the source count rate dips in successive energy channels .
it is necessary to assume a model of the source s spatial distribution ; we chose a gaussian in ecliptic longitude of fwhm @xmath60 , which harris et al .
( 1998 ) found to be the best estimate for the distribution of the diffuse 511 kev annihilation line ( which is compatible with a nova - like distribution of sources : milne et al .
2001 ) .
the energy channel binning in the region of interest to us around 478 kev was non - uniform and time - variable .
this is because , as noted in 3.2 , data were only returned at the optimum 1 kev binning in a single 64-kev wide energy window centered approximately on 511 kev , and unfortunately the 478 kev line was on the edge of this window .
energy resolution in the occulted data was therefore bad ( 8 kev ) on the red side , and moreover the transition between well- and coarse - binned data varied during the mission as the high - resolution window shifted with the instrument gain . rather than using the complete 19951997 harris et al .
( 1998 ) tgrs spectrum , we used a series of spectra summed over 90 d intervals which harris et al .
also generated ( for their search for 511 kev line variability ) .
this time - scale happens to be similar to that for significant variability of the gain .
we could thus treat the gain shift as a constant and `` fix '' the location of the high - resolution window boundary in a manner that facilitated spectrum modeling .
the 90 d spectra in the range 460490 kev were fitted by a model consisting of a power law plus a gaussian line at 478 kev .
a typical example of this fit is shown in fig .
4 . note the abrupt transition from coarse to fine binning in the middle of the energy range .
the measured 478 kev line fluxes during 90 d intervals are shown in fig .
5 . these results may be combined to give the total measured quasi - steady flux from the gc .
it is clear that no line is detected , within limits which are given in table 2 .
the sensitivity of this search is degraded by a factor 45 relative to that of the background - modeling method .
this is mainly due to the gain shift problem mentioned above .
table 2 also shows earlier measurements by _ smm _ for both individual novae and the quasi - steady integrated gc flux ( harris et al . 1991 ) . note that the integrated flux measurements are not exactly comparable , since the _ smm _ value comes from a very broad region @xmath61 across around the gc , whereas the tgrs value reflects the signal from the @xmath62 occulted region only
from the flux upper limits in table 1 we calculate the @xmath1be abundances in the sources , given the distances .
our limits improve upon the _ smm _ values by an order of magnitude for individual events , and by a factor 2 for the integrated flux from the gc .
unfortunately we do not obtain proportionate improvements in the limits on the @xmath1be abundances from these fluxes , for two reasons . in the case of individual novae , by misfortune
the tgrs sample lay at greater distances than the _ smm _ sample . in the case of the gc , the measurement methods ( _ smm _ or tgrs occulter )
return @xmath1be abundances integrated over the unknown number of novae that are simultaneously present , and thus depend on the uncertain galactic nova rate @xmath63 and on the aperture of the instrument . since the effective aperture of tgrs is smaller than the @xmath61 aperture of _ smm _ , the upper limit implied by a given @xmath0-ray flux
is divided among fewer novae .
comparisons with theory must take into account the shortcomings and uncertainties in nova models ( see 2 ) . for our purposes
, the prediction of any nova s @xmath1be mass may be factored into the product of the ejected mass @xmath64 , and the mass fraction @xmath65 of @xmath1be in it . in the favorable case of a massive co nova , `` optimistic '' models such as gmez - gomar et al .
( 1998 ) suggest values @xmath66
. both of these values may be underestimates , leading to yet more optimistic scenarios that were proposed earlier by clayton ( 1981 ) .
there are two arguments for this kind of scenario . and
@xmath65 can be varied independently .
this is unlikely to be the case in actual models , although starrfield et al .
( 1997 ) made a similar assumption in order to generate @xmath0-ray predictions .
] first , as pointed out in 2 , there is a `` missing mass '' problem ( starrfield et al .
2000b ) : nova models in general predict @xmath64 that are much too low . measured ejecta
masses range from @xmath67@xmath68 ( warner 1995 ) ; the mean value is probably @xmath69 ( gehrz et al .
1998 ) . as for @xmath65
, it was pointed out by starrfield et al .
( 1978 ) that , whereas models assume that the abundance of @xmath5he in the gas accreted onto the white dwarf is usually solar , this is probably not the case .
the @xmath5he abundance should be enhanced in the donor star due to synthesis of this isotope in the pp chain , and convection will mix the @xmath5he - rich material into the surface layers which are transported to the white dwarf .
older main sequence star models suggested @xmath5he enhancements @xmath70 times solar , while more recent models suggest a value about 50% of this ( morel et al .
1999 ) . however , boffin et al .
( 1993 ) showed that , contrary to what was assumed by starrfield et al .
( 1978 ) , the yield of @xmath1be from @xmath5he(@xmath6,@xmath0)@xmath1be in novae is not proportional to the @xmath5he abundance , but to its logarithm .
the model yields of @xmath1be are then probably enhanced by less than a factor 2.5 by the @xmath5he enhancement .
most of the possible variation in the product is thus due to uncertainty in @xmath64 , of which our upper limits may be a test . for the purpose of comparison with our upper limits ,
we show in table 2 the fluxes that would be expected from each nova ( and from the gc ) if optimistic values @xmath71 and @xmath72 , similar to the old values of clayton ( 1981 ) , are assumed . in all cases ,
our limits lie above the expected fluxes ; however some of the individual cases are of interest .
our limit for by cir is only a factor @xmath73 above the expected value , even better than the _ smm _ limit for the much closer event v482 cen .
these happen to be the only verified co novae in the two samples .
thus we can exclude highly optimistic estimates of the ejecta masses , exceeding 12@xmath74 , for two members of this class of nova . given
that cp cru is the only one event in the tgrs sample , with a limit on the flux 40 times the expected value , the same reasoning produces an upper limit @xmath75 for this class of event , improving on the _ smm _ limit by almost a factor 10 . from our result for the general case
, we also exclude any undiscovered nova with optimistic `` clayton - like '' parameters having occurred within 1.1 kpc during 19951997 ; or any undiscovered nova of the highest - yielding gmez - gomar et al .
( 1998 ) type ( co , @xmath76 ) within 134 pc .
this work has established the feasibility of the time series modeling method of background subtraction ( 4 ) , at least for resolvable lines ( not too severely blended ) , for @xmath0-ray emitters having appropriate time - scales and especially for stable background lines amenable to very simple temporal models .
an obvious improvement would be to model blended line complexes of astronomical interest , which would require much more sophisticated modeling than our semi - empirical approach .
thus , work is under way to model the tgrs background spectrum ab initio , from the mass model and the gcr flux ( weidenspointner 2001c ) .
some of the features that make this approach possible are included in the upcoming _ integral _ mission , in particular a high - resolution ge spectrometer ( spi ) , and a highly elliptical orbit which , like that of tgrs , avoids earth s trapped radiation belts ( winkler 1996 ) . if the background lines prove to be stable enough , it may be possible to apply our method to the off - source spi background .
relative to tgrs , spi has much greater sensitivity , close to @xmath77 photon @xmath78 s@xmath15 at 478 kev .
the best - case theoretical models in table 2 ( co , @xmath76 ) are then detectable at a distance of 500 pc ( hernanz & jos 2000 )
. the more optimistic parameters @xmath79 , @xmath72 , similar to those of clayton ( 1981 ) , that we discussed in 6.1 and assumed in column 5 of table 2 for the purpose of flux comparisons , would enable spi to make detections out to 3.4 kpc .
a detection distance of this order intercepts @xmath80% of the global nova distribution ( paper ii ) , two - thirds of which are of co subtype .
if _ integral _ continues beyond its 2-year nominal mission to the 5-year extended mission , the number of novae it can be expected to detect is @xmath81 , which is @xmath82 for realistic values of @xmath83 yr@xmath15 .
the prospects for a spi observation improving on our results are therefore quite good .
we thank albert jones and frank bateson of the royal new zealand astronomical society for prediscovery limits on some of the novae in table 1 .
we are grateful to theresa sheets ( lhea ) and sandhia bansal ( hstx ) for their work on the analysis software .
acua , m. h. , ogilvie , k. w. , baker , d. n. , curtis , s. a. , fairfield , d. h. , & mish , w. h. 1995 , space sci .
, 71 , 5 boffin , h. m. j. , paulus , g. , arnould , m. , & mowlavi , n. 1996 , a&a , 275 , 96 cass , m. , vangioni - flam , e. , & audouze , j. 2001 , to appear in proc .
conf . cosmic evolution ,
r. ferlet , m. lemoine , & e. vangioni - flam ( world scientific ) , and at ( _ astro - ph_/0101323 ) clayton , d. d. 1981 , apj , 244 , l97 della valle , m. 1996 , iau circ .
6532 downes , r. a. , & duerbeck , h. w. 2000 , aj , 120 , 2007 evans , a. , callus , c. m. , whitelock , p. a. , & laney , d. , 1990 , mnras , 246 , 527 gehrz , r. d. , truran , j. w. , williams , r. e. , & starrfield , s. 1998 , pasp , 110 , 3 gmez - gomar , j. , hernanz , m. , jos , j. & isern , j. 1998 , mnras , 296 , 913 harris , m. j. , leising , m. d. , & share , g. h. 1991 , apj , 375 , 216 harris , m. j. , teegarden , b. j. , cline , t. l. , gehrels , n. , palmer , d. m. , ramaty , r. , & seifert , h. 1998 , apj , 501 , l55 harris , m. j. , naya , j. e. , teegarden , b. j. , cline , t. l. , gehrels , n. , palmer , d. m. , ramaty , r. , & seifert , h. 1999 , apj , 522 , 424 ( paper i ) .
harris , m. j. , teegarden , b. j. , cline , t. l. , gehrels , n. , palmer , d. m. , ramaty , r. , & seifert , h. 2000 , apj , 542 , 1057 ( paper ii ) hatano , k. , branch , d. , fisher , a. , & starrfield , s. 1997 , mnras , 290 , 113 hernanz , m. , & jos , j. 2000 , in aip conf .
522 , cosmic explosions , ed .
s. holt & w. w. zhang , ( new york : aip ) , 339 hernanz , m. , smith , d. m. , fishman , j. , harmon , a. , gomz - gomar , j. , jos , j. , isern , j. , & jean , p. 2000 , in aip conf .
510 , the fifth compton symposium , ed .
m. l. mcconnell & j. m. ryan ( new york : aip ) , 82 hix , w. r. , smith , m. s. , mezzacappa , a. , starrfield , s. , & smith , d. l. 2000 , in aip conf .
522 , cosmic explosions , ed .
s. holt & w. w. zhang ( new york : aip ) , 383 iyudin , a. f. , et al .
1995 , a&a , 300 , 422 iyudin , a. f. , schnfelder , v. , strong , a. w. , diehl , r. , lichti , g. g. , bennett , k. , & ryan , j. 2001 , proc .
gamma 2001 high - energy astrophysics symposium , ed .
n. gehrels , c. r. shrader , & s. ritz ( new york : aip ) , in press johnson , w. n. , et al .
1993 , apjs , 86 , 693 kurczynski , p. , et al . 1999 , nucl . instr .
methods phys .
res . a , 431 , 141 leising , m. d. 1988 , in aip conf .
170 , nuclear spectroscopy of astrophysical sources , ed .
n. gehrels & g. h. share ( new york : aip ) , 130 leising , m. d. , share , g. h. , chupp , e. l. , & kanbach , g. 1988 , apj , 328 , 755 ling , j. c. , et al .
2000 , apjs , 127 , 79 livio , m. , mastichiadis , a. , gelman , h. , & truran , j. w. 1992 , 394 , 217 mahoney , w. a. , ling , j. c. , jacobson , a. s. , & lingenfelter , r. e. 1982 , apj , 262 , 742 milne , p. a. , kurfess , j. d. , kinzer , r. l. , leising , m. d. , & dixon , d. d. 2001 , proc .
gamma 2001 high - energy astrophysics symposium , ed .
n. gehrels , c. r. shrader , & s. ritz ( new york : aip ) , in press morel , p. , pichon , b. , provost , j. , & berthomieu , g. 1999 , a&a , 350 , 275 naya , j. e. , jean , p. , gehrels , n. , slassi - sennou , s. , teegarden , b. j. , tueller , j. , vedrenne , g. , & von ballmoos , p. 1996 , proc .
spie , 2806 , 472 owens , a. , et al .
1995 , space sci .
, 71 , 273 palmer , d. m. , et al .
1996 , in aip conf .
384 , gamma - ray bursts : 3rd huntsville symposium , ed . c. kouveliotou , m. f. briggs , & g. j. fishman ( new york : aip ) , 218 phillips , g. w. , & marlow , k. w. 1976 , nucl .
methods , 137 , 525 schwarz , g. j. 1999 , phd thesis , arizona state university , at http://brian.la.asu.edu/@xmath84schwarz/main/main.html seifert , h. , cline , t. l. , palmer , d. m. , ramaty , r. , teegarden , b. j. , hurley , k. , madden , n. w. , & pehl , r. h. 1998 , in aip conf .
428 , gamma - ray bursts : 4th huntsville symposium , ed . c. a. meegan , r. d. preece , & t. m. koshut ( new york : aip ) , 339 shafter , a. w. 1997 , apj , 487 , 226 shore , s. n. , starrfield , s. , & sonneborn , g. 1996 , apj , 463 , l21 starrfield , s. , truran , j. w. , sparks , w. m. , & arnould , m. 1978 , apj , 222 , 600 starrfield , s. , shore , s. n. , sparks , w. m. , sonneborn , g. , truran , j. w. , & politano , m. 1992 , apj , 391 , l71 starrfield , s. , shore , s. n. , sonneborn , g. , gonzalez - riestra , r. , sparks , w. m. , truran , j. w. , dopita , m. a. , & williams , r. e. 1993 , in aip conf .
280 , compton gamma - ray observatory , ed .
m. friedlander , n. gehrels & d. j. macomb ( new york : aip ) , 168 starrfield , s. , truran , j. w. , wiescher , m. c. , & sparks , w. m. 1997 , in aip conf .
410 , proceedings of the fourth compton symposium , ed . c. d. dermer , m. s. strickman , & j. d. kurfess ( new york : aip ) , 1130 starrfield , s. , schwarz , g. , truran , j. w. , & sparks , w. m. 2000a , in aip conf .
522 , cosmic explosions , ed .
s. holt & w. w. zhang ( new york : aip ) , 379 starrfield , s. , sparks , w. m. , truran , j. w. , & wiescher , m. 2000b , apjs , 127 , 485 stickland , d. j. , penn , c. j. , seaton , m. j. , snijders , m. a. j. , & storey , p. j. 1981 , mnras , 197 , 107 vedrenne , g. , et al . 1999 , apl&c , 39 , 793 warner , b. 1995 , cataclysmic variable stars , ( cup : cambridge ) , 257299 weidenspointner , g. , et al .
2001a , a&a , 368 , 347 weidenspointner , g. , harris , m. j. , jean , p. , & diallo , n. 2001b , presented at astronomy of radioactivities iii workshop , ringberg , germany , 2123 may 2001 , mpe report , in press weidenspointner , g. , harris , m. j. , sturner , s. j. , dean , a. j. , diallo , n. , & shaw , s. e. 2001c , exploring the gamma - ray universe ( proc .
4th integral workshop ) , in press winkler , c. 1996 , a&as , 120 , 637 lcccccc nova & white & possible date & @xmath19 at & @xmath18 , & distance , & distance , + & dwarf & of maximum & discovery & days & eq . ( 13 ) , pc@xmath85 & @xmath27 , pc@xmath86 + by cir & co & 1995 jan 21.0 jan 27.3 & 7.2 & @xmath87 & 32404000 & 23703030 + v888 cen & & 1995 feb 22.0 feb 23.3 & 7.2 & @xmath73 & 60606970 & 24503720 + v4361 sgr & & 1996 jun 19.6 jul 11.5 & 10.0 & 53 & 46406785 & 507010280 + cp cru & one & 1996 aug 22.0 aug 27.0 & 9.25 & 5.2 & 677016300 & 31007500 + v1141 sco & & 1997 jun 2.1 jun 5.1 & 8.5 & 4.5 & 68308340 & 24606860 + lcccccc target & distance & zenith & flux @xmath86 & expected @xmath88 & implied @xmath1be mass + & pc@xmath85 & angle & @xmath0 @xmath78 s@xmath15 & @xmath0 @xmath78 s@xmath15 & @xmath25 per nova@xmath86 + & & & & & + undiscovered nova & & @xmath51 & @xmath89 & & + by cir & 3160 & 45@xmath90 & @xmath91 & @xmath92 & @xmath93 + v888 cen & 4800 & @xmath94 & @xmath95 & @xmath96 & @xmath97 + v4361 sgr & 6700 & @xmath98 & @xmath99 & @xmath100 & @xmath101 + cp cru & 3180 @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 + nova sco & 6120 & @xmath107 & @xmath108 & @xmath109 & @xmath110 + v1370 aql @xmath111 & 3500 & & @xmath112 & @xmath113 & @xmath114 + qu vul @xmath111 & 3000 & & @xmath115 & @xmath116 & @xmath117 + v842 cen @xmath111 & 1100 & & @xmath118 & @xmath119 & @xmath120 + + _ gc integrated _ & & & & & + tgrs & 8000 & @xmath121 & @xmath122 & @xmath123 & @xmath124 @xmath125 + _ smm _ & 8000 & & @xmath126 & @xmath127 & @xmath128 @xmath125 + + _ theory _
@xmath129 & & & & & + co @xmath130 & 1000 & & @xmath113 & & @xmath131 + co @xmath76 & 1000 & & @xmath116 & & @xmath132 + one @xmath76 & 1000 & & @xmath133 & & @xmath134 + one @xmath135 & 1000 & & @xmath110 & & @xmath136 + + @xmath85 simple mean of estimates in columns 6 and 7 of table 1 , unless otherwise noted .
+ @xmath86 @xmath137 upper limit .
+ @xmath88 assumes `` optimistic '' parameters ( see 6 ) of @xmath138 of ejecta ( as observed : warner 1995 ) enriched to mass fractions @xmath139 for co subclass and @xmath140 for one subclass .
+ @xmath102 from expansion parallax measurement ( downes & duerbeck 2000 ) .
+ @xmath111 measured by _ smm _
( harris et al . 1991 ) .
subclasses : one ( v1370 aql 1982 and qu vul 1984 ) , co ( v842 cen 1986 ) .
+ @xmath125 the effective apertures of tgrs and _ smm _ will contain respectively @xmath141 and @xmath142 novae at any one time , where @xmath63 is the total galactic nova rate per year .
the total @xmath1be masses inferred from the flux upper limits have been divided by these factors to get the yield per nova .
+ @xmath129 hernanz & jos ( 2000 ) and gmez - gomar et al .
( 1998 ) . in column 2 1
kpc is an arbitrary choice for the purpose of line flux comparisons . in column 4
the ejecta masses are taken to be the actual model values 16 @xmath143 , not the `` optimistic '' value @xmath138 assumed elsewhere ( see note _ | the _ wind _ spacecraft carrying the transient gamma ray spectrometer ( tgrs ) moves in an extremely elliptical orbit that largely avoids earth s trapped radiation belts and albedo @xmath0-radiation .
the tgrs therefore enjoys a relatively low level of background that is also extremely stable .
we show how this stability enables modeling of the time variability of background lines , which in turn enables a novel technique of background subtraction to be used in the detection of transient astrophysical lines .
we apply a simple version of this method to the line at 478 kev that is expected to arise from nucleosynthesis of @xmath1be in nearby novae .
this search covers the entire southern ecliptic hemisphere during 19951997 , including five known individual events , and possible undiscovered individual events .
the tgrs design also uses _ wind _ s 3 s rotation period to modulate signals from the galactic center ( gc ) .
we use this feature of the instrument to search for a quasi - constant level of 478 kev emission from the accumulation of @xmath1be from several novae that are expected to occur in the direction of the gc during that isotope s 53 d half - life .
we derive upper limits on the transient ( single - nova ) emission which improve on previous limits by about an order of magnitude , and limits on the steady ( many - nova ) emission which represent a factor 2 improvement .
only weak limits can be placed on the key parameters in the nucleosynthesis and ejection of @xmath1be , however . |
we now believe in quantum chromodynamics ( qcd ) as _ the _ theory of strong interaction . yet , the centauro and anti - centauro events that were found in cosmic ray experiments @xcite indicate some puzzling features .
the events are characterized by large fluctuations in the ratio of number of neutral pions to that of charged pions .
it appeared to be unlikely that such feature can be realized inside the conventional picture of hadronic interactions with `` soft '' qcd interaction , which implies more or less an independent emission of particles .
however , several authors formulated a scenario which explains such features of the events based on the idea of formation of chirally misaligned domains , the disoriented chiral condensate ( dcc ) @xcite .
it requires that a `` hot '' matter forms during hadronic collisions in which the chiral symmetry is restored , and then it rapidly cools down so that chiral orientation of the pion fields can align to a random direction different from that of the vacuum .
if the low momentum mode of the pion fields are enhanced large domains of definite ( but random in direction ) isospin would form . then , a large isospin , and therefore , charge fluctuations naturally results .
the amplification of low momentum pion modes was demonstrated by rajagopal and wilczek @xcite by doing numerical simulation of the linear sigma model using the quench initial condition .
a more explicit confirmation of the formation of large domains was provided by asakawa , huang and wang @xcite who computed pion - field correlation functions by using an elaborated simulation code which takes into account expansion along the longitudinal as well as the transverse direction .
one should note , however , that the large charge - neutral fluctuation does not necessarily imply a random isotropic ( in isospin space ) rolling down from the top of the mexican hat potential . as was shown by greiner , gong and mller @xcite any isosinglet multipion states approximately have the well known neutral fraction distribution @xmath1 @xmath2 where @xmath3 we note that this distribution was in fact derived by only requiring the isospin singlet phase space @xcite .
therefore , the large charge - neutral fluctuation may merely imply isospin singlet dcc state , not a random isotropic rolling down .
the interpretation of the results of the simulation by rajagopal and wilczek @xcite also requires reexamination .
they interpreted the origin of the amplification of the low momentum pion mode as due to instability of the nambu - goldstone modes of chirally symmetric field configurations at around the top of the mexican hat .
however , the enhancement of the low momentum pion modes actually take place with much longer time scale than that of rolling down motion , @xmath4 , as one can clearly see in fig . 1 in their paper .
then , we need to identify certain mechanism which is responsible for the enhancement .
the most likely candidate , to our opinion , is the parametric resonance mechanism @xcite .
the mechanism , possibly being tied up with a new instability that arises within the approximation scheme we will take , seems to be capable of explaining qualitative features of long lasting amplification of low momentum modes .
( see sections 3 and 6 . )
some other aspects of formation mechanism of dcc were explored in refs .
@xcite . in a previous paper @xcite
, we briefly explored the parametric resonance mechanism emphasizing its characteristic features in the two pion correlations , and its possible relevance as a hunting tool of dcc . in this paper
we examine the mechanism in detail , hoping that it will shed some light to the origin of enhancement of the low momentum pion modes .
we focus on the pion production at around the bottom of the mexican hat potential
. it may be an appropriate setting at least for the late stage of non - equilibrium phase transition which would result in the formation of coherent dcc domains .
furthermore , we concentrate on the case that the sigma model fields oscillate along the @xmath0 direction .
in fact , it is shown in the numerical simulation @xcite that sigma model field configurations rapidly point to the @xmath0 direction in a time scale of the order of @xmath5 . under the background oscillation of the @xmath0 field
the quantum fluctuations of the pion and the @xmath0 fields can be excited .
if the frequencies of background and quantum fluctuations match with each other the fluctuations are parametrically amplified , leading to an explosive pion production .
we emphasize that the mechanism of pion production from dcc which we explore in this paper differs from that is usually adopted in the literature @xcite as we mentioned earlier . in the conventional picture of pion emission from dcc , one assumes that the sigma model fields roll down along the direction ( in most cases ) different from the @xmath0 direction .
then , the field configurations relax toward the @xmath0 direction , the orientation of the qcd vacuum , which entails the coherent pion emission through the relaxation process .
in contrast , pion production takes place even when the sigma model fields rolls down along the @xmath0 direction in our parametric resonance mechanism . to indicate this point ,
we propose in the next section a set of approximations that leads to a simplified but a concrete model field theory that allows us to construct explicitly the produced multi - pion quantum state .
we should remark that we do not claim that the parametric resonance is the whole story .
formation and decay of dcc is a complicated nonlinear phenomenon .
the instability associated with the nambu - goldstone modes is indispensable ingredient for triggering the growth of the low momentum pion modes .
but , we also stress that it is _ not _ enough either . in sec
vi we will calculate the power spectrum of pion and sigma fields in our framework and attempt a qualitative comparison with the result obtained in rajagopal - wilczek s simulation .
it will give us a feeling on the question of to what extent the amplification of power is due to the parametric resonance mechanism . even though the parametric resonance mechanism is the cause of the long lasting amplification of low momentum modes we still lack understanding how the initial instability is mediated to it @xcite .
leaving understanding of this last point to the future investigations our emphasis in this paper is that if the parametric resonance mechanism is operative at the late stage of the non - equilibrium phase transition it may give clear signatures which should be useful for its experimental hunting . in sec .
ii , we introduce the linear sigma model , formulate the system with parametric resonance by introducing successive approximations in various stages , and make clear that to what extent the approximations are valid . in sec .
iii , we quantize the system in a conventional way of quantizing a system of quantum fields , and clarify the relationship with the other formulation . in sec .
iv , we calculate the single particle distributions of pions and @xmath0 to demonstrate a characteristic feature of the parametric resonance mechanism . in sec .
v we discuss the two pion correlations which display another characteristic feature of the mechanism .
we argue that they can be used as powerful experimental hunting tool of dcc . in sec .
vi we compute the power spectrum of pion and sigma fields and examine the qualitative features of the amplification of power in our framework . in the last section
we will summarize our investigation , give concluding remarks , and discuss the limitations of our framework . in appendix
a , we summarize the feature of multiparticle states implied by the single mode squeezed state . in appendix b
, we derive sum rule obeyed by isospin invariance and the iso - singlet nature of the multiparticle state within the framework of single mode treatment .
the lagrangian of the linear sigma model is given by = _ _ a ^_a - ( _ a_a - v_0 ^ 2)^2 + h. where @xmath6 .
the typical values of the parameters which are relevant for phenomenology of qcd are : = 20 , v_0 = 90 , m_= = 140 .
these values of the parameters will be used in the numerical analysis in later sections . as discussed in the previous section we deal with the pion production that takes place when the sigma model fields are at around the bottom of the mexican hat potential , which may be realized in the time scale of @xmath5 after rolling down .
we assume that at this stage the sigma model fields oscillate along the @xmath0 direction . for clarity
we construct a further simplified model by doing successive approximations in the linear sigma model that allow us to explicitly construct the produced multi - pion quantum state .
we expand the sigma model fields at around the minimum of the mexican hat potential along the @xmath0 direction .
l = - v , + + = .
the linear sigma model lagrangian , then , takes the form @xmath7 = & { \displaystyle\frac{1}{2}(\partial_{\mu } \chi)^2 - { \lambda\over 2}(3v^2- v_0 ^ 2)\chi^2 - \lambda v \chi^3 - \frac{\lambda}{4}\chi^4 + \frac{1}{2}(\partial_{\mu}\vec{\pi})^2}\\ & \\ & -\displaystyle \frac{1}{2}m_{\pi}^2\vec{\pi}^2 { \displaystyle- \lambda v \chi\vec{\pi}^2 - { 1\over 2}\lambda \chi^2\vec{\pi}^2 - \frac{1}{4}\lambda\vec{\pi}^4}. \end{array } \label{eqn : lagrangian}\ ] ] following mrwczyski and mller @xcite , we make a further approximation of decomposing @xmath0 model fields into the classical time - dependent background fields and the quantum fluctuations around them @xmath8 we have assumed the spatial homogeneity of the background fields .
it may be a reasonable assumption if we work inside a single domain because the key feature of dcc is its coherence over a domain . of course , it is a simplifying assumption in real physical situation , but is an inevitable one in order to make the treatment technically manageable . substituting ( [ decompo1 ] ) into lagrangian ( [ eqn : lagrangian ] ) , we have rl & l[_0+,_0+]=l[_0,_0 ] + & + & + 12()^2- 12(m_^2 + 3(t))^2 -12_0 ^ 2 ^ 2 -(_0+v)^3 - 14 ^ 4&&+12()^2 -12(m_^2+(t))^2 -32_0 ^ 2 ^
2-_0 ^ 3 - 14 ^ 4 + & & -2v_0-(_0+v)^2 -_0 ^ 2 - 12 ^ 2 ^ 2 , [ new_l ] where @xmath9 and @xmath10 is the minimum value of the @xmath0 direction determined by the equation @xmath11 . for small @xmath12
, @xmath10 is given approximately by @xmath13 .
due to the background field oscillation , a time - dependent term @xmath14 arises in the mass terms of @xmath15 and @xmath16 fluctuations and is given by @xmath17 .
the terms linearly proportional to fluctuations , of course , vanish due to the equations of motion of the background fields : @xmath18 @xmath19 we work with the ansatz that the background fields oscillate along the @xmath0 direction , and set @xmath20 in the subsequent analyses .
we further restrict ourselves into the regime , @xmath21 , in which the nonlinear terms in @xmath22 s equation of motion are negligible .
then , the solution to the @xmath22 s equation of motion takes the harmonic form @xmath23 where @xmath24 is an initial phase which we set to vanish for simplicity . in the numerical analysis
performed in sec.iv , we will use @xmath25 ( @xmath26 1/2 , 1 , 2 ) for illustrative purpose .
we should note that the limitation of small amplitude oscillation that we have imposed , @xmath21 , restricts the validity of our treatment only to a qualitative level .
nevertheless , we believe that it is a meaningful starting point , and in fact it makes theoretical analysis transparent as we will see below . we differ from mrwczyski and mller @xcite by treating @xmath16 and @xmath15 quantum mechanically .
the equations of motion obeyed by fluctuations @xmath16 and @xmath15 are given as @xmath27 they are greatly simplified by the restriction of small oscillation amplitudes of background fields .
let us first focus on the pion fluctuations .
the @xmath28 term is ignorable compared to @xmath29 term because of the restriction of @xmath30 .
the @xmath29 term is also small compared with the mass term , but we shall keep it otherwise we lose the parametric resonance . the cubic term of @xmath15
is negligible until the time scale @xmath31 where the dimensionless time @xmath32 is measured by @xmath33 as @xmath34 .
for instance , it can be numerically estimated as @xmath35 for @xmath36 .
we have ignored the quantum back reactions to the pionic and sigma fluctuations due to particle production .
it could be taken into account , e.g. , by the hartree - fock approximation .
instead of going through this treatment , we estimate the time scale until which it can be negligible .
it is given roughly as @xmath37 and the same numerical examination indicates that it is at @xmath38 .
we emphasize , therefore , that under the approximation of small amplitude oscillation of background fields , ignoring nonlinear terms in the field fluctuations is a good approximation in a fairly long time even for such strong coupling as @xmath39 .
we end up with the equation of motion of sigma and pionic fluctuations : @xmath40 \eta_{\vecvar k}^{(1)}(z ) = 0 , \label{eqn : fluct1}\ ] ] @xmath41 \vec\xi_{\vecvar k}(z ) = 0 , \label{eqn : fluct2}\ ] ] where @xmath42 and @xmath43 @xmath44 and @xmath45 denote the fourier components of the fields @xmath46 and @xmath47 , respectively . the equations ( [ eqn : fluct1 ] ) and ( [ eqn : fluct2 ] ) are known as the mathieu equation and are known to admit unstable solutions for a wide range of parameters of @xmath48 s and @xmath49 s .
see , e.g. , ref . @xcite . such unstable solution may be interpreted as explosive particle production under the background of oscillating @xmath22 fields .
we discuss in this section the quantum evolution of the time - dependent states of @xmath50 and @xmath0 fluctuations .
thanks to the restrictions and approximations introduced in the previous section , the structure of quantum states formed by @xmath16 and @xmath15 quanta can be analyzed analytically .
we start by writing the hamiltonian in a form analogous to the harmonic oscillators but with time - dependent frequencies : rl h = d^3 k & + & + & + _ j d^3 k , [ effectiveh ] where l _ ^(t)^2 = k^2 + m_^2 + 6v_0(t ) , + + _
^(t)^2 = k^2 + m_^2 + 2v_0(t ) .
[ def_omega ] the variable @xmath51 and @xmath52 in ( [ effectiveh ] ) are defined by ( , t ) = e^i q_(t),^j(,t ) = e^i q_^j ( t ) , [ defq_k ] and @xmath53 and @xmath54 are , as usual , the conjugate momenta of @xmath55 and @xmath52 , respectively .
the index @xmath56 runs over 1 to 3 , and we take the adjoint representation for the iso - triplet pion fields . for its concrete form , we refer appendix b. one can see that the equations of motion ( [ eqn : fluct1 ] ) and ( [ eqn : fluct2 ] ) are derived from this hamiltonian . within the small background oscillation we can take that @xmath57 .
this restriction guarantees that the frequency @xmath58 is real , which is necessary in our present treatment which ignore quantum back reactions .
if the background oscillation amplitude is larger than the critical value the system have an instability which we referred to as a new instability in introduction .
it definitely arises within our approximation and is indicative of a remnant of the rajagopal - wilczek instability because it arises only for negative @xmath22 ( which means toward central maximum of the mexican hat ) .
nevertheless , the new instability has a characteristic feature that is quite different from that of the rajagopal - wilczek s .
the former can continue for long time while the latter lasts only a time scale of @xmath59 .
notice that both of the instabilities that we are talking about are the local instabilities , not the real global ones which totally destabilize the whole system .
the nature of the instability , however , is not transparent to us .
when one works with certain truncation in highly nonlinear systems one easily generate instabilities which may or may not be possessed by the original system .
there are instabilities that one can easily understand their origin from simple consideration .
for example , pion fields can have local instability by generating the winding motion encircling around the bottom of the wine bottle when the field energy of background @xmath0 field oscillation with amplitude @xmath22 exceeds the symmetry breaking energy , that is , @xmath60 .
it leads to the condition for onset of the instability @xmath61 but the instability we just encountered above occurs at much smaller amplitude , @xmath62 . to our knowledge there is no intuitive way of understanding the instability .
it can be the artifact of our approximation and truncation scheme and , if this is the case , may be cured by taking nonlinearities and back reactions into account . in the rest of the paper
we will restrict ourselves into the small amplitude regime of the background @xmath0 oscillation so that we are free from the new instability . in sec .
vii we will briefly discuss its role together with the parametric resonance as a possible candidate mechanism for understanding the long - lasting amplification of powers in sigma model simulations . the quantum theory of harmonic oscillators with time - dependent frequencies were developed since sometime ago . in a previous paper @xcite
we followed the formalism developed by shtanov , traschen and brandenberger ( stb)@xcite , but we shall present in this paper an equivalent but more conventional formalism of quantizing the same system . from the form of the hamiltonian ( [ effectiveh ] ) , we can discuss sigma and pion sector collectively ; we treat below the sigma sector but with suppressing the superscript @xmath0 , and it can be regarded as that of pion sector in which the superscript @xmath63 as well as iso - spin index are suppressed .
we decompose the dynamical variables in momentum space in terms of solutions of the equation of motions l q_(t)=q_1(t)a_0+q_2(t)a^_0- , + + p_(t)=_1(t)a_0- + _ 2(t)a^_0 , [ def_q ] where @xmath64 and @xmath65 are the solutions to the equation of motion of @xmath66 derived from the hamiltonian , _
( t ) + ^2_(t)q_(t ) = 0 .
[ eq_for_q ] notice that @xmath67 .
@xmath68 and @xmath69 are the annihilation and the creation operators , respectively .
they are time - independent , i.e. , they obey the heisenberg equation of motion as + i[h , a_0k]=0 . the consistency condition between the commutation relations , i.e. , = 1,=i ( ) .
is achieved by the wronskian condition q_1_2-_1q_2=i .
[ wronskian ] substituting ( [ def_q ] ) into ( [ effectiveh ] ) , we obtain the following form @xmath70 , % \end{array } \label{effh2}\ ] ] where the coefficients @xmath71 and @xmath72 are defined by @xmath73 note that @xmath71 is complex number , and that @xmath64 and @xmath65 are complex conjugate with each other . on the other hand , @xmath72 is real as far as @xmath74 is real , which is the case owing to our restriction to small background oscillations . in order to diagonalize the hamiltonian ( [ effh2 ] )
, we introduce the bogoliubov transformation as l a_0(t)=^*_k(t)a _ -^*_k(t)a^_- , + + a^_0(t)=_(t ) a^_-_(t)a_- , [ bogoliubov1 ] with _
( t)=e^i__,_(t)=e^i__. choosing @xmath75 as 2_k = |a_k|b_k , [ thetadiag ] one can diagonalize the hamiltonian into the form h_=_a^_(t)a_(t ) , [ effectiveh2 ] where use has been made of the wronskian relation ( [ wronskian ] ) to show that the coefficient of ( [ effectiveh2 ] ) is in fact given by @xmath76 . therefore , the operator @xmath77 defined in ( [ bogoliubov1 ] ) does represent annihilation operator of physical quanta at time @xmath78 . using ( [ thetadiag ] ) with ( [ def_ab ] )
is is easy to show that ^2_k = , ^2_k = .
[ chsh ] in the above treatment , we have relied on the fact that amplitudes of the background oscillation is small so that @xmath74 is real . if we remove this assumption the system experiences instability in a certain period of time .
it is not totally unstable system due to the nature of the instability and it would be interesting to work out such system .
however , we do not engage such business in the present paper .
instead , we argue that it is important to take account of back reactions in really addressing such problem .
it would even be the case that the quantum back reaction entirely cures the instability .
let us move on to the analysis of the eigenstates for the effective hamiltonian .
we define the vacua @xmath79 and @xmath80 as a_0|0 = 0,a_(t ) |0(t ) = 0 , [ def_vacuum ] and assume that the oscillator at @xmath81 is in the vacuum state @xmath79 , i.e. , @xmath82 .
it leads to the initial conditions for @xmath83 and @xmath84 , @xmath85 the physical quantum with momentum @xmath86 is created ( annihilated ) by @xmath87 ( @xmath77 ) , because the hamiltonian is diagonalized by them at each time @xmath78 .
as the system evolves , the state @xmath88 is no longer the vacuum annihilated by @xmath77 .
we rewrite the bogoliubov transformation ( [ bogoliubov1 ] ) into the form of unitary transformation : l a_0= e^-i_e^-g_(t)a_(t ) e^g_(t ) , + + a_0^= e^i_e^-g_(t)a^_(t ) e^g_(t ) , [ bogoliubov2 ] where g_(t)=_(a_^a_-^-a_a_- ) .
[ def_g ] from the definitions of vacua ( [ def_vacuum ] ) and of transformation ( [ bogoliubov2 ] ) , we have vacuum state of the form @xmath89 |0(t ) \rangle . \label{squeezed}\ ] ] the state is known as the squeezed state and widely used in quantum optics @xcite .
a significant feature of ( [ squeezed ] ) is that it has pairing correlation between back - to - back , namely @xmath86 and @xmath90 , modes , and we will refer to it as the squeezed state of bcs type in this paper . in the next two sections , we will discuss , in detail , what are the characteristic features of particle production due to the parametric resonance mechanism .
earlier discussions of the squeezed state in the context of dcc may be found in @xcite .
we also note that the particle production via the parametric resonance has been extensively discussed in the context of reheating in the inflationary universe @xcite-@xcite , and a mechanism for producing superheavy dark matter @xcite . before closing this section
, we would like to comment on the relationship between our formalism and the one by shtanov , traschen , and brandenberger @xcite which can provide description of quantum systems with parametric resonances .
since we have adopted their formalism in our previous paper , we will try to compare and clarify the relation between various quantities that appear in both formalism . in short
, there is a simple relationship between two formalisms : they start from the time - dependent operator ( our @xmath91 ) which diagonalizes the hamiltonian and then bogoliubov transform it to the time - independent operator @xmath92 .
therefore , such a correspondence is likely to exists . in order to make the correspondence more precise , we write down the equations obeyed by the coefficients of the bogoliubov transformation in both formalisms . in stb formalism
, they read @xmath93 on the other hand , the equations of @xmath94 and @xmath95 in our formalism are given by @xmath96,\\ & \\ \dot{\beta}_{\vecvar
k}(t ) & \displaystyle = { \dot{\omega}_{\vecvar k}\over { 2\omega_{\vecvar k } } } \alpha_{\vecvar k}\cdot { e^{-2i\varphi_\alpha}\over{\dot{\omega}_{\vecvar k}a_{\vecvar k}^ * } } \biggl[2i\dot{\varphi}_\beta(b_{\vecvar k}-\omega_{\vecvar k } ) \omega_{\vecvar k}+\dot{b_{\vecvar k}}\omega_{\vecvar k } -b_{\vecvar k}\dot{\omega}_{\vecvar k}\biggr].\\ \end{array}\label{eq_for_coefficient2}\ ] ] above two sets of equations are consistent with each other provided that the phases of the coefficients satisfy the transformation rule : @xmath97 namely , the phase parts , which are undetermined in our present formalism , are chosen in a particular way in the stb formalism .
notice that the right - hand - sides of ( [ consistency ] ) are pure phases and hence they are consistent .
let us discuss the qualitative features of the single particle distributions of particles produced to explore the parametric resonance mechanism .
the multi pion and sigma state in our mechanism is given by @xmath98 where @xmath79 stands for the bcs type squeezed state ( [ squeezed ] ) .
the factorized form in ( [ ketpsi ] ) stems from the linear approximations described in sec .
the average number of quanta @xmath99 with momentum @xmath86 at time @xmath78 is given by n _
( t ) = 0|a_^(t)a_(t)|0 .
[ one - point ] we suppress in this section the isospin index , since the single particle distributions is independent of isospin .
given the squeezed state ( [ squeezed ] ) , it can be readily shown to be n _ ( t ) = |_|^2 = , [ lbeta ] by using ( [ chsh ] ) with @xmath72 defined in ( [ def_ab ] ) . before presenting results of numerical computation ,
we discuss what would be the characteristic feature of the single particle distributions .
it is natural to expect that an enhancement occurs due to the parametric resonance mechanism . with physical values of the parameters , we take the @xmath49 parameters defined in ( [ def_aqsigma ] ) and ( [ def_aqpi ] ) are given by q_= 2.0 ( ) , q_= 6.0 ( ) . numerically , @xmath100 and @xmath101 are less than @xmath102 and @xmath103 , respectively , under the present approximation @xmath57 .
therefore , we are working with narrow resonance approximation .
one should keep in mind , however , that the narrow resonance approximation may not hold in the real world . at such small @xmath49 parameter ,
the resonance occurs in the narrow bands at around the discrete values of @xmath48 , @xmath104 .
the first resonance takes place for pion at @xmath105 and it correspond to @xmath106 mev .
for sigma , the first resonance takes place at @xmath107 at zero momentum .
the second resonance would be located at @xmath108 and at @xmath109 which imply @xmath110 mev and @xmath111 mev for pion and sigma , respectively .
since the parameter @xmath48 is independent of @xmath112 , we expect that the peak position is stable against varying @xmath22 , as will be demonstrated below .
we solve the mathieu equation ( [ eq_for_q ] ) subject to the boundary condition ( [ boundary ] ) , and then compute @xmath99 . in fig . 1 - 3 we plot the single particle momentum distributions of pion as a function of momentum @xmath113 and dimensionless time @xmath32 with background oscillation parameters @xmath114 = 0.025 , 0.05 , and 0.1 .
we observe that a prominent peak exists at the right position for the first resonance in every three figures indicating the parametric resonance enhancement . in fig .
3 the momentum range @xmath115 in which an exponential instability exists ( due to imaginary frequency ) is cut off . in fig.[s - amp0.05 ] , we give the single particle momentum distributions of sigma at @xmath114 = 0.05 . the first resonance , that is expected at @xmath116 ,
is barely seen in the figure .
the second resonance is invisible both in pion and sigma distributions .
we now discuss two particle distributions and correlations between pions .
the quantum pion and sigma state is summarized in ( [ ketpsi ] ) under the present approximations .
we focus on the pion distributions of various charge states . because of the factorized form of ( [ squeezed ] ) and ( [ ketpsi ] ) @xmath117-particle distributions of @xmath0
can readily be obtained from that of @xmath118 by an appropriate change of the parameters .
it is true that the interpretation of our results on sigma particles in real experiments is not very transparent .
furthermore , when the @xmath119 coupling is turned on the @xmath0 particles do affect the two pion correlations .
discussion of this point is , however , beyond the scope of this work .
two pion distributions are defined as _
^i_^j| a^i_(t ) a^i_(t ) a^j_(t ) a^j_(t ) | , because of the factorized form of ( [ ketpsi ] ) , there are no correlations between pion and sigma .
also , because of the factorized form of the squeezed state ( [ squeezed ] ) in momentum space , the nontrivial two particle correlations exist only between modes of identical momentum ( @xmath113 and @xmath113 ) , or between back - to - back momentum ( @xmath113 and @xmath120 ) configurations .
one of the most important feature of the state ( [ ketpsi ] ) is that it is the iso - singlet state .
it comes from the fact that the frequency @xmath121 is isospin singlet .
as we discuss below , it will give us nontrivial constraints on two pion correlations of various charge combinations .
at the same time it also generates the large charge - neutral fluctuations as discussed in sec .
i. in the computation of these quantities , we further recognize that the two - pion distributions at zero - momentum can _ not _ be obtained by taking the smooth limit @xmath122 of the expressions of either identical or back - to - back momentum configurations .
it is because @xmath123 and @xmath124 do not commute , whereas @xmath125 and @xmath126 do commute for @xmath127 .
therefore , we have to compute three types of the two - pion distributions separately .
we briefly describe some details of the computation of two pion distributions .
it is convenient to use adjoint representation operator @xmath128 for this purpose .
the two pion distribution of the same momentum @xmath113 goes as follows : rl _ _ & = + & + & = + + & + & = |__|^4 + |_|^2 + & + & = |__|^4 ( + 2)+ |_|^2 + & + & = |__|^4 ( + 3 + 2)+ |_|^2 + & + & = |__|^4 ( + 3|_|^2 + 2)+|_|^2 , [ twopi+ ] where the third equality stems from the relation a_k| = a^_- k| .
similarly , @xmath129 is given as = ( [ twopi+ ] ) and ( [ twopi- ] ) , we obtain two pion distributions of @xmath56-th isospin component as = = |_|^2(2|_|^2 + 1)= . via analogous ways , one can compute the following expressions of the elements of nontrivial two pion correlations .
for the same @xmath130 : l = |_|^2 , + + = |_|^4 ( i j ) , + + = |_|^2 ( 2|_|^2 + 1 ) , + + = 0 ( i j ) . for opposite @xmath130 : l = 0 , + + = |_|^4 ( i j ) , + + = |_|^2 ( 2|_|^2 + 1 ) , + + = |_|^2 ( |_|^2 + 1 ) . by using these formulae
, it is straightforward to compute the two pion distributions of definite isospin states . we transform them into the ones written by pion charge states , which are more suitable to gain insights for the experiments .
we will describe in appendix b the relationship between the adjoint representation pion operator ( with which we have worked ) and the charge - state operator , the ( slightly modified ) horn - silver representation ll a^(+)_= -(a^1 _ - ia^2 _ ) , + & + a^(-)_= ( a^1 _ + ia^2 _ ) , + & + a^(0)_= a^3 _
, [ chargestate ] where @xmath131 and @xmath132 stand for the annihilation operators of @xmath133 and @xmath134 with momentum @xmath113 , respectively .
we express the two pion distributions of various charge combinations in the form of the ratio @xmath135 of them to the single pion distributions as defined by r__1 , _ 2(p , q ) = , where @xmath136 and @xmath49 stand for @xmath137 and 0 .
they read : \(a ) identical momentum distribution : l r _ , ( + , + ) = r _ , ( 0 , 0 ) = 2 + , + + r _ , ( + , - ) = 1 .
\(b ) back - to - back momentum distribution : l r _ , -(+ , - ) = r _ , -(0 , 0 ) = 2 + , + + r _ , -(+ , + ) = 1 .
\(c ) zero - momentum distribution : l r_=0 , = 0(+ , + ) = r_=0 , = 0(+ , - ) = 2 + , + + r_=0 , = 0(0 , 0 ) = 3 + . as an another measure for two pion correlations ,
we define the correlation function in the following way : @xmath138 then , it is straightforward to obtain the following results .
\(a ) identical momentum correlations : l c(^+_,^+ _ ) = c(^0 _ , ^0 _ ) = n ^2 _ , + + c(^+ _ , ^- _ ) = 0 .
\(b ) back - to - back momentum correlations : l c(^+ _ ,
^-_- ) = c(^0 _ , ^0_- ) = n _ ( n _ + 1 ) , + + c(^+ _ , ^+_- ) = 0 .
\(c ) zero - momentum correlations : l c(^+_=0 , ^+_=0 ) = n ^2_=0 , + + c(^+_=0 , ^-_=0 ) = n _ = 0 ( n _ = 0 + 1 ) , + + c(^0_=0 , ^0_=0 ) =
n _ = 0 ( 2n _ = 0 + 1 ) . as will be discussed in appendix b , the features of the correlation functions in ( c ) can be understood partly as a consequence of the isospin singlet nature of ( [ ketpsi ] ) .
the identical and back - to - back momentum correlations are depicted in figures [ csame ] and [ copp ] . in these figures ,
the amplitude parameter is taken as @xmath139 and it is integrated over dimensionless time @xmath32 from 0 to 10 , which corresponds to the period of time from @xmath81 to 6.5 fm .
now we would like to discuss in some details the problem of the origin of the long lasting amplification of low - momentum pion modes seen in linear sigma model simulations , in particular in ref .
@xcite .
let us first try to make a rough estimate of @xmath140 in their simulation .
it is shown in fig.2 of @xcite that @xmath141 is given by a constant superposed by an oscillatory component during the time scale of about 10 - 40 in units of @xmath142 when large amplification takes place . here
@xmath142 is the lattice constant and it is taken as @xmath143 in @xcite . since the frequency of the oscillatory component reflect the pion mass not the sigma mass it can not be used to estimate the amplitude of oscillation in sigma direction .
we thus interpret ( a quarter of ) the constant component as a time and spatially averaged @xmath0 oscillation .
since the low momentum modes are dominant at later times it may be utilized to estimate the amplitude of uniform background @xmath0 oscillations in our scheme . using the constant component of @xmath144 in @xmath141
we obtain @xmath145 .
it is a small number and indicates that our approximation is not totally absurd .
but it is larger than @xmath146 , which implies that a new type of instability mentioned in sec .
iii must also be excited .
since we confine ourselves into @xmath57 in this paper , it is not possible to directly compare our parametric resonance mechanism to the rajagopal - wilczek simulation .
nevertheless , we attempt to make a bold comparison between these two by estimating the power spectrum , the corresponding quantity to the power calculated by them .
rajagopal - wilczek used the definition of power spectrum @xcite p_rw(,t)=1n^3|_n_1,n_2,n_3 e^ia ( , t)|^2,[powerrw ] where @xmath147 indicates ( one - dimensional ) box size and they used @xmath148 and @xmath149 denotes a single component of pion or @xmath0 fields at lattice cite @xmath150 .
it has the dimension of @xmath151 and hence the ordinates in their fig.1 must be understood to be plotted in units of @xmath152 .
we define the corresponding quantity , the power spectrum , in our framework .
the natural definition is : p_ours(,t)=|q_(t)|^2 .
[ powerhm ] the expectation value @xmath153 is to be evaluated by using the normalized squeezed state ( [ squeezed ] ) . when the fields are confined into a finite volume @xmath154 , fourier integral in ( [ defq_k ] ) is modified to the discrete summation of the form ( , t)=1_ke^-i q_(t ) , or inversely , q_(t)=1d^3xe^i ( , t ) .
then , using @xmath155 , our power spectrum ( [ powerhm ] ) is related with theirs via the following relation : p_rw(,t)=p_ours(,t)/a^3 .
[ power_relation ] it should be noticed that the number presented in @xcite is @xmath156 .
therefore , we must plot @xmath157 to compare our power with theirs .
we present our power spectra of pions and sigmas in fig.7 - 10 .
we have adjusted some parameters so that they accord with the values adopted in @xcite ; @xmath158mev , @xmath159mev , @xmath160mev , though its effect is tiny and does not affect our conclusions .
we employ the amplitude parameter of background oscillation as @xmath139 in order to avoid the instability
. in fig.7 plotted is the power spectrum of pions at @xmath81 .
one can see that there is a lot of power at relatively low momentum modes .
one may feel curious of the fact that the power is not small at @xmath81 at which we set the boundary condition of no particle production .
one can easily verify that it is natural because the latter condition implies that @xmath161 and @xmath162 . in fig.8
we present the time evolution of power of pion field at @xmath163mev , which is the longest wavelength case depicted in fig.1 in @xcite and it is where the largest amplification occurs .
we observe that the average power is of order unity as we expect from the above estimation .
it is a reasonable result because it is off resonance . with these parameters
, the first resonance appears in @xmath164mev and the amplification of the power in this momentum bin is depicted in fig.9 .
the @xmath165 is growing by a factor of 50 and reaches to @xmath166 which is similar order of magnitude with rajagopal - wilczek s . of course
, two numbers can not be compared directly because of the difference involved between two computations including the initial conditions .
nevertheless , the qualitative feature of the growing power to the same order of magnitude may be an indication that the parametric resonance mechanism plays an important role in the long lasting amplification of low momentum modes .
one may argue that our results and rajagopal - wilczek s are qualitatively different because the amplification occurs in low momentum modes in their simulation , whereas it happens at resonance in our case .
however , if we have not ignored the nonlinearity it might have mediated the enhancement at resonance to other modes . in fig.10 plotted
is the time evolution of power of sigma field at @xmath163mev as in fig.8 .
the power of sigmas is an order of magnitude smaller than that of pions , in agreement with the feature obtained in rajagopal - wilczek s simulation .
we now summarize our present understanding on possible origin of the long lasting amplification of low - momentum pion modes seen in the linear sigma model simulations .
we have suggested that the parametric resonance might be the cause .
the semi quantitative features of the power spectrum including its order of magnitude and the rate of growth is not inconsistent with our proposal .
we also pointed out that a new type of instability exists in our approximation scheme which , if real , would enhance the amplification of low - momentum modes . we , however , failed to achieve an intuitive understanding of the physical origin of the new instability .
we have discussed the pion production via the parametric resonance mechanism within the linear sigma model . in particular , stimulated by the feature of the numerical simulations of the linear sigma model , we focused on the scenario in which classical background oscillations of the sigma model fields are in the @xmath0 direction . assuming small amplitude oscillation which may be natural in the late stage of evolution of dcc , we have shown that one can ignore effects due to nonlinearity and quantum back reactions for sufficiently long time .
thanks to this fact , we were able to construct an explicit quantum pion ( and sigma ) state which allows us to calculate the two particle correlations as well as the single particle distributions . our treatment ,
though under very restrictive assumptions , may be good enough to illuminate characteristic features of the parametric resonance mechanism .
we formulated the quantum theory of the system with mathieu instability on a more conventional basis of quantum field theory and clarified the relationship between our and the formalism given by shtanov , traschen and brandenberger @xcite .
we have analyzed the computed single particle distributions and clarified the structure of narrow resonances characteristic to the parametric resonance mechanism .
we then discussed the two pion correlations as a possible experimental probe for disoriented chiral condensate .
since the two pion correlations have unique characteristics , the back - to - back ( in momentum space ) correlations , it must give a clear signature which should merit the experimental hunting of dcc .
in particular , it can not be masked by the identical particle interference , the hanbury - brown - twiss effect @xcite .
why two particle correlation ?
it is certainly far more difficult to measure compared with the multiplicity distributions , on which all the recent experimental search for dcc rely @xcite .
the global analysis using multiplicity distributions is powerful if large fractions of events are accompanied by the dcc domain formation . on the other hand ,
a different strategy is required for hunting if dcc is a rare phenomenon .
we have discussed possible origin of the long lasting amplification of low - momentum pion modes in the last section .
our discussion can not be a complete one , but we hope that it stimulates the readers interests in this problem .
we should mention about the limitation and the drawback in our treatment in this paper .
we ignored the quantum back reaction and the instability of @xmath0 meson which should exist in the real world .
it is the perfectly legitimate approximation if the amplitudes of background @xmath0 oscillations are really small .
however , it is possible that the amplitudes are sometimes large because of the prevailing thermal fluctuations in the initial stage . since we are dealing with the system in which the coupling is really strong , most probably , the peak in the single particle distributions will go away after the quantum back reaction is taken into account .
then , the key question is that if anything in our results remain valid after the quantum back reaction is taken into account .
we argue that the answer is yes ; it is the characteristic features of the two pion correlations which are discussed in detail in sec . v. we now engage in a computation to verify our expectation .
how about the @xmath119 coupling ?
it is clear that it also tends to smear out the resonance peaks and , more importantly , may obscure the signature of the back - to - back correlations .
again we need more elaborate treatment which includes the instability of @xmath0 to make definitive statement about how much the signature survive in the case with @xmath119 coupling .
the formalisms which may be suitable for such analysis have been investigated in detail @xcite .
there are also some recent proposals @xcite that the hadronic medium effects may induce the similar back - to - back correlations in momentum space , which would mimic the signature of dcc discussed in this paper .
one would hope that it should be possible to find observational features which discriminate these two mechanisms .
the task is , however , beyond the scope of this paper . during the long term investigation we benefited from numerous conversations in various stages of this work with masayuki asakawa , dan boyanovsky , berndt mller , krishna rajagopal , and valery rubakov .
in particular we thank krishna for his informative and kind correspondences in comparing power spectra obtained by our and their calculations .
the research of hm is partly supported by the grant - in - aid for international scientific research no .
09045036 , inter - university cooperative research , and the grant - in - aid for scientific research in priority areas no .
11127213 , ministry of education , science , sports and culture of japan .
we discuss in this appendix some aspects of multiparticle correlations among particles involved in a normalized single - mode squeezed state |= e^k_+|0 , where @xmath167 and @xmath168 , and @xmath169 denotes a complex number .
the operators @xmath170 and @xmath171 , together with @xmath172 , form a closed algebra , = n + , = 2k _ , which will play a role in our following computation . for convenience in the systematic treatment , we employ the generating function formalism developed by koba , nielson , and olesen , @xcite and by koba @xcite .
the generating function is defined by f[h ] = _ n=0^ ( 1+h)^n p_n , using the multiplicity distribution @xmath173 where |n= | 0 .
the generating function can generate whole set of various moments when expanded in various different manner : ll f[h ] & = + + & = . in these equations , @xmath174 denotes f^(k ) = n(n-1)(n-2)(n - k+1 ) , where @xmath175 implies the average over by the multiplicity distribution @xmath176 ; @xmath177 . as is familiar in cluster expansion in statistical mechanics
, @xmath178 represents the correlations ll c^(1 ) & = n , + & + c^(2 ) & = n(n-1 ) - n ^2 , + & + c^(3 ) & = n(n-1)(n-2 ) -3 n(n-1 ) n - n ^3 , and so on .
the expression of the generating function in a form of operator expectation value is give by koba @xcite ; f[h ] = | : e^h a^ a : | , where @xmath179 indicates to take the normal ordering .
it proves to be a very useful formula for our purpose . toward calculating @xmath180
$ ] , we define the quantity @xmath181 as @xmath182 . using the algebra applied to the vacuum state ll |0 & = a^m-1a^e^k_+ |0 , + & + 0| & = ^ * 0 | e^^*k_- a(a^)^m-1 , one can derive the recursion relation among @xmath181 ; ll a_0 & = 1 , + & + a_1 & = n , + & + a_m+2 & = n ( m0 ) . [ recursion ] one can show , by direct computation , that ll n & = | a^a | + & + & = . we can convert the recursion relation ( [ recursion ] ) into the differential equation obeyed by the generating function f[h ] = 0 .
we make a change of variable t= ( 1+h)^2 , to transform the differential equation into the standard form f[t ] = 0 .
the solution to the equation subject to the boundary conditions ll f[t]_t= & = 1 , + & + f[t]_t= & = , + is uniquely given by f[t]= 1 .
the boundary conditions correspond to @xmath183 , @xmath184 , and @xmath185 .
thus , we have obtained the generating function as f[h ] = .[gene ] the correlation moments can be easily computed as follows : ll c^(1 ) & = n , + & + c^(2 ) & = n ( 2n + 1 ) , + & + c^(3 ) & = 2n ^2 ( 4n + 1 ) , + & + c^(4 ) & = 48n ^2 ( n ^2 + n + ) . + since the generating function ( [ gene ] ) is the function of @xmath186 , the multiplicity distribution @xmath176 for odd @xmath117 vanishes . that for even @xmath117 is given by p_2k = ( )
^k , [ binomial ] which is nothing but the negative binomial distribution familiar in hadronic multiparticle phenomenology . the expression ( [ binomial ] )
reproduces the results obtained by yoshimura @xcite who also calculated all the off - diagonal elements of the density matrix . for completeness
, we calculate the kno scaling function for the multiplicity distribution ( [ binomial ] ) .
it is obvious , from the expression of the correlation moments , that the squeezed state has `` long range '' correlations _
n , the multiplicity distribution obeys the kno scaling behavior .
either by solving the moment problem or by using the explicit form ( [ binomial ] ) , one can show that ll ( z ) & = _ n n p_n + & + & = e^-z/2
we derive relations among two - pion correlation functions of various different charge states which follow from isospin invariance .
we restrict ourselves into the case with iso - singlet state which is of concern to us .
we also confine ourselves into the case of zero momentum pions ; the treatment for nonzero momentum pions is different and is much more involved .
we work with the ( slightly modified ) horn - silver @xcite representation of the isospin generator = a^i^ _ ija^j , ( i , j = + , -,0 ) where _ 1 = , _ 2 = , _ 2 = .
the explicit form of @xmath187 is as follows : where @xmath188 and @xmath189 are the annihilation operators of @xmath190 and @xmath191 , respectively , as defined in ( [ chargestate ] ) .
the relation between the charge eigenstate operators and the adjoint representation operators are given in ( [ chargestate ] ) .
let @xmath192 be the iso - singlet state .
then , it is annihilated by the isospin operator , @xmath193 . from this , it follows that | e^it a^k^a^l| = | a^k^ a^l | , for any @xmath194 .
this is the relation that should be obeyed for all orders in @xmath195 .
if we take the first - order term in @xmath195 , we obtain the sum rule |t a^k^ a^l | = 0 ( k , l ) . due to isospin invariance , it is easy to show that @xmath196 . then , the @xmath197 ( or @xmath198 ) sum rule produces the relations like ^+^0 = ^-^0 , which also follows from the @xmath199 sum rule .
the useful relations comes from @xmath199-sum rule ^+^+ = ^+^- = ^-^- . in these equations
, @xmath200 implies the two - pion distribution of charge state @xmath201 and @xmath202 .
then , we obtain the isospin sum rule , c(^+ , ^- ) = c(^+ , ^+ ) + , where @xmath203 .
since @xmath204 is positive , the inequality among two - pion correlations follows ; c(^+ , ^- ) > c(^+ , ^+ ) it may be surprising that the unlike - sign correlation is stronger than like - sign correlation .
but it is the consequence of the isospin invariance with iso - singlet nature of the multipion state .
y. fujimoto , c. m. g. lattes , and s. hasegawa , _ phys .
rep _ * 65 * 151 ( 1980 ) ; l. t. baradzei et al . , _ nucl . phys . _
* b370 * ( 1992 ) 365 ; j. lord and j. iwai , paper submitted to international conference on high energy physics , dallas , texas , 1992 .
early discussions on dcc may be found e.g. , in : a. a. anselm , _ phys .
lett . _ * b217 * 169 ( 1989 ) ; a. a. anselm and m. g. ryskin , _ phys .
lett . _ * b266 * 482 ( 1991 ) ; j .-
blaizot and a. krzywicki , _ phys .
* d46 * 246 ( 1992 ) ; _ phys .
_ * d50 * 442 ( 1994 ) .
j. d. bjorken , _ acta phys .
pol . _ * b23 * 561 ( 1992 ) ; j. d. bjorken , _ acta phys .
_ * b28 * 2773 ( 1997 ) ; j. d. bjorken , k. l. kowalski , and c. c. taylor , talk at 7th rencontres de physique de la valee daoste , _ la thuile rencontres 1993 _ , page 507 .
k. rajagopal and f. wilczek , _ nucl .
* b404 * 577 ( 1993 ) ; k. rajagopal , talk at 25th international workshop on gross properties of nuclei and nuclear excitation : qcd phase transitions , january 13 - 18 , 1997 hirschegg , austria , hep - ph/9703258 . for an attempt to model the mediation mechanism ,
see h. minakata , nonequilibrium aspects of dcc s , talk at riken bnl research center workshop on `` quantum fields in @xmath205 out of equilibrium '' , october 26 - 30 , 1998 , brookhaven national laboratory , upton , new york .
d. j. h. chung , e. w. kolb and a. riotto , _ phys .
lett . _ * 81 * , 4048 ( 1998 ) ; _ phys .
rev . _ * d59 * , 023501 ( 1999 ) ; v. kuzmin and i. tkachev , _ jetp lett .
_ * 68 * , 271 ( 1998 ) ; m. arai , h. hiro - oka , n. okada , and s. sasaki , _ phys . lett _ * b449 * , 230 ( 1999 ) .
r. hanbury - brown and r. q. twiss , _ phil . mag . _ * 45 * ( 1954 ) 633 ; g. goldhaber , s. goldhaber , w. lee , and a. pais , _ phys .
_ * 120 * ( 1960 ) 300 ; e. v. shuryak , _ phys .
lett . _ * b44 * ( 1973 ) 387 ; _ sov .
phys . _ * 18 * ( 1974 ) 667 ; g. i kopylov and m. i. podgoretsky , _ sov .
* 14 * ( 1971 ) 1084 ; * 18 * ( 1973 ) 656 ; g. i kopylov , _ phys .
_ * b50 * ( 1974 ) 472 .
z. koba , many - particle inclusive cross sections and correlations in hadronic production processes at very high energy , lecture note , nbi - he-72 - 9 , july 1972 ( unpublished ? ) .
does someone know if it was published , and if yes , in which journal ? | we discuss a dynamical mechanism of pion production from disoriented chiral condensates ( dcc ) .
it leads to an explosive production of pions via the parametric resonance mechanism , which is similar to the reheating mechanism in inflationary cosmology .
classically it is related with the instability in the solutions of the mathieu equation and we explore the quantum aspects of the mechanism . we show that nonlinearities and back reactions can be ignorable for sufficiently long time under the small amplitude approximations of background @xmath0 oscillations , which may be appropriate for the late stage of nonequilibrium phase transition .
it allows us to obtain an explicit quantum state of the produced pions and @xmath0 , the squeezed state of bcs type .
single particle distributions and two - pion correlation functions are computed within these approximations .
the results obtained illuminate the characteristic features of multi - pion states produced through the parametric amplification mechanism .
in particular , two - pion correlations of various charge combinations contain back - to - back correlations which can not be masked by the identical particle interference effect .
we suggest that the parametric resonance mechanism might be a cause of the long lasting amplification of low momentum modes in linear sigma model simulations .
# 1 # 1#1 # 1a^(#1)^ #1a^(#1 ) |
we thank c. blume for his help with the na49 data .
we acknowledge the support of the german bundesministerium fr bildung und forschung ( bmbf ) , the polish state committee for scientific research ( kbn ) grant 2p03 ( 06925 ) , the national research foundation ( nrf , pretoria ) and the urc of the university of cape town .
50 m. gadzicki , ( na49 collaboration ) , j. phys .
g : nucl . part .
phys . * 30 * ( 2004 ) s701 .
afanasiev et al . , ( na49 collaboration ) , phys . rev .
* c66 * ( 2002 ) 054902 .
t. anticic et al . , phys .
* 93 * ( 2004 ) 022302 ; c. blume et al . , sqm04 proceedings , to appear in j. phys .
g. l. ahle et al . , ( e802 collaboration ) , phys . rev .
* c57 * ( 1998 ) 466 . l. ahle et al . , ( e802 collaboration ) , phys
c60 * ( 1999 ) 044904 and 064901 . l. ahle et al . , ( e866/e917 collaboration ) , phys . lett .
* b490 * ( 2000 ) 53 .
s. albergo et al . , phys .
* 88 * ( 2002 ) 062301 .
s. ahmad et al .
, phys . lett . *
b381 * ( 1996 ) 3 . j. klay et al . , ( e895 collaboration ) , phys . rev . * c68 * ( 2003 ) 054905 . m. gadzicki and m.i .
gorenstein , acta phys .
polonica b * 30 * ( 1999 ) 2705 .
r. stock , j. phys .
g : nucl . part
30 * ( 2004 ) s633 .
alford , k. rajagopal and f. wilczek , phys .
b442 * ( 1998 ) 247 ; + y.h . atta and t. ikeda ,
d * 67 * ( 2003 ) 014028 .
z. fodor and s.d .
katz , j. high energy phys .
jhep03 ( 2002 ) 014 ; + hep - lat/0402006 .
f. karsch , k. redlich and a. tawfik , phys .
b * 571 * ( 2003 ) 67 .
j. cleymans and k. redlich , phys .
* 81 * ( 1998 ) 5284 .
j. cleymans , h. oeschler and k. redlich , phys . rev .
* c59 * ( 1999 ) 1663 .
f. becattini , j. cleymans , a. kernen , e. suhonen , k. redlich , phys .
c64 * ( 2001 ) 024901 .
p. braun - munzinger , j. cleymans , h. oeschler , k. redlich , nucl .
a697 * ( 2002 ) 902 .
p. braun - munzinger , k. redlich and j. stachel , nucl - th/0304013 in _ quark - gluon plasma 3 _ , ( eds . ) r. hwa and x.n .
f. becattini , m. gadzicki , a. kernen , j. manninen , r. stock , phys . rev .
* c69 * ( 2004 ) 024905 .
s. wheaton and j. cleymans , thermus - a thermal model package for root , hep - ph/0407174 . c. meurer et al .
, j. phys .
g * 30 * ( 2004 ) s14 ; + s.v .
afanasiev et al . , phys . lett .
* b358 * ( 2002 ) 275 .
p. chung et al . , phys . rev .
* 91 * ( 2003 ) 202301 . c. alt et al .
, nucl - ex/0409004 .
ratio as a function of beam energy .
b ) the @xmath0 and @xmath5 ratios as a function of energy . the solid and dashed lines are the predictions of the statistical model calculated using thermus @xcite .
the data points are from ref .
, width=510 ] ( full line ) and @xmath7 ( dashed line ) ratios as a function of beam energy calculated using thermus @xcite .
the data points are from ref .
the square points correspond to the @xmath20 ratio while the round points are for the @xmath6 ratio .
, width=510 ] | the recently discovered sharp peak in the @xmath0 ratio in relativistic heavy - ion collisions is discussed in the framework of the statistical model . in this model a rapid change is expected as the hadronic gas undergoes a transition from a baryon - dominated to a meson - dominated gas .
the transition occurs at a temperature @xmath1 = 140 mev and baryon chemical potential @xmath2 = 410 mev corresponding to an incident energy of @xmath3 = 8.2 gev .
the maximum in the @xmath4 ratio is well reproduced by the statistical model , but the change in the @xmath0 ratio is much less pronounced than the one observed by the na49 collaboration .
the calculated smooth increase of the @xmath5 ratio and the shape of the @xmath6 and @xmath7 ratios exhibiting maxima at different incident energies is consistent with the presently available experimental data .
we conclude that the measured particle ratios with @xmath8 deviations agree with a hadronic freeze - out scenario .
these deviations seem to occur just in the transition from baryon - dominated to meson - dominated freeze - out . , ,
, the na49 collaboration has recently performed a series of measurements of pb - pb collisions at 20 , 30 , 40 , 80 and 158 agev beam energies @xcite .
when these results are combined with measurements at lower beam energies from the ags @xcite they reveal an unusually sharp variation with beam energy in the @xmath9 , with @xmath10 , and @xmath0 ratios .
such a strong variation with energy does not occur in pp collisions and therefore indicates a major difference in heavy - ion collisions .
this transition has been referred to in ref .
@xcite as the `` horn '' .
a strong variation with energy of the @xmath9 ratio has been predicted on the basis of arguments put forward in @xcite .
it has also been suggested recently in ref .
@xcite that this is a signal of the special critical point of the qcd phase diagram @xcite at high baryon density . in this paper
we explore another , less spectacular , possibility for the origin of the sharp maximum , namely as being due to the transition from a baryon - dominated to a meson - dominated hadronic gas .
the distinction being based on whether the entropy of the hadronic gas is dominated by baryons or by mesons . for this purpose
we study various quantities along the freeze - out curve @xcite as a function of @xmath3 . in the statistical model a steep rise at low energies and a subsequent flattening off leading to a mild maximum in the @xmath0 ratio ,
was predicted many years ago @xcite .
the sharpness of the observed peak therefore comes as a surprise . on the other hand , a sharp peak in the @xmath9 ratio
was predicted by the statistical model @xcite and is in good agreement with the data . while the statistical model can not explain the sharpness of the peak in the @xmath0 ratio , there are nevertheless several phenomena , giving rise to the rapid change , which warrant a closer look at the model . in fig .
[ netbaryon ] we show the net baryon density calculated along the chemical freeze - out curve @xcite as a function of @xmath3 .
this curve shows a clear maximum with the net baryon density decreasing rapidly towards higher energies . to get a better estimate of the statistical parameters in the transition region we show in fig .
[ entropy_s ] the entropy density as a function of beam energy following the freeze - out curve given in @xcite .
the separate contribution of mesons and of baryons to the total entropy is also shown in this figure .
there is a clear change of baryon to meson dominance at @xmath3 = 8.2 gev . above this value
the entropy is carried mainly by mesonic degrees of freedom .
it is remarkable that the entropy density divided by @xmath11 is constant over the entire freeze - out curve , except for the low - energy region corresponding to the sis energy region .
the line denoting the transition from a baryon - dominated to a meson - dominated hadron gas is shown in fig .
[ entropy_pd ] .
this line crosses the freeze - out curve at a temperature of @xmath12 mev , when the baryon chemical potential equals @xmath2 = 410 mev .
the corresponding invariant energy is @xmath3 = 8.2 gev .
the strong decrease in the net baryon density seen in fig . [ netbaryon ] is due to the fact that low energies are characterized by a very low multiplicity of mesons and , correspondingly , a very large baryon - to - meson ratio . as a consequence ,
the baryon chemical potential is also very large .
as the beam energy is increased , meson production increases and the baryon chemical potential decreases .
the number of strange baryons produced in heavy - ion collisions at different collision energies will follow the net baryon density since a large baryon chemical potential will also enhance the number of hyperons .
the corresponding @xmath0 ratio is shown in fig .
[ cankpla ] .
as is well - known @xcite , the statistical model description leads to a mild maximum in this ratio which does not reproduce the so - called `` horn '' observed by the na49 collaboration @xcite .
the observed deviations at the highest sps energy have been interpreted as a lack of full chemical equilibrium in the strangeness sector , leading to a strangeness suppression factor , @xmath13 , deviating from its equilibrium value by about thirty percent .
detailed fits using the statistical model in the region of the `` horn '' show rapid variations in @xmath13 @xcite which do not lend themselves to any interpretation .
there is no corresponding peak in the @xmath5 ratio because the production of @xmath14 is not tied to that of baryons .
as the relative number of baryons decreases with increasing energy , there is no corresponding decrease in the number of @xmath14 as is the case with @xmath15 as these must be balanced by strange baryons .
it is worth noting that the maxima in the ratios for multi - strange baryons occur at ever higher beam energies .
this can be seen clearly in fig .
[ canxiomega ] for the @xmath6 ratio which peaks at a higher value of the beam energy .
the ratio @xmath7 also shows a ( very weak ) maximum , as can be seen in fig .
[ canxiomega ] .
the higher the strangeness content of the baryon , the higher in energy is the maximum .
this behavior is due to a combination of the facts that strangeness has to be balanced , the baryon chemical potential decreases rapidly with energy and the multi - strange baryons have successively higher thresholds . the values are listed in table 1 . .
maxima in particle ratios as predicted by the statistical model .
[ cols="<,^,^ " , ] it is to be expected that if these maxima do not all occur at the same temperature , i.e. at the same beam energy , then the case for a phase transition is not very strong . the observed behavior seems to be governed by properties of the hadron gas .
more detailed experimental studies of multi - strange hadrons will allow the verification or disproval of the trends shown in this paper .
it should be clear that the @xmath7 ratio is very broad and shallow and it will be difficult to find a maximum experimentally . in conclusion ,
while the statistical model can not explain the sharpness of the peak in the @xmath0 ratio , its position corresponds precisely to a transition from a baryon - dominated to a meson - dominated hadronic gas .
this transition occurs at a temperature @xmath16 140 mev , a baryon chemical potential @xmath17 410 mev and an energy @xmath18 8.2 gev . in the statistical model this transition leads to a sharp peak in the @xmath9 ratio , and to moderate peaks in the @xmath0 , @xmath6 and @xmath7 ratios .
furthermore , these peaks are at different energies in the statistical model .
the statistical model predicts that the maxima in the @xmath9 , @xmath6 and @xmath7 occur at increasing beam energies . if the change in properties of the above excitation functions were associated with a genuine deconfinement phase transition one would expect these changes to occur at the same beam energy .
it is clear that more data are needed to clarify the precise nature of the sharp variation observed by the na49 collaboration . |
campbell showed the theorem that any solution of the einstein equations in @xmath0 dimensions can be locally embedded in a ricci - flat manifold of @xmath1 dimensions whose field equations in terms of the ricci tensor are @xmath2 ( @xmath3).@xmath4 tavakol and coworkers have recently noted the relevance of this to the embedding of lower - dimensional ( possibly quantizeable ) gravity in 4d einstein gravity,@xmath5 and the technique is clearly applicable to the recovery of solutions of 4d general relativity from 10d superstrings , 11d supergravity and m - theory.@xmath6 a major application of campbell s theorem is to induced - matter theory , wherein solutions of the 4d einstein equations with matter are recovered from the 5d kaluza - klein equations in apparent vacuum.@xmath7 this approach has been applied to cosmology,@xmath8 clusters of galaxies@xmath9 and the solar system,@xmath10 where there is agreement with observational data .
it is valuable in application to general relativity , since solutions of the kaluza - klein equations can yield new solutions of the einstein equations . with regard to the latter
, spherically - symmetric sources such as stars can be modelled in the simplest way by the interior and exterior schwarzschild solutions . but to include the radiation outside a star , more complicated solutions are required.@xmath11 these include the vaidya metric which uses a retarded time coordinate to describe a radiating atmosphere,@xmath12 the metrics of herrera and coworkers wherein spheres of matter are matched to exterior spacetimes,@xmath13 and the metrics of glass and krisch which extend the vaidya solution to include both a radiation field and a string fluid.@xmath14 however , despite extensive work on star - like solutions of the 4d equations and the existence of campbell s theorem which shows that such can be embedded in the @xmath15d equations , not much work has been done on spherically - symmetric solutions of the 5d equations .
notable exceptions are solutions which have an isothermal equation of state in 4d@xmath9 and ones which are flat in 5d but curved in 4d.@xmath16 we will therefore present an analysis of a class of 5d metrics , and illustrate their relevance by isolating two 4d solutions .
one describes a source with the stiff equation of state believed to be applicable to dense astrophysical objects , and the other describes a spherical source with radiation .
in this section we let upper - case latin indices run 0 - 4 and lower - case greek indices run 0 - 3 . we absorb the speed of light and the gravitational constant through the choices of units @xmath17 ,
@xmath18 . the coordinates are @xmath19 with @xmath20 with ( 3d ) spherical symmetry .
consider the 5d line element with interval @xmath21 where @xmath22 and @xmath23
.. this metric is broad in scope and has been used in other contexts.@xmath24 it also has the advantage that the 5d field equations @xmath2 break down neatly into a set involving the 4d ricci tensor and a conservation - type equation for the scalar field @xmath25 : @xmath26 here @xmath27 @xmath28 and @xmath29 in ( 3 ) is made of @xmath30 .
equations ( 3 ) and ( 4 ) admit many different types of solution , but here we choose a time - dependent scalar field and a static 4d geometry : @xmath31 these require that the lhs of ( 3 ) should be independent of @xmath32 and the rhs of ( 3 ) should be independent of @xmath33 .
therefore both sides should equal to a constant , say , @xmath34 .
so we must have @xmath35 substituting this into equation ( 4 ) , we find that ( 4 ) is satisfied .
using ( 6 ) and ( 7 ) in ( 3 ) gives the non - vanishing components @xmath36 with @xmath37 .
these are three equations in two unknowns , and determine a class of solutions which is time - dependent in 5d via ( 7 ) but static in 4d and spherically symmetric in 3d .
let us manipulate ( 8)-(10 ) .
the sum of @xmath38 of ( 8) and @xmath39 of ( 9 ) gives @xmath40 and we can rewrite ( 10 ) as @xmath41 using ( 11 ) and ( 12 ) , we verify that ( 8) and ( 9 ) are satisfied .
therefore , we need to solve ( 11 ) and ( 12 ) for @xmath42 and @xmath43 . now
( 12 ) can be written as @xmath44 . integrating this we get @xmath45 \quad .\ ] ] substituting this into ( 11 ) , we obtain @xmath46 \quad .\ ] ] without loss of generality , let us introduce a mass function@xmath47 via @xmath48 then ( 14 ) becomes @xmath49 \quad .\ ] ] here @xmath50 is a fiducial radius that can be chosen as appropriate to a physical situation .
thus if @xmath51 we expect to recover the schwarzschild case .
the latter is indeed recovered for @xmath52 , when @xmath53 constant and ( 13 ) reads @xmath54 = 1-\frac{2m}r\;\;.\ ] ] that is , @xmath52 specifies the schwarzschild limit of a class of solutions determined by ( 16 ) and one or the other of ( 11 ) and ( 12 ) .
we can study close - to - schwarzschild approximation by expanding the mass function @xmath55 of ( 15 ) for @xmath56 small ( i.e. @xmath57 .
thus we write @xmath58 with @xmath59 . here
@xmath60 is the zero - order approximation of @xmath61 , @xmath62 is the first - order approximation , and so on . to obtain @xmath63 we substitute @xmath60 for @xmath61 in the rhs of ( 16 ) to obtain @xmath64 .
integrating @xmath65 gives @xmath66 .
then we substitute @xmath62 for @xmath61 in the rhs of ( 16 ) to obtain @xmath67 . in this way
we obtain the following recursion formulae for evaluating @xmath68 : @xmath69 \nonumber \\ & & .............. \nonumber \\ \mu _
n^{\prime } & = & -\sum_{m=0}^{n-1}\mu _ m^{\prime } + \frac 32\lambda ^2r\left ( r-2\sum_{m=0}^{n-1}\mu _ m\right ) .
\nonumber \\ & & \exp \left\ { 2\int_{r_0}^r\left [ \frac 1r-\left ( r-2\sum_{m=0}^{n-1}\mu _ m\right ) ^{-1}\right ] dr\right\ } \quad .\end{aligned}\ ] ] from this we find @xmath70 and so @xmath71 in the region @xmath72 and @xmath73 , we find @xmath74 where we have absorbed a constant term in @xmath60 without loss of generality , so the first - order approximation of @xmath75 in ( 15 ) is @xmath76 substituting this equation into ( 13 ) , keeping only terms up to first order in @xmath77 as well as in @xmath78 , and neglecting a constant factor in @xmath43 , we find @xmath79 from this coefficient and that of ( 23 ) we obtain the first order close - to - schwarzschild approximation of the 5d solution as @xmath80 -e^{-4\lambda t}dy^2.\ ] ] this metric is time - dependent .
when @xmath81 , it tends to the 5d schwarzschild solution .
so we call ( 25 ) the close - to - schwarzschild approximation for small @xmath82 implying that the time - variation of the field should be very slow .
we also find that the solution ( 25 ) belongs to the type d class of the general time - dependent 5d metrics.@xmath83 as regards the 4d part inside the square bracket in ( 25 ) , it is interesting to note that it does define the 4d schwarzschild - de sitter solution . in the latter , @xmath42 and @xmath43
both contain a term @xmath84 where @xmath85 is the cosmological constant .
by contrast , whereas @xmath77 has the same physical dimensions as @xmath85 ( namely @xmath86 or @xmath87 ) , @xmath42 and @xmath43 in ( 23 ) and ( 24 ) contain terms with different signs and different sizes .
this situation is analogous to another in kaluza - klein theory , where the exact solution of the 5d field equations for a charged point mass does not exactly reproduce the 4d reissner - nordstrom solution.@xmath88 since the 5d equations are richer than the 4d ones , such situations may be expected ; but even so , ( 25 ) defines a new solution .
there are , of course , many other solutions than those which are close to schwarzschild . in general
, we have a class of solutions of @xmath2 with a metric @xmath89 -e^{-4\lambda t}dy^2\ ] ] whose spacetime potentials @xmath90 are determined by ( 11),(12 ) .
alternatively , they are determined by ( 16 ) and one or the other of ( 11 ) , ( 12 ) .
we should recall , however , that while ( 8) , ( 9 ) , ( 10 ) are 4d relations and can be used as such , the class of metrics ( 26 ) is 5d in nature .
this means that we can use campbell s theorem@xmath4 and the induced - matter formalism@xmath7 to generate 4d solutions of einstein s equations with their appropriate matter .
we now proceed to show how the 5d metric ( 26 ) produces two exact 4d solutions .
the first is obtained by splitting off the part inside square brackets in ( 26 ) .
the 4d metric is then @xmath91 and is static .
we therefore expect that the source necessary to balance einstein s equations will also be static .
as mentioned above , the kaluza - klein equations @xmath92 contain as a subset the einstein equations @xmath93 ( @xmath94 ) . here
@xmath95 is the 4d energy - momentum tensor , which can always be constructed from the 5d geometry@xmath96 and whose form is now well known.@xmath97 for the present case , @xmath98 is given by ( 3 ) , so the induced energy - momentum tensor is @xmath99 using ( 7 ) and ( 27 ) , we find that the non - vanishing components of this in mixed form are @xmath100 this as expected represents a static source , which we can model as a perfect fluid with @xmath101 here @xmath102 is the density , @xmath103 is the pressure and the 4-velocity is @xmath104 . combining ( 29 ) and ( 30 ) , we see that the source has @xmath105 this is the stiff equation of state in which the speed of sound waves approaches the speed of light , and has been applied in previous studies@xmath106 to astrophysical situations such as collapsed stars and proto - galactic fluctuations . the second solution we look at is obtained by splitting off the whole of the first part of ( 26 ) .
the 4d metric is then @xmath107 , \ ] ] and is time - dependent .
we therefore expect that the source necessary to balance einstein s equations will also be time - dependent .
it should be noted that this property can not in general be removed by a coordinate transformation based on birkhoffsss theorem , because the 4d metric ( 32 ) will have a source constructed from the 5d geometry that will in general not be vacuum ; and because the 4d metric ( 32 ) is part of a 5d metric ( 26 ) and it is known that birkhoff s theorem in its standard form breaks down in the transition from 4d to 5d,@xmath108 as evidenced by the existence of both static and time - dependent soliton solutions.@xmath109 to investigate the time - dependence of ( 32 ) , let us make the coordinate transformation @xmath110 this brings ( 32 ) into the form @xmath111 we can use this with @xmath112 to evaluate the induced matter properties using the standard technique.@xmath97 we note that @xmath113 which can be used to form the induced energy - momentum tensor : @xmath114 the non - vanishing components of this in mixed form are @xmath115 we see from ( 38 ) that the density is inhomogeneous via @xmath116 and is time - dependent , as is the pressure by ( 39 ) though the latter is isotropic .
we also see from ( 40 ) that there is an off - diagonal component .
the latter can be accommodated by introducing a two - fluid model @xmath117 which as in other work@xmath118 we take to be the sum of a perfect fluid and a radial heat flow : @xmath119 the heat - flux vector and the 4-velocity must obey the orthogonality condition @xmath120 this we satisfy by taking @xmath121 where in ( 45 ) we have used the facts that @xmath33 is a comoving coordinate and that the 4-velocities are normalized via @xmath122 . substituting ( 45 ) , ( 46 ) into ( 42 ) , ( 43 ) and these into ( 41 ) gives the non - vanishing components of the last in mixed form : @xmath123 comparing these with ( 38 ) , ( 39 ) , ( 40 ) gives us the density , pressure and heat flow in explicit form : @xmath124 the equation of state is that of radiation or ultra - relativistic matter , and has been applied in previous studies@xmath125 to astrophysical situations such as fermion soliton stars and the early universe .
we have taken a 5d metric ( 1 ) which contains a 4d sub - metric ( 6 ) whose 3d part is spherically symmetric .
an expansion in the metric coefficient allows us to recover the schwarzschild case of general relativity in ( 17 ) as the zeroth approximation and a close - to - schwarzschild case in ( 25 ) as the first order approximation to a class of spherically - symmetric solutions .
any solution of the 5d kaluza - klein equations in apparent vacuum can be written as a solution of the 4d einstein equations with matter .
two solutions have then been shown , with matter properties corresponding to those of a stiff fluid ( 31 ) , and radiation or ultra relativistic particles with heat flow ( 50,51 ) .
these solutions can be applied to astrophysics , but are merely illustrative examples . in the wider scheme ,
it is clear that ( local ) embedding theorems are powerful tools , whether applied to @xmath126 ( possibly quantizeable ) gravity,@xmath5 4d einstein theory,@xmath11 5d kaluza - klein theory,@xmath127 or 10d superstrings , 11d supergravity and m - theory.@xmath6 campbell s theorem ensures that any solution in nd can be embedded in a ricci - flat solution in @xmath128 .@xmath4 the schwarzschild solution in 4d can of course be embedded in a flat space of @xmath129 .@xmath130 and any solution in 4d can be embedded in a flat space of @xmath131 @xmath132 .@xmath133 the implications of embedding theorems are diverse
. if the aim is to find new solutions of general relativity , the higher - dimensional field equations are often surprisingly tractable , and the method of reduction to 4d is straightforward.@xmath5 if the aim is to give meaning to higher - dimensional theories , the same method of reduction will inform about physicality in the 4d world .
we should recall that field equations like those of einstein or kaluza - klein allow the dimensionality to be freely chosen , which should be done partly with a view to what physics it is desired to describe and partly with a view to what technical constraints are involved . in the latter regard , it is well known that @xmath126 theories run into problems of formulation which are connected with the degenerate nature of lower - dimensional riemannian spaces . for @xmath134 theories ,
problems arise with the physics if the spaces are subject to arbitrary technical constraints.@xmath135 our opinion , therefore , is that future work should be focussed on higher - dimensional , fully covariant theory .
we thank the referee for valuable suggestions , and j. ponce de leon and w.n .
sajko for comments .
this work was supported by nsf of p.r . china and nserc of canada .
s. rippl , c. romero , r. tavakol , class .
quant . grav .
* 12 * , 2411 ( 1995 ) ; c. romero , r. tavakol , r. zalaletdinov , gen .
* 28 * , 365 ( 1996 ) ; j.e .
lidsey , c. romero , r. tavakol , s. rippl , class .
quant . grav .
* 14 * , 865 ( 1997 ) .
b. mashhoon , h. liu , p.s .
wesson , phys .
b * 331 * , 305 ( 1994 ) ; h. liu , b. mashhoon , ann .
( leipzig ) * 4 * , 565 ( 1995 ) ; p.s .
wesson , b. mashhoon , h. liu , mod .
a * 12 * , 2309 ( 1997 ) ; h. liu , b. mashhoon , phys .
a * 272 * , 26 ( 2000 ) .
lee , phys .
d * 35 * , 3637 ( 1987 ) ; r. frieberg , t.d .
lee , y. pang , phys .
d * 35 * , 3640 ( 1987 ) ; t.d .
lee , y. pang , phys .
d * 35 * , 3678 ( 1987 ) ; r. frieberg , t.d .
lee , y. pang , phys .
d * 35 * , 3658 ( 1987 ) ; p.s .
wesson , h. liu , astrophys .
j. * 440 * , 1 ( 1995 ) . | we study a time - dependent 5d metric which contains a static 4d sub - metric whose 3d part is spherically symmetric . an expansion in the metric coefficient
allow us to obtain close - to schwarzschild approximation to a class of spherically - symmetric solutions .
using campbell s embedding theorem and the induced - matter formalism we obtain two 4d solutions .
one describes a source with the stiff equation of state believed to be applicable to dense astrophysical objects , and the other describes a spherical source with a radial heat flow .
pacs number(s ) : 04.20.jb , 11.10 kk |
directed polymers in disordered media is a model of a variety of physical phenomena .
it describes vortex lines in superconductors @xcite , domain walls @xcite , roughness of crack interfaces @xcite , burgers turbulence @xcite and the kpz equation @xcite .
it provides a simple model in which the more difficult predictions of spin glasses can be tested and is related @xcite to the phase boundary in the 2d ising model . starting with @xcite
, they have been the subject of intensive study @xcite over the last twenty - five years . in the setting of the @xmath15-dimensional integer lattice ,
the polymer measure is a random probability measure on the set of paths of @xmath15-dimensional nearest neighbour lattice walks .
the randomness of the polymer measure is inherited from a random environment which manifests itself as an i.i.d .
collection of random variables placed on the sites of @xmath16 . given a fixed environment @xmath17 , the energy of an @xmath1-step nearest neighbour walk @xmath18 is @xmath19 the polymer measure on such walks
is then defined in the usual gibbsian way by @xmath20 where @xmath5 is the inverse temperature , @xmath21 is the uniform measure on @xmath1-step walks , and @xmath22 the point - to - point partition function @xmath23.\end{aligned}\ ] ] the partition function @xmath24 is the sum of the point - to - point versions over @xmath25 .
the random environment is a probability measure @xmath26 on the space of environments @xmath27 .
let @xmath26 be a product measure so that the variables @xmath28 are independent and identically distributed , and assume they have mean zero and variance one and @xmath29 .
the goal is to study the behaviour of the model as @xmath0 and @xmath15 vary . at @xmath4
the polymer measure is the simple random walk , hence the walk is entropy dominated and exhibits diffusive behaviour . for @xmath0 large the polymer measure
concentrates on paths with low energy and the diffusive behaviour is no longer guaranteed . a more precisely defined separation between these two regimes
is given in terms of the quenched and annealed free energies : @xmath30 weak disorder corresponds to equality and manifests itself through diffusive behavior . in the opposite , strong disorder case the system exhibits localization .
the transition occurs at a critical @xmath31 . for @xmath32 and @xmath33 , @xmath34 , while @xmath35 for @xmath36 @xcite .
we only consider @xmath37 .
the behaviour of the polymer is described in terms of fluctuation exponents for the polymer endpoint and the free energy . for @xmath38-dimensional polymers
there are long - standing predictions : for the polymer endpoint @xmath39 with @xmath40 for all @xmath0 ; for the free energy fluctuation exponent @xmath41 , with @xmath42 .
observe that @xmath40 and @xmath42 satisfy the relation @xmath43 , which is expected to hold whenever there is localization . for the free energy the limiting distribution of the fluctuations is the tracy - widom distribution @xcite corresponding to the gaussian unitary ensemble of random matrices , that is @xmath44 .
in fact , a stronger statement holds for the point - to - point partition functions .
after proper normalization the collection of partition functions parameterized by the renormalized target point @xmath25 converges to a universal stationary process called airy@xmath45 @xcite which has @xmath46 as its one - dimensional distributions .
precisely @xmath47 where @xmath48 is a positive constant which compensates for a convex dependence of the free energy on the target point , and @xmath22 is the point - to - point partition function with endpoint at @xmath25 .
we introduce the localization length exponent @xmath49 : @xmath50 \right)^2 \right ] \sim n^{2\nu}.\end{aligned}\ ] ] in the weak disorder ( diffusive ) case @xmath51 . in the strong disorder case @xmath52 for typical environments , i.e. the polymer is localized .
this does not contradict the known fact that the variance averaged over environments is of order @xmath1 @xcite .
for very few environments of total probability on the order of @xmath53 , @xmath54 is on the order of @xmath55 .
this explains the apparent discrepancy .
the central focus of this paper is to describe these exponents and the limiting distributions in what we have termed an intermediate disorder regime , accessed by scaling @xmath0 as we increase the polymer length , @xmath56 the @xmath9 case is the much studied polymer measure , with predictions given on the previous page . for different values of @xmath3
the `` @xmath8-polymers '' behave in ways that are quantifiably different from each other . in this sense
the intermediate disorder regime is actually an entire family of regimes with @xmath8 as an indexing parameter .
the value @xmath14 is critical and is also the most interesting ; it is described soon .
for @xmath57 the polymer is diffusive and behaves as a simple random walk , with @xmath58 fluctuations of the path , i.e. @xmath59 , @xmath51 .
the polymer , rescaled by @xmath58 , converges to brownian motion with diffusivity constant @xmath60 for almost all environments .
also @xmath61 . for @xmath62
the exponents @xmath63 and @xmath49 depend on @xmath8 as @xmath64 the leading behaviour of @xmath65 is of order @xmath66 .
the values of @xmath67 and @xmath68 still satisfy @xmath69 .
they linearly interpolate between the kpz @xcite scalings of @xmath70 at @xmath9 and the diffusive exponents @xmath71 at @xmath14 .
we now show how the interpolation can be derived from the values at @xmath9 together with the airy process asymptotics of the point - to - point partition function .
let @xmath72 $ ] and assume for the moment that @xmath73 is of smaller order than @xmath74 .
this assumption means that most of the contribution to the fluctuations of the free energy comes from paths with an endpoint in a neighbourhood of @xmath74 .
the logarithm of the number of such paths is @xmath75 ( by the normal approximation to simple random walk ) and from each path contributes energy on the order of @xmath76 $ ] to @xmath77 , so that the log of the point - to - point partition function at @xmath74 is @xmath78.\end{aligned}\ ] ] the polymer endpoint favors the value of @xmath74 which maximizes .
we obtain @xmath79.\end{aligned}\ ] ] it follows that @xmath74 is of a smaller order than @xmath80 , and since on small distances the @xmath81 process is similar to the wiener process we have @xmath82 .
hence @xmath83 which gives @xmath84 . substituting this back into implies that the fluctuations of the free energy are on the order of @xmath85 , which gives the formula for @xmath68 .
the localization length exponent is determined by the condition that the difference between logarithms of the point - to - point partition function with endpoints separated by the localization length @xmath86 should be of order @xmath60 . for distances
@xmath87 this difference is of the order of @xmath88,\end{aligned}\ ] ] which is of order @xmath89 .
this is of order @xmath60 only if @xmath90 .
we immediately see that for @xmath91 the polymer is localized ( @xmath92 ) . using the wiener process approximation the limiting distribution for the pair @xmath93}{c(\alpha , \beta ) n^{\zeta(\alpha ) } } \right)\end{aligned}\ ] ]
is the joint distribution of the maximum value @xmath94 of @xmath95 and the point @xmath96 where the maximum is achieved . here
@xmath97 is a @xmath33-sided brownian motion .
this probability distribution was calculated exactly in @xcite .
the joint density of @xmath98 at @xmath99 is @xmath100 , where @xmath101 has fourier transform @xmath102 @xmath103 has laplace transform @xmath104 and @xmath105 has fourier transform @xmath106 with @xmath107 . here
@xmath108 and @xmath109 are the airy functions @xcite . for @xmath14
the localization length is on the same order as the displacement of its endpoint @xmath110 .
the polymer is not localized anymore and the latter argument breaks down .
however the exponents are still correct .
the scaling behavior in this critical case will be discussed in the next section .
we verify the formulas for @xmath67 and @xmath68 by numerical simulations .
we fix @xmath111 and let @xmath8 vary from @xmath112 to @xmath113 in increments of @xmath114 .
exponents are computed for @xmath115 independent gaussian environments and polymers of length @xmath116 .
see the figures for results .
in the @xmath14 case the limiting distribution of the polymer endpoint is more involved but still computable .
from @xmath117 we do not require any normalization . consider the modified partition function @xmath118.\end{aligned}\ ] ] expanding the product and then applying the expectation with respect to the random walk yields @xmath119 , where @xmath120 here @xmath121 for random walks starting at zero , @xmath122 , and the sum is over ordered @xmath123 with @xmath124 , and free @xmath125 , @xmath126 , with @xmath127 .
we compute that for fixed @xmath128 @xmath129 where @xmath130 here @xmath131 is a gaussian white noise on @xmath132 with @xmath133 = \delta(t - s ) \delta(x - y)$ ] , @xmath134 , @xmath135 , and the integration is over @xmath136 with @xmath127 .
the @xmath137 term is necessary to keep the variance of each term bounded , and the @xmath138 terms come from the local limit theorem for simple random walk . in the case
@xmath139 the convergence is easily seen by a fourier transform computation .
we have @xmath140 as @xmath141 from the mean zero , variance one assumption on @xmath142 , so @xmath143 \\ & = -\frac{t^2 \beta^2}{2}(1+o(1 ) ) n^{-1/2 } \sum_{i=1}^n \sum_y q_n^x(i , y)^2 \to -\frac{2t^2 \beta^2}{\sqrt{\pi}}.\end{aligned}\ ] ] here @xmath144 .
therefore the limiting distribution is normal with mean zero and variance @xmath145 , and it is easily checked that @xmath146 has the same distribution . for the @xmath147 terms
the convergence is handled by the general theory of u - statistics @xcite . from these computations
we conclude that @xmath148 converges to the process @xmath149 .
the same convergence also holds replacing @xmath150 by @xmath151 . define the process @xmath152
it is very important that this process is universal , namely it appears as a limit for polymer models in the critical regime with an extremely general distribution for the random potentials @xmath142 .
it is also easy to see that the statistics of the density of the scaled polymer endpoint @xmath153 is the same as the statistics of the random density @xmath154 .
another important property of @xmath155 is that it interpolates between a gaussian process ( @xmath4 ) and the airy@xmath45 process ( @xmath156 ) . to see this crossover one has to properly rescale the process with @xmath0 to normalize the variance of its one - point distribution and its two - point correlation function at a fixed distance .
the interpolation property follows from @xcite and @xcite , where an exact formula for the distribution of @xmath155 was calculated ( see also @xcite ) . using the tracy - widom formula for asymmetric simple exclusion @xcite
, they obtain : @xmath157 where , @xmath158 with @xmath159 , @xmath160 , and @xmath161 here @xmath162 and @xmath163 in particular it follows that @xmath164 .
it also follows that @xmath155 is stationary in @xmath25 as its distribution is independent of @xmath25 .
finally , one can show that the process @xmath155 is continuous and locally brownian , i.e. @xmath165 is of order @xmath166 .
the limiting distribution is also characterized @xcite by observing that @xmath167 is the wick exponential of a stochastic integral .
let @xmath168 be a one - dimensional brownian motion that is independent of the white noise @xmath131 .
let @xmath169 denote expectation over brownian paths started at zero and ending at @xmath25 at time @xmath60
. then @xmath170.\end{aligned}\ ] ] the exact density of the polymer endpoint @xmath171 in the critical regime is part of the critical polymer measure on paths .
define the continuum partition function @xmath172.\end{aligned}\ ] ] the polymer measure @xmath173 in the critical @xmath14 case scales to a probability measure on paths of length @xmath174 given by @xmath175 where @xmath21 stands for the usual wiener measure .
write @xmath176 for expectation .
provides another derivation of the fluctuation exponents @xmath67 and @xmath68 . for @xmath5
let @xmath177 scaling properties imply that @xmath178 is a standard brownian motion and @xmath179 is a standard space - time white noise , so via the substitution @xmath180 we have the relation @xmath181 and so @xmath182 .
from @xmath183 we get @xmath184 which agrees with @xmath68 . for @xmath67
we have @xmath185 \sim t^{4/3}$ ] as @xmath186 for almost all @xmath187 and fixed @xmath0 . since @xmath188 we have @xmath189 = \beta^{-4 } { \operatorname{e}}_{\beta^4 t , 1}^{\tilde{w } } \big [ \tilde{b}(\beta^4 t)^2 \big ] .\end{aligned}\ ] ] therefore @xmath190 & = t^{4\alpha } { \operatorname{e}}_{t^{1 - 4\alpha } , 1}^{\tilde{w } } \big [ \tilde{b}(t^{1 - 4\alpha})^2 \big ] \sim t^{\frac{4}{3}(1-\alpha)}.\end{aligned}\ ] ]
we have identified a new disorder regime for directed polymers in @xmath191 and computed the corresponding wandering exponent @xmath6 and free energy fluctuation exponent @xmath7 , in addition to a localization length exponent @xmath49 .
these exponents vary linearly with the parameter @xmath8 between the kpz scaling exponents at @xmath9 and the simple random walk exponents at @xmath14 . at @xmath14
there is a phase transition from intermediate disorder to weak disorder .
for @xmath192 we are able to identify the joint limiting distribution of the free energy and the polymer endpoint , and give an explicit asymptotic formula for the distribution of the point - to - point free energy at @xmath14 . | we introduce a new disorder regime for directed polymers with one space and one time dimension that is accessed by scaling the inverse temperature parameter @xmath0 with the length of the polymer @xmath1 .
we scale @xmath2 for @xmath3 .
this scaling sits in between the usual weak disorder ( @xmath4 ) and strong disorder regimes ( @xmath5 ) .
the fluctuation exponents @xmath6 for the polymer endpoint and @xmath7 for the free energy depend on @xmath8 in this regime , with @xmath9 corresponding to the usual polymer exponents @xmath10 and @xmath11 corresponding to the simple random walk exponents @xmath12 . for @xmath13
the exponents interpolate linearly between these two extremes . at @xmath14
we exactly identify the limiting distribution of the free energy and the end point of the polymer . |
hercules x-1/hz herculis is a low mass x - ray binary system consisting of a @xmath3 neutron star and a @xmath4 a dwarf star locked in orbit about each other with a period of 1.7 d ( deeter et al . 1991 , and references therein ) .
the x - ray source is eclipsed by the a star for 0.24 d of each period , indicating an orbital inclination near 90@xmath5 .
the spectral type of the a star varies with orbital phase from late b / early a to late a / early f such that the later spectral types correspond to eclipse and the earlier spectral types correspond to 1/2 an orbit later ( when x - ray heating of the a star should be most apparent see milgrom & salpeter 1975 ) .
mass transfer , which may be periodic , feeds an accretion disk around the neutron star .
accretion onto the neutron star itself , which has a magnetic field of @xmath6 gauss , occurs on the magnetic poles ( gruber et al . 1980 ; voges et al .
1982 ) .
the x - rays from the system vary on a period near 34.85 d ( gelman 1987 , baykal et al .
1993 ) , with an x - ray high state typically lasting @xmath711 d , and an x - ray low state ( with a secondary maximum of 30%the high state ) lasting @xmath724 d. the time at which the x - ray high state begins varies unpredictably from the mean ephemeris by more than @xmath8 d , an effect that has led to the generation of a number of different models ( e.g. petterson 1975 ; gerend & boynton 1976 ; wolff & kondo 1978 ; kondo , van flandren , & wolff 1983 ; meyer & meyer - hofmeister 1984 ; schandl & meyer 1994 ) . as the 1.7 day modulation of the optical light continues during the x - ray low states
( bahcall & bahcall 1972 ) , the 35-day cycle probably results from a geometric effect , such as obscuration by a precessing accretion disk , and not an interruption of the accretion , which would cause the x - ray heating of hz her to cease .
in addition to the 35 d cycle , pulsed x - ray emission with a period of 1.24s is also seen , which is generally interpreted as the rotation period of the neutron star ( tananbaum et al .
1972 ; giacconi et al .
these pulsations have also been detected optically ( davidsen et al .
1972 ; middleditch 1983 ) , in the uv ( boroson et al . 1996 ) and at optical and infrared wavelengths ( middleditch , puetter , & pennypacker 1985 ) .
the ultraviolet spectrum above 1200 has been extensively observed with iue ( dupree et al .
1978 ; gursky et al .
1980 ; howarth & wilson 1983a , b ; vrtilek & cheng 1996 ) .
both the line and continuum components of the spectrum vary with orbital phase .
the ultraviolet light from her x-1 comes from at least two main regions within the system : the heated face of the normal star and the accretion disk . according to the analysis of howarth & wilson ( 1983a )
, roughly 1/2 of the hard x - rays striking the a star are reprocessed into uv and optical continuum emission . reprocessed x - rays are also responsible for most of the uv light from the disk .
using the goddard high resolution spectrograph ( _ ghrs _ ) aboard the hubble space telescope , boroson et al .
( 1996 ) showed that the uv lines have broad and narrow components which they interpreted in terms of emission from the disk and the star . in this paper
, we present a spectrum of her x-1 that extends the far - ultraviolet wavelength coverage to the lyman limit at 912 .
we have detected , for the first time , the ovi@xmath0 emission lines , and the continuum in this wavelength region . in 2 we discuss the observations and data reduction , and in 3 we discuss our data in the context of other observations of her x-1 , and we compare the data with models of the line and continuum emission .
the observations were carried out with the hopkins ultraviolet telescope ( hut ) as part of the astro-1 space shuttle mission in 1990 december .
hut consists of a 0.9 m mirror that feeds a prime focus spectrograph with a microchannel plate intensifier and reticon detector .
first order sensitivity covers the region from 850 to 1850 at 0.51 pix@xmath9 , with @xmath73 resolution .
the detector can be operated in any of several modes , including a high time resolution mode where individual photons can be time - tagged to a relative accuracy of 1 ms and an absolute accuracy of 3 ms .
this was the mode of operation during the observation described here .
further details of the spectrograph and telescope and their on - orbit performance and calibration can be found in davidsen et al .
( 1992 ) and kruk et al .
( 1997 ) .
the her x-1 observation discussed here took place on 1990 december 6 at 4:22 ut . according to the 1.7 day orbital ephemeris of deeter et al .
( 1991 ) , this corresponds to @xmath10 .
we do not know the x - ray state of her x-1 at the time of our observation . using a period @xmath11 d and the observed x - ray turn - on nearest in time to our observation ( jd 2448478.3 , as given in baykal et al .
1993 ) , our observation took place at @xmath12 , while x - ray turn - on occurs at @xmath13 .
this suggests that our observation took place in the x - ray off state ; however , the times of the x - ray turn - ons are known to vary by @xmath146 days ( gelman 1987 ; baykal et al . 1993 ) , so that without simultaneous x - ray observations , we can not be certain .
the hut observation was made with an 18@xmath15 aperture placed on the object , and guiding was accomplished using preplanned guide star positions on the hut tv guider .
the total integration was 1700 s. orbital viewing constraints required that we observe her x-1 entirely during the orbital day of hut , which resulted in substantial airglow contamination , leading not only to significant line emission ( including several euv lines seen in second order ) but also a pseudo - continuum from grating - scattered light ( chiefly from the brightest airglow lines at ly@xmath16 and oi@xmath17 ) .
the flux calibration is based on observations of the white dwarf star g191-b2b compared with model stellar atmospheres and laboratory calibrations as discussed by davidsen et al .
( 1992 ) and kruk et al . ( 1997 ) .
the relative fluxes are accurate to within 5% .
the count rate in the continuum from her x-1 ( in the 14001600 range , relatively free from airglow emission ) was not uniform , but peaked at 5.5 counts s@xmath9 during a five - minute interval .
this count rate should be compared with an average rate of 4.3 counts s@xmath9 ( 14001600 ) in the 1500 s interval that we used for our analysis .
data from the hut tv camera indicate sections at the beginning and end of the observation where pointing stability was disrupted .
post - flight analysis of video frames stored during the observation show that even during the portion of the observation where the pointing stability was good , the object was not well centered in the slit , especially in the coordinate perpendicular to the dispersion direction .
hence , even minor pointing errors could have affected the count rate ( and derived flux ) because of vignetting by the edge of the slit .
( the psf of the telescope produced star images with fwhm@xmath18 . )
we thus suspect that the fluctuating count rate was due to pointing errors , and that the maximum observed count rate is probably representative of the true flux of the object .
this analysis implies that the average count rate was low by @xmath727% ; the derived fluxes ( and errors ) were thus multiplied by 1.27 to correct for this effect .
since the hut detector is photon - counting , it is straightforward to generate and propagate errors along with the data as various operations are performed . to the extent that flux was being lost even during stable pointing ( due to vignetting ) this corrected flux could still be too low , but not by more than @xmath19% .
the corrected , flux - calibrated spectrum is shown as the upper curve in figure 1 . in order to subtract airglow contamination to first order ,
we have fit and subtracted a blank day sky observation obtained with the same aperture during astro-1 .
we used smooth fits to the known airglow lines in this spectrum instead of the observed airglow background to reduce the contribution of counting noise .
an actual smoothing of the background spectrum , instead of a fit to individual lines , would have increased the widths of the strong airglow lines where they overlap emission lines from her x-1 .
we have continued to use the error array associated with the counts in the original data .
the result of our fit to the background spectrum is shown as the lower curve in figure 1 .
the airglow - subtracted spectrum representing the summed intrinsic data from the source is shown in figure 2 . because the relative intensities of the fainter airglow lines change with orbital latitude and look - angle with respect to the earth limb , the subtraction of line emission is not perfect . also , a variable complex of airglow emissions affects the spectrum in the region near the lyman limit ( cf .
feldman et al .
. however , our technique should be sufficient to provide a reasonable representation of the intrinsic continuum longward of 930 .
this is the spectrum that will be used for continuum fitting below .
we have identified emission from the nv@xmath1 doublet and the civ@xmath2 doublet , which have been seen with iue ( boyle et al .
1986 , howarth & wilson 1983a , 1983b ) and hst ( anderson et al . 1994 ; boroson et al . 1996 ) , and also the [email protected],1037.6 resonance doublet and some fainter emission lines at a more marginal detection level .
this constitutes the first detection of o vi emission from the her x-1 system , and provides a significant new diagnostic of photoionized gas in this system .
in order to provide quantitative estimates of the line strengths , we have fit gaussian profiles to the data .
since some of the lines , including ovi@xmath0 , are affected by nearby airglow features , we first used our template airglow spectrum to carry out a local subtraction of the airglow in order to produce a flat continuum adjacent to each line .
typically , this required us to scale the template spectrum by of order 20% from the value used above to achieve a reasonable local fit near the lines .
each doublet was fit assuming an optically thick 1:1 ratio for the lines and a fixed separation of the doublet components ( although the central wavelength of each feature was allowed to vary ) .
observations with the hst show two line components ; for a narrow component , the doublet ratios are typically @xmath21 , while for the broad component , the doublet ratio is more nearly @xmath22 .
we found a superior fit ( reduced @xmath23 instead of @xmath24 with 42 degrees of freedom ) when we allowed a @xmath25 doublet ratio for ovi@xmath0 . allowing the doublet ratio to be a free parameter
does not improve the fit ( @xmath26 for a ratio of 1.65:1 ) .
alternately , the ovi@xmath0 lines could have a 1:1 ratio , but an interstellar cii@xmath27 line could absorb ovi@xmath28 .
allowing the wavelength of the interstellar line to vary , but fixing its width as the width of the ovi lines ( the instrumental width ) , we found @xmath29 with 42 degrees of freedom .
the @xmath30 value for all the ovi fits may be artificially high , as the errors do not reflect the uncertainty in the subtraction of the nearby strong airglow lines of h ly@xmath31 , oi@xmath32 , and oi@xmath33 .
figure 3 shows the region near ovi@xmath0 with the observed and modelled airglow lines , and figure 4 shows the best fits for all the lines .
in addition to ovi@xmath0 , nv@xmath1 , and civ@xmath2 , there was marginal ( 3@xmath34 ) evidence for an ov line at 1371 , at a strength relative to nv and civ similar to that seen with the hst .
heii@xmath35 may also have been detected at a marginal level .
there is no evidence for siiv@xmath36 , and while some features appear at @xmath37 , they are not significant enough to identify as ciii@xmath38 .
any possible ciii@xmath39 is lost in the airglow emission that affects this region of the spectrum .
all of the lines intrinsic to her x-1 are blueshifted .
after detailed inspection of the video frames , we were unable to determine unequivocably whether or not this shift is real .
based on analysis of two guide stars , we infer an offset within the aperture in the dispersion direction that is consistent with a blueshift of 0.0 1.9 ( 0.0 5.6@xmath15 miscentering ) .
the discrepant results from the different guide stars are due to a combination of distortions within the tv camera and measurement error .
if the blueshift caused by the miscentering is near 1.9 , then the apparent blueshifts of the lines are not real , and it becomes more likely that interstellar cii@xmath27 can absorb ovi@xmath28 .
we searched for pulsations in the uv lines from her x-1 , using the data as received in high time rate mode during the time period 1990 dec 6 04:29:00 to 04:48:20 ( ut ) .
we added counts in 0.2 s intervals , folded the data over several likely pulsation periods near the neutron star rotation period , and applied the analysis of variance method ( anova , davies 1990 ) to determine the likelihood of pulsations .
no significant pulsations were found in the lines or continuum .
however , due to the low count rate and the extensive airglow contamination , the limits on pulsation amplitude , 15 35% of the total flux , are not stringent .
the detection of ovi in her x-1 is important , but with a single observation at a particular @xmath40 and @xmath41 , we are limited in the conclusions we can draw about the system . in this section
we discuss the hut observation in the context of the more extensive iue and hst spectroscopy of her x-1 .
howarth & wilson ( 1983a , b ; hereafter hwa and hwb ) discuss a consistently reduced set of iue data on her x-1 that cover a range in both orbital period and the 35 day x - ray cycle .
the line and continuum strengths show a clear dependence on orbital phase ( figure 3 of hwa displays the 1500 continuum and figure 2 of hwb displays the nv and civ line strengths ) .
her x-1 has shown similar line strengths at corresponding orbital phases in subsequent hst observations ( anderson et al . 1994 , boroson et al . 1996 ) and iue observations ( vrtilek & cheng 1996 ) .
plotting the hut data on figure 3 of hwa , figure 2 of hwb , and figure 5 of vrtilek & cheng demonstrates that the hut fluxes are about a factor of 2 lower than those observed with iue at phase 0.34/0.66 ( i.e. 0.34 from mid - eclipse ) .
inspection of optical magnitudes provided by the american association of variable star observers ( aavso ) shows that the 1.7 day optical modulation continued during the period surrounding our observation ( v magnitudes @xmath42 at @xmath43 and @xmath44 , and @xmath45 at @xmath46 ) .
as x - ray heating is thought to cause both the optical modulation and the uv emission , this suggests that the decrease in uv flux that we observed was a short - lived phenomenon or unrelated to x - ray heating . from our examination of the data we are confident that the decrease is not due to any known shortcoming of the data acquisition and reduction .
in addition to the line fits described in 2 , we have produced fits to the continuum after airglow subtraction .
the continuum spectrum below @xmath47 , reported here for the first time , is important because the accretion disk should contribute more to the continuum flux at shorter wavelengths . in fitting the continuum , we used regions free of strong airglow emission lines .
we chose to fit the continuum with 1 ) a power law plus extinction component , and 2 ) a model of an x - ray illuminated accretion disk and roche lobe - filling star .
these fits will be described briefly below . for the power law fit , we specified initial guesses at a normalization , @xmath48 ( the flux at 1000 ) , and the exponent of the power law , @xmath16 , so that @xmath49 .
we also applied a seaton ( 1979 ) galactic extinction law , extropolated to the lyman limit in a manner that is consistent with voyager measurements ( longo et al .
for the her x-1 hut spectrum , we find @xmath50 erg s@xmath9 @xmath51 @xmath9 , @xmath52 , and @xmath53 .
the reduced @xmath30 for this fit was @xmath54 ( with 1125 points and three free parameters ) .
the @xmath55 value is inconsistent with uv observations of the 2200 feature ( gursky et al .
setting @xmath56 , we find @xmath57 and @xmath58 .
this fit is nearly as good ( the reduced @xmath30 is still @xmath54 ) as the fit with @xmath55 as a free parameter .
we plot this latter model against the data in figure 5 .
we also tried fitting the hut spectrum with a detailed physical model of an x - ray illuminated accretion disk and star ( assumed to fill its roche lobe ) .
the model is identical to that presented by cheng , vrtilek , & raymond ( 1995 ) , except that we have used kurucz model atmospheres ( kurucz 1992 ) instead of our library of stellar spectra , which does not extend to the lyman limit .
the model solves for the temperature structure of the disk using the method described by vrtilek et al .
( 1990 ) , and includes x - ray heating of the normal star , the eclipse of the star by the disk , and the x - ray shadow of the disk on the surface of the normal star using the methods described in hwa .
we used values for the fixed parameters describing her x-1 given in table 2 of cheng et al .
( 1995 ) , including @xmath59 . in figure 5
we compare the continuum spectrum produced by the detailed disk model with the power - law fit and with the spectrum observed with hut .
the disk contributes @xmath60% of the optical flux , @xmath61% of the flux near 1200 , @xmath62% of the flux near 1100 , and @xmath63% of the flux near 950 .
the only free parameter in the model is @xmath64 , the mass transfer rate onto the neutron star .
the best - fit value for @xmath64 is @xmath65 ( in units of solar masses per year ) .
this value of @xmath64 is lower than the values reported by vrtilek & cheng ( 1996 ) during an `` anomalous low '' state of her x-1 ( in their final fit , the lowest value is @xmath66 ) . with @xmath64
fixed at the best - fit value , the model predicts @xmath67 also at @xmath68 , when aavso observers report @xmath42 .
the fit of the detailed model is worse than the simple power - law ; we find a reduced @xmath30 of 1.64 .
the fit is especially poor below 1200 .
we investigated several possible systematic errors that could have caused this poor fit : the value we have used for the interstellar reddening , the flux calibration of hut , the timing of the 35-day cycle , and the subtraction of the background airglow spectrum . if @xmath69 , then a greater value of @xmath64 is required to fit the data .
the higher resulting temperature of the heated disk and star can compensate for the spectral change introduced by the reddening . for @xmath70 ,
the model predicts an observed flux from 10001100that is @xmath71% greater than in the case with no reddening ; however , the fit is still not good , and @xmath72 .
increasing the flux in the hut spectrum by a factor of 2 to reflect some unknown flux calibration problem results in a better fit ( @xmath73 ) and @xmath74 , which is more typical of her x-1 . at @xmath68 , @xmath75 .
the spectrum predicted by this model is shown in figure 5 .
the 35-day cycle of her x-1 does not repeat exactly ( baykal et al .
thus there is some uncertainty in our choice of @xmath41 , which determines the orientation of the accretion disk .
if we allow @xmath41 to vary as a free parameter , we find @xmath76 for the best - fit values @xmath77 , @xmath78 . although the fit is much improved , @xmath79 is further out of the range of previous observations . using this 35-day phase , more of the hot , central disk is visible , and so the flux at shorter wavelengths is increased .
the uncertainty in extrapolating the 35-day ephemeris through @xmath80 cycles can be approximated as 0.9 @xmath81 d ( baykal et al . 1993 ) .
the difference between the best - fit value of @xmath41 and that expected from the nearest reported x - ray turn - on is @xmath82d , or a deviation of @xmath83 .
as described in 2 , we subtracted from the her x-1 spectrum a smooth model of the airglow spectrum based on a blank - sky observation .
there is marginal evidence in the blank - sky spectrum for weak airglow lines of ly@xmath84 near 930,938,950 , where our fit to the data is poor .
we tested our method of background subtraction by smoothing the blank - sky spectrum with a savitzsky - golay filter . while this method introduces errors near the wings of the strong airglow lines , it can approximate the airglow background even when strong lines are not present .
using this method of background subtraction does not change our derived value of @xmath64 , and the model still predicts too low a flux below 1200 , but now @xmath85 .
furthermore , this shows that there may be significant errors introduced by the background - subtraction . in summary , we find that the her x-1 far - uv spectrum was unusually faint at the time of our observation , resulting in model fits with values of @xmath64 lower than previously reported
. the flux @xmath86 is stronger than expected from detailed models .
uncertainties in the 35-day phase , the centering of her x-1 in the aperture , or the background - subtraction may account for these discrepancies .
we have compared the observed emission line fluxes with those predicted by theoretical models of x - ray illuminated accretion disks ( raymond 1993 ) . the results are shown in table 1 .
the `` cos '' model for the emission lines uses cosmic abundances , while the `` cno '' model uses a fixed enhancement of he and n and a fixed depletion of c to represent abundances due to cno processing .
in computing the emission line strengths expected from the models , we have assumed that her x-1 is at a distance of 5.5 kpc , and have made no correction for interstellar reddening .
the spectrum confirms the basic predictions of the x - ray illuminated accretion models .
in particular , the interpretation of the nv / civ ratio as an element abundance effect , rather than a problem with the predicted ionization or density structure , is borne out by the line strengths .
while the nv@xmath1 doublet is stronger than civ@xmath2 , the ovi@xmath0 lines may be slightly weaker .
this supports the conclusion that the unexpected strength of the nv lines relative to civ does not result from a dependence of line flux on ionization potential .
the presence of ovi emission lines in her x-1 is interesting because ovi is formed in a hotter region than any of the other uv lines .
the ratio of ovi@xmath0 to ov@xmath87 provides a temperature diagnostic .
the ov line is produced from recombination of ovi and thus has the same dependence on the ovi ionization fraction and on the electron density as the ovi@xmath0 lines .
the strength of the ovi@xmath0 lines depends on the strongly temperature - dependent excitation rate of the upper level of the transition . using the ovi recombination rates of nussbaumer & storey ( 1984 ) and the standard excitation rate expressions and cross - section found in osterbrock ( 1989 ) , we find temperatures of @xmath88 k and @xmath89 k for the ovi@xmath0 fluxes inferred assuming , respectively , a 2:1 doublet ratio and a 1:1 ratio with interstellar absorption . correcting for reddening ( using @xmath70 ) increases ovi / ov by a factor of 1.44 .
this only increases the temperature estimate to 15,000 k. hwb model the line emitting region , and for the nv region find @xmath90 , and @xmath91 , assuming a cosmic abundance of n and that all n is in the nv state .
our temperature estimate has assumed that the ovi@xmath0 lines are optically thin ; if this assumption is not right , then the temperature may be much higher .
further , we have assumed that the ovi lines and the ov line are emitted in the same regions or in regions with similar temperatures and densities , which may not be the case .
both the cos model and the cno model predict observable ciii@xmath38 emission ( 7.5@xmath92 and 4.2@xmath92 erg s@xmath9 @xmath51 respectively ) .
the weak features that we saw had total flux @xmath93 erg s@xmath9 @xmath51 .
interstellar ciii@xmath38 absorption would probably decrease the strength of this line by @xmath94% .
we note that anderson et al .
( 1994 ) observed the ciii@xmath95 line with hst ( when the nv and civ lines were twice as strong as observed with hut ) , and found a flux of @xmath96 erg s@xmath9 @xmath51 , whereas cos and cno predict fluxes of @xmath97 and @xmath98 erg s@xmath9 @xmath51 , respectively . a further decrease in the c / n ratio beyond @xmath99 as assumed in model cno to c / n@xmath100 would improve the agreement between observations and models , but the evidence is not overwhelming .
this observation with the hopkins ultraviolet telescope was the first to study the flux from this interesting source below the lyman @xmath16 line , where a large fraction the continuum emission probably arises in the disk , and where there are strong emission lines especially sensitive to x - ray illumination .
future observations with fuse will be able to detect these features with greater sensitivity and resolution .
it is a pleasure to thank the many members of the hut team at jhu and at the jhu applied physics laboratory whose years of work paid off in two successful astro missions .
we also thank the spacelab operations support group at marshall space flight center and the crew of the astro-1 mission for their support before and during the flight .
we would like to thank the referee , ian howarth , for a careful reading and for suggestions on improving the manuscript .
the hopkins ultraviolet telescope project is supported by nasa contract 5 - 27000 to the johns hopkins university .
bb and sv were supported in part by nasa ( nag5 - 2532 , nagw-2685 ) , and nsf ( dge-9350074 ) . | we have obtained a far - ultraviolet spectrum of the x - ray binary hercules x-1/hz herculis using the hopkins ultraviolet telescope aboard the astro-1 space shuttle mission in 1990 december .
this is the first spectrum of her x-1 that extends down to the lyman limit at 912 .
we observed emission lines of ovi@xmath0 , nv@xmath1 , and civ@xmath2 , and the far uv continuum extending to the lyman limit .
we examine the conditions of the emitting gas through line strengths , line ratios , and doublet ratios .
the uv flux is lower by about a factor of 2 than expected at the orbital phase of the observation .
we model the uv continuum with a simple power - law and with a detailed model of an x - ray - illuminated accretion disk and companion star .
the power - law provides a superior fit , as the detailed model predicts too little flux below 1200 .
we note , however , that there are uncertainties in the interstellar reddening , in the background airglow spectrum , and in the long - term phase of the accretion disk .
we have searched the data for uv line and continuum pulsations near the neutron star spin period but found none at a detectable level . |
low - density parity - check codes ( ldpc ) have attracted significant interest in recent years due to their simplicity and exceptionally high performance @xcite .
their simplicity and inherent randomness make them amenable to analysis using established methods in the area of statistical physics .
these have been employed in a number of papers @xcite-@xcite to gain insight into the properties of ldpc codes and to evaluate their performance .
these studies include the evaluation of critical noise levels for given codes @xcite , an exact calculation of weight and magnetisation enumerators @xcite , the performance of irregular codes @xcite , properties of codes in real - valued channels @xcite , and the derivation of bounds for the reliability exponent @xcite , to name but a few .
these studies also represent the interdisciplinary nature of this research area and illustrate the successful interaction between researchers in the two disciplines .
the evaluation of error exponents has been a long - standing problem in information theory @xcite .
efforts to obtain exact expressions and/or bounds to the error exponent resulted in partial success ; although tight bounds have been derived in the case of random codes and ldpc with infinite connectivity @xcite , only limited results have been obtained for sparely connected codes .
main stream techniques to tackle the problem include sphere - packing and union - bound arguments @xcite . below a certain code - rate value
, the estimated bounds also become loose and require using the ` expurgated exponent ' techniques @xcite for obtaining a tighter bound . in this paper , we employ methods of statistical physics to evaluate directly the average error exponent and typical reliability exponent in gallager and mn @xcite ldpc codes . the _ average error exponent _
is obtained by carrying out averages over the ensemble of randomly generated ldpc codes of given rate and connectivity ; while the _ reliability exponent _ is obtained by selecting the best codes in that ensemble .
averages result in the emergence of macroscopic variables , representative of the ensemble properties , that can be obtained numerically and used to calculate the average error exponent ( in the current calculation we assume that short loops , which contribute polynomially to the block error probability in ldpc codes @xcite , have been removed ) .
average error exponent solutions have been obtained for both finite and infinite connectivity vector ensembles , while reliability exponent solutions have been obtained only in the case of infinite connectivity . as a reference point
to test our theory , we use known results obtained in the information theory literature for solvable limits ( e.g. codes of infinite connectivity ) , and find that our method reproduces them exactly . perhaps
not surprisingly , we also find that at fixed noise level and code rate , the reliability exponent for codes of finite connectivity is always upper - bounded by that of the infinite - connectivity case .
before we proceed , the distinction between the typical bounds found previously using methods of statistical physics @xcite , and the current calculation should be clarified . in the former ,
one employs methods of statistical physics to calculate the typical value of a _ bound _ based on inequalities introduced by gallager ; while in the current calculation , a direct estimation of the average error exponent , rather than a bound , is sought .
an additional advantage of the current approach is that it can be extended to provide reliability exponent values for ldpc codes by restricted averages over codes of high performance .
the paper is organised as follows : in section [ sec : definitions ] , we introduce the general coding framework and the technique used . in sections [ sec : solution ] and [ sec : exactsolution ] we present an outline of the derivation and the solutions obtained in both finite and infinite connectivity cases respectively . in section [ sec : mn ] we compare the error exponent results obtained for mn codes to those of gallager codes in both finite and infinite connectivity cases .
discussion and conclusions are presented in section [ sec : conclusions ] .
a regular ( @xmath0 ) gallager error - correcting code is defined by the binary @xmath1 ( parity check ) matrix @xmath2 $ ] , which is known to both sender and receiver .
the @xmath3 matrix @xmath4 is taken to be invertible .
the number of non - zero elements in each row of @xmath5 is given by @xmath6 , while the number of non - zero elements per column is given by @xmath7 .
gallager s encoding scheme consists of generating a codeword @xmath8 from an information ( message ) vector @xmath9 ( with @xmath10 ) via the linear operation @xmath11 ( mod 2 ) where @xmath12 is the generator matrix defined by @xmath13 $ ] ( mod 2 ) .
the code rate is then given by @xmath14 , and measures the information redundancy of the transmitted vector .
upon transmission of the codeword @xmath15 via a noisy channel , taken here to be a bsc , the vector @xmath16 ( mod 2 ) is received , where @xmath17 is the true channel noise .
the statistics of the bsc is fully determined by the flip rate @xmath18 $ ] : @xmath19 decoding is carried out by multiplying @xmath20 by @xmath5 to produce the syndrome vector @xmath21 , since @xmath22 by construction . in order to reconstruct the original message @xmath23
, one has to obtain an estimate @xmath24 for the true noise @xmath25 .
first we select the parity check set of @xmath5 and @xmath25 , i.e. all @xmath24 that satisfy the parity check equations : @xmath26 .
since all operations are performed in modulo 2 arithmetic , @xmath27 typically contains @xmath28 $ ] candidates for the true noise vector @xmath25 .
it was shown ( see e.g. @xcite for technical details ) that this problem can be cast into a statistical mechanics formulation , by replacing the field @xmath29 by ( @xmath30 ) , and by adapting the parity checks correspondingly . from the parity check matrix @xmath5 we construct the binary tensor @xmath31 , where @xmath32 if @xmath5 has a row in which the elements @xmath33 are all 1 ( i.e. when the bits @xmath34 are involved in the same parity check ) , and 0 otherwise . the fact that each bit @xmath35 is involved in exactly @xmath36 parity checks
is then expressed by @xmath37 and the parity check equations become @xmath38 , @xmath39 .
decoding now consists in selecting an @xmath24 from @xmath40 , on the basis of its noise statistics , which are fully described by its _ magnetisation _
@xmath41 ( corresponding to the weight in the information theory literature ) .
note that the number of flipped bits in a candidate noise vector @xmath24 is given by @xmath42 .
therefore , we introduce a hamiltonian or cost function for each noise candidate that is negatively proportional to its magnetisation : @xmath43 where we take @xmath44 , such that up to normalisation @xmath45 yields the correct prior for candidate noise vectors generated by the bsc @xcite . then , a vector @xmath24 from @xmath40 with the highest magnetisation ( lowest weight ) is selected as a solution ; this corresponds to maximum a posteriori ( map ) decoding .
we are now interested in the probability that other candidate noise vectors are selected from the parity check set @xmath40 , other than the correct ( i.e. true ) noise vector @xmath25 , for any given combination @xmath46 ; this is termed the _ block error probability_. in order to calculate this probability , we introduce an indicator function : @xmath47\right|_{\beta_1=\beta_2=\beta } \label{eq : delta_def}\ ] ] where @xmath48 the partition functions @xmath49 and @xmath50 differ only in the exclusion of @xmath25 from @xmath51 .
if the true noise @xmath25 has the highest magnetisation of all candidates in the parity check set ( decoding success ) , the boltzmann factor @xmath52 $ ] will dominate the sum over states in @xmath53 in the limit @xmath54 , and @xmath55 .
alternatively , if some other vector @xmath56 has the highest magnetisation of all candidates in the parity check set ( decoding failure ) , its boltzmann factor will dominate both @xmath51 and @xmath53 and @xmath57 .
note that the separate temperatures @xmath58 and @xmath59 , which are put to be equal to @xmath60 in the end , and the powers @xmath61 which are taken to be @xmath62 in the end , have been introduced in order to allow us to determine whether obtained solutions are physical or not .
the power @xmath63 have been introduced to restrict the indicator function results to 0/1 . in principle , this can be done by taking the limit @xmath64 ; however , in section [ sec : solution ] , we show that finite @xmath65 values will be used due to various constraints . to derive the _
average error exponent _ , we take the logarithm of the indicator function average with respect to all possible realisations of true noise vectors @xmath25 , and
the ensemble of regular @xmath66 codes @xmath67 : @xmath68 where @xmath69 and @xmath70~f(\mca ) } { \sum_{\mca}\prod_{i_1=1}^n \delta[\sum_{{i_2}<\cdots<{i_k}}\mca_{\bra i_1\cdots i_k\ket}-j ] } .
\label{eq : tensor_average}\ ] ] to obtain an expression for the _ reliability exponent _ one carries out a similar calculation with one main difference : prior to averaging the indicator function over the ensemble of regular @xmath66 codes @xmath67 , one takes the averaged expression with respect to realisations of true noise vectors @xmath25 to a power @xmath71 which favours code constructions with a low average error probability ( i.e. , @xmath72 ) .
the logarithm of the expression averaged over the ensemble of codes @xmath67 is then divided by @xmath71 to remove the exponent .
the expression calculated is : @xmath73^{r}\right\ket_{\mca } \label{eq : reliability_exp}\ ] ] since there are only discrete degrees of freedom , physically meaningful solutions must have a non - negative entropy , requiring the disorder - averaged entropies of the two partition functions ( [ eq : partition_sums ] ) to be non - negative .
note that due to the order of taking the logarithm vs the various averages , expressions ( [ eq : reliability ] ) and ( [ eq : reliability_exp ] ) are not equivalent to a ( quenched ) disorder - averaged free energy .
using general principles one can show that for general values of @xmath74 and @xmath61 , the disordered - averaged entropies ( with averages taken over the joint distribution of code - constructions @xmath75 , true- and candidate - noise @xmath76 as suggested by ( [ eq : reliability ] ) and ( [ eq : reliability_exp ] ) ) are given , for both calculations ( [ eq : reliability ] ) and ( [ eq : reliability_exp ] ) , by @xmath77 which have to be positive .
using standard statistical physics methods such as in @xcite , we perform the gauge transformation @xmath78 , and the averages over true noise ( [ eq : truenoise_average ] ) and code constructions ( [ eq : tensor_average ] ) . in the case of @xmath79 ,
each quantity carries two indices ( a replica index and another index coming from the power @xmath71 ) ; however , the two indices factorise _ unless an explicit , more complex , symmetry breaking structure is introduced_. here , we do not assume a more complex structure that entangles the two types of indices ; we also assume the simplest replica symmetric scheme @xcite to arrive at the following expression for the average error exponent ( @xmath80 ) , and for the reliability exponent ( optimised @xmath71 ) : @xmath81-j\log~i_2[\pi,\hat{\pi}]+\log~i_3[\hat{\pi } ] \right ] \label{eq : reliability_final_r}\ ] ] where @xmath82 @xmath83 @xmath84^{\lambda_1}\right .
\nonumber \\ & & \hspace*{37mm}\times\left .
\left[\sum_{v=\pm 1}e^{\beta_2 f n^0 v}\prod_{c=1}^j \left(\frac{1+v\hat{y}_c}{2}\right)\right]^{\lambda_2 } \right\ket_{n^0}^r \label{eq : i3_r}\end{aligned}\ ] ] where we have used the short - hand notation @xmath85=@xmath86 . for @xmath80 , functional extremisation of ( [ eq : reliability_final_r ] ) with respect to the densities @xmath87 and @xmath88 results in a closed set of equations ( reminiscent of ` density evolution ' equations @xcite ) : @xmath89\ \delta\left[\hat{y}-\prod_{c=1}^{k-1}y_c\right ] \label{eq : sp1}\ ] ] @xmath90 \delta\left[y-\frac{d_-(\hat{\by};\beta_2)}{d_+(\hat{\by};\beta_2)}\right ] \ket\!\ket ' } { \bra\!\bra1\frac{}{}\ket\!\ket'}. \label{eq : sp2}\ ] ] where @xmath91 \pm[e^{-\beta f n^0}\prod_{c=1}^{j-1}(1-z_c)].\end{aligned}\ ] ] for given @xmath92 in general , solutions to ( [ eq : sp1 ] ) and ( [ eq : sp2 ] ) can only be obtained numerically .
inserting these solutions into ( [ eq : reliability_final_r ] ) we then obtain @xmath93 , which becomes the _ average error exponent _ for @xmath94 , and for @xmath95 .
we must recall , however , that physically meaningful solutions must satisfy the conditions ( [ eq : entropies ] ) stating that the entropies related to the full and the restricted partition sums are non - negative .
we restrict ourselves to regions below the thermodynamic transition where the average case is dominated by the ferromagnetic solution , such that we can fix the system described by @xmath53 in ( [ eq : partition_sums ] ) to the ferromagnetic solution .
this dominance is guaranteed if the following constraint is satisfied @xmath96 it turns out that for given @xmath97 , the largest value of @xmath98 for which ( [ eq : condition3 ] ) is satisfied is given by the simple expression @xmath99 .
hence , in order to maximise @xmath98 , we must look for the smallest value @xmath100 that satisfies the conditions on the non - negativity of the entropies ( [ eq : entropies ] ) .
unfortunately , in general this value @xmath100 can only be obtained numerically .
the value obtained for the average error exponent by this analysis is then given by @xmath101 from ( [ eq : reliability_final_r ] ) . in figure [
fig : kcfinite ] we present the obtained average error exponent as a function of the flip rate for @xmath102 , ( @xmath103 ) and @xmath104 ( @xmath105 ) codes .
we observe that the error exponent indeed converges to zero , as it should , when the flip rate approaches its critical value .
notice the similarity between the equations obtained here and in @xcite in spite of the different starting points .
it has been shown in @xcite that the analysis should be refined in low rate regions by considering a more complex bound .
the refined analysis resulted in tight bounds of the error exponent even in the region of low code - rates , similar to those obtained using expurgated exponent methods .
in the next section we will show that the selection of best codes through the optimisation of the power @xmath71 , in calculating the reliability exponent , provides similar results to those obtained in @xcite .
( 120,55 ) ( 10 , 0)=40 ( 65 , 0)=40 ( 5,25)@xmath106 ( 62,25)@xmath106 ( 40,0)@xmath107 ( 95,0)@xmath107 the average error exponent can be calculated numerically for finite @xmath0 and @xmath144 values ; the average error exponent @xmath106 as function of the flip rate @xmath107 is shown in figure [ fig : mnfinite ] .
on the left , we show results for mn codes of fixed rate @xmath160 with three different sets of parameters @xmath161 ( circles ) , @xmath162 ( diamonds ) and @xmath163 ( upper line ) .
it is interesting to notice that average exponents for either @xmath164 or @xmath165 values coincide with that of the infinite connectivity case ( which can be obtained analytically ) .
this complements other interesting properties of mn codes , to do with their critical flip rate values , that have been obtained previously , distinguishing them from gallager ldpc codes @xcite . on the right ,
we see a comparison between average error exponents of gallager and mn codes ( @xmath105 ) .
the gallager code @xmath104 average error exponent ( circles ) is significantly below the random code @xmath163 value ( thin upper line ) and the equivalent mn code @xmath166 result ( diamonds ) .
( 120,55 ) ( 8 , 5)=40 ( 63 , 5)=40 ( 4,25)@xmath106 ( 61,25)@xmath106 ( 35,2)@xmath107 ( 90,2)@xmath107
whereas for finite density codes solutions for the average error exponent are obtained numerically , in the limit of @xmath108 ( while keeping the rate @xmath109 finite ) one obtains two types of analytic solutions to equations ( [ eq : sp1 ] ) and ( [ eq : sp2 ] ) , which can be verified by substitution . moreover , in this limit one also obtains solutions in the reliability exponent calculation ( [ eq : reliability_exp ] ) , which are generally difficult to obtain for finite @xmath6 and @xmath36 values .
solutions obtained in the average error exponent calculation take the following form : type i : @xmath110~\delta(y-1 ) \nonumber\\ \hat{\pi}(\hat{x},\hat{y})&=&\frac12 \left[\delta(\hat{x}-1)+\delta(\hat{x}+1)\right]~\delta(\hat{y}-1 ) \label{eq : typei}\end{aligned}\ ] ] type ii : @xmath111 \nonumber\\ \hat{\pi}(\hat{x},\hat{y})&=&\delta(\hat{y}-1)~\delta(\hat{x } ) \label{eq : typeii}\end{aligned}\ ] ] with @xmath112 $ ] .
taking @xmath113 and @xmath114 , the average error exponent as obtained from the type i solution is given by @xmath115 we find that the entropies ( [ eq : entropies ] ) are always identically zero , and that the constraint ( [ eq : condition3 ] ) requires that @xmath116 , such that @xmath117 and @xmath118+\log [ e^f+e^{-f}+2 ] \label{eq : kcinfty1}\ ] ] which is exactly the bhattacharyya limit @xcite .
the average error exponent as obtained from the type ii solution is given by @xmath119\right ] + \log [ 2\cosh(f-\beta f\lambda)]-\log 2\cosh f\ ] ] the condition on the entropy @xmath120 is satisfied for all @xmath121 , whereas the condition @xmath122 is violated below the critical ( freezing ) temperature @xmath123 obtained from @xmath124+\log2\cosh[\beta^*f]=0\ ] ] this negative entropy is an artifact of the assumption about the symmetry between replicas , and is easily remedied by considering a ` frozen rsb ' ansatz @xcite . using this ansatz and taking into account condition ( [ eq : condition3 ] ) , the ( frozen ) average error exponent obtained from the type ii solution , is finally given by @xmath125+\frac{j}{k}\log 2-\log 2\cosh f \label{eq : kcinfty2}\ ] ] what remains is to determine whether the type i or type ii solution is physically dominant , by using @xmath106 as a generating function for calculating the related free energies ( through its derivative with respect to @xmath126 ) .
results for the case of @xmath108 are presented in figure [ fig : kctoinfty ] for @xmath127 and @xmath128 . to obtain the reliability exponent we take equations ( [ eq : reliability_final_r])-([eq : i3_r ] ) and optimise with respect to @xmath71 . deriving a general set of equations similar to ( [ eq : sp1],[eq : sp2 ] ) , that can be solved iteratively , is difficult in this case . however , in the limit @xmath108 , we observe that we can restrict the possible solutions of @xmath88 to two different types : type i : @xmath129\delta(\hat{y}-1 ) \label{eq : typeir}\ ] ] type ii : @xmath130 in this case , knowledge of the solution for @xmath88 is sufficient for calculation the reliability exponent ( [ eq : reliability_final_r ] ) .
furthermore , the expression obtained from the type ii solution turns out to be identical to that of the average error exponent ( [ eq : kcinfty2 ] ) . on the other hand ,
the reliability exponent obtained from the type i solution is somewhat different , and takes the form : @xmath131+\frac{1}{r } \log [ \cosh^{r}(f ) + \cosh^{r}((2\beta\lambda-1)f ) ] \ .
\label{eq : kcinfty1r}\ ] ] given the relation ( [ eq : condition3 ] ) and @xmath99 , one obtains @xmath117 , @xmath116 , and the expression reduces to @xmath131+\frac{1}{r } \log [ \cosh^{r}(f ) + 1 ] \ .
\label{eq : kcinfty1rf}\ ] ] optimising the expression with respect to @xmath71 , one obtains a similar expression to the expurgated exponent result @xcite @xmath132 + \frac{1}{r}(1-r ) \ln 2 \right \ } \ , \ ] ] which is also identical to the result obtained for the average bound of the reliability exponent in @xcite .
the reliability exponent is therefore identical to the average error exponent except for very low @xmath133 values as shown in figure [ fig : kctoinfty ] for @xmath127 and @xmath128 ( marked by a dotted line in the two cases considered ) .
( 120,75 ) ( 15,7 ) 0 ( 7,35)@xmath134 ( 60,0)@xmath133 ( 28.5,30)@xmath135 ( 46.5,31.5)@xmath135 ( 60,21.5)@xmath136 ( 42,20.3)@xmath136 ( 73.5,10.8)@xmath137 ( 87.75,10.8)@xmath137
in this section , we extend our treatment of the average error and reliability exponent to regular mn codes @xcite , a variant of ldpc codes .
a regular mn code is defined by the binary @xmath138 matrix @xmath139 $ ] , concatenating two sparse matrices with the @xmath140 matrix @xmath141 assumed invertible .
the @xmath142 matrix @xmath143 has @xmath6 non - zero elements per row and @xmath36 per column while @xmath141 has @xmath144 non - zero elements per row and per column .
the code rate is given by @xmath145 .
the encoding scheme consists of generating a codeword @xmath146 from an ( unbiased ) message vector @xmath147 via @xmath148 . upon sending @xmath149 through the noisy channel the vector @xmath150
is received , where @xmath25 is the true channel noise ( [ noise ] ) .
decoding is carried out by multiplying the received vector @xmath20 by @xmath141 to produce the syndrome vector @xmath151 . in order to reconstruct the original message
, one selects the best estimate @xmath152 , for the true @xmath153 from the parity check set @xmath154 , on the basis of the message / noise statistics . note that since we take the message vector @xmath155 to be unbiased , the selection will only be based on the noise statistics .
since most calculation steps are completely analogous ( although lengthier ) to those of gallager codes , we only state the final general expression for mn codes : @xmath156^{r\lambda_1 } \left[\sum_{\s=\pm}\prod_{c=1}^j\left ( \frac{1+\s\hat{y}_c}{2}\right)\right]^{r\lambda_2 } \nonumber\\&&\hspace*{-40 mm } + \log\int\prod_{l=1}^t\left\{d\hat{\rho}(\hat{x}_l,\hat{y}_l ) \right\ } \left\bra\left [ \sum_{\tau=\pm}e^{\beta_1\tau f n^0}\prod_{l=1}^t \left(\frac{1+\tau\hat{x}_l}{2}\right)\right]^{\lambda_1 } \right . \nonumber\\&&\hspace * { 20 mm } \left.\left.\times \left [ \sum_{\tau=\pm}e^{\beta_2\tau f n^0}\prod_{l=1}^t \left(\frac{1+\tau\hat{y}_l}{2}\right)\right]^{\lambda_2 } \right\ket_{n^0}^r\right\ } \nonumber\end{aligned}\ ] ] with the short - hand notation @xmath85=@xmath86 . as for gallager codes , for @xmath113 , and @xmath114
, @xmath157 becomes the _ average error exponent _ for @xmath80 , while for optimised @xmath71 it becomes the _ reliability exponent_. furthermore , the conditions ( [ eq : entropies ] ) and ( [ eq : condition3 ] ) must always be satisfied . similarly to the case of gallager codes one can derive a set of functional equations ( reminiscent of density evolution equations @xcite ) for @xmath158 and @xmath159 .
the case of @xmath168 is solvable exactly for all transmission rates , and both average and reliability error exponents can be obtained analytically .
the solutions obtained as well as the average and reliability error exponents calculated are identical to those of gallager ldpc codes . retrospectively ,
this is not surprising as both codes become random codes in this limit .
in this paper we suggest a method for direct evaluation of the average and reliability error exponent over the ensemble of ldpc error - correcting codes of given rate and connectivity .
an analytical solution has been obtained , for both gallager and mn codes , using methods of statistical physics , which is in perfect agreement with known results in the limit @xmath169 ( with @xmath133 finite ) .
the results for mn and gallager codes become identical in this limit as both become random codes .
average error exponent results obtained by our method for codes of finite @xmath66 values can not be obtained using traditional approaches used in the information theory community . as expected , they seem to be upper bounded by the @xmath108 curves , but suggest a profoundly different behaviour for gallager and mn ldpc codes .
average error exponent results for gallager codes show a gradually improved performance as the parameters ( @xmath0 ) increase , until they finally coincide with the @xmath108 error exponent result .
the results for mn codes becomes identical to the @xmath108 error exponent result for all @xmath164 or @xmath165 . to some extent
, this is in agreement with previous results obtained for the critical flip rate of mn codes @xcite and is a result of the close - to - random codebook they generate . an interesting feature of the present study is the similarity of our equations to those obtained in @xcite in spite of the different approaches used .
an important advantage offered by the current approach is a potential extension to select high performance codes to obtain _ reliability exponent _ values for ldpc codes of finite connectivity ; obtaining such solutions remains a difficult task and is currently under study .
support from epsrc research grant gr / n63178 ( ds , ns ) , the royal society ( ds , ns , jvm ) , grants - in - aid , mext ( 13680400 and 13780208 ) and jsps ( yk ) is acknowledged .
99 t richardson , a shokrollahi and r urbanke ( 2001 ) , _ ieee trans . on info .
theory _ * 47 * 619 - 637 r vicente , d saad and y kabashima ( 1999 ) _ phys rev e _ * 60 * 5352 - 5366 r vicente , d saad and y kabashima ( 2000 ) _ j phys a _ * 33 * 6527 - 6542 j van mourik , d saad and y kabashima ( 2002 ) _ phys rev e _ * 66 * 026705 t tanaka and d saad , ( 2003 ) submitted in _
phys rev e _
y kabashima , n sazuka , k nakamura and d saad ( 2001 ) _ phys rev e _ ( 2001 ) * 64 * 046113 a montanari and n sourlas _ eur phys j b _ * 18 * 107 - 119 a montanari _ eur phys j b _ * 23 * 121 - 136 s. franz , m. leone , a. montanari and f. ricci - tersenghi , _ phys rev e _ * 66 * 046120 r g gallager ( 2001 ) information theory and reliable communication wiley & sons , ny a j viterbi and j k omura ( 1979 ) ` principles of digital communication and coding ' , mcgraw - hill int ed ( singapore ) d.j.c . mackay ( 1999 )
_ ieee trans .
* 45 * 399 g. miller and d. burshtein ( 2001 ) _ ieee trans .
theory _ * 47 * 2696 h nishimori ( 2001 ) statistical physics of spin glasses and information processing , oxford university press , uk k y m wong and d sherrington ( 1987 ) _ j phys a _ * 20 * l793-l799 | we present a theoretical method for a _ direct _ evaluation of the average and reliability error exponents in low - density parity - check error - correcting codes using methods of statistical physics .
results for the binary symmetric channel ( bsc ) are presented for codes of both finite and infinite connectivity . |
the avian and mammalian ears have similarities and differences .
the innervation pattern of avian tall hair cells resembles that of mammalian inner hair cells in having primarily afferent innervation , suggesting their role as the primary sensor .
in contrast , the innervation pattern of short hair cells in the avian ear is exclusively efferent and resembles that of outer hair cells , suggesting the role of motor element in the cochlear amplifier @xcite . since bending of hair bundles ( including tilting motion of the apical surface of hair cells @xcite ) is the only motile response in short hair cells , such a physiological role requires that the hair bundles of short hair cells exert force between the tectorial membrane and the basilar papilla .
there are marked differences .
one such difference is the absence of pillar cells and the tunnel of corti in the avian ear .
in addition , the presence of electrical tuning @xcite and absence of somatic motility in avian hair cells make them apart .
yet another is the morphology of the tectorial membrane ( tm ) .
is it true that the difference in the motile element of the cochlear amplifier is the reason for the structural difference ? here
this issue is examined by focusing upon the morphology of the tectorial membrane .
the stiffness of the avian tm would be low compared with the mammalian one , being devoid of collagen @xcite .
it shows perforations reminiscent of swiss cheese in electron micrographs due to fixation instead of continuous low density mass as observed in unfixed preparations @xcite . since collagen is largely responsible for the mechanical anisotropy of the mammalian tm @xcite , its absence would make the material properties of the avian tm very different from the mammalian one , in which stiffness gradient and the significance of the resonance frequency has been documented @xcite
the avian tm is tapered toward the abneural side ( fig .
[ fig : anatomy ] ) .
the steepness of tapering increases toward the basal end with higher best frequencies @xcite .
in contrast to the flat and featureless surface of the mammalian tectorial membrane @xcite , the avian tm has an elaborate structure on the side facing the basilar papilla .
a thin veil - like structure descends to the basilar papilla , surrounding each hair cell @xcite .
a dome like recess above each hair cell and each hair bundle makes contact with the tectorial membrane at the dome @xcite .
what is the functional significance of these structural features ? like outer hair cells in the mammalian ear , short hair cells are located nearly above the middle of the basilar membrane , the movement of which in the up - down direction @xcite would effectively transmit bending stress to the short hair cells hair bundles .
in contrast , tall hair cells , which functions as the sensor , are located away from the center of the basilar membrane , at a location seemingly unsuitable to sense the motion of the basilar membrane directly .
thus it is likely short hair cells are involved in amplification by a hair bundle active process @xcite , transmitting the mechanical vibration of the basilar membrane to the hair bundles of tall hair cells .
the problem is how that can happen ? to address this question , it requires to identify the vibrational mode , which involves the basilar membrane , the basilar papilla , and the tm .
since reported measurements are limited to the basilar membrane with mssbauer technique @xcite , we need to assume how other elements could move . since not much else is known for certain , the models considered herein are largely based on morphological observations and rather speculative
. since hair cell bundles are stimulated by shear , it would be logical to assume that shear motion between the basilar papilla and the tm stimulates the hair bundles of tall hair cells , similar to the mammalian ear . however , there are clear differences .
the orientation of hair bundles in the avian ear is not as uniform as in mammalian cochlea .
although the orientation is relatively uniform in the basal region , it forms domain structure in a more apical region @xcite . in the following ,
we examine the mode of motion that is consistent with the bundle alignment in the basal area , then consider a possible mechanism for domain formation in more apical region .
electron micrographs show that hair bundles are oriented primarily in the direction orthogonal to the longitudinal axis of the basilar papilla , particularly in the basal higher frequency region .
this orientation is consistent with the shear between the tectorial membrane and the basilar papilla , analogous to the mammalian ear ( fig .
[ fig : modes ] a ) . however , there are appreciable differences between the two
. the avian ear does not have pillar cells or a tunnel of corti .
thus it may not have well - defined pivoting points for creating shear from the up - down motion of the basilar membrane .
in addition , the tectorial membrane ( tm ) is tapered , making it thicker on top of tall hair cells .
the tm is attached to a wall , where it is thickest , nearer to tall hair cells .
how can such a morphology be consistent with the assumption that short hair cells serve as the cochlear amplifier ?
to explain this puzzle , charles steele suggested that the tm should have resonance near the best frequency @xcite .
indeed , resonance would make it work .
however , a question still remains .
why should it tapered toward short hair cells ?
the avian tm works despite its shape ? is there an advantage of this morphology ?
since so many factors remain uncharacterized , it would be advantageous to examine the simplest possible model to obtain a conceptual understanding of the issue rather than more complex ones , which would require many more details .
for this reason , two - mass models are used here for describing the tm ( fig .
[ fig : modes ] ) .
one has a smaller mass @xmath0 , located at the top of short hair cells and another with a larger mass @xmath1 , located at the top of tall hair cells .
the small mass @xmath0 is connected to the larger mass @xmath1 with a spring with stiffness @xmath2 and the larger mass @xmath1 is connected with the wall by another spring with stiffness @xmath3 .
let @xmath4 the displacement of the bigger mass from its equilibrium position , and that of smaller one be @xmath5 .
let @xmath6 be viscous drag on mass @xmath1 and @xmath7 be negative drag on mass @xmath0 to mimic the amplifying role of short hair cells ( fig .
[ fig : modes ] a ) .
the set of equations is given by , @xmath8 \\ \left(m\frac{d^2}{dt^2}+\eta_2\frac{d}{dt}+k\right)x - k(x - x)=0 , \label{eq : motion}\end{aligned}\ ] ] where @xmath9 $ ] is a periodic external force of angular frequency @xmath10 applied to mass @xmath0 , which is located just above the center of the basilar membrane .
here it is assumed that up - down motion of the basilar membrane results in this external force . following steele @xcite ,
let @xmath11 be the resonance frequency of the tectorial membrane , i.e. @xmath12 and seek a steady state solution . under this assumption ,
the force applied to the mass @xmath1 is plotted for a special case of @xmath13 ( fig .
[ fig : drag ] ) . for a given value of @xmath1 ,
this force is larger for smaller values of the reduced drag @xmath14 where the ratio @xmath15 is smaller .
since the reduced drag is defined by @xmath16 , this parameter is smaller for higher frequency @xmath11 .
thus the plot shows that the force is maximal for smaller ratio @xmath15 at larger frequencies .
this result is consistent with the experimental finding that the sharpness of tapering of the tm increases from the apical to the basal end @xcite .
one can argue that the force relevant to the transduction of tall hair cells would be better related to drag @xmath17 rather than the force @xmath18 .
that does not change the optimal mass ratio because drag is represented by @xmath19 , maximizing at the same mass ratio .
the condition for resonance is not very strict because those peaks exist for @xmath20 .
* domain structure * while hair bundles in the basal area of the basilar papilla are aligned perpendicular to the longitudinal axis of the papilla , those in more apical area are aligned in a patchwork fashion , particularly in the middle between neural and abneural ends . how can this pattern be explained ?
suppose the motion of the tectorial membrane relative to the basilar papilla is perpendicular to the surface of the basilar papilla ( fig .
[ fig : modes ] b ) , how do hair bundles bend ?
each hair bundles is attached to the tectorial membrane at the edge of a dome - like recess @xcite .
the fibrous veils , which descend from the tectorial membrane and reach the basilar papilla , surround each hair cell at its margins .
some parts of the seemingly extendable veils form triangles with hair bundles and parts of the basilar papilla @xcite , which is much denser and therefore expected to be much stiffer .
this structure could couple elastic elongation and shortening of the veils , resulting in a bending of the hair bundles . since bending of a hair
bundle results in local displacement of the tm , bending of the neighboring hair bundles in the same direction is energetically favorable if those hair bundles are oriented in the same direction .
for this reason , up - down motion of the tm favors domain formation in hair bundle orientations .
the size of domains is limited to even out those local deformations of tm to make the average motion in the up - down direction . +
* equations of motion * in this mode of motion , we assume for simplicity , that two masses do not interact directly , but only indirectly being mediated by the basilar papilla , which pivots at the supporting point ( fig .
[ fig : modes ] b ) .
let @xmath21 the distance of the large mass @xmath1 from the pivoting point , and @xmath22 that of the smaller mass @xmath0 .
notice that @xmath23 and for that reason , the torque is more effectively produced by the force applied by the mass @xmath0 .
the equations of motion would be given by , @xmath24,\end{aligned}\ ] ] where @xmath25 is the moment of inertia of the basilar papilla , @xmath26 the angle of rotation of the basilar papilla around its pivoting point and @xmath27 s ( j=1,2,3 ) are either positive or negative , depending on whether they have damping or amplifying effect .
the quantity @xmath28 is the perpendicular displacement of the mass @xmath0 , and @xmath29 that of the mass @xmath1 . here
@xmath30 and @xmath31 represent elasticity due to hair bundles and veils . the quantity @xmath32 represents the elastic elements not associated with hair bundles .
since this is a three body problem , which may not have a periodic solution in general , we examine a special case .
since we are interesting in the mass ratio @xmath15 , we regard the basilar papilla as mediating energy transfer between the smaller mass @xmath0 and the bigger mass @xmath1 , letting @xmath33 .
the longer lever arm of the hair bundles of short hair cells have is advantageous for short hair cells to function as an amplifier . under this condition ,
the force applied to the mass @xmath1 potted against the mass ratio @xmath15 has a maximum for a given @xmath14 , similar to the sliding mode ( fig .
[ fig : jump ] ) .
we assumed that short hair cells provide negative damping whereas tall hair cells do not .
this requires that active force generation at short hair cells is greater than in tall hair cells .
in addition , the source of viscous drag is important .
up - down motion of the tectorial membrane with respect to the basilar papilla requires volume changes in the gap between the tectorial membrane and the basilar papilla , even though changes are small .
since the soft and possibly porous structure of the tectorial membrane may not allow fluid movement at auditory frequencies , the fluid in the gap is likely go through veils surrounds each hair cells , creating an extra viscous drag .
in general , such drag would reduce the energy efficiency .
however , it could be beneficial in this particular situation if it makes the drag of tall hair cells net positive and that of short hair cell net negative .
the modes of motion described above are not mutually exclusive and are likely coexist .
the observation that basal hair bundles are well aligned than more apical bundles and that basal hair bundles have flatter appearance than more apical ones @xcite could indicate a relative significance of the sliding mode at higher frequencies .
it is likely that in more apical area the jumping mode , which is consistent with domain formation , is dominant .
the tapering of the tm is a feature optimizing the sensitivity of tall hair cells in the presence of active motion of the hair bundles of short hair cells .
one could argue that studying the modes of motion in the avian inner ear could be premature : first of all , very few physiological data are available .
the only data on the vibration of the basilar membrane was published in 1987 by gummer at al . using the mssbauer effect @xcite , showing traveling wave and that the amplitude is somewhat smaller compared with the mammalian counterpart .
the mechanical properties of the basilar papilla nor of the tm have not been measured . moreover , studying the modes of motion may not be easy . even in the mammalian cochlea , which has been studied much more extensively
, the details are becoming clearer only recently .
nonetheless the avian ear would be an interesting subject for a couple of reasons .
one such reason is that the avian basilar papilla has some resemblance to the mammalian ear , such as tonotopic organization and differentiated roles of two types of hair cells , and yet it has much simpler morphology , such as the absence of pillar cells and the tunnel of corti besides different tm morphologies , possibly suggesting simpler modes of motion .
in addition , its cochlear amplifier must depend solely on an active process in hair bundles , which may be based on fast adaptation @xcite and possibly , in part , avian prestin @xcite .
also the electrical properties of chick hair cells @xcite are quite different from mammalian hair cells in having an intrinsic electrical tuning mechanism that likely in part shapes the mechanical components , whether they are the hair bundle or prestin @xcite .
for this reason , the mechanical motion in the avian ear does not have to be as sharply tuned as in the mammalian one .
it is possible that the relatively simple structure of the avian ear could possibly be advantageous for identifying the modes of vibration . assuming a local tm resonance @xcite is attractive in view of its role in the mammalian cochlea @xcite . for the jumping mode of motion
, it would be reasonable because hair bundles , which have systematic gradient @xcite , would make a significant contribution to the stiffness .
how realistic is the assumption of resonance for the sliding mode of motion ?
could young s modulus , density , and the dimension of the tm be compatible with the assumption of resonance from the basal end to the apical end ?
can a gradient of the mechanical properties of the tm exist ?
these issues are perhaps experimentally testable .
the difference between short and tall hair cells is another requirement for our models models presented here .
specifically , our models require that the net effect of tall hair cells should be drag and that of short hair cells should be negative drag .
this requires substantial difference in the magnitude of active force production in the two hair cell types .
indeed , some difference in bending responses , both voltage - induced and not , in these hair cell types has been observed @xcite .
however , it remains unclear whether or not the differences are large enough to satisfy our requirement .
there are other possible modes of motion that are not considered here .
the degree of simplification of the present models in describing the tm is obvious .
in addition , there are other factors to consider .
for example , does the basilar membrane move together with the basilar papilla as a solid body or can they possibly have a difference in the phase as well as amplitude ?
photomicrographs show an appreciable space between the basilar papilla with embedded hair cells and the basilar membrane @xcite .
in addition , these two largely parallel structures appear to have separate anchoring points on the neural side .
this feature could possibly allow an external force to drive basilar papilla more than the tectorial membrane .
choe y , magnasco mo , hudspeth aj ( 1998 ) a model for amplification of hair - bundle motion by cyclical binding of ca@xmath34 to mechanoelectrical - transduction channel .
proc natl acad sci usa 95:1532115326 jones gp , lukashkina va , russell ij , elliott sj , lukashkin an ( 2013 ) frequency - dependent properties of the tectorial membrane facilitate energy transmission and amplification in the cochlea .
biophys j 104:13571366 steele cr ( 1997 ) three - dimensional mechanical modeling of the cochlea . in : lewis er ,
long gr , lyon rf , narins pm , steele cr , hecht - poinar e ( eds ) diversity in auditory mechanics , singapore : world scientific , pp . | while the mammalian- and the avian inner ears have well defined tonotopic organizations as well as hair cells specialized for motile and sensing roles , the structural organization of the avian ear is different from its mammalian cochlear counterpart .
presumably this difference stems from the difference in the way motile hair cells function .
short hair cells , whose role is considered analogous to mammalian outer hair cells , presumably depends on their hair bundles , and not motility of their cell body , in providing the motile elements of the cochlear amplifier .
this report focuses on the role of the avian tectorial membrane , specifically by addressing the question , `` why is the avian tectorial membrane tapered from the neural to the abneural direction ? ''
address = otolaryngology & head and neck surgery , stanford university , stanford , ca 93405 , usa , altaddress = nidcd , nih , bethesda , md 20892 , usa address = otolaryngology & head and neck surgery , stanford university , stanford , ca 93405 , usa , altaddress = molecular & cellular physiology , stanford university , stanford , ca 93405 , usa |
_ tensor _ is a generalization of _ vector _ and _ matrix_. a vector is a first - order ( also called one - way or one - mode ) tensor , and a matrix is a second - order tensor .
higher - order tensor arises in many applications such as 3d image reconstruction @xcite , video inpainting @xcite , hyperspectral data recovery @xcite , higher - order web link analysis @xcite , personalized web search @xcite , and seismic data reconstruction @xcite . in this paper
, we focus on the recovery of higher - order tensors that are ( exactly or approximately ) low - rank and have missing entries .
we dub the problem as _ low - rank tensor completion _ ( lrtc ) . the introduced model and algorithm
can be extended in a rather straightforward way to recovering low - rank tensors from their linear measurements .
lrtc can be regarded as an extension of low - rank matrix completion @xcite . to recover a low - rank tensor from its partially observed entries ,
one can unfold it into a matrix and apply a low - rank matrix completion algorithm such as fpca @xcite , apgl @xcite , lmafit @xcite , the alternating direction method @xcite , the @xmath0 minimization method @xcite , and so on .
however , this kind of method utilizes only one mode low - rankness of the underlying tensor .
we are motivated and convinced by the results @xcite that utilizing all mode low - ranknesses of the tensor gives much better performance .
existing methods for lrtc in @xcite employ matrix nuclear - norm minimization and use the singular value decomposition ( svd ) in their algorithms , which become very slow or even not applicable for large - scale problems . to tackle this difficulty , we apply _ low - rank matrix factorization _ to each mode unfolding of the tensor in order to enforce low - rankness and update the matrix factors alternatively , which is computationally much cheaper than svd .
our approach is non - convex , and the sizes of the matrix factors must be specified in the algorithm .
non - convexity makes it difficult for us to predict the performance of our approach in a theoretical way , and in general , the performance can vary to the choices of algorithm and starting point .
we found cyclic updates of the unknown variables in the model to perform well enough .
the sizes of the matrix factors dictate the rank of the recovered tensor .
if they are fixed to values significantly different from the true rank , the recovery can overfit or underfit . on the other hand ,
during the run time of our algorithms , there are simple ways to adaptively adjust the factor sizes . in short , the all - mode matricizations ,
cyclic block minimization , and adaptive adjustment are the building blocks of our approach . before introducing our model and algorithm , we review some notation and tensor operations .
following @xcite , we use bold lower - case letters @xmath1 for vectors , bold upper - case letters @xmath2 for matrices , and bold caligraphic letters @xmath3 for tensors .
the @xmath4-th component of an @xmath5-way tensor @xmath6 is denoted as @xmath7 . for @xmath8
, we define their inner product in the same way as that for matrices , i.e. , @xmath9 the frobenius norm of @xmath6 is defined as @xmath10 a _ fiber _ of @xmath6 is a vector obtained by fixing all indices of @xmath6 except one , and a _ slice _ of @xmath6 is a matrix by fixing all indices of @xmath6 except two .
for example , if @xmath11 has two frontal slices ( with the third index fixed ) @xmath12 then @xmath13^\top$ ] is a mode-1 fiber ( with all but the first indices fixed ) , @xmath14^\top$ ] is a mode-2 fiber ( with all but the second indices fixed ) , and @xmath15^\top$ ] is a mode-3 fiber ( with all but the third indices fixed ) .
its two horizontal ( with the first index fixed ) and two lateral slices ( with the second index fixed ) are respectively @xmath16 the mode-@xmath17 _ matricization _ ( also called _ unfolding _ ) of @xmath18 is denoted as @xmath19 , which is a matrix with columns being the mode-@xmath17 fibers of @xmath6 in the lexicographical order .
take the tensor in for example .
its mode-1 and mode-3 matricizations are respectively @xmath20 relating to the matricization process , we define @xmath21 and @xmath22 to reverse the process , i.e. , @xmath23 .
the @xmath17-rank of an @xmath5-way tensor @xmath6 , denoted as @xmath24 , is the rank of @xmath25 , and we define the rank of @xmath6 as an array : @xmath26 .
we say @xmath6 is ( approximately ) low - rank if @xmath25 is ( approximately ) low - rank for all @xmath17 .
we aim at recovering an ( approximately ) low - rank tensor @xmath27 from partial observations @xmath28 , where @xmath29 is the index set of observed entries , and @xmath30 keeps the entries in @xmath29 and zeros out others .
we apply low - rank matrix factorization to each mode unfolding of @xmath31 by finding matrices @xmath32 such that @xmath33 for @xmath34 , where @xmath35 is the estimated rank , either fixed or adaptively updated .
introducing one common variable @xmath36 to relate these matrix factorizations , we solve the following model to recover @xmath31 @xmath37 where @xmath38 and @xmath39 . in the model , @xmath40 , @xmath34 ,
are weights and satisfy @xmath41 .
the constraint @xmath42 enforces consistency with the observations and can be replaced with @xmath43 if @xmath44 is contaminated by noise with a known frobenius norm equal to @xmath45 . in this paper
, we do not assume the knowledge of @xmath45 and thus use for both noiseless and noisy cases .
the ranks @xmath46 in must be specified , yet we do not assume the knowledge of their true values . to address this issue , we dynamically adjust the rank estimates in two schemes .
one scheme starts from overestimated ranks and then decreases them by checking the singular values of the factor matrices in each mode .
when a large gap between the @xmath47th and @xmath48th singular values of the factors is found , @xmath35 is reduced to @xmath47 .
the other scheme starts from underestimated ranks and then gradually increases them if the algorithm detects slow progress .
we try to solve by cyclically updating @xmath49 , @xmath50 , and @xmath36 .
although a global solution is not guaranteed , we demonstrate by numerical experiments that our algorithm can reliably recover a wide variety of low - rank tensors .
in addition , we show that any limit point of the iterates satisfies the kkt conditions .
the details will be given in section [ sec : alg ] .
our model can be regarded as an extension of the following model @xcite from matrix completion to tensor completion @xmath51 where @xmath52 contains partially observed entries of the underlying ( approximately ) low - rank matrix @xmath53 . if @xmath54 in , i.e. , the underlying tensor @xmath31 is two - way , then it is easy to see that reduces to by noting @xmath55 . the problem is solved in @xcite by a successive over - relaxation ( sor ) method , named as lmafit . although is non - convex , extensive experiments on both synthetic and real - world data demonstrate that solved by lmafit performs significantly better than nuclear norm based convex models such as @xmath56 where @xmath57 denotes the nuclear norm of @xmath58 , defined as the sum of its singular values .
the work @xcite generalizes to the tensor case , and to recover the ( approximately ) low - rank tensor @xmath31 , it proposes to solve @xmath59 where @xmath60 are preselected weights satisfying @xmath41 .
different from our model , the problem is convex , and in @xcite , various methods are applied to solve it such as block coordinate descent method , proximal gradient method , and alternating direction method of multiplier ( admm ) .
the model utilizes low - rankness of all mode unfoldings of the tensor , and as demonstrated in @xcite , it can significantly improve the solution quality over that obtained by solving , where the matrix @xmath58 corresponds to some mode unfolding of the tensor .
the recent work @xcite proposes a more `` square '' convex model for recovering @xmath31 as follows : @xmath61}\| _ * , { \text { subject to } } { \mathcal{p}}_\omega({\boldsymbol{{\mathcal{z}}}})={\boldsymbol{{\mathcal{b}}}},\ ] ] where @xmath62 is a tensor by relabeling mode @xmath63 of @xmath36 to mode @xmath17 for @xmath34 , @xmath64}=\text{reshape}\left({\hat{{\mathbf{z}}}_{(1)},\prod_{n\le j}i_{i_n},\prod_{n > j}i_{i_n}}\right),\ ] ] and @xmath65 and the permutation @xmath4 are chosen to make @xmath66 as close as to @xmath67 .
the idea of reshaping a tensor into a `` square '' matrix has also appeared in @xcite for tensor principal component analysis . as the order of @xmath31 is
no more than three , is the same as with @xmath58 corresponding to some mode unfolding of the tensor , and it may not perform as well as .
however , for a low - rank tensor of more than three orders , it is shown in @xcite that can exactly recover the tensor from far fewer observed entries than those required by .
there are some other models proposed recently for lrtc .
for example , the one in @xcite uses , as a regularization term , a tight convex relaxation of the average rank function @xmath68 and applies the admm method to solve the problem .
the work @xcite directly constrains the solution in some low - rank manifold and employs the riemannian optimization to solve the problem .
different from the above discussed models that use tensor @xmath17-rank , the model in @xcite employs the so - called _ tubal - rank _ based on the recently proposed tensor singular value decomposition ( t - svd ) @xcite . for details about these models ,
we refer the readers to the papers where they are proposed .
the rest of the paper is organized as follows .
section [ sec : phase ] shows the phase transition of our proposed method and some existing ones .
section [ sec : alg ] gives our algorithm with two different rank - adjusting strategies , and the convergence result of the algorithm is given in section [ sec : convg ] . in section [ sec : numerical ] , we compare the proposed method with some state - of - the - art methods for tensor completion on both synthetic and real - world data .
section [ sec : conclusion ] conludes the paper , and finally , section [ sec : fig - table ] shows all figures and tables of our numerical results .
a phase transition plot uses greyscale colors to depict how likely a certain kind of low - rank tensors can be recovered by an algorithm for a range of different ranks and sample ratios .
phase transition plots are important means to compare the performance of different tensor recovery methods .
we compare our method ( called tmac ) to the following three methods on random tensors of different kinds . in section [ sec : numerical ] , we compare them on the real - world data including 3d images and videos . *
matrix completion method for recovering low - rank tensors : we unfold the underlying @xmath5-way tensor @xmath31 along its @xmath5th mode and apply lmafit @xcite to , where @xmath58 corresponds to @xmath69
. if the output is @xmath70 , then we use @xmath71 to estimate @xmath31 . *
nuclear norm minimization method for tensor completion : we apply falrtc @xcite to and use the output @xmath36 to estimate @xmath31 . * square deal method : we apply fpca @xcite to and use the output @xmath36 to estimate @xmath31 .
we call the above three methods as matcomp , falrtc , and squaredeal , respectively .
we chose these methods due to their popularity and code availability .
lmafit has been demonstrated superior over many other matrix completion solvers such as apgl @xcite , svt @xcite , and fpca @xcite ; appears to be the first convex model for tensor completion , and falrtc is the first efficient and also reliable for falrtc was set small , falrtc could also produce solutions of high accuracy . ] solver of ; the work @xcite about squaredeal appears the first work to give theoretical guarantee for low - rank higher - order tensor completion .
we set the stopping tolerance to @xmath72 for all algorithms except fpca that uses @xmath73 since @xmath72 appears too loose for fpca .
note that the tolerances used here are tighter than those in section [ sec : numerical ] because we care more about the models recoverability instead of algorithms efficiency . if the relative error @xmath74 the recovery was regarded as successful , where @xmath75 denotes the recovered tensor .
two gaussian random datasets were tested .
each tensor in the first dataset was 3-way and had the form @xmath76 , where @xmath77 was generated by matlab command ` randn(r , r , r ) ` and @xmath78 by ` randn(50,r ) ` for @xmath79 .
we generated @xmath29 uniformly at random .
the rank @xmath80 varies from 5 to 35 with increment 3 and the sample ratio @xmath81 from 10% to 90% with increment 5% . in the second dataset ,
each tensor was 4-way and had the form @xmath82 , where @xmath77 was generated by matlab command ` randn(r , r , r , r ) ` and @xmath78 by ` randn(20,r ) ` for @xmath83 .
the rank @xmath80 varies from 4 to 13 with increment 1 and sr from 10% to 90% with increment 5% . for each setting
, 50 independent trials were run .
figure [ fig : rate-3 g ] depicts the phase transition plots of tmac , matcomp , and falrtc for the 3-way dataset , and figure [ fig : rate-4 g ] depicts the phase transition plots of tmac and squaredeal for the 4-way dataset .
since lmafit usually works better than fpca for matrix completion , we also show the result by applying lmafit to @xmath84}\|_f^2 , { \text { subject to } } { \mathcal{p}}_\omega({\boldsymbol{{\mathcal{z}}}})={\boldsymbol{{\mathcal{b}}}},\ ] ] where @xmath85}$ ] is the same as that in . from the figures ,
we see that tmac performed much better than all the other compared methods .
note that our figure ( also in figures [ fig:4way - rand ] and [ fig:4way - plaw ] ) for squaredeal looks different from that shown in @xcite , because we fixed the dimension of @xmath31 and varied the rank @xmath80 while @xcite fixes @xmath86 to an array of very small values and varies the dimension of @xmath31 .
the solver and the stopping tolerance also affect the results .
for example , the `` square '' model solved by lmafit gives much better results but still worse than those given by tmac .
in addition , figure [ fig : diff - mode ] depicts the phase transition plots of tmac utilizing 1 , 2 , 3 , and 4 modes of matricization on the 4-way dataset .
we see that tmac can recover more tensors as it uses more modes .
this section tests the recoverability of tmac , matcomp , falrtc , and squaredeal on two datasets in which the tensor factors have uniformly random entries . in the first dataset , each tensor had the form @xmath76 , where @xmath77 was generated by matlab command ` rand(r , r , r)-0.5 ` and @xmath78 by ` rand(50,r)-0.5 ` .
each tensor in the second dataset had the form @xmath82 , where @xmath77 was generated by matlab command ` rand(r , r , r , r)-0.5 ` and @xmath78 by ` rand(20,r)-0.5 ` .
figure [ fig:3way - rand ] shows the recoverability of each method on the first dataset and figure [ fig:4way - rand ] on the second dataset .
we see that tmac with both rank - fixing and rank - increasing strategies performs significantly better than the other compared methods .
this section tests the recoverability of tmac , matcomp , falrtc , and squaredeal on two more difficult synthetic datasets . in the first dataset , each tensor had the form @xmath76 , where @xmath77 was generated by matlab command ` rand(r , r , r ) ` and @xmath78 by ` orth(randn(50,r))*diag([1:r].^(-0.5 ) ) ` .
note that the core tensor @xmath77 has nonzero - mean entries and each factor matrix has power - law decaying singular values .
this kind of low - rank tensor appears more difficult to recover compared to the previous random low - rank tensors .
each tensor in the second dataset had the form @xmath82 , where @xmath77 was generated by matlab command ` rand(r , r , r , r ) ` and @xmath78 by ` orth(randn(20,r))*diag([1:r].^(-0.5 ) ) ` .
for these two datasets , tmac with rank - decreasing strategy can never decrease @xmath35 to the true rank and thus performs badly .
figure [ fig:3way - plaw ] shows the recoverability of each method on the first dataset and figure [ fig:4way - plaw ] on the second dataset .
again , we see that tmac with both rank - fixing and rank - increasing strategies performs significantly better than the other compared methods .
we apply the alternating least squares method to . since the model needs an estimate of @xmath86 , we provide two strategies to dynamically adjust the rank estimates .
the model is convex with respect to each block of the variables @xmath87 and @xmath36 while the other two are fixed .
hence , we cyclically update @xmath87 and @xmath36 one at a time .
let @xmath88 be the objective of .
we perform the updates as [ eq : update ] @xmath89 note that both and can be decomposed into @xmath5 independent least squares problems , which can be solved in parallel .
the updates in can be explicitly written as [ eq : exupdate ] @xmath90 where @xmath91 denotes the moore - penrose pseudo - inverse of @xmath92 , @xmath93 is the complement of @xmath29 , and we have used the fact that @xmath94 in . no matter how @xmath95 is computed , only the products @xmath96 , affect @xmath36 and thus the recovery @xmath31 .
hence , we shall update @xmath49 in the following more efficient way @xmath97 which together with gives the same products @xmath98 , as those by and according to the following lemma , which is similar to lemma 2.1 in @xcite .
we give a proof here for completeness .
[ lem : equiv ] for any two matrices @xmath99 , it holds that @xmath100 let @xmath101 be the compact svd of @xmath102 , i.e. , @xmath103 , and @xmath104 is a diagonal matrix with all positive singular values on its diagonal .
it is not difficult to verify that @xmath105 . then @xmath106 where we have used @xmath107 for any @xmath108 of appropriate sizes in the second equality . on the other hand , @xmath109 hence , we have the desired result .
the problem requires one to specify the ranks @xmath46 . if they are fixed , then a good estimate is important for to perform well .
too small @xmath35 s can cause underfitting and a large recovery error whereas too large @xmath35 s can cause overfitting and large deviation to the underlying tensor @xmath31 . since we do not assume the knowledge of @xmath86 , we provide two schemes to dynamically adjust the rank estimates @xmath46 . in our algorithm , we use parameter @xmath110 to determine which one of the two scheme to apply . if @xmath111 , the rank - decreasing scheme is applied to @xmath35 ; if @xmath112 , the rank - increasing scheme is applied to @xmath35 ; otherwise , @xmath35 is fixed to its initial value .
this scheme starts from an input overestimated rank , i.e. , @xmath113 .
following @xcite , we calculate the eigenvalues of @xmath114 after each iteration , which are assumed to be ordered as @xmath115 .
then we compute the quotients @xmath116 .
suppose @xmath117 if @xmath118 which means a `` big '' gap between @xmath119 and @xmath120 , then we reduce @xmath35 to @xmath47 .
assume that the svd of @xmath121 is @xmath122 .
then we update @xmath95 to @xmath123 and @xmath124 to @xmath125 , where @xmath126 is a submatrix of @xmath127 containing @xmath47 columns corresponding to the largest @xmath47 singular values , and @xmath128 and @xmath129 are obtained accordingly .
we observe in our numerical experiments that this rank - adjusting scheme generally works well for exactly low - rank tensors .
because , for these tensors , @xmath130 can be very large and easy to identify , the true rank is typically obtained after just one rank adjustment . for approximately low - rank tensors , however , a large gap may or may not exist , and when it does not , the rank overestimates will not decrease . for these tensors ,
the rank - increasing scheme below works better .
this scheme starts an underestimated rank , i.e. , @xmath131 .
following @xcite , we increase @xmath35 to @xmath132 at iteration @xmath133 if @xmath134 which means `` slow '' progress in the @xmath35 dimensional space along the @xmath17-th mode . here
, @xmath135 is a positive integer , and @xmath136 is the maximal rank estimate . let the economy qr factorization of @xmath137 be @xmath138 .
we augment @xmath139 $ ] where @xmath140 has @xmath135 randomly generated columns and then orthonormalize @xmath141 .
next , we update @xmath142 to @xmath143 and @xmath144 $ ] , where @xmath145 is an @xmath146 _ zero _ matrix , appending any matrix of appropriate size after @xmath95 does not make any difference . ] .
numerically , this scheme works well not only for exactly low - rank tensors but also for approximately low - rank ones .
however , for exactly low - rank tensors , this scheme causes the algorithm to run longer than the rank - decreasing scheme .
figure [ fig : dec - inc ] shows the performance of our algorithm equipped with the two rank - adjusting schemes . as in section [ sec : randtest ] , we randomly generated @xmath31 of size @xmath147 with @xmath148 , and we started tmac with 25% overestimated ranks for rank - decreasing scheme and 25% underestimated ranks for rank - increasing scheme . from figure
[ fig : dec - inc ] , we see that tmac with the rank - decreasing scheme is better when @xmath80 is small while tmac with the rank - increasing scheme becomes better when @xmath80 is large .
we can also see from the figure that after @xmath35 is adjusted to match @xmath149 , our algorithm converges linearly .
in general , tmac with the rank - increasing scheme works no worse than it does with the rank - decreasing scheme in terms of solution quality no matter the underlying tensor is exactly low - rank or not . numerically , we observed that if the rank is relatively low , tmac with the rank - decreasing scheme can adjust the estimated rank to the true one within just a few iterations and converges fast , while tmac with the rank - increasing scheme may need more time to get a comparable solution .
however , if the underlying tensor is approximately low - rank , or it is exactly low - rank but the user concerns more on solution quality , tmac with the rank - increasing scheme is always preferred , and small @xmath135 s usually give better solutions at the cost of more running time .
the above discussions are distilled in algorithm [ alg : als ] .
after the algorithm terminates with output @xmath150 , we use @xmath151 to estimate the tensor @xmath31 , which is usually better than @xmath36 when the underlying @xmath31 is only approximately low - rank or the observations are contaminated by noise , or both .
@xmath29 , @xmath28 , and @xmath152 with @xmath153 .
@xmath155 with @xmath156 . in algorithm
[ alg : als ] , we can have different rank - adjusting schemes , i.e. , different @xmath110 , for different modes .
for simplicity , we set @xmath111 or @xmath112 uniformly for all @xmath17 in our experiments .
introducing lagrangian multiplier @xmath157 for the constraint @xmath42 , we write the lagrangian function of @xmath158 letting @xmath159 and @xmath160 , we have the kkt conditions [ eq : kkt ] @xmath161 our main result is summarized in the following theorem .
[ thm : main ] suppose @xmath162 is a sequence generated by algorithm [ alg : als ] with fixed @xmath35 s and fixed positive @xmath40 s .
let @xmath163 .
then any limit point of @xmath164 satisfies the kkt conditions in .
our proof of theorem [ thm : main ] mainly follows @xcite .
the literature has work that analyzes the convergence of alternating minimization method for non - convex problems such as @xcite .
however , to the best of our knowledge , none of them implies our convergence result . before giving the proof
, we establish some important lemmas that are used to establish our main result .
we begin with the following lemma , which is similar to lemma 3.1 of @xcite .
[ lem : diff1 ] for any matrices @xmath165 and @xmath166 of appropriate sizes , letting @xmath167 then we have @xmath168 from lemma [ lem : equiv ] , it suffices to prove by letting @xmath169 .
assume the compact svds of @xmath170 and @xmath102 are @xmath171 and @xmath172 .
noting @xmath173 and @xmath174 , we have @xmath175 similarly , noting @xmath176 and @xmath177 , we have @xmath178\nonumber\\ = & { \mathbf{u}}_{\tilde{a}}{\mathbf{u}}_{\tilde{a}}^\top({\mathbf{c}}-{\mathbf{a}}{\mathbf{b}})({\mathbf{i}}-{\mathbf{v}}_b{\mathbf{v}}_b^\top)\label{eq2}\end{aligned}\ ] ] where the third equality is from . summing and gives @xmath179 since @xmath180 is orthogonal to @xmath181 , we have @xmath182 in addition , note @xmath183 and @xmath184 hence , @xmath185 and thus @xmath186 then can be shown by noting @xmath187 this completes the proof .
we also need the following lemma .
[ lem : range ] for any two matrices @xmath102 and @xmath166 of appropriate sizes , it holds that @xmath188^\dagger ( { \mathbf{c}}{\mathbf{b}}^\top)^\top{\mathbf{c}}\right)\cr = & \ { \mathcal{r}}_r\left(\big[({\mathbf{c}}{\mathbf{b}}^\top({\mathbf{b}}{\mathbf{b}}^\top)^\dagger)^\top ( { \mathbf{c}}{\mathbf{b}}^\top({\mathbf{b}}{\mathbf{b}}^\top)^\dagger)\big]^\dagger \big[{\mathbf{c}}{\mathbf{b}}^\top({\mathbf{b}}{\mathbf{b}}^\top)^\dagger\big]^\top{\mathbf{c}}\right),\label{range - row}\end{aligned}\ ] ] where @xmath189 and @xmath190 denote the column and row space of @xmath92 , respectively . following the proof of lemma [ lem : equiv ] , we assume the compact svd of @xmath102 to be @xmath101 .
then @xmath191 for any vector @xmath192 of appropriate size , there must be another vector @xmath193 such that @xmath194 or equivalently @xmath195 .
hence @xmath196 , which indicates @xmath197 . in the same way
, one can show the reverse inclusion , and hence @xmath198 .
the result can be shown similarly by noting that @xmath199^\dagger ( { \mathbf{c}}{\mathbf{b}}^\top)^\top{\mathbf{c}}&={\mathbf{u}}{\boldsymbol{\sigma}}{\mathbf{v}}^\top({\mathbf{c}}^\top{\mathbf{c}})^\dagger{\mathbf{v}}{\mathbf{v}}^\top{\mathbf{c}}^\top{\mathbf{c}},\\ \big[({\mathbf{c}}{\mathbf{b}}^\top({\mathbf{b}}{\mathbf{b}}^\top)^\dagger)^\top ( { \mathbf{c}}{\mathbf{b}}^\top({\mathbf{b}}{\mathbf{b}}^\top)^\dagger)\big]^\dagger \big[{\mathbf{c}}{\mathbf{b}}^\top({\mathbf{b}}{\mathbf{b}}^\top)^\dagger\big]^\top{\mathbf{c}}&={\mathbf{u}}{\boldsymbol{\sigma}}^{-1}{\mathbf{v}}^\top({\mathbf{c}}^\top{\mathbf{c}})^\dagger{\mathbf{v}}{\mathbf{v}}^\top{\mathbf{c}}^\top{\mathbf{c}}.\end{aligned}\ ] ] this completes the proof . according to lemma [ lem : range ] ,
it is not difficult to get the following corollary .
[ cor : diff ] for any matrices @xmath165 and @xmath166 of appropriate sizes , let @xmath200 if the compact svds of @xmath170 and @xmath102 are @xmath171 and @xmath172 , then we have .
we are now ready to prove theorem [ thm : main ] .
let @xmath201 be a limit point of @xmath164 , and thus there is a subsequence @xmath202 converging to @xmath203 . according to , we have @xmath204 , and @xmath205 .
hence , and hold at @xmath206 for all @xmath207 and thus at @xmath203 . from lemma [ lem : diff1 ] , it follows that @xmath208 in addition , it is not difficult to verify @xmath209 summing up and and observing that @xmath210 is lower bounded by _ zero _ , we have @xmath211 and thus @xmath212 for each @xmath17 and @xmath207 , let the compact svds of @xmath213 and @xmath214 be @xmath215 and @xmath216 . letting @xmath217 , and @xmath218 in , we have @xmath219 which together with gives @xmath220 since @xmath221 and @xmath222 is bounded , then right multiplying @xmath223 for @xmath224 on both sides of yields
@xmath225 which indicates that is satisfied at @xmath203 . from and , we have @xmath226 which together with the boundedness of @xmath227 and @xmath228 gives @xmath229 and thus is satisfied at @xmath203 .
this completes the proof .
this section tests algorithm [ alg : als ] , tmac , for solving . to demonstrate its effectiveness , we compared it with matcomp and falrtc ( see section [ sec : phase ] ) on real - world data .
the parameters @xmath230 in were uniformly set to @xmath231 at the beginning of tmac . during the iterations
, we either fixed them or dynamically updated them according to the fitting error @xmath232 the smaller @xmath233 is , the larger @xmath40 should be . specifically ,
if the current iterate is @xmath234 , we set @xmath235^{-1}}{\sum_{i=1}^n\big[{\mathbf{fit}}_i({\mathbf{x}}_i^k{\mathbf{y}}_i^k)\big]^{-1}},\ n=1,\ldots , n.\ ] ] as demonstrated below , dynamic updating @xmath40 s can improve the recovery quality for tensors that have better low - rankness in one mode than others .
tmac was terminated if one of the following conditions was satisfied for some @xmath207 @xmath236 where @xmath237 is a small positive value specified below .
the condition checks the relative change of the overall fitting , and is satisfied if the weighted fitting is good enough .
this section compares tmac , matcomp , and falrtc on a @xmath238 brain mri data , which has been used in @xcite .
the data is approximately low - rank : for its three mode unfodings , the numbers of singular values larger than 0.1% of the largest one are 28 , 33 , and 29 , respectively .
one slice of the data is shown in figure [ fig : mri ] .
we tested all three methods on both noiseless and noisy data .
specifically , we added scaled gaussian noise to the original data to have @xmath239 and made noisy observations @xmath240 , where @xmath241 denotes the maximum absolute value of @xmath31 , and the entries of @xmath242 follow idendically independent standard gaussian distribution .
we ran all the algorithms to maximum 1000 iterations .
the stopping tolerance was set to @xmath243 for tmac and lmafit .
for falrtc , @xmath244 was set since we found @xmath245 was too loose .
both tmac and lmafit used the rank - increasing strategy .
for tmac , we initialized @xmath246 and set @xmath247 , and for lmafit , we set initial rank @xmath248 , increment @xmath249 , and maximal rank @xmath250 .
we tested tmac with fixed parameters @xmath251 and also dynamically updated ones by starting from @xmath252 .
the smoothing parameter for falrtc was set to its default value @xmath253 and weight parameters set to @xmath254 table [ table : mri ] shows the average relative errors and running times of five independent trials for each setting of @xmath255 and sr .
figure [ fig : mri ] shows one noisy masked slice and the corresponding recovered slices by different methods with the setting of @xmath256 and @xmath257 . from the results , we see that tmac consistently reached lower relative errors than those by falrtc and cost less time .
matcomp used the least time and could achieve low relative error as sr is large .
however , for low sr s ( e.g. , sr=10% ) , it performed extremely bad , and even we ran it to more iterations , say 5000 , it still performed much worse than tmac and falrtc .
in addition , tmac using fixed @xmath40 s worked similarly well as that using dynamically updated ones , which should be because the data has similar low - rankness along each mode .
this section compares tmac , matcomp , and falrtc on a @xmath258 hyperspectral data , one slice of which is shown in figure [ fig : hyp ] .
this data is also approximately low - rank : for its three mode unfoldings , the numbers of singular values larger than 1% of the largest one are 19,19 , and 4 , respectively .
however , its low - rank property is not as good as that of the above mri data .
its numbers of singular values larger than 0.1% of the largest one are 198 , 210 , and 18 .
hence , its mode-3 unfolding has better low - rankness , and we assigned larger weight to the third mode . for falrtc , we set @xmath259 . for tmac , we tested it with fixed weights @xmath259 and also with dynamically updated ones by starting from @xmath260 .
all other parameters of the three methods were set as the same as those used in the previous test . for each setting of @xmath255 and sr
, we made 5 independent runs .
table [ table : hyp ] reports the average results of each tested method , and figure [ fig : hyp ] shows one noisy slice with 90% missing values and 5% gaussian noise , and the corresponding recovered slices by the compared methods . from the results
, we see again that tmac outperformed falrtc in both solution quality and running time , and matcomp gave the largest relative errors at the cost of least time .
in addition , tmac with dynamically updated @xmath40 s worked better than that with fixed @xmath40 s as sr was relatively large ( e.g. , sr=30% , 50% ) , and this should be because the data has much better low - rankness along the third mode than the other two .
as sr was low ( e.g. , sr=10% ) , tmac with dynamically updated @xmath40 s performed even worse .
this could be explained by the case that all slices might have missing values at common locations ( i.e. , some mode-3 fibers were entirely missing ) as sr was low , and in this case , the third mode unfolding had some entire columns missing . in general , it is impossible for any matrix completion solver to recover an entire missing column or row of a matrix , and thus putting more weight on the third mode could worsen the recovery . that also explains why matcomp gave much larger relative errors than those by falrtc but the slice recovered by matcomp looks better than that by falrtc in figure [ fig : hyp ] .
note that there are lots of black points on the slice given by matcomp , and these black points correspond to missing columns of the third mode unfolding .
therefore , we do not recommend to dynamically update @xmath40 s in tmac when sr is low or some fibers are entirely missing . in this section ,
we compared tmac , matcomp , and falrtc on both grayscale and color videos .
the grayscale videosrbecker / escalator_data.mat ] has 200 frames with each one of size @xmath261 , and the color video has 150 frames with each one of size @xmath262 .
we treated the grayscale video as a @xmath263 tensor and the color video as three @xmath264 tensors , one for each channel .
for the grayscale video , the numbers of singular values larger than 1% of the largest one for each mode unfolding are 79 , 84 , and 35 .
hence , its rank is not low , and it is relatively difficult to recover this video .
the color video has lower rank . for each of its three channel tensors ,
the numbers of singular values larger than 1% of the largest one are about , @xmath265 , and @xmath266 respectively for three channel tensors . ]
50 , 50 , and 24 . during each run , the three channel tensors had the same index set of observed entries , which is the case in practice , and we recovered each channel independently .
we set @xmath267 for falrtc and tmac , while the latter was tested with both fixed @xmath40 s and dynamically updated one by .
all the other parameters of the test methods were set as the same as those in the previous test .
the average results of 5 independent runs were reported in table [ table : grayvideo ] for the grayscale video and table [ table : colorvideo ] for the color video .
figure [ fig : grayvideo ] shows one frame of recovered grayscale video by each method and figure [ fig : colorvideo ] one frame of recovered color video . from the tables
, we see that the comparisons are similar to those for the previous hyperspectral data recovery .
we have proposed a new method for low - rank tensor completion .
our model utilizes low - rank matrix factorizations to all - mode unfoldings of the tensor .
synthetic data tests demonstrate that our model can recover significantly more low - rank tensors than two nuclear norm based models and one model that performs low - rank matrix factorization to only one mode unfolding .
in addition , numerical results on 3d images and videos show that our method consistently produces the best solutions among all compared methods and outperforms the nuclear norm minimization method in both solution quality and running time .
numerically , we have observed that our algorithm converges fast ( e.g. , linear convergence in figure [ fig : dec - inc ] ) .
papers @xcite demonstrate that the sor technique can significantly accelerate the progress of alternating least squares .
however , we did not observe any acceleration applying the same technique to our algorithm . in the future
, we will explore the reason and try to develop other techniques to accelerate our algorithm .
we also plan to incorporate the objective term in to enrich , if the underlying low - rank tensor has more than three orders .
we give all figures and tables of our numerical results in this section . + + + + + + + [ cols="^,^ " , ]
the authors thank two anonymous referees and the associate editor for their very valuable comments .
y. xu is supported by nsf grant eccs-1028790 .
r. hao and z. su are supported by nsfc grants 61173103 and u0935004 .
r. hao s visit to ucla is supported by china scholarship council .
w. yin is partially supported by nsf grants dms-0748839 and dms-1317602 , and aro / arl muri grant fa9550 - 10 - 1 - 0567 . , _ a block coordinate descent method for regularized multi - convex optimization with applications to nonnegative tensor factorization and completion _
, siam journal on imaging sciences , 6 ( 2013 ) , pp . | higher - order low - rank tensors naturally arise in many applications including hyperspectral data recovery , video inpainting , seismic data reconstruction , and so on . we propose a new model to recover a low - rank tensor by simultaneously performing low - rank matrix factorizations to the all - mode matricizations of the underlying tensor .
an alternating minimization algorithm is applied to solve the model , along with two adaptive rank - adjusting strategies when the exact rank is not known .
phase transition plots reveal that our algorithm can recover a variety of synthetic low - rank tensors from significantly fewer samples than the compared methods , which include a matrix completion method applied to tensor recovery and two state - of - the - art tensor completion methods .
further tests on real - world data show similar advantages .
although our model is non - convex , our algorithm performs consistently throughout the tests and gives better results than the compared methods , some of which are based on convex models . in addition , subsequence convergence of our algorithm can be established in the sense that any limit point of the iterates satisfies the kkt condtions .
yangyang xu ruru hao wotao yin zhixun su |
optical lattices in 3d generated by counterpropagating laser fields provide a periodic array of microtraps for cold atoms , and implement quantum lattice models governed by hubbard hamiltonians @xcite .
this is of interest both in the study of strongly correlated systems from condensed matter physics , and in the implementation of quantum information processing .
two key elements of cold atom dynamics in 3d optical lattices are : ( i ) the lattice spacing is bounded below by the wavelength @xmath1 . therefore , all the relevant time , energy and temperature scales are determined by the recoil energy , @xmath2 , where @xmath3 is the wavenumber , and @xmath4 is the the mass of the atom .
in particular , the hopping matrix element between neighbouring sites within the region of validity of the single - band hubbard model is limited by @xmath5 .
this limits all corresponding timescales , for example the timescale over which entanglement can be generated using exchange interactions @xcite .
( ii ) varying the laser parameters provides _ global _ control of the lattice potential , but development of tools for local operations and readout remains a major challenge in light of applications in quantum computing and quantum simulation @xcite .
here we show how adiabatic potentials can be used to address these issues by creating addressable , state - dependent sub - wavelength lattices .
each lattice period can be divided into a set of wells , which can be frequency - addressed via their offset energies , reminiscent of quantum computing schemes with electrons in arrays of electrically gated quantum dots @xcite . in recent seminal experiments @xcite ,
optical superlattices have been used to provide addressability and to entangle atoms in a series of double - wells in one parallel operation .
our goal here is to achieve similar control on sub - wavelength scales by coupling optical lattice potentials with dressing fields , with the benefit that all corresponding timescales are significantly faster , and energy and temperature scales significantly higher .
single particle effects of such dressed potentials @xcite have been studied , e.g. , in the context of lithography @xcite , laser cooling @xcite , and radio frequency ( rf)-dressed potentials on atom chips @xcite . here
we show how the adiabatic potentials , created by coupling optical lattice potentials with additional dressing fields , can be used for subwavelength control in the context of manybody physics .
we will demonstrate how adiabatic potentials can be used to generate state - dependent , addressable subwavelength lattices .
in such a subwavelength lattice , each lattice period is subdivided into several potential wells , for which the relative well depth can be tuned by either applying external magnetic or electric fields , or by adjusting the laser parameters , e.g. the rabi - frequency and detuning .
atoms loaded into different wells within the subwavelength structure can also be frequency - addressed via their different offset energies . in view of recent seminal progress with alkaline - earth - like atoms , e.g. in the context of developing optical clocks @xcite , in many - body cold atom physics @xcite , and in applications in quantum information @xcite
, we investigate , in particular , the possibility for generating sub - wavelength potentials with these species .
the existence of metastable triplet states makes possible the production of near - resonant optical lattices , which allow for the creation of state - dependent sub - wavelength lattices .
we show how this can be used to generate and manipulate entanglement , making use of the spin dependence and/or shorter tunneling times within the subwavelength structure .
of course , in the context of adiabatic potentials , the question of non - adiabatic losses will always arise .
we analyse these losses in detail in the present context , and conclude that they can always be suppressed at the expense of using additional laser power .
this paper is organised as follows : we first give a general description in sec .
[ sec2 ] of two schemes to create adiabatic potentials with two - level atoms . in sec .
[ sec3 ] , we extend the idea to multi - level atoms , describing the additional control that is present when these lattices are generated for real atoms . in particular we discuss possibilities for state - dependent potentials with alkaline - earth - like atoms . in sec
. [ sec4 ] , we add interactions between atoms and give examples for possible applications of subwavelength lattices in quantum simulation and quantum information .
this includes a discussion of how entanglement within the subwavelength structure can be generated and controlled . in sec .
[ sec5 ] , we analyse the landau - zener - type loss rate from the adiabatic potentials in detail and discuss how it should be suppressed . in appendix
a we give additional details on the clock transition in alkaline - earth - like atoms , which are essential for the generation of subwavelength lattices with these atoms .
in this section we introduce the basic ideas behind addressable subwavelength potentials for two - level atoms .
this will be generalised to multi - level atoms in sec .
[ sec3 ] , where we also discuss their implementation with alkaline - earth - like and alkali atoms , and how they can be made internal state- ( or qubit- ) dependent .
we discuss these ideas on the single - particle level initially , with interactions between atoms to be added in sec .
[ sec4 ] . as shown in fig .
1a , our goal is to divide each period of a standard @xmath6 optical lattice into several potential wells with controllable barrier heights and well depths in this subwavelength structure .
a laser can then be used to address atoms on different sites independently based on energy selectivity , coupling to an unperturbed reference state .
in addition , we can control in this way the tunneling of atoms between adjacent wells in the subwavelength structure . the simplest way to achieve this goal
is shown in fig .
we consider a two - level atom with two long - lived internal states @xmath7 and @xmath8 .
we assume that we can generate counter - oscillating _ off - resonant _
potentials for these two internal states , for which the implemenation will be discussed in sec .
we then couple the internal states with additional dressing fields , which generate avoided crossings in the background potentials , and thus a subwavelength structure . here
, the well depth and barrier heights can be controlled by varying the detunings and rabi frequencies of the dressing fields , which shifts the locations of the avoided crossings . in fig .
1b , we show the simplest case with isolated avoided crossings .
it is also possible to generate background potentials with _ near - resonant _ coupling between @xmath8 and @xmath7 with a rabi frequency varying sinusoidally in space ( e.g. , as generated by a standing wave light field ) .
this results in counter - oscillating potentials due to ac - stark splitting , which can then be coupled by a series of dressing fields to produce the desired subwavelength structure ( see fig .
this scheme is motivated , in particular , by alkaline - earth - like atoms ( as discussed in sec .
[ sec3 ] ) .
we now discuss the details of the generation of subwavelength lattices with off - resonant background potentials for a two - level atom . in the rotating wave approximation with respect to the optical frequencies ,
we may write the single atom hamiltonian as : @xmath9 with @xmath10 here , @xmath11 is the kinetic energy , @xmath12 describes the background potentials , and @xmath13 contains the dressing fields with rabi frequencies @xmath14 . for dressing fields produced as an equally - spaced comb of sidebands ,
the detunings can be written as @xmath15 , where @xmath16 is the detuning of the first dressing field . in the off - resonant scheme ,
the background potentials for the two internal states are generated by standing wave fields and are opposite to one another .
for this , we may take @xmath17 , where @xmath3 is the wavenumber of the laser generating the background potential .
generation of the adiabatic potentials is based on the validity of a born - oppenheimer - type assumption .
this involves the fact that the kinetic energy of the atoms is small on a scale given by the separation of the resulting adiabatic potentials .
the wavefunction @xmath18 of a single atom satisfies the schrdinger equation , @xmath19 if we omit the kinetic energy term from the hamiltonian , we obtain an equation for adiabatic eigenstates , @xmath20 , @xmath21 where the @xmath20 are periodic functions and can be expanded as a fourier series . for a two - level atom
, it takes the form @xmath22.\ ] ] this results in a floquet eigenvalue equation for @xmath23 , which plays the role of the adiabatic potentials , @xmath24 the complete wavefunction can then be expanded in a basis of these adiabatic eigenstates , which play the role of born - oppenheimer channel functions , @xmath25 resulting in the equation @xmath26c_{\alpha}(x , t)+\sum_{\beta}h_{m}^{\alpha\beta}c_{\beta}(x , t),\ ] ] where @xmath27 .
the equations for different @xmath28 decouple provided that the mixing terms coupling the channels are small .
we expect this to be the case provided that the kinetic energy of the atom in the lattice , which is typically on the order of the recoil energy , is much smaller than the separation between the adiabatic potentials , which is characterised by laser parameters , e.g. the rabi frequency and detuning . below we will initially neglect these terms , and in sec .
[ sec5 ] we will discuss in detail non - adiabatic processes , and how they should be suppressed in experimental implementations . .
the potential in the figure is generated with the off - resonant scheme with the parameters : @xmath29khz , @xmath30khz , @xmath31khz .
similar potentials can also be generated with the near - resonant scheme .
( b ) subwavelength double - well structure generated with the off - resonant scheme with the parameters : @xmath29khz , @xmath30khz , @xmath31khz , @xmath32khz , @xmath33khz .
the relative depth of the wells in the subwavelength structure can be tuned by changing @xmath34 or @xmath35 .
( c ) subwavelength double - well structure generated with the near - resonant scheme with the parameters : @xmath36khz , @xmath37khz , @xmath38khz , @xmath39khz .
the central barrier height in the lower dressed potential can be adjusted by changing either @xmath35 or @xmath40 .
( d ) subwavelength double - well structure generated with the off - resonant scheme with @xmath41 dressing fields : @xmath29khz , @xmath42khz , @xmath43khz , @xmath44khz .
, width=302 ] in the case of two - level atoms , the eigenvalues @xmath23 of the equations appear in pairs , and may be identified as the upper @xmath45 and the lower @xmath46 dressed potentials for a given floquet manifold ( see fig .
the corresponding eigenstates for the adiabatic potentials in general have components from different floquet manifolds . for the simple case of one dressing field in the off - resonant scheme ( with only @xmath40 non - zero ) , the different manifolds in the floquet basis decouple , and we only need to diagonalise a two - by - two matrix . in this case , the expression for the adiabatic potential and its corresponding states are : @xmath47,\label{eq : dressedstates}\end{aligned}\ ] ] where @xmath48 , @xmath49 , @xmath50 .
the resultant dressed potential has a lattice spacing of @xmath51 .
the structure of the adiabatic potential becomes more complicated with more dressing fields . with @xmath52 dressing fields
, it is possible to create a lattice with spacing approximately @xmath53 .
this increase in the spatial resolution is at the expense of a decrease in the lattice depth . in the near - resonant scheme ,
the counter - oscillating background potentials are generated by directly coupling the two internal states with standing wave fields , producing a pair of ac - stark split states .
the hamiltonian describing the coupling between states is now given by [ cf .
( [ eq : hamiltonian ] ) ] @xmath54 where @xmath55 is the rabi frequency corresponding to a standing wave field , and @xmath14 ( @xmath56 ) are the rabi frequencies of the dressing fields with corresponding detunings @xmath57 .
note that the background potentials are associated with the eigenstates of @xmath12 , with the values of the the potentials and the states given by the same form as eq .
( [ eq : dressedstates ] ) , with @xmath58 , @xmath59 and @xmath60 @xcite .
similar to the procedure in the previous section , we derive the equations for the adiabatic potentials , @xmath61 with @xmath23 being the adiabatic potentials . in fig .
2 we give examples of the flexibility in constructing adiabatic potentials via the near - resonant scheme as well as the off - resonant scheme , including the realisation of double - well structures on a subwavelength scale . in fig .
2b , c , we demonstrate the simple case of one dressing field : by varying the rabi - frequency or the detuning of the dressing field , the spatial structure and the central barrier height of the subwavelength double - well structure can be controlled .
this is essential for entanglement manipulation and state preparation , as we will discuss later . in higher dimensions , adiabatic potentials
can also be used both to create shorter period lattices , and to change the geometry of the lattices .
two examples of 2d potentials are shown in figs .
[ 2ds ] , [ 2dh ] . in fig .
[ 2ds ] we show the form of the dressed potentials obtained with an initial square lattice with period @xmath62 . from initial off - resonant potentials
@xmath63 , and @xmath64 , and using a detuning @xmath65 , we obtain for @xmath66 a square lattice of period @xmath67 , as the minima in this dressed potential occur at each position of a minimum in @xmath68 or @xmath69 . in the upper adiabatic potential @xmath70
, we observe the same form inverted , giving rise to a potential consisting of interconnected rings . in fig .
[ 2dh ] , we show the dressed potentials arising from an initial triangular lattice , formed by three interfering laser beams at an angle of @xmath71 to each other .
this produces a well - defined honeycomb lattice structure , with auxiliary sites introduced in the lower dressed potential , @xmath66 . in the upper dressed potential , @xmath72 we observe a series of disconnected rings
( see fig .
[ 2dh]d ) .
such disconnected ring structures can also be produced from a square 2d lattice with two dressing frequencies .
note that the control over well depth here could have many useful applications , e.g. , in the case of the square lattice it could be well - suited to implementation of ring - exchange hamiltonians @xcite . , formed with a single dressing field , @xmath65 ( note that the potential energy represented by each colour is particular to the figure part , and in each case darker colours correspond to lower potential energy ) : ( a ) the initial off - resonant potential for state @xmath8 , @xmath73 .
( b ) the initial resonant potential for state @xmath7 , @xmath74 .
( c ) the lower dressed potential @xmath66 , corresponding to a square lattice with @xmath75 times the original period .
( d ) the upper dressed potential @xmath72 , corresponding to interconnected rings . note that the adiabatic potentials @xmath76 can be found from eq .
( 10 ) , and have a maximum depth @xmath77 , which decreases as the rabi frequency of the dressing field increases . , width=321 ] , formed with a single dressing field , @xmath65 ( note that the potential energy represented by each colour is particular to the figure part , and in each case darker colours correspond to lower potential energy ) : ( a ) the initial off - resonant potential for state @xmath8 , @xmath69 .
( b ) the initial resonant potential for state @xmath7 , @xmath74 .
( c ) the lower dressed potential @xmath66 , corresponding to a honeycomb lattice with additional sites in the centre of a ring - shaped barrier .
( d ) the upper dressed potential @xmath72 , corresponding to disconnected rings .
note that the adiabatic potentials @xmath76 can be found from eq .
( 10 ) , and have a maximum depth @xmath77 , which decreases as the rabi frequency of the dressing field increases . , width=321 ]
the ideas for creating addressable , subwavelength lattices for two level atoms as discussed in the previous section will now be extended to multi - level atoms , where it is possible to create internal state- ( qubit- ) dependent potentials .
we will discuss , in particular , the case of alkaline - earth - like atoms with nonzero nuclear spin , e.g. @xmath0yb .
in addition to the optical excitation to the metastable triplet levels , we have an extra degree of freedom in the form of different magnetic sublevels , which arise from the nuclear spin . as an example
, we consider the level structure of @xmath0yb , for which the nuclear spin @xmath78 .
we will make use of the states on the clock transition .
the states in the ground state manifold are @xmath79 , and those in the excited state manifold are @xmath80 .
it is also possible to generalise the scheme to other alkaline - earth - like atoms with non - zero nuclear spin , e.g. @xmath81sr ( @xmath82 ) , though one then needs to carefully choose the working states in the hyperfine structure @xcite .
yb from off - resonant background potentials , which are formed by ac - stark shifts represented by the red arrows . ( a ) formation of subwavelength lattices in two two - level systems with @xmath83-polarised dressing fields ( black arrows ) , where the magnetic field induces differential zeeman shifts in the @xmath84 ( @xmath85hz@xmath86 g ) , and the @xmath87 ( @xmath88hz@xmath86 g ) manifolds @xcite .
( b ) formation of subwavelength lattices in two two - level systems with circularly - polarised dressing fields ( black arrows ) .
, width=604 ] to generate off - resonant subwavelength lattices , we induce opposite background potentials for the states on the clock transition .
this is done by applying a set of standing wave laser fields at a specific `` anti - magic '' wavelength at which the ac polarisability of the manifolds @xmath89s@xmath90 and @xmath91p@xmath90 are exactly opposite .
this is exactly the opposite case as that used in optical lattice clock experiments @xcite , where a `` magic '' wavelength is chosen so that the ac stark shifts of the clock states are equal .
( we expect that the `` anti - magic '' wavelength in our case to be @xmath92 nm for @xmath0yb , and the resultant ac stark shift on the order of several hundred khz for a laser intensity of @xmath93kw/@xmath94 . )
once the background potentials are generated , we can couple these potentials by applying additional optical dressing fields tuned near the resonance of the clock transition @xmath95 at @xmath96 .
atoms can be loaded into the minima of the upper(lower ) adiabatic potential by first preparing them in one of the manifolds @xmath89s@xmath90 or @xmath91p@xmath90 , and then appropriately ramping up the dressing fields adiabatically . as shown in fig .
5a , the @xmath89s@xmath90 and @xmath91p@xmath90 manifolds for @xmath0yb are
each split into two sublevels .
this gives us a four - level system , and dressing fields coupling the different manifolds can be produced using either @xmath83-polarised or circularly polarised light . in appendix
a we summarise the origins of the corresponding matrix elements @xcite .
note , in particular , that the matrix elements for circularly polarised couplings are non - zero as a result of contributions in the @xmath91p@xmath90 manifold from other levels due to hyperfine mixing @xcite . in the case that we use
only circularly polarised dressing fields , we will produce two independent two - level systems , where the resulting subwavelength lattice can be manipulated for each by altering the appropriate component of each circular polarization in the dressing field . for @xmath83-polarised light
, we also obtain two independent two - level systems , in which we can offset the relative well depths by applying an external magnetic field . due to the differential zeeman shift in the @xmath89s@xmath90 and @xmath91p@xmath90 manifolds
@xcite , this will alter the detunings of the dressing fields , as well as the energies of the uncoupled internal states .
for example , with the laser parameters : @xmath29khz , @xmath97khz , @xmath98khz , and under a small magnetic field of @xmath99 g , the estimated energy difference between the potentials corresponding to the two different dressed states at @xmath100 is @xmath101khz , and at @xmath102 is @xmath103khz .
these numbers increase significantly as the external magnetic field is increased .
this allows the state to be rotated between dressed states of the different two - level systems selectively for different wells of the substructure using rf coupling fields .
this would be useful for applications to quantum information processing ( see below ) .
( a1 ) lights of different circular polarisation are tuned near the centre of the fine structure splitting to generate spin - dependent potentials .
( a2)(b1 ) rf fields are applied to couple different hyperfine states with counter - oscillating potentials , which are generated by shifting the spin - dependent potentials .
( b2 ) the resultant adiabatic potential in the case of one dressing field has a lattice spacing @xmath6 , with @xmath104 the wavelength of the laser generating the spin - dependent potentials in ( a1 ) .
, width=302 ] one may also generate off - resonant subwavelength lattices with alkali atoms . here
, the counter - oscillating background potentials should be produced for two different hyperfine ground states making use of well known ideas to produce spin - dependent lattice potentials @xcite , as shown in fig .
these spin - dependent lattice potentials can then be coupled using rf dressing fields to produce sub - wavelength optical lattices .
details of this scheme are summarised in the caption of fig .
. note that with alkali atoms it is not straightforward to find combinations of states that act as independently coupled two - level systems and that are simultaneously collisionally stable .
this will limit the application of alkali atoms in generating state - dependent subwavenlength potentials , as well as in entanglement operations , as we will discuss later for alkaline - earth - like atoms .
in addition , with alkali atoms , the two hyperfine ground states used to generate the dressed potentials decohere fast in the presence of magnetic field fluctuations , which is not the case with alkaline - earth - like atoms , where the dressed potentials are generated almost purely from nuclear spins .
yb using cross - coupling dressing fields . ( b ) ( above ) with only the coupling laser @xmath105 , there are two independent sets of spin - dependent lattice potentials @xmath106 ( @xmath107 ) , where @xmath108 labels the two different sets of potentials created by lasers with different polarisation and @xmath109 labels the upper / lower potential in a given set
. the relative position of the two sets of potentials can be adjusted by varying @xmath110 in the lin-@xmath111-lin configuration .
the lattice spacing in this case is still @xmath6 .
( below ) by adding one dressing field @xmath40 , the lower potentials become state - dependent subwavelength potentials , with lattice spacing @xmath51 .
the relative positions of the two lower potentials can be adjusted by varying @xmath110.,width=566 ] the long - lived @xmath91p@xmath90 level motivates the application of near - resonant lattices , as discussed previously in sec .
[ sec2b ] , with direct coupling with a standing wave field tuned to the clock transition producing counter - oscillating dressed potentials .
the great advantage here , though , is that we can now produce these background potentials with lights of two different circular polarisations , generating two sets of independent background potentials ( see fig .
we can take advantage of this in two ways : ( 1 ) we can couple these potentials with circularly polarised dressing fields ( see fig . 7 ) , so that we produce two independent sets of addressable sub - wavelength lattices ; or ( 2 ) we can couple these potentials with @xmath83-polarised fields , so that we form dressed states with all four levels ( see fig .
8a ) , and use this to engineer particular forms for the potentials .
for the convenience of discussion , we define the notation for the internal states @xmath112 , @xmath113 , @xmath114 , and @xmath115 .
in the first case , where the states are cross - coupled as shown in fig . 7 , we then produce state - dependent lattices in the sense that we have two independent two - level systems , in which we form dressed states of @xmath116 and @xmath117(@xmath107 ) for each @xmath108 independently .
both the near - resonant background potentials and the final sub - wavelength potentials are internal - state dependent , and can be easily manipulated independently .
for example , using lasers in a lin@xmath111lin configuration ( two counter - propagating , linearly polarized beams with a relative polarization angle @xmath110 ) , the lattices can be relatively shifted by varying @xmath110 @xcite ( see fig .
7b ) . such spin dependence could be used similarly to the case in alkali atoms @xcite , and we expect losses due to spontaneous emissions to be strongly suppressed in the present case
. note that the dipole matrix element for the cross coupling is given by @xmath118 , where @xmath119 is the frequency of the clock transition @xmath95 , and @xmath120mhz is the natural linewidth of the @xmath87 state in @xmath0yb @xcite(see appendix a for more details ) .
the two originally degenerate potentials ( the dashed and the solid curve ) are shifted relative to each other under a small bias magnetic field , and with a finite angle @xmath110 in the lin-@xmath111-lin configuration as shown in fig .
parameters : @xmath121khz , @xmath122khz , @xmath123khz , @xmath124khz , @xmath125 , @xmath99 g .
the energy shifts are @xmath126khz , and @xmath127khz , respectively.,width=302 ] subwavelength addressability can be achieved in this configuration in a manner similar to that discussed in the previous section . by preparing the atoms near the minima of the lower adiabatic potentials and applying an external magnetic field to lift the degeneracy of the states in the subwavelength structure , we can perform rotations between the independent two - level systems selectively by well within the subwavelength structure . in the second case we consider using @xmath83-polarised dressing fields as shown in fig .
all the states are coupled to make the adiabatic potentials , and changing the angle @xmath110 in the lin-@xmath111-lin configuration modifies the shape of all four adiabatic potentials in each floquet manifold . in this configuration , it is possible to produce a repeated double - well potential , in which the two wells can be individually addressed .
this is done by first setting the angle @xmath110 in the lin-@xmath111-lin configuration to a small but nonzero value .
this will change the shape of the potentials and lift the symmetry in the double - well structure .
an example is given in fig .
9 , where at @xmath121khz , @xmath128khz , @xmath129khz , @xmath123khz , @xmath125 , @xmath99 g , the energy shifts are @xmath130khz and @xmath131khz for the two sub - wells , respectively .
atoms in the left and right hand side of the double - well structure can then be rotated selectively by a rf coupling after the application of an external magnetic field to shift the potentials . in practice , strong coupling at intensity @xmath132 on the clock transition
will also give rise to ac - stark shifts @xmath133 of the states @xmath117 and @xmath116 , which , as @xmath134 and @xmath135 , will become important at large fields ( e.g , @xmath136kw/@xmath94 for @xmath0yb @xcite ) .
these have been neglected in the two - level atom picture presented in sec .
[ sec2b ] . at higher fields ,
we essentially obtain a combination of the off - resonant and near - resonant background potentials .
if the polarisability of @xmath91p@xmath90 and @xmath89s@xmath90 are opposite , it will simply enhance the depth of the resulting potentials .
these combined adiabatic potentials are still spin - dependent as the ac - stark shifts lose their spatial dependence as the angle @xmath110 is increased to @xmath137 . in the case that these polarisabilities have the same sign , the ac stark shift can also be cancelled or modified by an additional laser at a different wavelength . with alkali atoms one could also form near - resonant lattices by coupling different hyperfine ground states in the @xmath138 level , e.g. , via a raman process .
if the raman process beams have a sinusoidally varying effective rabi frequency , this will then generate two counter - oscillating dressed states .
these can be coupled by rf dressing fields to form subwavelength lattice potentials .
however , this does not give rise to simple internal state - dependent lattices as in the case with alkaline - earth - like atoms .
having established the methods to realise subwavelength lattice potentials , we now turn to discuss ways to utilise the flexibility in control of the potentials for applications in quantum simulation and quantum information .
this requires consideration of the many - body physics in these potentials .
we will show how a generalised hubbard model can be realised by engineering the subwavelength potential .
this generalised single - band hubbard model features fast tunnelling rates due to the small spatial separation of different wells in the potential , which can be utilised to improve the timescales associated with realisation of interesting quantum phases and entanglement via exchange interactions .
we begin by considering the simplest case of the off - resonant scheme with a sinusoidal lattice in 1d , beginning with two background counter - oscillating potentials @xmath139 , as discussed in sec .
[ sec2a ] .
we then choose the detuning and rabi frequency of @xmath52 dressing transitions so that the resulting bound states in each well of the adiabatic potential have approximately the same energy .
in this way a subwavelength lattice will be formed in which the lattice spacing scales approximately as @xmath140 $ ] , while the depth @xmath141 of the potential can be as large as @xmath142 and decreases with increasing rabi frequency @xmath14 .
motion can be considered in the tight - binding limit of the resulting lattice model provided @xmath143\gtrsim 5e_r$ ] , and the dynamics are described by the single - band hubbard model @xmath144 here , @xmath145 is the annihilation operator for the bosonic or fermionic atoms , and @xmath28 specifying the lattice site and internal atomic ( spin ) state , @xmath146 denotes the tunneling rates .
we denote by @xmath147 the energy offset for the internal state and lattice site specified by @xmath28 , and @xmath148 are the collisional interaction energies , which follow from our knowledge of the potentials and scattering properties .
a single - band hubbard model is valid provided all of these energies are smaller than the separation energy to higher bands , which , in the case that the lowest bound states in all wells have the same energy , is bounded above by the oscillation frequency in a typical well , @xmath149^{1/2}/h$ ] .
if we choose the detunings of the @xmath52 dressing transitions so that @xmath147 is constant , then in 1d , @xmath150 can be estimated via a simple rescaling of the recoil energy @xmath151 , as @xmath152 here , the tunneling rates increase exponentially with increasing @xmath52 . note , however , that these rates can vary within the substructure , because the wells with @xmath153 dressing frequencies will not be exactly symmetric , and this can give rise to small splittings within the lowest bloch band .
10 we plot values of @xmath154 for varying depths of the background potential , and number of dressing transitions @xmath52 .
we will typically have onsite interactions only , which can be controlled via the shape of each well and the scattering lengths for atoms in the different internal states @xmath7 and @xmath8 .
this can give rise to generalised hubbard models , where the onsite interaction changes values between neighbouring sites .
such models can support non - trivial phases , such as coexisting mott insulating and superfluid phases @xcite . note that when there is only one dressing frequency @xmath155 , the value of @xmath148 is approximately the same as the corresponding onsite interaction in the original background lattice potentials , as the curvature of the wells is essentially unchanged . in the presence of @xmath52 dressing fields as a function of @xmath156 .
the squares are numerically - obtained values for the case @xmath157(lowest curve ) and @xmath155(upper two curves ) , respectively .
the dotted lines are plots of the @xmath52-dependent scaling function ( see eq .
( 15 ) in the text ) .
the lowest curve is for the case @xmath157 , and the higher two curves are for the @xmath155 case.,width=302 ] as discussed in sec .
[ sec3b ] , we have with @xmath0yb the possibility to form different addressable subwavelength lattices from four internal states .
we have different possibilities then to encode qubits and generate entanglement depending on our choice of configurations for the dressing transitions . in the case
that we choose circularly polarised dressing transitions ( see fig .
7 ) , we obtain two independent two - level systems , for which we can control the subwavelength lattices independently ( see sec .
[ sec3b ] )
. we can then encode qubits on the dressed states of the two two - level systems .
these can be addressed individually within the substructure of the subwavelength lattice ( i.e. , modulo the period of the background potentials , where other recent ideas could be applied to provide addressing on the larger length scale @xcite ) .
we can then make use of the state - dependent lattice potentials to perform entangling operations , e.g. , using controlled collisions @xcite .
another possibility arises in the case that we use @xmath83-polarised dressing fields and form adiabatic eigenstates combining all four internal states ( see sec .
[ sec3b ] and fig .
here it is possible to create entanglement for fermionic atoms such as @xmath0yb within the double - well structures using exchange interactions @xcite , where the fast tunneling rates @xmath146 between neighbouring wells give rise to faster gate times than are possible without subwavelength lattices .
this leads to quantum gate schemes in the spirit of the loss - divincenzo proposal for quantum dots @xcite , which is based on encoding qubits on electron spins in electrically gated quantum dots , and performing entangling operations using exchange interactions between neighbouring dots by ramping down the corresponding potentials .
the same techniques can be applied here .
we consider the two dressed states of our four initial internal levels that have the same dressed potentials , and define these to be our qubit states .
we then consider the dynamics generated by the hamiltonian in eq .
( [ genbh ] ) , but now with the tunnelling parameters @xmath146 determined by our choice of subwavelength lattices with repeated double - well potentials .
the tunnelling between these wells can be made negligible on the timescale of the experiment , so that the dynamics within each double well is independent .
we then have a situation very similar to that discussed in the analysis of spin - exchange gates , e.g. , in ref .
@xcite , leading from the two - site double - well potential to exchange interactions between the spin states in each well , given by @xmath158 , where @xmath159 .
here , @xmath160 and @xmath161 represent the spin operators in the left and right well respectively , where the spin in each well corresponds to two - state system consisting of the two different dressed states we are condsidering .
the coefficient @xmath162 is determined from the tunneling amplitude between sites , @xmath163 [ see eq .
( [ genbh ] ) ] ; and the onsite interaction strengths @xmath164 and @xmath165 , between two atoms , one in each state , in the left well and right well respectively . in the context of electrically gated quantum dots , one of the most challenging aspects of implementing quantum computing is to achieve local addressing of the qubits . as described above , this
is made possible here by the addressability of the individual wells within the subwavelength structure , as discussed in sec .
[ sec3b ] .
there we specifically discussed rotation between different qubit states .
however , it is also possible to perform readout state - selectively by creating an energy offset between the two qubit states by introducing an external magnetic field , and then coupling them state - selectively to the @xmath91p@xmath166 level .
this is possible because the @xmath91p@xmath166 level has a very small natural linewidth @xmath167mhz @xcite , so that coupling to this level can be easily made frequency - selective .
coupling between the qubit states and @xmath91p@xmath166 can be generated with a raman pulse , making use of the larger coupling elements of both @xmath91p@xmath166 and @xmath91p@xmath90 to high - lying states such as @xmath91s@xmath168 .
as discussed in sec . [ sec2a ] , one of the major limitations of the adiabatic potentials is the loss of atoms via landau - zener - type processes .
these losses correspond to the tunneling of atoms from a given adiabatic potential into the continua of other adiabatic potentials . in the following we show that these processes can be made exponentially small by increasing the laser power .
landau - zener - type losses can be understood by looking at the simple scenario of a single avoided crossing of two adiabatic potentials .
this problem has been discussed in detail in ref .
@xcite , where the two adiabatic potentials @xmath169 are generated by coupling a two - level atomic system using a standing wave light field with spatially oscillating rabi - frequency .
the loss rate @xmath170 of a particle from the bottom of a well of the upper adiabatic potential @xmath171 into the continuum spectrum of @xmath172 can be estimated semiclassically for a small violation of adiabaticity : @xmath173,\label{lossrate}\ ] ] where @xmath174}$ ] , @xmath175 is the energy of the atom , and @xmath176 is determined by the saddle - point condition @xmath177 .
the quantity @xmath178 is the attempt frequency , which is of the order of the oscillation frequency @xmath179 of a particle at the bottom of the upper potential @xmath171 @xcite .
this analysis , in particular eq .
( [ lossrate ] ) , can be directly adopted for the estimation of the landau - zener - type loss rate from the upper adiabatic potential in the simple case of one dressing field in the off - resonant scheme , and no dressing field in the near - resonant scheme . for the off - resonant scheme with only one dressing field , and for a detuning @xmath180
, the adiabatic potentials can be written as @xmath169 , where @xmath181 , and @xmath182 . in the vicinity of the saddle point @xmath183
, the background potential can be linearised @xmath184 , and we take the energy of the atom @xmath185 .
we may then taylor expand the adiabatic potentials near the saddle point , @xmath186 , where @xmath187 with @xmath151 being the recoil energy in the original normal lattice . with these ,
one may carry out the integral in eq .
( [ lossrate ] ) explicitly and we get for the loss rate : @xmath188= \gamma_0\exp\left[-\alpha\frac{\omega_1^{3/2}}{2v_0e_r^{1/2}}\right],\ ] ] where the dimensionless parameter @xmath189 in this case .
note that @xmath40 characterises in this case the energy separation between the adiabatic potentials .
( 17 ) is different from the scaling relation of a typical landau - zener transition , where the exponent in the loss rate scales with the energy separation squared .
one may recover this typical scaling relation by examining the loss rate of atoms with higher energies in the adiabatic potential .
for the near - resonant lattices , different manifolds in the floquet basis only decouple in the absence of dressing fields . in this case
, we have only the background potentials @xmath169 resulting from the ac stark splitting , where @xmath190 .
we may also integrate eq .
( [ lossrate ] ) to calculate the loss rate from the upper background potential : @xmath191= \gamma_0\exp\left[-\alpha\frac{\delta_0^{3/2}}{\omega_0e_r^{1/2}}\right]\ ] ] where @xmath192 , with @xmath189 . in this case
, @xmath193 characterises the energy separation between the background potentials .
these results suggest that the loss rate from the upper adiabatic potential can be made exponentially small by increasing the ratio of the energy difference of the two potentials and the frequency of the atomic motion in the trap . in the simple cases above ,
this translates to large rabi - frequency @xmath194 in the off - resonant case , and large detuning @xmath195 in the near - resonant case .
lattice , formed with one dressing transition : ( a ) numerical simulation results for loss rates for atoms in the bound state of the higher - energy adiabatic potential in the off - resonant scheme .
the solid line is an exponential fit to the result .
( inset ) the corresponding adiabatic potentials with parameters : @xmath196 , @xmath197 , @xmath198 .
( b ) numerical calculations of the decay rate for atoms in the narrower wells in the low energy adiabatic potential in the near - resonant scheme . ( inset )
typical potentials in the limit where the floquet manifolds are weakly coupled ( @xmath199 ) , with a series of floquet manifolds each containing 2 dressed potentials .
parameters : @xmath200 , @xmath201 , @xmath202 , and @xmath203 . ,
title="fig:",width=302 ] lattice , formed with one dressing transition : ( a ) numerical simulation results for loss rates for atoms in the bound state of the higher - energy adiabatic potential in the off - resonant scheme .
the solid line is an exponential fit to the result .
( inset ) the corresponding adiabatic potentials with parameters : @xmath196 , @xmath197 , @xmath198 .
( b ) numerical calculations of the decay rate for atoms in the narrower wells in the low energy adiabatic potential in the near - resonant scheme . ( inset )
typical potentials in the limit where the floquet manifolds are weakly coupled ( @xmath199 ) , with a series of floquet manifolds each containing 2 dressed potentials .
parameters : @xmath200 , @xmath201 , @xmath202 , and @xmath203 .
, title="fig:",width=302 ] when more complicated lattice potentials are considered , the loss rate due to this landau - zener - type transition should in general be computed numerically .
however , simple estimates of the loss rates can be obtained by assuming that for a given adiabatic potential the loss is dominated by the leak into the continua of the adiabatic potentials which are closest in energy . if we assume that the loss into the continuum of each potential is independent , the matrix elements for the decay into the continuum of any one of these nearest adiabatic potentials will have the form eq .
( [ lossrate ] ) under the semiclassical approximation . an exponential dependence of the decay rate @xmath170 on the energy separation between the adiabatic potentials is then obtained , similar to eqs .
( 17 ) , ( 18 ) .
numerically , we find the dimensionless parameter @xmath28 to be on the order of unity or larger when the laser parameters , e.g. rabi - frequencies and detunings , are much larger than the recoil energy @xmath204 .
this implies that the loss rate @xmath170 can always be made exponentially small by increasing the laser power .
to check the general conclusion drawn above , we have also performed numerical simulations of the real time evolution of the atoms trapped in the adiabatic potentials by integrating the schrdinger equation for the full hamiltonian in the floquet basis .
the atom is initially prepared in the ground state of a potential well , and transition rates are extracted from the projection of the wavefunction onto the manifold corresponding to the initial adiabatic potential as a function of time .
for the simple case in fig .
11a , we vary @xmath40 for fixed trapping frequency in the wells of the adiabatic potential @xmath205 . for a large gap between the potentials , the loss rate scale exponentially as @xmath206 , as shown by the exponential fit ( solid line ) , which is in agreement with the semiclassical calculation . from the fit , the factor @xmath28 is estimated to be @xmath207 and @xmath208 for the exact and linearised potentials , respectively , slightly smaller than the results in the semiclassical approximation . in the case of near - resonant scheme , as in fig
11b , we fix @xmath209 , and vary the energy gap @xmath133 between the adiabatic potentials .
we again observe an exponential scaling , @xmath210 ( solid line ) , where the trapping frequency @xmath211 . in conclusion , we see one can always avoid large landau - zener - type losses from the adiabatic potential by increasing the laser power to make the background potential sufficiently deep .
to summarise , we have discussed various schemes for the generation of subwavelength lattices with either alkali or alkaline - earth - like atoms .
the lattice spacing in these adiabatic potentials can be much smaller than the wavelength of the laser that creates the lattice potential , depending on the number of the additional dressing lasers .
the enhanced tunneling rate one gets from the reduced lattice spacing may be useful for the engineering of generalised hubbard models for quantum simulation , or for the design of fast quantum gates for quantum computation .
the adiabatic potentials in dimensions higher in 1d may also have interesting structures , i.e. disconnected ring potentials in 2d , which is useful for the realisation of interesting phases .
the first experimental steps in this direction have been recently been reported in ref . @xcite . while we have discussed addressable lattices within a @xmath6 period , addressing on a larger scale
can be introduced using magnetic or electric field gradients , or other recently proposed techniques @xcite .
this , combined with the gate operation , holds the prospect for the realisation of general quantum computation in this system .
we have also discussed in detail the landau - zener - type losses from the adiabatic potentials .
we have shown that the loss rate can always be made small at the expense of laser power . for large number of dressing fields ,
i.e. for small lattice spacing , the landau - zener - type loss would become considerable , and would be the practical limitation on how small a lattice spacing one can get in generating the subwavelength lattices .
we thank h. p. bchler , a. v. gorshkov , s. kotochigova , i. lesanovsky , m. d. lukin , a. micheli , v. palchikov , t. porto , y. takahashi , j. ye , and t. zelevinsky for interesting discussions .
this work was supported by the austrian fwf through project sfb f15 , and by the eu networks olaqui and scala .
in this appendix , we examine in detail the matrix element for the clock transition @xmath95 in alkaline - earth - like atoms . for a detailed discussion , please refer to ref .
@xcite .
we consider the matrix element @xmath212 , where @xmath213 is the electric - dipole operator . in the presence of a hyperfine interaction
, @xmath214 is not an eigenstate of the system . under the first order perturbation
, the correction to the hyperfine sublevels in @xmath87 manifold can be written as : @xmath215 where @xmath216 labels the manifolds that the the state is coupled to , and @xmath217 is the hyperfine interaction . due to the scalar nature of @xmath218 ,
the total angular momentum @xmath219 and its projection on z - axis @xmath220 are good quantum numbers , as is shown in the expression above .
as we are only interested in the e1 transitions from @xmath84 to @xmath87 , we only consider the nuclear magnetic dipole contribution in the hyperfine interaction , which fixes @xmath221 in the expression above .
the matrix element for the clock states then becomes : @xmath222 where @xmath223 , and @xmath224 is the clebsch - gordon coefficient for the state @xmath225 .
for the convenience of discussion , we now consider the concrete example of @xmath0yb , which has a nuclear spin of @xmath78 .
it is obvious that the matrix element is nonzero for both the @xmath83-polarised laser field and the circularly - polarised laser field . for @xmath83-polarised lasers ,
the matrix element is effectively given by @xmath226 while for circularly - polarized lasers , the matrix element is effectively given by @xmath227 for the @xmath228- and @xmath229- polarised light , respectively .
the ratio between the matrix elements in the two cases is given by that of the corresponding clebsch - gordon coefficients in eq .
one may further relate these matrix element to the natural linewidth @xmath170 of the clock transition : @xmath230 where @xmath119 is the frequency for the clock transition .
the constant factor @xmath231 comes from the clebsch - gordon coefficient , which has @xmath232 for @xmath83-polarised lasers and
@xmath233 for circularly polarised lasers . for @xmath0yb , @xmath234mhz @xcite , and a typical laser intensity of @xmath235kw/@xmath94
gives a rabi - frequency of @xmath236khz for a circularly polarised field .
99 bloch i , dalibard j and zwerger w 2007 many - body physics with ultracold gases _ preprint _ arxiv:0704.3011v1 hein m , dr w , eisert j , raussendorf r , van den nest m and briegel h j 2006 _ proceedings of the international school of physics enrico fermi on quantum computers , algorithms and chaos ( varenna , italy ) _ vol
. 162 ed p zoller g casati d shepelyansky g benenti p115 | we discuss how adiabatic potentials can be used to create addressable lattices on a subwavelength scale , which can be used as a tool for local operations and readout within a lattice substructure , while taking advantage of the faster timescales and higher energy and temperature scales determined by the shorter lattice spacing . for alkaline - earth - like atoms with non - zero nuclear spin ,
these potentials can be made state dependent , for which we give specific examples with @xmath0yb atoms .
we discuss in detail the limitations in generating the lattice potentials , in particular non - adiabatic losses , and show that the loss rates can always be made exponentially small by increasing the laser power . |
the relation between the casimir - polder ( cp ) interaction of a single atom with a body @xcite and its van der waals ( vdw ) interaction with the atoms comprising the body @xcite was first discussed in detail by renne @xcite , with special emphasis on an atom interacting with a dielectric half space .
approaching the problem from the microscopic side and modelling all atoms by identical harmonic oscillators , he showed that the sum of all many - atom vdw interactions between the single atom and the body atoms corresponds to the result for the cp potential derived earlier on the basis of a macroscopic approach @xcite .
milonni and lerner generalised this result to bodies of arbitrary shapes @xcite .
using the ewald - oseen extinction theorem which follows from the clausius - mosotti relation , they demonstrated that the cp potential of an atom in the presence of a nonabsorbing homogeneous dielectric body can be obtained by summing over an infinite series of many - atom vdw potentials . in this paper , we approach the problem from the macroscopic side and under more general conditions . to that end , based on macroscopic qed in linear , causal media ( sec .
[ sec2 ] ) , we first consider the cp interaction of an atom with a macroscopic body as well as the vdw interaction of two atoms ( sec .
[ sec3 ] ) , and then make use of the born expansion and the clausius - mosotti relation to establish the microscopic origin of the cp potential ( sec .
[ sec4 ] ) , followed by a summary ( sec .
[ sec5 ] ) .
according to the multipolar scheme , the hamiltonian for a system of neutral atoms and/or molecules briefly referred to as atoms in the following interacting with the electromagnetic field in the presence of dispersing and absorbing bodies is given by @xcite @xmath0 where the hamiltonian @xmath1 governs the ( unperturbed ) internal dynamics of atom @xmath2 , @xmath3 is the hamiltonian of the system composed of the electromagnetic field and the magnetodielectric medium including dissipative interactions , with @xmath4 and @xmath5 being the dynamical variables of the system , which satisfy bosonic commutation relations , and the atom - field interaction in electric dipole approximation is given by @xmath6 ( @xmath7 : atomic electric dipole moment , @xmath8 : centre - of - mass position ) . in eq .
( [ eq6 ] ) , @xmath9 is the medium - assisted electric field expressed in terms of the dynamical variables , with the quantities @xmath10 being given in terms of the classical green tensor @xmath11 , @xmath12 \bm{g}(\mathbf{r},\mathbf{r}',\omega ) = \bm{\delta}(\mathbf{r}-\mathbf{r}').\ ] ] the ( macroscopic ) permittivity @xmath13 and permittivity @xmath14 satisfy the kramers - kronig relations and the conditions @xmath15 @xmath16 @xmath17 and @xmath18 @xmath16 @xmath17 imposed for absorbing media .
note that the green tensor has the useful properties @xcite @xmath19
dispersion forces can be derived from the associated potentials @xmath20 , which are commonly identified as the position - dependent part of the ground - state energy shift @xmath21 induced by the atom - field coupling . for a single atom @xmath2 the ground state of the system
is given by @xmath22 @xmath23 @xmath24 [ @xmath25 , atomic ground state ; @xmath26 @xmath23 @xmath17 ] , and the leading , second - order energy shift reads @xmath28 note that the primed sum includes ( principal - value ) integrals over the continuous degrees of freedom .
inspection of eq .
( [ eq6 ] ) reveals that only intermediate states of the type @xmath29 @xmath23 @xmath30 contribute . substituting the respective matrix elements @xmath31_i\ ] ] [ @xmath32 @xmath23 @xmath33 and energy denominators @xmath34 @xmath23 @xmath35 [ @xmath36 @xmath23 @xmath37 into eq .
( [ eq15 ] ) , using eq .
( [ eq13 ] ) , and separating the green tensor into its bulk and scattering parts according to @xmath38 one arrives , after some calculation , at [ @xmath39 @xmath40 @xmath41 @xcite @xmath42 where , for simplicity , an isotropic atomic ground - state polarizability is considered , @xmath43 the potential ( [ eq19 ] ) implies the cp force ( @xmath44 @xmath45 @xmath46 ) @xmath47 on a ground - state atom @xmath2 due to the presence of an arbitrary arrangement of dispersing and absorbing bodies [ which are accounted for by @xmath48 . to illustrate the use of eq .
( [ eq19 ] ) , consider an atom placed at distance @xmath49 from the centre of a small , homogeneous , magnetodielectric sphere of radius @xmath50 @xmath51 @xmath52 .
substituting the respective green tensor @xcite into eq .
( [ eq19 ] ) and retaining only the leading - order term in @xmath53 ( cf.ref .
@xcite ) , one obtains @xmath54,\\ \label{eq23 } g_{ee}(x ) = & \ ; 2e^{-2x}(3 + 6x+5x^2 + 2x^3+x^4),\\ \label{eq24 } g_{em}(x ) = & \ ; 2e^{-2x}(1 + 2x+x^2),\end{aligned}\ ] ] with the electric and magnetic polarizabilities of the sphere being given by @xcite @xmath55 in the nonretarded limit , where the atom - sphere separation is much larger than the characteristic transition wavelengths of the atom and the sphere medium , one may approximate @xmath56 @xmath57 @xmath58 , @xmath59 @xmath57 @xmath60 , so eq . ( [ eq22 ] ) reduces to @xmath61 and @xmath62 for electric and magnetic spheres , respectively , while in the opposite retarded limit the approximations
@xmath63 @xmath57 @xmath64 , @xmath65 @xmath57 @xmath66 , @xmath67
@xmath57 @xmath68 lead to @xmath69 } { 64\pi^3\varepsilon_0 ^ 2r_\mathrm{a}^7}\,.\end{aligned}\ ] ] to calculate the vdw interaction of two atoms @xmath2 and @xmath70 in the presence of dispersing and absorbing bodies , we start from the ground state @xmath22 @xmath23 @xmath71 and consider those contributions to the energy shift that depend on the positions of both atoms . the leading contributions are hence contained in the fourth - order perturbative shift @xmath72 where @xmath73 @xmath23 @xmath74 @xmath75 @xmath76 .
a typical set of possible intermediate states is given by @xmath77 upon using eq .
( [ eq16 ] ) as well as @xmath78_{i ' } \delta_{\lambda\lambda^{\prime\prime}}\delta_{ii^{\prime\prime } } \delta(\mathbf{r } -\mathbf{r}^{\prime\prime } ) \delta(\omega-\omega^{\prime\prime})\nonumber\\ & \hspace{4ex}\quad\,+\bigl[\mathbf{d}_a^{k0}\!\cdot\ !
\bm{g}_{\lambda^{\prime\prime}}(\mathbf{r}_a , \mathbf{r}^{\prime\prime},\omega^{\prime\prime } ) \bigr]_{i^{\prime\prime } } \delta_{\lambda\lambda'}\delta_{ii ' } \delta(\mathbf{r } -\mathbf{r } ' ) \delta(\omega-\omega')\bigr\}\end{aligned}\ ] ] and recalling eq .
( [ eq13 ] ) , substitution of the intermediate states ( [ eq31 ] ) into eq .
( [ eq30 ] ) leads , after some calculation , to @xmath79,\end{aligned}\ ] ] where we have assumed real dipole - matrix elements . under this assumption , the various two - atom contributions @xmath80 to the energy shift @xmath81 only differ by the denominators in the square brackets of eq .
( [ eq33 ] ) , and hence after a lengthy calculation one may show that [ @xmath81 @xmath40 @xmath82 @xcite @xmath83.\ ] ] from the two - atom potential ( [ eq37 ] ) one can calculate the vdw force on atom @xmath2(@xmath70 ) due to atom @xmath70(@xmath2 ) in the presence of arbitrary dispersing and absorbing magnetodielectric bodies according to @xmath84 as a simple example , consider two atoms embedded in bulk magnetodielectric material .
substitution of the respective green tensor @xcite into eq .
( [ eq37 ] ) leads to @xmath85}{\varepsilon^2(iu)}\,,\end{aligned}\ ] ] @xmath86 @xmath45 @xmath87 , @xmath88 @xmath23 @xmath89 , recall eq .
( [ eq23 ] ) , which reduces to @xmath90 for nonretarded and retarded interatomic separations , respectively . equations ( [ eq39])([eq41 ] ) show that the presence of a medium leads to a reduction of the potential w.r.t . its well - known free - space value @xcite , while comparison of eqs .
( [ eq22 ] ) and ( [ eq39 ] ) reveals that in free space the dispersion interaction of an atom with a small dielectric sphere has the same form as that of two atoms .
we now turn to the question how the cp interaction of a single atom with dielectric bodies can be related to its many - atom vdw interactions with the atoms comprising the bodies . for simplicity
, we will speak of a single dielectric body in the following .
we assume the dielectric body to be given by @xmath91 , and we allow for the presence of an arbitrary magnetodielectric background of additional bodies characterised by @xmath92 and @xmath14 , such that @xmath93 the green tensor corresponding to this scenario can formally be written as a born series @xcite @xmath94\nonumber\\ & \hspace{4ex}\times\ , \overline{\bm{g}}(\mathbf{r},\mathbf{s}_1,\omega)\!\cdot\ ! \overline{\bm{g}}(\mathbf{s}_1,\mathbf{s}_2,\omega)\!\cdots\ ! \overline{\bm{g}}(\mathbf{s}_k,\mathbf{r}',\omega),\end{aligned}\ ] ] where @xmath95 is the green tensor corresponding to the magnetodielectric background . substituting eq .
( [ eq43 ] ) into eq .
( [ eq19 ] ) , the cp potential can be written as @xmath96 where @xmath97 is the cp potential due to the magnetodielectric background , which is not of further interest here , and @xmath98 \nonumber\\ & & \times\mathrm{tr}\bigl [ \overline{\bm{g}}(\mathbf{r}_a,\mathbf{s}_1,iu)\!\cdot\ !
! \overline{\bm{g}}(\mathbf{s}_k,\mathbf{r}_a , iu)\bigr]\end{aligned}\ ] ] is the contribution to the potential that is of @xmath99th order in @xmath100 .
assuming the dielectric body described by @xmath100 to be comprised of atoms of polarizabilities @xmath101 and number densities @xmath102 , the gap between the macroscopic and microscopic descriptions can be bridged by means of the clausius - mosotti law @xmath103 note that since @xmath100 is the fourier transform of a linear response function , it must satisfy the condition @xmath104
@xmath16 @xmath105 @xmath16 @xmath17 for @xmath106 @xmath16 @xmath17 , which implies that the inequality @xmath107 must hold . substituting eq .
( [ eq48 ] ) into eq .
( [ eq47 ] ) and splitting off the singular part of the green tensor according to @xmath108 one obtains @xmath109 where @xmath110\\ \times\ , v_{ab_1\ldots b_l } ( \mathbf{r}_a,\mathbf{s}_1,\ldots,\mathbf{s}_l , iu)\end{gathered}\ ] ] with @xmath111\end{gathered}\ ] ] denotes the sum of all @xmath112-atom terms that are of order @xmath99 in @xmath100 , and each power of the factor @xmath113 is due to the integration of one term containing @xmath114 .
summing eq .
( [ eq51 ] ) over @xmath99 , one may rearrange the double sum as follows : @xmath115v_{ab_1\ldots b_l } ( \mathbf{r}_a,\mathbf{s}_1,\ldots,\mathbf{s}_l),\end{gathered}\ ] ] where we have performed the geometric sums @xmath116 by means of eq .
( [ eq54 ] ) .
note that the convergence of these sums requires @xmath117 @xmath118 @xmath119 , which by means of eq .
( [ eq49 ] ) is equivalent to @xmath120 finally , we symmetrize the many - atom terms by introducing the symmetrization operator @xmath121 where @xmath122 denotes the permutation group of the numbers @xmath123 . from the cyclic property of the trace together with eq .
( [ eq12 ] ) it follows that @xmath124 = \mathrm{tr}\bigl [ \overline{\bm{h } } ( \mathbf{r}_{\pi(1)},\mathbf{r}_{\pi(2)},\omega)\cdots \overline{\bm{h } } ( \mathbf{r}_{\pi(j)},\mathbf{r}_{\pi(1)},\omega)\bigr],\ ] ] if @xmath125 is either a cyclic permutation or the reverse of a cyclic permutation .
thus the sum on the r.h.s . of eq .
( [ eq59 ] ) contains classes of @xmath126 terms that give the same result . by forming a set @xmath127 @xmath128 @xmath129 containing exactly one representative of each class
, we can remove this redundancy , leading to @xmath130 \nonumber\\ & \qquad=\sum_{\pi\in \overline{p}(j)}\!\ !
\mathrm{tr}\bigl [ \overline{\bm{h } } ( \mathbf{r}_{\pi(1)},\mathbf{r}_{\pi(2)},\omega)\cdots \overline{\bm{h } } ( \mathbf{r}_{\pi(j)},\mathbf{r}_{\pi(1)},\omega)\bigr].\end{aligned}\ ] ] introducing the factor @xmath131 in eq .
( [ eq56 ] ) and summing over all @xmath132 possible ways of renaming the variables @xmath133 and @xmath134 , the representative of each class in eq .
( [ eq61 ] ) is generated exactly twice ( only once for @xmath135 @xmath23 @xmath136 @xmath23 @xmath137 ) , so that the cp potential of atom @xmath2 due to the dielectric body @xmath100 can be written as @xmath138 u_{ab_1\ldots b_l } ( \mathbf{r}_a,\mathbf{s}_1,\ldots,\mathbf{s}_l),\ ] ] where @xmath139\end{gathered}\ ] ] is nothing but the @xmath135-atom vdw potential on an arbitrary magnetodielectric background @xmath140 , @xmath14 .
we have hence proved that the cp interaction of an atom with a macroscopic dielectric body which is described within the framework of macroscopic qed is the result of all possible microscopic many - atom vdw interactions between the atom under consideration and the atoms forming the body , provided that the susceptibility is of clausius - mosotti type ( [ eq48 ] ) and the convergence condition ( [ eq58 ] ) holds generalising the result in ref .
conversely , our proof shows that when eqs .
( [ eq62 ] ) and ( [ eq63 ] ) hold and the convergence condition ( [ eq58 ] ) is satisfied , then the electric susceptibility must have the clausius - mosotti form .
in addition , the proof has delivered the general many - atom vdw potentials ( [ eq63 ] ) on an arbitrary magnetodielectric background . for @xmath135 @xmath23 @xmath137 , eq .
( [ eq63 ] ) agrees with the two - atom potential ( [ eq37 ] ) derived in sec .
[ sec3.2 ] , while for higher @xmath135 , it presents a generalisation of the free - space vdw potentials derived in ref .
@xcite .
the applicability of the microscopic expansion ( [ eq62 ] ) depends crucially on the validity of the convergence condition ( [ eq58 ] ) .
recalling eq .
( [ eq20 ] ) and estimating @xmath141
[ @xmath142 , electron charge ; @xmath143 , electron mass ; @xmath144 @xmath23 @xmath145 ; @xmath146 @xmath23 @xmath147 ; @xmath148 @xmath23 @xmath149 ; @xmath150 , volume accessible per atom within the body ; @xmath151 @xmath23 @xmath152 , , species - dependent factor ] , eq .
( [ eq58 ] ) can be reformulated as @xmath150 @xmath153 @xmath154 , stating simply that the atoms must be well - separated within the body .
if this is not the case , the microscopic expansion ( [ eq62 ] ) does not converge , while the more general macroscopic expression ( [ eq19 ] ) for the cp potential remains valid .
we have demonstrated that on the basis of macroscopic qed in linear , causal media , leading - order perturbation theory can be employed to derive general expressions for both the single - atom cp potential and the two - atom vdw potential in the presence of an arbitrary arrangement of magnetodielectric bodies .
moreover , starting from this very general , geometry - independent basis , we have used the born expansion of the green tensor together with the clausius - mosotti law to prove that the cp interaction of a single atom with inhomogeneous , dispersing and absorbing dielectric bodies in the presence of an arbitrary magnetodielectric background can be written as a sum of many - atom vdw potentials .
the proof demonstrates the equivalence of the microscopic and macroscopic descriptions provided that the microscopic picture is applicable , while at the same time delivering explicit expressions for the general many - atom vdw potentials in the presence of magnetodielectric media .
this work was supported by the deutsche forschungsgemeinschaft .
would like to thank the ministry of science , research , and technology of iran . h.t.d .
would like to thank the alexander von humboldt stiftung and the national program for basic research of vietnam . | we establish a general relation between dispersion forces . first , based on qed in causal media , leading - order perturbation theory
is used to express both the single - atom casimir - polder and the two - atom van der waals potentials in terms of the atomic polarizabilities and the green tensor for the body - assisted electromagnetic field .
endowed with this geometry - independent framework , we then employ the born expansion of the green tensor together with the clausius - mosotti relation to prove that the macroscopic casimir - polder potential of an atom in the presence of dielectric bodies is due to an infinite sum of its microscopic many - atom van der waals interactions with the atoms comprising the bodies .
this theorem holds for inhomogeneous , dispersing , and absorbing bodies of arbitrary shapes and arbitrary atomic composition on an arbitrary background of additional magnetodielectric bodies . |
it is well known that the density of states in a pure one dimensional interacting electron gas ( luttinger liquid ) vanishes as a power law in the vicinity of the fermi level @xcite : @xmath9 , where the exponent is determined by the interaction strength .
such a behavior should reveal itself in the dependence of the differential tunnel conductance on the applied bias : @xmath10 .
it is obvious that the presence of a disorder would perturb the local value of the density of states and , thus , cause some random correction , @xmath3 , to the conductance .
then the relevant questions are : \i ) what is the typical magnitude of @xmath7 for a given realization of the disorder ?
\ii ) how the values of @xmath7 at different voltages are correlated ?
\iii ) how the correlation between the values of @xmath7 at the same voltage , but at different points of the liquid , falls off with increasing separation between the points ?
these questions are addressed in the present paper .
we will consider the case of a smooth disorder .
namely , we will assume that the correlation radius is much larger than the fermi wavelength .
this assumption simplifies the problem drastically , since it permits one to neglect the backward scattering of electrons and , hence , to view the disorder as a random variation of the electron concentration with coordinate .
the key to understanding the role of the disorder is provided by the bosonization procedure @xcite , which allows to describe the low energy excitations of the system in terms of bosonic excitations ( plasmons ) , propagating along the liquid .
since the velocity of a plasmon depends on the concentration of electrons , the spatial variation of the concentration would give rise to the backscattering of plasmons .
in other words , in the presense of a disorder , a plasmon with frequency @xmath1 acquires a finite mean free path @xmath0 .
it is important to note that a smooth disorder , for which the backscattering of electrons at the fermi level is suppressed , might , ultimately , be quite efficient in backscattering of plasmons with wavelengths of the order of the correlation radius .
for non interacting electrons in one dimension it is established that even a weak disorder leads to the localization of all eigenstates @xcite . then the mean free path acquires the meaning of the localization radius .
the same conclusion applies , certainly , to plasmons .
the difference is , however , that an electron becomes more and more localized as its energy decreases , whereas for plasmons the situation is the opposite : the lower is the frequency , the weaker is the localization .
since the dispersion law of a plasmon is linear : @xmath11 , where @xmath12 is the sound velocity , the density of the plasmon states is frequency independent .
the reason for suppression of the localization for the low
frequency plasmons is that the matrix element of the baskscattering vanishes in the limit of long wavelengths . in particular , we show that for small @xmath1 the frequency dependence of the mean free path is @xmath2 and , correspondingly , the product @xmath13 , which determines the localization strength , increases as @xmath14 when @xmath1 goes to zero .
we also show that @xmath0 turns to infinity when the interactions are switched off , thus revealing that the backscattering of plasmons is possible only due to interactions .
if the relative variation of the concentration is small , the disorder induced correction to @xmath15 can be expressed in terms of the spectral characteristics of the localized plasmons ( such as local density of states ) . on the other hand ,
these characteristics were the subject of detailed studies in application to the localized electrons @xcite . by utilizing the approach developed in refs .
, we calculate the variance @xmath16 , and the two point correlator @xmath17 .
we find that the ratio @xmath18 increases with voltage as @xmath19 and that the characteristic `` period '' of change of @xmath3 with voltage is of the order of @xmath4 .
we also find that the two point correlator falls off with @xmath20 as a power law and oscillates with the period @xmath21 , which is one half of the wavelength of a plasmon with frequency @xmath8 .
the paper is organized as follows . in the next section the formula for the mean free path of a plasmon
is derived . in section [ ctc ]
we calculate the correlator of fluctuations of the tunnel conductance .
section [ con ] concludes the paper .
the hamiltonian of a luttinger liquid with concentration of electrons , @xmath22 , being a function of coordinate has the form @xcite @xmath23 , \ ] ] where @xmath24 and @xmath25 are , correspondingly , the displacement and conjugate momentum ( @xmath26=i\hbar \delta(x - x')$ ] ) ; @xmath27 is the effective interaction strength : @xmath28 .
the hamiltonian can be reduced to a system of harmonic oscillators @xmath29 , \ ] ] by means of the following transformation @xmath30 here @xmath31 are the eigenfunctions of the operator @xmath32 which is defined as @xmath33= \omega_{\mu}^2 \phi_{\mu}.\ ] ] the eigenvalues of @xmath32 determine the frequencies , @xmath34 , of the oscillators .
if the concentration is constant ( @xmath35 ) , the solutions of ( [ d ] ) are the plane waves @xmath36 with a linear spectrum @xmath37 ( @xmath38 is the normalization length ) . substituting ( [ pw ] ) into ( [ d ] ) , we get the standard expression for the sound velocity @xcite @xmath39 where @xmath40 is the fermi velocity . assuming that the relative variation of the concentration is small , @xmath41 , the expression for the mean free time , @xmath42 , for a plasmon with frequency @xmath1 is given by the golden rule @xmath43 where @xmath44 is the plane wave with @xmath45 and @xmath46 stands for the averaging over the fluctuations of @xmath22 .
the factor @xmath14 in ( [ gr ] ) appears since we define the mean free time as @xmath47 . keeping only the first order terms in the difference @xmath48 , we get the following expression for the matrix element @xmath49 .\end{aligned}\ ] ] the energy conservation , insured by the @xmath50-function in ( [ gr ] ) , requires that @xmath51 ( backscattering )
. then we have @xmath52 and , as it can be easily seen , the second term in the matrix element vanishes identically . as a result
, the matrix element takes the form @xmath53 we see that the matrix element for backscattering is proportional to the interaction strength , which reflects the fact that this process is possible only due to interactions .
the mean free path , @xmath0 , defined as @xmath54 , can be found after substituting ( [ fme ] ) into ( [ gr ] ) @xmath55 where we have introduced the correlator of the fluctuations of @xmath22 @xmath56 here @xmath57 is the correlation radius and @xmath58 is the mean square fluctuation of @xmath22 , so that @xmath59 .
it is instructive to rewrite ( [ mf ] ) for the product @xmath60 , which measures the effective scattering strength . assuming the gaussian form of the correlator , @xmath61
, we obtain @xmath62 note that the last factor in ( [ mfg ] ) is a function of the argument @xmath63 and this function does not exceed unity .
thus , the product @xmath13 is large for any @xmath1 . obviously , the backscattering is ineffective when @xmath64 , i.e. when the correlation radius @xmath57 exceeds the wave length of the plasmon . in the opposite limit , @xmath65 ,
the mean free path behaves as @xmath66 . as it was mentioned in the introduction ,
the real meaning of @xmath0 is the localization length of a plasmon with frequency @xmath1 .
the fact that @xmath67 allows one to apply to localized plasmons , the description developed earlier for localized electrons @xcite just in this limit .
this is done in the next section .
once the hamiltonian is diagonalized , the derivation of the formula for the density of states becomes standard @xcite .
the operator @xmath68 , creating an electron at point @xmath69 , can be presented in the form @xcite @xmath70 where the coefficients @xmath71 and @xmath72 are defined as @xmath73 the calculation of the green function @xmath74 does not differ from that for a pure luttinger liquid and leads to the following expression for the density of states @xmath75 where @xmath76 is the sum over eigenmodes @xmath77 if @xmath22 is constant so that the eigenfunctions are given by ( [ pw ] ) , eq.([dos ] ) reproduces the known result @xcite for the density of states : @xmath78 , with @xmath79 indeed , for plane waves we have latexmath:[$|\alpha_{\mu}|^2=1/k_{\mu}^2ln_0 , the function @xmath76 @xmath81 where @xmath82 is the cutoff parameter . substituting this expression into ( [ dos ] ) leads to ( [ kappa ] ) . to find the disorder - induced correction to @xmath83 , we treat the difference @xmath84 as a perturbation and expand the exponent in ( [ dos ] ) to the first order .
this gives @xmath85.\ ] ] to calculate the integral over @xmath86 , it is convenient to present the denominator in ( [ ddos ] ) in the form @xmath87 after substituting ( [ ga ] ) into ( [ ddos ] ) , both integrations , over @xmath88 and @xmath86 can be easily carried out , and one obtains @xmath89 ( \omega-\omega_{\mu})^{\kappa } \right .
\nonumber \\
\left.- ( \kappa+1 ) \int_0^{\omega
/ s}\frac{dq}{q}(\omega - sq)^{\kappa } \right\ } .\end{aligned}\ ] ] note that the constant @xmath90 can be expressed through the derivative @xmath91 by integrating the equation ( [ d ] ) from @xmath69 to @xmath92 .
then , one gets @xmath93 to study the correlation properties of @xmath94 , we introduce the following local densities @xmath95 the average values of these densities are equal @xmath96 then the correction @xmath97 can be rewritten in terms of the fluctuations @xmath98 and @xmath99 @xmath100 .\ ] ] correspondingly , the correlator of @xmath97 at different points is expressed through the correlators of the fluctuations @xmath101
@xmath102 ^ 2 \times \nonumber \\ & & \int_0^{\omega } \frac{d\omega_1}{\omega_1 } \int_0^{\omega } \frac{d\omega_2}{\omega_2 } ( \omega - \omega_1)^{\kappa } ( \omega - \omega_2)^{\kappa } \left(\frac{v_f^2}{s^2 } f_1 + \frac{s^2}{v_f^2 } f_2 + 2 f_{12 } \right ) \ ; , \end{aligned}\ ] ] where the correlators @xmath103 and @xmath104 are defined as @xmath105 the correlator @xmath106 is , in fact , the correlator of the fluctuations of the local density . in application to localized electrons
, it was studied in refs .
using , correspondingly , the berezinski technique @xcite and the technique developed by berezinski and gorkov @xcite . in the limit @xmath107 ,
the correlator is nonzero only if the difference @xmath108 is small enough : @xmath109 . for distances
@xmath110 , the expression for @xmath106 , obtained in ref . , reads @xmath111 \ ; , \ ] ] where @xmath112 is the difference between the two frequencies and the function @xmath113 is given by @xmath114 as it was noted in refs .
, at @xmath115 and @xmath116 , we have @xmath117 , i.e. the correlation is absent .
it is easy to establish that the correlator @xmath118 is equal to @xmath106 .
concerning the correlator @xmath104 , we did not find the expression for this correlator in the literature .
so , we have calculated it using the berezinski - gorkov technique and obtained the following expression @xmath119 \ ; .\ ] ] it is seen that in contrast to @xmath106 , the correlator @xmath104 is non - zero at @xmath115 and @xmath120 @xcite . for finite @xmath121 ,
both correlators are proportional to @xmath122 and decay with increasing @xmath121 as @xmath123 . to calculate the double integral in ( [ nucor ] ) , we make use of the fact that the correlators @xmath124 and @xmath104 are sharp functions of @xmath121 , i.e. the major contribution to the integral comes from the domain @xmath125 .
this allows to put @xmath126 in all other factors and to extend the integration over @xmath127 to @xmath128 . noting that the integral of @xmath129 is equal to @xmath130
, we obtain @xmath131 note that the coefficients in front of @xmath124 and @xmath104 in ( [ nucor ] ) have combined into the factor @xmath132 which is proportional to @xmath133 at small @xmath27 . using ( [ int_e ] ) , the expression for the correlator @xmath134 takes the form @xmath135 ^ 2 \int_0^{\omega } \frac{d \omega_1}{\omega_1 ^ 2 \tau_{\omega_1 } } ( \omega -\omega_1)^{2 \kappa } \cos \left(\frac{2 \omega_1 z}{s } \right ) \ ; .\ ] ] apparently , the factor @xmath136 in ( [ nnucor ] ) diverges at small @xmath137 .
however , this divergency is compensated by the frequency dependence of the mean free time , which at small @xmath137 behaves as @xmath136 .
substituting @xmath138 in ( [ nnucor ] ) and introducing the new variable @xmath139 , we get the final result @xmath140 ^ 2 \left(1-\frac{v_f^2}{s^2}\right)^4 \frac{\overline{(\delta n)^2}}{n_0 ^ 2 } \frac{s r_c \omega^{2\kappa+1}}{v_f^2 } r(\tilde{z } ) \ ; , \ ] ] where the function @xmath141 is defined as @xmath142 here , @xmath143 is the dimensionless distance @xmath144 the function @xmath141 is defined in such a way that , in the only interesting limit @xmath145 , it turns to @xmath146 at @xmath147 .
this function determines the coordinate dependence of the correlator ( [ nucor ] ) and , consequently , the coordinate dependence of the correlator of the fluctuations of the tunnel conductance @xmath148 the function @xmath141 is ploted in fig.1 . with increasing distance , it falls off and oscillates .
the asymptotic behavior of @xmath149 at @xmath150 is as follows @xmath151 the spatial period of the oscillations of @xmath141 is @xmath152 and it decreases with increasing voltage .
it is to be emphasized that such a behavior of the correlator of @xmath7 ( slow decay and oscillations ) is entirely due to interactions . in the absence of the interactions ,
the correlator ( [ tcf ] ) would simply reproduce the correlator ( [ dc ] ) of the fluctuations of @xmath22 , i.e. it would decay monotonously at distance @xmath153 which is much smaller than @xmath154 . at @xmath147 , the formula ( [ nucor ] ) defines the variance of @xmath97 .
it is convenient to rewrite it for the relative magnitude of the fluctuations of @xmath7 @xmath155 naturally , the variance @xmath156 is proportional to the magnitude of the fluctuations of the density .
less trivial is that ( [ tcv ] ) contains two additional factors : @xmath157 which is proportional to @xmath133 at small @xmath27 , and a small factor @xmath158 ( note that ( [ tcv ] ) is written in the limit @xmath159 ) . this factor determines the voltage dependence : @xmath160 .
qualitatively , this dependence can be interpreted as follows .
formula ( [ fddos ] ) shows that all plasmon modes with frequencies smaller than @xmath1 contribute to @xmath161 .
obviously , the major contribution comes from plasmons for which the center of localization lies within the localization radius from the point @xmath69 .
let us consider some frequency strip @xmath162 centered at @xmath163 and with the width @xmath164 .
then , the localization radius for all modes within this strip is approximately @xmath165 . for a spatial interval of length @xmath0 ,
the typical frequency spacing between modes is @xmath166 .
correspondingly , the average number of modes within the strip is @xmath167 and , hence , the relative fluctuation of this number is @xmath168 .
the final estimate for the fluctuation of @xmath94 emergies if we set @xmath169
. then , one gets @xmath170 . substituting ( [ mfg ] ) for @xmath171
, we reproduce the voltage dependence in ( [ tcv ] ) .
however , the estimate and the result of the calculation still differ by a factor @xmath172 .
this factor originates from the specific details of the strucure of the eigenfunctions ( see the comment after eq .
( [ int_e ] ) ) , and we can not interpret it qualitatively .
note in conclusion of the section , that since all frequencies from @xmath173 to @xmath1 give , roughly , comparable contributions to @xmath161 , it will change significantly only when the frequency doubles .
this means that at fixed @xmath69 the characteristic period of fluctuations in @xmath7 as a function of voltage is of the order of @xmath4 .
the main result of the present paper is that the randomness in the concentration of electrons in the luttinger liquid causes a random component in the tunnel conductance which changes semiperiodically along the liquid , with the period depending on the applied bias .
this behavior results from the fact that at a fixed bias , v , the frequencies of plasmons , responsible for the correction to the density of states , are strictly limited by @xmath174 .
correspondingly , the oscillating behavior of the correlator of the fluctuations reflects the distribution of density in a plasmon with maximal frequency .
if the conductance is studied as a function of bias , the disorder would cause fluctuations with characteristic period of the order of @xmath4 .
the fluctuations disappear as the wavelength of the plasmon with frequency @xmath174 becomes smaller than the spatial scale of the change of the concentration , @xmath57 .
there is also a limitation from low voltages , imposed by the finite length of the liquid , @xmath38 .
namely , our calculation applies when the mean free path of a plasmon , @xmath0 , is smaller than @xmath38 . at low enough frequencies
@xmath0 exceeds @xmath38 and the disorder does not play any role . for such frequencies the oscillations of the tunnel conductance with voltage have their origin in the size quantization of plasmons , the period of oscillations being @xmath175 . these oscillations were studied in ref . .
note finally , that the approach developed in the present paper can be extended to the case of the multichannel luttinger liquid @xcite . in the latter case
, the scattering of plasmons between the channels , caused by a disorder , should be taken into account .
y. oreg and a. m. finkelstein , phys .
lett . * 76 * , 4230 ( 1996 ) .
d. c. mattis and e. h. lieb , j. math .
* 6 * , 304 ( 1965 ) .
f. d. m. haldane , j. phys .
c * 14 * , 2585 ( 1981 ) . n. f. mott and w. d. twose , adv .
* 10 * , 107 ( 1961 )
. v. l. berezinski , zh .
* 65 * , 1251 ( 1973 ) [ sov .
jetp * 38 * , 620 ( 1974 ) ] .
v. l. berezinski and l. p. gorkov , zh .
. fiz . * 77 * , 2489 ( 1979 ) [ sov .
jetp * 50 * , 1209 ( 1979 ) ] .
e. p. nakhmedov , v. n. prigodin , and yu .
a. firsov , zh .
* 92 * , 2133 ( 1987 ) [ sov .
jetp * 65 * , 1202 ( 1987 ) ] .
l. p. gorkov , o. n. dorokhov , and f. v. prigara , zh .
. fiz . * 84 * , 1440 ( 1983 ) [ sov .
jetp * 57 * , 838 ( 1983 ) ] . yu .
v. nazarov , a. a. odintsov , and d. v. averin , bull .
. soc . * 41 * , 376 ( 1996 ) .
a. gramada , m. e. raikh , preprint cond - mat/9604090 g. mahan _ many - particle physics _ , 2nd ed .
( plenum , new york , 1990 ) . b. l. altshuler and v. n. prigodin , zh . eksp .
fiz . * 95 * , 348 ( 1989 ) [ sov .
jetp * 68 * , 198 ( 1989 ) ] .
actually there is no qualitative understanding by now , why @xmath106 turns to zero at coinciding points @xcite .
k. a. matveev and l. i. glazman , phys .
* 70 * , 990 , ( 1993 ) . | the luttinger liquid in which the concentration of electrons varies randomly with coordinate is considered .
we study the fluctuations of the tunnel conductance , caused by the randomness in the concentration .
if the concentration changes slowly on the scale of the fermi wavelength , its prime role reduces to the scattering of the plasmon waves , propagating along the system . as a result of such a scattering , plasmons get localized .
we show that the localization length , @xmath0 , of a plasmon with frequency @xmath1 is inverse proportional to the square of the interaction strength and changes with frequency as @xmath2 .
if the relative variation of the concentration is small , the randomness induced correction to the tunnel conductance , @xmath3 , where @xmath4 is the applied bias , can be expressed through the spectral characteristics of the localized plasmons .
the magnitude of the correction , @xmath5 , increases with @xmath4 as @xmath6 .
the typical period of the fluctuations in @xmath3 is of the order of @xmath4 . at a fixed @xmath4 , the correlator of @xmath7 at different points of the liquid falls off with distance as a power law and oscillates with the period which is one half of the wavelength of a plasmon with frequency @xmath8 . |
this paper is concerned with the mathematical analysis of an electromagnetic scattering problem in periodic structures , where the wave propagation is governed by the time - domain maxwell equations . the scattering theory in periodic diffractive structures ,
also known as diffraction gratings , has applications in many cutting - edge scientific areas including ultra - fast and high - energy lasers , space flight instruments , astronomy , and synchrotron spectrometers .
a good introduction can be found in @xcite to diffraction grating problems and various numerical approaches .
the book @xcite contains descriptions of several mathematical problems that arise in diffractive optics modeling in industry .
some more recent developments are addressed in @xcite on theory , analysis , and computational techniques of diffractive optics .
the time - harmonic grating problems have been extensively studied by many researchers via either the integral equation methods or the variational methods @xcite .
a survey may be found in @xcite for mathematical studies in rigorous grating theory .
the general result may be stated as follows : the diffraction problem has a unique solution for all but a countable sequence of singular frequencies .
unique solvability for all frequencies can be obtained for gratings which have absorbing media or perfectly electrically conducting surfaces with lipschitz profiles .
numerical methods are developed for both the two - dimensional helmholtz equation ( one - dimensional gratings ) and the three - dimensional maxwell equations ( crossed or two - dimensional gratings ) @xcite .
the time - domain scattering problems have attracted considerable attention due to their capability of capturing wide - band signals and modeling more general material and nonlinearity @xcite .
comparing with the time - harmonic problems , the time - domain problems are much less studied due to the additional challenge of the temporal dependence .
rigorous mathematical analysis is very rare
. the analysis can be found in @xcite for the time - domain acoustic and electromagnetic obstacle scattering problems .
we refer to @xcite for the analysis of the time - dependent electromagnetic scattering from a three - dimensional open cavity .
numerical solutions can be found in @xcite for the time - dependent wave scattering by periodic structures / surfaces .
the theoretical analysis is still lacking for the time - domain scattering by periodic structures .
the goal of this work is to analyze mathematically the time - domain scattering problem which arises from the electromagnetic wave propagation in a periodic structure . specifically , we consider an electromagnetic plane wave which is incident on a one - dimensional grating in @xmath0 .
so the structure is assumed to be invariant in the @xmath1-direction and periodic in the @xmath2-direction .
the three - dimensional maxwell equations can be decomposed into two fundamental polarizations : transverse electric ( te ) polarization and transverse magnetic ( tm ) polarization , where maxwell s equations are reduced to the two - dimensional wave equation .
we shall study the wave equation in two dimensions for both polarizations .
the structure can also be characterized by the medium parameters : the electric permittivity and the magnetic permeability .
they are periodic in @xmath2 and assumed only to be bounded measurable functions .
hence our method works for very general gratings whose surfaces / interfaces are allowed to be lipschitz profiles or even graphs of some lipschitz continuous functions .
there are two challenges of the problem : time dependence and unbounded domain . in the frequency domain
, various approaches have been developed to truncate unbounded domains into bounded ones , such as absorbing boundary conditions ( abcs ) , transparent boundary conditions ( tbcs ) , and perfectly matched layer ( pml ) techniques .
these effective boundary conditions are being extended to handle time - domain problems @xcite . utilizing the laplace
transform as a bridge between the time - domain and the frequency domain , we develop an exact time - domain tbc and reduce the problem equivalently into an initial boundary value problem in a bounded domain . using the energy method with new energy functions ,
we show the well - posedness and stability of the time - dependent problem .
the proofs are based on examining the well - posedness of the time - harmonic helmholtz equations with complex wavenumbers and applying the abstract inversion theorem of the laplace transform .
moreover , a priori estimates , featuring an explicit dependence on time and a minimum regularity requirement of the data , are established for the wave field by studying directly the time - domain wave equation .
the paper is organized as follows . in section [ frp ]
, we introduce the model problem and develop a tbc to reduce it into an initial boundary value problem .
section [ srp ] is devoted to the analysis of the reduced problem , where the well - posdeness and stability are addressed and a priori estimates are provided .
we conclude the paper with some remarks and directions for future work in section [ cl ] .
in this section , we introduce the mathematical model of interest and develop an exact tbc to reduce the scattering problem from an unbounded domain into a bounded domain .
consider the system of time - domain maxwell equations in @xmath0 for @xmath3 : @xmath4 where @xmath5 is the electric field , @xmath6 is the magnetic field , @xmath7 and @xmath8 are the dielectric permittivity and magnetic permeability , respectively , and satisfy @xmath9 here @xmath10 are constants .
we assume that the structure is invariant in the @xmath1-direction and thus focus on the one - dimensional grating .
the more complicated problem in biperiodic structures will be considered in a separate work .
there are two fundamental polarizations for the one - dimensional structure : \(i ) te polarization .
the electric and magnetic fields are @xmath11^{\top},\quad \boldsymbol h(x , z , t ) = [ h_1(x , z , t ) , 0 , h_3(x , z , t)]^{\top}.\ ] ] eliminating the magnetic field from ( [ tmw ] ) , we get the wave equation for the electric field : @xmath12 \(ii ) tm polarization .
the electric and magnetic fields are @xmath13^{\top},\quad \boldsymbol h(x , y , z , t)= [ 0 , h(x , z , t ) , 0]^{\top}.\ ] ] we may eliminate the electric field from and obtain the wave equation for the magnetic field : @xmath14 it is clear to note from and that the te and tm polarizations can be handled in a unified way by formally exchanging the roles of @xmath7 and @xmath8 . we will just present the results by using as the model equation in the rest of the paper . now let us specify the problem geometry , which is shown in figure [ pg ] . since the structure and medium are assumed to be periodic in the @xmath2 direction , there exists a period @xmath15 such that @xmath16 we assume that @xmath7 and @xmath8 are constants away from the region @xmath17 , where @xmath18 are constants .
denote @xmath19 and @xmath20 .
there exist constants @xmath21 and @xmath22 such that @xmath23 throughout we also assume that @xmath24 , which is usually satisfied since @xmath25 and @xmath26 are the electric permittivity and magnetic permeability in the free space @xmath27 .
finally we define @xmath28 and @xmath29 .
consider an incoming plane wave @xmath30 which is incident on the structure from above .
explicitly we have @xmath31 where @xmath32 is a smooth function and its regularity will be specified later , and @xmath33 . here
@xmath34 , satisfying @xmath35 , is the incident angle , and @xmath36 is the light speed in the free space . clearly , the incident field @xmath37 satisfies the wave equation ( [ te ] ) when @xmath38 .
although the incident field @xmath30 may not be a periodic function in the @xmath2-direction , we can verify that @xmath39 motivated by the uniqueness of the solution , we assume that the total field satisfies the same translation property , i.e. , @xmath40 we define @xmath41 it follows from ( [ cv ] ) that we get @xmath42 which shows that @xmath43 is a periodic function in the @xmath2-direction with period @xmath44 .
similarly , we can verify that the incident field @xmath45 is a trivially periodic function of @xmath2 ( independent of @xmath2 ) since @xmath46 using the change of variables , we have @xmath47 the equation ( [ te ] ) becomes @xmath48 a simple calculation yields that @xmath49 which shows that the equation is a well - defined wave equation .
it is easy to verify that the incident field @xmath45 satisfies with @xmath50 . to impose the initial conditions , we assume that the total field and the incident field vanish for @xmath51 so that the incident field @xmath52 and the scattered field @xmath53 for @xmath51 .
the initial conditions are @xmath54 in addition @xmath43 is @xmath44-periodic in the @xmath2-direction .
this paper aims to study the well - posedness and stability of the scattering problem .
we introduce some notation . for any @xmath55 with @xmath56 ,
define by @xmath57 the laplace transform of the function @xmath58 , i.e. , @xmath59 define a weighted periodic function space @xmath60 which is sobolev space with the norm characterized by @xmath61 given @xmath62 it has a fourier expansion with respect to @xmath2 : @xmath63 a simple calculation yields an equivalent norm in @xmath64 via fourier coefficients : @xmath65 for a periodic function @xmath66 defined on @xmath67 with fourier coefficients @xmath68 , we define a weighted trace functional space @xmath69 where @xmath70 .
it is clear to note that the dual space of @xmath71 is @xmath72 under the @xmath73 inner product @xmath74 the weighted sobolev spaces @xmath75 and @xmath76 are equivalent to the standard sobolev spaces @xmath77 and @xmath78 since @xmath79 .
hereafter , the expression @xmath80 stands for @xmath81 , where @xmath82 is a positive constant and its specific value is not required but should be always clear from the context .
we introduce a tbc to reformulate the scattering problem into an equivalent initial - boundary value problem in a bounded domain .
the idea is to design a dirichlet - to - neumann ( dtn ) operator which maps the dirichlet data to the neumann data of the wave field . subtracting the incident field @xmath45 from the total field @xmath43 in and
, we obtain the equation for the scattered field @xmath83 and the initial conditions @xmath84 let @xmath85 be the laplace transforms of @xmath86 with respect to @xmath87 .
recall that @xmath88 taking the laplace transform of @xmath89 and using the initial conditions ( [ icv ] ) , we have @xmath90 which reduces to @xmath91 since @xmath92 is a periodic function in @xmath2 , it has the fourier expansion @xmath93 substituting the fourier expansion of @xmath94 into ( [ lse1 ] ) , we obtain an ordinary differential equation for the fourier coefficients : @xmath95 where @xmath96 using the outgoing radiation condition , we have @xmath97 thus we get the rayleigh expansion for the scattered field in @xmath27 : @xmath98 taking the normal derivative of the above equation on @xmath99 yields @xmath100 where @xmath101^\top$ ] is the unit normal vector on @xmath99 . similarly , we can obtain the rayleigh expansion for the total field in @xmath102 : @xmath103 where @xmath104 taking the normal derivative of @xmath105 on @xmath106 gives @xmath107 where @xmath108^\top$ ] is the normal vector on @xmath106 . for any function @xmath109 defined on @xmath67
, we define the dtn operators @xmath110 [ tt ] there exists a positive constant @xmath111 such that @xmath112 first we have @xmath113 using the fact that @xmath114 with @xmath115 , we have @xmath116 letting @xmath117 we can show that @xmath118 the proof is completed by combing the above estimates and the definition ( [ hpn ] ) . [ dtn ] the dtn operator @xmath119 is continuous , i.e. , @xmath120 where @xmath121 is a constant . for any @xmath122 , it follow form ( [ htn ] ) that @xmath123 where we have used @xmath124 here @xmath125 which completes the proof .
[ tp ] we have the estimate @xmath126 it follows from the definitions of ( [ dtm ] ) and ( [ htn ] ) that we have @xmath127 with @xmath128 . taking the real part of the above equation gives @xmath129 recalling @xmath130 , we have @xmath131 and @xmath132 using , we get @xmath133.\ ] ] plugging into gives @xmath134.\ ] ] adding and , we obtain @xmath135.\ ] ] substituting into yields @xmath136|u_n(h_j)|^2\leq 0,\end{aligned}\ ] ] which completes the proof . using the dtn operators ( [ dtm ] ) ,
we obtain the following tbc in the @xmath137-domain : @xmath138 where @xmath139 . taking the inverse laplace transform of ( [ tbc ] )
yields the tbc in the time domain : @xmath140 where @xmath141 is the inverse laplace transform of @xmath142 , i.e. , @xmath143 , and @xmath144
in this section , we present the main results of this work , which include the well - posedness and stability of the scattering problem and related a priori estimates . taking the laplace transform of and using the tbc ( [ tbc ] ) , we may consider the following reduced boundary value problem : @xmath145 next we introduce a variational formulation of the boundary value problem ( [ rp ] ) and give a proof of its well - posedness in the space @xmath146 . multiplying ( [ rp ] ) by the complex conjugate of a test function @xmath147 , using the integration by parts and tbcs , we arrive at the variational problem : to find @xmath148 such that @xmath149 where the sesquilinear form @xmath150 { \rm d}x { \rm d}z\notag\\ & - \sum\limits_{j=1}^2 \langle(s\mu_j ) ^{-1 } \mathscr b_j \breve u , v \rangle_{\gamma_j}.\end{aligned}\ ] ] [ vp ] the variational problem ( [ uvf ] ) has a unique solution @xmath151 , which satisfies @xmath152 it suffices to show the coercivity of the sesquilinear form of @xmath153 , since the continuity follows directly from the cauchy schwarz inequality , lemma [ tt ] , and lemma [ dtn ] .
letting @xmath154 in , we get @xmath155 { \rm d}x { \rm d}z\notag\\ & - \sum\limits_{j=1}^2 \langle ( s\mu_j ) ^{-1 } \mathscr b_j \breve u , \breve u \rangle_{\gamma_j}.\end{aligned}\ ] ] taking the real part of the above equation yields @xmath156 since @xmath157 and @xmath158 are periodic in @xmath2 , we have from the integration by part that @xmath159 which gives @xmath160 combining the above estimate and lemma [ tp ] , we obtain @xmath161 where @xmath162 .
it follows from the lax
milgram lemma that the variational problem ( [ uvf ] ) has a unique solution @xmath163 moreover , we have from ( [ uvf ] ) and lemma [ tt ] that @xmath164 combing ( [ cp ] ) and ( [ ae ] ) leads to @xmath165 which completes the proof after applying the cauchy schwarz inequality . using the time - domain tbc , we consider the reduced initial - boundary value problem : @xmath166 the following lemma ( cf .
* ; * ? ? ?
* theorem 43.1 ) ) is an analogue of paley wiener
schwarz theorem for fourier transform of the distributions with compact support in the case of laplace transform .
[ a2 ] let @xmath167 denote a holomorphic function in the half - plane @xmath168 , valued in the banach space @xmath169 .
the two following conditions are equivalent : 1 .
there is a distribution @xmath170 whose laplace transform is equal to @xmath167 ; 2 .
there is a real @xmath171 with @xmath172 and an integer @xmath173 such that for all complex numbers @xmath137 with @xmath174 it holds that @xmath175 , where @xmath176 is the space of distributions on the real line which vanish identically in the open negative half line .
the initial - boundary value problem ( [ trp ] ) has a unique solution @xmath177 , which satisfies @xmath178 and the stability estimate @xmath179 } & \big(\|\partial_t u\|_{l^2(\omega)}+\|\partial_t ( \nabla u)\|_{l^2 ( \omega)^2 } \big ) \lesssim \big ( \|\rho\|_{l^1(0 , t ; h^{-1/2}(\gamma_1))}\nonumber\\ & + \max \limits_{t \in [ 0 , t ] } \|\partial_t\rho\| _ { h^{-1/2}(\gamma_1)}+\|\partial^2_t\rho\|_{l^1(0 , t ; h^{-1/2}(\gamma_1))}\big).\end{aligned}\ ] ] first we have @xmath180 hence it suffices to estimate the integral @xmath181 taking the laplace transform of ( [ trp ] ) yields @xmath182 the well - posedness of @xmath183 follows directly from theorem [ vp ] . by the trace theorem in lemma [ tt ]
, we get @xmath184 it follows from ( * ? ? ?
* lemma 44.1 ) that @xmath158 is a holomorphic function of @xmath137 on the half plane @xmath185 where @xmath186 is any positive constant .
hence we have from lemma [ a2 ] that the inverse laplace transform of @xmath158 exists and is supported in @xmath187.$ ] one may verify from the inverse laplace transform that @xmath188 where @xmath189 is the fourier transform with respect to @xmath190 . recall the plancherel or parseval identity for the laplace transform ( cf .
* ( 2.46 ) ) ) @xmath191 where @xmath192 and @xmath193 is abscissa of convergence for the laplace transform of @xmath66 and @xmath194 using ( [ pi ] ) , we have @xmath195 since @xmath196 in @xmath197 , we have @xmath198 in @xmath199 it is easy to note that @xmath200 hence we have @xmath201 using the parseval identity again gives @xmath202 which shows that @xmath203 next we prove the stability . let @xmath204 be the extension of @xmath205 with respect to @xmath87 in @xmath206 such that @xmath207 outside the interval @xmath208.$ ] by the parseval identity ( [ pi ] ) , we follow the proof of lemma [ tp ] and get @xmath209 which yields after taking @xmath210 that @xmath211 for any @xmath212 consider the energy function @xmath213 it follows from ( [ trp ] ) that we have @xmath214 since @xmath8 and @xmath43 are periodic functions in @xmath2 , integrating by parts yields @xmath215 which gives @xmath216 since @xmath217 , we obtain from that @xmath218}\|\partial_t u\|_{h^1_{\rm p}(\omega ) } \big)\|\rho\|_{l^1(0 , t ; h^{-1/2}(\gamma_1))}. \end{aligned}\ ] ] taking the derivative of ( [ trp ] ) with respect to @xmath219 we know that @xmath220 also satisfies the same equations with @xmath141 replaced by @xmath221 . hence , we may consider the similar energy function @xmath222 and get the estimate @xmath223}\|\partial_t u\|_{h^1_{\rm p } ( \omega)}\big ) \big(\max
\limits_{t \in [ 0 , t ] } \|\partial_t\rho\| _ { h^{-1/2}(\gamma_1)}+\|\partial^2_t\rho\|_{l^1(0 , t ; h^{-1/2}(\gamma_1))}\big ) .
\end{aligned}\ ] ] combing the above estimates , we can obtain @xmath224}\|\partial_t u\|^2_{h^1_{\rm p } ( \omega)}\lesssim \max\limits_{t\in [ 0 , t ] } e_1(t)+e_2(t)\\ & \lesssim \big ( \|\rho\|_{l^1(0 , t ; h^{-1/2}(\gamma_1))}+\max \limits_{t \in [ 0 , t ] } \|\partial_t\rho\| _ { h^{-1/2}(\gamma_1)}+\|\partial^2_t\rho\|_{l^1(0 , t ; h^{-1/2}(\gamma_1))}\big ) \| \partial_t u\|_{h_p^1(\omega ) } , \end{aligned}\ ] ] which give the estimate ( [ st ] ) after applying the cauchy
schwarz inequality . in this section
, we derive a priori estimates for the total field with a minimum regularity requirement for the data and an explicit dependence on the time .
the variation problem of ( [ trp ] ) in time domain is to find @xmath225 for all @xmath3 such that @xmath226 to show the stability of its solution , we follow the argument in @xcite but with a careful study of the tbc .
let @xmath227 be the solution of ( [ trp ] ) .
given @xmath228 , we have for any @xmath229 that @xmath230 and @xmath231 let @xmath232 and define an auxiliary function @xmath233 it is clear that @xmath234 for any @xmath235 we have @xmath236 indeed , using integration by parts and ( [ ap ] ) , we have @xmath237 next , we take the test function @xmath238 in ( [ tvp ] ) and get @xmath239 it follows from ( [ ap ] ) and the initial conditions in ( [ trp ] ) that @xmath240 integrating ( [ tvp1 ] ) from @xmath241 to @xmath242 and taking the real part yield @xmath243 in what follows , we estimate the three terms of the right - hand side of ( [ ee1 ] ) separately
. by the property ( [ ap1 ] ) , we have @xmath244 let @xmath245 be the extension of @xmath43 with respect to @xmath87 in @xmath206 such that @xmath246 outside the interval @xmath247.$ ] we obtain from the parseval identity and lemma [ tp ] that @xmath248 where we have used the fact that @xmath249 after taking @xmath210 , we obtain that @xmath250 for @xmath251 we have from ( [ ap1 ] ) that @xmath252 using integration by parts and ( [ ap ] ) , we have @xmath253 by the periodicity of @xmath8 and @xmath43 in @xmath2 , it yields that @xmath254 thus @xmath255 substituting ( [ i1])([i3 ] ) into ( [ ee1 ] ) , we have for any @xmath256 $ ] that @xmath257 taking the derivative of ( [ trp ] ) with respect to @xmath87 , we know that @xmath220 satisfies the same equation with @xmath141 replaced by @xmath258 . define @xmath259 we may follow the same steps as those for @xmath260 to obtain @xmath261 integrating by parts yields that @xmath262 the first and the third terms on the right - hand side of ( [ ee3 ] ) are discussed as above .
we only have to consider the second term . by ( [ ap ] ) , lemma [ tt ] , and lemma [ dtn ] , we get @xmath263 substituting ( [ i4 ] ) and ( [ i5 ] ) into ( [ ee3 ] )
, we have for any @xmath264 $ ] that @xmath265 combing the estimates ( [ ee2 ] ) and ( [ ee4 ] ) , we obtain @xmath266 taking the @xmath267- norm with respect to @xmath268 on both side of ( [ uh1 ] ) yields @xmath269 which gives the estimate ( [ es1 ] ) after applying the young inequality . integrating ( [ uh1 ] ) with respect to @xmath270 from @xmath271 to @xmath272 and using the cauchy
schwarz inequality , we obtain @xmath273 which implies the estimate ( [ es2 ] ) by using the young inequality again .
in this paper , we studied the time - domain scattering problem in a one - dimensional grating .
the te and tm cases were considered in a unified approach .
the scattering problem was reduced equivalently into an initial - boundary value problem in a bounded domain by using the exact time - domain dtn map .
the reduced problem was shown to have a unique solution by using the energy method .
the stability was also presented .
the main ingredients of the proofs were the laplace transform , the lax milgram lemma , and the parseval identity .
moreover , by directly considering the variational problem of the time - domain wave equation , we obtained a priori estimates with explicit dependence on time . in the future
, we plan to investigate the time - domain scattering by biperiodic structures where the full three - dimensional maxwell s equations should be considered .
the progress will be reported elsewhere . | this paper is devoted to the mathematical analysis of a time - domain electromagnetic scattering by periodic structures which are known as diffraction gratings .
the scattering problem is reduced equivalently into an initial - boundary value problem in a bounded domain by using an exact transparent boundary condition .
the well - posedness and stability of the solution are established for the reduced problem . moreover , a priori energy estimates are obtained with minimum regularity requirement for the data and explicit dependence on the time . |
one of the major open problems with ultraluminous infrared galaxies ( ulirg ) is whether they host active nuclei at their centers . in at least a fraction of objects
, there is circumstantial evidence that this is the case .
one of the best examples in this respect is ngc 6240 is the limiting criterion for ulirg , strictly speaking ngc 6240 belongs to this class only if @xmath565 km s@xmath3 mpc@xmath3 ; genzel et al .
( 1998 ) . ] , a well known source in which the presence of two nuclei ( fried & schulz 1983 ) suggests an ongoing galaxy merger .
indications of the presence of an agn come mainly from x rays : the flat 210 kev x ray spectrum and the prominent iron line complex observed by asca ( iwasawa & comastri 1998 ) are very similar to those of ngc 1068 ( iwasawa , fabian & matt 1997 ; matt et al . 1997 and references therein ) and suggest reflection from cold and warm matter of an otherwise invisible nuclear component . on the other hand , in the optical spectrum all the diagnostic line ratios point towards a classification as a liner ( e.g. veilleux et al . 1995 ) , and the iso sws diagnostic diagram places ngc 6240 in the region of star formation dominance ( genzel et al . 1998 ) . in soft x
rays the spectrum is fairly complex , and best modeled by multi - temperature plasma emission , very likely associated , at least partly , with the powerful starburst and related wind present in this galaxy , a hypothesis strengthened by the rosat hri discovery of rather luminous ( @xmath610@xmath7 erg @xmath1 s@xmath3 ) emission extended over tens of kpc ( komossa , schulz & greiner 1998 ; schulz et al . 1998 ) . on the other hand , the point like soft x ray component is possibly due to scattering of the nuclear radiation ( komossa et al . 1998 ) .
a direct measurement of the nuclear luminosity is needed to assess the relative importance of agn and starburst emission . the only way to do that is to explore the hard x ray band , which may reveal the direct nuclear component .
if the absorbing matter is moderately thick to compton scattering ( i.e. @xmath8a few@xmath9 @xmath1 ) , the photons emerge above 10 kev , where the most sensitive instrument presently available is the pds onboard _ beppo_sax .
this has already added two sources ( circinus galaxy , matt et al .
1999 ; mrk 3 , cappi et al . 1999 ) to the small list of moderately thick " seyfert 2s , which was previously composed of only one object , ngc 4945 ( iwasawa et al .
1993 ; done et al .
_ beppo_sax ( boella et al . 1997 ) observed ngc 6240 on 1998 august 1417 . in this paper ,
data from three instruments are discussed : the lecs ( 0.110 kev ) , the mecs ( 1.510 kev ) , and the pds ( 13300 kev ) .
data reduction and analysis follow the standard criteria , as described in matt et al .
the total effective exposure time was 52.5 ks in the lecs , 119.4 ks in the mecs and 114.4 ks in the pds .
we accumulated and examined images for the mecs instruments . comparing the radial profiles in three energy ranges ( 1.54 , 47.5 , 7.510 kev ) with the corresponding point spread function
, no evidence for extended emission is found .
light curves and spectra were extracted within regions of 4@xmath10 and of 2@xmath10 radius centered on the source for mecs and lecs , respectively ( the small lecs extraction radius is chosen to avoid contamination from a nearby , soft source , not visible in the mecs ) .
the light curves are consistent with being constant .
the resulting count rate is ( [email protected])@xmath1210@xmath13 counts s@xmath3 ( 0.14.5 kev ) for the lecs and ( [email protected])@xmath1210@xmath14 counts s@xmath3 ( 1.510 kev ) in the mecs .
no significant variability was found in the pds light
curve either , with a count rate of [email protected] counts s@xmath3 .
we restricted the spectral analysis to the 0.14.5 kev and 1.510 kev energy bands for lecs and mecs respectively , where the latest released ( september 1997 ) response matrices are best calibrated .
standard blank sky files provided by the _ beppo_sax science data center ( sdc ) were used for the background subtraction .
in fig . 1 we compare the @xmath152 kev spectra of ngc 6240 and ngc 1068 .
both spectra have been normalized to the crab spectrum ; since the latter is a power law with photon index @xmath16 2 , we are basically showing the @xmath17 spectrum .
the mecs spectra of the two sources are similar in shape , not surprisingly as we know that in this energy range both sources have been successfully fitted by two reflection components , warm and cold ( e.g. iwasawa et al .
1997 ; iwasawa & comastri 1998 ) .
while below 10 kev the flux of ngc 6240 is the lowest , the opposite occurs in the pds band . clearly , in ngc 6240 a new component emerges at high energies .
no known bright sources are present in the field of view of the pds , and the probability of a serendipitous source in the @xmath18 pds field with a flux equal or larger than that of ngc 6240 is @xmath19 , if the 210 kev asca logn logs ( cagnoni et al .
1998 ) is adopted , and a power law spectrum with photon index 1.8 is assumed .
thus the best explanation for this component is nuclear emission piercing through an absorber with column density @xmath20 a few@xmath1210@xmath21 cm @xmath14 , i.e. compton thick but still permitting partial transmission above @xmath610 kev . in fig . 2 ,
the broad band spectrum is fitted following iwasawa & comastri ( 1998 ) : an excess is evident in the pds band .
we therefore modeled the continuum above 2 kev with : _ i _ ) a power law , representing the intrinsic emission , with a photon index fixed at a typical value of @xmath22=1.8 , absorbed by a column density allowed to be free ( leaving the index as a free parameter , a value consistent with the adopted one , but very loosely constrained , is found ) ; _ ii _ ) a second , unabsorbed power law with the same index , standing for the warm reflection component ; _ iii _ ) a pure cold reflection component ( pexrav model in xspec ) generated by a power law with the same index .
as we are interested here in the high energies , and because below 2 kev asca is superior to _ beppo_sax as far as both sensitivity and spectral resolution are concerned , we content ourselves with adopting in this band the best fitting model of iwasawa & comastri ( 1998 ) .
it consists of a multi - temperature optically thin plasma , which probably originates in a powerful starburst .
in particular we use a two temperature mekal model , where the absorption of the coldest mekal was fixed at the galactic value ( n@xmath23 = 5.8@xmath12 10@xmath24 @xmath1 ) , while the absorption of the other mekal was allowed to be a free parameter .
we applied the latter absorption to all the agn components .
the envisaged scenario is then one in which an external starburst region , associated with the cooler mekal component , absorbs emission from the inner regions of ngc 6240 ( e.g. fabian el al .
1998 ) .
a prominent and broad iron k line complex , very similar to that seen in ngc 1068 , is also evident .
because of this similarity we assume that the line is actually a blend of : the 6.4 kev k@xmath25 and 7.06 kev k@xmath26 fluorescent lines from cold iron ( the intensity of the latter has been forced to be equal to 1/9 that of the former , as expected from atomic physics ) ; the 6.7 and 6.97 k@xmath25 recombination lines from he and h like iron . in summary the complete model , @xmath27 , we used to fit our data
can be written as : @xmath28 + m_{\rm c}\},\ ] ] where : @xmath29 is the absorption associated with the galactic column , @xmath30 is the absorption related to the starburst , @xmath31 is the absorption acting on the nuclear emission ( including the compton cross section ) , possibly associated with the molecular torus shrouding the nucleus , @xmath32 is the power
law modeling the nuclear component , @xmath33 and @xmath34 are respectively the warm ( optically thin ) and cold ( optically thick ) reflection components , @xmath35 and @xmath36 are respectively the cold and hot thermal components and @xmath37 are the four gaussian lines modeling the iron blend . &
n@xmath23 ( 10@xmath38 @xmath1 ) & 1.99@xmath39 @xmath31 & n@xmath23 ( 10@xmath38 @xmath1 ) & 218@xmath40 @xmath32 & @xmath22 & 1.8 ( fixed ) & flux at 1 kev@xmath41 & 4.1(@xmath42)@xmath43 @xmath44 & e@xmath45 & 6.4 ( fixed ) & i@xmath45@xmath46 & 2.23(@xmath47)@xmath48 @xmath49 & e@xmath45 & 6.7 ( fixed ) & i@xmath45@xmath46 & 0.70(@xmath50)@xmath48 @xmath51 & e@xmath45 & 6.97 ( fixed ) & i@xmath45@xmath46 & 0.53(@xmath52)@xmath48 @xmath53 & e@xmath45 & 7.06 ( fixed ) & i@xmath45@xmath46 & 0.11@xmath12i@xmath45(6.4 kev ) @xmath33 & @xmath22 & 1.8 ( fixed ) & flux at 1 kev@xmath41 & 0.24(@xmath54)@xmath55 @xmath34 & @xmath22 & 1.8 ( fixed ) & flux at 1 kev@xmath41 & 2.4(@xmath56)@xmath55 @xmath35 & @xmath57 ( kev ) & 0.51@xmath58 & flux@xmath59 & 9.4(@xmath60)@xmath61 @xmath36 & @xmath57 ( kev ) & 0.70@xmath62 & flux@xmath59 & 1.3(@xmath63)@xmath64 & & @xmath41photons @xmath1 s@xmath3 kev@xmath3 ; @xmath46photons @xmath1 s@xmath3 ; @xmath59bolometric flux in erg @xmath1 s@xmath3 in fig .
3 we show the deconvolved best fitting model with all the components .
the best fit parameters are listed in table 1 .
the fit is acceptable , @xmath65/d.o.f.=131/119 . if we do not include the transmission component , and try therefore to account for the spectral component emerging in the pds band with pure compton reflection ( as in ngc1068 , matt et al .
1997 ) , the fit is significantly worse ( @xmath66 dof ; see fig . 2 ) .
it must be noted that the absorber of the nuclear emission , @xmath31 , is compton thick . a proper model , including compton scattering within the matter ( see matt et al . 1999 and matt , pompilio & la franca 1999 for details on this model ) should have then been used , if this absorber has a significant covering factor ( as for example for a torus geometry with no significant vertical density gradient ) .
when this is done , however , the parameters of the various components do not differ significantly from those obtained with the simpler model , apart from the flux associated with the intrinsic emission , which turns out to be about one third that reported in table 1 .
this is because a fraction of the photons emitted towards other directions are now scattered into the line of sight , and the nuclear luminosity required to explain the observed flux is therefore lower . the same fit but without @xmath34 is also acceptable ( @xmath67 dof ) .
in this case , the observed 6.4 kev iron line would entirely originate in the thick absorbing medium , which is possible provided that the covering factor is not much smaller than unity , as we have verified by monte carlo simulations .
therefore , the presence of the cold reflection component is not certain in this source .
if @xmath33 is instead excluded , the fit is significantly worse ( @xmath68 dof ) .
the main result of the _ beppo_sax observation of ngc 6240 is the discovery and measurement of the nuclear emission .
the 210 kev nuclear luminosity , after correction for absorption , is 3.6@xmath1210@xmath69 erg s@xmath3 ( assuming @xmath70=50 km s@xmath3 mpc@xmath3 ) , or about one third of this if the absorber covers a large solid angle , like in a torus . as discussed above ,
the presence of cold reflection is not unambiguously established in this source .
let us , however , assume that it does exist , and take the best fitting parameters listed in table 1 .
the ratio between the 210 kev cold reflection and intrinsic fluxes is about 0.3% for the higher value of the intrinsic luminosity and 3 times larger for the lower value . assuming a torus geometry like that adopted by ghisellini , haardt & matt ( 1994 ) , and consistently the lower figure for the intrinsic luminosity
, we derive an inclination angle of the torus axis with respect to the observer of @xmath635@xmath71 ( @xmath645@xmath71 if the upper figure is instead adopted ) .
it is worth noting that both ngc 6240 and circinus galaxy ( guainazzi et al .
1999 ) appear to be observed at moderate inclinations , while ngc 1068 , which is totally obscured , is possibly seen almost edge - on ( greenhill et al . 1996 , matt et al . 1997 ; but see kishimoto 1999 for a different point of view ) .
this would suggest a decrease in the density of the x
ray absorbing medium with increasing distance from the equatorial plane .
the ratio between 210 kev warm reflection and intrinsic fluxes is about 0.5% ( 1.5% for the lower luminosity ) .
the he and h like lines have ew with respect to the warm reflection component of about 1 and 0.8 kev , respectively . from these values
we estimate the column density of the warm mirror ( e.g. matt , fabian & brandt 1996 ; guainazzi et al .
1999 ) , @xmath72 , to be between 10@xmath73 and 10@xmath38 atoms @xmath1 , roughly consistent with the value estimated from the ratio between the normalizations of the warm reflection and of the intrinsic emission .
the 2 - 10 kev x
ray luminosity is 10 to 30 times lower than the ir luminosity ( calculated as prescribed by mulchaey et al .
1994 ) , consistent with the usual values for seyferts ( e.g. mulchaey et al . 1994 ; awaki 1997 ) . in the case of ngc 6240
, therefore , a ulirg turns out by no means to be a peculiar agn , as far as x ray properties are concerned .
if @xmath740.03 , as typical for qso ( elvis et al . 1994 ) , our result implies that it is the agn that dominates the energy output , and not the starburst as deduced by genzel et al .
( 1998 ) on the basis of the iso spectrum . in the optical , ngc 6240
is classified as a liner .
possibly , the nlr is obscured by the starburst region , whose column density , from our analysis , corresponds to a@xmath7510 .
it is worth noting that ngc 6240 is not the first liner that , when observed in x - rays , turns out to be an agn ( see e.g. , ngc 4945 , done et al .
1996 ; ngc 1052 , guainazzi & antonelli 1999 ) .
differently from them ( but similarly to two other ulirg , iras 23060 + 0505 : brandt et al . 1997 and iras 20460 + 1925 : ogasaka et al .
1997 ) , ngc 6240 has a high x
ray luminosity , and it may be called an obscured qso . while it would be premature to draw conclusions on ulirgs as a class , one can not help noting that the so far elusive type 2 qso should be better searched for in x
rays rather than in the optical band .
we thank the referee , r. antonucci , for valuable comments .
we acknowledge the bepposax sdc team for providing pre processed event files and the support in data reduction .
gm and gcp acknowledge financial support from asi , wnb from nasa ltsa grant nag5 - 8107 . mg acknowledges an esa research fellowship .
bepposax is a joint italian - dutch program . | the direct nuclear x ray emission of the ultraluminous iras galaxy ngc 6240 has been observed for the first time by _ beppo_sax .
it is seen through an absorber with @xmath0 @xmath1 . the 2 - 10 kev x
ray luminosity of @xmath2 erg s@xmath3 definitely proves that a powerful agn is hosted in this galaxy , despite the liner optical appearance . assuming the typical @xmath4 ratio for quasars
, our result implies that it is the agn and not the starburst that dominates the energy output of ngc 6240 . |
it has recently been pointed out @xcite that many phenomena taking place in the sun , especially in its corona and in its magnetic field , are far from being completely understood .
the production by the sun of axion like particles ( alps ) and their subsequent interactions in the solar environment could provide a key to interpreting the physical mechanisms underlying these phenomena . these , and other considerations led to starting a search with the cast magnetic helioscope @xcite for hypothetical alps emitted by the sun in the energy range below 100 ev . the first step of this search , which will be reported here , has involved looking for 2 - 4 ev photons produced in the cast magnet bore by the primakoff @xcite conversion into photons of solar alps in the latter energy range . the short term objective was to efficiently couple a detector system sensitive in the ev energy range to a cast magnet bore and evaluate its background in normal operating conditions .
the long term objective of the effort is attempting to detect , using sensors with the appropriate spectral sensitivity and good enough background , `` low''-energy ( tens of ev s ) photons generated in the cast helioscope by possible interactions of low - energy solar alps .
we will briefly describe the detector system , which has been developed for this purpose under the barbe project financed by the italian istituto nazionale di fisica nucleare ( infn ) , along with the coupling of this system to the cast magnet .
finally , the data taking campaigns will be discussed and a summary of the data presented .
the starting idea of the barbe project is to begin with readily available photon detectors sensitive in the visible range , test them in ideal laboratory conditions , and then design and build an optical system to couple the detectors to one of the bores of the cast magnet .
the devices used in this initial phase were a photomultiplier tube ( pmt ) ( model 9893/350b made by emi - thorn ) and an avalanche photodiode ( apd ) ( model id100 - 20 made by idquantique ) .
the pmt had an active area dia . of 9 mm , with peak sensitivity at 350 nm .
the apd had an active area dia . of 20 @xmath0 m and peak sensitivity at 500 nm .
for both detectors the dark count rate ( dcr ) measured in the laboratory during preliminary tests was about 0.4 hz . to measure the dcr , the pmt was biased at 1950 v and instrumented with an electronic readout chain consisting of nim standard modules .
the apd readout module gave a 2 v , 10 ns wide , output signal which was trasformed into a ttl pulse via a custom circuit .
both detectors were cooled by means of their built in peltier coolers , operating at -20 @xmath1c , and were illuminated by a suitably attenuated blue led source . to demonstrate single photon operation , the count histogram for each detector
was fitted with a poissonian distribution with average equal to 1 .
the dcr was then obtained after cutting the electronic noise pedestal using the fitted curve .
the basic elements of the coupling system were a galileian telescope , a 40 m multimode optical fiber complete with input collimator and an optical switch ( mod .
1x2 made by leoni ) .
the galileian telescope consisted of a 2 inch dia .
, f = 200 mm , convex lens and of a 1 inch dia . , f = - 30 mm , concave lens and was designed to optically couple the 40 mm dia .
bore of the cast magnet into the 9 mm dia .
input collimator of the multimode fiber .
the telescope was mounted directly onto one of the sunset side ports of the cast magnet .
the detectors were placed far away from the magnet in the cast experimental hall .
the optical switch , which can be triggered by external ttl pulses , was used in order to share the light coming from the fiber between the pmt and the apd : each detector could then look at the magnet bore for 50% of the time and at the background for the remaining 50% .
figure [ fig : figure1 ] shows a block - diagram of the layout of the system as mounted on cast .
the overall light collection efficiency was about 50% for the pmt and less than 1% for the apd .
the low efficiency of the apd channel was due to unresolved focussing difficulties .
since the apd data are of inferior quality , only the pmt measurements are considered .
the total number of counts in each measurement is affected by afterpulses generated either by the pmt itself or by its readout electronic chain .
it was found that afterpulses account for 11% of total counts . to eliminate the effect of the afterpulses the mean rate of counts
is calculated , fot both light and dark counts by solving for @xmath2 the equation @xmath3 , where x is the channel number , a is the total number of occurences in all channels , @xmath4 is the number of occurences in the x - th channel measured experimentally . in this way occurrences in channel 0
are not affected by afterpulses .
two measurement campaigns were conducted .
the first one in november 2007 with the telescope attached to the v2 port of the cast magnet , while the second one was in march 2008 with the telescope on the v1 port .
r0.75 in the first run each detector , pmt and apd , looked at the magnet bore for 50% of the time and at the background for the other 50% .
environmental checks and background measurements in different magnet positions and with field on and off were performed .
a total of 45000 s of `` live '' data with the magnet on were taken , 10000 s of dummy solar tracking data and 35000 s of actual sun tracking data .
the second run was conducted using the pmt only , but keeping the switching system in order to have again the detector share its live time equally between signal and background . in this case 45000
s of live data were taken , of which 5000 s of dummy solar tracking data , 20000 s of actual solar tracking data and 20000 s of data with the magnet pointing off the sun center ( 10000 s pointing 0.25@xmath1 to the right and 10000 s pointing 0.25@xmath1 to the left ) .
figure [ fig : figure2 ] shows a plot of the difference of the measured average count rates between `` light '' , when the pmt was looking at the magnet bore , and `` dark '' , when the optical switch was toggled on the other position .
the abscissa axis refers to the different conditions in which data were taken .
taking into account the 1 @xmath5 error bars reported in the plot , no statistically significant difference is found between `` light '' and `` dark '' count rates .
the average background count rate measured for 3 - 4 ev photons during solar tracking was 0.35 @xmath6 0.02 hz for a total of 75000 s of data .
the measurement runs conducted with the barbe detector system demonstrated that it is possible to couple two detectors to the cast magnet via an optical fiber , while preserving a reasonable light collection efficiency ( 50% in the pmt case , that corresponds to 10% of overall system efficiency when taking into account the pmt spectral response curve ) , and without introducing additional noise sources . in 12 data sets taken during solar tracking ( including 2 sets pointing off center ) the background count rate for 3 - 4 ev photons was 0.35 @xmath6 0.02 hz and no significative excess counts over background were observed .
this is the first time such a measurement has been done with an axion helioscope .
the challenge is now to progress to a new detector(s ) with lower intrinsic background , possibly extending the spectral sensitivity to other regions of the energy interval below 100 ev . in the case of visible photons , one could also envision enclosing the cast magnet bore in a resonant optical cavity in order to enhance the axion - photon conversion probability @xcite .
this would however require solving rather complex compatibility problems with the rest of the cast apparatus .
three types of detectors have at this moment been considered for future developments , a transition edge sensor ( tes ) @xcite , a silicon sensor with depfet readout @xcite and an apd cooled to liquid nitrogen temperatures .
the tes sensor promises practically zero background , spectroscopic capability and sensitivity from less than 1 ev up to tens of ev s .
it however requires operation at 100 mk and it has a small sensitive area ( about @xmath7 ) .
the depfet sensors could reach a very low background if used in the repetitive non destructive readout ( rndr ) mode , however they also have a small sensitive area . finally , the cooled apd could be operated relatively easily if one accepts afterpulsing events , which should not pose a problem in a low expected rate environment . on the other hand , the sensitive area would only be about 300 @xmath8 and the spectral sensitivity limited to the visible region .
the present plan is to initially pursue all three possibilities in the hope of identifying the one where progress is faster .
funding for the barbe project has been provided by infn ( italy ) . the support by the gsrt in athens is gratefully acknowledged .
this research was partially supported by the ilias project funded by the eu under contract eu - rii3-ct-2004 - 506222 .
special thanks to m. karuza , v. lozza and g. raiteri of infn trieste for their excellent detector work , and to the entire cast collaboration for wonderful support and assistance .
we also warmly thank the cern staff for their precious help .
99 see for instance k. zioutas , `` the enigmatic sun : a crucible for new physics '' , cern courier , may 2008 .
s. andriamonje _ et al .
_ [ cast collaboration ] , jcap * 0704 * ( 2007 ) 010 [ arxiv : hep - ex/0702006 ] ; kuster , m. _ et al .
_ , new j. phys . ,
* 9 * ( 2007 ) 169 ; abbon , p. _ et al .
_ , new j. phys .
* 9 * ( 2007)170 ; autiero , d. _ et al .
_ , new j. phys .
* 9 * ( 2007 ) 171 ; zioutas , k. _ et al . _ [ cast collaboration ] , phys .
* 94 * ( 2005 ) 121301 .
see for instance g. raffelt and l. stodolsky , phys .
d * 37 * ( 1988 ) 1237 - 1249 .
p. sikivie , d. b. tanner and k. van bibber , phys .
* 98 * ( 2007 ) 172002 [ arxiv : hep - ph/0701198 ] .
b. cabrera , j low temp phys * 151 * ( 2008 ) 82 - 93 ; d. bagliani et al .
, j low temp phys * 151 * ( 2008 ) 234 - 238 . j. kemmer and g. lutz , nucl . instrum . meth . a * 253 * ( 1987 ) 365 ; p. fischer _ et al .
_ , nucl .
instrum .
a * 582 * ( 2007 ) 843 . | we have started the development of a detector system , sensitive to single photons in the ev energy range , to be suitably coupled to one of the cast magnet ports .
this system should open to cast a window on possible detection of low energy axion like particles emitted by the sun .
preliminary tests have involved a cooled photomultiplier tube coupled to the cast magnet via a galileian telescope and a switched 40 m long optical fiber .
this system has reached the limit background level of the detector alone in ideal conditions , and two solar tracking runs have been performed with it at cast .
such a measurement has never been done before with an axion helioscope .
we will present results from these runs and briefly discuss future detector developments . |
delay sensitive applications such as audio / video conferencing and real - time control require low end - to - end delay @xcite . in these applications ,
when the ultimate delay of a data packet exceeds a certain threshold , the packet is useless and has to be dropped .
so , the delay violation probability is a suitable quality of service ( qos ) performance metric here @xcite .
in wireless telecommunication systems and a proper channel condition , transmission with high data rate can be accomplished .
in contrast , in a severe condition , the data rate decreases and probably becomes zero .
accordingly , data must tolerate longer stop at the transmitter buffer before transmission .
therefore , in the wireless channel , deterministic guarantee of total transmission delay for each data packet is not possible .
however , with a genuine design , statistical delay constraint can be meet .
in this respect , the interesting theory of effective capacity has been presented recently @xcite-@xcite .
in contrast to shannon capacity without any restrictions on complexity and delay , effective capacity is defined as the maximum constant arrival rate that a wireless channel can support in order to guarantee quality - of - service ( qos ) requirements such as the statistical delay constraint @xcite-@xcite .
the theory of effective capacity in a general time correlated rayleigh fading channel is introduced in @xcite and in the low signal to noise ratio ( snr ) regime , a closed - form solution for the effective capacity is extracted .
this concept with the statistical qos constraint is completely discussed with a different viewpoint in @xcite for a generic traffic source over a time correlated rayleigh fading channels .
the authors assume a gaussian distribution for the accumulated service rate of the channel , which leads to a solution for the effective capacity of the channels . following these basic researches ,
the effective capacity has been conducted in various communication links such as rate and power adoption in single - input single - output ( siso ) and multiple - input multiple - output ( mimo ) systems @xcite-@xcite , antenna selection ( as ) mimo systems @xcite-@xcite and cooperative and cognitive radio channels in @xcite-@xcite . a mimo wireless communication system is defined as a transmission link where the transmitter and the receiver are equipped with multiple antenna elements to increase channel capacity and data throughput .
digital communication using mimo wireless link , has recently emerged as one of the most significant technical breakthroughs in modern communications such as ieee 802.16 wimax , ieee 802.11 wifi , cellular third and forth generation ( 3g/4 g ) systems @xcite .
hence , for high data rate , multiple antenna structure is assumed here .
study of effective capacity in the mimo channel is a new research topic which is discussed in receive / transmit antenna selection mimo systems in @xcite-@xcite . in @xcite , the effective capacity of a system in two different cases with single receive antenna selection ( ras ) and single transmit antenna selection ( tas ) schemes in quasi - static fading channel is investigated . when the channel state information ( csi ) is not available at the transmitter , single ras scheme is employed . on the other hand , with the available csi at the transmitter ,
single tas is executed .
moreover , an optimal power - control policy is applied to the selected antenna and the effective capacity of the mimo system is derived . in @xcite , a maximum constant arrival rate with the delay quality - of - service ( qos ) guarantee
is obtained in a mimo orthogonal space time block coding ( mimo - ostbc ) system .
the authors have assumed spatially correlated transmit antenna and simple receiver which employs single ras technique . in both @xcite and @xcite , snr maximization
is considered as a criteria in the antenna selection procedure . regarding day by day increase of mobile users and request of high frequency reuse in the networks ,
interference mitigation is considered as one of the main challenge in modern communication systems @xcite-@xcite . in this way
, we discuss mimo - ostbc downlink scenario with co - channel interference here .
this point is a reasonable assumption when the mobile station is near the cell edge or even when cognitive radio users are active in the same frequency @xcite-@xcite . in the downlink ,
the transmitter usually have sufficient area , but , the receiver dimension is small .
therefore , ostbc is employed at the transmitter side and single ras technique is applied at the receiver . for this purpose , the receiver examines their antennas in each utilization of the channel , and select the antenna which maximizes the signal to noise plus interference ratio ( sinr ) .
single ras can simplify the receive rf front end @xcite while some performance of multiple antenna is preserved .
finally , the effective capacity of this mimo - ostbc system with single ras scheme is investigated and approved by numerical simulations .
the contribution of this paper can be summarized as : * closed - form calculation of the effective capacity in the mimo - ostbc downlink channel with the ras scheme , when there are some co - channel interferers working simultaneously * comparing the effective capacity in the mimo - ostbs systems with ras and also miso - ostbc systems which shows the necessity of using ras technique when high qos is required ( see fig .
7 ) this paper is organized as fallows : first , the system model is introduced in section [ sec:2 ] .
section [ sec:3 ] provides a brief introduction to the effective capacity concept .
the effective capacity of a mimo - ostbc system with the ras scheme is explained in section [ sec:4 ] .
finally , the simulation results are presented in section [ sec:5 ] , and section [ sec:6 ] concludes the paper .
the downlink mimo - ostbc system with co - channel interferers is illustrated in fig .
[ fig:1 ] .
the mimo system has @xmath0 transmit and @xmath1 receive antennas and there are @xmath2 single antenna interferers which disturb our mimo receiver in the same frequency .
so , in addition to the additive white gaussian noise , we encounter the co - channel interference at the mimo receiver . in each utilization of the channel ,
@xmath3 symbols are packed , encoded and construct the desired ostbc .
next , this code transmits from @xmath0 antennas at @xmath4 time slots .
therefore , the code rate is considered as @xmath5 .
we assume quasi - static rayleigh flat fading channel in both mimo link and interferer channel .
according to this model , the received symbols by @xmath1 receive antennas at @xmath4 time slots is given by @xmath6 where @xmath7 represents the ostbc transmitted symbols and @xmath8 contains transmitted symbols from @xmath2 co - channel interferers during @xmath4 time slots .
similarly , @xmath9 determine the quasi - static flat fading channels between @xmath0 transmit and @xmath1 receive antennas , @xmath10 is the channel coefficients between @xmath2 interferers and mimo receiver and @xmath11 represents additive white gaussian noise components at @xmath1 antennas and @xmath4 time slots . note that , @xmath12 displays @xmath13 complex matrix set .
we assume single ras scheme in the receiver .
therefore , we encounter multiple - input single - output ( miso ) system for more simplicity . regarding ( [ eq:1 ] ) , the received symbols by the selected antenna at @xmath4 time slots is simplified to @xmath14 where @xmath15 depicts miso channel coefficients ans @xmath16 shows interferers channel components.we will show the elements of @xmath17 and @xmath18 by @xmath19 and @xmath20 respectively . in the miso - ostbc channels without any distortion , received snr is expressed as @xcite @xmath21 where @xmath22 is the average snr of the link .
however , when some interferers are available , the transmitted signals by the interferer users are added to the noise and similar to ( [ eq:3 ] ) , we can write signal to interference plus noise ratio ( sinr ) as @xmath23 here , @xmath24 represents the received power by the @xmath25-th interferers at the miso receiver , @xmath26 is the white noise power and @xmath27 shows the power of @xmath25-th interferer to noise ratio .
note that , in a dense environment or when we have very close interferers , we can assume @xmath28 , so , the sinr in ( [ eq:4 ] ) is approximated as @xmath29 if all interferers have nearly the same received power , we have @xmath30 and then , @xmath31 has gamma distribution . however , when @xmath32 is not the same for @xmath2 interferers , @xmath33 is very tightly approximated as a gamma distributed random number . this claim is demonstrated for the cumulative distribution function ( cdf ) of @xmath34 interferers in fig .
[ fig:2 ] .
the values of @xmath32 for different @xmath2 , have displayed in fig .
[ fig:2 ] .
now , for more simplicity , we assume @xmath35 in the proceeding . .
] since @xmath36 and @xmath37 have gamma distribution ( @xmath38 , @xmath39 ) , then , @xmath40 has beta - prime distribution ( @xmath41 ) @xcite with the probability density function ( pdf ) of @xmath42 and cdf of @xmath43 where @xmath44 stands for the beta function @xcite .
next , we summarize the sinr as @xmath45 where @xmath46 is replaced by @xmath47 , and @xmath47 denotes the signal to interference ratio as well . to obtain the solution of ( [ eq:7 ] )
, we can use ( * ? ? ?
* eq . 2.111 - 2 )
and we get @xmath48 where @xmath49
despite the time - varying nature of wireless channels , network service providers must guarantee a specific qos to satisfy their customers who need real - time multimedia communication . in @xcite , wu and negi defined the effective capacity as the maximum constant arrival rate that a given service process can support in order to guarantee a qos requirement specified by the qos exponent @xmath50 . for a dynamic queuing system with stationary and ergodic arrival and service processes , under sufficient conditions , the queue length process @xmath51 converges in distribution to a random variable @xmath52 such that @xcite @xmath53 where @xmath54 determines a certain threshold , and therefore , we have the following approximation : @xmath55 for a large @xmath54 . for a small @xmath54
the following approximation is shown to be more accurate @xcite : @xmath56 where @xmath57 represents the non - empty buffer probability .
in addition , for the delay experienced by a packet as the main qos metric , we have a similar probability function as @xcite @xmath58 where @xmath59 indicates the tolerated delay , @xmath60 is a delay - bound , and @xmath61 is jointly determined by both arrival and service processes . the statistical delay constraint in ( [ eq:13 ] )
represents the qos which has to be guaranteed for the delay sensitive traffic sources .
it is apparent that the qos exponent @xmath50 has an important role here .
larger @xmath50 corresponds to more strict qos constraint , while smaller @xmath50 implies looser qos requirements .
effective capacity provides the maximum constant arrival rate that can be supported by the time - varying wireless channel under the statistical delay constraint ( [ eq:13 ] ) .
since the average arrival rate is equal to the average departure rate when the buffer is in a steady - state @xcite , the effective capacity can be viewed as the maximum throughput in the presence of such a constraint .
the effective capacity is defined in ( * ? ? ?
12 ) and ( * ? ? ?
6 ) in the correlated channels . however in the uncorrelated case it reduces to @xmath62 where @xmath63 is the time - varying rate of the channel and @xmath64 denotes the expectation . for a specific application with a given statistical delay requirement , the qos exponent @xmath50 can be determined from ( [ eq:13 ] ) .
then , the maximum constant arrival rate of the sources that a wireless channel can support in order to guarantee the given qos , is determined from ( [ eq:14 ] ) . @xmath63 and the effective capacity @xmath65
are further discussed in the next section .
the physical layer of downlink mimo - ostbc system is introduced in section [ sec:2 ] where for obtaining simple and cheap handsets , we assumed single ras technique at the receiver .
in addition , single ras can eliminate practical issues such as mutual coupling and spatial correlation between antennas @xcite .
in our assumed case , in each utilization of the channel , one of the @xmath1 receive antenna is selected .
this antenna has maximum sinr among the all @xmath1 antennas as well . here
, @xmath66 denotes the sinr at @xmath1 different receive antennas , and @xmath67 represents these random numbers in an ordered form where @xmath68 .
therefore , the channel capacity in this scenario is written as @xmath69 where @xmath70 is the total spectral bandwidth , @xmath71 is the frame duration , and @xmath72 is the maximum value in the set of @xmath73 . by inserting ( [ eq:15 ] ) into ( [ eq:14 ] ) , we obtain the effective capacity as @xmath74 and , @xmath75 .
since @xmath76 has beta - prime distribution @xmath77 , the pdf of @xmath72 can be found as @xcite @xmath78^{n-1}\ ] ] where @xmath79 and @xmath80 were determined in ( [ eq:6 ] ) and ( [ eq:8 ] ) respectively .
now , by employing multinomial theorem , we get @xmath81^{n-1}=\nonumber\\ & \left[\frac{(-1)^{m}(m-1)!}{\mathcal{b}(m , k)\prod_{m=1}^{m}(m - m - k)}\right]^{n-1}\sum_{k = n-1}^{(n-1)(m+k-1)}q_{k-(n-1)}\frac{x^k}{(1+x)^{(n-1)(m+k-1)}},\end{aligned}\ ] ] where @xmath82 represents the @xmath83-th coefficient from the multinomial expansion of @xmath84^{n-1}.\ ] ] for simple calculation of @xmath85 , we define @xmath86^t$ ] where @xmath87 shows the transpose operator and @xmath88 and @xmath89 represents the discrete convolution function . here
, @xmath85 indicates the @xmath25-th element of @xmath90 . according to ( [ eq:17 ] ) and ( [ eq:18 ] ) ,
@xmath91 is simplified to @xmath92^{n-1}\sum_{k = n-1}^{(n-1)(m+k-1)}q_{k-(n-1)}\frac{x^{m+k-1}}{(1+x)^{n(m+k-1)+1}}\end{aligned}\ ] ] and we have @xmath93^{n-1}\nonumber\\&\times\sum_{k = n-1}^{(n-1)(m+k-1)}\left[q_{k-(n-1)}\mathcal{b}(m+k , n(m+k-1)+1+\breve{\theta}-(m+k))\right.\nonumber\\&\left.\qquad\qquad\qquad\qquad\times{}_2f_1(\breve{\theta},m+k;n(m+k-1)+1+\breve{\theta};1-\xi_0/mr_c)\right]\end{aligned}\ ] ] where @xmath94 shows gaussian hypergeometric function @xcite .
note that , a closed - form solution for ( [ eq:20 ] ) can be found by using the integral in ( * ? ? ?
* eq . 3.197 - 5 ) .
it is visible that ( [ eq:20 ] ) contains a finite sum of two special math functions and can be evaluated by commercial numeric software . by applying this expression into ( [ eq:16 ] ) ,
eventually , the effective capacity of mimo - ostbc with ras scheme is attained when @xmath2 co - channel interferers are alive .
here we assume @xmath95 , @xmath96 , @xmath97 and for satisfactory results , mont - carlo simulation is repeated @xmath98 times in each step . moreover , for more simple drawing , @xmath99 is plotted in the followings . the normalized effective capacity is plotted in fig .
[ fig:3 ] , [ fig:4 ] and fig .
[ fig:5 ] versus the signal to interference ratio @xmath47 in a mimo - ostbc systems and single ras scheme .
also we assume @xmath100 and @xmath101 , so that , one antenna is chosen in each utilization of the channel .
in addition , the interference to noise ratio is equal to @xmath102 , @xmath103 and @xmath104 in fig . [ fig:3 ] , [ fig:4 ] and fig . [ fig:5 ] respectively and the simulation is executed for different number of co - channel interferers @xmath105 and @xmath106 .
tight agreement between theory and simulation , specially at high interference to noise ratio @xmath107 , or large number of interferers @xmath2 is clear .
therefore , the approximation of sinr in ( [ eq:5 ] ) is quit precise .
furthermore , we observe that increasing @xmath47 leads to increase of the effective capacity as might be expected . again
, increment of interferers causes some grade of decrease in the amount of effective capacity .
hence , large number of interferers have to be avoid for obtaining suitable capacity with the qos guarantee .
mimo - ostbc with ras scheme versus the signal to interference ratio @xmath47 , @xmath102 and different number of interferers . ]
mimo - ostbc with ras scheme versus the signal to interference ratio @xmath47 , @xmath103 and different number of interferers . ]
mimo - ostbc with ras scheme versus the signal to interference ratio @xmath47 , @xmath104 and different number of interferers . ] in fig .
[ fig:6 ] the normalized effective capacity is plotted versus the signal to interference ratio @xmath47 and @xmath103 , @xmath108 and @xmath100 in fig . [ fig:6 ] ( a ) and @xmath109 in fig . [ fig:6 ] ( b ) .
fine agreement between theory and simulation results are clear . in addition , performance improvement is observed when the number of receive antenna is increased and one of them is chosen for signal reception . for more details , the normalized effective capacity of a miso system without receive antenna selection
is also plotted .
therefore , high performance is obtained with the low complexity of miso system when ras technique is applied at our mimo systems . , @xmath103 and @xmath108 when ( a ) @xmath100 ( b ) @xmath109 . ]
+ , @xmath103 and @xmath108 when ( a ) @xmath100 ( b ) @xmath109 . ]
the advantage of ras scheme in the mimo - ostbc compared to the miso - ostbc systems is depicted in fig .
[ fig:7 ] .
we assume a @xmath110 mimo - ostbc with ras scheme and a @xmath111 miso - ostbc systems and plot the normalized effective capacity for three different qos requirement . also , @xmath103 and @xmath108 .
it can be seen that the ras is more efficient when high qos ( high value of @xmath50 ) at moderate to high snr is requested .
this point is specified in the figure where we have @xmath112 effective capacity gain at @xmath113 , @xmath114 effective capacity gain at @xmath106 and @xmath115 effective capacity gain at @xmath116 respectively at @xmath117 .
consequently , ras technique can improve the performance , specially at high qos request , and keep the complexity as low as a miso system .
mimo - ostbc with ras scheme and a @xmath111 miso - ostbc system versus the signal to interference ratio @xmath47 when @xmath103 and @xmath108 . ]
finally , we plot the normalized effective capacity versus @xmath50 in fig .
[ fig:8 ] .
we assume @xmath100 , @xmath101 , @xmath118 and @xmath103 when we have @xmath119 , @xmath108 or @xmath120 co - channel interferers .
similar to the previous figures , the simulation results follow the theory tightly .
the normalized effective capacity decreases as the qos exponent @xmath50 increases .
also , we observe that , by increase of co - channel interferers , the effective capacity decreases dramatically . therefore , using some interference cancellation techniques is necessary when good performance is expected .
note that , all curves converge to a single point at large qos exponent @xmath50 .
mimo - ostbc with ras scheme versus the qos exponent @xmath50 and @xmath118 , @xmath103 with @xmath119 , @xmath108 or @xmath120 co - channel interferers . ]
effective capacity is an interesting topic , so , study of this subject in a practical mimo system with ras scheme under co - channel interference circumstance is considered here . to obtain simple and cheap mimo receiver in downlink
, we adopt ras technique and a closed - form solution for the maximum constant arrival rate with the qos guarantee is extracted .
the solution is tightly accurate when large number of co - channel interferers or some strong co - channel interferers are active in mimo - ostbc environment .
we show that the effective capacity decreases by the increase of interferers or their interference power .
in addition , operation of mimo - ostbc with ras was compared with miso - ostbc systems .
we indicate that ras scheme is more efficient in high qos demand .
1 g. m. su _
_ , qoe in video streaming over wireless networks : prespectives and research challenges , " _ springer wireless netw .
_ , doi:10.1007/s11276 - 015 - 1028 - 7 , aug
et al . _ , exploiting device - to - device communications in joint scheduling of access and backhaul for mmwave small cells , " _ ieee j. sel .
areas commun .
10 , pp . 2052 - 2069 , oct
d. wu and r. negi , effective capacity : a wireless link model for support of quality of service , " _ ieee trans .
wireless commun .
_ , vol . 2 no .
630 - 643 , jul . 2003 .
b. soret , c. a. torres , and j. t. entrambasaguas , capacity with explicit delay guarantees for generic source over correlated rayleigh channel , " _ ieee trans .
wireless commun .
_ , vol . 9 , no . 6 , pp
1901 - 1911 , jun . 2010 .
j. tang and x. zhang , quality - of - service driven power and rate adaptation over wireless links , " _ ieee trans .
wireless commun .
_ , vol . 6 , no . 8 , pp
. 3058 - 3068 , aug .
2007 . j. tang and x. zhang , quality - of - service driven power and rate adaptation for multichannel communications over wireless links , " _ ieee trans .
wireless commun .
_ , vol . 6 , no . 12 , pp
. 4349 - 4360 , dec .
m. lari , a. mohammadi , a. abdipour and i. lee , characterization of effective capacity in antenna selection mimo systems , " _ j. commun . networks _ ,
476 - 485 , oct . 2013 .
m. lari , a. mohammadi , a. abdipour and i. lee , effective capacity in receive antenna selection and spatially correlated mimo - ostbc systems , " in _ 6th int . symp .
telecommun . ,
( ist12 ) _ , tehran , iran , nov .
2012 , pp .
117 - 122 .
m. lari , a. mohammadi , a. abdipour and i. lee , characterization of effective capacity in af relay systems , " _ ieice electron . express _
, vol . 9 , no . 7 , pp .
679 - 684 , apr . 2012 .
l. musavian , s. aissa and s. lambotharan , effective capacity for interference and delay constrained cognitive radio relay channels , " _ ieee trans .
wireless commun . _ , vol . 9 , no .
1054 - 1062 , may 2010 .
d. l. perez _
et al . _ , on distributed and coordinated resource allocation for interference mitigation in self - organizing lte networks , " _ ieee / acm trans .
1145 - 1158 , aug .
d. l. perez _
et al . _ , power minimization based resource allocation for interference mitigation in ofdma femtocell networks , " _
ieee j. sel .
areas commun .
333 - 344 , aug . 2014 .
et al . _ , full duplex techniques for 5 g networks : self - interference cancellation , protocol design , and relay selection , " _ ieee commun . mag .
5 , pp . 2 - 10 , may 2015 .
s. rini and a. goldsmith , on the capacity of the multiantenna gaussian cognitive interference channel , " _ ieee j. sel .
areas commun .
2252 - 2267 , dec . 2014 .
m. youssef _ et al .
_ , routing metrics of cognitive radio networks : a survey , " _ ieee commun .
survey tutorials .
92 - 109 , sept .
m. lari , s. a. bassam , a. mohammadi and f. m. ghannouchi , time - multiplexed single front - end multiple - input multiple - output receivers with preserved diversity gain , " _ iet commun .
_ , vol . 5 , no . 6 ,
789 - 796 , apr . 2011 .
h. shin and j. h. lee , performance analysis of space time block codes over keyhole nakagami - m fading channels , " _ ieee trans .
351 - 362 , mar .
y. yao , j. zheng and z. feng , small - scale variations of cross - polar discrimination in nakagami - m fading channels , " _ ieee commun .
1820 - 1823 , nov . 2012 .
i. s. gradshteyn and i. m. ryzhik , _ table of integrals , series , and products _ , 7th ed .
, academic press , 2007 . c. s. chang and j. a. thomas ,
effective bandwidth in high - speed digital networks , " _ ieee j. select .
areas commun .
_ , vol . 13 , no . 6 , pp . 1091 - 1100 , aug
s. catreux , p. f. driessen and l. j. greenstein ,
simulation results for an interference - limited multiple - input multiple - output cellular system , " _ ieee commun .
4 , no . 11 , pp .
334 - 336 , nov . 2000 .
h. a. david , _ orderstatistics _
, @xmath121 ed . , john wiley&sons , 1981 . | in this paper , delay constrained performance of a multiple - input multiple - output ( mimo ) communication system in a dense environment with co - channel interference is investigated .
we apply orthogonal space - time block coding ( ostbc ) at the transmitter , and for alleviating the high complexity and cost of the mimo system , receive antenna selection ( ras ) scheme is employed in the downlink . here , for simple and cheap mobile handsets , one antenna is chosen at the receiver in each utilization of the channel . under these assumptions , a maximum constant arrival rate with the delay quality - of - service ( qos ) guarantee in a wireless channel
is extracted .
we obtain a closed - form solution for the effective capacity of the mimo - ostbc channel with the ras scheme in a quasi - static rayleigh fading conditions and co - channel interference .
after all , the numerical simulations are provided and verified the theoretical results .
example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore |
understanding the processes that lead to lithium destruction on the main sequence ( ms ) provides powerful diagnostics of stellar structure and evolution and of mixing mechanisms in stars .
rccccl + & & & & & + + & & & & & + name & ( b@xmath4v)@xmath5 & ew(li i ) & t@xmath2 & @xmath1 n(li ) & rv + ( 1 ) & & ( m ) & ( k ) & & info . ( 2 ) + eg 7 & 0.56 & 70 @xmath6 15 & 6061 & 2.7 @xmath7 & s + eg 45 & 0.57 & 44 @xmath6 7 & 6016 & 2.4 @xmath7 & s + at 38 & 0.54 & 35 @xmath6 8 & 6122 & 2.4 @xmath8 & + at 39 & 0.54 & 49 @xmath6 10 & 6114 & 2.6 @xmath8 & + at 1108 & 0.53 & 47 @xmath6 10 & 6189 & 2.6 @xmath8 & sb + at 1109 & 0.62 & 61 @xmath6 15 & 5814 & 2.4 @xmath9 & + at 1225 & 0.51 & 64 @xmath6 7 & 6235 & 2.8 @xmath9 & s + at 2207 & 0.59 & 44 @xmath6 10 & 5908 & 2.3 @xmath9 & + at 2105 & 0.54 & 48 @xmath6 10 & 6110 & 2.5 @xmath9 & s + at 3226 & 0.58 & 49 @xmath6 8 & 5967 & 2.4 @xmath8 & nm + at 4226 & 0.62 & 31 @xmath6 10 & 5795 & 2.1 @xmath9 & + & & & & & + + + & & & & & + + & & & & & + name & ( b@xmath4v)@xmath5 & ew(li i ) & t@xmath2 & @xmath1 n(li ) & rv + & & ( m ) & ( k ) & & info .
+ 23 & 0.58 & 46@xmath6 5 & 5971 & 2.41 @xmath6 0.04 & sb1 + 60 & 0.59 & 48@xmath6 6 & 5924 & 2.38 @xmath6 0.05 & m ?
+ 70 & 0.60 & 42@xmath6 5 & 5884 & 2.28 @xmath6 0.04 & m + 4114 & 0.58 & 51@xmath6 6 & 5951 & 2.44 @xmath6 0.05 & m ?
+ & & & & & + + surveys of li among cluster and field stars have shown that standard models ( i.e. , those that take into account convecting mixing only ) are in contradiction with many of the observational features ( see deliyannis @xcite ; jeffries @xcite ; pasquini @xcite , and references therein for recent reviews ) . focusing on solar type stars , at least two results in strong disagreement with model predictions have been found : _
i ) _ these stars appear to deplete li while on the ms , in spite of the fact that their convective zones bases are too cool to burn li ; _ ii ) _ even more surprisingly , old , otherwise similar solar type stars show different amounts of li depletion .
type stars in the solar age , solar metallicity cluster m 67 ( 4.5 gyr ) show a dispersion in li abundances larger than a factor of 10 ( spite et al .
@xcite ; garca lpez et al .
@xcite ; pasquini et al .
@xcite ; jones et al .
@xcite ) , in contrast with the tight relationship between li abundance and effective temperature ( or mass ) observed for single stars in the younger hyades ( age 600 myr ) . the evolution from the hyades to the age of
the sun seems to be ` bimodal ' , with a fraction ( @xmath3 60 % ) of stars depleting a relatively small amount of lithium , and another fraction , virtually similar to the other one as far as mass and chemical composition are concerned , that undergoes a severe li depletion .
it is important to mention that , whereas no li data for other clusters coeval to m 67 are available , a similar behavior is observed among old solar - type stars in the field ( e.g. , pasquini et al .
the sun with @xmath1 n(li ) = 1.1 belongs to the class of li - poor stars . in summary ,
not only we do not understand the mechanism(s ) that lead to li depletion on the ms , but these mechanism(s ) seem to work differentially for otherwise similar stars . as a possible solution for the dispersion observed among m 67 stars , jones et al .
( @xcite ) speculated that the mixing is driven by ms spin down and angular momentum loss ; in this case , the dispersion in li would be due to different initial rotation rates .
if this were the case , a spread in li should be seen already at ages of 2 gyr or younger ( e.g. , pinsonneault @xcite ) .
more in general , in order to explain the observational evidences , several models have been developed in the last decade , involving more complex physics or / and additional mixing processes ; the validation of these models still needs strong observational support , in particular at ages intermediate between the hyades and m 67 where very few data are available .
determining li in a significant sample of stars in intermediate age clusters is thus extremely important : so far li measurements for clusters of this age have been obtained only for very few objects in ngc 752 ( hobbs & pilachowski , @xcite ) .
we present here li abundances for a sample of solar type stars in the open clusters ic 4651 and ngc 3680 .
these clusters , for which metallicities close to solar have been derived ( friel @xcite ) , have estimated ages of 1.5 2 gyr ( e.g. , meynet et al .
@xcite ; friel @xcite ) ; these data , therefore , allow us to address the problem of li depletion from the age of the hyades to that of m 67 and to put observational constraints on the mechanisms responsible for ms li depletion and on their timescales .
the observations of both clusters were carried out at eso , la silla .
most of ic 4651 stars were observed during three observing campaigns in june 1997 , june 1998 , and may 1999 .
the 3.6 m telescope with the caspec spectrograph was used ; the standard grating ( 31.6 lines / mm ) , the red cross - disperser , the long camera , and eso ccd # 37 ( 1024 @xmath10 1024 24@xmath11 pixels ) were employed .
a few additional ic 4651 spectra were taken in sparse observing runs with feros at the 1.5 m eso telescope and with caspec .
the spectra of ngc 3680 were obtained using caspec .
different slit widths ( from 1.4 to 2 arcsec ) were used in the various caspec runs due to different seeing conditions .
correspondingly , we obtained spectra with resolving powers ranging from @xmath12 to @xmath13 .
the feros spectra have resolution @xmath14 .
the s / n of the spectra is in the range @xmath3 30 to 60 .
data reduction for the caspec spectra was performed using the echelle context within midas and following the usual steps .
feros spectra were reduced using the feros data reduction pipeline .
lithium abundances were obtained in the same fashion as in jones et al .
( @xcite ) for m 67 .
namely , effective temperatures were derived from dereddened b@xmath4v colors using the relationship of soderblom et al .
( @xcite ) . reddening values e(b v)@xmath15 and @xmath16 were assumed for ngc 3680 and ic 4651 , respectively .
abundances were derived from measured equivalent widths ( ews ) using the curves of growth ( cogs ) of soderblom et al .
( @xcite ) .
when the li i 6707.81 doublet was blended with the fe i @xmath17 6707.44 + line , we estimated the contribution of the latter using the prescription of soderblom et al .
( @xcite ) .
information on the sample stars together with our li measurements are given in table 1 .
we carried out a similar analysis for ngc 752 and ngc 188 , using the published li ews ( from hobbs & pilachowski @xcite , @xcite ) and the most updated sources for photometry and membership .
hyades data from thorburn et al .
( @xcite ) were also re - analyzed in a consistent way .
we stress therefore that the li abundances used in this letter are all on the same abundance scale . besides dwarf stars ,
we observed giants and turn off stars in the two clusters ; these data will be used to address different issues and they will be presented in a separate paper .
in figure 1 we plot @xmath1 n(li ) vs. effective temperature for ic 4651 and ngc 3680 together with the hyades , ngc 752 , m 67 , and ngc 188 .
four major features are evident in the figure : * 1 . * the three about coeval clusters , ic 4651 , ngc 3680 , and ngc 752 , have very similar li vs. t@xmath2 distributions , with only a couple of ic 4651 stars lying somewhat above the average trend ( but within 1@xmath18 error bar from it ) .
their average li at a given t@xmath2 is about 0.4 dex ( or a factor 2.5 in abundance ) below the hyades ; * 2 . *
if we consider the upper envelope of the li vs. t@xmath2 patterns , it is evident that the less li depleted stars in m 67 have very similar abundances as the stars in the @xmath0 gyr clusters .
the ngc 188 ( age 67 gyr ) sample is too small and the error bars too large to allow definite statements , but there might be an additional slight li decrease with respect to m 67 ; * 3 . * _ none _ of the three intermediate age clusters shows any significant dispersion for stars warmer than @xmath3 5800 k ; there are no stars cooler than this in ngc 3680 and ic 4651 , while the three ngc 752 stars cooler than 5800 k have significantly different li abundances ( but one is possibly a non member ) ; * 4 . *
as now well known , a large fraction of stars in m 67 have suffered much more li depletion than stars defining the upper envelope of this cluster . under the assumption that all the clusters had the same initial li abundance , the four points listed above in turn suggest that
: * i. * li depletion among solar type stars older than the hyades is a single function of age up to @xmath3 1.52 gyr and t@xmath2 @xmath3 5800 k ; * ii . * the mechanism which drives li depletion between the hyades and the 2 gyr old clusters appears to saturate , or to considerably slow down , at ages of 2 gyr . if stars continued to deplete li at the same rate as between the hyades and the 2 gyr clusters
, one would expect a maximum li abundance ( in the range @xmath19t@xmath2@xmath20 k ) of the order of @xmath1 n(li)@xmath21 at the m 67 age and of the order of @xmath22 at the ngc 188 age .
both clusters show a higher maximum li abundance ; the alternative hypothesis can be made however that li - rich m 67 stars suffered a slower overall li destruction than inferred from younger clusters ; * iii . * li depletion appears to stop or to slow down only for a fraction of the stars , while the other fraction suffers significant additional depletion .
additional depletion may start at ages younger than 2 gyr for stars cooler than 5800 k. on the other hand , we can not exclude that the sun itself at 2 gyr still had a li abundance a factor of 10 higher than its present value . before trying to interpret these results
, we must check whether and to which extent the lack of a major dispersion among the ic 4651 and ngc 3680 clusters may be due to low number statistics .
37 m 67 members are included in the 6200 @xmath23 t@xmath2 @xmath24 k range : excluding the sb2 binary s1045 , 11 of them ( @xmath3 30 % ) are li poor , 23 ( @xmath3 62 % ) are li - rich , and one has a very high li upper limit .
the ic 4651 @xmath25 ngc 3680 merged sample in the same temperature range contains 14 stars ; more specifically , our sample contains about half of the total population of solar type stars in ic 4651 and all the presently known solar type stars in ngc 3680 . in order to have
a similar fraction of li - poor / li rich stars as in m 67 one would have to assume that all the li - rich stars fell in our ( unbiased ) sample or , viceversa , that most of the stars that were not observed by us are li - poor .
whereas this possibility can not be entirely excluded , it would represent a very unusual coincidence .
basically four possible processes have been suggested in the literature for li depletion on the ms : namely , microscopic diffusion , mass loss , and slow mixing either induced by rotation or driven by gravity waves .
* a ) * diffusion is a slow mechanism and could in principle explain the smooth decay of the maximum li abundance between the hyades and the 2 gyr clusters , as well as the lack of a star - to - star scatter up to at least 2 gyr .
diffusion has been shown to be able to explain li in the sun , where he diffusion is validated by heliosismology ( vauclair @xcite ) .
however , diffusion is probably far too slow : model calculations suggest that due to the increasing depth of the convective zone , below 6400 k the diffusion timescales are very large , comparable to the evolutionary timescale , and thus no significant li depletion should occur before the star reaches the very end of its ms lifetime ( michaud @xcite ) .
in addition , it would be difficult to explain the saturation " of li depletion that seems to occur for clusters older than 2 gyr .
* b ) * mass loss has been shown not to be able to reproduce the hyades li vs. t@xmath2 distribution ( swenson & faulkner @xcite ) and , according to the same authors , it would be even more difficult to reproduce the ngc 752 pattern ; by extension , it is reasonable to expect that diffusion would not be able to account for the ic 4651 and ngc 3680 li patterns . in any case , contrary to diffusion , mass loss is probably a too fast mechanism to account for the slow decrease of the maximum li abundance .
* c ) * both rotation ( due to angular momentum loss ) and waves driven mixing mechanisms are rather slow processes and , in this sense , have the right timescales to explain the decay between the hyades and the 2 gyr clusters . *
c1 ) * montalbn & schatzman ( @xcite ) were able to reproduce the hyades pattern at 600 myr with somewhat simplified models including gravity waves and a certain amount of pms depletion ; the speculation therefore can be made that the distribution of older clusters may be fitted as well .
however , it is not clear whether this mechanism would saturate or not ; * c2 ) * as to rotational mixing , quoting from pinsonneault ( @xcite ) , theoretical models of rotational mixing in solar analogs produce a rate of depletion that decreases with age as stars spin down " .
thus , it is plausible that li destruction eventually stops ( or becomes very slow ) when stars are completely spun - down .
however , whereas the existence of a dispersion is a prediction of rotational mixing " , our results indicate that a significant dispersion develops only very late during ms evolution . in figure
2 we show again @xmath1 n(li ) vs. t@xmath2 for ic 4651 , ngc 3680 , ngc 752 , and m 67 .
the rotationally induced mixing isochrones at 1.7 and 4 gyr of deliyannis & pinsonneault ( @xcite ; see also stephens et al .
@xcite ) for two initial rotational velocities are also shown in the plot .
the figure indicates that the virtually null li destruction for li - rich stars observed between the 2 gyr clusters and m 67 is not predicted by the models and , more in general , that the upper envelopes of the observed distributions are not well reproduced by the isochrones corresponding to initial velocities of 10 km / sec , too much li depletion being predicted by the latter models . in addition , as the figure shows , the models predict that a large difference in li between stars with initial velocities of 10 and 30 km / sec should be present already at ages of @xmath3 1.7 gyr . on the other hand , the 4 gyr model with an initial rotational velocity of 30 km / sec
fairly well reproduces the lower envelope of m 67 . in summary
, both diffusion and mass loss can probably be excluded as processes affecting in a major way li depletion on the ms . if any , diffusion may have some effect between m 67 and ngc 188 , if the maximum li abundance in the latter cluster is really slightly below that of m 67 .
gravity waves are a viable mechanism , but they can not reproduce the spread observed in m 67 ; more work needs to be carried out on model calculations to check whether they could fit the intermediate age cluster patterns .
rotational mixing has in principle the requirements to meet the observational constraints , but currently available models do not fit quantitatively the observational data . in order to get a better agreement between observations and models , the latter ones should predict an even slower li depletion for stars with low initial velocities ( of the order of 10 km / sec models with even lower initial velocities should also be computed ) and , furthermore , they should predict no ( or very small ) additional depletion after 2 gyr for these stars ( it is conceivable that stars starting their ms lifetimes with low rotational velocities should have a small degree of internal differential rotation by that age ) . in order to reproduce the lack of a spread at 2 gyr , stars that start their ms lifetimes as rapid rotators
should deplete the same amount of li as initially slow rotators up to @xmath3 2 gyr ; afterwards rotation driven mixing should become much faster . the above discussion is based on the assumption that the three middle aged clusters are representative of stars at 2 gyr , while m 67 is representative of stars at the solar age . in other words , we gave for granted that the m 67 and the middle aged clusters had similar initial properties ( in particular a similar distribution of rotation rates ) and that m 67 at 2 gyr had a similar li distribution as the intermediate clusters or , viceversa , that the latter clusters would show at the m 67 age a li pattern similar to that of m 67 .
however , we can not exclude that m 67 and the intermediate age clusters were characterized by different initial conditions and , consequently , underwent a different rotational history and li evolution .
for example , the intermediate age clusters may have had a very small initial rotational spread with most of the stars being very slow rotators ( but available rotational data for young clusters do not support this ideas ) ; or , as originally suggested by garca lpez et al .
( @xcite ) , there may have been more than one burst of star formation within m67 .
obviously , additional intermediate age and old clusters should be observed in order to test these hypotheses .
we are grateful to drs .
j. andersen and b. nordstrm for providing information on radial velocities and membership for ic 4651 prior to publication .
we thank the referee , dr .
f. spite , for useful comments on the manuscript .
anthony - twarog , b.j . , mukherjee , k. , caldwell , n. , and twarog , b.a .
1988 , aj 95 , 1453 deliyannis , c.p .
2000 , in : stellar clusters and associations : convection , rotation , and dynamos , asp conf .
r. pallavicini , g. micela , and s. sciortino , in press deliyannis , c.p . , and pinsonneault , m.h .
1997 , apj 488 , 833 eggen , o.j .
1971 , apj 166 , 87 friel , e.d .
1995 , ara&a 33 , 381 garca lpez , r.j . , rebolo , r. , and beckman , j.e . 1988 , pasp 100 , 1489 hobbs , l.m . , and pilachowski , c. 1986 , apj 309 , l17 hobbs , l.m . , and pilachowski , c. 1988 , apj 334 , 734 jeffries ,
r. 2000 , in : stellar clusters and associations : convection , rotation , and dynamos , asp conf .
r. pallavicini , g. micela , and s. sciortino , in press jones , b.f . , fisher , d. , and soderblom , d.r .
1999 , aj 117 , 330 meynet , g. , mermilliod , j .- c . , & maeder , a. 1993 , a&as 98 , 477 michaud , g. 1986 , apj 302 , 650 montalbn , j. , & schatzman e. 1996 , a&a 305 , 513 nordstrm , b. , andersen , j. , and andersen , m.i .
1997 , a&a 322 , 460 pasquini , l. 2000 , iau symp .
198 , asp conf .
ser . , eds . m. spite , l. da silva , and j.r .
de medeiros , in press pasquini , l. , liu , q. , and pallavicini , r. 1994 , a&a 287 , 191 pasquini , l. , randich , s. , and pallavicini , r. 1997 , a&a 325 , 535 pinsonneault , m.h . , 1997 , ara&a 35 , 557 soderblom , d.r . , jones , b.f .
, balachandran , s. , et al .
1993 , aj 106 , 1059 spite , f. , spite , m. , peterson , r.c . , chaffee , f.h.jr .
1987 , a&a 171 , l8 stephens , a. , boesgaard , a.m. , king , j.r . , and deliyannis , c.p . 1997 ,
apj 49 , 339 swenson , f.j . , and faulkner , j. 1992 , apj 395 , 654 thorburn , j.a . ,
hobbs , l.m .
, deliyannis , c.p . , and pinsonneault , m.h .
1993 , apj 415 , 150 vauclair , s. 1998 , space sci . | we present li abundances for 14 solar type stars in the intermediate age ( @xmath0 gyr ) clusters ic 4651 and ngc 3680 .
the @xmath1 n(li ) vs. effective temperature distributions are compared with those of the similar age cluster ngc 752 , of the younger hyades ( 600 myr ) and of the older m 67 ( 4.5 gyr ) and ngc 188 ( 67 gyr ) clusters .
neither ic 4651 nor ngc 3680 show the dispersion in li which is observed in m 67 .
the 2 gyr clusters have very similar li vs. t@xmath2 distributions ; in addition , stars in the upper envelope of the m 67 distribution have the same li content as stars in the 2 gyr clusters , suggesting that either they have not suffered any significant depletion between @xmath3 2 and 4.5 gyr or they had a much slower li depletion .
mechanisms that lead to li depletion on the main sequence are discussed in the light of these observations .
none of the existing models seem to reproduce well the observed features . |
in optical communication systems , information is often carried by states with gaussian statistics ; the coherent state is a simple and very important example .
the processing of these states in the quantum domain is now a major concern toward attaining ultimate capacity of bosonic channels @xcite , and also the construction of quantum networks and computers @xcite . toward these applications , it is important to generate and manipulate superpositions of pure gaussian states .
in fact , certain types of such superposition states can be used as ancillary states to build up a universal set of operations to construct an arbitrary processor @xcite .
( see also ref @xcite for its experimental perspectives . )
these states are essentially non - gaussian states , and are easily destroyed even by a small amount of losses . in practice , it is not easy to prepare in labs the state that preserves enough purity and useful non - gaussian properties
. therefore , one needs to develop the methods how to stabilize and/or purify such states .
a simple and more experimentally feasible example is a superposition of two coherent states with small amplitudes .
such a state can be conditionally generated from a squeezed vacuum ( gaussian state ) by using the photon number measurement @xcite .
recently , a proof - of - principle experiment of this idea was successfully presented in ref @xcite , where an on / off - type detector ( which clicks upon the arrival of a photon ) was used rather than a photon number resolving detector .
there is , however , still a distance between the observed state and the ideal one due to various losses . in this paper , we develop a feasible scheme to purify such coherent superposition states ( csss ) that has undergone linear dissipation .
we consider a specific scenario of the purification , but it meets the current experimental situations and is based only on the use of a single beam splitter and a single homodyne detector , which probabilistically selects out favourable events .
the scenario we treat is the following ; consider a css of an equal superposition of two distinct coherent states , @xmath0 , where the amplitude @xmath1 and the phase @xmath2 are a priori known and @xmath1 is relatively small .
this state decoheres into a mixed state by the linear loss and represented by a classical mixture of a css and two coherent states .
our task is then to amplify the probability of the css as efficiently as possible .
the css itself has also been discussed as a resource of various optical quantum information processing protocols @xcite . as a related topic to our study , glancy _ et al . _
have proposed a method to restore a css , whose weights and phase information of the superposition are unknown , by applying a three - bit quantum error correcting code @xcite .
however , this proposal requires almost perfect photon number resolving detectors and pure csss as ancillary states . in our study
we make no use of non - gaussian ancillary states in the purifaction method .
the schematic of our proposal is shown in fig .
[ fig : scheme ] .
a fraction of the input state is tapped off by a beam splitter and subsequently measured with a homodyne detector . from the measurement outcomes we try to distinguish the css from the mixture of coherent states through the differences of their wave functions .
based on the success of this estimation , the coherence of the css is retrieved through selection of favourable events .
we show that homodyne measurement is one of the optimal measurements for the most efficient purification given such a simple setup . using the purification protocol in fig . [
fig : scheme ] , the amplitude of the coherent state components in the ccs is inevitably degraded due to the conditional selection of events .
so we then address the possibility to recover the amplitude by using the css amplification scheme proposed in refs @xcite , where it has been shown that a css with small amplitudes of its constituents is equivalently generated by squeezing a single photon state and the amplitudes can be amplified using linear optics and on / off detectors .
this paper is organized as follows . in section [ sec:2 ] , the influence of linear dissipation to the css is briefly reviewed . in section [ sec:3 ] ,
we discuss the purification scheme in general terms without specifying the exact measurement strategy applied to the tapped off state . in section [ sec:4 ] ,
we consider the example where a homodyne detector is used and show that it is one of the optimal measurement strategies .
in section [ sec:5 ] , we discuss the amplification of css using two csss that has suffered from linear loss , and finally in section [ sec:6 ] we conclude . . ]
in this section we briefly summarize the evolution of css s under linear dissipation .
let us first define a superposition of two coherent states ( css ) with the amplitudes @xmath3 and the relative phase @xmath2 as @xmath4 where @xmath5 we assume @xmath6 to be real without loss of generality . in practice ,
the most dominant loss in optical channels is due to a linear interaction between the signal mode and the vacuum environment .
this is modeled by coupling the signal mode to the environment , letting them evolve for some time and finally trace out the states of the environment .
a completely positive ( cp ) map describing such a linear interaction between the signal and the environment is given by @xcite @xmath7 , \end{aligned}\ ] ] where @xmath8 and @xmath9 are the superoperators defined as @xmath10 for an arbitrary operator @xmath11 .
the transmission of the environment is denoted by @xmath12 and thus the linear loss is given by @xmath13 .
using the transformation @xmath14
\nonumber\\ & & \times| \alpha_1 \sqrt{\eta } \rangle \langle \alpha_2 \sqrt{\eta } | .\end{aligned}\ ] ] we find that the css , @xmath15 where @xmath16 is a mixed state of two coherent states @xmath17 . here
the fraction of the original state in this mixed state is given by @xmath18 as expected , the linear lossy interaction with the environment transforms the original pure css state into a classical mixture .
the aim of this paper is to develop a method by which the decohered state can be probabilistically purified to produce a smaller number of states with higher purity , or in other words ; the aim is to increase the value of @xmath19 .
finally , we note that the purity defined by @xmath20 $ ] is not a suitable figure of merit for our purpose .
one can easily show that , for the mixed css in eq .
( [ eq : decohered_cat ] ) , a smaller amplitude gives a higher purity of the state even for small @xmath19 .
this is because such a state is close to a vacuum state which is completely pure .
however , it is obvious that it does not mean the state is close to a pure css .
our proposed scheme for probabilistic purification of csss is illustrated in fig .
[ fig : scheme ] .
the state under interrogation is mixed with the vacuum at a beam splitter with the transmission coefficient @xmath21 .
the reflected beam is subsequently measured using a positive operator - valued measure ( povm ) generally described by a set of positive operators @xmath22 with @xmath23 .
depending on the outcome of this measurement , the transmitted state is probabilistically purified : if the outcomes fall within a certain detection range , the transmitted state is purified , otherwise the state turns out to be worse and is discarded .
let us first describe the function of the purification scheme when an already pure css @xmath24 is injected into it . in this case
, the output of the beam splitter ( described by the unitary operator @xmath25 ) is given by the entangled state @xmath26 where @xmath27 is the reflectivity of the beam splitter .
if we now consider only one output and trace out the other , the pure css is transformed into a mixed state as a result of the coupling to the vacuum environment .
the purification protocol now works by making a measurement on the `` environmental '' mode ( mode b ) and selecting the outcomes that map @xmath28 back onto a pure css at the output .
it implies that the measurement outcomes for successful events have to satisfy @xmath29 where the superscript @xmath30 denotes the successful events . the unnormalized conditional output for such outcomes is given by @xmath31 ) , it is possible to recover the purity of the input css .
let us now consider the purification of a decohered css .
the input to the purifier is now the mixture @xmath32 the reflected part of the state is measured using the above mentioned povm and events corresponding to condition ( [ eq : condition ] ) are selected .
the resulting density operator of the output state conditioned on this measurement is given by @xmath33 where @xmath34 \nonumber\\ & = & \left|\langle\pi_j^s|\sqrt{r}\alpha\rangle\right|^2 \frac{1+\cos(\varphi+\theta_j ) \ , e^{-2t\alpha^2 } } { 1+\cos\varphi \ , e^{-2\alpha^2 } } \end{aligned}\ ] ] and @xmath35 \nonumber\\ & = & \left|\langle\pi_j^s|\sqrt{r}\alpha\rangle\right|^2 \label{eq : det_probability_0}\end{aligned}\ ] ] correspond to the probability distributions for detecting the outcomes @xmath36 in the states @xmath37 and @xmath38 , respectively .
as a consequence , one finds the fraction of the css after the purification as @xmath39 the condition for successful purification @xmath40 is then given by @xmath41 the maximum purification efficiency is achieved when @xmath42 is minimized .
there are two free parameters controlling the magnitude of this fraction , namely @xmath43 and @xmath21 .
first we note that the phase @xmath43 is optimized , that is @xmath42 is minimized when @xmath44 .
second , it is clear that the fraction ( [ eq : purif_codition ] ) decreases with decreasing transmissions @xmath21 .
this , on the other hand , means that the amplitude of the css is decreasing .
therefore , there is a trade - off between the efficiency of the purification protocol and the amplitude of the purified css .
since large amplitude csss are desired for many applications , we consider the possibility of amplifying the degraded css in sec . v.
the optimal povm for the purification protocol is therefore the one that obeys eq .
( [ eq : condition ] ) and simultaneously minimizes the fraction ( [ eq : purif_codition ] ) by satisfying the constraint @xmath45 .
these conditions provide a nice flexibility to chose among different measurement strategies .
two well known measurement strategies that satisfy the requirements are photon number resolving detectors and homodyne detectors .
if however the photon number resolving detector is used to execute the purification protocol , successful demonstration will only be obtained if the phase between the coherent state components in the css , is @xmath46 or @xmath47 .
furthermore , these detectors are experimentally very challenging to construct and thus , to date , they are not an integral part of the standard lab . fortunately , homodyne detectors ( which have matured much further than photon number resolving detectors ) also meet the requirements for successful purification .
in fact by using homodyne detectors there are no restrictions on the phase @xmath2 . in the following section we investigate in greater details the purification of css using homodyne detection .
a homodyne detector measures a quadrature of the electromagnetic field , with the specific quadratures being determined by the phase , @xmath48 , of the intrinsic local oscillator .
it is mathematically described by an operator @xmath49 which projects the state onto a continuous set of quadrature eigenstates @xmath50 with @xmath48 denoting the quadrature phase .
specific and important examples are the position and momentum eigenstates corresponding to @xmath51 and @xmath52 , respectively .
the probability amplitude associated with the detection of a quadrature eigenvalue , @xmath53 , from the coherent state @xmath54 by homodyne detection is given by @xmath55.\end{aligned}\ ] ] the necessary condition for our purification scheme to work as quoted in eq .
( [ eq : condition ] ) is fulfilled when @xmath52 , that is when the momentum eigenstate is measured . in that case
we obtain @xmath56 from which we see that the phase shift introduced by the measurement depends on @xmath57 as @xmath58 note that the homodyne measurement always allows us to choose the appropriate @xmath57 to optimize @xmath59 for any given @xmath2 . in the following we set @xmath60 . the unnormalized conditional output corresponding to eq .
( [ eq : cond_output ] ) is then given by @xmath61 and one obtains the probability distributions for detecting the @xmath62 and @xmath63 as @xmath64 and @xmath65 respectively . with @xmath66 ( solid line ) and a mixture of coherent states @xmath67 ( dotted line ) .
@xmath68 and @xmath69 .
the purification protocol is successful when the measurement outcome falls within the interval indicated by the dashed vertical lines . ] with @xmath70(solid line ) and a mixture of coherent states @xmath67 ( dotted line ) . @xmath68 and @xmath69 .
the purification protocol is successful when the measurement outcome falls outside the interval indicated by the dashed vertical lines . ]
[ fig : cond_k_p ] and [ fig : cond_k_m ] show these probability distributions for @xmath46 and @xmath47 , respectively .
the probability distributions @xmath67 , which is represented by the dashed curves , are gaussian , while the probability distributions for the css , @xmath71 , ( represented by solid lines ) are non - gaussian due to the quantum coherence between the two coherent state components . as discussed above
, these distinct differences between the probability distributions enable the purification protocol : it succeeds when the absolute quadrature value falls within a certain threshold for @xmath66 and outside a certain threshold for @xmath72 , since in these regions the probability for a css to occur is larger than the probability for @xmath63 to occur .
the efficiencies of the purification protocol for @xmath46 and @xmath73 are illustrated in figs .
[ gainp ] and [ gainm ] , respectively . in case of @xmath46 , the purification condition of eq .
( [ eq : condition ] ) is satisfied in the range given by @xmath74 and the efficiency of the purification protocol is optimised for @xmath75 .
the protocol works more efficiently when the initial css has the phase @xmath73 . in that case , purification is achieved in the range given by @xmath76 and is maximized at @xmath77 where the initial input phase @xmath73 is transformed into @xmath46 after purification .
the maximum increase of @xmath78 as a function of @xmath79 for @xmath80 and @xmath81 is depicted in fig .
[ gain - p ] , and we see that the improvement for these parameters is highest around @xmath82 . in figs .
[ gain - alpha ] and [ gain - t ] , we show how the purification efficiency depends on the amplitude @xmath1 and the transmittance of the beam splitter @xmath21 , respectively , for an input state with @xmath83 . from fig .
[ gain - alpha ] , we see that the purification of states with @xmath66 ( represented by the solid curve ) is most effective around @xmath84 whereas for states with @xmath70 the efficiency increases with smaller amplitudes . in fig .
[ gain - t ] , where the purification efficiency is plotted versus the transmission coefficient , it is evident that small transmission coefficients are advantageous .
however , as already pointed out above , small transmission coefficients are on the other hand not desirable since it degrades the amplitude of the state .
note that , obviously the purification does not work at @xmath85 .
although , in fig
. [ gain - t ] , @xmath86 is still larger than 1 for @xmath73 at @xmath85 , the probability amplitude @xmath87 goes to zero and thus this event never happens .
these results , therefore , show that our scheme can be used to purify small amplitude csss , which are exactly the ones that recently has been produced @xcite .
however , the reduction of amplitude of the css is unavoidable in our scheme and it is therefore desirable to amplify the amplitude of the csss after purification .
this is the topic of the next section . before ending this section
, we address the effect of practical losses in homodyne detection . in practice ,
the efficiency of homodyne detection is less than unity due to the imperfect quantum efficiency of photodiodes , misalignment , mode mismatch , losses of optics etc .
such non - ideal homodyne detection can be modeled by a lossy quantum channel defined in eq .
( [ eq1.1 ] ) plus a perfect homodyne detector @xcite .
the effect of this imperfection in our scheme is estimated by replacing @xmath12 with @xmath88 in eq .
( [ eq : p ] ) where @xmath89 is the effective efficiency of the homodyne detection .
however , it should be noted that the recent quantum network experiments in cw domain at the wavelength of si - photodiode can achieve sufficiently high homodyne effciency , e.g. @xmath90 in @xcite .
for example , with @xmath73 , @xmath68 , @xmath69 , and @xmath83 , the purified outputs will be @xmath91 and @xmath92 for @xmath93 and @xmath94 , respectively .
( although the experiment in @xcite was performed in pulsed regime , the feasibility of non - gaussian experiments with cw sources has also been discussed in @xcite for example . ) conditioned on the homodyne detection outcomes .
the phase of the local oscillator is chosen to measure the quadrature @xmath60 .
@xmath68 , @xmath69 , and @xmath83 . ] conditioned on the homodyne detection outcomes .
the phase of the local oscillator is chosen to measure the quadrature @xmath60 .
@xmath68 , @xmath69 , and @xmath83 . ] on @xmath79 for the css with @xmath46 ( solid line ) and @xmath73 ( dashed line ) where the outputs are conditioned on the measurement outcomes @xmath95 and @xmath96 , respectively . @xmath68 and @xmath69 .
the dotted line denotes @xmath97 as a reference . ]
( solid line ) and @xmath73 ( dashed line ) .
@xmath68 and @xmath83 . the measurement outcome is selected to satisfy the optimal condition @xmath98=0 . ] ( left ) and the probability density @xmath99 ( right ) on the transmittance of the tapping beam splitter @xmath21 for @xmath46 ( solid line ) and @xmath73 ( dashed line ) . @xmath83 and @xmath81 .
the measurement outcome @xmath100 is selected to satisfy the optimal condition @xmath101 . ]
lund _ et al . _ have proposed a scheme to conditionally produce a large amplitude css from two small amplitude csss by using two 50/50 beam splitters , an ancillary coherent state and two on / off photon detectors @xcite , and later on the scalability of this scheme was studied in details @xcite .
they pointed out that when one uses a weakly squeezed single photon state corresponding to a small css ( which in practice inevitably consists of a mixture between the single photon state and a vacuum state ) it is possible to reduce the squeezed vacuum component during the amplification process . in this section
, we discuss the application of such a scheme to obtain a larger amplitude css from two decohered csss .
we show that , the decoherence is , unfortunately , in most cases increased as a result of the amplification process .
the scheme proposed in ref .
@xcite is illustrated in fig .
[ fig : amplification ] .
two copies of the csss with small amplitude are combined on a 50/50 beam splitter .
then one of the outputs is mixed with a coherent state with amplitude @xmath102 on a 50/50 beam splitter .
subsequently the two outputs are measured using two on / off photon detectors .
if both detectors click , the remaining quantum state is projected onto a larger css .
in addition to the amplification of the amplitude , the projection process also filter out the squeezed vacuum component when the inputs are two mixed states of weakly squeezed single photon and vacuum . now , suppose that the input is two copies of the decohered css @xmath103 defined in eq .
( [ eq : initial_state ] ) .
then , the input state is described by the tensor product @xmath104 and in the conditioned amplified output state , only the first term is mapped onto a large css whereas all other terms are transformed into a mixture of coherent states . as a result
, the fraction of a css in the conditioned output states are approximately degraded as @xmath105 .
more precisely , the output probabilities @xmath106 and @xmath107 with inputs @xmath108 and @xmath109 , respectively , are given by @xmath110 note that the conditioned output css with the probability @xmath111 is always @xmath112 @xcite .
the parameter region where the inputs are purified during the conditional process can be evaluated from the condition @xmath113 .
first , we find that if the phase @xmath2 is @xmath114 ( that is @xmath108 is amplified ) , the inequality can not be satisfied .
therefore , in this case , the output probability @xmath106 is always smaller than the input probability , or in other words , the decoherence is always amplified .
however , if the amplitude of decohered css with @xmath70 ( that is @xmath109 ) are amplified , then the inequality @xmath115 can be satisfied if @xmath116 this implies that the average photon number of the input css must be @xmath117 for successful amplification of the amplitude without amplifying the decoherence .
finally , we investigate the possibility to repeat the purification process by concatenating this amplification scheme .
the situation considered here is illustrated in fig .
[ fig : concatenation ] . two initial states
@xmath118 are purified in parallel by the purification processes with @xmath80 .
once the purified outputs @xmath119 from both processes are simultaneously obtained , the outputs are injected into the amplification process to obtain the final output which has the original amplitude @xmath1 .
let us now determine whether the output probability @xmath78 is larger than the input probability @xmath79 for such a machine : since the phase of the css obtained from the amplification process is always @xmath120 , we assume the initial states to have @xmath46 .
we then look for parameters where @xmath40 by replacing @xmath79 in eq .
( [ eq : amp_out ] ) by @xmath121 ( which is the output probability from the purification step ) .
however , unfortunately , a parameter interval fulfilling such criterion does not exist , and we conclude that for all possible amplitudes @xmath1 , the decoherence is amplified along with the amplitudes .
it suggests that when the goal is to generate large amplitude csss from smaller ones by using the conditional processes with linear optics proposed in ref .
@xcite , we have to be careful about the effects of linear loss even if the initial states contain only a very small fractions of mixtures of coherent states .
as we have shown , the decoherence is in most cases amplified during the conditional generation of large csss .
in this paper , we have investigated a simple scheme to recover the quantum coherence of decohered coherent state superpositions .
the scheme we have examined is based on partial estimation of the state by tapping off a part of it using a beam splitter , measuring the reflected part by homodyne detection and finally select out favourable events that maps the decohered state onto a purer state .
we derived the condition for the measurement to maximize the efficiency of the purification and showed that homodyne measurement is one of the optimal strategies .
since homodyne measurement is a well developed technique , our purification scheme is experimentally easier to implement than previously proposed purification protocols .
we therefore believe that our scheme can be directly implemented in current on - going experiments on the generation of css to compensate for losses and increase the nonclassicality of the states .
in addition , we showed that this simple measurement - induced scheme inevitably causes further amplitude degradation and is not suitable for the purification of large amplitude css with @xmath122 .
moreover , we have discussed the possibility to restore the degraded amplitude by the linear optical scheme proposed in @xcite .
if the initial csss are decohered due to a linear loss , the scheme is increasing not only the amplitude but also the degree of the mixture .
we showed that , even combining this scheme with the purification process , it is still hardly possible to recover the coherence of css .
note that , also in the context of entanglement purification , it has been suggested in @xcite that the purification of entangled coherent states decohered by linear loss also seems to be difficult by using only linear optics and photon counting without ancillary csss .
these results emphasize the importance of the further investigations of the purification scheme . | we present a simple protocol to purify a coherent state superposition that has undergone a linear lossy channel .
the scheme constitutes only a single beam splitter and a homodyne detector , and thus experimentally feasible . in practice
, a superposition of coherent states is transformed into a classical mixture of coherent states by linear loss , which is usually the dominant decoherence mechanism in optical systems .
we also address the possibility of producing a larger amplitude superposition state from decohered states , and show that in most cases the decoherence of the states are amplified along with the amplitude . |
the spontaneous appearance of cooperative motions called dynamic heterogeneity ( dh ) when liquids are supercooled @xcite has been extensively studied during the last decade , as scientists were searching for a cooperative mechanism at the origin of the glass - transition . despite large efforts in that direction of research however , the origin of the dynamic heterogeneity and the effect of these cooperative motions on the liquid dynamics still remain elusive .
it is widely thought that dhs appear due to the presence of some defects inside the low temperature liquid , that could be structural ( a local packing fluctuation ) or purely dynamical ( excitations ) .
the resulting excitations ( arising directly or from structural defects ) then grow due to facilitation@xcite processes , eventually disappearing at large timescales due to thermal diffusion . in 2004 ,
asaph widmer - cooper and coworkers@xcite showed that dhs appear preferentially in some regions of the liquid , a result that strongly suggests a connection between the local structure and the dhs . in 2007 ,
laura kaufman s group @xcite studied the effect of probes larger than the medium molecules on the dynamics of supercooled liquids .
they found that the dynamics slows down or accelerates depending on the probe s roughness , while the dhs increase in both cases , and concluded that probes can promote dynamic heterogeneities . in 2009
, we found that dynamical defects created by probes isomerizing inside the medium can strongly promote dhs@xcite while accelerating the dynamics , results that were confirmed by experiments and simulations@xcite . following these works , results on colloids@xcite , and theoretical findings@xcite , in order to test the hypothesis of a packing fluctuation origin of the dhs
, we investigate in this paper the effect of small structural defects on the dynamics of a model supercooled liquid and on the strength of dynamic heterogeneity .
we focus our work on probes of sizes comparable with the size of the medium s molecules in order to model small cages fluctuations .
we also focus on the possibility for some probes to destroy the dynamic heterogeneities . in agreement with the results of kaufman
s group we find that large probes promote the dhs in our liquid , however we also find that small probes can destroy the heterogeneities .
in this work we use molecular dynamics simulations@xcite ( md ) to simulate the dynamics of a fragile@xcite supercooled liquid when small or large probe molecules are diluted inside .
this simulation method permits to gain information on the motion of each molecule of the medium provided that the interatomic potentials are known with enough accuracy , and is thus an invaluable tool to unravel condensed matter physics phenomena@xcite at the microscopic level .
we model the molecules@xcite of the medium as constituted of two atoms ( @xmath0 ) that we rigidly bond fixing the interatomic distance to @xmath1@xmath2 .
each atom of our linear molecule has a mass @xmath3 g/@xmath4 where @xmath4 is the avogadro number .
atoms of the set of linear molecules constituting our liquid interact with the following lennard - jones potentials : @xmath5 with the parameters : @xmath6 kj / mol , @xmath7 kj / mol , @xmath8 and @xmath9@xmath2 .
our mean molecule is thus @xmath10@xmath2 long and @xmath11@xmath2 wide . with these parameters
the liquid does not crystallize when supercooled even during long simulation runs@xcite .
this model has been described and studied in detail previously@xcite and was found to display the typical behaviors of fragile supercooled liquids .
as they are modeled with lennard - jones atoms , the potentials are quite versatile . due to that property , a shift in the parameters
@xmath12 will shift all the temperatures by the same amount , including the glass - transition temperature and the melting temperature of the material .
we add to that medium @xmath13 smaller or larger molecules ( the probes ) intended to create small localized perturbations in the medium structure .
we have @xmath14 probe molecules diluted within @xmath15 medium molecules inside the simulation box .
this number of molecules lead to a simulation box size of @xmath16 , that insures the disappearance of size effects@xcite in our liquid at the temperature of study .
the probes are similar diatomic molecules defined by the parameters : @xmath17 , @xmath18 , @xmath19 , @xmath20 , @xmath21@xmath2 , @xmath22 , @xmath23 and @xmath24 .
the density is set constant in our calculations at @xmath25 .
when rescaled , or in dimensionless units , that density value is larger than the density of the original model@xcite , and thus leads to a more viscous medium .
we use the gear algorithm with the quaternion method@xcite to solve the equations of motions with a time step @xmath26 @xmath27 above @xmath28 and @xmath29 @xmath27 below that temperature .
the temperature is controlled using a berendsen thermostat@xcite . to prevent aging processes we equilibrate the supercooled liquid during @xmath30 ns , a time much larger than the @xmath31 relaxation time at the lower temperature studied , before recording the simulation results .
we use periodic boundary conditions .
we expect slower motions around the large inclusions and more rapid motions around the small inclusions .
this modification of the molecular motions around the defects will result in a local heterogeneity of the mobilities , that may induce or destroy dynamic heterogeneities if facilitation mechanisms are present in the system .
we thus expect a modification of the liquid s dynamic heterogeneity around the inclusions .
( color online ) radial distribution function @xmath32 between the centers of masses ( com ) of the medium molecules ( dashed blue line ) and between the average molecules and the defects ( continuous red line ) .
( a ) large defects ; ( b ) small defects .
the @xmath32 between the average molecules does nt change in our simulations when defects are included , due to the small proportion of defects .
the temperature is t=120k . + _ due to the small percentage of probes
, we do not find any modification of the mean structure of the liquid when we replace a few ( @xmath13 ) molecules of the medium by larger or smaller probe molecules .
the local structure is however modified around the probe as shown in figure 1a and 1b .
figure 1 suggests that the cage is smaller around the large probes and slightly larger around the small probe , as the first peak of the radial distribution function is shifted to a smaller distance for large probes than for medium molecules .
the figure also shows that the local structure around the probe is only slightly modified when the small probes are used .
( color online ) radial distribution function @xmath32 between the centers of masses ( com ) of the medium molecules and of the defects ( continuous red line ) at a temperature t=120k ( continuous red line ) and t=165k ( dashed green line ) .
( a ) large defects ; ( b ) small defects . + _ figure 2 shows that the temperature modifies the structure around the large probes , increasing the probability of unfavored geometrical configurations .
by contrast , for small probes the structure remains the same around the probes at low and high temperature .
however figures 3 and 4 show that both probes modify the dynamics .
this modification of the dynamics begins around @xmath33 and increases when the temperature drops . _
( color online ) mean square displacement of ( inset ) the defects or ( main figure ) the medium molecules , when the defects included are : small ( blue dashed line ) , large ( continuous red line ) , or we are in a pure liquid ( dotted green line).t=120k .
+ _ because the cages around the large probes are smaller than the medium s cages , and these probes are larger than the medium s molecules , we expect a decrease of the free volume around large probes .
the decrease of the free volume around large probes then results in a decrease of the diffusion around these probes .
similarly we expect an increase in the free volume around small probes resulting in an acceleration of the diffusive motions . to verify that picture we evaluate the free volume @xmath34 around the probe from the height of the plateau of the mean square displacement of the probe ( see the inset of figure 3 ) .
if @xmath35 is the plateau s height , we have @xmath36 .
figure 3 shows that @xmath37 ( @xmath38@xmath39 , @xmath40@xmath39 , @xmath41@xmath39 ) leading to the same relation for the free volumes , with @xmath42@xmath43@xmath44 and @xmath45@xmath44 .
the free volume for small probes is thus larger than for the mean medium molecules , leading to the increase of the diffusion for small probes that we observe in the inset of figure 3 . as a result
the inclusion of small or large molecules respectively increase or decrease the mean square displacements around the perturbation .
the defects thus slow down or accelerate the dynamics around them for large or small inclusions respectively .
the effect of these modifications of the local dynamics around the probes results in a modification of the whole dynamics as shown in figure 4 .
we see on the figure that the global diffusion coefficient follows the same trend and decreases for large probes ( or increase for small ones ) . we note that the modifications of the msds appear at the end of the plateau regime , around the characteristic time @xmath46 for which the cooperative motions are maximum
, however the probes behave differently as shown in the inset of figure 3 . for the probes
the msds change at the beginning of the plateau .
we also note that this effect is temperature dependent and disappears at high temperature as shown in figure 4 .
the fragility , seen here as the diffusion coefficient variation from the pure exponential arrhenius law @xmath47 , increases when large probes are included and decreases for small probes . as the increase of the activation energy when the temperature drops
is usually explained by the increase of cooperative motions , that result suggests that the cooperativity increases for large probes and decreases for small ones .
this result suggests that the cooperative motions increase when large probes are used and decreases for small probes .
we will see below that the dhs follow that behavior .
( color online ) diffusion coefficient ( @xmath48 ) for the medium molecules , versus temperature when small or large probes are included , compared with the diffusion coefficient of the pure liquid in the same conditions . from top to bottom : large probes ( red solid circles ) , pure medium ( green solid circles ) , and small probes ( blue solid circles ) .
the medium molecules diffuse less when the probes are large .
( color online ) maximum value of the dynamic susceptibility @xmath49 for the medium molecules , versus temperature when small or large probes are included , compared with values from the pure liquid in the same conditions .
the function @xmath50 reaches its maximum for @xmath51 .
from top to bottom : large probes ( red solid circles ) , pure medium ( green solid circles ) , and small probes ( blue solid circles ) .
the dynamic susceptibility is normalized with its maximum value for the pure liquid at t=200k ( @xmath52 ) . + _ _ fig.6 .
( color online ) dynamic susceptibility @xmath50 for the medium molecules , versus time when small or large probes are included , compared with the dynamic susceptibility of the pure liquid in the same conditions . from top to bottom : large probes ( red continuous line ) , pure medium ( green dotted line ) , and small probes ( blue dashed line ) .
t=120 k. the dynamic susceptibility is normalized with its maximum value for the pure liquid at t=200k ( @xmath52 ) . + _ the 4-point dynamic susceptibility@xcite measures the autocorrelation of the fluctuations of the mobility and as such is an efficient measure of the dynamic heterogeneity .
we calculate the dynamic susceptibility @xmath53 from the equation@xcite : + @xmath54 with @xmath55 in these equations , @xmath56 denotes the volume of the simulation box , @xmath57 denotes the number of molecules in the box , and @xmath58 .
also , the symbol @xmath59 stands for a discrete mobility window function , @xmath60 , taking the values @xmath61 for @xmath62 and zero otherwise .
we use the value @xmath63@xmath2 in this work , which maximizes @xmath53 in our liquid at the density of the study . note that some caution must be taken if small values of @xmath64 are chosen in the calculation ( that may probe vibrational motions inside the cages instead of cage escaping motions ) while larger values of the parameter @xmath64 lead to qualitatively similar results than with the parameter we chose .
0.5 cm figures 5 and 6 show that the dynamic susceptibility is larger when we add large probes inside the medium , and smaller when we add small probes . as the dynamic susceptibility measures the dhs ,
these results show that the dhs increase or decrease depending on the defects we add to the medium .
these results strongly suggest that the defects associated with the small molecules destroy the dhs , while the large molecules defects facilitate the dhs .
when we include defects ( probes ) inside the medium , figure 5 shows that for both sort of probes the dhs modification from the pure medium values , increases when the temperature drops . adding large probes to the medium , results in an increase of the dhs while the dynamics slow down , following the same trend than the inverse of the diffusion coefficient in figure 4 .
( color online ) maximum value of the non gaussian parameter @xmath65 for the medium molecules , versus temperature when small or large probes are included , compared with values from the pure liquid in the same conditions . from top to bottom : large probes ( red solid circles ) , pure medium ( green solid circles ) , and small probes ( blue solid circles ) .
the non gaussian parameter is normalized with its maximum value for the pure liquid at t=200k ( @xmath66 ) . + _
( color online ) non gaussian parameter @xmath67 for the medium molecules , versus time when small or large probes are included , compared with the non gaussian parameter of the pure liquid in the same conditions .
from top to bottom : large probes ( red continuous line ) , pure medium ( green dotted line ) , and small probes ( blue dashed line ) .
t=120 k. the non gaussian parameter is normalized with its maximum value for the pure liquid at t=200k ( @xmath66 ) . + _
we display in figure 7 the non gaussian parameter ( ngp ) @xmath68 evolution with temperature , and in figure 8 its time variation at a temperature t=120k .
the non - gaussian parameter @xmath67 is defined as : @xmath69 the ngp @xmath67 measures the variations from the gaussian shape of the self van hove correlation function , that is predicted by brownian motion .
as one of the signature of the dynamic heterogeneity is the appearance of a tail in the van hove originating from fast cooperative motions , @xmath67 measures the dynamic heterogeneity .
we see in figures 7 and 8 an evolution of @xmath68 with temperature and with the probe that is qualitatively similar than the evolution we have observed for the dynamic susceptibility . at high temperature the probes have no effect on the dh and the circles merge in figures 7 and 5 for @xmath70 . then , as the temperature decreases the different probes lead to increasingly different dhs .
the small probes decrease the dh while the large probes increase them . we observe a smaller variation of the non gaussian parameter than of the dynamic susceptibility when decreasing the temperature in our liquid . note
that while the characteristic time for the susceptibility depends on the choice of the parameter @xmath64 , the characteristic time of the ngp @xmath46 is an important characteristic of the supercooled liquid at the temperature of study .
however we observe in both cases ( figures 6 and 8) that the maximum value of the ngp and of the susceptibility are shifted to larger times for large probes and to smaller times for small probes . _
( color online ) inverse of the diffusion coefficient @xmath71 for the medium molecules normalized with the diffusion coefficient @xmath72 of the pure liquid in the same conditions , versus the relative size @xmath73 of the probe . t=120k .
the red ( dark ) circle corresponds to the pur medium .
this point was obtained with high accuracy . +
_ _ fig.10 .
( color online ) maximum value of the non gaussian parameter @xmath65 for the medium molecules , normalized with the non gaussian parameter of the pure liquid @xmath66 at the same temperature t=120k , versus the relative size @xmath73 of the probe .
the red ( dark ) circle corresponds to the pur medium .
this point was obtained with high accuracy .
note that the temperature for the normalization ( t=120 k ) is here different from that of the normalization in figures 7 and 8 ( t=200k ) . + _
we have studied the effect of two particular probes ( a small probe and a large one ) diluted inside a supercooled medium .
we will now show that the effects observed are not qualitatively particular to the probes we chose previously .
for that purpose we will now vary the probe size and study the resulting effects on the dynamics and heterogeneity .
in the following , we use a probe that is similar to the medium molecule but with a size that has been multiplied by the factor @xmath73 ( i.e. @xmath74 and @xmath75 ) .
we show in figures 9 and 10 the evolution of the diffusion coefficient @xmath48 and of the non gaussian parameter @xmath68 with the relative size of the probe @xmath73 . the diffusion coefficient decreases strongly when the size of the probe is larger than the medium molecule ( @xmath76 increases in the figure ) .
larger probes result in smaller diffusion coefficients .
simultaneously we observe in figure 10 an increase of the non gaussian parameter for large probes . for probes smaller than the medium molecules ,
the diffusion is larger than for the pure medium , and roughly constant .
the non gaussian parameter is also roughly constant and smaller than the bulk value .
in this work , using molecular dynamics simulations we have investigated the effect of small packing defects on the dynamic heterogeneity and diffusion processes in a supercooled liquid .
our objective was to test the hypothesis of dynamic heterogeneities created by packing defects and the resulting modification of the liquid s dynamics .
a secondary purpose was to better understand how diluted molecules modify the viscosity of a liquid below @xmath77 .
for these purposes we have used a simple diatomic glass - former with a few ( 1@xmath78 ) probe molecules that were similar to the medium molecules but of a different size . due to the simplicity of the molecules we were able to access very long time scales ( of the order of the micro - second ) .
we found that the induced defects do not modify the dynamics at high temperature but below @xmath77 the modification increases rapidly when the temperature drops .
this result shows that the effects are linked to the physics of the glass - transition .
we found that the relative size @xmath73 between the diluted probe and the medium molecule strongly determines the dynamics when @xmath79 , but not when @xmath80 .
when @xmath80 the cooperativity decreases and the dynamics is accelerated , while when @xmath79 we observe the opposite effect . in summary , our results show that a few ( 1@xmath78 ) packing defects are able to strongly modify the dynamical heterogeneity in a supercooled liquid , in most cases increasing the heterogeneity and in some cases destroying them .
these results come in support of the hypothesis of dynamic heterogeneities created by packing fluctuations in supercooled liquids .
similarly , the packing defects also strongly affect the viscosity of the liquid , with larger defects leading to larger effects . | we use molecular dynamic simulations to investigate the relation between the presence of packing defects in a glass - former and the spontaneous cooperative motions called dynamic heterogeneity . for that purpose
we use a simple diatomic glass - former and add a small number of larger or smaller diatomic probes .
the diluted probes modify locally the packing , inducing structural defects in the liquid , while we find that the number of defects is small enough not to disturb the average structure .
we find that a small packing modification around a few molecules can deeply influence the dynamics of the whole liquid , when supercooled .
when we use small probe molecules , the dynamics accelerates and the dynamic heterogeneity decreases .
in contrast , for large probes the dynamics slows down and the dynamic heterogeneity increases .
the induced heterogeneities and transport coefficient modification increase when the temperature decreases and disappear around the onset temperature of the cage dynamics . |
fully compensated half - metallic ferrimagnets ( also known as half - metallic antiferromagnets ) @xcite have promising properties for spintronics . due to their internal spin compensation ,
the total moment is ( nearly ) zero at low temperature , yet the possibly strong local moments allow for high curie temperatures .
however , approaching the curie temperature one expects to find a non - zero total magnetic moment @xcite . in conjunction with possible appearance of half - metallicity
, such materials would be ideal electrode materials for spin transfer torque switching devices and other advanced applications @xcite . among the inverse heusler compounds ( space group @xmath1 , prototype hg@xmath0cuti ) @xcite , several compositions with 24 valence electrons
have been identified which could have these unusual properties , in particular cr@xmath0coga and cr@xmath0fege @xcite .
in contrast to regular heusler compounds , the inverse heusler compounds do not possess inversion symmetry , such that the cr atoms become nearest neighbors , which leads to an antiparallel coupling of their magnetic moments due to direct exchange interaction .
these compounds prefer the inverted heusler structure over the regular one and the local moments are predicted to be quite large . in consequence ,
high curie temperatures of 1520k and 748k are expected , respectively @xcite .
experiments indicate that cr@xmath0coga crystallizes in a cubic structure , which might be the inverse heusler structure @xcite . however , magnetization measurements reveal that the cr@xmath0coga compound is susceptible to atomic disorder , which generates a non - zero moment @xcite .
further , the curie temperature was observed to be only about 300k and the electrical resistivity was semimetallic @xcite .
both experimental findings indicate that severe structural disorder must be present and that the actual crystal structure of cr@xmath0coga can not simply be portrayed as the inverse heusler structure .
recent experiments on cr@xmath0fege indicate that this compound crystallizes in a more complex tetragonal structure @xcite .
these experimental results prompted us to perform a stability analysis of these compounds and their relatives of the type cr@xmath0yz ( y = co , fe ; z = al , ga , in , si , ge , sn ) with 24 valence electrons .
our analysis is based on formation energies computed within density functional theory ( dft ) @xcite , similar to recent analyses for a large number of heusler and inverse heusler compounds given by gilleen and dronskowski @xcite .
the dft calculations with the perdew - burke - ernzerhof ( pbe ) @xcite exchange - correlation functional were performed with the elk code @xcite , an implementation of the full - potential linearized augmented plane - wave method ( flapw ) .
the muffin - tin sphere radii were set to 2.0bohr for all elements and the augmented plane - wave expansion was taken to @xmath2bohr@xmath3 .
@xmath4-point meshes with @xmath5 points were used .
we obtained formation energies ( zero - temperature formation enthalpies ) with respect to the elemental solids from the total energies of the compounds and the total energies per atom of the constituent elements in their respective ground - state structures : @xmath6 the total energy differences were converged to within 10mev .
structural relaxations of the constituent elements were performed with the gpaw code @xcite .
decompositions into binary phases were studied with the help of the _ materials project _ database @xcite .
[ cols= " < , > , > , > , > , > , > , > " , ] for comparison we have calculated the formation energies of the co@xmath0fesi , co@xmath0fege , co@xmath0fesn series of full heusler compounds .
we find @xmath7 , @xmath8 , and @xmath9ev / f.u .
, respectively .
indeed , co@xmath0fesn is known to be thermodynamically unstable and to disintegrate into binary phases , but it can be synthesized in the heusler phase in a non - equilibrium process @xcite . its binary decomposition into cofe + cosn has a formation energy of @xmath10ev / f.u .
in contrast , co@xmath0crsi has a formation energy of @xmath11ev / f.u . , yet it is found to be unstable and to form disordered and binary phases @xcite .
the most favorable binary decomposition is co@xmath0si + cr , with a formation energy of -1.35ev / f.u .
searching for possible binary decompositions of the cr based compounds , we find that also cr@xmath0coal and cr@xmath0fesi are unstable .
cr@xmath0coal would most likely decompose into 2cr + coal ( @xmath12ev / f.u . ) , whereas 3cr@xmath0fesi would decompose into 2cr@xmath13si + fe@xmath13si ( @xmath14ev / f.u . ) .
the most favorable binary reactions for the six compounds are given in table [ tab2 ] .
for completeness , we study the magnetism as a function of the lattice constant of cr@xmath0fesi and cr@xmath0coal in fig .
[ fig1 ] , where the absolute moment denotes the sum of the absolute values of muffin - tin moments .
the energy zero is chosen as the sum of the elemental total energies , i.e. , the second term on the right - hand side of eq .
cr@xmath0fesi is nonmagnetic at its equilibrium lattice constant , but turns into a weak ferrimagnet at slightly increased volume .
cr@xmath0coal has a ferrimagnetic ground - state in the inverse heusler structure .
due to the similarly large magnetic moments as in cr@xmath0coga , a high curie temperature may be expected for this compound @xcite .
in summary , we have studied the energy of formation for six chromium based ternary materials with inverse heusler structure , which potentially could form compensated ferrimagnets .
we have shown that all of these compounds are thermodynamically unstable , having positive formation energy or because of energetically more favorable binary decompositions
. however , the inverse heusler structure may be metastable at finite temperature , eventually opening a path to the synthesis of the compounds we have studied .
we thank the developers of the elk and gpaw codes and of the _ materials project _ for their efforts .
we further thank gnter reiss for his support .
financial support by the deutsche forschungsgemeinschaft ( dfg ) is gratefully acknowledged .
20 h. van leuken , r. a. de groot , physical review letters 74 ( 1995 ) 1171 .
e. aiolu , physical review b 79 ( 2009 ) 100406 .
w. e. pickett , j. s. moodera , physics today 54 ( 2001 ) 39 .
x. hu , advanced materials 24 ( 2011 ) 294 . s. skaftouros , k. zdoan , e. aiolu , i. galanakis , physical review b 87 ( 2013 ) 024420 .
i. galanakis , e. aiolu , applied physics letters 99 ( 2011 ) 052509 .
t. graf , f. casper , j. winterlik , b. balke , g. h. fecher , c. felser , zeitschrift fr anorganische und allgemeine chemie 635 ( 2009 ) 976 . j. f. qian , l. feng , w. zhu , e. k. liu , x. d. tang , w. h. wang , g. h. wu , f. b. meng , h. y. liu , h. z. luo , acta physica sinica 60 ( 2011 ) 056402 .
m. hakimi , m. venkatesan , k. rode , k. ackland , j. m. d. coey , journal of applied physics , 12th joint mmm / intermag conference 2013 proceedings .
p. hohenberg , w. kohn , physical review 136 ( 1964 ) b865 . w. kohn , l. j. sham , physical review 140 ( 1965 ) a1133 .
m. gilleen , r. dronskowski , journal of computational chemistry 30 ( 2009 ) 1290 .
m. gilleen , r. dronskowski , journal of computational chemistry 31 ( 2010 ) 612 .
j. p. perdew , k. burke , m. ernzerhof , physical review letters 77 ( 1996 ) 3865 .
the elk fp - lapw code , http://elk.sourceforge.net , version 1.4.22 .
j. j. mortensen , l. b. hansen , k. w. jacobsen , physical review b 71 ( 2005 ) 035109 .
s. r. bahn , k. w. jacobsen , computing in science and engineering 4 ( 2002 ) 56 .
a. jain , g. hautier , c. moore , s. p. ong , c. fischer , t. mueller , k. persson , g. ceder , computational materials science 50 ( 2011 ) 2295 . s. p. ong , a. jain , g. hautier , m. kocher , s. cholia , d. gunter , d. bailey , d. skinner , k. persson , g. ceder , http://materialsproject.org/. m. a. tanaka , y. ishikawa , y. wada , s. hori , a. murata , s. horii , y. yamanishi , k. mibu , k. kondou , t. ono , s. kasai , journal of applied physics 111 ( 2012 ) 053902 .
m. aftab , g. hassnain jaffari , s. k. hasanain , s. ismat shah , journal of applied physics 110 ( 2011 ) 053903 . | chromium based inverse heusler compounds of the type cr@xmath0yz ( y = co , fe ; z = al , ga , in , si , ge , sn ) have been proposed as fully compensated half - metallic ferrimagnets .
such materials are of large interest for spintronics because they combine small magnetic moment with high spin polarization over a wide temperature range .
we assess their thermodynamic stability by their formation enthalpies obtained from density functional theory calculations .
all compounds under investigation are unstable .
cr@xmath0fesi and cr@xmath0coal are stable with respect to the elemental constituents , but decompose into binary phases .
cr@xmath0fege , cr@xmath0coga , cr@xmath0fesn and cr@xmath0coin are found to be unstable with respect to their elemental constituents .
we identify possible binary decompositions . |
the lhc collaborations have reported new results on the search for the standard model higgs , @xmath4 @xcite . by combining analyses ,
the atlas collaboration has excluded the sm higgs in the mass range @xmath5 and @xmath6 at 95%cl with an integrated luminosity up to @xmath7 .
similarly , the cms collaboration excluded a sm higgs in the range @xmath8 at 95%cl with integrated luminosity of @xmath9 .
furthermore , atlas reported an excess at 2.8@xmath10 in the @xmath11 decay channel , consistent with a higgs mass @xmath12 gev @xcite .
this signal is larger than expected from a sm higgs ( with signal strength @xmath13 ) and might favour new physics .
an excess in the @xmath0 channel was also reported by the cms experiment @xcite but at a slightly lower mass @xmath14 gev .
although atlas and cms statistics are not sufficient to claim a higgs discovery , it is interesting to assume that these are possible evidence for the higgs @xcite and investigate whether this is compatible with the predictions which arise in supersymmetric extensions of the standard model such as the minimal supersymmetric standard model ( mssm ) or next - to - mssm ( nmssm ) . to fit the data , the higgs candidate must have both a mass in the favoured range ( that is , [ 122,128 ] gev when taking into account the theoretical and experimental uncertainties @xcite ) and a signal strength that is consistent with the value of @xmath15 measured by atlas and cms . in the mssm , the higgs mass is required to be light @xmath16 @xcite .
while a higgs mass of 125 gev or larger is still possible it requires appreciable fine - tuning @xcite .
in addition , the question of whether the signal strength can be as large as suggested by the data must be answered . assuming that the higgs is produced by gluon fusion , the signal strength is defined by the ratio @xmath17 where @xmath18 can be either photons , @xmath19 or @xmath20 .
it is generally expected that , after applying other constraints on the model , @xmath21 is at most as large as unity @xcite .
indeed , the large mixing in the stop sector ( which is required to have a large enough higgs mass to fit the data ) means that the stop contribution to the higgs - gluon - gluon ( @xmath22 ) loop - induced vertex interferes destructively with the dominant top contribution .
this reduces the higgs production cross section and is not completely compensated by the increase in the higgs-@xmath23-@xmath23 coupling ( even though an exception was found in a corner of the parameter space where heavily mixed staus can lead to an increase in the @xmath24 partial decay width @xcite ) .
large susy corrections to the @xmath25 partial decay width may nevertheless reduce the higgs total width and hence lead to an increase of the @xmath24 branching ratio .
the presence of a light neutralino could also have a strong impact on the higgs signal since this would increase the higgs decay width into invisible particles @xcite and therefore decrease the signal in visible channels such as @xmath26 . in the nmssm ( a singlet extension of the mssm @xcite ) , the higgs mass receives additional corrections from the singlet sector . as a result
the fine - tuning problem is less severe and one more naturally obtains higgs masses of about 125 gev @xcite . in @xcite
, it was found that the signal strength could reach at most the sm value within the framework of the constrained nmssm . as in the mssm , the contribution of susy particles in the loop - induced @xmath22 and @xmath27 couplings lead to an overall suppression of the signal strength .
hence , it would be difficult to explain the central value @xmath28 reported by both collaborations .
values of @xmath29 larger than unity were nevertheless found in the nmssm when the higgs doublet was heavily mixed with the lightest higgs singlet @xcite . indeed , in this case ,
the singlet component of the higgs can cause a greater suppression of the partial decay width for @xmath30 than for @xmath31 , thus boosting the @xmath32 branching ratio .
( here @xmath33 are the two lightest scalar higgses of the nmssm . )
the results in @xcite therefore raise hopes to find higgs candidates in the nmssm which could fit recent lhc data
. however , there is no indication of how likely these sets of parameters are with respect to the rest of the parameter space and whether they fit both the mass and signal strength when constraints on neutralino dark matter are applied .
for example it may be that the lsp in these configurations leads to a too large relic density with respect to cosmological observations , thus excluding the model .
furthermore , direct detection limits from the xenon100 experiment @xcite or indirect detection limits from gamma - rays from dwarf spheroidal galaxies @xcite ( which are known to constrain the properties of the neutralino lsp , especially if it is light @xcite ) might have an impact on higgs properties . in this paper , we focus on the nmssm and address the question of the higgs mass and signal strength in the @xmath34 and @xmath35 modes when both dark matter and lhc constraints on the higgs sector and on supersymmetric particles are simultaneously taken into account .
we will assume friedman - robertson - walker ( frw ) cosmology , standard thermodynamics and the freeze - out mechanism to compute the relic density .
the paper is organised as follows . in section
ii , we recall the parameter space for the nmssm and the relevant constraints from colliders and astroparticle physics .
we recall the lhc searches useful for our analysis in section iii .
we determine the parameter space and give the characteristics of the higgs which can fit the latest lhc data when neutralinos are required to be lighter than 15 gev in section iv and when this condition is relaxed in section v. we conclude in section vi .
the nmssm is a simple extension of the mssm that contains an additional gauge singlet superfield .
the vev of this singlet induces an effective @xmath36 term which is naturally of the order of the electroweak scale , thus providing a solution to the naturalness problem @xcite .
the model contains three cp - even ( @xmath37 ) and two cp - odd ( @xmath38 ) higgs bosons : one more than the mssm in both cases , due to mixing of the singlet with the two doublets . among the interesting features of the nmssm
, there is a coupling @xmath39 in the superpotential which leads to a positive contribution to the mass of the sm - like higgs boson for small values of @xmath40 .
in addition , the presence of singlet / doublet mixing can lead to an @xmath41 higgs with a mass within the observed range and a light eigenstate , @xmath42 , compatible with all collider bounds despite the fact that @xmath43 is much smaller than @xmath2 120 gev .
this happens if the singlet component of @xmath42 is large enough to reduce its couplings to the sm fields @xcite .
when this singlet component is not large , @xmath42 becomes sm - like and @xmath41 can be much heavier than @xmath2120 gev . in the nmssm ,
it is therefore possible to either have @xmath42 or @xmath41 in the [ 122 - 128 ] gev mass range ( or even both ) , as favoured by the latest lhc data .
the nmssm also contains an additional neutralino , referred to as the singlino . due to its singlet nature
the singlino can be very light , hence providing a potential light dark matter candidate .
however many of these scenarios are excluded , as shown in @xcite , because they overproduce gamma rays ( or radio emission ) in the galaxy or appear to be in conflict with latest direct detection experiment constraints . in order to make sure that the higgs scenarios that we consider are all relevant , we base our present analysis on @xcite in which we explored the nmssm parameter space in light of particle physics and astroparticle physics constraints . in these studies , based on two different markov chain monte carlo ( mcmc ) analyses , we required that the lsp relic density does not exceed the wmap observed value ( but it can be much lower than this , hence calling for another type of particles to solve the dark matter problem ) and also took into account limits from b- physics , @xmath44 , as well as lep and tevatron limits on the higgs and susy particles .
these include in particular limits from the invisible width of the z as well as from neutralino production at lep .
lhc limits on the higgs sector computed with nmssmtools @xcite were also included as well as theoretical constraints on the model ( landau pole or unphysical global minimum ) .
additional constraints such as direct detection limits from xenon100 @xcite , gamma rays from dwarf spheroidal ( dsph ) galaxies probed by fermi - lat @xcite and the radio emission in the milky way ( mw ) and in galaxy clusters @xcite were superimposed on the parameter space selected by the mcmc . in the first analyses
@xcite , we imposed the condition that the neutralino be relatively light ( between 1 and 15 gev ) .
this was motivated by hints of a signal in direct detection experiments @xcite . however since these signals have not been established yet , we relaxed this condition on the mass in the second study @xcite . even though the second study encompass all lsp masses , a separate analysis with a prior on the neutralino mass is needed to ensure a complete coverage of the parameter space of this rather fine - tuned scenario and investigate thoroughly the special features associated with light neutralinos .
for this reason we continue to keep the two analyses separate in this work .
the model that we considered in both analyses has input parameters which are defined at the weak scale .
the free parameters are taken to be the gaugino masses @xmath45 , the higgs sector parameters @xmath46 , @xmath47 , the common soft masses for left- and right - handed sleptons @xmath48 , @xmath49 and the common soft masses for the squarks of the first and second generation ( @xmath50 ) and third generation ( @xmath51 ) .
at last , we take one non - zero trilinear coupling , @xmath52 . for more details , see @xcite . for the analysis with the @xmath53 gev requirement
, we consider only scenarios with one common soft mass for sleptons ( @xmath54 ) , one common mass for squarks ( @xmath55 ) , and the gaugino mass relation @xmath56 .
the latter is imposed to reduce the number of free parameters knowing that the gluino does not play an important role in dark matter observables for light neutralinos . for the analysis _ without _ the @xmath53 gev requirement ( which we explored in the later work @xcite ) we relaxed these three conditions , giving three further free parameters . details of the mcmc ( including the range of parameters for the scans and limits on observables ) were given in table i of @xcite for the light neutralino case and in @xcite for the general case .
was published @xcite .
however we have checked _ a posteriori _ that the bulk of the points fall below this limit , thus having little impact on our analysis , see @xcite in the light neutralino case . ]
physical quantities are computed by using micromegas2.4 @xcite that we interfaced with nmssmtools @xcite for the computation of the nmssm spectrum and particle physics constraints . for the case of the heavy neutralino we extended our previous analysis in order to explore more precisely the region outlined in @xcite where the light higgses have a large singlet component and are heavily mixed .
for this we started some mcmc chains in the region with @xmath57 and @xmath58 .
in these scans , computing the relic density provides an important constraint on the higgs sector .
the latter was required to satisfy @xmath59 with @xmath60 @xcite .
this sets a constraint on the neutralino pair annihilation cross section in the primordial universe . since neutralinos lighter than 15 gev
can either be binos ( as in the mssm ) or singlinos , the easiest way to ensure significant annihilations is through resonant exchange of a light scalar or pseudoscalar higgs @xcite . hence light @xmath61 or @xmath42 ( @xmath62 gev )
are preferred .
however , when the neutralino mass is large enough , annihilation mechanisms through z - exchange or light sleptons can also be efficient and a very light higgs singlet is no longer important .
for a higgs lighter than 140 gev as preferred by the lhc results , the main search channel is @xmath63 , while the channel @xmath64 also contributes ( v denoting either a w or a z ) .
therefore we compute @xmath65 , @xmath66 as well as @xmath67 , see section [ sec : intro ] , where the decays also include the virtual w / z .
note that @xmath68 is taken to be proportional to @xmath69 even though qcd corrections are different for the two processes .
one can reasonably assume that the effect of qcd corrections cancels out when taking the ratio of the nmssm to the sm value .
in what follows , we impose up - to - date lhc constraints on the higgs sector using higgsbounds 3.6.1beta @xcite susy * ] . ] . in particular , the recent results in the two - photon mode are taken into account as well as the limits from the search for @xmath70 which impact the heavy higgs doublet sector of supersymmetric models @xcite .
to find higgs candidates compatible with the latest lhc data , we impose ( on top of all the particle physics and astroparticle constraints set in @xcite ) lhc limits on sparticles . in particular , we take into account the exclusion limit coming from the atlas @xmath71 search for squarks and gluinos via jets and missing @xmath3 @xcite . for each susy point
, signal events were generated using @xcite .
experimental cuts of each search channel were then applied using @xcite . for the atlas jets and missing @xmath3 searches ,
these are included @xcite in the package
. in general these limits exclude the first and second generation squarks lighter than @xmath72 tev and gluinos lighter than @xmath73 tev .
however they rely on the fact that , in the constrained mssm , there are large branching ratios of the gluinos and squarks into jets and the neutralino lsp ( for the rh squarks this branching ratio is nearly 100@xmath74 ) . in the nmssm , this is not always the case : when the lsp is purely singlino , the squarks and gluinos can not decay to this lsp directly but must do via an intermediate particle , frequently the second - lightest neutralino . as noted in @xcite this reduces the acceptance into jets + missing @xmath3 search channels , as the extra step reduces the missing @xmath3 and may result in leptons .
( susy searches with leptons would have in fact greater sensitivity but they do not compensate for the loss of sensitivity in the 0-lepton search @xcite ) .
if the intermediate state decays into the lsp and a jet , there will also be greater alignment between the missing @xmath75 and one of the jets failing the angular separation trigger @xmath76 . in the analysis where we require sub-@xmath77gev neutralinos ,
whenever the coloured sparticles are light enough to be within reach of the lhc they are associated with a singlino - like @xmath78 . here
the jets + missing @xmath3 search is observed to be less sensitive for all of the aforementioned reasons and excludes very few of the points . ] . in the analysis with unconstrained neutralino mass , however , the singlet sector particles are generally much heavier , so that the lsp is not singlino - like . thus the usual @xmath79 and
@xmath80 decays take place and the familiar jets + missing @xmath3 exclusion is observed : @xmath81 tev , @xmath82 tev .
search , for points in the light neutralino mcmc scan .
red points have a bino - like lsp , blue points a singlino - like lsp .
the exclusion zone ( @xmath83 ) is shaded.,scaledwidth=47.0% ] finally , it is interesting to note that the relative acceptance into different search channels can help to distinguish a bino - like lsp from a singlino - like lsp : the latter produces a higher average number of jets in the cascade .
[ fig:23jets ] illustrates this : we plot the signal cross sections in the two- and three - jets channels of the jets + missing @xmath3 search , showing both bino- and singlino - like lsp points from the analysis with @xmath84 gev .
the graph also shows the low level of exclusion singlino - like lsp points for the reasons discussed above , bino - like lsp points simply because the squarks are heavier and out of reach .
for a model to be compatible with the data , it is required that the sm - like higgs be in the observed mass range ( say [ 122 - 128 ] gev ) and the signal strength be consistent with the data . as a criteria for @xmath65 ,
we choose a 2@xmath10 error bar around the central value determined by atlas , thus @xmath85 .
this is also compatible with cms results and sm expectations ( where @xmath86 ) . in this subsection , we present the results for the higgs sector when the neutralino mass satisfies @xmath87 gev and one applies only an upper limit on the relic density .
imposing the condition of a light neutralino leads the mcmc to select models containing a light higgs ( @xmath88 gev ) which is mostly a singlet . as mentioned earlier
, this could be either a scalar @xmath42 or a pseudoscalar @xmath61 . as illustrated in fig .
[ fig : mh12 ] , we found that @xmath42 is typically much below the electroweak scale ( thus with a large singlet component ) when @xmath41 is sm - like . when @xmath41 is much heavier than 125 gev , @xmath42 is sm - like ( reaching at most up to @xmath2122 gev ) and the pseudoscalar @xmath61 is basically a light singlet , see fig .
[ fig : mh12 ] .
it is nevertheless possible for both scalars to be heavily mixed and have a mass around 100 - 130 gev ( in this case @xmath61 has to be light ) . in fig .
[ fig : mh12 ] blue and black points show the scenarios with at least one of the scalars within the range preferred by atlas and cms .
as one can see , this is generally @xmath41 since @xmath89 barely exceeds 122 gev . and pseudoscalar @xmath61 .
red points are ruled out either by higgsbounds constraints or the atlas @xmath90 jets and missing @xmath3 susy search .
green points have no higgs with a mass in @xmath91 gev , blue points have a higgs ( @xmath42 and/or @xmath41 ) within this mass range , and black points have such a higgs with @xmath92.,title="fig:",scaledwidth=47.0% ] and pseudoscalar @xmath61 .
red points are ruled out either by higgsbounds constraints or the atlas @xmath90 jets and missing @xmath3 susy search .
green points have no higgs with a mass in @xmath91 gev , blue points have a higgs ( @xmath42 and/or @xmath41 ) within this mass range , and black points have such a higgs with @xmath92.,title="fig:",scaledwidth=47.0% ] the predictions for @xmath65 as a function of the @xmath41 mass are displayed in fig . [
fig : gggg_nmssm ] ( bottom panel ) .
we only display the region where this channel is relevant , that is when @xmath93 gev .
as one can see , all the configurations that were selected by the mcmc ( and which were compatible with the constraints that we imposed for the scans ) have @xmath94 .
an explanation is that , although @xmath41 couplings are usually sm - like , large suppressions in @xmath65 are possible because the width of the higgs is enhanced by many new non - standard decay channels @xcite . in particular
, @xmath41 can decay into two neutralinos , two light scalar higgses ( @xmath42 ) or two light pseudoscalar higgses ( @xmath61 ) which reduces significantly the branching ratio into two - photons . in fig .
[ fig : gggg_nmssm ] , all the points which do not satisfy either the newest higgsbounds limits or the susy searches in jets plus missing @xmath3 are colored in red . as mentioned above very few of these points are excluded by susy searches .
the points which fall within the higgs observed mass range are highlighted in blue .
scenarios where the strength of the signal in @xmath34 is also compatible with the 2@xmath10 range reported by atlas ( @xmath95 ) are represented by black squares .
we have also computed @xmath65 for @xmath42 and found that it is usually much below unity because @xmath42 has a large singlet component . only a few points have @xmath96 and they correspond to either a sm - like @xmath42 with a mass near 122 gev or to a very light singlet . as a function of the mass of @xmath42 ( top panel ) and of @xmath41 ( the more usual candidate ; bottom panel ) in the light neutralino lsp model .
same colour code as fig .
[ fig : mh12 ] .
, title="fig:",scaledwidth=47.0% ] as a function of the mass of @xmath42 ( top panel ) and of @xmath41 ( the more usual candidate ; bottom panel ) in the light neutralino lsp model .
same colour code as fig .
[ fig : mh12 ] .
, title="fig:",scaledwidth=47.0% ] let us now examine the predictions for @xmath97 ( with @xmath98 or @xmath99 ) as a function of the @xmath41 mass .
again the maximum value is unity .
@xmath97 can be much suppressed especially because of a large branching ratio into invisibles , hence a correlation between the large suppression in the @xmath100 channel and in the @xmath34 channel .
the results for @xmath97 are very similar to those for @xmath65 shown in fig .
[ fig : gggg_nmssm ] .
it is interesting to note that for all the scenarios where the @xmath41 mass is above 130 gev ( i.e. where the ww / zz modes are the dominant decay channels ) , the signal strength @xmath97 is suppressed by at least a factor 3 and even often suppressed by one order of magnitude .
hence , a non sm - like @xmath41 higgs with a mass of about 140 gev ( corresponding to a scenario similar to the mssm ) is still allowed .
interestingly enough , these nmssm scenarios could not be distinguished from the mssm based on the two - photon channel ( and/or the ww channel ) since both have a maximum signal strength of about unity .
however a search for the singlet higgs would give a distinctive nmssm signature . as we have mentioned , these light neutralino models must also have a light singlet higgs and thus non - standard decays occur .
the effect of these decays on the signal strength is shown in fig .
[ fig : signalstrength_v_bsmdecays ] and the distribution of the decays in fig .
[ fig : distrib ] for the points compatible with the observed excess .
clearly , too large branching ratios into non standard modes ( such as @xmath101 ) would render the two - photon mode invisible or suppressed with respect to the sm prediction @xcite .
in fact , in order for the signal strength @xmath102 to be compatible with @xmath103 , the branching ratio @xmath104 must be lower than @xmath105 . competing with new bsm decays . here only points with a higgs mass between 122 - 128 gev and with @xmath95 are included.,scaledwidth=47.0% ] ( where @xmath106 is negligible ) .
histograms have unit area.,title="fig:",scaledwidth=25.0% ] ( where @xmath106 is negligible ) .
histograms have unit area.,title="fig:",scaledwidth=25.0% ] ( where @xmath106 is negligible ) .
histograms have unit area.,title="fig:",scaledwidth=25.0% ] ( where @xmath106 is negligible ) .
histograms have unit area.,title="fig:",scaledwidth=25.0% ] the existence of decay modes such as @xmath107 or @xmath108 , with the singlet higgs further decaying into standard model particles remains nevertheless interesting because such modes give a distinctive signature which could be searched for at lhc and would constitute evidence for new physics if they are found .
extraction of this signal from background via jet substructure techniques has been studied for @xmath109 in @xcite , @xmath110 in @xcite , and @xmath111 ( less relevant for susy due to @xmath40 suppression ) in @xcite .
these decays with an intermediate scalar ( @xmath42 ) instead of an intermediate pseudoscalar ( @xmath61 ) give the same signal .
a reduced two - photon signal could also be due to a sizable branching ratio of the higgs into @xmath112 and @xmath113 .
indeed this generally dominates over the decay to lighter higgses in this sample .
@xmath114 occurs mainly when the lsp is singlino : the bino / higgsino / singlino nlsp can be significantly lighter than 100 gev thus allowing sufficient phase space for the decay .
this mode is also possible for a bino lsp and a higgsino nlsp but the phase space is quite limited .
note that the nlsp @xmath115 can also have decay modes into light higgses which further decay in fermion pairs .
another distinctive feature of the nmssm higgs sector would be the direct search for @xmath42 .
we have seen that the @xmath116 channel is suppressed .
this is also true for other channels ; indeed the couplings of @xmath42 to sm particles is suppressed by the singlet component . for the @xmath117 production mode , the suppression can be in part compensated by a @xmath40 enhancement .
however in our scans we found that the @xmath118 coupling could reach at most its sm value .
thus the production of @xmath42 in association with b - quarks followed by the decay of the higgs into tau pairs does not benefit from an enhancement over the sm expectations .
if we now select only the points which predict a relic density at the wmap observed value ( more precisely we impose @xmath119 where @xmath120 is @xmath121 from the central measured value of wmap ) , we find strong constraints on the parameter space as light higgses in the s - channel or in the final states may reduce the relic density too much .
however , the results for the higgs sector which are compatible with the latest lhc data are similar to the ones that were discussed in the previous subsection . for completeness
we show the expectations for @xmath65 for these points in fig .
[ fig : gggg_wmap ] . as a function of @xmath122 in the light neutralino lsp model
when the relic density is compatible with wmap .
same colour code as fig .
[ fig : mh12 ] .
, scaledwidth=47.0% ]
we repeated the analysis for the mcmc scan in which there is no @xmath123 gev requirement .
we first consider the impact of higgs searches at lhc .
the effect of also imposing the relic density condition and susy searches will be discussed in section [ sec : heavy_susy ] .
as opposed to the previous case with a light neutralino we find that in general the higgs in the mass region preferred by the lhc is a sm - like @xmath42 . indeed , without very light neutralinos , a very light singlet sector is not needed for resonant annihilations .
thus the associated values for @xmath65 are naturally of order unity ( see the black squares in fig . [
fig : gggg_all ] , top panel ) .
nevertheless we found cases where @xmath124 , when invisible decay modes ( such as @xmath125 or @xmath126 ) are kinematically accessible .
we also found points with @xmath127 ; these were mostly associated with a higgs mass below the lhc preferred range .
indeed , the light higgs has a large singlet component , which leads to suppressed couplings to sm particles , in particular to @xmath117 .
the suppression of the partial decay width for @xmath128 , and thus of the total width , leads to a larger branching ratio in the @xmath129 @xcite . for @xmath130 gev ,
the increase in @xmath65 is generally modest ( below 20% ) except for a few points where both higgses are heavily mixed and have a large singlet component .
an enhancement with respect to the sm expectations is found for large values of @xmath131 and small values of @xmath36 ( below @xmath132 gev ) . under these conditions
the mixing between the singlet and doublet higgs is large and the singlet higgs is light .
recall that the mass of the singlet higgs depends on @xmath36 with @xmath133 @xcite .
note that in fig .
[ fig : gggg_all ] top panel , we found scenarios where @xmath41 is in the observed range while @xmath42 is lighter and has @xmath134 .
this means that even though @xmath42 has evaded present constraints from lhc , these points offer good prospects for discovery of a second higgs scalar in the next run of the lhc .
such a signal would allow to distinguish the nmssm higgs sector from the mssm one .
@xmath41 masses extend over a wide range ( all the way to several tev s ) and include some points in the mass region preferred by the lhc .
the values of @xmath65 for @xmath41 are displayed in fig .
[ fig : gggg_all ] ( bottom panel ) for the range of masses where the two - photon search mode is relevant .
we found that the signal strength reaches values as high as @xmath135 .
this enhancement with respect to the sm expectations is found when @xmath41 has some singlet component and a suppressed partial width to @xmath117 .
( @xmath42 , conversely , has an enhanced @xmath117 partial width and a reduced signal strength @xmath65 . ) as mentioned above , when @xmath136 for @xmath41 , the signal strength for the lighter higgs can be enhanced .
we have also found points where @xmath137 for both @xmath42 and @xmath41 .
this can be due to the presence of invisible modes or to the presence of a singlet component for which the suppression in the higgs coupling to @xmath138 compensates the increase of the @xmath34 branching ratio . as a function of the mass of @xmath42 ( top panel ) and of @xmath41 ( bottom panel ) in the arbitrary neutralino lsp model .
in the top panel most points are on top of one another at high mass and high @xmath65 .
same colour code as fig .
[ fig : mh12 ] except red points are denoting only exclusion by higgsbounds.,title="fig:",scaledwidth=47.0% ] as a function of the mass of @xmath42 ( top panel ) and of @xmath41 ( bottom panel ) in the arbitrary neutralino lsp model . in the top panel
most points are on top of one another at high mass and high @xmath65 .
same colour code as fig .
[ fig : mh12 ] except red points are denoting only exclusion by higgsbounds.,title="fig:",scaledwidth=47.0% ] finally we analyse the impact of requiring the relic density condition . imposing a strict lower limit on the relic density strongly constrains the scenarios where the singlet component of the light higgs leads to @xmath127 . as mentioned above , these points were found for @xmath139 gev . hence the lsp has a non - negligible higgsino component , usually leading to a relic density below the preferred wmap range . after imposing the strict lower bound on the relic density
, we found that @xmath42 was in general within the lhc preferred range , while @xmath41 was heavy , its mass extending above the tev scale . @xmath65 for @xmath42
is displayed in fig .
[ fig : gggg_heavy ] : the relic density constraint has removed most of the points where the signal strength was enhanced ( the maximum value is now 1.06 ) .
we also considered the constraint from the jets and missing @xmath3 susy search for those points with a higgs of mass @xmath140 .
exclusion is observed in this mass range for any signal strength , showing as expected that there is no direct correlation between the first and second generation squarks and the higgs sector .
finally we have checked that these points were compatible with the latest upper limit on @xmath141 from lhcb @xcite . as a function of the mass of @xmath42 ( the more usual candidate ) for points with a relic density within the wmap range . a handful of points
are also seen at masses lower than those shown , with reduced signal strength .
same colour code as fig .
[ fig : mh12].,scaledwidth=47.0% ]
in this paper , we have investigated the higgs signal strength expected in scans of the nmssm scenarios in which we have applied particle physics as well as astroparticle physics constraints .
in particular we have imposed that the neutralino relic density does not exceed the wmap observed value nor the limit imposed by direct detection experiments and does not overproduce gamma rays and radio waves in the galaxy .
we also took into account limits from b- physics , @xmath142 , as well as lep , tevatron and lhc limits on the higgs and susy particles .
we found many scenarios where either @xmath42 or @xmath41 had a mass in the range [ 122 - 128 ] gev and a signal strength compatible with the standard model .
we also found scenarios where the signal strength in the two - photon mode was as large as the excess reported by atlas and cms .
however for most of these points the neutralino would form only a fraction of the observed dark matter . if a higgs is confirmed at the lhc with a signal strength in excess of the sm expectations , it would be a clear indication of bsm physics , while if the signal strength is compatible with sm expectations , it might be necessary to search for light pseudoscalar or scalar higgs in order to determine whether one has discovered the sm higgs or bsm physics .
when insisting on a light neutralino , we found that the most promising configurations favour a sm - like @xmath41 rather than a @xmath42 sm - like higgs and therefore predict the existence of a light higgs dominantly singlet .
the possibility of observing a second light higgs provides a distinct signature of the nmssm higgs sector .
this could be done directly , e.g. via the diphoton search for an sm - like higgs with higher luminosities ; or indirectly via the decay of the heavier higgs , e.g. @xmath143 .
furthermore we note that the traditional jets + missing @xmath3 signature of squarks and gluinos in the mssm can be modified in a very interesting way in the nmssm , when the lsp is singlino - like and the second - lightest neutralino @xmath115 can decay to a higgs .
the @xmath115 is expected to be boosted ( generically being much lighter than the squarks and gluinos ) , and the lighter higgs which it decays to ( @xmath42 or @xmath61 ) is expected to further decay into @xmath117
. would be dominant for @xmath144 this extra step in this cascade compared to the mssm @xmath145 adds jets and halves the missing @xmath3 , dimming discovery prospects .
however a distinctive topology results : the boosted nature of the intermediate @xmath115 means the missing @xmath75 vector will point in between the two doubly - tagged @xmath146-jets . we illustrate this scenario in fig .
[ fig : fatjets ] for disquark production .
gluinos tend to decay to a neutralino plus two jets rather than one , as is well known ; otherwise the signal remains the same as for squarks .
calculation of the standard model background and estimation of the systematic error in measuring such a signal are clearly vital , but beyond the scope of this work
. the nmssm higgs sector can however be very similar to the mssm , or even the sm , with only a light sm - like scalar and much heavier scalars and pseudoscalars . in this case
searches for the superpartners , and in particular peculiar signatures associated with the singlino lsp ( e.g. bino @xmath147 singlino + soft @xmath148 @xcite ) , offer the only possibility to identify the nmssm .
signal in the nmssm with a singlino - like lsp , when a decay @xmath149 is kinematically accessible .
because the @xmath150 are boosted , the missing @xmath75 vector will point in between the two @xmath146-jets.,scaledwidth=47.0% ]
gb and cw thank the lpsc for its hospitality . the authors would like to thank alexander pukhov , ulrich ellwanger , michael spannowsky and david grellscheid for useful discussions ; also tim stefaniak and oscar stl for communications concerning higgsbounds .
this work was partly supported by the stfc and the cnrs - pics grant `` propagation of low energy positrons '' held by cb .
jd acknowledges the cmira 2011 explora doc program of the french region rhne - alpes . | recent lhc data suggest an excess in the higgs decay channels into @xmath0 and @xmath1 at @xmath2125 gev .
the current excess in the diphoton channel is twice that expected from a standard model higgs ; whilst this may well change with more statistics , it is interesting to consider the implications should the result persist . here
, we assess whether the nmssm with a neutralino dark matter candidate could explain this excess when astrophysical constraints ( e.g. no overproduction of gamma rays and radio emission in the galaxy , no anomalous excess in the dark matter direct detection experiments and no dark matter overabundance ) are imposed on the neutralino .
this enables us to disregard unphysical regions of the parameter space even though the higgs signal is compatible with the observed excess .
the result of our analysis is that there are configurations of the parameter space which can explain the signal strength reported by the atlas and cms collaborations for a higgs mass within the required range .
should the observed signal strength finally be compatible with standard model expectations , it would be difficult to distinguish between the discovery of standard model higgs and a sm - like higgs from the nmssm , unless one performs dedicated searches of very light higgs bosons and possibly investigate peculiar signatures of supersymmetric particles .
we also propose a new jets + missing @xmath3 signal for the case where the lsp is a singlino - like neutralino . |
web service ( ws ) is a new distributed component which emerged about ten years ago , which uses wsdl@xcite as its interface description language , soap@xcite as its communication protocol and uddi@xcite as its directory service . because ws uses the web as its provision platform , it is suitable to be used to develop cross - organizational business integrations .
cross - organizational business processes are usual forms in e - commerce that orchestrate some business activities into a workflow .
ws orchestration ( wso ) provides a solution for such business process based on ws technologies , hereby representing a business process where business activities are modeled as component wses . from a ws viewpoint
, wso provides a workflow - like pattern to orchestrate existing wses to create a new composite ws , and embodies the added values of ws .
in particular , we use the term wso , rather than another term
ws composition , because there are also other ws composition patterns , such as ws choreography ( wsc ) @xcite .
however , about wsc and the relationship of wso and wsc@xcite , we do not explain more , because it is not the focus of this paper . in this paper , we focus on wso , exactly , the qos - aware wso engine ( runtime of wso ) and its formal model .
a qos - aware wso enables the customers to be satisfied with not only their functional requirements , but also their qos requirements , such as performance requirements , reliability requirements , security requirements , etc . a single execution of a wso is called a wso instance .
a qos - aware wso engine provides runtime supports for wsos with assurance of qos implementations .
these runtime supports include lifetime operation on a wso instance , queue processing for request from the customers and incoming message delivery to a wso instance .
ws and wso are with a continuously changing and evolving environment .
the customers , the requirements of the customers , and the component wses are all changing dynamically . to assure safe adaptation to dynamically changing and evolving requirements , it is important to have a rigorous semantic model of the system : the component wses , the wso engine that provides wso instance management and invocation of the component wses , the customer accesses , and the interactions among these elements .
using such a model , designs can be analyzed to clarify assumptions that must be met for correct operation .
we design a typical qos - aware wso engine , called qos - wsoe in this paper .
an architecture of qos - wsoe is given , and more importantly , a formal model of qos - wsoe is established based on actor systems theory@xcite@xcite@xcite . in the formal model ,
we introduce the notion of * qos - aware wso service * , * qos - aware wso system * , and * qos - aware wso behavior*. and we draw conclusions that : ( 1 ) if a qos - aware wso engine is a * qos - aware wso system * , then it provides * qos - aware wso service * for customers ; ( 2 ) if a qos - aware engine has * qos - aware wso behavior * , then it is a * qos - aware wso system * and further provides * qos - aware wso service*. this paper is organized as follows . in section 2 ,
we introduce the related works .
the actor computational model and the three layered pyramidal architecture are introduced in section 3 .
we illustrate a ws composition example called buyingbooks in section 4 .
we design qos - wsoe in section 5 . and in section 6 , formal model of qos - wsoe
is established .
finally , we conclude our works and point out future works .
the main efforts on wso of the industry are trying to establish a uniform wso description language specification , such as the early wsfl@xcite , xlang@xcite , and lately converged ws - bpel@xcite .
such wso description languages based on different mathematical models have constructs to model invocation of wses , manipulate information transferred between wses , control execution flows of these activities and inner transaction processing mechanisms .
the wso description language can be used to define various wsos under different requirements and acts as a so - called meta language .
wsos described by such meta languages actually are pure texts and must be enabled by the meta language interpreter called wso engine , such as the famous open source activebpel@xcite , respect tuple centres based ws - bpel engine@xcite , a multi - agent system based ws - bpel engine@xcite and event - driven architecture based ws - bpel engine@xcite . in industry , there are many research works to give the meta languages correctness verifications based on different theoretical tools@xcite .
@xcite formalizes ws - bpel@xcite with petri - net .
@xcite uses process algebra to give ws - bpel a theoretical foundation .
@xcite establishes a calculus to verify correctness of ws - bpel .
semantics and verifications of ws - bpel are researched in @xcite .
@xcite uses a kind of formal specification to orchestrate wses . unlike these formalizations
, our formal model focuses on the correctness of the wso engine , that is , correctness of a runtime system of wso , but not correctness of the meta languages .
qos - aware wso engine can process not only the functional requirements modeled by the wso description language , but also the qos requirements of the customers .
for example , a wso must complete within three hours and the cost running some wso must be under twenty dollars .
a qos - aware wso translates the qos requirements of the customers into qos requirements of component wses .
such a translation is accomplished by the implementation of the so - called qos - aware service selection algorithm ( qssa ) .
there are many kind of such service selection algorithms , such as @xcite , which uses the so - called local service selection approach and the global allocation approach based on integer programming , and another one in @xcite , which models the service selection problem ( ssp ) in two ways : one defines ssp as a multi - dimensional multi - choice 0 - 1 knapsack problem ( mmkp ) based on combinatorial model and the other captures ssp as a multi - constraint optimal path problem ( mcop ) based on graph model , and gives heuristic algorithms of the two ways .
we do not research ssp itself , but use the implementation of a qssa above as a component of our wso engine qos - wsoe .
note that qos - aware wso implies that the wses involved in this wso also must be qos - aware . in this paper , we assume that the wses are all qos - aware .
about qos of ws , the readers please refer to @xcite and @xcite .
@xcite also involves implementations of qos - aware wses .
actor@xcite@xcite is a basic concurrent computing model and can be used in reasoning about open distributed systems@xcite . in @xcite and @xcite , actors are used to reason about middleware of resource management in distributed computing and even qos - aware middleware .
our works follow the works above , especially @xcite .
that is , in this paper , we adopt the way for formalization of open distributed systems in @xcite which uses a pyramidal refinement to capture the concepts of customer requirements , system requirements , system behaviors and their relationships .
our works focus on qos - aware wso engine , a software system different to the multimedia resource management middleware in @xcite with respect to different qos aspects and qos management , different system functions and architecture , different system components and behaviors .
that is , the formal model in this paper is a different one from that in @xcite .
an actor@xcite@xcite is a basic concurrent computation unit which encapsulates a set of local states , a control thread and a set of local computations .
it has a unique mail address and maintains a mail box to receive messages sent by other actors . through processing the messages stored in the main box sequentially
, an actor computes locally and blocks when its mail box is empty . during processing a message from its mail box ,
an actor may perform three candidate actions : ( 1)(*send * ) sending messages asynchronously to other actors ; ( 2)(*create * ) creating new actors with new behaviors ; ( 3)(*ready * ) becoming ready again to process the next message from the mail box or block if the mail box is empty .
note that synchronization@xcite can also be achieved among actors .
also there are many works to abstract at a high - level from aspects of distributed computing , such as policy management@xcite , interaction policies@xcite , resource management@xcite , communication and coordination of agents@xcite , worldwide computing@xcite , etc .
the actor computational model can be used to reason about the behavior of a computer system , especially the behavior of a distributed system@xcite .
@xcite uses actors to reason about the behavior of qos - aware distributed middleware and presents a so - called pyramidal structure to capture the concepts of customer requirements , system requirements , system behavior and their relationships .
our works are greatly inspired by this pyramidal structure and are with the similar goal but different system .
that is , our qos - aware wso engine called qos - wsoe has similar requirements , but different behavior , with the qos - aware distributed middleware in @xcite , just because they are different systems in nature .
in this section , we give a so - called buyingbooks example for the scenario of cross - organizational business process integration and use a so - called bookstore wso to illustrate some related concepts , such as wso , activity , etc .
and we use the bookstore wso to explain the formal model we established in the following .
a further example is buyingbooks as fig.[fig.buyingbooksexample ] shows .
we use this buyingbooks example throughout this paper to illustrate concepts and mechanisms in ws composition . in fig.[fig.buyingbooksexample ] , there are four organizations : buyeragent , bookstore , railwaycorp , and airlinecorp . and
each organization has one business process .
exactly , there are two business processes , the business processes in railwaycorp and airlinecorp are simplified as just wses for simpleness without loss of generality .
we introduce the business process of bookstore as follows , and the process of buyeragent can be understood as contrasts . 1 . the bookstore receives request of list of books from the buyer through buyeragent .
2 . it sends the list of books to the buyer via buyeragent .
3 . it receives the selected book list by the buyer via buyeragent .
4 . it calculates the price of the selected books .
it sends the price of the selected books to the buyer via buyeragent .
it gets pays for the books from the buyer via buyeragent .
if the pays are greater than 100 $ , then the bookstore calls the shipment service of airlinecorp for the shipment of books . 8 .
otherwise , the bookstore calls the shipment service of railwaycorp for the shipment of book .
then the process is ended .
each business process is implemented by a wso , for example , the bookstore wso and buyeragent wso implement bookstore process and buyeragent process respectively .
each wso invokes external wses through its activities directly . and
each wso is published as a ws to receive the incoming messages .
the bookstore wso described by ws - bpel is given as follows .
@xmath0process name=``bookstore '' targetnamespace="http://example.wscs.com /2011/ws - bp / bookstore" ... @xmath1 @xmath0partnerlinks@xmath1 @xmath0partnerlink name=``bsandba '' ... /@xmath1 @xmath0partnerlink name=``bsandrc '' ... /@xmath1 @xmath0partnerlink name=``bsandac '' ... /@xmath1 @xmath0/partnerlinks@xmath1 @xmath0variables@xmath1 @xmath0variable name=``requestlistofbooks '' messagetype=``lns : requestlistofbooks''/@xmath1 @xmath0variable name=``requestlistofbooksresponse '' messagetype=``lns : requestlistofbooksresponse''/@xmath1 @xmath0variable name=``listofbooks '' messagetype=``lns : listofbooks''/@xmath1 @xmath0variable name=``listofbooksresponse '' messagetype=``lns : listofbooksresponse''/@xmath1 @xmath0variable name=``selectlistofbooks '' messagetype=``lns : selectlistofbooks''/@xmath1 @xmath0variable name=``selectlistofbooksresponse '' messagetype=``lns : selectlistofbooksresponse''/@xmath1 @xmath0variable name=``price '' messagetype=``lns : price''/@xmath1 @xmath0variable name=``priceresponse '' messagetype=``lns : priceresponse''/@xmath1 @xmath0variable name=``pays '' messagetype=``lns : pays''/@xmath1 @xmath0variable name=``paysresponse '' messagetype=``lns : paysresponse''/@xmath1 @xmath0variable name=``shipmentbytrain '' messagetype=``lns : shipmentbytrain''/@xmath1 @xmath0variable name=``shipmentbytrainresponse '' messagetype=``lns : shipmentbytrainresponse''/@xmath1 @xmath0variable name=``shipmentbyair '' messagetype=``lns : shipmentbyair''/@xmath1 @xmath0variable name=``shipmentbyairresponse '' messagetype=``lns : shipmentbyairresponse''/@xmath1 @xmath0/variables@xmath1 @xmath0sequence@xmath1 @xmath0receive partnerlink=``bsandba '' porttype="lns : bookstore4buyeragent- interface " operation=``oprequestlistofbooks '' variable=``requestlistofbooks '' createinstance=``yes''@xmath1 @xmath0/receive@xmath1 @xmath0invoke partnerlink=``bsandba '' porttype="bns : buyagent4bookstore- interface " operation=``opreceivelistofbooks '' inputvariable=``listofbooks '' outputvariable=``listofbooksresponse''@xmath1 @xmath0/invoke@xmath1 @xmath0receive partnerlink=``bsandba '' porttype="lns : bookstore4buyeragent- interface " operation=``opselectlistofbooks '' variable=``selectlistofbooks''@xmath1 @xmath0/receive@xmath1 @xmath0reply partnerlink=``bsandba '' porttype="lns : bookstore4buyeragent- interface " operation=``opselectlistofbooks '' variable=``selectlistofbooksresponse''@xmath1 @xmath0/reply@xmath1 @xmath0!inner activity : calculate the price of selected books@xmath1 @xmath0invoke partnerlink=``bsandba '' porttype="bns : buyagent4bookstore- interface " operation=``opreceiveprice '' inputvariable=``price '' outputvariable=``priceresponse''@xmath1 @xmath0receive partnerlink=``bsandba '' porttype="lns : bookstore4buyeragent- interface " operation=``oppays '' variable=``pays''@xmath1 @xmath0/receive@xmath1 @xmath0reply partnerlink=``bsandba '' porttype="lns : bookstore4buyeragent- interface " operation=``oppays '' variable=``paysresponse''@xmath1 @xmath0if@xmath1@xmath0condition@xmath1 getvariable(price ) @xmath0 100 @xmath0/condition@xmath1 @xmath0invoke partnerlink=``bsandac '' porttype="ans : airlinecorp4bookstore- interface " operation=``opshipmentbyair '' inputvariable=``shipmentbyair '' outputvariable=``shipmentbyairresponse''@xmath1 @xmath0else@xmath1 @xmath0invoke partnerlink=``bsandrc '' porttype="rns : railwaycorp4bookstore- interface " operation=``opshipmentbytrain '' inputvariable=``shipmentbytrain '' outputvariable="shipmentbytrain- response"@xmath1 @xmath0/else@xmath1@xmath0/if@xmath1 @xmath0/sequence@xmath1 @xmath0/process@xmath1 there are several receive - reply activity pairs and several invoke activities in the bookstore wso .
the qos requirements are not included in the ws - bpel description , because these need an extension of ws - bpel and are out of the scope of this paper . in the request message from the buyeragent wso , the qos requirements , such as the whole execution time threshold and the additional charges , can also be attached , not only the functional parameters .
another related specification is the wsdl description of the interface ws for buyingbooks wso .
because we focus on ws composition , this wsdl specification is omitted .
in this section , we firstly analyze the requirements of a wso engine .
and then we discuss problems about qos management of ws and define the qos aspects used in this paper .
finally , we give the architecture of qos - wsoe and discuss the state transition of a wso instance .
as the introduction above says , a wso description language , such as ws - bpel , has : * basic constructs called atomic activities to model invocation to an external ws , receiving invocation from an external ws and reply to that ws , and other inner basic functions ; * information and variables exchanged between wses ; * control flows called structural activities to orchestrate activities ; * other inner transaction processing mechanisms , such as exception definitions and throwing mechanisms , event definitions and response mechanisms .
therefore , a wso described by ws - bpel is a program with wses as its basic function units and must be enabled by a wso engine .
an execution of a wso is called an instance of that wso .
the wso engine can create a new wso instance according to information included in a request of a customer via the interface ws ( note that a wso is encapsulated as a ws also . ) of the wso .
once a wso is created , it has a thread of control to execute independently according to its definition described by a kind of description language , such as ws - bpel . during its execution
, it may create activities to interact with wses outside and also may do inner processings , such as local variable assignments .
when it ends execution , it replies to the customer with its execution outcomes . in order to provide the adaptability of a wso , the bindings between its activities and wses outside are not direct and static
that is , wses are classified according to ontologies of specific domains and the wses belonging to the same ontology have same functions and interfaces , and different access points and different qos . to make this possible , from a system viewpoint , a name and directory service
uddi@xcite is necessary .
all wses with access information and qos information are registered into a uddi which classifies wses by their ontologies to be discovered and invoked in future .
uddi should provide multi interfaces to search wses registered in for its users , for example , a user can get information of specific set of wses by providing a service ontology and specific qos requirements via an interface of the uddi . the above mechanisms make qos - aware service selection possible . in a qos - aware wso engine , after a new wso instance is created , the new wso instance firstly selects its component wses according to the qos requirements provided by the customer and ontologies of component wses defined in the description file of the wso by ws - bpel . about qos of a ws@xcite@xcite , there are various qos aspects , such as performance qos , security qos , reliability qos , availability qos , and so on . in this paper
, we use a cost - effective qos approach .
that is , cost qos is used to measure the costs of one invocation of a ws while response time qos is used to capture effectiveness of one invocation of a ws . in the following ,
we assume all wses are aware of cost - effective qos .
according to the requirements of a wso engine discussed above , the architecture of qos - wsoe is given as fig.[fig.1 ] shows . in the architecture of qos - wsoe , there are external components , such as client , ws of a wso , uddi and component wses , and inner components , including wso instance manager , wso instances , activities , and service selector . among them , ws of a wso , uddi , wso instance manager and service selector are permanent components and client , component wses , wso instances , activities are transient components .
component wses are transient components since they are determined after a service selection process is executed by service selector . through a typical requirement process
, we illustrate the functions and relationships of these components . 1
. a client submits its requests including the wso ontology , input parameters and qos requirements to the ws of a wso through soap protocol .
2 . the ws transmits the requirements from a soap message sent by the client to the wso instance manager using private communication mechanisms .
3 . the wso instance manager creates a new wso instance including its activities and transmits the input parameters and the qos requirements to the new instance .
the instance transmits ontologies of its component wses and the qos requirements to the service selector to perform a service selection process via interactions with a uddi .
if the qos requirements can not be satisfied , the instance replies to the client to deny this time service .
if the qos requirements can be satisfied , each activity in the wso instance is bound to an external ws .
the wso instance transmits input parameters to each activity for an invocation to its binding ws .
after the wso instance ends its execution , that is , every invocation to its component wses by activities in the wso instance is returned , the wso instance returns the execution outcomes to the client .
an execution of a wso is called a wso instance ( wsoi ) .
a wsoi is created when the wso instance manager receive a new request ( including the functional parameters and the qos requirements )
. as fig.[fig.wsoi ] shows , once a wsoi is created , the wsoi is in the _ waiting _ state .
then the wso instance manager requires the service selector to perform a service selection process to select suitable component services .
if the qos requirements can not be satisfied , the wsoi transmits to _ denied _ state , otherwise , the wsoi transmits to the _ granted _ state .
when the wsoi starts to execute , that is , the first activity is executed , the wsoi is in the state of _ servicing _ and executes according to the definition of the wso .
finally , the wsoi ends to execute , that is , every activity of the wsoi is completed , the wsoi is with the state of _
completed_. every instance of an activity ( we do not distinguish the uses of an activity and an activity instance )
is included in a wsoi . when a wsoi is granted , all activities are in the state of _
preparing_. in addition , an activity is with the state of _ invoking _ when it is the executing turn of this activity .
once this activity is completed , that is , it gets the outcomes of the invocation to the external ws , it is in the _ returned _ state .
the state transitions is illustrated in fig.[fig.aa ] .
based on the state transitions of a wsoi and an activity as shown in fig.[fig.wsoi ] and fig.[fig.aa ] , we can get the state transitions of a bookstore wsoi as fig.[fig.bookstore ] illustrates when the qos requirements of the request are satisfied .
components in fig.[fig.1 ] are all implemented as actors .
some of these actors used in the following are client actor ( ca ) , wso instance manager actor ( wsoim ) , wso instance actor ( wsoi ) , activity actor ( aa ) , service selector actor ( ss ) , component ws actor ( ws ) .
we follow the symbol convention in @xcite and @xcite .
some are following , and others are introduced when we need .
@xmath2 denotes a system configuration .
@xmath3 denotes the set of names of actors existing in @xmath2 .
@xmath4 denotes the state of actor @xmath5 in @xmath2 .
@xmath6 gets the value of tag @xmath7 of actor @xmath5 in @xmath2 .
@xmath8 sets the value of tag @xmath7 of actor @xmath5 in @xmath2 with a value @xmath9 .
@xmath10 denotes an actor with a name @xmath5 and a state @xmath11 .
@xmath12 denotes a message with a type @xmath13 sent from actor @xmath5 to actor @xmath14 with optional parameters @xmath15 .
@xmath16\xrightarrow[effect]{trigger}\langle a : s'\rangle , mc \quad if \quad condition$ ] denotes a transition rule , where @xmath10 is an actor with name @xmath5 and state @xmath11 , @xmath17 is a message to actor @xmath5 with content @xmath18 , @xmath19 is the new state of actor @xmath5 , @xmath20 is the event triggering this rule , @xmath21 is the effect of this rule such as @xmath8 , @xmath22 is the set of ( possible empty ) messages sent in this rule , @xmath23 is the condition of occurrence of this rule .
@xmath24 denotes a transition , @xmath25 is the source configuration , @xmath26 is the target configuration , and @xmath27 is the label of the transition rule applied . @xmath28 $ ] denotes a computation path , which is a possibly infinite sequence of transitions .
@xmath29 is called @xmath30th transition of @xmath31 or @xmath30th stage of @xmath31 .
in this section , we will introduce the formal model of qos - wsoe .
firstly , we introduce actors and the symbols used in this section .
then we establish a pyramidal formal structure of qos - wsoe , including * qos - aware wso service * to model the requirements of the customers , * qos - aware wso system * to capture the properties of qos - aware wso engine , and * qos - aware wso behavior * to capture the behaviors of qos - aware wso engine .
let us use a wso ontology @xmath32 range over the set of wso ontologies @xmath33 , and let a request from the customer @xmath34 range over the set of requests @xmath35 . * definition 4.1 .
( wso request ) * a wso request actor @xmath34 is defined as a 4-tuples @xmath36 @xmath37 @xmath38 @xmath39 , where @xmath40 is the client i d , @xmath32 is the wso ontology requested , @xmath41 is the input parameters of the wso ontology , and @xmath42 is the qos requirements . * * definition 4.2 .
( wso request functions)**to manipulate a wso request @xmath34 , we define the following functions : ( 1)@xmath43 denotes the function to get the client i d ; ( 2)@xmath44 denotes the function to get the wso ontology ; ( 3)@xmath45 denotes the function to get the input parameters ; ( 4)@xmath46 denotes the function to get the qos requirements ; ( 5)@xmath47 denotes the function to get a wso request with a client i d @xmath40 from a configuration @xmath2 .
* * definition 4.3 .
( qos - aware wso service)**a system @xmath48 provides a qos - aware wso service over @xmath33 and @xmath35 iff for every configuration @xmath2 of @xmath48 , if there is an undelivered request @xmath34 in @xmath2 , then along any path @xmath31 of @xmath2 , exactly one of the following properties holds : ( 1)there is a unique transition @xmath49 in @xmath31 where @xmath34 is accepted for service and the qos requirements @xmath42 of @xmath34 can be satisfied ; ( 2)or there is a unique transition @xmath49 in @xmath31 where @xmath34 is rejected , only because the qos requirements @xmath42 of @xmath34 can not be satisfied when @xmath34 arrives .
when a @xmath34 message is sent to the wsoim , the wsoim creates a new wsoi .
now we give the definition of wsoi .
* * definition 4.4 .
( wsoi)**a wsoi is defined as a 4-tuples @xmath50 , where @xmath34 is the wso request , @xmath51 is the state of the wsoi , @xmath52 denotes the output parameters of the wsoi , and @xmath53 is activity actors included in the wsoi .
@xmath51 ranges over the set @xmath54 , where @xmath55 denotes that a wsoi is created and is waiting for further processing , @xmath56 denotes that a wsoi is accepted and the qos requirements can be satisfied , @xmath57 denotes that a wso is rejected because the qos requirements can not be satisfied , @xmath58 denotes that a wsoi is under running , and @xmath59 denotes that a wsoi is completed in running and the qos requirements are satisfied . * * definition 4.5 .
( wsoi functions)**to manipulate a wsoi , the following functions are defined : ( 1)@xmath60 denotes the function to get the wso request contained in the wso instance @xmath61 ; ( 2)@xmath62 denotes the function to get the state of the wso instance @xmath61 ; ( 3)@xmath63 denotes the function to get the set of activity actors contained in the wso instance @xmath61 ; ( 4)@xmath64 denotes the function to get the output parameters of the wso instance @xmath61 ; ( 5)@xmath65 creates a new aa name ; ( 6)@xmath66 denotes the function to get a wso instance @xmath61 with a client i d @xmath40 from the configuration @xmath2 , since a wso instance and a wso request are with 1:1 relation .
we define aa as follows .
* * definition 4.6 .
( aa)**an aa is defined as a 7-tuples @xmath67 , where @xmath68 is the name of the aa , @xmath40 denotes the i d of a wso instance , @xmath69 is the qos requirements of the aa , @xmath70 is the input parameters of the aa , @xmath71 is the output parameters of the aa , @xmath72 is the state of the aa , and @xmath73 is the ws bound after a service selection process .
@xmath72 ranges over @xmath74 , where @xmath75 denotes that an aa is created by a wsoi and is waiting for invoking a ws , @xmath76 denotes that an aa is now invoking a ws , and @xmath77 denotes that the invocation from an aa to a ws is completed . *
* definition 4.7 .
( aa functions)**to manipulate an aa , we define the following functions : ( 1)@xmath78 denotes the function to get an activity actor @xmath79 with a name @xmath68 contained in the wso instance @xmath61 ; ( 2)@xmath80 denotes the function to get an activity actor @xmath79 with a name @xmath68 contained in the wso instance @xmath40 from a configuration @xmath2 ; ( 3)@xmath81 denotes the function to get the qos requirements of an aa @xmath79 ; ( 4)@xmath82 denotes the function to get the input parameters of an aa @xmath79 ; ( 5)@xmath83 denotes the function to get the output parameters of an aa @xmath79 ; ( 6)@xmath84 denotes the function to get the state of an aa @xmath79 ; ( 7)@xmath85 denotes the function to get the name of an aa @xmath79 ; ( 8)@xmath86 denotes the function to get the wsoi i d of an aa @xmath79 ; ( 9)@xmath87 denotes a function to get the ws bound to an aa @xmath79 .
ws is defined as follows . *
* definition 4.8 .
( ws)**a ws is determined by a pair @xmath88 , where @xmath40 denotes the i d the a wsoi , and @xmath68 denotes the name @xmath68 of aa that the ws bound . since a ws is out of the management domain of the wsoi , we assume that a ws is always able to process the invocation of its customers , such as an aa . * * definition 4.9 . (
ws functions)**to manipulate a ws , we define the following functions : ( 1)@xmath87 denotes the function to get the ws bound of an aa @xmath79 ; ( 2)@xmath89 denotes the function to get wses included in a wsoi @xmath61 ; ( 3)@xmath90 denotes the function to get the ws bound to the aa with a name @xmath68 contained in the wso instance @xmath61 .
( 4)@xmath91 denotes the function to get a ws @xmath73 bound to an aa with a name @xmath68 contained in the wso instance @xmath40 from a configuration @xmath2 ; * * definition 4.10 .
( qos requirements of ws and wsoi)**we use a cost - effectiveness couple @xmath92 to capture the qos requirements of ws and wsoi .
if a wsoi has @xmath93 wses : @xmath94 , then there is the following relation between the wsoi and its wses : @xmath95 , @xmath96 .
since all aas included in a wsoi are executed in parallel by default , we can get the above equations if we ignore the causalities of messages sending among the aas . *
* definition 4.11 .
( qos - allocate function)**the qos - allocate function translates the qos requirements of a wsoi into the qos requirements of its wses implemented by some kind of qssa , such as in @xcite and @xcite , with an effect of binding aas in the wsoi to wses selected .
the qos - allocate function is defined as follows : @xmath97 @xmath98 where @xmath61 has @xmath93 wses and @xmath99 . * * definition 4.12 .
( aa - inputparametersgenerate function)**the aa - inputparametersgenerate function maps the input parameters of a wsoi to those of aas . the aa - inputparametersgenerate is defined as follows : @xmath100 where @xmath61 has @xmath93 wses and @xmath99 . *
* definition 4.13 .
( wsoi - outputparametersgenerate function)**the wsoi - outputparametersgenerate function maps the output parameters of aas to those of wsoi .
the wsoi - outputparametersgenerate is defined as follows : @xmath101 where @xmath61 has @xmath93 wses and @xmath99 .
* * definition 4.14 .
( wsoi constraints @xmath102)**the wsoi constraints @xmath102 includes : ( 1)a system @xmath48 satisfies @xmath102,@xmath103 , just if for every computation @xmath31 of @xmath48 , @xmath104 holds .
( 2)@xmath104holds , just if for every transition @xmath49 of @xmath31 , @xmath105 holds , and ( 2-a)a wsoi in some configuration @xmath2 with the state @xmath55 will eventually be granted or be denied .
@xmath106 ( 2-b)a wsoi in some configuration @xmath2 with the state @xmath56 will eventually be in servicing and be completed .
@xmath107 ( 3)@xmath105 holds for @xmath108 , just if ( 3-a)for a wsoreq in some configuration @xmath2 , once the corresponding wsoi is created , then the wsoreq and all aas in the wsoi are constant .
@xmath109 ( 3-b)for any wsoreq in some configuration @xmath2 , if the state of the corresponding wsoi is @xmath55 , then the wsoi may be in a state @xmath55 , or @xmath56 , or @xmath57 ; if the state of the corresponding wsoi is @xmath57 , then the wsoi will keep in the state @xmath57 ; if the corresponding wsoi is in a state @xmath56 , the it will still be in a state @xmath56 or in a new state @xmath58 ; if the corresponding wsoi is still in the state @xmath58 , the it will still be in a state @xmath58 or in a new state @xmath59 ; if the state of the corresponding wsoi is @xmath59 , then the wsoi will keep in the state @xmath59 .
@xmath110 @xmath111 , and @xmath112 , and @xmath113 , and @xmath114 , and @xmath115 .
( 3-c)if the corresponding wsoi of any wsoreq in some configuration @xmath2 is denied , then the wses included in the wsoi will always be @xmath116 , that is , after a service selection process executed for this wsoi , the qos requirements of the wsoreq can not be satisfied .
@xmath117 * * definition 4.15 .
( wsoi - ws constraints @xmath118)**a system @xmath48 satisfies @xmath118,@xmath119 , just if for every computation @xmath31 of @xmath48 , @xmath120 holds , iff for any ws included in the corresponding wso of any wsoreq in some configuration @xmath2 is not @xmath116 , then the wsoi will in a state @xmath56 , or @xmath58 , or @xmath59 , that is , @xmath121 * * definition 4.16 .
( qos - aware wso system)**a system @xmath48 is a qos - aware wso system , with respect to request @xmath34 in @xmath35 , the above functions including qos - allocate function , aa - inputparametersgenerate function , wsoi - outputparametersgenerate function , wso request functions , wsoi functions , aa functions and ws functions , iff ( 1)@xmath48 satisfies @xmath103 and @xmath119 ; ( 2)for @xmath122 , if there is an undelivered request @xmath34 with parameters @xmath123 , then along any path @xmath31 of @xmath2 there is a unique stage @xmath30 transition : @xmath124 to process @xmath34 , and there is a newly created wso instance @xmath125 , with parameters : * @xmath60 ; * @xmath126 ; * @xmath127 ; * for @xmath128 , * * @xmath129 ; * * @xmath130 ; * * @xmath131 ; * * @xmath132 ; * * @xmath86 ; * * @xmath133 ; * * @xmath134 . * * theorem 4.1 .
( service2system)**if a system @xmath48 is a qos - aware wso system as definition 4.16 shows , then this system @xmath48 provides qos - aware wso service as defined in definition 4.3 .
we firstly see that if there is an undelivered request @xmath34 for any @xmath2 in @xmath48 , according to definition 4.16 , a new wso instance @xmath61 is created and @xmath135 .
then @xmath136 or @xmath137 according to definition 4.14 , if in @xmath138 , a reply is sent to the customer to reject the request ; if in @xmath139 , then @xmath140 and @xmath141 .
thus , @xmath34 is assured to be just accepted or just rejected .
qos - allocate function , @xmath102 , @xmath118 and definition 4.8 guarantee that the qos requirements are satisfied in case that @xmath34 is accepted . in this section
, we will specify the behavior of actors in fig.[fig.1 ] , including their states , messages exchanged among them and their transition rules .
then we define the concept of qos - aware wso behavior and draw two conclusions .
wsoi , aa and ws are defined in the above section including their states .
now we will explain wsoim and ss .
the functions of wsoim are creating wsois when incoming requests from ca arrive .
we assume that wsoim are always ready for servicing .
and the functions of ss are providing qos - aware service selection processing for wsois and ss is also always ready for servicing .
we define the messages exchanged among these actors as table.[table.1 ] illustrates .
[ cols="<,<",options="header " , ] [ table.1 ] the followings are the transition rules for wsoim and wsoi .
( 1)rule for wsoim creating new wsoi : when the wsoim receives a incoming request message , then a new wsoi is created and a service selection processing message will be sent from the new wsoi to the ss .
@xmath142 { } \langle wsoim : wsoimstate'\rangle,\\ ss\lhd select(qos , wso)@wsoi\nonumber $ ] .
where @xmath143 , and @xmath144 ( 2)rules for wsoi processing reply from ss : after a service selection process is executed by the ss , a reply message from the ss to the wsoi is sent .
if the qos requirements can not be satisfied , then the wsoi is in a new state @xmath57 and a _ denied _ reply message is sent from the wsoi to the customer .
if the qos requirements can be satisfied , then the wsoi is in a new state @xmath56 and a _ granted _ reply message is sent from the wsoi to the customer and a _ invoke _ message will be sent to any aa included in the wsoi from the wsoi .
@xmath145)@ss\\ \rightarrow \langle wsoi : denied\rangle,\\ ca\lhd denied(wso , qos)@wsoi $ ] .
@xmath146)@ss\\ \xrightarrow[effect_{grantedwsoi } ] { } \langle wsoi : granted\rangle , \\
ca\lhd granted(wso , qos)@wsoi,\\ ( \forall aa\in { getaas(wsoi ) } ) aa\lhd invoke()@wsoi $ ] . where @xmath147 , and @xmath148 , where @xmath149 denotes the requirements after a service selection processing by ss , @xmath150 denotes input parameters generation of aa from the wsoi , and @xmath151 denotes the ws bound to the aa after a service selection processing by ss .
( 3)rule for wsoi processing reply from aa : when a wsoi in a state @xmath56 or @xmath58 receives a reply message _ invokeack _ from an aa , then it will be in a new state @xmath58 .
@xmath152 .
( 4)rules for wsoi processing notification from aa : when a wsoi in a state @xmath58 receives a notification message from an aa , if all aas included in the wsoi are in state @xmath77 , then the wsoi will be in a new state @xmath59 and a _ completed _ reply message will be sent from the wsoi to the customer .
@xmath153)@aa\nonumber\\ \xrightarrow[effect_{completedwsoi } ] { } \langle wsoi : completed\rangle , \\
ca\lhd completed(wso , qos , ops)@wsoi $ ] .
where @xmath154 and @xmath155 , and @xmath156 , where @xmath157 denotes the results after a ws invocation , @xmath158 denotes output parameters generation of wsoi from the aas .
when a wsoi in a state @xmath58 receives a notification message from an aa , if there is an aa included in the wsoi is not in state @xmath77 , then the wsoi will keep the state @xmath58 .
@xmath159)@aa\nonumber\\ \rightarrow \langle wsoi : servicing\rangle $ ] . where @xmath160 .
the following is the transition rule for ss to process request from wsoi : when the ss receives a
request message from a wsoi , then a reply message _ selectreply _ will be sent to the wsoi for denying the request or granting the request .
@xmath161)@ss\\ \quad or wsoi\lhd selectreply\\\quad(wsoi=[state = granted])@ss ) $ ] .
the followings are the transition rules for aa .
( 1)rule for aa to process request from wsoi : an aa in a state @xmath75 receives a request message _ invoke _ from its wsoi , then it will be in a new state @xmath76 , send an _ invokeack _ reply message to the wsoi , and send an _ invoke _ request message to its binding ws .
@xmath162 .
( 2)rule for aa to process reply from ws : when an aa in a state @xmath76 receives an _ invokereply _ reply message from its binding ws , it will be in a new state @xmath77 and send a _
notify _ notification message to its wsoi .
@xmath163 { } \langle aa : returned\rangle , \\ wsoi\lhd notify(aa=[state = returned])@aa\nonumber $ ] . where @xmath164 , and @xmath165 , where @xmath157 denotes the results after a ws invocation .
the following is the transition rule for ws to accept an invocation from an aa : when a ws receives an _ invoke _ request message from its binding aa , it will do some inner computations and send an _ invokereply _ reply message to its binding aa .
@xmath166 . * * definition 4.17 .
( qos - aware wso behavior)**a system @xmath48 has qos - aware wso behavior , with respect to architecture as fig.[fig.1 ] shows and actors wsoim , wsoi , aas , wses , ss , if ( 1)for @xmath2 in @xmath48 , states of actors in @xmath2 are just in accordance with definitions of states of actors ; ( 2)for @xmath2 in @xmath48 , messages in @xmath2 are just as table.[table.1 ] define ; ( 3)for every computation @xmath31 in @xmath48 , @xmath31 obeys the transition rules defined above . * * theorem 4.2 .
( system2behavior)**if a system @xmath48 has qos - aware wso behavior as definition 4.17 defines , then @xmath48 is a qos - aware wso system as definition 4.16 defines . by the states definition of actors
wsoim , wsoi , aas , wses , ss , message definitions as table.[table.1 ] defined and the transition rule definitions above , we can see that(1)@xmath103 and @xmath119 hold ; ( 2)for @xmath122 , if there is an undelivered request @xmath34 , a new wso instance @xmath61 is assured to be created . * * theorem 4.3 .
( service2behavior)**if a system @xmath48 has qos - aware wso behavior as definition 4.17 defines , then @xmath48 provides qos - aware wso service as defined in definition 4.3 . by theorem 4.1 and theorem 4.2 .
the behavior of qos - aware bookstore wso is embodied by the following transition rules .
( 1 ) @xmath142 { } \langle wsoim : wsoimstate'\rangle,\\ ss\lhd select(qos , wso)@wsoi\nonumber $ ] .
@xmath61 is a bookstore wsoi , and ( 2 ) @xmath161)@ss ) $ ] .
@xmath61 is the bookstore wsoi , and ( 3 ) @xmath145)@ss\\ \rightarrow \langle wsoi : denied\rangle,\\ ca\lhd denied(wso , qos)@wsoi $ ] .
@xmath61 is the bookstore wsoi . or , ( 1 ) @xmath142 { } \langle wsoim : wsoimstate'\rangle,\\ ss\lhd select(qos , wso)@wsoi\nonumber $ ] .
@xmath61 is the bookstore wsoi , and ( 2 ) @xmath167)@ss ) $ ] .
@xmath61 is the bookstore wsoi , and ( 3 ) @xmath146)@ss\\ \xrightarrow[effect_{grantedwsoi } ] { } \langle wsoi : granted\rangle , \\
ca\lhd granted(wso , qos)@wsoi,\\ ( \forall aa\in { getaas(wsoi ) } ) aa\lhd invoke()@wsoi $ ] .
@xmath61 is the bookstore wsoi , @xmath168 , and ( 4 ) @xmath162 .
@xmath61 is the bookstore wsoi , @xmath168 , @xmath169 , and ( 5 ) @xmath166 .
@xmath168 , @xmath169 , and ( 6 ) @xmath163 { } \langle aa : returned\rangle , \\
wsoi\lhd notify(aa=[state = returned])@aa\nonumber $ ] .
@xmath61 is the bookstore wsoi , @xmath168 , @xmath169 , and ( 7 ) @xmath152 .
@xmath61 is the bookstore wsoi , @xmath168 , and ( 8) @xmath153)@aa\nonumber\\ \xrightarrow[effect_{completedwsoi } ] { } \langle wsoi : completed\rangle , \\
ca\lhd completed(wso , qos , ops)@wsoi $ ] .
@xmath61 is the bookstore wsoi , @xmath168 .
naturally , qos - aware bookstore wso has qos - aware wso behavior , and further provides qos - aware wso service according to theorem 4.3 .
in this paper , we have discussed issues on qos - aware wso and design a typical qos - aware wso engine called qos - wsoe . mainly ,
a formal model of qos - wsoe based on actor systems theory is established . in the formal model
, a three - layered pyramidal structure is adopted to capture the requirements of the customers with a concept named qos - aware wso service , characteristics of qos - wsoe with a concept named qos - aware wso system , and behaviors of qos - wsoe with a concept named qos - aware wso behavior and a fine relationship among these three layers is established .
i hope this paper can be a guidance for implementing a real qos - aware wso engine with correctness assurance .
m. coabano , e. denti , a. ricci , m. viroli , _ designing a bpel orchestration engine based on respect tuple centres _ , proceedings of the 4th international workshop on the foundations of coordination languages and software architectures ( foclasa 2005 ) , 139158 .
m. viroli , e. denti , a. ricci , _ engineering a bpel orchestration engine as a multi - agent system _ , special issue on the 4th international workshop on foundations of coordination languages and software architectures ( foclasa 2005 ) , 226245 .
g. agha , t. prasanna and z. reza , _ actors : a model for reasoning about open distributed systems _
, formal methods for distributed processing : a survey of object - oriented approaches , cambridge university press , 2001 .
m. c. astley , _ customization and composition of distributed objects : policy management in distributed software architectures _ , ph.d .
dissertation , university of illinois , urbana - champaign . for computer science , 1999 . c. a. varela , _ worldwide computing with universal actors : linguistic abstractions for naming , migration , and coordination _ , ph.d .
dissertation , university of illinois , urbana - champaign . for computer science , 2001 . | qos - aware applications can satisfy not only the functional requirements of the customers , but also the qos requirements .
qos - aware web service orchestration translates the qos requirements of the customers into those of its component web services . in a system viewpoint , we discuss issues on qos - aware web service orchestration and design a typical qos - aware web service orchestration engine called qos - wsoe .
more importantly , we establish a formal model of qos - wsoe based on actor systems theory . within the formal model
, we use a three - layered pyramidal structure to capture the requirements of the customers with a concept named qos - aware wso service , characteristics of qos - wsoe with a concept named qos - aware wso system , and structures and behaviors of qos - wsoe with a concept named qos - aware wso behavior .
conclusions showing that a system with qos - aware wso behavior is a qos - aware wso system and further can provide qos - aware wso service are drawn .
[ firstpage ] web services ; web service orchestration ; web service orchestration engine ; actor systems ; qos ; formal model |
why is gravity so weak ? or equivalently , why are the proton ( and other hadrons ) so light compared to the planck scale ? as wilczek has explained nicely @xcite , the solution to this hierarchy puzzle results from asymptotic freedom . without any fine - tuning of the bare gauge coupling at distances as short as the planck length ,
a 4-d non - abelian gauge theory like qcd produces a correlation length @xmath3 that is larger than the shortest length scale in the problem , by a factor exponentially large in the inverse coupling .
the inverse correlation length defines a mass scale @xmath4 the proton mass in wilczek s example that is hence exponentially smaller than the planck scale .
since we consist mostly of these very light particles ( protons and neutrons ) in natural units of their mass we experience gravity as a very weak force .
lattice gauge theorists benefit from the absence of a hierarchy problem in numerical simulations of gauge theories . in this case the shortest length scale ( and hence the analog of the planck length ) is the lattice spacing .
again , thanks to asymptotic freedom , without any fine - tuning of the bare gauge coupling , the physical correlation length can be made arbitrarily large in lattice units .
hence , putting aside practical difficulties due to the limited power of computers , there is no problem of principle in approaching the continuum limit in 4-d yang - mills theories on the lattice .
unfortunately , the situation is not as simple in full lattice qcd including quarks .
in fact , for a long time lattice field theorists have suffered from a hierarchy problem in the fermion sector .
this problem arises when one removes the unwanted doubler fermions by breaking chiral symmetry explicitly , for example , by introducing a wilson term in the lattice fermion action .
recovering chiral symmetry in the continuum limit then requires a delicate fine - tuning of the bare fermion mass .
this is not only a pain in practical numerical computations , but should be considered a serious problem of principle at the heart of the nonperturbative regularization of theories with a chiral symmetry .
when one works in the continuum , one often takes chiral symmetry for granted because it can be maintained in dimensional regularization .
however , the subtleties related to the definition of @xmath5 in the framework of dimensional regularization are just another aspect of the same deep problem of chiral symmetry that is obvious on the lattice . returning to the hierarchy puzzle
, we should ask why the quarks inside the proton are light .
in particular , we should still be puzzled by the fact that we consist of not just of gluons , but also of light quarks .
certainly , if one imagines wilson s lattice qcd as a ( highly simplified ) model for the short distance physics of our world , without unnatural fine - tuning , quarks would live at the lattice spacing scale , while gluons ( or , more precisely , glueballs ) are naturally light .
fortunately , the long - standing hierarchy problem of lattice fermions and hence of the nonperturbative regularization of chiral symmetry has recently been solved .
based on work of callan and harvey @xcite , the first step in this direction was taken by kaplan @xcite who realized that massless lattice fermions arise naturally , i.e. without fine - tuning , as zero - modes localized on a domain wall embedded in a higher - dimensional space - time .
for example , left - handed fermions in four dimensions can be localized on a domain wall that represents a 3-brane embedded in a 5-d space - time .
similarly , right - handed fermions can be localized on an anti - wall . by keeping the wall and the anti - wall at a sufficiently large distance , i.e. by spatially separating left- and right - handed fermions in the extra dimension , they are prevented from mixing strongly with one another .
thus , they are protected from picking up a large mass and are naturally light .
since @xmath5 appears in its action , the 5-d theory itself does not even have a chiral symmetry .
hence , in contrast to four dimensions , a wilson term in the 5-d lattice action removes the doubler fermions without doing damage to chiral symmetry .
overlap lattice fermions @xcite also live in an extra dimension and are closely related to domain wall fermions .
when one integrates out the extra dimension , the 4-d version of overlap fermions satisfies the so - called ginsparg - wilson relation @xcite . as lscher first realized , this relation implies a lattice version of chiral symmetry @xcite which led him to a spectacular breakthrough : the nonperturbative construction of lattice chiral gauge theories @xcite .
perfect and classically perfect lattice fermion actions @xcite also obey the ginsparg - wilson relation @xcite .
however , the explicit construction of such actions is a delicate problem that can itself be considered a form of fine - tuning .
fermion actions that naturally obey the ginsparg - wilson relation without fine - tuning , on the other hand , can be related to the physics in a higher - dimensional space - time .
the existence of light 4-d fermions may hence be a hint to the physical reality of extra dimensions .
indeed , brane worlds embedded in a higher - dimensional space - time provide a very interesting perspective on ordinary 4-d physics @xcite . in present lattice qcd applications of domain wall or overlap fermions the extra dimension is not taken physically seriously . in particular , the gluon field is usually frozen in the fifth dimension , thus violating locality in the extra dimension . for left- and right - handed domain wall fermions localized on a wall and an anti - wall ,
one takes the chiral limit by separating the wall from the anti - wall in the extra dimension , while the gluonic correlation length is kept fixed .
consequently , one approaches the chiral limit before the continuum limit , and the extra dimension does not have a physically meaningful extent . in order to cancel singularities resulting from the 5-d bulk , one introduces ghost fields violating the spin - statistics theorem . while all this is not wrong if one only wants to construct 4-d qcd , it is unnatural if one takes the fifth dimension seriously .
there is an alternative nonperturbative formulation of qcd ( and other field theories ) d - theory @xcite in which the extra dimension is used in a more natural way . in d
- theory the familiar classical fields emerge dynamically from discrete variables ( such as quantum spins or quantum links ) which undergo dimensional reduction . in this formulation of qcd
, gluons emerge dynamically from higher - dimensional physics . starting with a 5-d quantum link model from which a non - abelian coulomb phase with massless gluons emerges
, compactification of the fifth dimension leads to a correlation length ( i.e. an inverse glueball mass ) that is exponentially large in the size of the extra dimension . as a consequence of asymptotic freedom of 4-d qcd
, the size of the extra dimension then shrinks to zero in physical units as one approaches the continuum limit .
domain wall fermions fit naturally into this framework .
in particular , the chiral and continuum limits are reached simultaneously , the theory is fully local in the extra dimension , no unphysical ghost fields are needed , and the extra dimension completely disappears in the continuum limit via dimensional reduction . if one imagines the d - theory regularization of qcd as an ( again highly simplified ) short distance model of our world , not just gluons , but also quarks are naturally light .
this solves the second part of the hierarchy puzzle : if we assume the existence of a fifth dimension , we need no longer be puzzled why we also consist of light quarks .
in this paper we use the 3-d gross - neveu model @xcite at large @xmath0 as an analytically soluble toy model for illustrating the issues discussed above . using the standard procedure ,
domain wall fermions were applied to the gross - neveu model in ref .
@xcite . in that calculation the fermionic left- and right - handed zero - modes localized on a wall and an anti - wall are coupled by hand .
while this is sufficient if one just wants to construct the 2-d gross - neveu model , it violates locality in the extra dimension .
this is unnatural if one takes that dimension seriously . in this paper , in the spirit of d - theory
, we do take the extra dimension physically seriously and let the left- and right - handed zero - modes communicate through the bulk of the extra dimension by a locally implemented 4-fermion interaction .
then , just as in qcd , the d - theory mechanism of dimensional reduction leads to naturally light fermions without fine - tuning . in our calculation
not even the domain walls on which fermion states can be localized are put by hand .
they arise dynamically from the spontaneous breakdown of the @xmath6 chiral symmetry of the gross - neveu model .
the walls play the role of branes and hence our construction can be viewed as an attempt to construct stable brane worlds .
we find exact brane world solutions just relying on the dynamics of the underlying higher - dimensional theory , without making ad - hoc assumptions about where the branes shall be located .
in realistic brane world constructions stabilizing the branes is a highly non - trivial issue . in our model
we construct an exact solution for a fully consistent and stable wall - anti - wall configuration .
although our toy model is very simple , it reveals an interesting mechanism for brane stabilization : if branes are made of fermions ( and thus have a baryon asymmetry ) baryon number conservation in the higher - dimensional theory may imply brane stability .
unfortunately , in our toy `` world '' the baryon asymmetry is so large that all states are occupied with fermions and any potentially non - trivial physics is completely pauli - blocked . in section 2
basic features of the 2-d gross - neveu model are reviewed , while section 3 discusses the 3-d gross - neveu model . in section 4 the 3-d gross - neveu model
is considered in the chirally symmetric phase .
after compactification of the third dimension , the model undergoes dimensional reduction to the 2-d gross - neveu model by the generation of a correlation length that is exponentially large in the size of the extra dimension .
section 5 discusses dimensional reduction from the chirally broken phase of the 3-d model , using configurations with either one wall or a wall - anti - wall pair . in the wall - anti - wall case
a naturally light dirac fermion is generated with its left- and right - handed components being localized on the two walls .
again , the theory undergoes dimensional reduction to the 2-d gross - neveu model .
the emerging chiral order parameter of the 2-d theory as well as the brane tensions and the issue of brane world stability are also discussed .
finally , section 6 contains our conclusions .
a synopsis of this work is given in ref .
in this section , we introduce the 2-d gross - neveu model @xcite as the target theory which we will later obtain from dimensional reduction of the corresponding 3-d model .
various 2-d field theories , including the gross - neveu model , were recently reviewed in @xcite .
we consider the gross - neveu model at large @xmath0 in the continuum .
its euclidean action is given by @xmath7 = \int d^2x \left[{\overline \psi}\gamma_{\mu } \partial_{\mu } \psi - \frac{g}{2n } \left({\overline \psi}\psi \right)^2\right].\ ] ] we have suppressed a flavor index that runs from 1 to @xmath0 and gives rise to a global @xmath8 flavor symmetry .
the parameter @xmath9 is the dimensionless 4-fermion coupling which remains fixed in the t hooft limit @xmath10 .
the explicit factor @xmath11 ensures that the interaction term stays of the same magnitude as the kinetic term in this limit .
in addition to the @xmath8 flavor symmetry , the model has a @xmath6 chiral symmetry @xmath12 where @xmath13 let us consider the @xmath14 limit and derive a gap equation that describes the spontaneous breakdown of the @xmath6 symmetry of eq.([symmetry2 ] ) .
first , we introduce an auxiliary scalar field @xmath15 representing the chiral order parameter @xmath16 .
it linearizes the 4-fermion interaction , such that @xmath17 = \int d^2x \left[{\overline \psi}\gamma_{\mu } \partial_{\mu } \psi - { \overline \psi}\psi \phi + \frac{n}{2 g } \phi^2 \right].\ ] ] the partition function then takes the form @xmath18 ) = \int { \cal d}{\overline \psi}{\cal d}\psi { \cal d}\phi \exp(- s[{\overline \psi},\psi,\phi]).\ ] ] in the large @xmath0 limit we may restrict @xmath19 to a space - time - independent constant @xmath20 @xcite . doing so , the action in momentum space takes the form @xmath17 = \frac{1}{(2\pi)^2 } \int d^2k \
{ \overline \psi}(- k)[i \gamma_\mu k_\mu - \phi ] \psi(k ) + \frac{n v}{2 g } \phi^2,\ ] ] where @xmath21 is the volume of space - time .
we integrate out the fermion fields @xmath22 ) = \exp(- n v v_{eff}(\phi)),\ ] ] in order to obtain the effective potential @xmath23 of the chiral order parameter . in the infinite volume limit the fermion integration yields @xmath24 using eq.([epot2 ] ) , the minimum @xmath25 of the effective potential is determined by @xmath26 which implies the gap equation @xmath27 in order to solve the gap equation , a cut - off @xmath28 is introduced in 2-d momentum space .
the ultraviolet limit @xmath29 is reached for @xmath30 and one then obtains @xmath31 the exponential factor in eq.([massgap2 ] ) is a manifestation of asymptotic freedom of the 2-d gross - neveu model .
the factor @xmath32 in the exponent is the inverse of the corresponding 1-loop @xmath33-function coefficient . due to spontaneous chiral symmetry breaking
, the fermions pick up a mass @xmath34 .
the cut - off @xmath28 is removed by varying the bare coupling @xmath9 such that the physical fermion mass @xmath1 remains fixed . after removing the cut - off
, the effective potential takes the form @xmath35 which is depicted in figure 1 .
next we consider the 3-d gross - neveu model @xcite with the euclidean action @xmath36 = \int d^3x \left[{\overline \psi}\gamma_\mu \partial_\mu \psi + { \overline \psi}\gamma_3 \partial_3 \psi - \frac{g}{2n } \left({\overline \psi}\psi \right)^2\right].\ ] ] in three dimensions the 4-fermion coupling @xmath37 has the dimension of a length .
we will soon compactify the third direction or endow it with domain walls and obtain the 2-d gross - neveu model by means of dimensional reduction . hence , in this case , the 3-direction is not euclidean time but just an additional spatial dimension which will ultimately become invisible .
instead , we choose the 2-direction to represent euclidean time . in three dimensions
there is no chiral symmetry because @xmath38 appears explicitly in the action .
still , the 3-d action has a @xmath6 symmetry which reduces to the chiral symmetry of the 2-d gross - neveu model after dimensional reduction , but which also involves a spatial inversion in the 3-direction , @xmath39 later we will consider the @xmath40-independent zero - mode that determines the physics of the dimensionally reduced 2-d gross - neveu model . then eq.([symmetry3 ] )
reduces to the usual 2-d @xmath6 chiral symmetry of eq.([symmetry2 ] ) . as in the 2-d case , we take the large @xmath0 limit and derive a gap equation @xmath41 introducing a cut - off @xmath42 in 3-d momentum space , and assuming @xmath43 , in the ultraviolet limit @xmath44 we obtain @xmath45 we have introduced the critical coupling constant @xmath46 . at strong coupling , @xmath47 ,
one has @xmath48 , so we are in the broken phase . for weak coupling , @xmath49 , on the other hand
, we are in the symmetric phase . for large @xmath42 ( and up to a trivial additive constant ) the effective potential reduces to @xmath50 in the symmetric phase the effective potential @xmath51 has a single minimum at @xmath52 , while in the broken phase it has two degenerate minima at @xmath53 .
the effective potential in the symmetric and broken phases is depicted in figures 2 and 3 , respectively .
in this section we start from the symmetric phase of the 3-d gross - neveu model . by compactifying the third dimension to a circle of circumference @xmath33 ,
the system is dimensionally reduced to the 2-d gross - neveu model .
interestingly , the massless fermion that exists in the bulk of the 3-d gross - neveu model can not remain massless after compactification .
if it would , the size @xmath33 of the extra dimension would be negligible compared to the then infinite fermionic correlation length , and the model would become a massless 2-d gross - neveu model . however , as we have seen , in the 2-d gross - neveu model the @xmath6 chiral symmetry is spontaneously broken for all values of the coupling constant @xmath9 .
hence , as the third dimension is compactified , the previously massless 3-d fermion must necessarily pick up a mass . consequently
, its correlation length @xmath2 will become finite .
then the question arises if @xmath3 is large or small compared to @xmath33 .
let us study the mechanism of dimensional reduction from the symmetric phase in some detail .
the 3-d gap equation with periodic boundary conditions in the third direction takes the form @xmath54 to evaluate the sum , we use the poisson resummation formula and we obtain @xmath55 the gap equation then reads @xmath56 again , we introduce the cut - off @xmath28 in 2-d momentum space and perform the integral . for large @xmath28
we obtain @xmath57 in order to match the 2-d cut - off @xmath28 to the 3-d cut - off @xmath42 , for a moment we consider the broken phase of the 3-d model .
then @xmath58 approaches its constant bulk value as @xmath59 , such that eq.([massgap3to2 ] ) implies @xmath60 we match this with the previous result of eq.([phi0 ] ) by identifying @xmath61 returning to the symmetric phase of the 3-d model , @xmath58 vanishes as @xmath59 and eq.([massgap3to2 ] ) implies @xmath62 interestingly , the correlation length @xmath3 becomes exponentially larger than @xmath33 as @xmath33 itself goes to infinity .
hence , counter - intuitively , as the size of the third dimension becomes large ( in units of the inverse cut - off @xmath63 ) the physical correlation length of the fermion increases exponentially and the low - energy physics reduces to the one of the 2-d gross - neveu model .
there is a hierarchy of three separate length scales in this problem .
the shortest length scale is determined by the inverse cut - off @xmath63 .
the next scale @xmath64 is the size of the extra dimension . finally , the largest length scale @xmath65 is dynamically generated by spontaneous chiral symmetry breaking in the dimensionally reduced 2-d model .
in fact , the size @xmath33 of the extra dimension plays the role of the inverse cut - off for the 2-d gross - neveu model that arises by means of dimensional reduction . identifying the coupling constant of the dimensionally reduced model as @xmath66 eq.([xi ] ) turns into @xmath67 which is nothing but the asymptotic freedom relation eq.([massgap2 ] ) of the 2-d gross - neveu model .
in particular , we identify @xmath68 as the effective cut - off of the dimensionally reduced model .
the same subtle mechanism of dimensional reduction was first observed in the 3-d @xmath69 sigma model @xcite .
however , in that case dimensional reduction results only when one starts in the broken phase of the 3-d model , which contains massless goldstone bosons . as a consequence of the mermin - wagner - coleman theorem
, these particles can not remain massless as the theory dimensionally reduces to two dimensions .
since the @xmath6 symmetry of the gross - neveu model is discrete , the theorem does not apply in this case . in the 3-d gross - neveu model
the massless phase is the one without spontaneous symmetry breaking .
still , both the 2-d @xmath69 model and the 2-d gross - neveu model are massive , although for different reasons . in the next section
we will see how dimensional reduction arises starting from the broken phase of the 3-d gross - neveu model .
figure 4 shows the chiral condensate ( or equivalently the mass gap ) as a function of @xmath37 both for finite and for infinite @xmath33 . from the point of view presented in the introduction , the way we have obtained the 2-d gross - neveu model by means of dimensional reduction from the symmetric phase of the 3-d model is not really satisfactory .
since we studied the large @xmath0 limit , an analytic calculation was possible in the continuum using a momentum cut - off that does not break chiral symmetry explicitly . on the other hand , a fully nonperturbative treatment at finite
@xmath0 would require a formulation on the lattice , which leads to the fermion doubling problem .
for even @xmath0 this problem can be avoided using staggered fermions . however , for odd @xmath0 one would use wilson fermions and thus break chiral symmetry explicitly by the regulator . recovering chiral symmetry in the continuum limit then requires unnatural fine - tuning of the bare fermion mass .
the idea of domain wall fermions is to construct naturally light fermions via dimensional reduction from a higher dimension . in the next section
, we will see explicitly how this works in the broken phase of the 3-d gross - neveu model .
dimensional reduction from the symmetric phase , on the other hand , requires a 3-d massless fermion to begin with . in a fully nonperturbative lattice calculation at finite @xmath0
this would require unnatural fine - tuning at the level of the 3-d theory .
hence , the above mechanism of dimensional reduction from the symmetric phase does not work naturally ( i.e. without fine - tuning ) when the cut - off violates chiral symmetry .
this is why we now turn to the chirally broken phase .
in this section we describe how the 2-d gross - neveu model can arise via dimensional reduction from the broken phase of the 3-d model .
if one would again compactify the third dimension to a circle , the finite correlation length @xmath70 of the fermion in the 3-d bulk would never become large compared to @xmath33 .
hence , with periodic boundary conditions , dimensional reduction would not happen from the broken phase , even in the @xmath71 limit , because the ratio @xmath72 remains non - zero in this limit .
light fermions in @xmath73 dimensions can arise naturally by dimensional reduction from a massive theory in @xmath74 dimensions as zero - modes localized on a domain wall @xcite .
interestingly , domain walls indeed exist as stable topological objects in the broken phase of the 3-d gross - neveu model .
we will see that a system with a single domain wall has a free massless left - handed 2-d fermion living on the wall .
a system with a stable wall - anti - wall pair separated by a distance @xmath33 , on the other hand , supports a left - handed fermion living on the wall as well as a right - handed fermion living on the anti - wall .
the left- and right - handed fermionic zero - modes communicate with each other through the 3-d bulk between the wall and the anti - wall and pick up a mass @xmath1 .
interestingly , in analogy to the mechanism of dimensional reduction discussed before , the corresponding correlation length @xmath2 grows exponentially with @xmath33 when one separates the wall from the anti - wall .
hence , in units of the 2-d fermion correlation length @xmath3 , the size @xmath33 of the bulk between the walls shrinks to zero .
consequently , from the low - energy point of view of a 2-d observer living on the walls , the wall - anti - wall system although truly 3-dimensional looks exactly like a 2-d space - time .
first we consider a system with a single domain wall separating two distinct broken phases with chiral condensate values @xmath58 and @xmath75 .
the domain wall itself is described by a field configuration @xmath76 that interpolates between the two vacuum states , @xmath77 . determining the shape @xmath76 of the domain wall is a non - trivial problem .
fortunately , a similar problem has been solved a long time ago by dashen , hasslacher , and neveu @xcite in the 2-d gross - neveu model . in that case , the topological object is not a domain wall but simply a solitonic particle . as we will see , the 2-d soliton solution of ref .
@xcite naturally extends to three dimensions . in order to determine @xmath76 we apply the following strategy
first , we make an educated guess ( the 2-d soliton solution ) for @xmath76 .
second , we integrate out the fermions in the given non - trivial background , and we verify explicitly that the resulting @xmath78 self - consistently reproduces @xmath76 . our approach is a 3-d generalization of work by pausch , thies , and dolman @xcite and by feinberg @xcite , and is also related to work by chandrasekharan @xcite .
the ansatz for the domain wall profile which is obvious from what we know about 2-d solitons is given by the standard kink profile @xmath79 we have arbitrarily centered the domain wall at @xmath80 . of course , there is a trivial translational zero - mode which does not concern us here . in minkowski space - time
the fermions propagating in the non - trivial domain wall background field @xmath76 are described by the single - particle dirac hamiltonian @xmath81.\ ] ] we choose the euclidean dirac matrices to coincide with the pauli matrices , @xmath82 .
after dimensional reduction the 1-direction will remain and from the low - energy point of view of a 2-d observer the spatial 3-direction becomes invisible .
using translation invariance of the dirac equation in both , the spatial 1-direction and in time , we make the separation ansatz @xmath83 which implies @xmath84 there is a localized state @xmath85 with the energy eigenvalue @xmath86 .
this state describes a free massless left - handed fermion with spatial momentum @xmath87 localized on the domain wall .
note that the localized states of positive energy are left - movers with @xmath88 . since in 2-d
there is no spin , a left - handed particle is indeed simply moving to the left .
similarly , the particles localized on an anti - wall with the profile @xmath89 are right - movers .
the wall profile as well as the wave function of the localized state are illustrated in figure 5 .
the spectrum of the dirac equation also includes a continuum of bulk states @xmath90 with the energy eigenvalues @xmath91 .
the localized and bulk states together form a complete orthonormal basis of the single particle hilbert space . in the next step ,
we verify explicitly that the solutions from above self - consistently reproduce the domain wall profile of eq.([profile ] ) . for this purpose
we evaluate @xmath92 summing over the filled dirac sea of negative energy states , in the limit @xmath93 one finds @xmath94 i.e. the ansatz of eq.([profile ] ) for the profile of the domain wall is indeed reproduced by the actual @xmath78 of the fermions living in the corresponding background field .
the factor @xmath11 has canceled because each negative energy state is filled with @xmath0 fermions .
we have also used eq.([cutoff ] ) to eliminate the cut - off @xmath28 in favor of @xmath37 and @xmath58 .
since the modes localized on the wall have @xmath95 , they have no effect on the self - consistency of the domain wall solution itself .
in particular , one can also fill the negative energy states for those modes and thus construct the usual vacuum for the 2-d physics on the brane . by occupying a localized state @xmath96 with spatial momentum @xmath97 ,
we then obtain a left - moving particle with energy @xmath98 . at low energies @xmath99 the fermions that are localized on the brane
do not feel the extra dimension and see just a 2-d space - time . at sufficiently large energies
@xmath100 , on the other hand , fermions can escape into the extra dimension .
from the point of view of a 2-d observer , this process seems to violate fermion number conservation .
at this point we have explicitly constructed a topologically stable brane on which massless left - handed fermions can propagate as free particles . in the case of a single domain wall ,
the dimensional reduction of fermions from three to two dimensions is straightforward . in the next subsection
we will construct a brane world consisting of a wall - anti - wall pair . in that case
, the mechanism of dimensional reduction is more subtle .
as we have seen , left - handed fermions can be localized on a domain wall .
similarly , right - handed fermions can be localized on an anti - wall .
a world that contains both left- and right - handed fermions should hence contain a wall - anti - wall pair . in order to keep the fermions light ,
the wall and the anti - wall must be separated in the extra dimension by a sufficient distance .
if the low - energy physics of the dimensionally reduced theory happens simultaneously on the wall and the anti - wall , the question arises how the brane separation @xmath33 compares with the natural length scale @xmath3 on the branes .
interestingly , in analogy to the mechanism of dimensional reduction from the symmetric phase that was discussed in section 4 , we will find that , as @xmath33 increases , the length scale @xmath3 grows exponentially such that the brane separation becomes small in the natural physical units of a low - energy observer .
again inspired by the corresponding soliton solution of the 2-d gross - neveu model , we now make the ansatz @xmath101.\ ] ] as before , we need to solve the dirac equation in the given background field .
remarkably , this is still possible in closed form .
the following relations are important for the derivation of the results presented below @xmath102[1 - a \tanh(y + c ) ] = 1 - a^2 , \nonumber \\ & & \frac{\cosh(y + c)}{\cosh(y - c ) } = \frac{1}{\sqrt{1 - a^2 } } [ 1 + a \tanh(y - c ) ] , \nonumber \\ & & \frac{\cosh(y - c)}{\cosh(y + c ) } = \frac{1}{\sqrt{1 - a^2 } } [ 1 - a \tanh(y + c ) ] , \nonumber \\ & & \frac{a}{\cosh(y + c ) \cosh(y - c ) } = \sqrt{1 - a^2}[\tanh(y + c ) - \tanh(y - c ) ] .
\end{aligned}\ ] ] here we have put @xmath103 . the dirac equation , as well as the above relations , are satisfied only if @xmath104 the parameter @xmath105 $ ] determines the brane separation @xmath33 and will later be fixed self - consistently . at zero distance
we have @xmath106 and hence @xmath107 , so that we are in one of the two vacua of the 3-d theory .
as @xmath108 approaches 1 , on the other hand , @xmath33 goes to infinity .
then @xmath109 and we are in the other vacuum state of the 3-d theory .
hence , by varying @xmath108 we can change the brane separation and interpolate smoothly between the two vacua . in the wall - anti - wall case one
obtains the states @xmath110 localized on the branes . here
@xmath111 is the mass of the particles propagating on the branes and @xmath112 is the corresponding energy .
these states describe a dirac fermion of momentum @xmath87 whose left - handed component propagates on the wall , while its right - handed component propagates on the anti - wall .
the wall - anti - wall profile , as well as the wave function for the left- and right - handed components of a localized state at rest , are depicted in figure 6 . for a particle moving to the left at high speed one
has @xmath113 .
then the upper component of the wave function , which is localized on the anti - wall , tends to zero , and the particle is almost entirely located on the wall .
this is illustrated in figure 7 .
the wave function of a fast right - mover , on the other hand , is concentrated on the anti - wall . at @xmath106 ,
i.e. at distance @xmath114 between the branes , we end up in the vacuum @xmath115 and the mass of the dirac fermion is simply given by the 3-d fermion mass @xmath116 .
when the branes are separated by a large distance @xmath117 , i.e. for @xmath118 , on the other hand , the mass of the dirac fermion goes to zero as @xmath119 consequently , when the branes are separated at a large distance @xmath33 , the 2-d correlation length @xmath120 again grows exponentially with @xmath33 and the system dimensionally reduces . in particular , from the point of view of a low - energy 2-d observer , the left- and right - handed modes propagating simultaneously on the wall and the anti - wall are indistinguishable from an ordinary point - like dirac fermion . at the same time , a 3-d high - energy observer whose natural length scale is @xmath121 would say that the dirac particle consists of a left - handed and a right - handed constituent , both separated by a large distance @xmath117 . in the wall - anti - wall configuration
the bulk states take the form @xmath122 \\ - ( \phi_0 + i k_3)[a
\phi_0 \tanh(a \phi_0 x_3 + c ) - i k_3 ] \end{array } \right),\ ] ] with the energy eigenvalues @xmath91 . we still need to verify the self - consistency of the wall - anti - wall profile . as before
, we need to sum @xmath78 over all occupied states .
first of all , we fill all bulk states with negative energy , and we obtain @xmath123 - 1 \right ) = \nonumber \\ & & - \frac{1}{e } \phi(x_3 ) - \frac{1}{e } \frac{a \phi_0 m^2}{k_3 ^ 2 + a^2 \phi_0 ^ 2 } [ \tanh(a \phi_0 x_3 + c ) - \tanh(a \phi_0 x_3 - c)].\end{aligned}\ ] ] as in the case of a single wall , the first term , @xmath124 , contributes @xmath125 which is just what we need for self - consistency .
however , in the wall - anti - wall case , there is also the second term . performing the @xmath126-integration
, that term contributes @xmath127 which needs to be canceled by an appropriate contribution from the states localized on the branes . for these states
one obtains @xmath128.\ ] ] first , let us fill all negative energy states localized on the branes .
those have energy @xmath129 and exactly cancel the contribution @xmath32 in eq.([contribution ] ) . in order to cancel the @xmath130-term in eq.([contribution ] ) as well , we also need to fill some positive energy states ( or , equivalently , empty some negative energy states ) . hence ,
in contrast to the single wall case , the wall - anti - wall configuration is unstable if the 2-d brane world is in its vacuum state . in order to stabilize the brane configuration ,
we occupy all positive energy states localized on the wall ( with energy @xmath131 ) up to some fermi momentum @xmath132 .
the cancellation condition , which determines the value of @xmath132 , then takes the form @xmath133 consequently , the energy of the particles at the fermi - surface @xmath134 is equal to the lowest energy @xmath58 of states propagating in the 3-d bulk of the extra dimension .
any fermion that one adds to the brane world has enough energy to escape into the extra dimension .
thus , our brane world which indeed contains naturally light fermions is necessarily completely packed with these particles .
hence , any potentially interesting physics of the light fermions is entirely pauli - blocked , and this `` world '' can only exist in one ( physically quite uninteresting ) state .
this is in contrast to the single - wall case , where any fermion occupation of the localized states was self - consistent with the wall profile .
it should be noted that we have assumed the configuration to be constant in the spatial 1-direction along the brane .
the results of ref .
@xcite show that , at non - zero fermion density , translation invariance may break by the spontaneous formation of a crystal lattice .
if this happens here as well , it is still possible that our brane world displays some non - trivial physics .
this could be investigated along the lines of ref .
@xcite .
as we have seen , in the wall - anti - wall case a light fermion mode is localized on the branes . in this subsection
we raise the question about the dynamical origin of the fermion mass .
in particular , we ask if it arises as a consequence of chiral symmetry breaking in the dimensionally reduced theory .
let us first discuss the single wall case .
obviously , the wall configuration arises as a consequence of the broken @xmath6 symmetry in the 3-d bulk .
this @xmath6 symmetry is a combination of a 2-d chiral transformation and a reflection about a 2-d plane perpendicular to the 3-direction .
it is interesting to note that the wall profile @xmath135 is invariant against the transformation @xmath136 which combines a 2-d chiral transformation with a reflection at the domain wall center . together with the transformation of eq.([symmetry3 ] ) , this particular @xmath6 symmetry indeed remains intact even in the presence of the domain wall . from the point of view of a 2-d observer living on the brane , this @xmath6 symmetry is nothing but 2-d chiral symmetry . as a consequence of this unbroken @xmath6 chiral symmetry , the fermion localized on the brane is massless .
next we discuss the wall - anti - wall case in which the wall profile is no longer invariant under the transformation of eq.([trans ] ) . as a consequence , there is no exact @xmath6 chiral symmetry from the point of view of the 2-d brane world .
still , as the wall and the anti - wall are separated by a large distance @xmath33 , an approximate chiral symmetry emerges which turns into an exact symmetry at infinite brane separation .
the fermion mass @xmath137 decreases exponentially with @xmath33 .
is this mass due to explicit or spontaneous breaking of the emerging approximate @xmath6 symmetry ? in order to clarify this question , we now compute the value of the chiral order parameter of the dimensionally reduced 2-d theory on the brane . first , we define an effective 2-d fermion field @xmath138 by smearing the 3-d fermion field with the wave function of the modes localized on the right or left side of the brane world . note that the 2-d fermion field , defined in this way , indeed has the correct dimension .
next , we calculate the 2-d chiral condensate @xmath16 . it is straightforward to show that only the modes localized on the branes contribute to this quantity .
the contribution takes the form @xmath139 where again @xmath140 and @xmath141 .
the domain wall height @xmath58 serves as a natural cut - off for the physics in the brane world .
hence , we occupy all states with negative energies @xmath142 , and ( for @xmath143 ) we obtain @xmath144 using eq.([massgap ] ) we identify @xmath145 as the effective 2-d coupling constant , and we then indeed obtain @xmath146 just as in the 2-d gross - neveu model . as we have seen in the previous subsection , for reasons of consistency , our brane world can not exist in its vacuum state .
thus , the actual ( non - vacuum ) value of the chiral condensate also receives contributions from states with positive energies @xmath147 and , in fact , even vanishes .
still , since for large @xmath33 the vacuum value of the chiral condensate @xmath25 agrees with the dynamically generated fermion mass @xmath1 , we conclude that this mass actually results from the spontaneous breakdown of the emergent @xmath6 chiral symmetry .
let us consider the question of stability of the brane world . obviously , unlike the single wall , the wall - anti - wall configuration is not topologically stable .
for example , one might worry that the wall and the anti - wall attract each other and annihilate through a continuous deformation of the profile @xmath76 into the trivial vacuum @xmath148 .
the energy stored in the brane tension would then be released and turned into heavy 3-d fermions .
such a catastrophic event would clearly be the end of our brane world .
alternatively , if the wall and the anti - wall repel each other , they would simply drift apart in the extra dimension .
remarkably , in the gross - neveu model the wall - anti - wall configuration is stable against both annihilation and drifting apart .
what is the dynamical mechanism responsible for the stability ?
the key observation is that the wall and the anti - wall themselves consist of fermions what else could they possibly be made of in this model ? in the wall - anti - wall case ,
the fermion density is given by @xmath149 let us also calculate the brane tension @xmath150 , i.e. its energy per unit length .
the brane tension receives contributions from three different sources : the filled bulk states , the filled surface states localized on the branes , and the @xmath151 term .
the latter contributes @xmath152 = - \frac{2 n a \phi_0}{g},\ ] ] while the surface states localized on the branes yield @xmath153 the calculation of the fermionic contribution of the bulk states requires special care .
following refs .
@xcite , we impose periodic boundary conditions in the 3-direction over a finite extent @xmath154 .
note that , in contrast to a single wall , the wall - anti - wall configuration is indeed consistent with periodic boundary conditions . in the finite box
only discrete values for @xmath155 are allowed .
these values are determined by the scattering phase shift @xmath156 as @xmath157 with @xmath158 . for sufficiently large @xmath154 ( such that @xmath159 ) the @xmath155 values are given by @xmath160 in the next step we sum the energy differences between occupied bulk states in the wall - anti - wall and the vacuum configuration @xmath161 \nonumber \\ & = & - \frac{n}{2 \pi } \int dk_1 \sum_n \left(k_{3,n } - \frac{2 \pi n}{l}\right ) \frac{d\sqrt{k_1 ^ 2 + k_3 ^ 2 + \phi_0 ^ 2}}{dk_3}|_{k_3 = 2 \pi
n / l } \nonumber \\ & = & \frac{n}{2 \pi } \int dk_1 \ \frac{1}{l } \sum_n \delta(\frac{2 \pi n}{l } ) \frac{d\sqrt{k_1 ^ 2 + k_3 ^ 2 + \phi_0 ^ 2}}{dk_3}|_{k_3 = 2 \pi
n / l } \nonumber \\ & = & \frac{n}{(2 \pi)^2 } \int dk_1 \int dk_3 \
\delta(k_3 ) \frac{d\sqrt{k_1 ^ 2 + k_3 ^ 2 + \phi_0 ^ 2}}{dk_3}.\end{aligned}\ ] ] in these manipulations , we have taylor expanded around @xmath162 and , at the end , we have taken the @xmath163 limit .
performing a partial integration , using @xmath164 and summing up all contributions to the brane tension , one finally obtains @xmath165 remarkably , the energy per fermion , @xmath166 , is independent of the separation of the walls and equals the mass of a fermion in the 3-d bulk .
consequently , given a fixed number of fermions , they can be divided arbitrarily into some that form the brane world and others that remain at rest in the 3-d bulk .
this explains how our brane world is protected from the dooms - day scenario of wall - anti - wall annihilation .
the branes can accrete fermions that fall onto them , coming from the bulk of the extra dimension .
this process increases the brane separation @xmath33 as well as the fermi - momentum @xmath132 , but it also decreases the fermion mass @xmath1 , such that the energy of the states at the fermi - surface @xmath58 remains fixed .
the thickness of the world @xmath33 increases only logarithmically with the fermion density . at infinite brane separation , @xmath59 , i.e. for @xmath118 , the fermion density and the energy density approach two times the corresponding values of a single wall brane world that is completely packed with fermions .
the brane tension for this object is given by @xmath167 .
interestingly , once the wall and the anti - wall are separated by an infinite distance , they can support configurations with any value of the fermi - momentum @xmath168 .
the corresponding brane tension is then given by @xmath169 for instance , an empty single - wall brane world ( in its vacuum state ) costs an energy @xmath170 per unit length .
as we discussed in the introduction , light fermions can arise naturally from higher dimensions as states localized on domain walls .
following the ideas behind d - theory , in contrast to standard applications of domain wall fermions in lattice field theory , we have taken the extra dimension physically seriously .
in particular , we have maintained locality in the bulk of the extra dimension , and we have used dimensional reduction to make the extra dimension invisible to a low - energy observer . the 3-d gross - neveu model at large @xmath0 has been used as a toy model in which one can obtain analytical insight into these phenomena . in particular , the gross - neveu model with a discrete @xmath6 chiral symmetry
provides domain walls dynamically through spontaneous chiral symmetry breaking .
interestingly , fermionic zero - modes can be localized on these walls which themselves consist of fermions .
an exact analytic solution was found for a stable wall - anti - wall configuration , and indeed light fermions arise naturally via dimensional reduction in this brane world . remarkably , the wall - anti - wall configuration is stable ( although not topologically stable ) due to the baryon asymmetry of the fermions that the brane is made of .
this mechanism of brane stabilization may be interesting for realistic brane world constructions .
ironically , in our toy model the `` world '' is stable only if the baryon asymmetry is so large that all fermion states are occupied and all physics is completely pauli - blocked .
we take this as a lesson for brane world model building .
our toy `` world '' was free to follow its own dynamics , without us making any ad - hoc assumptions about where branes should be located .
perhaps not surprisingly , the `` world '' then behaved in some but not in all respects as its builders had intended .
we would like to thank s. chandrasekharan , n. graham , p. hasenfratz , and r. l. jaffe for very interesting discussions .
thanks the universitt bern for kind hospitality during several visits .
this work was supported in part by funds provided by the schweizerischer nationalfond ( snf ) , by the european community s human potential programme under contract hprn - ct-2000 - 00145 , and by the dfg sonderforschungsbereich transregio 9 , `` computergesttzte theoretische teilchenphysik '' .
s. chandrasekharan and u .- j .
wiese , nucl .
b492 ( 1997 ) 455 ; + r. c. brower , s. chandrasekharan , and u .- j .
wiese , phys .
d60 ( 1999 ) 094502 ; + u .- j .
wiese , nucl .
b ( proc . suppl . ) 73 ( 1999 ) 146 . | we study the dimensional reduction of fermions , both in the symmetric and in the broken phase of the 3-d gross - neveu model at large @xmath0 . in particular ,
in the broken phase we construct an exact solution for a stable brane world consisting of a domain wall and an anti - wall . a left - handed 2-d fermion localized on the domain wall and a right - handed fermion localized on the anti - wall
communicate with each other through the 3-d bulk . in this way they are bound together to form a dirac fermion of mass @xmath1 . as a consequence of asymptotic freedom of the 2-d gross - neveu model , the 2-d correlation length @xmath2 increases exponentially with the brane separation .
hence , from the low - energy point of view of a 2-d observer , the separation of the branes appears very small and the world becomes indistinguishable from a 2-d space - time .
our toy model provides a mechanism for brane stabilization : branes made of fermions may be stable due to their baryon asymmetry .
ironically , our brane world is stable only if it has an extreme baryon asymmetry with all states in this `` world '' being completely filled .
addtoresetequationsection preprint hu - ep-03/58 , sfb - cpp-03 - 40 |
cosmogony is defined as a study of the origin ( cosmogenesis ) of the universe in the physical aspect .
it asks the question how the universe came into being . by comparing this process with human life , we called the birth of the universe cosmogenesis @xcite . among different conceptions of cosmogenesis which offer the possibility of describing the origin of the universe as a physical process
, there are cosmogenic theories of the spontaneous creation from vacuum instability . according to these theories ,
the universe was created in a spontaneous way , from a quantum fluctuation .
edward tryon @xcite is usually considered as the scientist who first supported such theories in 1973 by arguing that total mechanical energy of the closed universe is zero and effects of quantum fluctuation of the vacuum were important when the universe was coming to the existence .
the second approach developed independently by fomin @xcite in 1973 was based on the idea that the closed universe originated from gravitational instability of the vacuum is realized in the background of general relativity theory , rather than on newtonian one , like in tryon s paper .
fomin s contribution is only addressed to in a footnote in vilenkin s paper on quantum tunneling @xcite .
the main aim of this paper is to show that fomin s idea offers deeper understanding of the origin of the universe from the vacuum fluctuation for two reasons : the mechanism proposed by fomin is based on einstein general relativity in contrast to tryon s idea formulated on the background of the newtonian theory of gravity .
moreover , tryon s idea can not concern the curved closed universe .
nevertheless , the newtonian theory can be useful in the context of discovery , but of course , it does not describe the early epoch of the universe when both gravitational and quantum effects plays a crucial role .
we can discover a new contemporary context for fomin s theory by considering the cosmological models in which his idea is realized , providing the conservation condition is postulated .
we find two basic categories of the cosmological models in which the universe is generally relativistic and originated from quantum fluctuation via fomin s proposition .
we also demonstrate that in contrast to tryon s model , the models obtained in this paper can be confronted with modern cosmological observations of the distant supernovae snia .
the critical analysis of tryon s paper is not the subject of the present paper .
some interesting remarks and critiques of tryon s idea can be found in mccabe s papers @xcite .
we share mccabe s comments on tryon s paper in many points but the correct definition of the energy seems to be crucial for the investigation of the creation of the universe .
many authors have concluded recently that energy of the flat and closed frw universes are equal to zero locally and globally .
unfortunately such conclusions originate from coordinate dependent calculations which are performed in the special comoving coordinates ( called the cartesian coordinates ) , by using energy momentum tensor of matter and gravity @xcite . however , let us note that all expressions for the conserved energy - momentum tensor in general relativity can be written as a divergence of a superpotential .
the integral over space can then be expressed as a surface integral over the boundary , which is zero for the closed universe .
this conclusion does not depend on which metric or which coordinate system is used the organization of this paper is as follows . in section 2
we present fomin s idea in details .
section 3 is devoted to the construction of cosmological models from basic fomin s assumptions first principles . in section 4
these cosmological models are tested by the data from observation of distant supernovae type ia , which offer the possibility of explaining the present acceleration of the universe .
fomin s main goal ( like tryon s ) was to describe the origin of the universe as a whole without violation of any conservation laws . since at that moment gravitational fields were very strong , the gravitational interaction of the primordial vacuum with gravity should be taken into account . in both tryon s and fomin s papers ,
the vacuum is chosen as an initial state of the universe ( a metagalaxy in fomin s terminology ) .
note that if strong gravitational fields are important , then effects of general relativity ( not special relativity ) should be included at the very beginning .
this is the very reason why fomin formulated the problem in the context of general relativity . in the introduction to his paper
we can find an analogy to the newtonian cosmology , but he noted the absence of the kinetic term @xmath0 in the energy balance equation @xmath1 where @xmath2 is the mass of the universe particle and @xmath3 is the potential energy .
it would be worth mentioning at this point that the notion of energy for the frw models is well defined from the physical point of view because we are not dealing with isolated systems which are asymptotically flat spacetimes .
note that formally , from pirani s or komar s formula we can obtain zero energy for the friedmann - robertson - walker ( frw ) models but there is no time killing vector and we have geodesic congruence of privileged observers .
that is why , in our opinion , this result does not make much physical sense .
moreover , the result of energy calculations following pirani s or komar s formula depends on the choice of a coordinate system . in the introduction
to fomin s paper we find remarks that the process of quantum cosmogenesis should be considered in the presence of a strong gravitational field and therefore the effects of general relativity should be included at the very beginning .
in contrast to tryon , fomin considered an evolving non - static universe .
it is well known that the frw dynamics can be represented by the notion of a particle of unit mass moving in the potential , following an equation analogous to the newtonian equation of motion , namely [ eq:1 ] @xmath4 where @xmath5 and @xmath6 are generalized momenta and coordinates , respectively , @xmath7 is the hamiltonian and @xmath8 is the potential function of the scale factor @xmath6 , if @xmath8 is a decreasing function of @xmath6 , the universe is accelerating @xcite . system ( [ eq:1 ] ) should be considered on the energy level @xmath9 where @xmath10 is the curvature constant ( we use here and in next sections the system of units in which @xmath11 .
therefore , only for a flat relativistic system we have the energy level @xmath12 equals to zero .
it is interesting that the frw model based on fomin s conception can be formulated in terms of hamiltonian systems of a newtonian type in the form ( [ eq:1 ] ) with the corresponding function of the scale factor @xmath13 . in fomin s scenario ,
the universe emerged from a physical state , which he called the `` null vacuum '' system @xmath14 .
vacuum is assumed as an initial state of further evolution of the universe .
this system produced particles and antiparticles in a spontaneous manner without violating any conservation laws .
then emerging matter of mass @xmath15 , ( called bimatter by fomin ) , gives rise to the growth of entropy .
fomin argued that such a process of decaying vacuum is energetically preferred and both entropy and the number of particles and antiparticles will be growing at the cosmological time .
the growth of the number of particles is strictly related to the size of the system , as we assume that the created universe is closed ( see fig .
[ fig:1 ] ) . from vacuum
@xmath14 ( the idea of picture comes from fomin s papers @xcite).,scaledwidth=80.0% ] fomin pointed out that for the description of the contribution of decaying vacuum ( in a phenomenological way ) , the einstein field equations should be generalized .
hence the vacuum is characterized by an energy momentum tensor of the form @xmath16 where the energy density of vacuum @xmath17 depends on the scalar curvature of the spacetime @xmath18 where @xmath19 and @xmath20 is a dimensionless parameter . in fomin s
paper @xmath17 was assumed to be positive ( however we will not restrict it in the further sections ) .
there are some physical reasons for such a choice of the parameterization by considering a heuristic argument based on the conjecture about quantum gravity .
let us suppose that @xmath21 raises due to a quantum gravity fluctuation .
then the action @xmath22 might be nonzero due to quantum @xmath23 .
so @xmath24 @xcite .
therefore @xmath25 due to energy - time uncertainty relation @xmath26 .
let us assume that a quantum process proposed by fomin can be modelled , in a phenomenological way , by the energy momentum tensor in the form ( [ eq:2 ] ) and when the source of gravity is a perfect fluid with energy density @xmath27 and pressure @xmath28 .
therefore the einstein equation assumes the form @xmath29 where @xmath30 and energy momentum tensor is @xmath31 @xmath32 where @xmath33 is a four - vector of velocity .
the form of the einstein equation ( [ eq:4 ] ) is more suitable for our aims because the contribution from the energy density of the vacuum was shifted to the energy momentum tensor .
it is clear from divergence of the field equations @xmath34_{;\nu } = 0.\ ] ] hence the energy - momentum tensor is given in the diagonal form @xmath35 where @xmath36 , @xmath37 and @xmath38 , @xmath39 , @xmath40 which is obtained following fomin s assumption from the basic equation @xmath41 with the natural system of units @xmath42 .
since we are going to construct the cosmological model , as the first approximation , we consider the model , which is spatially homogeneous and isotropic , therefore its spacetime metric is given in the form @xmath43\ ] ] where @xmath44 is the cosmological time ; @xmath45 is a scale factor ; @xmath46 , @xmath47 , @xmath48 are standard spherical coordinates ; @xmath49 , @xmath50 , because it is assumed a positive value of curvature spatial slices of constant time of hypersurfaces . in fomin s
original paper , the evolution of the spacetime is governed by a modified einstein equation rather than the modified energy - momentum tensor ( [ eq:7 ] ) .
in our extension of fomin s model we shift the contribution coming vacuum instability to the energy - momentum tensor . while both formulations are formally equivalent ( gives the same formula @xmath51 ) , the presented approach offers the possibility of adding the conservation condition to the basic equation , which makes the model in a closed form . due to homogeneity and isotropy
we have @xmath52 and dynamics is equivalent to the dynamics of frw cosmological models with some effective energy density @xmath53 and pressure @xmath54 @xmath55 note that if we treat @xmath56 as an energy density of dark energy then @xmath57 this means that so - called coincidence problem ( i.e. why dark matter and dark energy are of the same order today ) is solved naturally in the framework of the fomin model .
if we require @xmath58 then it is satisfied for @xmath59 .
therefore if we choose the form of the equation of state for matter @xmath60 then one can obtain effective values of pressure and energy density . as a result
we obtain that the fomin model belongs to a category of kinematical model with a dynamic value of cosmological constant . obviously , a solution with dynamic cosmological constant is possible only if @xmath61 ( and @xmath62 ) . in the absence of matter @xmath63 ,
@xmath17 has got to remain a constant .
@xcite denoted that this category models which are invoked to solve a cosmological constant problem are in fact consistent with mach s ideas because an empty spacetime can not be a solution of general relativity with the dynamic cosmological constant ( @xmath64 in our case ) .
it is a consequence of the fact that a conserved quantity is of usual matter and vacuum ( and not these two separately ) following @xmath65 which reduces to the following equation @xmath66 or @xmath67 + \left [ \frac{d}{dt } \left ( \lambda a^{3 } \right ) - \lambda \frac{d}{dt } a^{3 } \right ] = 0.\ ] ] among the dynamical models of the cosmological constant ( for review see @xcite ) there are different choices of cosmological constant parameterization . in the fomin model the dependence of @xmath68 is determined by the form of the equation of state @xmath60 through equation ( [ eq:14 ] ) .
some different constraints on a decaying cosmological term from the astronomical observation have been found , e.g. @xcite . as a consequence the fomin model with the decaying term @xmath69
can be tested through the different astronomical observations like distant type ia supernovae data and wilkinson microwave anisotropy probe ( wmap ) data as well as compared with the concordance @xmath70cdm model .
one can say that the fomin model is a prototype of the model that incorporates a cosmic time variation of the cosmological term which plays a crucial role in the mechanism of creation of the universe from the vacuum . if @xmath69 is positive then @xmath71 .
the evolution of the universe is governed by two basic equations which constitute the closed system of basic equations for the frw cosmology ( with incorporated fomin s mechanism of gravitational instability of vacuum ) when we postulate dependence of energy density and pressure on the scalar factor ( @xmath72 ) @xmath73 and the conservation condition ( [ eq:14 ] ) . to integrate
( [ eq:14 ] ) and ( [ eq:16 ] ) the form of @xmath74 should be postulated .
hereafter for simplicity of presentation we assume that @xmath75 where @xmath76 .
of course equation ( [ eq:16 ] ) possesses the first integral ( called the friedmann first integral ) @xmath77 where @xmath78 is a curvature constant .
one can simply check this by differentiation of both sides of equation ( [ eq:17 ] ) with respect of cosmological time and substitution ( [ eq:17 ] ) .
the model under consideration contains a new term in comparison with the standard frw perfect fluid cosmology which is related to the existence of the parameter @xmath20 not equal zero .
this parameter measures the effectiveness of the process of creation of the universe from the null system .
note that in the radiation epoch , when @xmath79 , the effects of gravitational instability of vacuum vanish .
we assume following fomin that quantum effects of creation of the universe are manifested at the epoch in which @xmath80 is positive ( this condition is distinguished from fomin s condition of positiveness of @xmath69 . therefore , if @xmath81 , this guarantees the positiveness of @xmath68 .
fomin pointed out that effect of dissociation of vacuum originates in the nonhomogeneous region where @xmath82 . during evolution of the universe ,
the effects of vacuum instability stay small because of local expansion of the space . with the assumption ( [ eq:15 ] ) the solution of equation ( [ eq:16 ] ) takes the form @xmath83 from ( [ eq:14 ] ) we obtain as a solution @xmath84 ( in the special case of dust matter we obtain @xmath85 . ) hence we determine the equation of state coefficient @xmath86 we obtain from formula ( [ eq:21 ] ) that the universe is accelerating for dust filled matter if the parameter @xmath20 belongs to the interval @xmath87 ; @xmath88 . for our further analysis of constraints from snia observational data it is useful to represent an evolutional scenario of models in terms of the hubble function and dimensionless density parameters .
for the two distinguished cases we obtain @xmath89 where @xmath90 ; @xmath91 is the present value of the scale factor and @xmath92 where @xmath93 are the density parameters for the @xmath94-th component of the fluid . of course @xmath95 from condition @xmath96 . in the special case of dust matter
we obtain the useful relation in the context of constraining the model parameters against the astronomical data @xmath97 where @xmath98 .
we obtain that universe is still accelerating if @xmath99 and @xmath100 , i.e. @xmath101 for dust it means that that @xmath102 i.e. that @xmath103 . therefore the fomin model predicts acceleration of the universe driven by the mechanism of origin of the universe from gravitational instability of vacuum . in the next section we concentrate on the constraints from recent snia measurements of the effects arising from possible fomin s quantum creation mechanism which can be phenomenologically manifested by dark energy through a modified friedmann equation .
today cosmology appears to be one of the fastest growing parts of physics due to new experiments from measurements of the cosmic microwave background radiation to measurements of the apparent magnitudes of several high redshift supernovae of type ia , published recently by riess et al .
for distant supernovae , one can directly observe the apparent magnitude ( i.e. , log of flux @xmath104 ) and its redshift .
as the absolute magnitude @xmath2 of any supernovae is related to the present luminosity @xmath105 , then relation @xmath106 can be written as @xmath107 usually instead of @xmath108 dimensionless quantity @xmath109 is used , and then equation ( [ eq:23 ] ) changes to @xmath110 where the parameter @xmath111 is related to @xmath2 by the relation @xmath112 we know the absolute magnitude of snia from its light curve .
therefore we can obtain @xmath108 for these supernovae as a function of redshift because the apparent magnitude @xmath2 can be determined from low @xmath113 apparent magnitude . finally , it is possible to probe dark energy which constitutes main contribution to the matter content from @xmath114 . in our further analysis of snia data , we estimate models with value of @xmath115 determined from low redshift relation ( @xmath114 is then linear ) without any prior assumption on @xmath116 .
we use the standard statistical approach to obtain the best fitting model minimizing the @xmath117 function .
this analysis is supplemented by the maximum likelihood method to find confidence ranges for the estimated model parameters .
it is assumed that supernovae measurements came with uncorrelated gaussian errors and the likelihood function @xmath118 could be determined from the @xmath119 statistic @xmath120 . in our analysis the latest compilation of snia prepared by riess et al .
is used ( the gold sample ) @xcite .
recent measurements of type ia supernovae as well as other wmap and extragalactic observations , suggest that the expansion of the universe is an accelerating expansion phase .
there are different attempts to explain the present acceleration of the universe .
dark energy of unknown form has usually been invoked as the most feasible mechanism .
also effects arising from exotic physics , like extra dimensions , some modification of the frw equation can also mimic dark energy through a modified friedmann equation
. it would be interesting to check the consistency of fomin s model with snia observations .
we begin by evaluating the luminosity distance as a function of redshift @xmath113 as well as the parameters of the model .
we define the redshift dependence of @xmath7 as @xmath121 . for cardassian stylization of fomin s model of the closed universe with matter ( baryonic and cold matter )
we get that the luminosity distance is given by @xmath122 the standard measure of acceleration in cosmology is the dimensionless deceleration parameter @xmath123 defined as @xmath124 we can calculate the value of this parameter for both classes of models under consideration
. then we obtain @xmath125}.\ ] ] we estimate the two - parameter model in the following form @xmath126 where @xmath127 is positive and less than one , and the parameter @xmath128 is positive . the latter parameter is related to the searched parameter @xmath20 @xmath129 where we assume @xmath20 to be @xmath130 .
we estimate two cases where different regions for @xmath131 are assumed .
the first model corresponds to the case when the curvature of the space is undetermined for which @xmath132 .
the second model exactly corresponds to the original fomin model with the positive curvature assumed at the very beginning for which @xmath133 . in both cases
we assume that @xmath134 . using the snia union sample we estimate both parameters @xmath131 and @xmath20 to find the best fit +
for the first model @xmath135 and for the fomin model @xmath136 the expected value of estimated parameters with a @xmath137 confidence interval + for the first model @xmath138 and for the fomin model @xmath139 these posteriori probability distribution functions are shown in fig .
[ fig:2 ] for the first model and in fig .
[ fig:3 ] for the fomin model . and @xmath20 for the model with @xmath132 : left panel posteriori probability distribution functions ; right panel
posteriori probability confidence levels @xmath137 and @xmath140.,title="fig:",scaledwidth=45.0% ] and @xmath20 for the model with @xmath132 : left panel posteriori probability distribution functions ; right panel posteriori probability confidence levels @xmath137 and @xmath140.,title="fig:",scaledwidth=45.0% ] and @xmath20 for the model with @xmath141 : left panel posteriori probability distribution functions ; right panel posteriori probability confidence levels @xmath137 and @xmath140.,title="fig:",scaledwidth=45.0% ] and @xmath20 for the model with @xmath141 : left panel posteriori probability distribution functions ; right panel posteriori probability confidence levels @xmath137 and @xmath140.,title="fig:",scaledwidth=45.0% ] the comparison these two models with the @xmath70cdm model gives that there is the moderate evidence in favour of the @xmath70cdm model in both cases .
this conclusion is based on the bayes factors analysis .
for the first model with respect to the @xmath70cdm model we obtain the bayes factor @xmath142 . for the second model with respect to the @xmath70cdm model
we obtain the bayes factor @xmath143 .
our estimation shows that the fomin model is consistent with observations of snia .
the dynamics of the model can be represented with the help of a particle - like description , namely it can be reduced to the form of a particle of unit mass moving in a potential well .
the forms of potential function for dust matter and matter with @xmath144 models respectively are @xmath145 there are two different scenarios of an initial state of the universe namely a singularity or a bounce .
it depends on the exponent in the potential form ( [ eq:31 ] ) .
note that if @xmath146 is negative this term will dominate the curvature term and we have a standard singularity ( @xmath8 goes to minus infinity ) .
this happens when @xmath147 for dust matter it is satisfied if @xmath148 . in the opposite case
if @xmath149 we obtain a bounce as a scenario of an initial state . this scenario is the generic feature of cosmology inspired by quantum gravity effects @xcite .
note that for @xmath150 effects of decaying @xmath17 vanish and the model evolution follows the radiation dominating universe scenario .
the motion of the system is defined on the zero energy level @xmath151 . in the bouncing scenario the standard evolution with a singularity at @xmath152
is replaced by a bounce .
note that negative contribution to @xmath153 formula can not dominate the material term and the universe starts from some @xmath154 a finite value of the scale factor . on the other hand
, evolution can be prolonged into the domain @xmath155 ( a pre - bounce region ) because of the reflectional symmetry @xmath156 .
if @xmath157 , i.e. the universe is closed , then we obtain in the configuration space returning points . as a consequence
we obtain characteristic type of evolution without initial singularity .
let us illustrate this situation .
for the model with dust we have @xmath158 \leq 0.\ ] ] this function has always zero : @xmath159 and @xmath160 for @xmath161 and @xmath162 otherwise .
the derivative of @xmath13 determines a domain of acceleration @xmath163 therefore for any @xmath44 the scale factor acceleration @xmath164 takes place if @xmath165 i.e. , @xmath166 . in the opposite case v(a ) is a growing function of @xmath6 ( @xmath167 ) and @xmath168 . at @xmath169 @xmath170 because @xmath171 . if @xmath172 then @xmath173 and it is a bouncing point at @xmath174 . in the opposite case @xmath175 , i.e. there is a singularity at @xmath174 .
summing up , in the acceleration regime @xmath8 is a decreasing function of @xmath6 , @xmath166 , and there is a bouncing point at @xmath174 . in the opposite
, deceleration regime @xmath8 is an increasing function of @xmath6 , @xmath176 and there is an initial singularity and a maximum size of the universe @xmath174 .
the phase portrait of model for two cases is in fig .
[ fig:4 ] . in the literature
there are well known propositions of explanation of acceleration of the universe using cosmological models with modified friedmann equations .
freese and lewis @xcite showed that the cardassian modification of friedmann equation ( which contains a term of type @xmath177 , @xmath178 ) may give rise to acceleration without any dark energy of unknown origin contribution ( see also @xcite for the estimation of the model ) . in fomin s model the age of the universe in the approximation of small @xmath179
is given by formula @xmath180 therefore @xmath181 is required to explain the problem of age of the universe .
we reviewed , clarified , and critically analyzed early approaches to tryon s and fomin s description of the origin of the universe .
we showed that the idea of the universe s creation as a vacuum fluctuation event was discovered by them independently at around the same time . while tryon s idea seems to be interesting only from the historical point of view , fomin s approach is still heuristically fertile in description of latest issues in cosmology .
fomin s model brings the interesting explanation of the accelerating expansion of the universe in terms of the ricci scalar dark energy @xmath182 @xcite .
this idea is important in the context of holographic principle of quantum gravity theory which regards black holes as the maximally entropic objects . for a given region of size @xmath105
effective field theory with uv cut - off @xmath70 gives the bekenstein entropy bound @xmath183 entropy of a black hole .
in this paper we study fomin s idea of quantum cosmogenesis through the dynamical analysis .
it is possible , in contrast to tryon , because fomin s conception is embedded in the environment of general relativity rather than in the newtonian theory .
we demonstrated that it is possible to construct the dynamical cosmology with incorporation of fomin s mechanism of gravitational decaying of vacuum .
this cosmology can be treated as a cosmology with varying cosmological constant . in fomin model
appears the free parameter @xmath20 .
there are some interval of this parameter for which the model accelerates and possesses a bounce instead of an initial singularity .
it is necessary then to estimate it from the astronomical observations . as recent supernovae of type ia measurements
indicate the expansion of our universe is presently accelerating . while the cosmological constant offers the possibility of reconstruction of the effective theory of acceleration , the presence of fine - tuning difficulties motivates theorists to investigate alternative forms of dark energy .
all these propositions can be divided into two groups . in the first group ,
unusual properties of dark matter with negative pressure violating the strong energy condition are postulated . in the other one , the modification of the friedmann equation is postulated . in this approach , instead of a new hypothetical energy component of unknown form
, some modifications of the frw equations are proposed a priori ( unfortunately without any fundamental justification ) .
the status of fomin s cosmology seems to be similar to the status of the brane cosmological models where our observable universe is a surface or a brane embedded in a higher dimensional bulk spacetime in which gravity could spread .
note , that this model with radiation can be recovered if parameter @xmath20 is @xmath184 .
we point out that fomin s idea can be tested ( in contrast to tryon s idea ) by using type ia supernovae data .
it is also important that fomin s approach is free from the assumption of zero energy which is crucial for tryon .
recently many authors concluded that energy - momentum tensor of the frw universes are equal to zero locally and globally ( flat universe ) or just globally ( closed universe ) .
however , such a conclusion originated from coordinate - dependent calculations performed in the special comoving coordinates ( cartesian coordinates ) .
it is interesting that predictions of this theory can be also tested by both wilkinson microwave anisotropy probe ( wmap ) observations as well as by observations of distant supernovae type ia ( snia ) .
we confront the fomin frw model with snia observation and estimate all parameters of models .
our analysis showed that fomin s idea plays not only a historical role but is related to the modern cosmological models appeared in the context of dark energy .
the idea that sum of energies of all particles in the universe is the same order of magnitude like gravitational energy , and consequently the total energy vanishes is very old .
overduin and fahr @xcite argued that haas @xcite and jordan @xcite introduced this idea in 1936 and 1947 , respectively , and called it the haas - jordan idea .
the main difficulty in extending it to relativistic cosmology is caused by finding the correct definition of the energy in the frw universe .
the problem was not solve so far . on the other hand
it was proposed to use the newtonian framework to show this idea at work . but this approach basing on the concept of infinite @xmath185 space as a model of real space is inconsistent with homogeneous and isotropic static distribution of matter .
the frw general relativistic cosmological models can be represented as a motion of a fictitious particle moving in the potential well @xcite .
the shape of the potential function is determined by both matter content and curvature . in the simplest case
the potential function is a function of the scale factor only .
the balance energy relation @xmath186 corresponds to the hamiltonian energy constraint @xmath170 .
note that division on kinetic and potential parts of energy has purely conventional character .
the system under consideration has a natural lagrangian . from the potential function
the curvature term can be extracted in such a way that motion of the system is restricted to the energy level @xmath187 . in the newtonian cosmology
the kinetic part is zero and the solution is admissible only for a closed universe because the gravity is attracting .
hence the problem of extension of the haas - jordan idea for the general relativistic case seems to be crucial for deeper understanding of the cosmogenesis mechanism .
the weakness of the fomin model is that it does not specify what kinds of matter are created .
if you have several components , you get only the overall balance equation , but not how the created matter is divided between the components
. however the modern cosmological context of ricci scalar dark energy suggests that created energy may corresponds to dark energy component .
we obtain from observation that value of the basic parameter @xmath20 is @xmath188 and @xmath189 as the best fit .
this means that formally fomin model is equivalent to the original cardassian one with @xmath190 .
the value of density parameter of curvature is in good agreement with current cmb data @xcite .
we also calculated the pdfs and confidence level of two parameters of our analysis showed that in contrast to the tryon model the fomin model can be statistically estimated and in the light of snia data can explain accelerating phase of expansion of the current universe . from the philosophical point of view
it means that there is some class of models of quantum cosmogenesis which are susceptible to falsify .
falsificability of fomin s conception means that this theory becomes scientific in popper s sense @xcite .
the authors are very grateful to prof .
m. heller , dr a. krawiec and a. kurek for remarks and comments .
marek szydowski is especially very grateful to prof .
y. shtanov for information on a detailed publication history of fomin s paper and the fruitful discussion on quantum cosmogenesis .
the authors also thank prof .
j. overduin and prof .
a. vilenkin for comments and historical remarks .
this work was supported in part by the marie curie actions transfer of knowledge project cocos ( contract mtkd - ct-2004 - 517186 ) and polish state committee for scientific research ( kbn ) grant no . 1 h01a 019 30 . | the main aim of this paper is to extend the early approach to quantum cosmogenesis provided by fomin .
his approach was developed independently to the well - known tryon description of the creation of the closed universe as a process of quantum fluctuation of vacuum .
we apply the fomin concept to derive the cosmological observables .
we argue that fomin s idea from his 1973 work , in contrast to tryon s one has impact on the current universe models and the proposed extension of his theory now can be tested by distant supernovae snia .
fomin s idea of the creation of the universe is based on the intersection of two fundamental theories : general relativity and quantum field theory with the contemporary cosmological models with dark energy . as a result of comparison with contemporary approaches concerning dark energy
, we found out that fomin s idea appears in the context of the present acceleration of the universe explanation : cosmological models with decaying vacuum .
contemporary it appears in the form of ricci scalar dark energy connected with the holographic principle .
we show also that the fomin model admits the bounce instead of the initial singularity .
we demonstrate that the fomin model of cosmogenesis can be falsified and using snia data the values of model parameters is in agreement with observations . |
the multiple access channel ( mac ) is one of the most well - studied problems in network information theory @xcite .
the capacity region of the discrete memoryless mac was independently derived by ahlswede @xcite and liao @xcite in the early 1970s . in this paper , we are interested in the gaussian version of this problem for which the channel output @xmath1 corresponding to the inputs @xmath2 is @xmath3 where @xmath4 is standard gaussian noise . we assume an average transmission power constraint of @xmath5 corresponding to each transmitter @xmath6 .
the capacity region was derived by cover @xcite and wyner @xcite and is the set of all rate tuples @xmath7 that satisfy @xmath8 for all subsets @xmath9 . for the @xmath10 case , the pentagonal region of rate tuples in is known as the _ cover - wyner _ region and
is illustrated in figure [ fig : cap_reg ] . despite our seemingly complete understanding of fundamental limits of the gaussian mac ,
it is worth highlighting that in the above - mentioned seminal works @xcite , it is assumed that the average error probability tends to zero as the length of the code grows without bound .
this implies that those established converses are , in fact , _
weak converses_. fano s inequality ( * ? ? ?
2.1 ) is typically used as a key tool to establish such weak converses . in this work ,
we strengthen the results of cover @xcite and wyner @xcite and show that any rate tuple that can be supported by a sequence ( in the blocklength ) of gaussian multiple access codes with asymptotic average error probability _ strictly less than one _ ( and not necessarily tending to zero ) must lie in the cover - wyner region .
this is a _
strong converse _ statement , akin to the work on strong converses for point - to - point channels by wolfowitz @xcite .
it indicates that the boundary of the cover - wyner region designates a sharp phase transition of the smallest achievable asymptotic error probability , which is zero for any rate tuple inside the capacity region and one for any rate tuple outside the capacity region .
thus , this work augments our understanding of the first - order fundamental limit of the gaussian mac .
additionally , it may also serve as a stepping stone for studying the second - order asymptotics @xcite or upper bounds ( e.g. , the sphere - packing bound ) on the reliability function of the gaussian mac ( cf .
* th . 4 ) ) .
( 115 , 135 ) ( 10,60)(4,0)18(1,0)2 ( 60,10)(0,4)18(0,1)2 ( 0 , 10)(1 , 0)110 ( 10 , 0)(0,1)110 ( 110,8 ) @xmath11 ( -6 , 105 ) @xmath12 ( 53 , 0)@xmath13 ( 80 , 0)@xmath14 ( -5,55)@xmath15 ( -1,79)@xmath16 ( 35,35)@xmath17 ( 3,3)@xmath18 ( 10 , 80)(1 , 0)50 ( 80 , 10)(0,1)50 ( 60 , 80)(1 , -1)20 the study of macs has a long history and we refer the reader to the excellent exposition in el gamal and kim ( * ? ? ? * ch .
4 ) for a thorough discussion .
dueck @xcite proved the strong converse for the ( two - source ) discrete memoryless mac using the the technique of blowing up decoding sets originally due to ahlswede , gcs and krner @xcite , combined with a novel strategy known as the _ wringing technique_. the technique of blowing up decoding sets uses the so - called _ blowing - up lemma _
@xcite ( see also ( * ? ? ?
5 ) or ( * ? ? ?
* sec . 3.6 ) ) . this technique is useful for establishing strong converse results for memoryless channels with finite output alphabets .
dueck s proof proceeds in three steps .
first , dueck expurgates an appropriate subset of codeword pairs to convert any given sequence of codes with asymptotic _ average _ error probability bounded away from one to a sequence of codes with asymptotic _
maximal _ error probability bounded away from one .
this expurgation step is performed so that the blowing - up lemma to be applied in the third step yields tight upper bounds on the sum - rate , which will then lead to the desired strong converse result
. unfortunately , the expurgation step introduces undesirable correlations among the codewords transmitted by the @xmath0 encoders .
second , a wringing technique is introduced to wring out any residual dependence among the symbols transmitted by the @xmath0 encoders by choosing a further subcode from each subcode obtained in the expurgation step .
wringing is necessary for establishing a tight sum - rate bound , because the sum - rate capacity of the mac is expressed as the supremum of mutual information terms over all independent input distributions ( the independence is due to the fact that the @xmath0 encoders do not cooperate ) .
third , the blowing - up lemma is applied to the resultant subcode to yield a tight upper bound on the sum - rate .
ahlswede @xcite presented another strong converse proof for the ( two - source ) discrete memoryless mac by modifying dueck s wringing technique as well as replacing the use of the blowing - up lemma in dueck s proof with an application of augustin s non - asymptotic converse bound @xcite .
however , the proofs of dueck and ahlswede are specific to the discrete ( finite alphabet ) setting and it is not clear by examining the proofs that the same strong converse statement follows in a straightforward way for the gaussian mac with peak power constraints .
another approach to proving the strong converse for a general mac is due to han @xcite , who used the information spectrum technique @xcite to provide a general formula for macs and stated a condition ( * ? ?
6 ) for the strong converse to hold .
however , unlike for the point - to - point setting ( * ? ? ?
3.63.7 ) , the property is difficult to verify for various classes of memoryless macs .
in view of the above works and the practical and theoretical importance of strong converse theorems , we are motivated to provide a self - contained proof for the strong converse of the gaussian mac . in this subsection
, we discuss the challenges of leveraging existing techniques to prove the strong converse for the gaussian mac .
in particular , we highlight the difficulties in directly using the ideas contained in dueck s @xcite and ahlswede s @xcite proofs .
we also describe , at a high level , the strategy we employ to overcome these difficulties .
finally , we discuss some other auxiliary proof techniques . in dueck
s paper @xcite , he used a version of the _ blowing - up lemma _ , together with other tools , to prove the strong converse theorem for the discrete memoryless mac . a crucial step in dueck s proof
involves the establishing of an upper bound on the list size of possible messages for every output sequence based on the blown - up decoding sets .
if the resultant list size is too large ( e.g. , contains an exponential number of messages ) , the dueck s technique can not lead to the strong converse theorem .
since this crucial step heavily relies on the finiteness of the output alphabet and the output alphabet of the gaussian mac is uncountably infinite , it is not immediately apparent how to extend this step to the gaussian case .
as mentioned in the previous section , ahlswede s proof @xcite is based on a modification of dueck s wringing technique and augustin s non - asymptotic converse bound @xcite .
however , it is not apparent how to adapt his techniques to obtain a strong converse bound on the sum - rate .
more specifically , ahlswede s wringing technique ( see equation ( 5.3 ) in @xcite ) leads to the following sum - rate bound for any sequence of length-@xmath19 codes whose asymptotic average error probability is bounded away from one : @xmath20 in , @xmath21 and @xmath22 are _ independent _ random variables .
however , the bound in is sensitive to the sizes of the input and output alphabets , which prevents us from directly extending ahlswede s proof to the gaussian case .
furthermore , there are no cost constraints in the discrete memoryless mac and incorporating cost constraints does not seem to be trivial .
a nave strategy to extend ahlswede s proof to the gaussian case is to quantize the input and output alphabets of the gaussian mac so that @xmath23 , @xmath24 and @xmath25 depend on @xmath19 and their cardinalities grow with @xmath19 .
say we denote the quantized alphabets as @xmath26 , @xmath27 and @xmath28 .
this sequence of quantized alphabets and the corresponding channels will be designed to provide increasingly refined approximations to the gaussian mac as @xmath19 increases . in designing @xmath26 , @xmath27 and @xmath28
, we would also like to ensure that the power constraints are satisfied and the term @xmath29 in vanishes as @xmath19 tends to infinity .
however , quantization arguments that are used to prove information - theoretic statements for continuous - valued alphabets are usually applied to the achievability parts of coding theorems .
for example , a quantization argument is used in ( * ? ? ?
3.4.1 ) for leveraging the achievability proof for the discrete memoryless channel ( dmc ) with cost constraints to prove the achievability part of the capacity of the awgn channel .
to the best of our knowledge , standard quantization arguments for achievability parts do not work for strong converse proofs because upon quantization , one has to ensure that the resultant asymptotic error probability is bounded away from one .
the reader is also referred to ( * ? ? ?
* appendix d.6 ) for a complementary explanation of why ahlswede s original wringing technique works for only macs with finite alphabets but not the gaussian mac .
the difficulties in directly using dueck s and ahlswede s techniques led the authors to combine a novel quantization argument together with ahlswede s wringing idea .
we use a scalar quantizer of increasing precision in the blocklength to discretize ( only ) the input alphabets of the channel so that the ahlswede s wringing technique can be performed _ on the quantized channel inputs _ for any given code whose asymptotic error probability is bounded away from one . in doing so , we obtain a sequence of subcodes whose asymptotic error probability is bounded away from one such that the resultant correlations among the codeword symbols transmitted by the different sources vanish as @xmath19 increases . note that if the quantizer s precision is too small or too large , the resultant upper bound on the sum - rate will be too loose and hence not useful in proving the strong converse .
we discuss feasible choices of the quantizer s precision and the parameters used in the wringing technique in section [ sectiondiscussionquantizer ] . in our proof
, the quantizer s precision is chosen in such a way that the quantized input alphabets @xmath30 grow no faster than @xmath31 .
it turns out that this choice of quantization also allows us to control the approximation errors between the true channel inputs and the quantized ones uniformly . in ahlswede
s proof of the strong converse for the discrete memoryless mac , he appealed to a non - asymptotic converse bound by augustin @xcite . in our proof
we use a conceptually similar non - asymptotic converse bound that is motivated by modern techniques relating binary hypothesis testing to channel coding .
in particular , we use a form of the _ meta - converse _ ( * ? ? ? * sec .
iii - e ) due to wang , colbeck and renner ( * ? ? ?
* lemma 1 ) .
we derive a multi - user version of this non - asymptotic converse bound .
after doing so , we choose the auxiliary conditional output distributions therein to be product distributions that approximate the _ quantized _ code distribution .
we note that the flexibility of the choice of the output distributions is essential for proving the strong converse for the gaussian mac as we can allow these distributions to depend not only on the peak powers but also the chosen precision of the scalar quantizer ( cf .
section [ sec : remedy ] ) . in the next subsection , we state the notation used in this paper . in section [ sectiondefinition ]
, we describe the system model and define the @xmath32-capacity region of the gaussian mac . in section [ sec : main_res ] , we present the main result of the paper .
we present a few preliminaries for the proof in section [ sectionprelim ] .
the complete proof is then presented in section [ sec : prf_main_result ] .
section [ sectionic ] extends our strong converse result to the two - source two - destination gaussian ic under strong interference .
we use the upper case letter @xmath33 to denote an arbitrary ( discrete or continuous ) random variable with alphabet @xmath34 , and use a lower case letter @xmath35 to denote a realization of @xmath33 .
we use @xmath36 to denote the random tuple @xmath37 .
the following notations are used for any arbitrary random variables @xmath33 and @xmath1 and any mapping @xmath38 whose domain includes @xmath34 .
we let @xmath39 and @xmath40 denote the probability distribution of @xmath41 ( can be both discrete , both continuous or one discrete and one continuous ) and the conditional probability distribution of @xmath1 given @xmath33 respectively .
we let @xmath42 and @xmath43 be the evaluations of @xmath39 and @xmath40 respectively at @xmath44 . to avoid confusion , we do not write @xmath45 to represent @xmath42 unless @xmath33 and @xmath1 are both discrete . to make the dependence on the distribution explicit ,
we let @xmath46 denote @xmath47 for any real - valued function @xmath38 and any set @xmath48 .
the expectation and the variance of @xmath49 are denoted as @xmath50 $ ] and @xmath51=\e_{p_x}[(g(x)-\e_{p_x}[g(x)])^2]$ ] respectively , where we again make the dependence on the underlying distribution @xmath52 explicit .
we let @xmath53 denote the probability density function of a gaussian random variable whose mean and variance are @xmath54 and @xmath55 respectively .
this means that @xmath56 we will take all logarithms to base 2 throughout this paper .
the euclidean norm of a vector @xmath57 is denoted by @xmath58 .
we consider a gaussian mac that consists of @xmath0 sources and one destination .
let @xmath59 be the index set of the sources ( or encoders ) , and let @xmath60 denote the destination ( or decoder ) .
the @xmath0 message sources transmit information to the destination in @xmath19 time slots ( channel uses ) as follows . for each @xmath61 ,
node @xmath62 chooses message @xmath63 and sends @xmath64 to node @xmath60 where @xmath65 denotes the message size .
based on @xmath66 , each node @xmath62 prepares a codeword @xmath67 to be transmitted and @xmath68 should satisfy @xmath69 where @xmath5 denotes the power constraint for the codeword transmitted by node @xmath62 . then for each @xmath70 ,
each node @xmath62 transmits @xmath71 in time slot @xmath72 and node @xmath60 receives the real - valued symbol @xmath73 where @xmath74 are i.i.d .
and @xmath75 is a standard gaussian random variable .
after @xmath19 time slots , node @xmath60 declares @xmath76 to be the transmitted @xmath77 based on @xmath78 .
to simplify notation , we use the following convention for any @xmath79 . for any random tuple @xmath80 , we let @xmath81 be its subtuple , whose generic realization and alphabet are denoted by @xmath82 and @xmath83 respectively .
similarly , for any @xmath70 and any random tuple @xmath84 , we let @xmath85 be its subtuple , whose realization is denoted by @xmath86 .
the following five definitions formally define a gaussian mac and its capacity region .
[ defcode ] let @xmath87 be a non - empty subset in @xmath88 .
@xmath89-code _ for the gaussian mac , where @xmath90 and @xmath91 , consists of the following : 1 . a message set @xmath92 at node @xmath62 for each @xmath61 .
2 . a support set of the message tuple @xmath93 denoted by @xmath94 where @xmath93 is uniform on @xmath48 .
in addition , all the @xmath95 s in @xmath48 have the same @xmath96 , i.e. , there exists a @xmath97 such that for all @xmath98 , we have @xmath99 .
define @xmath100 to be the support of @xmath101 . consequently , the message tuple @xmath101 is uniform on @xmath102 .
an encoding function @xmath103 for each @xmath61 , where @xmath104 is the encoding function at node @xmath62 such that @xmath105 and @xmath106 for all @xmath107 .
the set of codewords @xmath108 is called the _
codebook for @xmath66_. for each @xmath109 , the finite alphabet @xmath110 is called the _ support _ of symbols transmitted by @xmath62 because @xmath111 .
note that @xmath112 for each @xmath61 by .
a ( possibly stochastic ) decoding function @xmath113 which is used by node @xmath60 to estimate the message tuple @xmath114 , i.e. , @xmath115 . if @xmath116 and @xmath117 , then @xmath93 is uniformly distributed on @xmath118 , which implies that the @xmath0 messages are mutually independent .
since @xmath119-codes are of our main interest , they are also called @xmath120-codes for notational convenience . however , in the present work , it is necessary to allow @xmath48 and @xmath87 to be strict subsets of @xmath118 and @xmath88 respectively so the generality afforded in the above definition is necessary . in this case , the @xmath0 messages need not be independent . in the rest of this paper ,
if we fix a code with encoding functions @xmath121 , then @xmath122 as defined in denotes the support of symbols transmitted by each @xmath109 .
[ defgaussianmac ] a _ gaussian mac _ is characterized by the conditional probability density function @xmath123 satisfying @xmath124 for all @xmath125 and all @xmath126 such that the following holds for any @xmath127-code : let @xmath128 be the probability distribution induced by the @xmath127-code .
then , @xmath129 for all @xmath130 where @xmath131 since @xmath132 does not depend on @xmath72 by and , the channel is stationary . for
any @xmath127-code defined on the gaussian mac , let @xmath133 be the joint distribution induced by the code .
since @xmath134 is a function of @xmath78 by definition [ defcode ] , it follows that @xmath135 which implies from that @xmath136 [ deferror ] for an @xmath137-code defined on the gaussian mac , we can calculate according to the _ average probability of decoding error _ which is defined as @xmath138 an @xmath137-code with average probability of decoding error no larger than @xmath32 is called an @xmath139-code . similarly for an @xmath127-code , we can calculate the _ maximal probability of decoding error _ defined as @xmath140 an @xmath127-code with maximal probability of decoding error no larger than @xmath32 is called an @xmath141-code .
[ defachievablerate ] a rate tuple @xmath142 is _
@xmath32-achievable _ for the gaussian mac if there exists a sequence of @xmath143-codes on the gaussian mac such that @xmath144 for each @xmath109 and @xmath145 [ defcapacityregion ] for each @xmath146 , the _ @xmath32-capacity region _ of the gaussian mac , denoted by @xmath147 , is the set consisting of all @xmath32-achievable rate tuples @xmath148 .
the _ capacity region _ is defined to be the @xmath18-capacity region @xmath149 .
the following theorem is the main result in this paper .
[ thmmainresult ] define @xmath150{2 in}{$\sum_{i\in t}r_i\le \frac{1}{2}\log\big(1+\sum_{i\in t}p_i\big)$ } \right.\right\}. \label{rout}\ ] ] then for each @xmath146 , @xmath151 we now present three remarks concerning theorem [ thmmainresult ] . 1 . note that @xmath152 is the cover - wyner @xcite region for an @xmath0-source gaussian mac .
the theorem says that regardless of the admissible average error probability ( as long as it is strictly smaller than @xmath153 ) , all achievable rate tuples must lie in @xmath152 .
since all rate tuples in @xmath152 are @xmath18-achievable ( * ? ? ?
4.7 ) , we have for every @xmath154 @xmath155 2 .
in fact , the proof allows us to additionally assert the following : for any non - vanishing average error probability @xmath154 and any subset @xmath79 , it can be shown that the sum - rate of the messages indexed by @xmath87 of any sequence of @xmath156-codes satisfying the constraint in also satisfies @xmath157 \le\overline{\upsilon}(\varepsilon , t , p_{\mathcal{i}})<\infty\label{eqn : sec_order}\ ] ] for some finite constant @xmath158 .
see in the proof of theorem [ thmmainresult ] . even though the normalizing speed of @xmath159 is not the desired @xmath160 ( as usually defined in second - order asymptotic analyses @xcite )
, the techniques in this work may serve as a stepping stone to establish an outer bound for the second - order coding rate region @xcite for the gaussian mac .
the best inner bound for the second - order coding rates for the gaussian mac was established independently by scarlett , martinez , and guilln i fbregas @xcite and molavianjazi and laneman @xcite . according to the inner bounds in @xcite and the relation between second - order coding rates and second - order asymptotics of sum - rates in @xcite , @xmath161 \ge\underline{\upsilon}(\varepsilon , t , p_{\mathcal{i } } ) > -\infty \label{eqn : sec_inner}\ ] ] for some finite constant @xmath162 .
our normalizing speed of @xmath163 in is slightly better than in ahlswede s work on the discrete memoryless mac @xcite , which is @xmath164 .
we have attempted to optimize ( reduce ) the exponent of the logarithm @xmath165 in the normalizing speed @xmath166 . however , as we will discuss in section [ sectiondiscussionquantizer ] in the sequel , we are unable to use our proof technique to further reduce ( improve ) @xmath167 from @xmath168 .
for both the discrete and gaussian macs , it is challenging to prove that the exact normalizing speed of the second - order term is @xmath169 .
this is , in part , due to the use of wringing technique in the converse part , which prevents one from obtaining a converse that matches the achievability in the rate of growth of the second - order term .
unless new techniques are invented to replace the wringing argument in the strong converse proof for the mac ( such techniques have remained elusive for over 30 years ) , the exact normalizing speed of the second - order term for the discrete and gaussian macs will remain an open problem . in the next section
, we will present a few preliminaries for the proof of theorem [ thmmainresult ] , which will be detailed in section [ sec : prf_main_result ] .
the following lemma is based on the technique of expurgating message tuples introduced by dueck ( * ? ? ?
ii ) , and the proof is provided in the appendix for completeness . [ lemmaexpurgation ]
let @xmath170 .
suppose an @xmath139-code for the gaussian mac is given .
then for each nonempty @xmath79 such that @xmath171 there exist a set @xmath94 and an @xmath172-code such that @xmath173 where @xmath102 is as defined in . as a consequence ,
if we let @xmath174 denote the probability distribution induced on the gaussian mac by the @xmath175-code , then we have for each @xmath176 @xmath177 lemma [ lemmaexpurgation ] says that restricted to the set @xmath102 , the @xmath178 ( for @xmath179 ) codebooks have almost the same sizes as the original codebooks .
in addition , the conditional probability of decoding error for each message tuple in this restricted codebook is upper bounded by @xmath180 , which is still smaller than one because @xmath146 . according to
, the probability of each message tuple can not be greater than its original value by a factor of @xmath181 .
the following lemma forms part of the wringing technique proposed by ahlswede and its proof can be found in ( * ? ? ?
* lemma 4 ) .
[ lemmafromahlswede ] let @xmath34 be a finite alphabet , let @xmath182 and @xmath183 be two probability mass functions defined on @xmath184 and let @xmath185 be a real number such that @xmath186 for all @xmath187 . fix any @xmath188 .
then for any @xmath189 , there exist @xmath190 natural numbers in @xmath191 , denoted by @xmath192 , and @xmath190 elements of @xmath34 denoted by @xmath193 , such that the following three statements hold : 1 .
2 . @xmath195 3 . for all @xmath196
, we have @xmath197 for all @xmath198 .
the crux of lemma [ lemmafromahlswede ] is in the identification of the event @xmath199 such that conditioned on @xmath200 , the distributions of the resultant codeword symbols transmitted in each time slot @xmath72 can be approximated by @xmath201 ( cf . ) . in the sequel where each @xmath202 in lemma [ lemmafromahlswede ] is substituted by @xmath203 where @xmath203 is some quantized version of @xmath204 to be specified later
, the joint distribution @xmath205 that approximates @xmath206 will be chosen to be a product distribution ( cf . ) with marginals @xmath207 . in order to use lemma [ lemmafromahlswede ] for proving theorem [ thmmainresult ] ,
an important step involves controlling the size of @xmath34 in lemma [ lemmafromahlswede ] . to this end
, we use the following scalar quantizer to quantize the alphabet @xmath122 ( in ) which is exponential in the blocklength @xmath19 ( cf . )
so that its quantized version is an alphabet whose size is polynomial in the blocklength .
[ scalarquantizer ] let @xmath208 be a natural number and @xmath209 be a positive real number , and let @xmath210 be a set of @xmath211 quantization points where @xmath209 specifies the quantization precision .
a scalar quantizer with domain @xmath212 $ ] and precision @xmath209 is the mapping @xmath213\rightarrow \mathbb{z}_{l,\delta}\ ] ] such that @xmath214 in other words , @xmath215 maps @xmath35 to the closest quantized point whose value is smaller than or equal to @xmath35 if @xmath216 , and to the closest quantized point whose value is larger than or equal to @xmath35 if @xmath217 .
in addition , define the scalar quantizer for a real - valued tuple as @xmath218^n\rightarrow \mathbb{z}_{l,\delta}^n\ ] ] such that @xmath219 @xmath220 by our careful choice of the quantizer in definition [ scalarquantizer ] , we have the following property for all @xmath221 : @xmath222 although the following lemma looks similar to ( * ? ? ?
* corollary 2 ) and they both rely on lemma [ lemmafromahlswede ] , the proof of the following lemma is more involved due to the additional consideration of the quantizer s precision and the quantized input symbols .
if the quantizer s precision is too small or too large , then the resultant bound obtained from the following lemma will not be useful in proving the strong converse .
see section [ sectiondiscussionquantizer ] for a detailed discussion on the appropriate choice for the quantizer s precision .
[ lemmawringing ] suppose we are given an @xmath223-code such that @xmath224 and @xmath225 for each @xmath226 where @xmath227 denotes the probability distribution induced on the gaussian mac by the @xmath228-code .
then , there exists an @xmath172-code with @xmath229 such that the following holds : let @xmath174 denote the probability distribution induced on the gaussian mac by the @xmath175-code .
in addition , let @xmath230 define the alphabet @xmath231 for each @xmath232 ( @xmath233 is always in the domain of @xmath234 because of , and , and hence @xmath235 ) , define @xmath236 and define @xmath237 for all @xmath238 .
then there exists a distribution @xmath239 defined on @xmath240 where @xmath241 such that for all @xmath242 , we have @xmath243 for all @xmath244 and @xmath245 \le \sum_{i\in t}np_i . \label{corollarywringingst5}\ ] ] before presenting the proof of lemma [ lemmawringing ] , we would like to stress the following two important implications of lemma [ lemmawringing ] . 1 . by identifying a certain event @xmath246 ( whose probability is quantified in in the following proof ) , we can find a subcode such that for each time slot @xmath72 , the resultant probability distribution of the quantized vector of transmitted symbols @xmath247 can be approximated by a product distribution @xmath248 as in .
this is the essence of the wringing technique @xcite which involves approximating the joint distribution of the random variables corresponding to the different encoders with a product distribution . by approximating @xmath203 with a product distribution , we effectively _ wring _ out
the dependence among the collection of random variables @xmath249 .
the alphabet size of the quantized transmitted symbol @xmath203 grows no faster than polynomially in @xmath19 as in .
our quantization strategy that results in the polynomial growth of the alphabet sizes of the quantized symbols appears to be an important and necessary step , because the original alphabet size @xmath250 could be exponentially large in @xmath19 ( cf . ) .
furthermore , the controlled growth of @xmath251 ensures that @xmath252 does not decay to zero exponentially fast as shown in in the following proof and hence the asymptotic rates of the resultant subcode are the same as that of the original code .
an important point to note here is the following : we are able to lower bound the probability @xmath252 because we defined @xmath253 in terms of the _ quantized _ random variables ( rather than the original ones ) .
the application of the wringing technique on the quantized random variables is one of the major contributions of the present work .
let @xmath227 be the probability distribution induced on the gaussian mac by the @xmath254-code that satisfies and , and let @xmath255 define a probability mass function @xmath256 as @xmath257 for all @xmath258 ( cf . and ) , where @xmath259 represents the encoding function for @xmath66 of the @xmath228-code ( cf . definition [ defcode ] ) .
the distribution @xmath256 is well - defined ( the probability masses sum to one ) through because @xmath260 using , we obtain @xmath261 where @xmath262 for all @xmath263 .
we will use lemma [ lemmafromahlswede ] to prove the existence of a subcode of the @xmath228-code such that the subcode satisfies , and for some @xmath264 defined on @xmath240 . to this end , we first consider the following chain of inequalities for each @xmath265 such that @xmath266 : @xmath267 where ( a ) follows from and .
it follows from and lemma [ lemmafromahlswede ] with the identifications @xmath268 that there exist @xmath190 natural numbers in @xmath191 , denoted by @xmath192 , and @xmath190 real - valued @xmath269-dimensional tuples in @xmath270 , denoted by @xmath271 , such that the following three statements hold : 1 .
2 . 3 . for all @xmath196
, we have @xmath273 for all @xmath244 . using statement ( ii ) , statement ( iii ) and
, we can construct an @xmath172-code by collecting all the codewords @xmath274 for the @xmath223-code which satisfy @xmath275 such that the following two statements hold : 1 .
2 . let @xmath174 denote the probability distribution induced on the gaussian mac by the @xmath277-code , and let @xmath278 then , @xmath279 and we have for all @xmath280 @xmath281 for all @xmath244 . since for each @xmath282 @xmath283 for all @xmath244
, it follows from that the following statement holds : 1 . for all @xmath242
, we have @xmath284 for all @xmath244 .
consequently , follows from statement ( i ) and statement ( i ) , and follows from statement ( iii ) by letting @xmath285 it remains to prove the upper bounds on @xmath251 and @xmath286 $ ] in and respectively .
to prove , we consider @xmath287 to prove , we first use and to obtain @xmath288 since @xmath289 for all @xmath232 and all @xmath242 by and , it follows from that @xmath290 consequently , @xmath291\notag\\ & \quad \stackrel{\eqref{defu}}{= } \sum_{i\in t}\sum_{k=1}^n\e_{u_{\hat x_{i , k}|\hat x_{i , t_1}=\bar x_{i , t_1 } , \hat x_{i , t_2}=\bar x_{i , t_2 } , \ldots , \hat x_{i , t_\ell}=\bar x_{i , t_\ell}}^\prime}\left[\hat x_{i , k}^2\right]\\ & \quad \stackrel{\eqref{distuproductform}}{=}\sum_{i\in t}\sum_{k=1}^n\e_{u_{\hat x_{t , k}|\hat x_{t , t_1}=\bar x_{t , t_1 } , \hat x_{t , t_2}=\bar x_{t , t_2 } , \ldots , \hat x_{t , t_\ell}=\bar x_{t , t_\ell}}^\prime}\left[\hat x_{i , k}^2\right]\\ & \quad = \sum_{i\in t}\sum_{k=1}^n\e_{u_{\hat x_t^n|\hat x_{t , t_1}=\bar x_{t , t_1 } , \hat x_{t , t_2}=\bar x_{t , t_2 } , \ldots , \hat x_{t , t_\ell}=\bar x_{t , t_\ell}}^\prime}\left[\hat x_{i , k}^2\right]\\ & \quad = \e_{u_{\hat x_t^n|\hat x_{t , t_1}=\bar x_{t , t_1 } , \hat x_{t , t_2}=\bar x_{t , t_2 } , \ldots , \hat x_{t , t_\ell}=\bar x_{t , t_\ell}}^\prime}\left[\sum_{i\in t}\sum_{k=1}^n\hat x_{i , k}^2\right]\\ & \quad \stackrel{\eqref{uprimexhatpowerconstraint}}{\le}\sum_{i\in t } np_i\ , .
\end{aligned}\ ] ] the following definition concerning the non - asymptotic fundamental limits of a simple binary hypothesis test is standard .
see for example ( * ? ? ?
iii - e ) . [ defbhtdivergence ]
let @xmath292 and @xmath293 be two probability distributions on some common alphabet @xmath34 .
let @xmath294 be the set of randomized binary hypothesis tests between @xmath292 and @xmath293 where @xmath295 indicates the test chooses @xmath296 , and let @xmath297 $ ] be a real number .
the minimum type - ii error in a simple binary hypothesis test between @xmath292 and @xmath293 with type - i error no larger than @xmath298 is defined as @xmath299 the existence of a minimizing test @xmath300 is guaranteed by the neyman - pearson lemma .
we state in the following lemma and proposition some important properties of @xmath301 , which are crucial for the proof of theorem [ thmmainresult ] .
the proof of the following lemma can be found in , for example , the paper by wang , colbeck , and renner ( * ? ? ? * lemma 1 ) . [ lemmadpi ]
let @xmath292 and @xmath293 be two probability distributions on some alphabet @xmath34 , and let @xmath38 be a function whose domain contains @xmath34 .
then , the following two statements hold : 1 .
data processing inequality ( dpi ) : @xmath302 2 . for all @xmath303 , @xmath304
the proof of the following proposition is similar to lemma 3 in @xcite and therefore omitted .
[ propositionbhtlowerbound ] let @xmath305 be a probability distribution defined on @xmath306 for some finite alphabet @xmath307 .
in addition , let @xmath308 be a distribution defined on @xmath307 , and let @xmath309 be a real number in @xmath310 where @xmath311 is distributed according to @xmath305 . then for each @xmath312 , @xmath313
let @xmath170 and suppose @xmath314 is an @xmath32-achievable rate tuple . by definition [ defachievablerate ] ,
there exists a @xmath315 and a sequence of @xmath143-codes such that @xmath316 for all sufficiently large @xmath19 and @xmath317 for each @xmath109 .
fix a non - empty set @xmath79 .
our goal is to prove that @xmath318 since holds trivially if @xmath319 , we assume without loss of generality that @xmath320 it follows from and that @xmath321 for all sufficiently large @xmath19 . fix a sufficiently large @xmath19 and the corresponding @xmath143-code for the gaussian mac such that and hold . using lemma [ lemmaexpurgation ] ,
lemma [ lemmawringing ] and definition [ defcode ] , there exists an @xmath322-code , which induces a probability distribution on the gaussian mac denoted by @xmath174 , such that the following four statements hold : 1 . for all @xmath98 and all @xmath176 , @xmath323 2 .
there exists a @xmath97 such that for all @xmath98 , we have @xmath99 .
3 . the support of @xmath101 satisfies @xmath324 4 .
define @xmath325 for all @xmath326 , where @xmath327 and @xmath328 then there exists a distribution @xmath239 defined on @xmath240 such that for all @xmath242 , we have @xmath329 for all @xmath244 and @xmath330 \le \sum_{i\in t}np_i \ , .
\label{eqn : statement(v ) } \end{aligned}\ ] ] note that @xmath174 is not the distribution induced by the original @xmath143-code but rather it is induced by the expurgated @xmath322-code . now ,
let @xmath331 be a distribution such that for each @xmath242 , the auxiliary conditional output distribution is chosen to be @xmath332 + \sum_{j\in t^c}x_{j , k},1+\sum_{i\in t}p_i \right ) \label{defsimulatingdistsksumrate}\ ] ] for all @xmath333 and @xmath334 .
it can be seen from and that @xmath335 depends on the choice of @xmath87 we fixed at the start of the proof and the distribution @xmath239 in statement ( iv ) .
we shall see later that this choice of @xmath335 , in particular the mean of the distribution in namely @xmath336 + \sum_{j\in t^c}x_{j , k}$ ] , combined with proposition [ propositionbhtlowerbound ] and lemma [ lemmadpi ] enables us to prove .
we do not index @xmath335 by @xmath87 nor @xmath239 for notational brevity . to simplify notation ,
let @xmath337 be the maximal probability of decoding error of the @xmath322-code , where @xmath338 because @xmath339 .
then for each @xmath340 , since @xmath341 it follows from proposition [ propositionbhtlowerbound ] and definition [ defcode ] with the identifications @xmath342 , @xmath343 , @xmath344 , @xmath345 and @xmath346 that @xmath347 consider the following chain of inequalities for each @xmath348 : @xmath349 where 1 . follows from the dpi of @xmath350 by introducing the channel output @xmath78 .
2 . follows from the fact that @xmath351 forms a markov chain under the distribution @xmath352 .
3 . follows from the dpi of @xmath350 by introducing the channel input @xmath353 .
4 . follows from definition [ defcode ] , which says @xmath354 is a function of @xmath355 . following , we consider @xmath356 and we obtain from lemma [ lemmadpi ] and that for each @xmath98 and each @xmath357 , @xmath358 combining , and , we obtain for each @xmath98 and each @xmath357 @xmath359 which implies that @xmath360 for each @xmath98 , let @xmath361 \notag \\ & \qquad + \sqrt{\frac{2}{1-\bar \gamma}\var_{p_{x_{\mathcal{i}}^n , y^n|w_\mathcal{i}=w_\mathcal{i}}}\left[\sum_{k=1}^n \log\left ( \frac{p_{y_k|x_{\mathcal{i},k}}(y_k|x_{\mathcal{i},k})}{s_{y_{k}|x_{t^c , k}}(y_k|x_{t^c , k})}\right)\right]}\ , \ , .
\label{defxiwi}\end{aligned}\ ] ] using chebyshev s inequality , it follows from that for each @xmath98 @xmath362 which implies from that @xmath363 since @xmath364 is convex for @xmath365 , by jensen s inequality @xmath366 we have @xmath367 where ( a ) follows from the definition of @xmath102 in and the fact stated in statement ( ii ) that @xmath99 for all @xmath98 . using and ,
we obtain @xmath368 taking expectation with respect to @xmath369 on both sides of and applying , we obtain @xmath370 in order to simplify , we will simplify the log - likelihood term in @xmath371 defined in . to this end
, we first let @xmath372 ( @xmath104 is the encoding function at node @xmath62 defined in definition [ defcode ] ) and we also let @xmath373 denote the @xmath374 component of @xmath375 for each @xmath109 and each @xmath376 such that @xmath377 in addition , we let @xmath378 and we let @xmath379 be a subtuple of @xmath380 . similarly , let @xmath381 and let @xmath382 be a subtuple of @xmath383 . using the fact that @xmath68 is a function of @xmath66 for all @xmath109 and the notations defined above , we obtain from that @xmath384 \notag \\ & \qquad + \sqrt{\frac{2}{1-\bar \gamma}\var_{p_{y^n|w_\mathcal{i}=w_\mathcal{i},x_{\mathcal{i}}^n = x_\mathcal{i}^n(w_\mathcal{i})}}\left[\sum_{k=1}^n \log\left ( \frac{p_{y_k|x_{\mathcal{i},k}}(y_k|x_{\mathcal{i},k}(w_\mathcal{i}))}{s_{y_{k}|x_{t^c , k}}(y_k|x_{t^c , k}(w_{t^c}))}\right)\right]}\ , \ , , \end{aligned}\ ] ] which implies from that @xmath385 \notag \\ & \qquad + \sqrt{\frac{2}{1-\bar \gamma}\var_{\prod_{k=1}^np_{y_k|x_{\mathcal{i},k}=x_{\mathcal{i},k}(w_\mathcal{i } ) } } \left[\sum_{k=1}^n \log\left ( \frac{p_{y_k|x_{\mathcal{i},k}}(y_k|x_{\mathcal{i},k}(w_\mathcal{i}))}{s_{y_{k}|x_{t^c , k}}(y_k|x_{t^c , k}(w_{t^c}))}\right)\right]}\ , \ , , \end{aligned}\ ] ] which then implies that @xmath386 \notag \\ & \qquad + \sqrt{\frac{2}{1-\bar \gamma}\sum_{k=1}^n\var_{p_{y_k|x_{\mathcal{i},k}=x_{\mathcal{i},k}(w_\mathcal{i } ) } } \left [ \log\left ( \frac{p_{y_k|x_{\mathcal{i},k}}(y_k|x_{\mathcal{i},k}(w_\mathcal{i}))}{s_{y_{k}|x_{t^c , k}}(y_k|x_{t^c , k}(w_{t^c}))}\right)\right]}\ , \ , .
\label{defxiwi*}\end{aligned}\ ] ] following , we use , and to obtain @xmath387)\right)\left(y_k-\sum_{i\in\mathcal{i}}x_{i , k}(w_i)\right ) + \left(\sum_{i\in t}(x_{i , k}(w_i)-\e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2\bigg ) .
\label{eqnbht8thchain}\end{aligned}\ ] ] for each @xmath98 and each @xmath242 , it follows from definition [ defgaussianmac ] that @xmath388 is a standard normal random variable if @xmath389 is distributed according to @xmath390 , which then implies that @xmath391 \notag\\ * & \stackrel{\eqref{eqnbht8thchain}}{= } \frac{1}{2}\log\left(1+\sum_{i\in t}p_i\right)+\frac{\log e}{2(1+\sum_{i\in t}p_i)}\left(-\left(\sum_{i\in t}p_i\right ) + \left(\sum_{i\in t}(x_{i , k}(w_i)-\e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2\right ) \label{infospectrumexptimek}\end{aligned}\ ] ] and @xmath392 \notag\\ * & \stackrel{\eqref{eqnbht8thchain}}{=}\bigg ( \frac{\log e}{2(1+\sum_{i\in t}p_i ) } \bigg)^2 \var_{p_{y_k|x_{\mathcal{i},k}=x_{\mathcal{i},k}(w_\mathcal{i})}}\bigg [ -\bigg(\sum_{i\in t}p_i\bigg)\bigg(y_k-\sum_{i\in\mathcal{i}}x_{i , k}(w_i)\bigg)^2 \notag\\ * & \qquad + 2\left(\sum_{i\in t}(x_{i , k}(w_i)-\e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)\bigg(y_k-\sum_{i\in\mathcal{i}}x_{i , k}(w_i)\bigg ) \bigg ] \\ * & = \frac{\left((\sum_{i\in t}p_i)^2 + 2\left(\sum_{i\in t}(x_{i , k}(w_i)-\e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2\right)(\log e)^2}{2(1+\sum_{i\in t}p_i)^2}. \label{infospectrumvarexptimek}\end{aligned}\ ] ] define @xmath393 and @xmath394
. \label{defbarxik}\ ] ] combining , , , , and , we obtain for each @xmath98 @xmath395 which implies from jensen s inequality ( @xmath396 is concave for @xmath397 ) that @xmath398 in the following , we will obtain an upper bound on the crucial term @xmath399 which appears in the second and third terms on the right - hand - side of .
following , we consider for each @xmath242 @xmath400 since @xmath68 is a function of @xmath66 for each @xmath179 , it follows from that for each @xmath242 @xmath401\right)\right)^2,\ ] ] which implies from that @xmath402\right)\right)^2 . \label{eqnbht10thchain*}\ ] ] recalling the definition of @xmath403 and @xmath240 in and respectively , we write for each @xmath242 @xmath404\right)\right)^2 \notag\\ * & = \sum\limits_{x_{t , k}\in \mathcal{x}_t , \hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{x_{t , k } , \hat x_{t , k}}(x_{t , k } , \hat x_{t , k})\left(\sum\limits_{i\in t } \left(x_{i , k}-\hat x_{i , k}+\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}]\right)\right)^2 \\ & = \sum\limits_{x_{t , k}\in \mathcal{x}_t , \hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{x_{t , k } , \hat x_{t , k}}(x_{t , k } , \hat x_{t , k})\left(\sum\limits_{i\in t } ( x_{i , k}-\hat x_{i , k})\right)^2 \notag\\ * & \qquad + 2\sum\limits_{x_{t , k}\in \mathcal{x}_t , \hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{x_{t , k } , \hat x_{t , k}}(x_{t , k } , \hat x_{t , k})\left(\sum\limits_{i\in t } ( x_{i , k}-\hat x_{i , k})\right)\left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)\notag\\ * & \qquad + \sum\limits_{\hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{\hat x_{t , k}}(\hat x_{t , k})\left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2 \\ & \le \sum\limits_{x_{t , k}\in \mathcal{x}_t , \hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{x_{t , k } , \hat x_{t , k}}(x_{t , k } , \hat x_{t , k})\left|\sum\limits_{i\in t } ( x_{i , k}-\hat x_{i , k})\right|^2 \notag\\ * & \qquad + 2\sum\limits_{x_{t , k}\in \mathcal{x}_t , \hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{x_{t , k } , \hat x_{t , k}}(x_{t , k } , \hat x_{t , k})\left|\sum\limits_{i\in t } ( x_{i , k}-\hat x_{i , k})\right|\left|\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right|\notag\\ * & \qquad + \sum\limits_{\hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{\hat x_{t , k}}(\hat x_{t , k})\left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2 \\ & \le \sum\limits_{x_{t , k}\in \mathcal{x}_t , \hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{x_{t , k } , \hat x_{t , k}}(x_{t , k } , \hat x_{t , k})\left(\sum\limits_{i\in t } |x_{i , k}-\hat x_{i , k}|\right)^2 \notag\\ * & \qquad + 2\sum\limits_{x_{t , k}\in \mathcal{x}_t , \hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{x_{t , k } , \hat x_{t , k}}(x_{t , k } , \hat x_{t , k})\left(\sum\limits_{i\in t } |x_{i , k}-\hat x_{i , k}|\right)\left(\sum\limits_{i\in t}(|\hat x_{i , k}| + \e_{u_{\hat x_{i , k}}}[|\hat x_{i , k}|])\right)\notag\\ * & \qquad + \sum\limits_{\hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{\hat x_{t , k}}(\hat x_{t , k})\left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2 \\ & \stackrel{\text{(a)}}{\le}\frac{|t|^2}{n^2 } + \frac{4|t|}{\sqrt{n}}\left(\sum_{i\in t}\sqrt{p_i}\right ) + \sum\limits_{\hat x_{t , k}\in \hat{\mathcal{x}}_t}p_{\hat x_{t , k}}(\hat x_{t , k})\left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2 \label{eqnbht11thchain}\end{aligned}\ ] ] where ( a ) follows from the facts below for each @xmath179 , each @xmath242 and each @xmath405 ( recall the definition of @xmath406 in ): @xmath407 and @xmath408 in order to bound the last term in , we use the bound in for bounding @xmath409 in terms of @xmath410 to obtain @xmath411)\right)^2 \notag\\ & \quad \le \sum\limits_{\hat x_{t , k } \in \hat{\mathcal{x}}_t}\left(\left(1+\sqrt{\frac{\log n}{n}}\right)\prod_{i\in t}u_{\hat x_{i , k}}(\hat x_{i , k})+ \frac{1}{n^{4|t|}}\right)\left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2 \label{eqnbht12thchain_a}\\ & \quad = \sum\limits_{\hat x_{t , k } \in \hat{\mathcal{x}}_t}\bigg[\left(1+\sqrt{\frac{\log n}{n}}\right)\prod_{i\in t}u_{\hat x_{i , k}}(\hat x_{i , k})\left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2 \notag\\ * & \qquad\qquad\qquad\qquad+ \frac{1}{n^{4|t| } } \left(\sum\limits_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])\right)^2 \bigg]\label{eqnbht12thchain}\end{aligned}\ ] ] for each @xmath242 . the bound in consists of two distinct terms which we now bound separately .
consider the following two chains of inequalities for each @xmath242 : @xmath412)\right)^2 \notag \\ * & = \sum_{i\in t } \e_{u_{\hat x_{i , k}}}\left [ ( \hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])^2\right ] \\ * & \le \sum_{i\in t } \e_{u_{\hat x_{i , k } } } \left[\hat x_{i , k}^2\right ] \label{eqnbht13thchain}\end{aligned}\ ] ] and @xmath413)\right)^2 \notag\\ * & \quad \le \sum\limits_{\hat x_{t , k } \in \hat{\mathcal{x}}_t } \left(|t|\max\limits_{i\in t}\left\{|\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}]|\right\ } \right)^2 \\ & \quad = |t|^2\sum\limits_{\hat x_{t , k } \in \hat{\mathcal{x}}_t } \max\limits_{i\in t}\left\{(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])^2\right\ } \\ & \quad \le |t|^2\sum\limits_{\hat x_{t , k } \in \hat{\mathcal{x}}_t } \sum_{i\in t}(\hat x_{i , k}- \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])^2 \\ & \quad \stackrel{\text{(a)}}{\le } 2|t|^2\sum\limits_{\hat x_{t , k } \in \hat{\mathcal{x}}_t } \sum_{i\in t}\left(\hat x_{i , k}^2 + ( \e_{u_{\hat x_{i , k}}}[\hat x_{i , k}])^2\right ) \\ & \quad \stackrel{\eqref{boundonhatxik}}{\le } 2|t|^2\sum\limits_{\hat x_{t , k } \in \hat{\mathcal{x}}_t } \sum_{i\in t}2np_i \\ & \quad \stackrel{\text{(b)}}{\le } 4n|t|^2 |p_t| | \hat{\mathcal{x}}_t|\\ & \quad \stackrel{\eqref{eqn : size_xhat } } { < } 4n^{3|t|}|t|^2 |p_t|\prod\limits_{i \in t}(2\sqrt{p_i } + 3 ) , \label{eqnbht14thchain}\end{aligned}\ ] ] where 1
. follows from the fact that @xmath414 for all real numbers @xmath415 and @xmath416 .
2 . follows from the definition of @xmath417 in . combining , and
, we obtain for each @xmath242 @xmath418)\right)^2 \notag\\ * & \quad \le \left(1+\sqrt{\frac{\log
n}{n}}\right)\sum_{i\in t } \e_{u_{\hat x_{i } } } \left[\hat x_{i , k}^2\right ] + 4n^{-|t|}|t|^2 |p_t|\prod\limits_{i \in t}(2\sqrt{p_i } + 3),\end{aligned}\ ] ] which implies from and that @xmath419 + 4n^{-|t|}|t|^2 |p_t|\prod\limits_{i \in t}(2\sqrt{p_i } + 3 ) .
\label{eqnbht15thchain}\end{aligned}\ ] ] using and and recalling that @xmath420 ( because @xmath87 is non - empty ) , we obtain @xmath421 to simplify notation , let @xmath422 be two constants that are independent of @xmath19 . then , we combine and to yield @xmath423 combining and , we obtain @xmath424 dividing both sides of by @xmath19 and taking limit inferior as @xmath19 goes to infinity , we obtain from that holds as desired .
this completes the proof of theorem [ thmmainresult ] .
our choice of @xmath425 in has been optimized in the following sense .
if @xmath425 is chosen such that @xmath426 , then the second - order term on the rhs of would be @xmath427 ( cf . and ) , which then leads to an upper bound on @xmath428 with a looser ( larger ) second - order term @xmath429 ; if @xmath425 is chosen such that @xmath430 , then the magnitude of the first term on the lhs of would be @xmath427 ( cf . ) , which then leads to an upper bound on @xmath428 with a looser second - order term @xmath429 .
hence our choice of @xmath431 balances " the rates of growth of the two second - order terms in . in this sense
, our choice of @xmath425 is optimal .
we now discuss the choice of the quantizer s precision @xmath432 as shown in . based on this choice of @xmath433
, we note that any choice of @xmath434 in satisfying @xmath435 does not affect the second - order term of the resultant upper bound on @xmath428 implied by .
in particular , the current choice @xmath436 stated in leads to the rightmost term in , which contributes to the fourth constant term in as well as the constant term on the rhs of .
if the quantizer s precision is chosen to be some other @xmath437 , then it can be seen by inspecting , the upper bound obtained at step ( a ) in the chain of inequalities leading to , and that the second - order term of resultant upper bound on @xmath428 is @xmath438 . in particular ,
if @xmath437 is chosen such that @xmath439 for any fixed @xmath440 , we can follow similar calculations ( with a slight modification of @xmath434 ) to conclude that the second - order term of the upper bound on @xmath428 is proportional to @xmath441 . as explained in the second remark after lemma [ lemmawringing ] , as long as @xmath437 decays to zero no faster than polynomially in @xmath19
, then @xmath251 grows at most polynomially fast in @xmath19 , which will ensure that the asymptotic rates of the resultant sequence of subcodes obtained from the wringing step are the same as that of the original sequence of codes .
however , if @xmath437 decays to zero exponentially fast ( i.e. , @xmath442 for some @xmath443 ) , then @xmath251 will grow exponentially fast in @xmath19 and the rhs of will decay exponentially rather than polynomially fast .
this in turn causes the asymptotic rates of the resultant sequence of subcodes to decrease by a positive quantity , thus resulting in a loose first - order term on the rhs of the final inequality ( which does not match the corresponding term in the cover - wyner capacity region ) .
therefore , with this choice of @xmath437 , the strong converse can not be shown .
the capacity region of a two - source two - destination gaussian interference channel ( ic ) under strong interference was derived by han and kobayashi @xcite and sato @xcite .
let @xmath444 be the received signal - to - noise ratios and let @xmath445 be the received interference - to - noise ratios ( * ? ? ?
6.4 ) . under the formulation of the gaussian ic under _ strong interference _ , it is assumed that @xmath446 and @xmath447 . under this condition , the capacity region was shown in ( * ? ? ?
5.2 ) to be the han - kobayashi region @xmath448{3.6
in}{$r_1\le \frac{1}{2}\log(1+p_1)$ , \vspace{0.04 in}\\ $ r_2\le \frac{1}{2}\log(1+p_2 ) $ , \vspace{0.04 in } \\ $ r_1+r_2\le \min\{\frac{1}{2}\log(1+p_1+i_1),\frac{1}{2}\log(1+p_2+i_2 ) \ } $ } \right.\right\}. \label{strong_interference}\ ] ] by applying theorem [ thmmainresult ] to each of the decoders of the two - source two - destination gaussian ic , we can show that the corresponding @xmath449-capacity region @xmath450 is outer bounded as @xmath451 as long as @xmath452 , where @xmath453 characterizes the asymptotic average probability of destination @xmath62 decoding message @xmath62 wrongly .
since the rate pairs in @xmath454 are @xmath455-achievable via simultaneous non - unique decoding ( * ? ? ?
6.4 ) , we have @xmath456 as long as @xmath452 . the strong converse ( in fact , the complete second - order asymptotics ) for the gaussian ic under the more restrictive condition of strictly very strong interference
was shown by le , tan , and motani @xcite . in the rest of this section , we will describe the formulation of the gaussian ic under strong interference and present in section [ sectionproofofthmic ] the corresponding strong converse result .
we follow the standard setting of the gaussian ic under strong interference as given in ( * ? ? ?
the gaussian ic under strong interference consists of two sources , denoted by @xmath457 and @xmath458 respectively , and two destinations , denoted by @xmath459 and @xmath460 respectively . for each @xmath461
, @xmath462 chooses a message @xmath66 and transmits @xmath68 in @xmath19 time slots , and @xmath463 receives @xmath464 in @xmath19 time slots and declares @xmath465 to be the transmitted @xmath66 .
the channel law in each time slot @xmath72 is @xmath466 = \left[\begin{array}{cc}1 & g_{12}\\ g_{21 } & 1\end{array}\right]\left[\begin{array}{c}x_{1,k } \\
x_{2,k}\end{array}\right ] + \left[\begin{array}{c}z_{1,k } \\ z_{2,k}\end{array}\right ] , \label{channellawic}\ ] ] where @xmath467 and @xmath468 are two real constants characterizing the channel gains of the interference links , and @xmath469 are @xmath19 independent copies of a gaussian random vector denoted by @xmath470 ( @xmath75 and @xmath471 need not be independent ) such that @xmath472=\e\left[z_{2}\right ] = 0 \label{channellawic1 } \end{aligned}\ ] ] and @xmath473=\e\left[z_{2}^2\right ] = 1 .
\label{channellawic2 } \end{aligned}\ ] ] for each @xmath461 , the codewords transmitted by @xmath462 should satisfy the peak power constraint @xmath474 for some @xmath475 .
we assume that the ic is under strong interference , i.e. , @xmath476 and @xmath477 , which implies that @xmath478 and @xmath479 where @xmath480 and @xmath481 characterize the interference power received at @xmath459 and @xmath460 respectively ( cf . ) .
the gaussian ic is characterized by some conditional probability density function @xmath482 and we define the gaussian ic in a similar way to a gaussian mac ( cf .
definition [ defgaussianmac ] ) such that , and hold .
in addition , we define a length-@xmath19 code for the gaussian ic as follows .
[ defcodeic ] an _ @xmath483-code _ for the gaussian ic consists of the following : 1 . a message set @xmath92 at node @xmath62 for each @xmath484 , where @xmath66 is uniform on @xmath485 .
an encoding function @xmath103 for each @xmath484 , where @xmath104 is the encoding function at node @xmath62 such that @xmath105 and @xmath486 for all @xmath107 .
3 . a ( possibly stochastic ) decoding function @xmath487 for each @xmath461 , where @xmath488 is used by node @xmath463 to estimate @xmath66 , i.e. , @xmath489 we define an @xmath490-code as follows .
[ deferroric ] for an @xmath483-code defined on the gaussian ic , the _ average probability of decoding error for @xmath66 _ is defined for each @xmath461 as @xmath491 an @xmath483-code with @xmath492 and @xmath493 is called an @xmath494-code . for each @xmath495 and each @xmath496
, we define an @xmath449-achievable rate pair as in definition [ defachievablerate ] , and we define the @xmath449-capacity region , denoted by @xmath450 , to be the set of @xmath449-achievable rate pairs .
the following theorem is the main result in this section .
[ thmmainresultic ] for each @xmath495 and each @xmath496 such that @xmath452 , @xmath497 we need the following definitions and lemma before presenting the proof of theorem [ thmmainresultic ] .
the definition below concerning a multicast code differs from definition [ defcodeic ] in the decoding functions only , but we state the whole definition for clarity .
essentially , a multicast code for the gaussian ic is the same as a standard code except that each decoder must output estimates of _ both _ messages .
[ defcodeicandmac ] an _ @xmath483-multicast code _ for the gaussian ic consists of the following : 1 . a message set @xmath92 at node @xmath62 for each @xmath484 , where @xmath66 is uniform on @xmath485 .
an encoding function @xmath103 for each @xmath484 , where @xmath104 is the encoding function at node @xmath62 such that @xmath105 and @xmath498 for all @xmath107 .
3 . a ( possibly stochastic ) decoding function @xmath499 for each @xmath461 , where @xmath488 is used by node @xmath463 to estimate both @xmath500 and @xmath501 such that the pair of message estimates is @xmath502 .
we define an @xmath490-multicast code as follows .
note that the multicast code is used for the gaussian ic but not a general multicast channel .
[ deferroricandmac ] for an @xmath483-multicast code defined on the gaussian ic , the _ average probability of decoding error at destination @xmath463 _ is defined for each @xmath461 as @xmath503 an @xmath483-multicast code with average probability of decoding error at destination @xmath463 no larger than @xmath453 for each @xmath461 is called an @xmath490-code . the following lemma plays a crucial role in extending our strong converse result for the gaussian mac to the gaussian ic under strong interference , because it relates the error probabilities for standard codes defined for the gaussian ic in definition [ deferroric ] to the error probabilities for multicast - codes defined for the gaussian ic in definition [ deferroricandmac ] .
[ lemmaic ] for each @xmath490-code for the gaussian ic , there exists an @xmath504-multicast code for the gaussian ic .
suppose we are given an @xmath490-code whose encoding and stochastic decoding functions are denoted by @xmath505 and @xmath506 respectively ( cf . definition [ defcodeic ] ) .
let @xmath507 be the probability distribution induced by the @xmath490-code . by definition [ deferroric ] , we have for each @xmath461 @xmath508 which implies from that @xmath509 and @xmath510 in the rest of the proof , we construct new stochastic decoding functions at @xmath459 and @xmath460 , denoted by @xmath511 and @xmath512 respectively , such that @xmath513 and @xmath514 can be viewed as the stochastic decoding functions of an @xmath515-multicast code . to this end , we first define @xmath516 and @xmath517 to be @xmath19 independent copies of the standard normal random variable such that @xmath516 , @xmath518 and @xmath519 are independent . in addition , there exist @xmath520 and @xmath521 such that @xmath522 and @xmath523 which implies from and that @xmath524 and @xmath525 then , we define the stochastic decoders @xmath526 and @xmath527 where the randomness properties of the stochastic functions originate from not only @xmath528 and @xmath529 but also @xmath516 and @xmath518 . since @xmath530 and @xmath531 have the same distribution by , it follows from and that @xmath532 combining and , we obtain @xmath533 following similar procedures for deriving , we obtain the following inequality by using , and : @xmath534 replacing the decoding functions of the @xmath490-code with @xmath513 and @xmath514 and keeping the encoding functions unchanged , we conclude from and that the resultant code is an @xmath535-multicast code
. we are now ready to prove the strong converse theorem for the gaussian ic under strong interference .
fix @xmath536 and @xmath537 such that @xmath538 as discussed at the beginning of section [ sectionic ] , it follows from theorem 5.2 in @xcite that @xmath539 where the quantities @xmath480 and @xmath481 in @xmath540 are defined in and respectively . since @xmath541 for all non - negative real numbers @xmath542 and @xmath543 by definition , @xmath544 therefore , it suffices to prove @xmath545 to this end , fix a rate pair @xmath546 . by definition , there exists a sequence of @xmath547-codes such that @xmath548 and @xmath549 for each @xmath461 .
it then following from lemma [ lemmaic ] and that there exists a sequence of @xmath550-multicast codes such that @xmath551 for each @xmath461 .
construct a subnetwork of the gaussian ic formed by deleting @xmath460 as well as the links connecting to it . by inspection ,
the resultant subnetwork is a two - source gaussian mac and the sequence of @xmath552-multicast codes for the gaussian ic induces a sequence of @xmath553-codes for the two - source gaussian mac .
it then follows from and that @xmath554 is @xmath555-achievable for the two - source gaussian mac , which implies from theorem [ thmmainresult ] , and that @xmath556 @xmath557 and @xmath558 similarly , if we repeat the above procedures for the other two - source gaussian mac resulting from deleting @xmath459 from the gaussian ic , we obtain @xmath559 @xmath560 and @xmath561 combining the bounds in , , , , the capacity region in , and the strong interference conditions in and , we have @xmath562 .
consequently , the outer bound in holds , and the theorem follows from and the inner bound stated in .
suppose an @xmath139-code is given for some @xmath170 , and let @xmath563 be the probability of decoding error given that @xmath95 is the message tuple transmitted by the sources .
then by choosing @xmath95 one by one in an increasing order of @xmath564 , we can construct a set @xmath565 such that @xmath566 for all @xmath567 and @xmath568 this is essentially an expurgation argument .
the bound in means that there exists an @xmath569-code such that holds .
fix a nonempty @xmath79 .
define @xmath570 for each @xmath571 such that @xmath572 since @xmath573 , it follows from and that there exists a @xmath97 such that @xmath574 or otherwise we would obtain the following chain of inequalities which would eventually contradict : @xmath575 which contradicts . due to , we can construct an @xmath576-code based on the @xmath569-code such that they have the same message sets , encoding functions and decoding function and differ in only the support set of the message tuple @xmath93 ( cf . definition [ defcode ] ) .
in particular , the second statement in definition [ defcode ] is satisfied because of the following reasons : 1 . by construction , @xmath93 is uniform on @xmath577 .
2 . for all @xmath578
, we have @xmath99 by .
let @xmath579 .
it remains to show that and hold for the @xmath175-code .
recalling the definition of @xmath102 in , we obtain from that @xmath580 which implies from that @xmath241 consequently , follows from , and .
it remains to prove . to this end , let @xmath581 denote the probability distribution induced on the gaussian mac by the @xmath175-code , where @xmath582 for all @xmath176 by definition [ defcode ] . using and , we obtain @xmath583 for each @xmath176 .
the authors are extremely grateful to yury polyanskiy for pointing out an error in an earlier version of the manuscript .
they would also like to thank the associate editor prof .
sandeep pradhan and the two anonymous reviewers for the useful comments which greatly improve the presentation of this paper .
v. y. f. tan , `` asymptotic estimates in information theory with non - vanishing error probabilities , '' _ foundations and trends in communications and information theory _ , vol . 11 , no . 1 - 2 , pp . 1183 , 2014 .
j. scarlett , a. martinez , and a. guilln i fbregas , `` second - order rate region of constant - composition codes for the multiple - access channel , '' _ ieee trans . on inf .
theory _ , vol .
61 , no . 1 ,
pp . 157172 , 2015 .
e. molavianjazi and j. n. laneman , `` a second - order achievable rate region for gaussian multi - access channels via a central limit theorem for functions , '' _ accepted to ieee trans . on inf .
theory _ , 2015 .
r. ahlswede , p. gcs , and j. krner , `` bounds on conditional probabilities with applications in multi - user communication , '' _ z. wahrscheinlichkeitstheorie verw .
gebiete _ , vol .
34 , no . 3 , pp . 157177 , 1976 .
m. raginsky and i. sason , `` concentration of measure inequalities in information theory , communications , and coding , '' _ foundations and trends in communications and information theory _ , vol .
10 , no . 1 - 2 , pp .
1246 , 2013 . | we prove the strong converse for the @xmath0-source gaussian multiple access channel ( mac ) .
in particular , we show that any rate tuple that can be supported by a sequence of codes with asymptotic average error probability less than one must lie in the cover - wyner capacity region .
our proof consists of the following .
first , we perform an expurgation step to convert any given sequence of codes with asymptotic average error probability less than one to codes with asymptotic maximal error probability less than one .
second , we quantize the input alphabets with an appropriately chosen resolution . upon quantization , we apply the wringing technique ( by ahlswede ) on the quantized inputs to obtain further subcodes from the subcodes obtained in the expurgation step so that the resultant correlations among the symbols transmitted by the different sources vanish as the blocklength grows .
finally , we derive upper bounds on achievable sum - rates of the subcodes in terms of the type - ii error of a binary hypothesis test .
these upper bounds are then simplified through judicious choices of auxiliary output distributions .
our strong converse result carries over to the gaussian interference channel under strong interference as long as the sum of the two asymptotic average error probabilities less than one .
gaussian multiple access channel , strong converse , binary hypothesis testing , expurgation , wringing technique |
so far , warm absorbers revealed their existence mainly in the soft x - ray spectral region by imprinting absorption edges on the x - ray spectra of many agn .
this highly ionized material provides an important new diagnostic tool to investigate the physical conditions in the nuclei of active galaxies .
recently , evidence has accumulated that some warm absorbers contain significant amounts of dust .
this possibility was first suggested by brandt et al .
( 1996 , b96 hereafter ) to explain the lack of excess x - ray _ cold _ absorption despite strong optical reddening of the quasar iras 13349 + 2438 . a point emphasized by komossa & fink ( kofi hereafter , e.g. 1997a , b ) and komossa & bade ( koba ; 1998 ) is the strong influence of the presence of dust on the x - ray absorption spectrum which becomes drastic for high column densities @xmath0 .
signatures of the presence of ( galactic - ism - like ) dust , as opposed to the dust - free case are , e.g. , * a strong carbon edge in the x - ray spectrum , * a stronger temperature gradient across the absorber with more gas in a ` colder ' state , shifting the dominant absorption edges to more lowly ionized species * increased sensitivity of dusty gas to radiation pressure , which may drive strong outflows of the warm material .
the first two effects lead to an _ effective flattening _ of the observed x - ray spectrum .
thus , for both , the derivation of the physical properties of the warm absorber as well as the intrinsic spectral shape it is important to apply self - consistent models that take into account the presence of dust .
.properties of the dusty warm absorber in iras 13349 derived from x - ray spectral fits . for comparison results for a dust - free warm absorber
are shown .
the derived properties of the ionized material and the intrinsic continuum are rather different in this case .
[ cols="<,<,^ , < , < , < , < " , ] [ tab1 ] @xmath1 fixed to the value @xmath2 determined from optical reddening this quasar received a lot of attention , recently . a detailed optical study was presented by wills et al . (
1992 ) . in a thorough study of the x - ray spectrum of
, b96 suggested the presence of a dusty warm absorber .
brinkmann et al .
( 1996 ) detected changes in the spectrum as compared to the earlier data ; the warm - absorption features remained present ( brandt et al .
further candidates for dusty warm absorbers quickly followed , including ngc3227 ( kofi 1996 , 1997b ) , ngc3786 ( kofi 1997c ) , and iras17020 + 4544 ( leighly et al .
1997 , koba ) .
whereas the model of a warm absorber that _ includes _ the presence of dust has been successfully fit to these objects , it has not yet been applied to ( first results of the present study were reported by komossa 1998 , and komossa & greiner 1999 ) .
given the potentially strong modifications of the x - ray absorption spectrum in the presence of dust , it is important to scrutinize whether a dusty warm absorber is consistent with the observed x - ray spectrum . since some strong features of dusty warm absorbers appear outside the sensitivity range , data are best suited for this purpose ; we used the two pointed pspc observations of of jan .
1992 and dec .
1992 . as ionizing continuum illuminating the absorber we adopted a mean seyfert spectrum of piecewise powerlaws ( as in kofi 1997b ) with @xmath3 .
we use as definition for the ionization parameter @xmath4 , where @xmath5 is the rate of photons above the lyman limit .
the photoionization calculations were carried out with ferland s ( 1993 ) code _
cloudy_. in a first step , we fit a dust-_free _ warm absorber ( as in b96 , but using the additional information on the hard x - ray powerlaw available from the asca observation , @xmath6 ) .
this gives an excellent fit with log @xmath0=22.7 ( @xmath7 = 0.8 ) .
if this same model is re - calculated by fixing @xmath0 and the other best - fit parameters but _ adding dust _ to the warm absorber the x - ray spectral shape is drastically altered and the data can not be fit at all ( @xmath7 = 150 ) .
this still holds if we allow for non - standard dust , i.e. , selectively exclude either the graphite or silicate species . whereas this first approach was mainly to demonstrate the strong influence of the presence of dust
, it has to be kept in mind that the expected column density @xmath2 derived from optical extinction is less than the x - ray value of @xmath0 determined under the above assumptions .
therefore , in a next step , we allowed all parameters ( except ) to be free and checked , whether a dusty warm absorber could be successfully fit at all .
this is not the case ( e.g. , if @xmath0 is fixed to log @xmath2 = 21.2 we get @xmath7 = 40 ) .
the bad fit results can be partially traced back to the ` flattening ' effect of dust .
in fact , if we allow for a steeper intrinsic powerlaw spectrum , with @xmath8 much steeper than the value , a dusty warm absorber with @xmath9 fits the spectrum well ( @xmath7 = 1.2 , tab .
we also analyzed the all - sky survey data of this source and , again , a steep intrinsic spectrum is required . at present , there are several possible explanations for the _ rosat_-spectral differences : ( i ) variability in the intrinsic powerlaw .
we consider this solution unlikely , since the spectrum is steep in all three observations taken at different times .
( ii ) remaining _ rosat_-cross - calibration uncertainties .
this is well possible , given previous reports of a tendency of _ steeper _ spectra when compared with data of the same object .
( iii ) variability in a _
two_-component warm absorber .
there are increasing indications of more than one x - ray warm absorber in mcg6 - 30 - 15 , and might be a similar case . in summary ,
the presence of a _ dusty _ warm absorber in iras13349 + 2438 seems consistent with the data if one allows for some spectral differences between the and data .
the necessity of dust survival then locates the ionized material _
outside _ the bulk of the broad line region ( blr ) of .
clear signatures of the presence of dust are the carbon edge at 0.28 kev and the oxygen edge at 0.56 kev ( not yet individually resolved by current x - ray instruments ) , and the confirmation and study of dusty warm absorbers will certainly be a prime goal of future x - ray satellites like _
axaf _ and _ xmm_. in particular , _ dust_-created absorption edges will play an important role not only in probing components of the active nucleus , like the dusty torus , but also are they a very useful diagnostic of the dust properties in other galaxies .
further , measurements of edge energies and widths will allow to determine the dynamical state of the warm absorber .
brandt w.n .
et al . , 1996 ,
mnras 278 , 326 ( b96 ) ; 1997 , mnras 292 , 407 brinkmann w. et al . , 1996 ,
a&a 316 , l9 ferland g. , 1993 , univ .
kentucky , phys .
report komossa s. , fink h. , 1996 , astron . ges . abs .
12 , r. schielicke ( ed . ) , 228 komossa s. , fink h. , 1997a , a&a 322 , 719 ; 1997b , a&a 327 , 483 komossa s. , fink h. , 1997c , a&a 327 , 555 ( kofi ) komossa s. , 1998 , in ` science with xmm ' , m. dahlem ( ed . ) ; astro - ph/9811049 komossa s. , bade n. , 1998 , a&a 331 , l49 ( koba ) komossa s. , greiner j. , 1999 , in ` high energy processes in accreting black holes ' , j. poutanen & r. svensson ( eds ) , asp conf .
161 , 228 ; astro - ph/9810105
leighly k. et al . , 1997 ,
apj 489 , l137 wills b.j .
et al . , 1992 , apj 400 , 96 | warm absorbers are an important new probe of the central region of active galaxies ( agn ) .
observing and modeling this component provides a wealth of information on the nature of the warm absorber itself , its relation to other components of the active nucleus , and the intrinsic agn x - ray spectral shape .
we briefly review the general properties of dusty warm gas .
for the first time , we then apply such a model to the ir loud quasar .
it was the first to be suggested to host a _ dusty _ warm absorber ( brandt et al .
1996 ) , but has not yet been modeled as such .
.... contribution to the proceedings of ` astrophysical dynamics ' ( evora , april 14 - 16 , 1999 ) ; to appear in ap&ss .
preprint available at http://www.xray.mpe.mpg.de/~skomossa .... |
the helical magnetic dynamo is the basis for a promising class of mechanisms to explain large scale magnetic fields observed in stars and galaxies @xcite .
the basic `` @xmath0 '' dynamo is the specific version most relevant for strongly sheared rotators ( fig . 1 ) .
interface @xmath0 dynamos ( parker 1993 ; charbonneau & macgregor 1996 ; markiel & thomas 1999 ) include the fact that , unlike for galaxies and disks , the dominant shear layer is beneath the dominant turbulent region .
focusing on the simplest `` @xmath0 '' picture ( fig . 1 ) ,
consider an initially weak toroidal ( = encircling the rotation axis ) loop of the magnetic field embedded in the astrophysical plasma rotator . the magnetic field is coupled to the plasma so the field is carried or stretched in response the plasma motion .
now imagine , as in the sun , that there is an outwardly decreasing density gradient .
consider the motion of a rising turbulent swirl of gas , threaded by a magnetic field .
conservation of angular momentum dictates that the swirl will writhe oppositely to the underlying rotation of the system .
this means that the initially toroidal field threading the swirl gains a radial component .
statistically , rising swirls in the northern ( southern ) hemisphere writhe the field clockwise ( counterclockwise ) .
this is the `` @xmath1 '' effect and is shown by the writhed loop of fig .
1a for the northern hemisphere .
differential rotation at the base of the loop shears the radial field ( the `` @xmath2''-effect ) .
the bottom part of the loop amplifies the initial toroidal seed loop as shown in fig .
1b , whilst the top part of the loop diffused away ( the `` @xmath3 '' effect ) .
in doing so , magnetic flux is amplified : the flux penetrating the tilted rectangular surface is zero in fig .
1a , but finite in fig . 1b .
this process is represented mathematically by averaging the magnetic induction equation over a local volume and breaking all quantities ( velocity @xmath4 , magnetic field @xmath5 in alfvn velocity units , and normalized current density @xmath6 ) into their mean ( indicated by an overbar ) and fluctuating ( indicated by lower case ) components .
the result is @xcite : @xmath7 where @xmath8 is the microphysical diffusivity .
the @xmath9 term incorporates the @xmath2-effect , the @xmath3 term incorporates the turbulent diffusion ( assuming constant @xmath3 ) and the first term on the right incorporates the @xmath1-effect . in the kinematic theory @xcite @xmath1
is given by @xmath10 . here
@xmath11 is a turbulent damping time and @xmath12 is the kinetic helicity , which dictates the @xmath1-effect described physically above .
usually , @xmath1 and @xmath3 are prescribed as input parameters .
a long standing problem has been the absence of properly incorporating the ( time - dependent ) backreaction from the growing magnetic field on the driving turbulent motions .
this stimulated criticisms of mean - field dynamos @xcite and motivated interface dynamo models @xcite .
but the backreaction is now much better understood .
steady - state studies of @xmath1-quenching @xcite apply only at fully saturated dynamo regimes , not at early times , when the backreaction is just beginning to be important .
there the growth is fast , and most relevant for astrophysics .
demonstrating this requires including the time - evolution of the turbulent velocity , subject to magnetic forces . carrying this out formally @xcite and using a closure in which triple correlations act as a damping term , amounts to replacing @xmath13 by @xmath14 , where the second term is the backreaction .
it arises from @xmath15 , the force associated with the action of the small scale current and the large scale field .
this residual form of @xmath1 has long been thought @xcite to be the real driver of the helical dynamo and has been employed in attempts to understand its quenching @xcite from the form of @xmath1 , it is evident that a large current helicity can offset the kinetic helicity and quench the dynamo @xcite . in section 2
we summarize the successful backreaction theory , and show how it predicts ejection of twist and writhe of opposite sign . in section 3
we give a new pictorial of dynamo action that includes magnetic helicity conservation , and discuss a simulation of a rising flux ribbon .
observational implications are discussed in section 4 , and we conclude in section 5 .
the principle of magnetic helicity conservation determines the strength of the current helicity correction term in @xmath1 discussed in the previous section .
the magnetic helicity , defined by a volume integral @xmath16 , satisfies @xcite @xmath17 where the magnetic field @xmath18 is related to @xmath19 by @xmath20 , and the current helicity @xmath21 is defined by @xmath22 . without the surface terms ( which represent flow through boundaries ) @xmath23 is well conserved : for @xmath24 the @xmath8 term in ( [ magcons ] ) converges to zero @xcite .
the magnetic helicity is a measure of `` linkage '' and `` twist '' of field lines @xcite .
equation ( [ magcons ] ) then means that in a closed system , the total amount of twist and writhe is conserved .
if the large scale field is writhed one way , then the small scale field must twist oppositely . in the sun ,
differential rotation and cyclonic convection ( the @xmath1-effect ) are both sources of helicity , @xcite but here we focus on the @xmath1-effect , which generates large scale poloidal structures . the @xmath1-effect does not produce a net magnetic twist but produces simultaneously positive and negative magnetic twists on different scales @xcite .
the importance of this scale segregation of @xmath23 for the backreaction term in @xmath1 is easily seen in the two - scale approach .
here we write @xmath25 , where @xmath26 and @xmath27 correspond to the volume integrated large and small scale contributions respectively . for @xmath21
, we then have @xmath28 , where @xmath29 and @xmath30 represent the wavenumbers ( inverse gradients ) associated with the large and small scales respectively , and the second equality follows rigorously for a closed system .
the current helicity backreaction in @xmath31 is thus @xmath32 .
we now relate @xmath33 to @xmath34 .
we define @xmath35 such that the large scale magnetic energy @xmath36 and where @xmath37 , where @xmath38 only for a force - free helical large scale field ( i.e. for which @xmath39 so that the force @xmath40 ) . in the northern hemisphere @xmath41 . by writing conservation equations analogous to ( [ magcons ] ) for @xmath34 and @xmath42 respectively ,
we obtain @xmath43 and @xmath44 where we have used @xmath45 and @xmath46 . the case without surface terms and with @xmath47
represents a dynamo without differential rotation .
the solution @xcite shows that for initially small @xmath42 but large @xmath48 , @xmath34 grows .
growth of @xmath34 implies the oppositely signed growth of @xmath42 .
this @xmath42 backreacts on @xmath48 , ultimately quenching @xmath31 and the dynamo .
since the sun is a differentially rotating open system , differential rotation and surface terms are important .
the former forces @xmath49 , and @xmath35 a function of time to reflect the solar cycle .
the presence of surface terms generally requires the use of the relative magnetic helicity @xcite but to capture the key points , we can instead treat them as diffusion terms @xcite .
we combine the @xmath8 and surface terms of both ( [ h1 ] ) and ( [ h2 ] ) into the forms @xmath50 and @xmath51 respectively .
we consider the volume average @xmath52 to be taken over one hemisphere and assume that surface terms represent diffusion into the corona . on time
scales much shorter than the 11 year solar half cycle , the left sides of ( [ h1 ] ) and ( [ h2 ] ) are negligible and the system is in a relatively steady state .
we then see that the boundary terms are equal and opposite @xcite . here
this implies @xmath53 .
since the boundary diffusion terms represent a flux of ( relative ) magnetic helicity to the exterior , these quantities are connected to measurable observables .
_ we therefore predict that the shedding rates of small scale twist and large scale writhe from the @xmath1 effect are equal in magnitude and opposite in sign . _
the helicity conservation , shedding , and magnetic backreaction are represented simply in figs .
1c e for the northern hemisphere .
the key is representing the field by a two - dimensional ribbon instead of a one - dimensional line . comparing figs .
1a & b ( @xmath23 conservation not included ) with figs .
1c & d ( @xmath23 conservation included ) we see that in the latter , as the @xmath1-effect produces its large scale writhe ( the loop corresponding to gradient scale @xmath54 ) it also twists the ribbon ( corresponding to gradient scale @xmath55 ) .
the large - scale writhe is right handed , but the twist along the ribbon is left - handed and so the total @xmath23 is conserved .
this is the simplest generalization of the picture in fig . 1a , to include helicity conservation .
the ribbon should be thought of as a mean field , averaged over smaller fluctuations , and so the actual field need not be so smooth .
the top view is shown in fig .
1e for comparison to observations . in the northern hemisphere
we expect an n shaped ( or reverse - s ) sigmoid prominence and in the southern hemisphere we would expect an s shaped sigmoid . in fig .
1e we also show the backreaction force corresponding to the small - scale magnetic twist along the ribbon : it fights against writhing or bending the ribbon .
eventually this twisting would suppress the @xmath1 effect ( and thus statistically , its hemispheric average @xmath31 entering ( [ h1 ] ) and ( [ h2 ] ) ) , which thrives on being able to writhe the ribbon . in the sun ,
such sigmoid structures precede cmes @xcite which on time scales of order days or weeks dissipate both the writhe and twist . in doing so
, they help alleviate the backreaction on the @xmath1 effect , and allow a net amplification of magnetic flux inside the sun as shown in fig . 1c . some loops produced by the dynamo
may not escape , implying that some of the simultaneous diffusion of @xmath34 and @xmath42 is hidden in the solar interior .
but even so , the helicity fluxes of @xmath34 and @xmath42 from the loops which do escape , should still be equal and opposite .
getting rid of @xmath42 simultaneously with @xmath34 , either internally or externally , is what alleviates the backreaction , external removal by means of cmes is one important aspect of how this occurs .
removal of @xmath42 implies that by solar minimum the @xmath1 effect is again driven by @xmath56 , allowing the cycle to repeat .
we have performed a numerical simulation to measure the magnetic helicity spectrum of a buoyant magnetic flux ribbon in the presence of rotation which confirms the basic ideas that twist and writhe emerge with opposite sign .
though rising flux tube simulations have been carried out in the past @xcite they have not focused on the magnetic helicity spectrum .
we started with a toroidal , horizontal flux tube and a vertically dependent sinusoidal modulation of the entropy along the ribbon .
this destabilizes the ribbon to buckle and rise in one portion .
the boundaries were sufficiently far away to use a fourier transform to obtain power spectra ( fig .
[ fpspec ] ) .
after @xmath57 free - fall times the spectrum shows mostly positive magnetic helicity together with a gradually increasing higher wavenumber component of negative spectral helicity density .
the latter is the anticipated contribution from the twist along the ribbon .
in comparing the above results with observations note that the magnetic helicity is a volume integral , so the ribbon on which the twisted prominence arises may have a hidden twist elsewhere inside the sun .
this subtlety is surmounted by use of the relative magnetic helicity @xcite which allows a quantifiable interpretation of locally twisted structures .
observations typically measure the current helicity density , @xmath58 , within a single structure , from which hemispheric averages can be computed , or the surface - integrated relative magnetic helicity fluxes .
existing observations seem to be consistent with our basic picture .
first , the observed n sigmoids outnumber s sigmoids by a ratio of 6:1 in the northern hemisphere with the expected reverse relation in the southern hemisphere @xcite .
studies of sigmoids such as gibson et al .
2002 and refs .
therein ) do seem to show qualitative agreement with our picture .
figure 2a of gibson et al .
( 2002 ) shows a trace image of an n - sigmoid ( right - handed writhe ) with left - handed twisted filament of the active region noaa ar 8668 , typical of the northern hemisphere just as we predict .
( as our theory is statistical in nature , it is also not surprising that occasionally n sigmoids such as ar8100 ( as opposed to s sigmoids ) appear in the southern hemisphere @xcite .
but even for ar8100 , the writhe is opposite in sign to the twist along the prominence . )
second , recent work confirms a basic hemispheric dependence of the sign of small scale current helicity , corresponding to the twist along the ribbon in fig .
measurements of small scale current helicity densities and surface - integrated relative magnetic helicity fluxes @xcite as well as fits to line of sight magnetograms of solar active regions @xcite show primarily negative values in the north and positive in the south .
these studies measure the sign of the twist along the ribbon not the writhe of the ribbon itself .
we therefore agree that signs of current helicity measured by these line of sight magnetograms are characteristic of the small scale rather than the large scale field @xcite .
twist is expected at the apex of a writhed prominence since the density is lowest ( parker 1974 ; choudhuri 2002 ) . in sum , fig . 1e , showing an n sigmoid is consistent with the dominant structures of the northern hemisphere .
large scale positive writhe dominates in the north , and large scale negative writhe dominates in the south .
small scale twists along the prominences are predominately negative in the north and positive in the south , so as to produce a very small net helicity in each hemisphere .
this is complementary to demoulin et .
al ( 2002 ) where oppositely signed twist an writhe from shear were shown to be able to largely cancel , producing a small total magnetic helicity .
here we focused on the @xmath1-effect which has the same effect .
finally note that our @xmath33 represents a local averaging over the small scale twist so that @xmath33 has only the writhe ( fig .
1a is thus applicable to @xmath33 , whereas fib 1c shows both @xmath33 and @xmath59 ) . on even larger scales , the globally averaged field computed by an
azimuthal average @xmath60 , is weaker than @xmath61 in a local structure , due to the small filling fraction .
we have suggested that a new understanding of how helical dynamos conserve magnetic helicity may help resolve several mysteries of the solar magnetic field .
we have shown that large and small scale helicities of approximately equal magnitude should be ejected into the solar corona as part of the sustenance of the solar cycle .
the large scale helicity corresponds to the writhe of a prominence , whilst the small scale helicity corresponds to the twist along the prominence .
we emphasize the importance of simultaneously detecting large and small scale contributions to the losses of helical magnetic fields in pre - cme sigmoid structures at the solar surface .
the observations seem to be roughly consistent with our simple picture at present but more studies and detailed modeling will be needed to test these ideas .
our fig .
1 illustrates the basic concepts through a new pictorial representation of the mean field dynamo that includes magnetic helicity conservation and the backreaction .
the implications are also relevant for large scale dynamos in other stars and disks in astrophysics .
eb acknowledges doe grant de - fg02 - 00er54600 , and the dept . of astrophys .
sciences at princeton for hospitality during a sabbatical .
we acknowledge pparc supported supercomputers at leicester and st andrews , and the odense beowulf cluster is acknowledged . | two questions about the solar magnetic field might be answered together once their connection is identified .
the first is important for large scale dynamo theory : what prevents the magnetic backreaction forces from shutting down the dynamo cycle ?
the second question is : what determines the handedness of twist and writhe in magnetized coronal ejecta ?
magnetic helicity conservation is important for answering both questions .
conservation implies that dynamo generation of large scale writhed structures is accompanied by the oppositely signed twist along these structures .
the latter is associated with the backreaction force .
we suggest that coronal mass ejections ( cmes ) simultaneously liberate small scale twist and large scale writhe of opposite sign , helping to prevent the cycle from quenching and enabling a net magnetic flux change in each hemisphere .
observations and helicity spectrum measurements from a numerical simulation of a rising flux ribbon in the presence of rotation support this idea .
we show a new pictorial of dynamo flux generation that includes the backreaction and magnetic helicity conservation by characterizing the field as a 2-d ribbon rather than a 1-d line . * subject headings * : mhd sun : activity sun : magnetic fields mhd turbulence ; stars magnetic fields ; galaxies magnetic fields ; methods |
[ intro ] since the discovery of the high - temperature ( high-@xmath1 ) superconductors in 1986 , there have been many advances in the field of superconductivity due to intense worldwide research efforts .
several sets of ceramic cuprate materials have been synthesized and the transition temperature @xmath1 in some of these has exceeded @xmath2 .
they all share a common quasi - two - dimensional structure and the superconductivity is believed to originate from the electron correlations in the @xmath3 planes . in spite of intensive study of these compounds during the past decade , there is no consensus on a possible mechanism for superconductivity . at the same time , researchers have focused on out - of - plane effects as an explanation in the cuprate superconductors , because all the cuprate families share the same @xmath3 plane , but the maximum @xmath4 varies dramatically from one cuprate family to another @xcite .
this variation is unlikely to be caused by the electronic interactions in the @xmath3 plane , but it is most probably due to different atomic structures between the planes and their indirect effect on the electronic structure in the @xmath3 plane .
recently , it has been argued that the non - planar apical oxygen atoms in the @xmath5 pyramid or @xmath6 octahedron play an important role in the out - of - plane atomic - scale lattice effects in high @xmath7 superconductors ( htscs ) @xcite .
more systematic experimental studies have been done on @xmath0 @xcite ( @xmath8-@xmath9 ) . because of the natural cleavage plane between two @xmath10 layers , the above system has been ideal for probing the planar electronic structure .
this material also has a complicated superstructure modulation ( supermodulation ) which is known to have a strong effect on the position of the apical oxygens but almost no effect on the @xmath3 plane @xcite .
slezak et al .
measured the local superconducting gap using stm with atomic precision and found that the energy gap also has a modulation with the same periodicity as the structural supermodulation @xcite .
superconductivity in cuprate materials is often discussed in the context of the quantum critical point that controls the optimal doping and maximum in the transition temperature @xcite .
most measurements of magnetic correlation length near putative quantum critical point ( qcp ) reveal that it remains finite .
scanning probes ubiquitously reveal electronic nanoscale inhomogeneity ( see ref .
@xcite and references therein ) .
all these findings cause us to question the assumption of a smooth evolution to the quantum critical point .
a possible alternative explanation of these findings is that the electronic matter develops a spontaneously inhomogeneous modulations ( stripes , checkerboards , etc . ) near qcp @xcite .
this propensity to spontaneous charge inhomogeneity can significantly benefit from the electronic modulations induced by the apical oxygen positions . from this perspective
it is natural to expect that apical oxygen position modulations provide a perturbation that leads to in - plane charge and spin modulation and therefore would lead to a spatially modulated superconducting state with finite correlation lengths both for charge and spin excitations .
in addition , scanning tunneling microscopy ( stm ) measurements in underdoped @xmath8-@xmath9 provide information about two energy scales or gaps , coherent vs. incoherent pairings , metal - insulator transition , etc .
driven by temperature and electron concentration @xcite .
for most htscs , the changes in physical properties are obtained by doping carriers through changes of composition .
apart from the usual doping cases , some htscs with bi , such as @xmath8-@xmath9 , can experience self - doping , where the ions in non - conducting layers can act as a charge reservoir , which provides free carriers to the @xmath3 plane @xcite . here
we investigate possible supermodulation of a charge redistribution ( self - doping ) between the charge reservoir and the @xmath3 plane at fixed stoichiometry .
inspired by the observation of gap modulation , several theoretical proposals have been put forward by introducing variations of coupling constants on a phenomenological level , or applying the conventional bcs theory @xcite .
here we report the results of a microscopic study of pairing gap modulation for repulsive electrons within individual unit cells in real space for various spatial displacements of the apical atoms on superconducting pairing . in this paper
we discuss the physical nature of the observed local pairing gap modulation and possible charge redistribution in connection with the structural changes , the change of chemical potential and self - doping effect .
we argue that the superconducting pairing correlations are strongly driven by pairing modulations while charge density and the chemical potential in the supermodulation phase can change slightly with the periodicity of structural changes .
the paper is organized as follows : following the introduction to the various aspects of pairing modulations , we present the microscopic model and methodology of exact cluster calculations of pyramidal structures in section [ method ] . in section
[ result ] , we show the results of charge gap variations and pairing modulation in @xmath0 .
the concluding summary constitutes section [ summary ] . in appendixes , we present the details of calculations related to various aspects of modulation properties .
[ model ] slezak et al . showed a direct link between the spatial modulation of the out - of - plane atomic structure within an individual unit cell and the variation of the local gap maximum @xcite .
the non - planar atoms , especially the apical oxygen atoms , are playing an indirect but important role on modulating the local electronic structure of the @xmath3 plane .
merz et al .
@xcite concluded from their x - ray absorption spectroscopy that increasing @xmath11 dopant can induce holes both on oxygens in @xmath3 plane and the apical oxygen atom .
this result suggests that both the planar and apical oxygens can be electronically connected and might be controlling the local electronic state by exchanging holes between @xmath3 plane and other planes
. this could be the origin of interlayer interactions and local gap variations .
therefore , in order to simulate the spatial gap modulation , we need a model which can treat the local electronic state exactly and simultaneously include an out - of - plane interaction that can exchange electrons / holes with the basal atoms .
this requirement suggests the use of a quantum cluster treatment to deal with local correlations and an extra out - of - plane site to control the electron / hole concentration on the @xmath3 plane .
we begin with the canonical single band hubbard model on a pyramid cluster ( as shown in fig .
[ fig : fig1 ] ) .
@xmath12 here , @xmath13 ( @xmath14 ) is the electron destruction ( creation ) operator at the basal sites with spin @xmath15 ( or magnetic sublevel ) , while @xmath16 ( @xmath17 ) is the same operator at the apical site .
the first term indicates the hopping between the nearest basal sites @xmath18 and @xmath19 ( @xmath20 denotes the nearest neighbor sites ) with the same hopping parameter @xmath21 .
the second term describes hopping between the apical site and the basal site @xmath18 with a hopping amplitude @xmath22 .
in addition , @xmath23 is the on - site coulomb interaction . in practice
, the hopping within the plane should be much bigger than that of the out - of - plane atom , _
i.e. _ @xmath24 .
the energies are measured with respect to @xmath25 , unless otherwise stated .
the cluster can be solved exactly for all the eigenvalues and eigenvectors @xcite .
the apical site can exchange holes with all the basal sites , so displacements of the top site within the cluster will affect the planar electronic structures . in the absence of electron - hole symmetry , the sign of the hopping amplitude between the nearest neighbor sites leads to distinct changes of the electronic structure @xcite . in nonbipartite structures , such as the pyramid or tetrahedron
, the results strongly depend on the sign of @xmath21 and @xmath22 . here
the pyramidal non - bipartite geometry is defined by choosing @xmath25 as the hopping connecting square ( plaquettes ) sites in the basal plane , and we consider various coupling parameters @xmath26 as hoppings between the apical site and the planar site @xmath18 in the square base ( fig .
[ fig : fig1 ] ) . and
the hopping parameters between the top site and basal sites are @xmath27 . here , all the energy parameters are measured in units of @xmath21 . ]
[ method ] an essential element for understanding our results on pairing modulation is the formation of coherent pairing state of @xmath28 electrons with a negative charge gap and positive spin gap obtained earlier in refs .
@xcite and references therein . for completeness
here we first briefly summarize the key results of exact diagonalization in small clusters at weak and moderate @xmath29 values .
the square lattices , generated recently on optimized ( finite ) betts unit cells , provide strong evidence for the phase separation instabilities found in generic @xmath30 , @xmath31 and other clusters @xcite .
numerical ( exact ) calculations of the energy levels @xmath32 for the relevant electron number @xmath28 and a given spin @xmath33 are used to define charge and spin gaps .
the charge excitation gap @xmath34 at finite temperature can be written as an energy difference of canonical energies , @xmath35 , where @xmath36 and @xmath37 .
the charge gap determines the stability of an @xmath38electron state compared to an equal admixture of @xmath39 and @xmath40electron states ( the average number of electrons for this mixture is still @xmath28 ) .
depending on the strength of the on - site electron - electron repulsion @xmath29 , this charge excitation gap can be positive @xmath41 or negative @xmath42 .
the pair binding energy @xmath43 determines whether a total 2@xmath28 electrons on pair of isolated clusters prefer to distributed uniformly ( @xmath41 ) or give rise to local charge inhomogeneity ( @xmath42 ) . for odd @xmath28
this corresponds also to an effective electron pairing , since clusters are more likely have an even number of electrons .
when the charge gap is negative , the system energetically prefers the mixed state with @xmath44 and @xmath45 electron configurations .
the spin gaps @xmath46 , defined in the grand canonical ensemble , can provide essential insights into the response of the system to applied magnetic field @xmath47 .
the magnetic susceptibility @xmath48 is defined as a function of the magnetic field @xmath47 near the critical chemical potential ( say @xmath49 ) and , in the negative charge gap region ( @xmath42 ) , we introduce the spin gap @xmath46 as the magnetic field which gives rise to the first magnetic susceptibility peak , _
i.e. , _ the magnetic field required to drive the system into its lowest exited state . the zeros of canonical and the grand canonical gaps as a function of @xmath29 at @xmath50 describe qcps for continuous and abrupt quantum phase transitions .
the change of sign in canonical gaps at the level crossing point away from half filling is crucial for electron ( charge ) phase separation and pairing instabilities .
as temperature approaches zero , the possible sign change in canonical gap @xmath51 at @xmath52 signifies the existence of a qcp related to first order ( dramatic ) changes .
this qcp describes an abrupt ( discontinuous ) change in the ground state of a many - body system due to strong quantum fluctuations .
the negative charge gap ( @xmath42 ) shows the regions of electron instability with a tendency toward phase separation . for one hole
off half filling the mott - hubbard like insulating ground state for electron - hole pairing with @xmath41 is stable at all @xmath53 , while for @xmath54 , the ground state consists of paired electrons ( @xmath42 ) . this pairing instability which redistributes electrons signals the formation of spatially inhomogeneous , hole - rich @xmath55 and hole - poor @xmath56 regions .
[ instabilities ] many cooperative phenomena invoked in the approximate treatments of the large " concentrated systems , are also seen in exact analysis of pairing instabilities in the canonical and grand canonical ensembles of the small clusters in the thermodynamic equilibrium .
the negative sign of the canonical charge gap ( @xmath42 ) is a signature of phase separation instabilities .
however , instead of a full phase separation , the spin gap rigidity ( @xmath57 ) provokes local inhomogeneities in electron distribution .
the spatial inhomogeneities in the charge distribution imply different electron configurations , close in energy .
the quantum mixing of these hole - rich @xmath55 and hole - poor @xmath56 clusters for one hole off the half filling case provides a stable spatial inhomogeneous media that allows the pair charge to fluctuate .
thus , the negative sign of charge gap implies phase ( charge ) separation of the clusters on hole - poor and hole - rich regions .
we find that at rather low temperatures the calculated positive gap @xmath58 in grand canonical method can have equal amplitude with a negative charge gap @xmath34 derived in the canonical method , _
i.e. _ , with identical gaps , @xmath59 .
such behavior is similar to the existence of a single quasiparticle gap in the conventional bcs state .
we identify such an opposite spin ( singlet ) coupling and electron charge pairing as spin coherent electron pairing .
this is an important characteristic of ( quantum ) phase coherence in the ground state associated with simultaneous coherent pairing of independent charge and spin ( fermion ) entities and the corresponding full bose - einstein condensation of these preformed " bosons with a single energy gap .
such behavior is similar to coherent pairing with a unique quasiparticle gap in the conventional bcs theory .
however , unlike in the bcs theory , the charge gap differs significantly from the spin gap as temperature increases above the superconducting temperature , @xmath1 .
it is important to note that the coherent pairing formation through _ separate _ condensation of bound holes and coupled opposite spins ( both boson entities ) at different critical temperatures @xmath60 and @xmath61 is somewhat different from the cooper pairs ( composite bosons ) condensation with a unique critical temperature .
the magnitudes of the spin gap and charge gap also differ with increasing temperature here , unlike in the bcs case .
this implies the emergence of two gaps and two energy scales or two characteristic critical temperatures leading correspondingly to incoherent pairing of pairs with unpaired ( opposite spins ) far above @xmath1 seen in htscs using stm measurements @xcite .
microscopic spatial inhomogeneities and incoherent pairing detected in these experiments are correlated remarkably with predictions of charged pairs without quantum phase coherence above @xmath61 in nanoscale .
the positive change gap @xmath41 describes charge redistribution and mott - hubbard like transition above @xmath60 .
the positive spin gap calculated in the grand canonical approach implies coherent pairing with a homogeneous spatial distribution below @xmath61 .
this picture is consistent with some stm measurements in @xmath62 @xcite .
[ result ] [ sec1 ] high temperature superconductors have complex layered structures . for example , there are various layers , bilayers or trilayers that affect the superconductivity attributed to the cu - o planes acting as _ reservoirs _ of electrons or holes . here
we attempt to model such complex structures with a simple pyramid having a square base and and an out - of - plane atom . as discussed in our previous publications ,
a negative charge gap @xmath34 is essential for having coherent pairing states at zero temperature @xmath63 .
we first consider the simplest case to testify the validity of the model : a symmetric pyramidal cluster with equal hopping terms @xmath64 @xmath65 between the apex atom and the all the basal atoms .
the charge gap @xmath34 at @xmath63 as a function of coupling parameter @xmath66 in fig .
[ fig : fig2 ] is shown at various @xmath29 values and low temperatures , @xmath67 ( in units of @xmath21 ) close to optimal doping for one hole of half filling , @xmath68 ( @xmath69 ) .
the energy gap vanishes in a multiparameter space of @xmath29 and @xmath66 , where its sign changes define the quantum critical points for electron charge and spin pairing instabilities . in the fig .
[ fig : fig2 ] plots with @xmath70 provides strong evidence for the existence of a qcp of level crossing instability , associated with phase separation instability .
for example , the charge gap at @xmath71 vanishes ( @xmath72 ) at @xmath73 .
this qcp describes an abrupt change in the ground state of a many - body system due to strong quantum fluctuations , _
i.e. , _ a transition from electron pairing into mott - hubbard behavior . the negative charge gap ( @xmath42 ) at @xmath74 corresponds to coherent electron pairing @xmath75 in a vertically deformed square pyramid ( with @xmath76 in the basal plane ) , while for @xmath77 , the mott - hubbard insulating ground state for electron - hole pairing with @xmath41 is stable . as shown in fig .
[ fig : fig2 ] , as the hopping @xmath66 increases but still small ( for example , @xmath78 for @xmath71 ) , the value of the charge gap changes very slowly , which corresponds to the case with the top site far from the basal plane .
the result indicates that the presence of the top site induces very small perturbations on the electronic state at the planar electronic state when it is far away .
however , when @xmath66 is large enough ( for example , @xmath79 for @xmath71 ) , there is a dramatic change and the charge gap becomes positive .
this illustrates the case with the top site close enough to the base , so the electronic state on the top site greatly affects the electronic state on the basal plane and tends to destroy the electron pairing on the basal as the apical atom gets closer to the plane . comparing with results of the elementary square geometries @xcite
, we can conclude that the electronic structure of the basal sites contributes significantly to the formation of electron pairs , while the apical site tries to destroy such pairing by sharing " holes / electrons with the basal sites .
this result is consistent with the conclusion of slezak et al .
@xcite who stated that the pair density " is anti - correlated with the height of apical @xmath11 atom .
given the consistency of this simple result , the pyramid may be a good candidate for studying local gap modulations .
the pairing gap has the maximum magnitude and the widest range at @xmath71 , so we here consider the gap modulation for only @xmath71 .
another feature of the pairing gap behavior in the pyramidal structure is that when the variations of @xmath66 are small , the gap varies linearly with a very small slope .
it suggests that the apical effects at small @xmath66 values can be treated by perturbations .
this feature resembles the effect of small next nearest neighbor hopping in the square lattice . actually , the two effects are equivalent when @xmath66 is small ( see appendix [ s3 ] ) .
this equivalence indicates that the influence of a far - away top site is like a bridge connecting the next nearest neighbor sites so that it perturbs the electronic state at the plane but does not significantly destroy the paired state .
slezak et al .
observed a gap modulation of about @xmath80 in their experiments which is large enough to address the destruction of electron pairs @xcite .
such a big gap modulation is not suitable for a perturbation treatment . . when hoppings between the apical site and the base sites are all the same @xmath81 , the charge gap @xmath34 is calculated as a function of @xmath66 at several @xmath29 values
the region @xmath42 has a ground state with coherent pairing @xmath82 . at relatively small @xmath66 ,
the gap remains almost constant , which indicates that when the apical site is far from the base , the electronic state on the apical site can not affect the pairing state on the base .
when @xmath66 is large enough , the dependence of @xmath83 becomes rather steep and beyond the critical value ( @xmath84 ) gap @xmath34 becomes positive .
this implies that coupling between the top and base ( when apical site gets closer to the plane ) can destroy electron pairing at the base .
note that at @xmath71 , the system has the largest magnitude of the negative gap covering the widest range of @xmath66 .
] is plotted as a function of the vertical displacement of the apical site from its equilibrium position at various @xmath85 values ( @xmath85 is the gradient of hopping @xmath86 on the vertical direction ) .
the hopping parameter between apex and base at the equilibrium position is @xmath87 .
the plot shows two regions with different linear slopes .
the apical site should be close enough to the base so that it can provide sufficient sufficient effect on the electronic state on the base . in the figure
, this requires @xmath88 to have a large slope . as @xmath85 increases
, the linear regime with higher becomes smaller .
for @xmath89 , the linear range with higher slope up to about @xmath90 yields a modulation amplitude of about @xmath91 of the gap at equilibrium position which means that @xmath92 is the suitable region for the simulation . ] as a function of the planar @xmath93 coordinates of the apical site assuming @xmath94 ( vertical coordinate ) is fixed .
the gradient of hopping @xmath86 on the @xmath95 and @xmath96 direction of the basal plane @xmath97 is set to @xmath98 following the result in fig .
[ fig : fig3 ] and at the equilibrium position @xmath99 , which is the center of the basal plane , the hopping between apex and base is set to @xmath100 .
the gap modulation along the horizontal direction is not linear .
the amplitude of changes of the gap is about @xmath91 of the gap magnitude at the equilibrium position .
the amplitude is close to that in vertical displacements , this suggests that the abilities to interact with basal electronic states caused by vertical and horizontal displacements are almost the same .
however , considering the symmetry of lattice , the displacements on the horizontal direction are much more difficult , so apical site mainly oscillates along the vertical direction . ]
[ displacement ] from the analysis of the previous section , we conclude that the best simulation conditions for the charge gap are @xmath101 and @xmath102 .
the next step is to find out the position dependence of the hopping parameters .
the hopping parameter between the top and the base is closely related to the distance between the apical site and the basal site , _ i.e. _ @xmath103 , where @xmath104 is the distance between the apical site and the basal site @xmath18 . in an ideal crystal , there is an equilibrium position @xmath104 with the hopping parameter @xmath105 that makes the system most stable .
when the layers of atoms are present above the cu - o planes , the position of the top site can be perturbed with a lattice distortion represented by a very small displacement @xmath106 .
the new hopping @xmath107 can be expanded as @xmath108 where @xmath109 is the slope of @xmath86 : @xmath110 , so the value of @xmath109 is strongly dependent on the functional form of @xmath111 and the value of @xmath104 .
notice , when the variation of the position of the top site is relatively small , the change of the hopping parameter can be well approximated by a linear function . by adjusting the value of @xmath109 and range of @xmath106 , we can easily model a @xmath80 gap variation .
the variation of the apical site can be the displacement along vertical and/or horizontal directions .
a vertical displacement @xmath88 is equivalent to the case discussed in the sec .
[ sec1 ] except for converting the dependence of the hopping parameters to that of a spatial displacement . because @xmath112 , @xmath113 where @xmath114 .
@xmath85 is the gradient of hopping @xmath86 along the vertical ( @xmath94 ) direction .
the pairing gap can be calculated as a function of the vertical displacement based on eq .
( [ eqn : perth ] ) .
we choose @xmath86 to be @xmath115 and confine the change of the position of apical site @xmath116 .
[ fig : fig3 ] plots the charge of gap @xmath34 as a function of @xmath88 at different @xmath85 values .
in order to simulate the sinusoidal gap behavior , we consider @xmath117 as a linear function of @xmath88 . the plots in fig .
[ fig : fig3 ] show two regions with different slopes , but only in the region with higher slope , the apical atom is close enough to interact with the basal electronic pairing states .
therefore the useful region for our simulation is only confined to a small region of displacement @xmath88 . as discussed in appendix [ s5 ] , we need to have the range of linear function smaller than @xmath118 in the simulation . according to fig .
[ fig : fig3 ] , at @xmath89 , the linear range with higher slope up to ( approximately ) @xmath90 achieves an amplitude about @xmath91 of the equilibrium gap magnitude , which is close to the result in the experiment ( @xmath80 ) , so @xmath92 is probably the suitable region for @xmath85 in the simulation . at fixed @xmath47 the horizontal distortion
provide the relationship between the distance change of apical atom with regard to the atoms in basal plane or corresponding planar @xmath119 and @xmath120 displacements @xmath121 and the corresponding hopping parameters between the apex and the basal sites are reduced to @xmath122 where @xmath123 .
@xmath97 is the gradient of hopping @xmath86 on the @xmath95 and @xmath96 direction of the basal plane .
we confine the change of @xmath95 and @xmath96 to a small range @xmath124 and @xmath125 and @xmath86 to be @xmath115 for the same reason .
we calculated the charge gap of the system at different horizontal positions @xmath93 with @xmath126 in fig .
[ fig : fig4 ] . the gap behaviors with horizontal displacements are not linear and the amplitude of the gap change is also about @xmath91 of the gap magnitude at @xmath99 position . by comparing fig .
[ fig : fig3 ] and fig .
[ fig : fig4 ] , the displacements on vertical and horizontal directions have almost the same effects on the basal electronic states , so the gap change due to the planar displacements appears to be equally important to that caused by vertical distortions .
the simulation become rather cumbersome by combining linear vertical and nonlinear horizontal gap behaviors . however , the measurements clearly show the sustained square atomic structure in the @xmath3 plane while other layers are strongly distorted by supermodulation @xcite .
this implies that the planar structure of our model should somehow preserve a square symmetry . therefore , the horizontal displacements of the apical site must be strongly limited to maintain such a square symmetry . since the vertical distortions basically do not influence the planar square symmetry , then only vertical @xmath88 displacements of the apical site are essential in practice for simulation of pairing gap modulation .
is defined as the phase of lattice structure supermodulation which can be expressed as a sine wave function .
the black diamonds with error bars are experimental data with uncertainties from ref .
the red triangles are fitted data from our model .
the fitting gives the following parameters for modeling : @xmath127 , @xmath128 and @xmath129 .
note that there is an excellent agreement between experiment and the results from the simple microscopic model . ]
below we show the results of numerical calculations of pairing gap variation within individual unit cells for various displacements of the apical sites on superconducting pairing .
the pairing ( sinusoidal variation of gap ) modulation has been found in @xmath0 using scanning tunneling microscopy .
the local pyramidal structures affect the superconducting crystalline supermodulation " of the energy - gap in bulk @xmath0 .
such supermodulation displacements can be seen for dopant apical oxygen displacement in both the a " and c " directions and tilt the @xmath5 pyramid .
the local gap modulations here can merely be related to the dopant density modulation with the same period as the supermodulation .
the @xmath5 pyramidal coordination of oxygen atoms surrounding each copper atom in @xmath0 ( see fig . 1 in ref .
@xcite ) is equivalent to the square pyramid @xmath3 geometry built on the basal @xmath130 squares by a rotation of 90 degrees with respect to the vertical axis ) .
thus a translation along a " direction for @xmath5 in @xmath0 is equivalent to displacements along the diagonal direction in the basal cu unit cell .
the local supermodulation displacements in both the a " and c " directions are equivalent to vibrational distortions along the diagonal directions in the square pyramid and vertical c " direction .
experimental results show explicitly that variations found near the dopant atoms as the primary effect of the interstitial dopant atom which is to displace the apical oxygen so as to diminish the distance
d " between atoms or distort ( slightly tilt ) the @xmath5 pyramid .
we apply our exact results on the square pyramids near the optimal @xmath71 values to fit into the data for the experimentally observed variation of gap with the unit cell distortions .
we study the impact of the apical atom on superconductivity by monitoring the local negative gap as a function of the vertical displacement @xmath88 of the apical atom from its average position .
experiments have shown that the modulation of the lattice structure can be expressed as a sinusoidal wave function with respect to a phase @xmath131 which is defined according to the period of the supermodulation . to simulate the modulation
, we introduce the modulation of top site along the vertical direction : @xmath132 , where @xmath133 is the equilibrium position of the apical atom where the hopping between top and base is set to @xmath134 .
@xmath135 here is a phase according to the period of the modulation of the top site and @xmath136 is a constant phase shift .
we use the absolute value of charge gap as the pairing gap , so we can derive the pairing gap modulation @xmath137 as a function of @xmath135 and by properly setting parameters we can fit our results into experimental data . the fitting with experimental observation of the sinusoidal gap behavior is shown in fig .
[ fig : fit ] for optimally doped conditions .
the details of the fitting procedure are given in appendix [ s5 ] .
our calculations are consistent with results in the ref .
@xcite which shows a @xmath80 modulation of average gap value for optimally doped @xmath0 .
the fitting parameters are @xmath128 and @xmath127 .
@xmath85 , one of the free parameters , is within the range @xmath92 so it is consistent with our discussion in the sec .
[ displacement ] .
the value of @xmath21 , which is the other free parameter , is close to values derived from other theoretical methods @xcite .
this provides additional evidence for the validity of our model .
one interesting question has to do with the intrinsic origin of this pairing gap modulation .
there are several possible answers here : the modulation of ( 1 ) electron / hole pair strength , ( 2 ) the electronic density and ( 3 ) the electron / hole pair density . in order to verify the first option
, we calculate the change of the chemical potential @xmath138 of the system with the modulation , where @xmath138 is calculated as an average of @xmath139 and @xmath140 ( defined in sec .
[ method ] ) : @xmath141 .
the chemical potential represents the average energy needed to break an electron / hole pair , so it shows the average binding energy of the hole / electron pairs .
we find that modulation amplitude of @xmath138 is smaller than @xmath142 of the average chemical potential .
this result indicates that although the pairing gap experiences a strong modulation , the binding energy of the hole / electron pairs is not affected , i.e. , the pairing gap modulation is a strong coherent effect not strictly relevant to the modulation of ( hole / electron ) pairing strength .
we have also studied the change of electron density on the apical site with modulation .
when the apical site is very far away from basal plane ( all @xmath143 ) , the gap is maximum and there is exactly one electron on the apical atom , i.e. , decoupling of the apical atom with the plane restores the two - dimensional character of electron pairing . for our simulated case , when the apical site is spatially close to the plane , we find that the electron density of the top site increases , which supports the experimental observation that the dopant on oxygen can provoke a transfer of holes / electrons on both @xmath3 plane and @xmath11 atoms .
however , our calculations show that the amplitude of modulations of the electronic density on apical site is smaller than @xmath144 of the average value .
this result indicates that the electronic density on the pairing plane is insignificantly affected by modulation of the out - of - plane structure .
therefore , the gap modulation does not originate from the modulated electronic density .
therefore , we propose the modulation of electron / hole pair density is the source of the pairing gap modulation .
this kind of modulation of pair density is somewhat similar to the hole - pair density wave proposed in ref .
[ summary ] the effects of disorder in the position of the apical atom outside the cuo@xmath145 plane on the superconducting gap have been investigated in @xmath0 ( bi2212 ) within the framework of small single band hubbard clusters .
we have shown that exact cluster calculations in appropriate geometries can quantitatively describe the displacements of the atoms in the bulk of @xmath0 material with its well - known supermodulation behavior .
our calculations at low temperatures can be used to fit and understand experimental data under variation of the position of the apex atom in the vertical c " direction or along a " , b " or square diagonal directions .
independent of the character of the quasiperiodic distortion , we find modulation in the variation of the pairing gap around its central position .
the theory provides strong evidence that the supermodulation is correlated with the out - of - plane displacement of the apical site and shows how variations of the inter - atomic distances affect the coupling and maximum of the superconducting energy gap .
the current , exact solution - based study offers strong evidence that the inhomogeneity observed in the stm experiments is driven by distortions of the dopant oxygen atoms , located away from the cuo@xmath145 plane , whose primary effect is to suppress local pair formation . the structural changes in our microscopic model in the vicinity of qcp provide the locally enhanced pairing modulation where the nanoscale variation of the pairing strength in the superconducting state have a dramatic effect on the pairing gap .
@xmath0 type materials contain the cuo@xmath145 planes and a `` reservoir '' of carriers in other layers .
a charge redistribution may occur between @xmath3 plane and nonplanar atoms driven by supermodulation , i.e. , out - of - plane displacements of the apical atom .
usually one discusses the charge redistribution by changing composition , such as oxygen stoichiometry , etc .
thus , one of the most direct ways to determine the structural changes in couplings can be done through a detailed analysis of the charge distribution and the valence states of the respective out - of- and in - plane ions . in particular , the position of the apex oxygen or the corresponding bond length ( coupling ) is a sensitive probe of these effects . to summarize ,
this effect may explain structural changes such as the shift of the apical oxygen atom under modulation , and therefore one would expect that there should be a charge density modulation accompanying the formation of a modulated gap in @xmath0 type materials .
however , our results strongly suggest that there should not be significant charge transfer from the reservoir into the superconducting plane .
probably the most direct confirmation of the rigidity of charge density distribution can be obtained for the hole concentration from positron annihilation experiments or x ray absorption spectra @xcite .
the gap modulation or corresponding quantum oscillations in @xmath8-@xmath9 can be attained by tuning pressure or doping due to the direct access they provide to the structural changes that can be traced in the vicinity of qcp .
these calculations can also be extended to finite temperatures , different doping levels or electron concentrations or can be applied to slightly different ( octahedron , tetrahedron , etc . ) geometries .
the obvious next step is to see how the charge gap variations change with temperature and its relationship with the critical temperature t@xmath146 of the cuprate superconductors to explore methods of enhancing and maximizing t@xmath146 . in a nutshell ,
the spectroscopy assessed by cluster studies of repulsive electrons possesses both energy and symmetry information on the local pairing interaction and consequently contains the essential elements for understanding the variations of superconducting gap , coherent pairing and modulation in inhomogeneous , large systems . heavily relying on gained insights from superconductivity effects driven by out - of - plane distortions of the apical atoms
, one can understand also the role of local out - of - plane impurities tied to the mechanisms of superconductivity in other layered high t@xmath146 superconductors , impurities in graphene intercalated compounds , etc .
we are grateful to a. bishop , j.c .
davis , k. fujita , h. she , and j.x .
zhu for useful discussions .
the authors acknowledge the computing facilities provided by the center for functional nanomaterials , brookhaven national laboratory supported by the u.s .
department of energy , office of basic energy sciences , under contract no.de-ac02-98ch10886 .
the work was performed also , in part , at the center for integrated nanotechnologies , a u.s .
department of energy , office of basic energy sciences user facility at los alamos national laboratory ( contract de - ac52 - 06na25396 ) and sandia national laboratories ( contract de - ac04 - 94al85000 ) .
< 11 > h. eisaki , n. kaneko , d. l. feng , a. damascelli , p. k. mang , k. m. shen , z .- x .
shen , and m. greven , phys .
b 69 , 064512 ( 2004 ) hua zhou , yizhak yacoby , vladimir y. butko , gennady logvenov , ivan boovi , and ron pindak , proc .
107 , 8103 ( 2010 ) m. merz , n. ncker , p. schweiss , s. schuppler , c. t. chen , v. chakarian , j. freeland , y. u. idzerda , m. klser , g. mller - vogt , and th .
wolf , phys .
80 , 5192 ( 1998 ) y. ohta , t. tohyama , and s. maekawa , phys . rev .
b 43 , 2968 ( 1991 ) x. b. kan , s. c. moss , acta crystallogr .
b48 , 122 ( 1992 ) j. lee , k. fujita , k. mcelroy , j. a. slezak , m. wang , y. aiura , h. bando , m. ishikado , t. masui , j .- x .
zhu , a. v. balatsky , h. eisaki , s. uchida , and j. c. davis , nature 442 , 546 ( 2006 ) j.a .
slezak , j. lee , m. wang , k. mcelroy , k. fujita , b.m .
andersen , p.j .
hirschfeld , h. eisaki , s. uchida , j.c .
davis , proc .
105 , 3203 ( 2008 ) subir sachdev , science 288 , 475 ( 2000 ) s. l. sondhi , s. m. girvin , j. p. carini and d. shahar , rev .
69 , 315 ( 1997 ) jian - huang she , jan zaanen , alan r. bishop , and a.v .
balatsky , phys . rev .
b 82 , 165128 , ( 2010 ) s. e. sebastiana , , n. harrison , m. m. altarawneh , c. h. mielke , r. liang , d. a. bonnc , w. n. hardyc , and g. g. lonzaricha , proc .
107 , 60175 ( 2010 ) k. k. gomes , a. n. pasupathy , a. pushp , s. ono , y. ando and a. yazdani , nature * 447 * ( 2007 ) 569 m. c. boyer , w. d. wise , k. chatterjee , ming yi , takeshi kondo , t. takeuchi , h. ikuta and e. w. hudson , nat
. phys . * 3 * 802 ( 2007 ) k. tanaka , _ et al_. , science , * 314 * 1910 ( 2006 ) w. s. lee , _ et al_. , nature ( london ) * 450 * 81 ( 2007 ) m.c .
boyer , nat . phys . * 3 * 802 ( 2007 ) d. i. khomskii , arxiv : cond - mat/0101164v1 ( 2001 ) tamara s. nunner , brian m. andersen , ashot melikyan , and p. j. hirschfeld , phys .
rev . lett .
95 , 177003 ( 2005 ) brian m. andersen , p. j. hirschfeld and james a. slezak , phys .
b 76 , 020507(r ) ( 2007 ) k .- y .
yang , t. m. rice , and f .- c .
zhang , phys .
b 76 , 100501(r ) ( 2007 ) y. he , s. graser , p. j. hirschfeld , and h .-
cheng , phys .
b 77 , 220507(r ) ( 2008 ) t. machida , t. kato , h. nakamura , m. fujimoto , t. mochiku , s. ooi , a. d. thakur , h. sakata , and k. hirata , phys . rev .
b 84 , 064501 ( 2011 ) a. n. kocharian , g. w. fernando , k. palandage , and j. w. davenport , phys .
b * 74 * 024511 ( 2006 ) a. n. kocharian , g. w. fernando , k. palandage , and j. w. davenport , phys . rev .
b 78 , 075431 ( 2008 ) a.
n. kocharian , g. w. fernando , c. yang 2010 _ spin and charge pairing instabilities in nanoclusters and nanomaterials , in scanning probe microscopy in nanoscience and nanotechnology _ ( springer ) p 507 - 570 .
a. n. kocharian , g. w. fernando , k. palandage and j. w. davenport , phys .
lett . a 373 , 1074 ( 2009 ) g. w. fernando and k. palandage , a. n. kocharian , j. w. davenport , phys .
b 80 , 014525 ( 2009 ) kun fang , g. w. fernando , a. n. kocharian , phys .
lett . a 376 , 538 ( 2012 ) .
a. n. kocharian , kun fang , g. w. fernando , j. magn .
magn . mater .
324 , 3427 ( 2012 ) a. k. mcmahan , richard m. martin and s. satpathy , phys . rev .
b 38 , 6650 ( 1988 ) d. m. newns , p. c. pattnaik , and c. c. tsuei , phys . rev .
b 43 , 3075 ( 1991 ) han - dong chen , jiang - ping hu , sylvain capponi , enrico arrigoni , and shou - cheng zhang , phys .
89 , 137004 ( 2002 ) y. c. jean , c. s. sundar , a. bharathi , j. kyle , h. nakanishi , p. k. tseng , p. h. hor , r. l. meng , z. j. huang , c. w. chu , z. z. wang , p. e. a. turchi , r. h. howell , a. l. wachs , and m. j. fluss , phys .
* 64 * , 1593 ( 1990 ) .
yasuharu hirai , izumi waki , atsushi momose , tokuumi fukazawa , toshiyuki aida , kazumasa takagi and tatsumi hirano , phys .
rev . b,*45 * , 2573 ( 1992 ) a. bianconi , m. lusignoli , n. l. saini , p. bordet , a. kvick and p. g. radaelli , phys .
b 54 , 4310 ( 1996 )
[ s3 ] the effect on the charge gap of a small apical hopping is similar to that of a small next nearest hopping .
here are a little simple investigates of the relation between these two small extra hopping based on the first order perturbation . for the simplest case , assume all the apical hoppings are the same @xmath86 , and @xmath147 is used to denote next nearest hopping .
both @xmath86 and @xmath147 are small relative to @xmath21 .
assume that the ground state of a four - site cluster without next nearest neighbor is @xmath148 and the ground state of an isolated apical atom is @xmath149 , so the first order correction due to the next nearest hopping is @xmath150 where @xmath151 means summation over all the next nearest neighbors .
while , the first order perturbation due to the apical hopping is @xmath152 where @xmath153 is the @xmath154 state of the top atom . comparing eq . [ eq : apc ] and eq . [ eq : nnn ] , @xmath155 multiplication is the only difference . if @xmath156 , the two equation are equivalent . usually , @xmath153 can be state with no electron @xmath157 , one up spin @xmath158 , one down spin @xmath159 , or two spins @xmath160 and these four states build up a complete basis , so @xmath161 .
however , @xmath162 only include three terms @xmath157 , @xmath158 , and @xmath159 , so @xmath163 therefore , as long as the ground state of apical atom is not a two spin state , the effect on the ground state energy within the first order correction due to the apical atom is same as the next nearest hopping .
it is hard to determine all the cases that the condition is satisfied , but some specific cases can be empirically found : ( @xmath164 ) when hopping between the apex and the plane is small , the holes have very small probability to occupy the apical atom at the same time ; ( @xmath165 ) when the electron interaction is strong enough , the double occupancy is prohibited , but this case also causes the spin frustration because of the non - bipartite structure .
[ s5 ] the experiment ( see @xcite ) shows a cosinusoidal superconducting gap behavior .
the average gap value at optimal doping is about @xmath166mev and the ( peak - to - peak ) amplitude of gap modulation is about @xmath167 of the average gap value .
the sinusoidal behavior is caused by the super modulation in the structure of the bscco system .
the detail of the modulation structure can be found in ref .
@xcite . from these papers ,
the @xmath3 layer does not change too much from the original square structure , but the @xmath168 and @xmath10 layers are strongly distorted .
therefore , the modulation is mainly coming from the changes of @xmath168 and @xmath10 layers and the modulation of the @xmath8 and @xmath169 atoms are found to be combination of sinusoidal waves . in our model , the effects of the @xmath168 and @xmath10 layers
are included in the apical site of the pyramid hubbard model .
the distortion of the two layers can be simulated by the displacement of the apical site along vertical and horizontal direction . as mentioned in sec .
[ displacement ] , only the vertical displacement is essential in our model due to the requirement of a strong square symmetry , so the horizontal distortions can be ignored .
the equilibrium hopping term @xmath170 between top and base is set to @xmath134 at which the pairing gap is about @xmath171 . in the experiment
the average gap value is @xmath172 , so @xmath128 .
experiments @xcite shows that the vertical amplitudes of the modulation layers are about @xmath173 where @xmath66 is the vertical lattice constant .
the amplitude of vertical displacement of the apical site in our model should be much smaller than experimental results because the simplified apical site averages in matter of fact describe displacements of several ( coupled ) layers in real material . in the paper
, we set the amplitude to @xmath174 .
the fitting with experimental results also provides @xmath175 and @xmath176 by the amplitude and the phase shift of gap modulation . | there is growing evidence that the unconventional spatial inhomogeneities in the doped high - tc superconductors are accompanied by the pairing of electrons , subsequent quantum phase transitions ( qpts ) , and condensation in coherent states .
we show that these superconducting states can be obtained from phase separation instabilities near the quantum critical points .
we examine electron coherent and incoherent pairing instabilities using our results on exact diagonalization in pyramidal and octahedron hubbard - like clusters under variation of chemical potential ( or doping ) , interaction strength , temperature and magnetic field .
we also evaluate the behavior of the energy gap in the vicinity of its sign change as a function of out - of - plane position of the apical oxygen atom , due to vibration of apical atom and variation of inter - site coupling .
these results provide a simple microscopic explanation of ( correlation induced ) supermodulation of the coherent pairing gap observed recently in the scanning tunneling microscopy experiments at atomic scale in @xmath0 .
the existence of possible modulation of local charge density distribution in these materials is also discussed . |
electromagnetic waves in a material medium propagate according to maxwell s equations complemented by certain relations linking strengths and induced fields the constitutive relations . depending on the dielectric properties of the medium and also on the presence of applied external fields , a variety of optical effects can be found .
one of wide interest is the birefringence @xcite , occurring when electromagnetic waves propagate in media exhibiting two distinct refractive indexes @xcite in a same wave vector direction .
this effect occurs naturally in some well known crystalline solids , as quartz and sapphire for instance . in nonlinear media , where the dielectric coefficients are field dependent , birefringence could also be induced by the presence of external fields , leading to the known kerr , cotton mouton and magnetoelectric effects @xcite .
birefringence has been widely used in technology of optical devices @xcite and as a powerful experimental technique for investigating properties of physical systems , including biological @xcite and astrophysical ones @xcite , among others .
the phenomenon of triple refraction has been much less investigated . by trirefringence in a given wave vector direction , we mean the existence of three distinct refractive indexes in that direction . in the realm of linear electrodynamics
, trirefringence does not occur @xcite .
nevertheless , nothing was studied when the dielectric tensors are field dependent ( nonlinear electrodynamics ) .
multirefringent properties have been measured in tailored photonic crystals @xcite .
such materials are constructed to manipulate light propagation and hence can lead to a variety of applications @xcite .
however , in these media , structural details ( lattice constants , defects , etc ) are imperative and hence they can not be described in terms of effective dielectric tensors @xcite .
developments in photonic band gap materials and the so - called metamaterials have enabled the discovery of several new phenomena @xcite .
for instance , it was experimentally shown @xcite that nearly transparent isotropic metamaterials allow light propagation when both effective dielectric coefficients ( permittivity and permeability ) are negative .
in fact , this phenomenon and other unusual properties displayed by isotropic media with negative dielectric coefficients were long ago proposed theoretically @xcite .
it is worth emphasizing these media must be dispersive and the negative coefficients are obtained for convenient frequency ranges . some other unusual properties exhibited by specific metamaterials are negative refractive index @xcite , trapping of light @xcite , perfect lens devices @xcite , the electromagnetic cloaking effect @xcite and the occurrence of asymmetry for propagation of light in opposite wave vector directions @xcite .
wave propagation in indefinite metamaterials ( where not all the principal components of the dielectric tensors have the same sign ) has been also considered @xcite , showing that effects already proposed in the context of isotropic metamaterials can be obtained and possibly improved .
indefinite metamaterials can also be used for investigating certain aspects of general relativity @xcite . in this paper , we show that nonlinear metamaterials described in terms of effective dielectric tensors , may display trirefringence .
analytical expressions describing this effect are formally obtained from maxwell s electromagnetism and a simple theoretical model is numerically examined .
a possible experimental realization of the media expected to display this effect is also addressed .
the vectorial three dimensional formalism @xcite is used .
the units are set such that @xmath0 . in the next section electromagnetic wave propagation in nonlinear materials
is examined .
the eigenvalue problem is stated and formally solved for a class of nonlinear materials presenting non - isotropic but constant permittivity tensor ( for a given frequency ) and isotropic permeability dependent on the modulus of the resultant electric field . in sec .
iii , trirefringence phenomenon is theoretically described and proposed to occur in nearly transparent nonlinear uniaxial metamaterials . the analysis is performed in terms of phase and group velocities .
final remarks and conclusions are presented in sec .
the electrodynamics of a continuum medium at rest in the absence of sources is governed by the maxwell field equations @xmath1 taken together with the constitutive relations between the fundamental fields @xmath2 and @xmath3 , and the induced ones @xmath4 and @xmath5 , written here as @xmath6 the dielectric coefficients @xmath7 and @xmath8 are the components of the permittivity and the inverse permeability tensors , respectively , and they encompass all information about the electromagnetic properties of the medium .
further , for any vector @xmath9 we denote its @xmath10th component by @xmath11 ( @xmath12 ) .
the propagation of monochromatic electromagnetic waves is here examined within the limit of geometrical optics @xcite using the method of field disturbances @xcite .
this method can be summarized as follows @xcite .
let @xmath13 , defined by @xmath14 , be a smooth ( differentiable of class @xmath15 ) hypersurface .
the function @xmath16 is understood to be a real - valued smooth function of the coordinates ( @xmath17 ) and regular in a neighborhood @xmath18 of @xmath13 .
the spacetime is divided by @xmath13 into two disjoint regions @xmath19 , for which @xmath20 , and @xmath21 , corresponding to @xmath22 .
the discontinuity of an arbitrary function @xmath23 ( supposed to be a smooth function in the interior of @xmath24 ) on @xmath13 is a smooth function in @xmath18 , and is given by @xcite @xmath25_{\sigma } \doteq \lim_{\{p^\pm\}\rightarrow p } \left[f(p^+ ) - f(p^-)\right],\ ] ] with @xmath26 and @xmath27 belonging to @xmath28 and @xmath13 , respectively .
the electromagnetic fields are supposed to be smooth functions in the interior of @xmath21 and @xmath19 and continuous across @xmath13 ( @xmath16 is now taken as the eikonal @xcite of the wave ) .
however they have a discontinuity in their first derivatives , such that @xcite @xmath29}_{\sigma}&\!\!\!=\!\!&\omega e_i\,,\;\;\;\;\;\ ; { \left[\partial_{t}b_i\right]}_{\sigma}=\omega b_i\ , , \label{q1}\\ { \left[\partial_i e_j \right]}_{\sigma}&\!=\!&-q_i e_j,\;\;\ ; { \left[\partial_i b_j \right]}_{\sigma}\,\!=\!-q_i b_j , \label{q2}\end{aligned}\ ] ] where @xmath30 and @xmath31 are related to the derivatives of the electric and magnetic fields on @xmath13 , and correspond to the components of the polarization of the propagating waves @xcite .
the quantities @xmath32 and @xmath33 are the angular frequency and the _ i_th component of the wave vector .
( incidentally , we note that the negative signs appearing in eq .
( [ q2 ] ) are missing in the corresponding equations in ref .
@xcite . ) for the cases of interest in this work , the permittivity of the media under study will be described by real diagonal tensors ( losses have been neglected ) whose components @xmath7 are dependent only upon the constant frequency of the wave .
we set the magnetic permeability of these media to be real functions of the modulus of the electric field , such that @xmath34 where @xmath35 .
thus , applying the boundary conditions stated by eqs .
( [ q1]-[q2 ] ) to the field equations ( [ 1a]-[1b ] ) we obtain the eigenvalue equation @xcite @xmath36 where the fresnel matrix @xmath37 is given by @xmath38 with @xmath39 and @xmath40 non - trivial solutions for the eigenvalue problem in eq .
( [ 1 ] ) can be found if , and only if , @xmath41 , which is known as the generalized fresnel equation .
this equation gives also the dispersion relations of the media under study . using the covariant formulae for the traces of linear operators @xcite the fresnel equation can be cast as @xmath42 where we defined the traces @xmath43 as a requirement of the geometrical optics limit , the wave fields are considered to be negligible when compared with the external fields .
thus , we assume from now on that the fields are approximated by their external counterparts @xmath44 and @xmath45 .
we set @xmath46 and @xmath47 which could be arbitrary functions of space and time coordinates .
let us examine the particular case of uniaxial media @xcite with permittivity @xmath48 using eqs .
( [ 16b ] ) and ( [ tracesa]-[tracesc ] ) , straightforward calculations show that eq .
( [ 23 ] ) results in the following algebraic fourth degree equation for the phase velocity @xmath49 of the propagating waves , @xmath50 with @xmath51 , \label{56}\\ & d = & -\frac{6\,\mu'}{\mu^3}\ , \varepsilon_{\scriptscriptstyle\perp}\,e\,b\,\hat{q}_z , \label{60}\\ & e = & \frac{6}{\mu^2 } \left[\varepsilon_{\scriptscriptstyle\parallel}\,\hat{q}_x{}^2 + \varepsilon_{\scriptscriptstyle\perp}\,(\hat{q}_y{}^2+\hat{q}_z{}^2)\right ] .
\label{61}\end{aligned}\ ] ] we defined @xmath52 , where @xmath53 , for any unit vector @xmath54
. then , @xmath55 and @xmath56 .
notice that when the propagation occurs in the @xmath57-plane ( spanned by the external electric and magnetic fields ) , the coefficients @xmath58 and @xmath59 are null and the generalized fresnel equation reduces to a quadratic equation in @xmath60 , therefore allowing only birefringence .
the same behavior occurs if @xmath61 .
in fact in this situation we recover linear electrodynamics . solving eq .
( [ 40 ] ) we obtain @xmath62 @xmath63 where @xmath64 the solution @xmath65 does not depend on direction of the wave propagation and will be called the ordinary wave , whereas @xmath66 depend on direction of wave propagation and will be called extraordinary waves @xcite . by definition ,
the velocities of the waves are given by @xmath67 , where @xmath68 is given by eqs .
( [ 75]-[76 ] ) . in order to achieve more simplicity in the following analysis
we assume the external fields to be constant and set the wave vector in a given direction of the @xmath69-plane , i.e. , @xmath70 , @xmath71 and @xmath72
. in this notation @xmath73 indicates the angle between @xmath74 and the @xmath75 direction .
two distinct solutions for @xmath76 in a same given direction @xmath74 can be obtained from eq .
( [ 76 ] ) if @xmath77 thus , taking into account the ordinary wave , eq .
( [ 84 ] ) defines a region inside which trirefringence occurs in any chosen direction @xmath74 .
let us examine this effect closer .
in order to guarantee the existence of an ordinary wave we must set @xmath78 , otherwise @xmath65 is not real .
this is true when both coefficients @xmath79 and @xmath80 present the same sign , which can be positive for usual media or negative for left handed materials ( in this case a negative refractive index occurs @xcite ) .
let us set these coefficients to be positive .
now , in order to satisfy eq .
( [ 84 ] ) we set @xmath81 .
hence , trirefringence occurs in directions determined by @xmath82 the above discussed phenomenon is displayed in fig .
( [ fig1 ] ) where the normal surfaces @xcite associated with the ordinary and extraordinary waves are depicted for some specific values of the quantities appearing in eq .
( [ angles ] ) . for
any given direction encompassed by the angles between the two dashed straight lines , defined by eq .
( [ angles ] ) , there are three distinct solutions : the dashed and dot - dashed curves representing the extraordinary waves , and the circular solid curve representing the ordinary wave . for angles between the dashed and the solid straight lines only birefringence occurs , and finally only one refraction occurs for directions encompassed by the angles between the two solid straight lines .
it is also worth noticing from fig .
( [ fig1 ] ) that in the sectors where more then one refractive index occur , the medium under consideration behaves as a positive or negative medium @xcite , depending on sub - sectors and extraordinary waves . in geometrical optics
the directions of light rays are given by the directions of the group velocities @xmath83 which are considered as the physical velocities of propagation of the rays @xcite .
as we see from eq .
( [ 76 ] ) , the extraordinary waves depend on the wave vector .
thus , the directions of the extraordinary light rays do not in general coincide with the directions of the extraordinary phase velocities , as explicitly shown by eq .
( [ groupvs ] ) .
taking @xmath84 into the definition of the group velocity we obtain that the ordinary group velocity is identified with the ordinary phase velocity @xmath85 .
nevertheless , the extraordinary group velocities are given by @xmath86 with @xmath87 and where we defined @xmath88 in these equations @xmath89 represents the extraordinary wave velocities in the direction @xmath74 , which can be obtained from fig .
[ fig1 ] by taking the distances from the associated points on its wave vector surfaces to the origin of the coordinate system . for the particular model set in fig .
[ fig1 ] the corresponding ray surfaces @xcite are depicted in fig .
[ fig2 ] as a function of the angle @xmath90 between @xmath91 and the z axis , where it was assumed for simplicity that @xmath92 .
notice that this assumption resides in the realm of nonlinear media . in the context of linear electrodynamics , for which @xmath93 , eqs .
( [ group]-[groupz ] ) show that @xmath92 leads to superluminal group velocity solutions .
therefore , this assumption is not in disagreement with the so - called causality requirement @xcite . for the ordinary waves we have @xmath94 , as anticipated . for the extraordinary waves
, the relationship between these angles can be obtained directly from eqs .
( [ groupx]-[groupz ] ) as @xmath95 $ ] . from fig .
[ fig2 ] , we see that there is a sector inside which there exist two extraordinary rays in any chosen direction of observation . for the complementary sector
there are none extraordinary rays .
we point out that any chosen direction for the rays inside the trirefringent sector is related to two different extraordinary wave vectors in fig [ fig1 ] .
similarly , for any given wave vector inside the two dashed straight lines in fig .
[ fig1 ] , there are two different extraordinary rays associated with it in fig [ fig2 ] . for instance , let us take the wave vector in the direction @xmath96 , as indicated by the arrow drawn in fig .
[ fig1 ] . as is clear , in this direction there are three solutions for the phase velocity , indicating that trirefringence occurs .
these solutions are represented by the points on the @xmath97 , @xmath98 and @xmath99 curves , found in this particular direction , and their moduli are @xmath100 , @xmath101 and @xmath102 , respectively . from eqs .
( [ group]-[groupz ] ) , the corresponding group velocities present magnitudes @xmath103 , @xmath104 and @xmath105 .
their associated directions are indicated by the arrows appearing in fig .
[ fig2 ] , where the dashed and dot - dashed arrows correspond to @xmath106 and @xmath107 , respectively .
if the principal permittivity components are also dependent upon the modulus of the electric field , then analogous calculations leading to the equations for the phase and group velocities and conditions for having trirefringence derived in this paper will still hold by identifying @xmath108 and @xmath109 once we work in the regime of high frequencies ( required by geometrical optics ) losses must be assumed to be low @xcite .
metamaterials with low - losses are still a matter of current investigations @xcite , with some achievements already obtained in such regime @xcite . in what concerns
the role played by the nonlinearities eqs .
( [ 76]-[76ib ] ) show us that if they exist ( i.e. @xmath110 ) , irrespective of their strengths , then they will lead to the presence of two extraordinary waves ( hence allowing trirefringence ) .
given the physical quantities appearing in eqs .
( [ group]-[groupz ] ) , the thresholds for the strengths of the nonlinearities are such that causality is not violated . the important quantity related to the nonlinearities
is @xmath111 , given by eq .
( [ 76ib ] ) , which is experimentally controllable , thereby making the predictions here derived feasible .
layered media @xcite seem good candidates for experimental realizations of our assumptions .
consider a system constituted by a repetition of two thin low - loss layers .
one of them is composed by a nonmagnetic ( @xmath112 ) and dispersive ( @xmath113 ) medium .
the other one is composed of a liquid medium whose permeability and permittivity are dependent upon the modulus of the resultant electric field as @xmath114 and @xmath115 , where @xmath116 and @xmath117 are usually much smaller than unity @xcite . additionally , if convenient external fields and layer parameters are present , then it is always possible to guarantee that our constraints are fulfilled .
refraction analyzes were disregarded in our work once they are just a straightforward extension of the analyzes valid for birefringent crystals @xcite .
the only difference now is the existence of two extraordinary waves . summing up , working in the limit of geometrical optics we studied the propagation of electromagnetic waves in nearly transparent nonlinear uniaxial metamaterials in the presence of external electric and magnetic fields .
we assumed a constant non - isotropic permittivity tensor ( for a given frequency ) presenting a negative principal component and an isotropic permeability dependent on the modulus of the electric field .
we solved the corresponding eigenvalue problem and obtained the general forth degree polynomial equation , whose solutions describe the propagation of waves . in this context , we showed that trirefringence is a phenomenon allowed to occur and could be described both in terms of wave and ray propagation . with the present technology in manipulating the dielectric coefficients in metamaterials @xcite , we hope that the effect here derived can be experimentally tested and if verified could lead to applications .
this work was partially supported by the brazilian research agencies cnpq , fapemig and capes .
jpp thanks fapemig for supporting his msc studies and erasmus mundus irap phd program for supporting his current phd studies .
j. hadamard , in _ leons sur la propagation des ondes et les quations de lhydrodynamique _ ( ed .
hermann , paris , 1903 ) ; g. boillat , j. math .
11 * , 941 ( 1970 ) ; a. papapetrou , in _ lectures on general relativity _
( dordrecht , holland,1974 ) . | we study the propagation of electromagnetic waves in the limit of geometrical optics for a class of nearly transparent nonlinear uniaxial metamaterials for which their permittivity tensors present a negative principal component .
their permeability are assumed positive and dependent on the electric field .
we show that light waves experience triple refraction trirefringence . additionally to the ordinary wave ,
two extraordinary waves propagate in such media . |
entanglement plays an important role in distinguishing the nature of quantum versus classical systems .
it is an essential ingredient for quantum computation .
it also connects quantum information theory to the traditional quantum many - body systems , for example , quantum critical phenomena @xcite and topological systems such as fractional quantum hall effects @xcite , topological insulators @xcite and graphene @xcite . in recent developments entanglement
has also been related to numerical methods based on tensor network algorithms , ranging from the density - matrix renormalization group and matrix product state@xcite , to the projected entangled pair states , and variational renormalization group methods @xcite .
one can measure the entanglement of a pure state using bipartite entanglement entropy : consider a pure state @xmath1 of a bipartite total system @xmath2 that consists of the system @xmath3 and the environment @xmath4 .
the reduced density matrix of the system a is @xmath5 .
the entanglement entropy @xmath6 , has been widely used to measure the bipartite entanglement between the system @xmath3 and the environment @xmath4 .
the entanglement entropy is especially useful for studying quantum criticality . in one dimension ,
scaling of entanglement entropy is well understood both for fermions and for bosons .
for one - dimensional quantum chains at zero temperature , it is generally known that the entanglement entropy @xmath7 of a system @xmath3 saturates away from criticality , however , it scales logarithmically when the system becomes quantum - critical , that is , when the correlation length diverges . in the latter case , conformal field ( cft )
theory @xcite yields @xmath8 where @xmath0 and @xmath9 are holomorphic and antiholomorphic central charges of the cft and @xmath10 is a model - dependent constant . for a quantum critical system with an unknown central charge
, one can estimate @xmath0 by calculating the scaling of entanglement entropy .
for the conveniency of numerical calculations , the scaling law of entanglement entropy for finite sizes @xcite @xmath11 + k , \label{eqn : s_scalingfinite}\ ] ] for a subsystem of size @xmath12 and total size @xmath13 ( the size of the system plus environment ) is often used instead of the infinite size one eq .
( [ eqn : s_scalinginfinite ] ) . in our system discussed below @xmath14 .
however , the general knowledge about the scaling law of entanglement entry for non - critical systems is not 100 percent correct .
there are some exceptions .
it is pointed out recently that even when the system loses the conformal invariance , the entanglement entropy @xmath7 can still scale logarithmically with the system size .
examples include the ferromagnetic heisenberg model @xcite and the infinite random fixed point ( irfp ) of several random spin chain models @xcite .
those systems are at the crossover points between critical and gapped systems and , most essentially , their ground states are highly degenerate .
therefore it is possible to define an effective central charge for those gapped but symmetric systems , according to eq .
( [ eqn : s_scalinginfinite ] ) , as three times the coefficient in front of the logarithmic scaling . in this work
we are interested in the situation where the cft regime ends at a ferromagnetic point with an effective central charge that is larger than the central charge of the cft regime .
we found that , if one only uses eq .
( [ eqn : s_scalingfinite ] ) to calculate central charges for finite systems , the central charges are influenced by the ferromagnetic points .
hence , we study the effects of such a nearby ferromagnetic point on the estimation of the value of the central charge from the numerical simulation .
we find that , depending on the algorithms and the physical quantities used to extract the value of central charge , different crossover behavior arises as the system approaches the ferromagnetic point from within the critical regime .
based on these results we show how to determine if the central charge estimation is strongly influenced by the crossover behavior and how to properly evaluate the central charge .
we consider the 1d spin-1/2 anisotropic heisenberg
( xxz ) model with periodic boundary conditions ( pbc ) .
the hamiltonian reads : @xmath15 where @xmath16 is the anisotropy . for @xmath17
the system is in the nel phase which spontaneously breaks the lattice translation symmetry and the ground states are two - fold degenerate . for @xmath18
the system is in the ferromagnetic ising phase which spontaneously breaks the spin reflection symmetry .
the ground states are two - fold degenerate and fully polarized in @xmath19 directions . when @xmath20 the system is in the gapless critical xy phase .
it is known that the critical xy phase is described by a @xmath21 cft .
the point @xmath22 is very special and is the main interest of this work .
we will refer this point as the _ ferromagnetic point _ in the rest of the manuscript .
at @xmath22 the symmetry of the hamiltonian is enlarged to isotropic ferromagnet with full rotational symmetry .
the ground states are infinitely degenerate in the thermodynamic limit .
however the system is not conformally invariant at this point .
when the system is in the ferromagnetic phase @xmath23 the entanglement saturates as one increases the block size .
however , it has been shown when @xmath22 there is an essential singularity in the entanglement entropy @xcite .
it has been also shown that in the limit of @xmath24 the entanglement entropy scales logarithmically with the block size but with a coefficient that is larger than the critical regime @xcite . for @xmath25 $ ]
the ground state is @xmath26 symmetric and has @xmath27 . at the ferromagnetic point
the symmetry of the hamiltonian is enlarged to su(2 ) , but this is broken by the ground state and hence there are infinitely many degenerate ground states in the thermodynamic limit .
it is exactly this ground stage degeneracy that gives rise to the logarithmic scaling of the entanglement entropy . in this work
we are interested in the entanglement entropy scaling of the ground state that is smoothly connected to the ground state in the critical xy phase .
this particular ground state at the ferromagnetic point can be reached by taking the @xmath28 limit , which corresponds to the ferromagnetic ground state with @xmath29 = 0 .
it is predicted that as @xmath24 one has @xmath30 this corresponds to an effect central charge @xmath31 which is larger than the @xmath21 in the xy phase .
consequently from entanglement entropy scaling point of view , there is a jump from @xmath21 for @xmath25 $ ] to @xmath31 as @xmath24 . for any numerical simulation ,
however , it is expected that such an abrupt jump is smeared out .
the resulting crossover behavior may depend on the algorithms and finite - size effects .
consequently , conventional methods to extract the value of central charge may be influenced by the nearby ferromagnetic point .
to the best of our knowledge , such an influence is not widely studied in the literature .
furthermore , it is pointed out in ref.@xcite that at @xmath22 it is possible to have @xmath32 where @xmath33 .
the exact value of @xmath34 depends on the particular ground state one choose and @xmath34 can be interpreted as the ( not necessarily an integer ) number of zero - energy goldstone bosons describing the ground state .
it is , however , not clear how such a prediction manifest itself in the conventional calculation of the entanglement entropy . to investigate these issues
, we use three different methods to evaluate the central charge in the regime @xmath35 $ ] in this work .
the first two methods study the entanglement scaling of a finite and an infinite system respectively while the third method studies the scaling of the ground state energy .
two numerical algorithms are used to evaluate relevant quantities for finite systems and infinite systems .
they are the density matrix renormalization group ( dmrg ) algorithm @xcite for finite systems and idmrg algorithm @xcite for the infinite systems respectively .
( note that idmrg is not the infinite size dmrg algorithm that is used in the warm up stage of the dmrg algorithm . )
we pay special attention to the behavior when @xmath36 . for the rest of the manuscript we will express this limit as @xmath37 for clarity .
the manuscript is organized as follows : in sec.[sec : ent_finite ] , we use the entanglement entropy scaling of a finite - system to estimate the value of the central charge . in sec.[sec : ent_infinite ] , we study the entanglement entropy scaling of an infinite - system . in sec.[sec : ent_gs ] , we use finite - size scaling of the ground state energy to extract the central charge value .
spin - wave velocity and the scaling dimension of the primary field are also estimated in sec.[sec : scaling ] . in sec.[sec : sum ] we discuss our results and suggest a strategy to determine accurately the value of the central charge when there is a ferromagnetic point nearby .
as a function of @xmath38 with system size @xmath39 , and @xmath40 respectively .
inset : @xmath41 with system size @xmath42 , extracted from dmrg while keeping @xmath43 states respectively . ] in this section we use the entanglement entropy scaling of a finite - size system to estimate the central charge @xmath0 .
most of the numerical studies in the literation to confirm the cft predictions on xxz chain concentrates on the regime of @xmath44 .
some works extend the study to @xmath45 @xcite but are not focused on the limit of @xmath46 , which is the main interest of this work . here
the dmrg algorithm @xcite is used to obtain the ground state wave - function of a finite system with length @xmath13 with periodic boundary condition ( pbc ) .
it is then straightforward to calculate the entanglement entropy @xmath47 between a block of length @xmath12 and the rest of the system . in this work
we use dmrg that preserves the @xmath26 symmetry and target the @xmath27 sector .
this is to ensure that the proper ground state is reached as we take the @xmath24 limit .
we define an @xmath13-dependent effective central charge @xmath41 by fitting the data using eq.[eqn : s_scalingfinite ] .
the subscripts dmrg is used to distinguish from the central charge obtained by other methods .
since the accuracy of the dmrg ground state depends on the number of states kept ( denoted as @xmath48 ) , it is important to study how @xmath41 depends on @xmath48 . for a given @xmath13 , the entanglement entropy
@xmath47 reaches its maximum at half - chain @xmath49 and grows logarithmically with @xmath13 in the critical regime .
on the other hand the maximal half - chain entanglement entropy attainable by dmrg is @xmath50 .
it is then natural to expect that for larger @xmath13 the value of @xmath41 may depend strongly on the @xmath48 used . in inset of fig .
[ fig : c_eff ] we show the value of @xmath51 evaluated using @xmath43 , and @xmath52 respectively .
we observe that for a fixed @xmath16 the value of @xmath0 monotonically increases as @xmath48 increases .
we find that while the data of @xmath53 seem to fit eq.[eqn : s_scalingfinite ] well ( not shown here ) , the fitted @xmath54 seems to deviate substantially from the results obtained with larger @xmath48 .
this indicates that @xmath53 is too small to accurately determine the value of @xmath54 .
in contrast for @xmath55 the value starts to converge .
we also find that for smaller @xmath56 , @xmath53 and @xmath57 lead to similar values of @xmath54 and it is unnecessary to go to larger @xmath48 .
these results suggest that in order to estimate reliably the central charge , one should increase @xmath48 until the value of @xmath54 start to saturate or perform the extrapolation of @xmath48 to infinity . in fig .
[ fig : c_eff ] , we plot @xmath41 as a function of @xmath38 for various @xmath13 . in this plot
we ensure that sufficiently large @xmath48 is used for each @xmath13 . for a fixed @xmath13
, we observe that @xmath41 monotonically increases as @xmath38 decreases to zero from the positive side .
consequently the central charge deviates more and more from the cft prediction @xmath21 when approaching the ferromagnetic point .
furthermore , as @xmath37 , @xmath41 seems to approachf @xmath58 , the value predicted in ref.@xcite regardless the @xmath13 used . recall that we use dmrg with u(1 ) symmetry and we target the ground state with @xmath27 , which is exactly the state considered in ref.@xcite .
hence at @xmath22 one expects @xmath59 with @xmath60 .
we also observe that for a fixed @xmath16 , @xmath41 decreases monotonically as @xmath13 increases and the crossover from @xmath21 to @xmath31 become sharper and shaper as @xmath13 increases .
based on all the observations above , we expect that in the thermodynamic limit the function @xmath54 becomes non - analytic at @xmath22 , resulting in @xmath21 for @xmath61 $ ] and jump to @xmath62 at @xmath22 .
this picture is consistent with the theoretical prediction of ref.@xcite .
this is also consistent with the calculation in ref.@xcite , in which by exact diagonalization of small @xmath13 systems it is found that the entanglement spectrum deviates substantially from the cft prediction when @xmath63 . from the numerical point of view
, the important observation is that for smaller @xmath13 the crossover regime is larger and one can overestimate the central charge by a large amount .
it is also important to note that if the number of states kept for the dmrg calculation is too small , one can underestimate the central charge . since the error due to finite @xmath13 and finite @xmath48
partially cancel each other , a systematic study of @xmath48 and @xmath13 dependence of the central charge is necessary to reliably extract the value of the central charge near a ferromagnetic point .
as a function of @xmath38 .
inset : entanglement entropy @xmath64 of a half - infinite chain as a function of @xmath65 for various @xmath16 . here
@xmath66 is the correlation the length .
black solid line is obtained by fitting data in the regime of @xmath67 with @xmath68 . here
@xmath69 is a @xmath16-dependent length scale as described in the main text . ] in this section we use the entanglement entropy of an infinite system to estimate the value of central charge in the thermodynamic limit .
we employ the idmrg algorithm to obtain the optimal ground state wavefunction of an infinite system @xcite .
the wavefunction is in the form of the matrix product state ( mps ) with truncation dimension @xmath70 .
if the true ground state is critical , the ground state obtained by idmrg corresponds to a nearly critical system with a large correlation length @xmath66 . the ground state obtained by idmrg can be imagine as the ground state of a nearly critical hamiltonian , which is obtained by adding certain relevant perturbation to the critical hamiltonian . to ensure that the ground state is smoothly connected to the ground state in the critical xy phase
, we always preserve @xmath26 symmetry in the idmrg calculation and target the @xmath27 sector .
for such a system the entanglement entropy of a half - infinite chain scales as @xcite @xmath71 in idmrg both the half - infinite chain entropy @xmath64 and the correlation length @xmath66 can be easily calculated from the transfer matrix .
eq.[eq : idmrg_xi ] then can be used to estimate the central charge @xmath0 .
we note that for such an infinite size algorithm , it is believed that the simulation at infinite @xmath66 and finite @xmath70 reproduces the results at finite @xmath66 and infinite @xmath70 with a scaling law @xmath72 .
is is shown in ref @xcite that @xmath73 consequently one has @xmath74 which can also be used to evaluate the value of central charge without evaluating the correlation length .
we find that fitting using eq.[eq : idmrg_xi ] converges faster , but the results are always consistent with each other .
in the inset of fig . [
fig : idmrg_all ] we show the half - infinite chain entanglement entropy @xmath64 as a function of @xmath65 for various @xmath35 $ ] .
we observe an interesting phenomenon : when the system is far from the ferromagnetic point ( for example @xmath75 ) all the data fall on a straight - line . by fitting the data with eq .
[ eq : idmrg_xi ] one finds @xmath76 as predicted by the cft .
when the system gets closer to the ferromagnetic point , however , the data starts to spilt into two regions which are separated by a @xmath16-dependent crossover length scale @xmath69 . in both regions the data fall on a straight line but with different slopes . in fig .
[ fig : idmrg_all ] we plot @xmath77 which is obtained by fitting with eq.[eq : idmrg_xi ] but using only data with @xmath78 as a function of @xmath38 .
we always find @xmath79 , but larger and larger @xmath70 is needed to access the regime of @xmath78 as the system gets closer to the ferromagnetic point .
we conclude that @xmath76 provided that only data with @xmath78 are used for the fitting .
this is consistent with the cft prediction and the results in the proceeding section ( after taking the limit of @xmath80 ) .
we find that when the system gets very close to the ferromagnetic point , the deviation from the expected result @xmath21 becomes larger .
this is due to ( 1 ) larger @xmath70 is needed to have enough data with @xmath78 and ( 2 ) there are many nearly degenerate low energy states when the system approaches the ferromagnetic point , making it more difficult for idmrg to converge .
furthermore , it is surprising to observe that all the data with @xmath81 seem to fall on a universal straight line regardless the value of @xmath16 .
since the crossover length scale @xmath69 diverges as @xmath24 we expect that the scaling behavior of the ferromagnetic point can be obtained by fitting the data with @xmath81 .
we find that @xmath82 leading to an effective central charge @xmath83 .
to understand this result we note that u(1 ) symmetry is preserved in our idmrg calculation . similar to the finite - size dmrg we also target the @xmath27 ground state .
we conjecture that for infinite size xxz chains near the ferromagnetic point , when @xmath81 the entanglement scales as @xmath84 if this scaling is interpreted as @xmath85 , one find @xmath86 which is very close to the numerical value we obtain .
our results also agree with the physical picture proposed in ref @xcite , where it is proposed that when @xmath24 , there is an increasing characteristic length scale such that ( 1 ) on the scales much below it , sites are randomly up and down with appropriate coefficients that locally reproduce the ground state of @xmath22 and ( 2 ) on the scale above , a spin singlet that is expected for the ground state of the xxz model for @xmath61 $ ] .
we believe that this characteristic length scale is exactly the @xmath69 we identified by imdrg .
consequently when @xmath81 the entanglement should scale according as eq.[eq : scaling_fm_xi ] with an effective central charge @xmath86 as observed in idmrg calculation .
as a function of @xmath38 by fitting eq.[eq : scaling_d_gs ] with data from @xmath87 and @xmath88 ( purple left triangle ) , @xmath89 ( blue right triangle ) , @xmath90 ( green diamond ) , @xmath91 ( red square ) , and @xmath92 ( black circle ) respectively .
inset : ground state energy per site for @xmath93 ( black circle ) .
fitted lines using eq.[eq : scaling_d_gs ] with data from @xmath94 to @xmath95 ( purple line ) and @xmath96 to @xmath95 ( blue line ) respectively . ] in this section we use the finite - size scaling of the ground state energy to estimate the central charge .
it is well known that when the critical system is described by a cft with central charge @xmath0 in the continuum limit , the ground state energy of a finite system of length @xmath13 scales as @xmath97 where @xmath98 is the ground state energy per site in the thermodynamics limit and @xmath99 is the spin - wave velocity @xcite . in this work
we use ed ( for @xmath100 ) and finite - size dmrg ( for @xmath101 ) to obtain the ground state energy @xmath102 of size @xmath13 . in order to obtain @xmath103 by fitting @xmath102 with eq .
[ eq : scaling_d_gs ] , it is better to have an independent estimation of the spin - wave velocity @xmath99 .
since we are mainly interested in the influence of the ferromagnetic point , we first analyze our data using the exact spin - wave velocity , @xmath104 which is obtained from the bethe ansatz @xcite . here
@xmath105 is defined via @xmath106 with @xmath107 .
the scenario in which the spin - wave velocity is obtained numerically will be discussed later . in the inset of fig .
[ fig : gs ] we show the fitted lines for @xmath93 using data from @xmath94 to @xmath95 ( purple line ) and @xmath96 to @xmath95 ( blue line ) respectively .
it is clear from the figure that the higher order corrections to eq .
[ eq : scaling_d_gs ] are large .
furthermore , we observe that for @xmath108 $ ] all the data behave in a similar fashion .
these higher order corrections can be suppressed by removing smaller sizes data in the fitting procedure . in fig .
[ fig : gs ] we show the fitted @xmath103 using data with @xmath87 with @xmath109 , and @xmath92 respectively .
we find that when the system is far away from the ferromagnetic point , one has @xmath110 regardless the @xmath111 used .
when approaching the ferromagnetic point , fitted @xmath103 begins to deviate from 1 and monotonically decreases . however , larger @xmath111 allows one to reach closer to the ferromagnetic point while maintaining @xmath110 .
this behavior is consistent with the observation in the preceding section that there is a @xmath16 dependent length scale , which grows larger as one approaches the ferromagnetic point .
the system only behaves like one with the xxz ground state when the system size is larger than this length scale .
this is why larger and larger @xmath111 is needed to obtain the proper central charge as @xmath112 .
however it is unclear to us how to define a proper @xmath111 from the @xmath69 obtained in the preceding section . in order to ensure that the behavior above is due to the ferromagnetic point and not due to the inaccuracy of the data , it is important to have some independent check on the quality of the data
. we have performed the following two tests .
first , we investigate how our results depend on the @xmath48 used . for finite - size dmrg calculation
, it is expected that the energy is very accurate once @xmath48 is large enough . in our calculation
we ensure that @xmath48 is large enough so that the fitted values of @xmath103 and @xmath113 are not sensitive to @xmath48 .
second , we compare the fitted @xmath114 to the exact energy per site from the bethe ansatz @xcite : @xmath115},\ ] ] where @xmath105 was defined below equation ( [ v - th ] ) . we find that the absolute error @xmath116 is at most at the order of @xmath117 .
in the analysis in the preceding section , we used the exact spin - wave velocity for the fitting of the central charge . in general , however , the spin - wave velocity can only be estimated numerically . in the conventional approach
one first obtains an @xmath13-dependent velocity from the ground state energy and the lowest energy with momentum @xmath118 @xmath119,\ ] ] then one uses the following scaling ansatz @xmath120 to obtain an extrapolated value @xmath121 of the spin - wave velocity in the thermodynamic limit@xcite . while it is difficult to keep momentum quantum number in conventional dmrg , we find that for the xxz model the lowest energy state with @xmath118 is always the first excited state in the @xmath27 sector , which can be calculated reliably with dmrg . in the inset of fig .
[ fig : scaling ] we plot the absolute error @xmath122 of the spin - wave velocity .
we find that the error is very small for @xmath123 but grows rapidly as one approaches the ferromagnetic point .
this is understandable since the spin - wave velocity should approach zero at the ferromagnetic boundary which makes it very difficult to be accurately determined by the numerical simulation .
obtained by using @xmath124 ( black circle ) and @xmath121 ( red square ) respectively .
inset : absolute error of the spin wave velocity @xmath122 . ] from the results in the preceding three sections we find that the value of central charge extracted from typical simulation might deviate from the cft prediction when the system is close to the ferromagnetic point .
this might make it more difficult to identify the underlying cft from the value of central charge alone .
another quantity which can be used to identify the cft is the scaling dimension of the primary field .
it is known that the excited state energies @xmath125 are related to the scaling dimension of a certain primary field of the cft via @xmath126 where @xmath127 is the scaling dimension and @xmath48 and @xmath128 are integers . for the xxz model
the smallest scaling dimension is @xmath129 the corresponding excited state is the lowest energy state in the @xmath130 sector and has momentum @xmath131 . by using finite - size dmrg with @xmath26 symmetry the energy @xmath132 can be evaluated accurately . to remove the sub - leading correction in eq.[eq : scaling_dimension ]
we first obtain a @xmath13-dependent scaling dimension defined as @xmath133 and then extrapolate to the thermodynamic limit using @xmath134 in fig .
[ fig : scaling ] we plot the absolute errors @xmath135 of the results obtained by using exact velocity ( black dot ) and estimated velocity ( red dot ) respectively .
we find that the error is extremely small when the exact velocity is used in the whole parameter regime including the ferromagnetic point .
this suggests that eq .
[ eq : scaling_dimension ] , the finite - size scaling formula of the scaling dimension , is less influenced by the ferromagnetic point provided that the exact spin - wave velocity is used .
however , as shown above , the ferromagnetic point does influence the estimation of the spin - wave velocity . as shown in fig .
[ fig : scaling ] this makes the estimated scaling dimension less accurate near the ferromagnetic point .
in summary we investigate how the numerical estimation of the central charge @xmath0 is influenced by the non - trivial logarithmic scaling of the entanglement entropy of the nearby ferromagnetic point . from entanglement point of the view
, the nontrivial scaling at the ferromagnetic point gives rise to a jump of the central charge from @xmath21 predicted by cft to @xmath136 of the ferromagnetic point .
in particular we use the entanglement entropy scaling of a finite and an infinite system , as well as the finite - size scaling of the ground state energy to estimate @xmath0 .
we find that all methods are influenced by the nearby ferromagnetic point but different crossover behavior appears .
we also find that the nontrivial scaling of the ferromagnetic point only manifests itself in the first two methods , which are entanglement based . in the following we
briefly summarize the crossover behavior and the proper procedure to estimate @xmath0 within each method .
we also suggest that one should employ all three methods and use the consistency between different methods as an additional check for the accuracy of the value of @xmath0 . in any finite - size calculation
we observe a smooth crossover between these two values .
we show that the proper procedure is that for a fixed @xmath13 one need to increase @xmath48 until a saturated value of @xmath0 is reached .
one then increases @xmath13 until @xmath0 is not sensitive to the changes in @xmath13 . if the value of @xmath0 is still changing when one reaches the largest @xmath13 one can calculate , the corresponding @xmath0 value can not be trusted .
when idmrg is used , however , a different crossover behavior appears .
here we show that there is a @xmath16 dependent crossover length scale @xmath69 such that for scales above the system scales according to the cft prediction while for scales below the system scales according to the ferromagnetic point . for any model with a nearby ferromagnetic like point we suggest that one should increase @xmath48 ( hence @xmath66 ) gradually to check if such a crossover behavior appears or not .
we also show that the conventional method to use finite - size scaling of the ground state energy to determine @xmath0 also suffers from the influence of the nearby ferromagnetic point .
it induces a sub - leading correction which seems to have an opposite sign with respect to the logarithmic correction due to the marginally irrelevant operator . for the unknown model
, the sign of the correction can be used to detect the existence of such an influence . here
we suggest that one should just discard smaller size data .
in addition , we show that the finite - size scaling relation between the excited state energy and the scaling dimension of the primary field is less influenced by the ferromagnetic point , provided that exact spin - wave velocity is used .
the numerical estimation of the velocity itself , however , is also influenced by the ferromagnetic point .
we thank andreas m. lachli for fruitful discussions . we acknowledge the support from national center for theoretical science ( ncts ) and national science council ( nsc ) , taiwan .
ian mcculloch acknowledges support from the australian research council discovery projects funding scheme ( project number dp1092513 ) . | we study the entanglement entropy scaling of the xxz chain . while in the critical xy phase of the xxz chain the entanglement entropy scales logarithmically with a coefficient that is determined by the associated conformal field theory , at the ferromagnetic point , however , the system is not conformally invariant yet the entanglement entropy still scales logarithmically albeit with a different coefficient .
we investigate how such an nontrivial scaling at the ferromagnetic point influences the estimation of the central charge @xmath0 in the critical xy phase .
in particular we use the entanglement scaling of the finite or infinite system , as well as the finite - size scaling of the ground state energy to estimate the value of @xmath0 .
in addition , the spin - wave velocity and the scaling dimension are also estimated .
we show that in all methods the evaluations are influenced by the nearby ferromagnetic point and result in crossover behavior .
finally we discuss how to determine whether the central charge estimation is strongly influenced by the crossover behavior and how to properly evaluate the central charge . |
a rich variety of novel phenomena perceptible in ce - based strongly correlated electron systems due to the duality between the localized and the itinerant nature of @xmath8-electrons , such as heavy electron and mixed valence behavior , kondo insulator or semiconductors , unconventional superconductivity , spin and charge density waves , spin and charge gap formation and metal - insulator transition of special recent interest , is the fascinating phenomena associated with magnetic quantum criticality and non - fermi liquid behavior arising from cooperative behavior at a zero temperature phase transition .
@xcite 0.4 cm rnic@xmath0 ( r = rare earth ) compounds crystallize in the noncentrosymmetric ( ncs ) orthorhombic cenic@xmath0 type structure ( space group amm2 ) in which a mirror plane is missing along the @xmath9-axis .
@xcite diverse ground states such as the superconducting , antiferromagnetic ( afm)/ferromagnetic ( fm ) and the charge density - wave ( cdw ) state are reported so far in rnic@xmath0(r = la , ce , pr , sm , gd , tb and er ) , reflecting the competition among different electronic states .
@xcite no structural phase transition has been reported in these compounds , and hence the spin and charge ordering phenomena in rnic@xmath0 compounds are thus native to the ncs crystal structure .
lanic@xmath0 is a superconductor below about 2.7 k. @xcite the absence of an inversion center is believed to give rise to mixed wave pair states ( s - wave singlet and p - wave triplet ) in ncs superconductors .
it has been argued that heavy fermion effects are not significant except perhaps for r = ce ( perhaps mixed valence effects for sm ) .
murase et al @xcite found that rare - earth intermetallic compounds rnic@xmath0 with r = nd , sm , gd and tb show anomalous temperature dependences of electrical resistivity and lattice constants .
@xcite they proposed that these anomalies are attributed to cdw transitions .
x - ray scattering studies of smnic@xmath0 reveal satellite peaks corresponding to an incommensurate wave vector ( 0.5 , 0.52 , 0 ) below 148 k at which the resistivity shows an anomaly , signing development of a charge - density wave .
@xcite the rare earth local moments order antiferromagnetically in most of the rnic@xmath0 compounds ( apart from r = pr ) , and smnic@xmath0 undergoes a first - order ferromagnetic transition at @xmath10 = 17.5 k. @xcite those results indicate that rnic@xmath0 compounds having various magnetic orders are promising candidates for the systematic investigation of the interplay of cdw and magnetic order . in this system ,
nickel atoms are nonmagnetic and rare - earth elements mainly contribute to the magnetic properties showing the character of local magnetic moments interacting through the ruderman@xmath11kittel@xmath11kasuya@xmath11yosida ( rkky ) interaction .
the magnetic properties of the rnic@xmath0 series are affected strongly by the orthorhombic crystalline electric field ( cef ) .
in this paper we report a detailed investigation on cenic@xmath0 by dc magnetic susceptibility ( @xmath12 ) , isothermal magnetization ( @xmath13 ) , heat capacity ( @xmath14 ) , neutron diffraction , muon spin relaxation ( @xmath1sr ) and inelastic neutron scattering ( ins ) measurements . below 20
k a sharp anomaly is observed in our @xmath15 and @xmath14 data which corresponds to a transition from the paramagnetic ( pm ) to afm state .
@xmath16 data also show a second anomaly around 2.5 k and a further weak anomaly near 10 k. the presence of long - range magnetic order is also revealed by @xmath1sr measurements where oscillations are observed in the spectra below 20 k. a doublet ground state is inferred both from @xmath16 and ins data .
ins reveals two well - defined cef excitations at 9 and 28 mev , indicating the localized nature of the @xmath17electrons in cenic@xmath0 .
.a summary of the results obtained from the refinement of the room - temperature neutron powder diffraction data : the lattice parameters and atomic positions .
the site occupancy for all the atoms was fixed to 100 % .
the reliability factors ( weighted profile factor @xmath18 = 5 @xmath19 ) were calculated by comparing the fit to the data ( space group no .
38 , amm2 ) .
[ cols="<,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] the solid lines in figs .
12 ( a)-(f ) represent the fit to the cef model for simultaneous refinement of all six data sets for 8 , 20 , and 50 mev incident energies ( we first refined 25 and 100 k data ) .
the phenomenological crystal field parameters @xmath20 obtained from the best fit are given in table ii .
the cef wave functions obtained from the simultaneous fit to 25 and 100 k , ins data are given by , @xmath21 @xmath22 @xmath23 @xmath24 , is the gs wave function , whereas , @xmath25 is for the first excited state at 92 k and @xmath26 is for the second excited state at 342 k above the gs . the gs magnetic moments of the ce atoms @xmath27 , @xmath28 and @xmath29 , along the three crystallographic axes @xmath30 ,
@xmath31 and @xmath9 , respectively can be calculated from @xmath32 @xmath33 @xmath34 calculated magnetic moments in presence of an applied magnetic field ( @xmath35 = 10 koe ; @xmath36 = 1 k ) using gs wave function are @xmath27 = 0.583 @xmath7 , @xmath28 = 0.6876 @xmath7 and @xmath29 = 0.341 @xmath7 .
this is in agreement with neutron diffraction data , which gives possible direction of moment along @xmath5axis .
@xcite however it is to be noted that the neutron diffraction study was unable to find accurate direction and magnitude of the moment due to limited numbers of magnetic bragg reflections .
@xcite 0.4 cm after obtaining cef parameters from the simultaneous fit of the six ins data sets at 25 and 100 k , we have analyzed 20 mev ins data at 5 k ( i.e. below the magnetic ordering ) adding the molecular field term in the hamiltonian .
we kept the values of cef parameters ( which were obtained from 25 and 100 k as mentioned above ) fixed for the analysis of 5 k data and varied molecular fields .
first we varied only molecular field along @xmath37axis ( @xmath38 ) , while keeping zero - molecular fields along @xmath5 ( @xmath39 ) and @xmath40axes ( @xmath41 ) .
this method did not give good fit to the data .
then we tried to vary @xmath39 ( while keeping @xmath38=0 and @xmath41=0 ) , however we did not obtain good fit for this method either .
hence finally we varied @xmath38 and @xmath39 simultaneously ( keeping @xmath41=0 ) , which gave a reasonably good fit to the data ( it is to be noted that when we also allowed to vary @xmath41 it gave almost zero value ) .
the values of the molecular field estimated from the analysis are @xmath38= 79.3(1.8 ) t and @xmath39=76.5(1.7 ) t. the quality of the fit can be seen in fig . 13 ( a ) .
the presence of molecular fields along @xmath42 and @xmath5axes may suggests that the ce moment is confined to the @xmath43-plane and not along @xmath42 or @xmath5 axis . a detail neutron diffraction investigation using a high neutron flux instrument on cenic@xmath0 single crystal
will be helpfully to understand the presence of the molecular field along @xmath30 and @xmath31-axes . to check reliability of our set of cef parameters we model the single crystal magnetic susceptibility data from literature .
@xcite observed anisotropic magnetic susceptibility data correctly fits with our calculated @xmath15 using cef parameters [ see fig . 13 ( b ) ]
we have studied the magnetic properties of cenic@xmath0 using magnetic susceptibility , heat capacity , inelastic neutron scattering and @xmath1sr measurements . a magnetic transition around 20 k is observed with the emergence of oscillations in zero field @xmath1sr spectra .
we fitted the temperature dependence of the internal magnetic fields to a model of a mean field magnet , which revealed 3d nature of magnetic interactions .
ins measurements of polycrystalline cenic@xmath0 at low temperatures indicate two cef excitations at 8 and 30 mev . at 5 k
, we observe an additional peak at 4.5 mev due to spin - wave excitations . above @xmath44 ,
this peak is not present , but quasielastic scattering is observed .
a linear fit to the temperature dependence of the quasielastic linewidth gives an estimate of @xmath45 = 11 k. from an analysis of ins and magnetic susceptibility data with a cef model , we propose a cef scheme for cenic@xmath0 .
the cef scheme correctly predicts the direction of the ordered moment , but the observed magnetic moment at 2 k is 0.6876 @xmath7/ce which is higher compared to the moment as observed from neutron diffraction ( 0.25 @xmath7/ce ) .
we believe that the observed reduced moment is due to presence of hybridization between the localized ce@xmath46 @xmath8-electrons and the conduction band .
further the susceptibility analysis reveals the anisotropic molecular fields , stronger along the @xmath31-axis , which is also in agreement with the direction of moment along the @xmath31-axis .
the spin wave measurements on cenic@xmath0 single crystal will help to shade more light on the anisotropic exchange interactions in cenic@xmath0 .
we note furthermore that the estimated on - site kondo exchange ( 11 k ) is of the same order of magnitude as the magnetic phase transition temperatures , which classifies this compound as a magnetically ordered kondo lattice .
the complexity of spin ordering in cenic@xmath0 is inferred as a consequence of admixing of crystal electric field energies together with kondo and rkky exchange , and we believe that our study and determination of cef parameters presented in this work establishes the essential ingredients with which to formulate an understanding of the ground state in cenic@xmath0 .
ab thanks the frc of uj and isis - stfc for funding support .
dta and adh would like to thank cmpc - stfc , grant number cmpc-09108 , for financial support .
ams thanks the sa - nrf ( grant 78832 ) and uj research committee for financial support .
m. smidman , d. t. adroja , a. d. hillier , l. c. chapon , j. w. taylor , v. k. anand , r. p. singh , m. r. lees , e. a. goremychkin , m. m. koza , v. v. krishnamurthy , d. m. paul , and g. balakrishnan , phys . rev .
b * 88 * , 134416 ( 2013 ) . | the magnetic state of the noncentrosymmetric antiferromagnet cenic@xmath0 has been studied by magnetic susceptibility , heat capacity , muon spin relaxation ( @xmath1sr ) and inelastic neutron scattering ( ins ) measurements .
cenic@xmath0 exhibits three magnetic phase transitions at @xmath2 = 20 k , @xmath3 = 10 k and @xmath4 = 2.5 k. the presence of long range magnetic order below 20 k is confirmed by the observation of oscillations in the @xmath1sr spectra between 10 and 20 k and a sharp increase in the muon depolarization rate .
ins studies reveal two well - defined crystal electric field ( cef ) excitations around 8 and 30 mev .
ins data have been analyzed using a cef model and the wave functions were evaluated .
we also calculated the direction and magnitude of the ground state moment using cef wave functions and compare the results with that proposed from the neutron diffraction .
our cef model correctly predicts that the moments order along the @xmath5axis ( or @xmath6-axis ) and the observed magnetic moment is 0.687(5 ) @xmath7 , which is higher than the moment observed from the neutron diffraction ( 0.25 @xmath7/ce ) .
we attribute the observed reduced moment due to the kondo screening effect . |
according to recent observational data sets@xcite , our current universe is flat and undergoing a phase of the accelerated expansion which started about five billion years ago . in principle
, this phenomenon can be explained by either dark energy ( see , for instance , ref.@xcite for reviews ) , in which the reason of this phenomenon is due to an exotic component with large negative pressure , or modified theories of gravity@xcite .
unfortunately , up to now a satisfactory answer to the question that what dark energy is and where it came from has not yet to be obtained .
alternative to dark energy , modified theories of gravity is extremely attractive , such as f(r ) gravity ( see , for instance , ref.@xcite for reviews ) , here f(r ) is an arbitrary function of the ricci scalar r. cosmic acceleration can be explained by f(r ) gravity@xcite , and the conditions of viable cosmological models have been derived in @xcite .
a general model of @xmath1 gravity has been proposed in ref.@xcite , which contains a non - minimal coupling between geometry and matter .
this coupling term can be considered as a gravitational source to explain the current acceleration of the universe . as a result of
the coupling the motion of the massive particles is non - geodesic , and an extra force , orthogonal to the four - velocity , arises .
different forms for the matter lagrangian density @xmath2 , and the resulting extra - force , were considered in @xcite , and it was shown that more natural forms for @xmath2 do not imply the vanishing of the extra - force .
the implications of the non - minimal coupling on the stellar equilibrium were investigated in @xcite , where constraints on the coupling were also obtained . an inequality which expresses a necessary and sufficient condition to avoid the dolgov - kawasaki instability for the model
was derived in @xcite . however , a more general model , in which the coupling style is arbitrary and the lagrangian density of matter only appears in coupling term , has been proposed in ref.@xcite , i.e. , the so - called the generalized f(r ) gravity with arbitrary coupling between matter and geometry . in this class of models
the energy - momentum tensor of the matter is generally not conserved , and the matter - geometry coupling can induce a supplementary acceleration of the test particles . moreover , the energy conditions and the dolgov - kawasaki criterion for the model have been derived in ref.@xcite , which are quite general and can degenerate to the well - known energy conditions in gr and f(r ) gravity with non - minimal coupling and non - coupling as special cases .
another interesting alternative modified theory of gravity is the modified gauss - bonnet gravity , i.e. , f(g ) gravity , where f(g ) is a general function of the gauss - bonnet ( gb ) term@xcite . at present specific models of f(g ) gravity have been proposed to account for the late - time cosmic acceleration@xcite , and the respective constraints on the parameters of the models have also analyzed in ref.@xcite . for more general forms of f(g ) ,
the most crucial condition to be satisfied is @xmath3 , which is required to ensure the stability of a late - time de sitter solution as well as the existence of standard radiation / matter dominated epochs@xcite , and solar system constraints on f(g ) gravity models have been studied in ref.@xcite .
recently the energy conditions in f(g ) gravity have been also discussed@xcite , but they are only adapted to @xmath4 gravity without coupling between matter and geometry . hence , in this paper the @xmath4 gravity models with curvature - matter coupling will be proposed , and some relevant issues , such as the energy conditions , the stability criterion and the conditions for late - time cosmic accelerated expansion , will be studied . this paper is organized as follows . in section 2 ,
the @xmath4 gravity models with curvature - matter coupling are proposed , which is here called the modified @xmath4 gravity . and some fundamental elements of the modified @xmath4 gravity are given . in section 3 ,
the well - known energy conditions , namely , the strong energy condition ( sec ) , the null energy condition ( nec ) , the weak energy condition ( wec ) and the dominant energy condition ( dec ) , will be derived .
concretely , the two models are applied to the weak energy condition in order to understand the meaning of these energy conditions .
furthermore , we will study stability criterion at the de sitter point , by which the parameters in the specific model in the modified f(g ) gravity can be constrained in section 4 .
in addition , by using the conditions of power - law accelerated expansion and the equation of state of matter less than @xmath0 , the conditions for late - time cosmic accelerated expansion in the modified f(g ) gravity are discussed in section 5 .
conclusions and discussions on our work are given in the last section .
as we know , the modified einstein - hilbert action@xcite is , @xmath5,\ ] ] in which @xmath6 , @xmath7 is the gravitational constant , @xmath8 is the ricci scalar , and @xmath2 is the lagrangian density of matter .
the gauss - bonnet invariant is defined as @xmath9 ( @xmath10 and @xmath11 are the ricci tensor and the riemann tensor , respectively ) .
below , we consider f(g ) gravity with curvature - matter coupling , which is here called the modified f(g ) gravity .
the lagrangian density of matter only appears in coupling term and the action is given by @xmath12l_m\},\ ] ] where we have chosen @xmath13 , which will be adopted hereafter .
varying the action ( [ 2 ] ) with respect to the metric @xmath14 yields the field equations : @xmath15 where @xmath16 , and @xmath17 is the energy - momentum tensor of matter , which is defined as : @xmath18 by assuming that the lagrangian density @xmath2 of the matter depends only on the metric tensor components , and not on its derivatives , we obtain @xmath19 . by taking the covariant derivative of eq.([3 ] ) , with the use of the bianchi identities , @xmath20 ( here @xmath21 is the einstein tensor ) , we can obtain the following relation : @xmath22 , \end{array}\ ] ] from which we see that the conservation of the energy - momentum tensor of matter is violated due to the coupling between matter and geometry . for the perfect fluid ,
the energy - momentum tensor is @xmath23 where @xmath24 is the four - velocity , and satisfies the conditions @xmath25 and @xmath26@xcite .
thus , the covariant derivative of eq.([6 ] ) can be given as @xmath27 by imposing the condition of the conservation of the matter current , @xmath28 , and with the use of the identity @xmath29@xcite , we have the equation of motion of a test particle in the model as @xmath30 where the extra - force @xmath31 has the following expression @xmath32.\ ] ] by substituting the relation ( [ 5 ] ) into eq.([9 ] ) , we can point that due to the presence of the coupling between matter and geometry , the motion of the massive particles is non - geodesic , and the extra - force @xmath31 is not orthogonal to the four - velocity @xmath24 , that is , there exists an angle between the extra - force @xmath31 and the four - velocity @xmath24 , and @xmath33 .
the energy conditions arise when one refers to the raychaudhuri equation for the expansion@xcite . under these energy conditions
, one allows not only to establish gravity which remains attractive , but also to keep the demands that the energy density is positive and can not flow faster than light .
below , following ref.@xcite we simply review the raychaudhuri equation which is the physical origin of the null energy condition(nec ) and the strong energy condition(sec)@xcite . in the case of a congruence of timelike geodesics defined by the vector field @xmath34 , the raychaudhuri equation is given by @xmath35 where @xmath36 and @xmath37 are the ricci tensor , the expansion parameter , the shear and the rotation associated with the congruence , respectively .
while in the case of a congruence of null geodesics defined by the vector field @xmath38 , the raychaudhuri equation is given by @xmath39 from above expressions , it is clear that the raychaudhuri equation is a purely geometric statement and independent of the gravity theory . in order to constrain the energy - momentum tensor by the raychaudhuri equation , one can use the ricci tensor from the field equations of gravity to make a connection .
namely , through the combination of the field equations of gravity and the raychaudhuri equation , one can obtain physical conditions for the energy - momentum tensor .
since @xmath40 ( the shear is a spatial tensor ) and @xmath41 ( hypersurface orthogonal congruence ) , from eqs .
( [ 10 ] ) and ( [ 11 ] ) , the conditions for gravity to remain attractive ( @xmath42 ) are @xmath43 @xmath44 thus by means of the relationship ( [ 12 ] ) and einstein s equation ( @xmath45 ) , one obtains @xmath46 if one considers a perfect fluid with energy density @xmath47 and pressure p , @xmath48 the relationship ( [ 14 ] ) turns into the well - known sec in general relativity , i.e. , @xmath49 similarly , by using the relationship ( [ 13 ] ) and einstein s equation , one has @xmath50 thus , by considering eq.([15 ] ) , the familiar nec in general relativity can be reproduced as : @xmath51 the einstein tensor resulting from the field equation ( [ 3 ] ) can be written @xmath52 where @xmath53 contracting the above equation , we have @xmath54 thus , we can write @xmath10 in terms of an effective stress - energy tensor and its trace , i.e. , @xmath55 by using the relationship ( [ 12 ] ) and eq.([22 ] ) , the sec can be given as : @xmath56 by using the eqs.([20 ] ) and ( [ 21 ] ) , the sec in eq.([23 ] ) can be expressed as @xmath57 where @xmath58 is the hubble parameter .
the nec in the modified f(g ) gravity can be expressed as : @xmath59 by the same method as the sec , the nec in eq.([25 ] ) can be changed into @xmath60 furthermore , by means of eqs.([20 ] ) and ( [ 22 ] ) , the effective energy density and the effective pressure can be derived as follows : @xmath61 @xmath62 note that the above expressions of the sec ( [ 24 ] ) and the nec ( [ 26 ] ) are directly derived from raychaudhuri equation . however , equivalent results can obtained by the transformations @xmath63 and @xmath64 into @xmath65 and @xmath66 . thus by extending this approach to @xmath67 and @xmath68 , the corresponding dec and the wec in the modified f(g ) gravity can be respectively given as : @xmath69 @xmath70 it is worth stressing that when taking @xmath71 , the energy conditions in general relativity can be reproduced .
furthermore , by defining the deceleration , jerk , and snap parameters as@xcite @xmath72 we have @xmath73 and the gb term is given by @xmath74 in addition , we consider @xmath75 , we can rewrite eqs .
( [ 27 ] ) and ( [ 28 ] ) as follows : @xmath76,\ ] ] @xmath77.\ ] ] hence , the energy conditions , i.e. , the sec , nec , dec and wec can be rewritten as : @xmath78 + 144\lambda h^{4}\rho[q f'+4h^{4}(-2q^{3}\nonumber\\-8q^{2}+q-2j-6qj+s+3)f''+96h^{8}(2q^{2}+3q+j)^{2}f''']\geq0,\end{aligned}\ ] ] @xmath79 + 48\lambda h^{4}\rho[q f'+4h^{4}(-2q^{3}\nonumber\\-8q^{2}+q-2j-6qj+s+3)f '' + 96h^{8}(2q^{2}+3q+j)^{2}f''']\geq0,\end{aligned}\ ] ] @xmath80 - 48\lambda h^{4}\rho[q f'+4h^{4}(-2q^{3}\nonumber\\-8q^{2}+q-2j-6qj+s+3)f '' + 96h^{8}(2q^{2}+3q+j)^{2}f''']\geq0,\end{aligned}\ ] ] @xmath81\geq0.\ ] ] it is worth stressing that when taking @xmath82 , eqs.([27 ] ) and ( [ 28 ] ) change into @xmath83,\ ] ] @xmath84.\ ] ] and the corresponding energy conditions are as follows : @xmath85 - 144\lambda h^{4}p[q f'+4h^{4}(-2q^{3}\nonumber\\-8q^{2}+q-2j-6qj+s+3)f''+96h^{8}(2q^{2}+3q+j)^{2}f''']\geq0,\end{aligned}\ ] ] @xmath86 - 48\lambda h^{4}p[q f'+4h^{4}(-2q^{3}\nonumber\\-8q^{2}+q-2j-6qj+s+3)f '' + 96h^{8}(2q^{2}+3q+j)^{2}f''']\geq0,\end{aligned}\ ] ] @xmath87 + 48\lambda h^{4}p[q f'+4h^{4}(-2q^{3}\nonumber\\-8q^{2}+q-2j-6qj+s+3)f '' + 96h^{8}(2q^{2}+3q+j)^{2}f''']\geq0,\end{aligned}\ ] ] @xmath88\geq0.\ ] ] in order to exemplify how to use the energy conditions to constrain the f(g ) theories of gravity , below , we study the realistic models of f(g ) gravity , which have been found to reproduce the current acceleration@xcite : @xmath89 @xmath90 where @xmath91 are all constants .
since there has been no reliable measurement for the snap parameter @xmath92 up to now , we only focus on the wec ( [ 39 ] ) and ( [ 45 ] ) in this particular case . since the inequalities are too complicated to find exact analytical expressions , so we have to take some specific values of the parameters , such as @xmath93@xcite . also , when @xmath94 or @xmath95 , the model ( [ 47 ] ) can be changed in the form @xmath96@xcite , which means @xmath97
the above two models can be rewritten as : @xmath98 @xmath99 \(1 ) when @xmath75 , we can get the corresponding wec as follows : @xmath100(-2+b)}{(2g^{n}+b)^{3}}\}\geq0 ~~(for f_{1}(g)),\end{aligned}\ ] ] @xmath101\geq0 ~~(for f_{2}(g)).\end{aligned}\ ] ] after a series of simplification , taking some present values @xmath102@xcite ,
@xmath103 and @xmath104@xcite , and @xmath105 , we can obtain the restrictions on the parameters @xmath106 , @xmath107 and @xmath108 , which satisfy the wec in eqs.([50 ] ) and ( [ 51 ] ) , respectively ( see figs.1 and 2 ) .
fig.1 shows that the wec @xmath109 can give the constraints on the parameters @xmath106 and @xmath107 in the @xmath110 model , i.e. , @xmath111 and @xmath112 , but except @xmath113 owing to the non - continuity of @xmath109 .
similarly , fig.2 illustrates that only when @xmath114 and @xmath115 , the wec @xmath109 can be satisfied in the @xmath116 model .
\(2 ) when @xmath82 , let us consider a perfect fluid composed of non - relativistic or relativistic particles with constant barotropic equation of state ( eos ) @xmath117 , @xmath118 is a constant relating to the eos by @xmath119@xcite , the corresponding wec changes into : @xmath120(-2+b)}{(2g^{n}+b)^{3}}\}\geq0~~(for f_{1}(g)),\end{aligned}\ ] ] @xmath121\geq0~~(for f_{2}(g)).\end{aligned}\ ] ] by the similar discussions to the case of @xmath75 , we can obtain the restrictions on the parameters @xmath106 , @xmath107 and @xmath108 , which respectively satisfy the wec in eqs.([52 ] ) and ( [ 53 ] ) , and are illustrated in figs.3 and 4 ( here taking @xmath122 ) . from fig.3
it is easy to see that the constraints on the parameters @xmath106 and @xmath107 for the @xmath110 model are @xmath123 and @xmath124 , which are nearly the same as the results of the @xmath75 , but in the fig.4 @xmath125 and @xmath126 for the @xmath116 model are quite different from the results of the @xmath75 .
+ ( a ) ( b ) + + ( a ) ( b ) + + ( a ) ( b ) + + ( a ) ( b ) +
modified gravity must be stable at the classical and quantum level .
there are in principle several kinds of instabilities to consider , such as dolgov - kawasaki criterion in f(r ) gravity@xcite .
below , following ref.@xcite , we will focus on the stability criterion at de sitter point in the modified f(g ) gravity . in a flat flrw background with the metric @xmath127 where @xmath128 is the scale factor and @xmath129 contains the spacial part of the metric .
the 00 component of the field equation ( [ 3 ] ) gives @xmath130 let us first discuss the stability around the de sitter point in the modified f(g ) gravity by neglecting the contribution of pressure - less matter @xmath131 and radiation @xmath132 .
the hubble parameter , @xmath133 ( at the de sitter point ) , satisfies @xmath134 where @xmath135 , and the relations @xmath136 and @xmath137 are used .
considering a linear perturbation @xmath138 about the de sitter point , eq.([55 ] ) gives @xmath139 substituting the relations @xmath140 and @xmath141 into eq.([57 ] ) , we obtain @xmath142 it follows that the effective mass squared is @xmath143 , which must be non - negative for stability
. therefore , we can obtain @xmath144 which is just the stability criterion at the de sitter point .
it follows that if the exact value of @xmath145 and a suitable form of @xmath2 can be given in the modified f(g ) gravity models , then the constraints on the parameters in the specific model can be obtained .
it is known that late - time cosmic accelerated expansion occurs under the conditions of either a power - law expansion or the equation of state of matter less than @xmath0 .
to exemplify how to use these conditions to realize the phase of accelerating expansion in the modified f(g ) gravity , now we concentrate on the model @xmath110 in ( [ 48 ] ) .
thus , by means of the action ( [ 2 ] ) and the energy density @xmath2 of perfect fluid@xcite , i.e. , @xmath146 where @xmath147 is the equation of state of perfect fluid and is taken to be a constant , the field equation ( [ 55 ] ) becomes @xmath148.\ ] ] note that the relation @xmath149 is used in the derivation of the eq.([61 ] ) .
if assuming the solution of ( [ 61 ] ) is @xmath150@xcite , then we have @xmath151 , @xmath152 , @xmath153 . substituting all these relations into ( [ 61 ] )
, we can get the following equation : @xmath154^{3 } } \times \{b^{3}(r-1)\nonumber \\ & & + 8\frac{(r-1)^{2n}r^{6n}}{t^{8n}}[-13824^{n}\times\frac{(r-1)^{n}r^{3n}}{t^{4n } } + 13824^{n}\times \frac{r(r-1)^{n}r^{3n}}{t^{4n } } + 5576^{n}n\lambda \nonumber\\ & & + 4^{3n+1}\times9^{n}n^{2}\lambda-576^{n}nr\lambda ] + 2^{3n+1}\times3^{n}b^{2}\frac{(r-1)^{n}r^{3n}}{t^{4n } } [ -3 - 5n\lambda+4n^{2}\lambda \nonumber\\&&+r(3+n\lambda ) ] + 4b\frac{(r-1)^{n}r^{3n}}{t^{4n}}\{-3^{2n+1}\times64^{n}\frac{(r-1)^{n}r^{3n}}{t^{4n } } + 5n\lambda[24^{n}\nonumber\\ & & -576^{n}\frac{(r-1)^{n}r^{3n}}{t^{4n } } ] -4n^{2}\lambda[24^{n}+2^{6n+1}\times9^{n}\frac{(r-1)^{n}r^{3n}}{t^{4n}}-576^{n } \frac{(r-1)^{n}r^{3n}}{t^{4n}}]\nonumber\\&&+r[3^{2n+1}\times64^{n}\frac{(r-1)^{n}r^{3n}}{t^{4n } } -24^{n}n\lambda+576^{n}n\lambda\frac{(r-1)^{n}r^{3n}}{t^{4n}}]\}\}.\end{aligned}\ ] ] we find six kinds of possible relationships among r , @xmath147 and n , namely , @xmath155 , @xmath156 , @xmath157 , @xmath158 , @xmath159 and @xmath160 . under the condition of a power - law expansion ( i.e. , @xmath161 ) , the corresponding regions of @xmath147 are @xmath162 for @xmath155 , @xmath163 for @xmath156 , @xmath164 for @xmath157 , @xmath165 for @xmath158 , @xmath166 for @xmath159 and @xmath167 for @xmath160 , respectively .
furthermore , by considering the equation of state of matter less than @xmath0 ( i.e. , @xmath168 ) , the relationship among @xmath169 , @xmath147 and @xmath106 , condition and candidate for late - time cosmic accelerated expansion are shown in table 1 .
the candidate for late - time cosmic accelerated expansion can be either the effective quintessence ( @xmath170 ) or the effective phantom ( @xmath171 ) . from the above discussions , it is easy to see that the results in the model are interesting . compared with the f(r ) models ,
f(g ) models are even more complicated .
since the hubble parameter can be expressed as @xmath151 , the gb term turns into @xmath172 . if @xmath173 , the early universe is in deceleration phase , which corresponds to the matter dominated phase with @xmath174 , and if @xmath175 , the late universe is in acceleration phase .
note that for the case of @xmath176 we can make similar discussions to ones in the case of @xmath75 , and obtain the same results as the ones shown in table 1 due to the constant @xmath147 .
in the present paper we have considered a modified f(g ) gravity model with coupling between matter and geometry , described by the product of the lagrange density of the matter and an arbitrary function of the gauss - bonnet term .
the proposed action represents the general extension of the standard hilbert action for the gravitational field , @xmath177l_m\}$ ] .
the field equations and the equations of motion corresponding to this model show the non - conservation of the energy - momentum tensor , the presence of an extra - force acting on test particles and the non - geodesic motion .
moreover , in the modified f(g ) gravity we have derived the energy conditions ( sec , nec , dec , wec ) when we consider @xmath75 and @xmath82 , respectively . for the sec and the nec ,
the raychaudhuri equation , which is the physical origin of them , has been used . from the derivation
, we found equivalent results can be obtained by taking the transformations @xmath178 and @xmath179 into @xmath65 and @xmath66 . by means of these transformations ,
the dec and wec in the modified f(g ) gravity have been also obtained . in order to exemplify how to use these energy conditions to constrain the modified f(g ) gravity models ,
we have considered two specific models of f(g ) gravity , i.e. , @xmath110 and @xmath116 and given the corresponding constraints on the parameters in the @xmath110 and @xmath116 models . since there has been no reliable measurement for the snap parameter @xmath92 up to now , we only focus on the wec in this particular case . by analysis on figs.1 and 2
we have given the constraints on the parameters in the @xmath110 and @xmath116 models satisfying the weak energy conditions when @xmath75 . by the similar discussions to the case of @xmath75 ,
when @xmath82 the restrictions on the parameters @xmath106 , @xmath107 and @xmath108 have been also illustrated in figs.3 and 4 , from which we have found that in the two different forms of @xmath2 the constraints on the parameters for the @xmath110 model are nearly the same , but quite different for the @xmath116 model .
furthermore , we have derived the stability criterion at the de sitter point for the modified f(g ) gravity models , which means that the modified f(g ) gravity models may be stable .
in addition , we have researched the conditions for late - time cosmic accelerated expansion in the modified f(g ) gravity . concretely , for
the two different forms of @xmath2 , the relationship among r , @xmath147 and n have been respectively given in the model @xmath180 , and by using the conditions of power - law accelerated expansion and the equation of state of matter less than @xmath0 , the constraints on the parameter n have been obtained , which are exactly the same in the two different forms of @xmath2 . the candidate for late - time
cosmic accelerated expansion would be either the effective quintessence ( @xmath181 ) or the effective phantom ( @xmath171 ) , which could be determined by choosing n properly .
of course , other forms of f(g ) gravity models with curvature - matter coupling will be considered in our following investigations .
* acknowledgments * the research work is supported by the national natural science foundation of china ( 11147150 ) , the natural science foundation of education department of liaoning province ( l2011189 ) , the natural science foundation of liaoning province , china ( grant no.20102124 ) the nsfc ( 11175077 ) of p.r . china .
riess , et al .
j. 116 ( 1998 ) 1009 .
s. perlmutter , et al .
, astrophys .
j. 517 ( 1999 ) 565 .
copeland , m. sami , s. tsujikawa , int .
d 15 ( 2006 ) 1753 - 1936 . c. brans , r.h .
dicke , phys .
124 ( 1961 ) 925 ; + v. faraoni , cosmology in scalar - tensor gravity ( kluwer academic , dordrecht ) 2004 ; + dvali , et al .
b 485 ( 2000 ) 208 ; + r. maartens , living rev . rel . 7
( 2004 ) 7 ; + j.d .
bekenstein , phys . rev .
d 70 ( 2004 ) 083509 ; + t. jacobson , d. mattingly , phys .
d 64 ( 2001 ) 024028 .
s. capozziello , s. nojiri , s.d .
odintsov and a. troisi , phys .
b 639 ( 2006 ) 135 ; + l. amendola , d.polarskiands .
tsujikawa , phys .
98 ( 2007 ) 131302 ; + s. capozziello , s. nojiri , s.d .
odintsov and a. troisi , phys .
b 639 ( 2006 ) 135 ; + s. nojiri and s.d .
odintsov , phys .
d 74 ( 2006 ) 086005 ; + j. santos , j.s .
alcaniz , m.j .
reboucas and f.c .
carvalho , phys .
d 76 ( 2007 ) 083513 ; + s. nojiri and s.d .
odintsov , phys .
b 652 ( 2007 ) 343 ; + s. tsujikawa , phys . rev .
d 77 ( 2008 ) 023507 ; t.p .
sotiriou , v. faraoni , rev .
82 ( 2010 ) 451 - 497 ; + s. nojiri , s.d .
odintsov , econf c0602061 ( 2006 ) 06 ; int . j. geom .
4 ( 2007 ) 115 - 146 ; + s. capozziello , int .
d 11 ( 2002 ) 483 ; + s. nojiri , s.d .
odintsov , gen .
rel . grav .
36 ( 2004 ) 1765 - 1780 ; + a.d .
dolgov , m. kawasaki , phys .
b 573 ( 2003 ) 1 ; + v. faraoni , phys . rev .
d 74 ( 2006 ) 104017 ; + v. faraoni , phys . rev .
d 76 ( 2007 ) 127501 .
s. nojiri , s.d .
odintsov , phys .
b 631 ( 2005 ) 1 ; + g. cognola , e. elizalde , s. nojiri , s.d .
odintsov , s. zerbini , phys .
d 73 ( 2006 ) 084007 ; + g. cognola , e. elizalde , s. nojiri , s.d .
odintsov , s. zerbini , phys .
d 75 ( 2007 ) 086002 ; + b.j .
barrow , d.f .
mota , phys .
d 76 ( 2007 ) 044027 ; + m. alimohammadi , a. ghalee , phys . rev .
d 79 ( 2009 ) 063006 ; + c.g .
boehmer , f.s.n .
lobo , phys .
d 79 ( 2009 ) 067504 ; + s.y .
zhou , e.j.copeland , p.m. saffin , j. cosmol . astropart .
phys . 0907 ( 2009 ) 009 ; + n. goheer , r. goswami , p.k.s .
dunsby , k.ananda , phys .
d 79 ( 2009 ) 121301 ; + j. sadeghi , m.r .
setare , a. banijamali , phys .
b 679 ( 2009 ) 302 . | a modified f(g ) gravity model with coupling between matter and geometry is proposed , which is described by the product of the lagrange density of the matter and an arbitrary function of the gauss - bonnet term .
the field equations and the equations of motion corresponding to this model show the non - conservation of the energy - momentum tensor , the presence of an extra - force acting on test particles and the non - geodesic motion .
moreover , the energy conditions and the stability criterion at de sitter point in the modified f(g ) gravity models with curvature - matter coupling are derived , which can degenerate to the well - known energy conditions in general relativity . furthermore ,
in order to get some insight on the meaning of these energy conditions , we apply them to the specific models of f(g ) gravity and the corresponding constraints on the models are given .
in addition , the conditions and the candidate for late - time cosmic accelerated expansion in the modified f(g ) gravity are studied by means of conditions of power - law expansion and the equation of state of matter less than @xmath0 . |
the main aim of the author in this paper is to present a new tool which can make easier the study of the real dynamics of families of iterative methods which depends on a certain parameter or even the study of an iterative method applied to a uniparametric family of polynomials .
this tool can be modified in order to extend , amongst other ones , to methods which needs two approximations as for example , secant - type methods , modified newton s method , etc .
the dynamics of iterative methods used for solving nonlinear equations in complex plane has been studied recently by many authors @xcite .
there exists a belief that real dynamics is included in the complex dynamics but this is not true at all .
for example in the real dynamics one can proof the monotone convergence which does not exist in the complex plane , there exists also asymptotes in the real dynamics but in the complex one that concept has no sense or the point @xmath0 in the complex plane can be studied as another point but in the real line it is not possible . as a consequence
real dynamics is not contained in the complex dynamics and both must be studied separately .
taking into account that distinction , many authors @xcite have begun to study it since few years ago .
if one focus the attention on families of iterative methods studied in the complex plane , parameter spaces have given rise to methods whose dynamics are not well - known .
these parameter spaces consist on studying the orbits of the free critical points associating each point of the plane with a complex value of the parameter .
several authors @xcite have studied really interesting dynamical and parameter planes in which they have found some anomalies such as convergence to @xmath1-cycles , convergence to @xmath2 , or even chaotical behavior . in the real line , there exists tools such as feigenbaum diagrams or lyapunov exponents that allow us to study what happens with a concrete point , but it is really hard to study each point in a separate way .
this is the main motivation of the author to present the new tool called _ the convergence plane_. _ the convergence plane _ is obtained by associating each point of the plane with a value of the starting point and a value of the parameter .
that is , _ the convergence plane _ is based on taking the vertical axis as the value of the parameter and the horizontal axis as the starting point , so every point in the plane represents an initial estimation and a member of the family .
if one draws a straight horizontal line in a concrete value of the parameter , the dynamical behavior , for that value , for every starting point is on that line . on the other hand ,
if the straight line is vertical , the dynamical behavior for that starting point and every value of the parameter is on that line , this is the information that gives feigenbaum diagrams or lyapunov exponents , so both tools are included .
the rest of the paper is organized as follows : in section [ s2 ] the algorithm of _ the convergence plane _ is shown and in section [ s3 ] an example of _ the convergence plane _ associated to the damped newton s method applied to the polynomial @xmath3 is provided in order to validate the tool .
finally , the conclusion are shown in the concluding section [ s4 ] .
as it is said in the introduction each point of the plane corresponds to a starting point and a value of the parameter , in other words the pair @xmath4 represents that the study is developed using @xmath5 as the starting point and @xmath6 as the value of the parameter .
the algorithm of this new tool is the following one : * the fixed points of the method must be computed . * then , a color is assigned to each fixed point .
* moreover , the region in which one wants to study the family , @xmath7 , the maximum number of iterations , @xmath8 , and the tolerance , @xmath9 , are prefixed . * then , a grid of @xmath10 points in @xmath7 of the initial points and values of the parameter must be chosen . * if after @xmath8 iterations of the family with @xmath6 as the value of the parameter , the point @xmath5 does not converge to any of the fixed points that point must be black . * if one is interested on representing the convergence to any cycle or other behavior such as convergence to extraneous fixed points , divergence etc .
, there exists the possibility of assigning a color to that behaviors too .
once _ the convergence plane _ has been computed it is easy to distinguish the pairs @xmath11 for which the element of the family is convergent to any of the roots using @xmath5 as a starting point .
so this tool provides a global vision about what points converges and shows what are the best choices of the parameters to ensure the greatest basin of attraction .
moreover , it can be used also as a tool that show how the basins of attraction of a family changes with the value of the parameter or even to study the convergence of methods which depend on 2 points such as , for example , secant - type methods . in this moment
, the author is going to explain how figure [ fig1 ] and figure [ fig2 ] of this paper were generated . to do this , it is shown the mathematica programs that has been used which are a modification of the ones that appears in @xcite . in concrete ,
in the example the region will be @xmath12\times [ 2,2.26]$ ] ( the season of taking that values is in @xcite and will appear in section 3 ) .
moreover , a number of iterations @xmath13 , a tolerance of @xmath14 and a grid of @xmath15 have been taken .
first of all , the function , the roots and the procedure that identifies the root to which converge the iterations have been defined as .... f[x_]:=x^3-x ; rootf[1]=-1 ; rootf[2]=0 ; rootf[3]=1 ; rootposition = compile[{{z,_real } } , which [ abs[z - rootf[1 ] ] < 10.0^(-6 ) , 1 , abs[z - rootf[2 ] ] <
10.0^(-6 ) , 2 , abs[z - rootf[3 ] ] < 10.0^(-6 ) , 3 , abs[z ] >
10.0^(3 ) , 11 , true , 0 ] , { { rootf[_],_real } } ] .... then , the iteration method is the following .... iternewtonlamda = compile[{{x,_real},{k,_real}},x - k*f[x]/f'[x ] ] .... the algorithm used to show if the iteration of _ itermethod _ of a point converges to a root or a cycle is the following .... itercoloralgorithm[itermethod_,xx_,yy_,lim _ ] : = block[{z , z2,kk , ct , r } , z = xx ; kk = yy ; ct = 0;r = rootposition[z ] ; while[(r==0 ) & & ( ct < lim),++ct ; z = itermethod[z , kk ] ; r = rootposition[z ] ; ] ; if[r==0,z2= itermethod[z , kk];z2= itermethod[z2,kk ] ; if[abs[z2-z]<10^(-6),r=4,z2=itermethod[z2,kk ] ; if[abs[z2-z]<10^(-6),r=5,z2=itermethod[z2,kk ] ; if[abs[z2-z]<10^(-6),r=6,z2=itermethod[z2,kk ] ; if[abs[z2-z]<10^(-6),r=7,z2=itermethod[z2,kk ] ; if[abs[z2-z]<10^(-6),r=8,z2=itermethod[z2,kk ] ; if[abs[z2-z]<10^(-6),r=9,z2=itermethod[z2,kk ] ; if[abs[z2-z]<10^(-6),r=10,z2=itermethod[z2,kk ] ; ] ] ] ] ] ] ] ] if[head[r]==which , r = 0 ] ; ( * " which " unevaluated * ) return[n[r+ct/(lim+0.001 ) ] ] ] .... the palette of colors used is defined as .... convergencecolor[p _ ] : = switch[integerpart[11p ] , 11 , cmykcolor[0.0,0.0,0.0,0.0],(*white * ) 10 , cmykcolor[0.0,1.0,1.0,0.8],(*dark red * ) 9 , cmykcolor[0.0,0.0,1.0,0.5],(*dark yellow * ) 8 , cmykcolor[1.0,0.0,1.0,0.5],(*dark green * ) 7 , cmykcolor[1.0,0.0,0.0,0.5],(*dark blue * ) 6 , cmykcolor[0.0,0.5,1.0,0.0],(*orange * ) 5 , cmykcolor[1.0,0.0,1.0,0.0],(*green * ) 4 , cmykcolor[0.0,1.0,1.0,0.0],(*red * ) 3 , cmykcolor[0.0,0.0,1.0,0.0],(*yellow * ) 2 , cmykcolor[0.0,1.0,0.0,0.0],(*magenta * ) 1 , cmykcolor[1.0,0.0,0.0,0.0],(*cyan * ) 0 , cmykcolor[0.0,0.0,0.0,1.0](*black * ) ] .... the function used by the author to plot the convergence plane is the following .... plotconvergenceplane[itermethod_,points _ ] : = densityplot[itercoloralgorithm[itermethod , x , k , limiterations ] , { x , xxmin , xxmax } , { k , kkmin , kkmax } , plotrange->{0,11 } , plotpoints->points , mesh->false , colorfunction->convergencecolor ] .... then , a graphic is obtained in this way ( notice that we avoid overflow and underflow errors and other errors by means of using the instruction _ off _ ) : .... numberpoints = 1024 ; limiterations = 4000 ; xxmin = -2.0 ; xxmax = 2 ; kkmin = 0.0 ; kkmax = 2.6 ; off[general::ovfl ] ; off[general::unfl ] ; off[infinity::indet ] off[compiledfunction::cccx ] ; off[compiledfunction::cfn ] ; off[compiledfunction::cfcx ] ; off[compiledfunction::cfex ] ; off[compiledfunction::crcx ] ; off[compiledfunction::cfse ] ; off[compiledfunction::ilsm ] ; off[compiledfunction::cfsa ] ; plotconvergenceplane[iternewtonlamda , numberpoints ] ....
to show the goodness of _ the convergence plane _ we are going to study the dynamics of the damped newton s method which has the following form @xmath16 where @xmath17 is a polynomial with real coefficients and we take @xmath18 . in concrete , in @xcite , the author has made an extensive and deep study about the real dynamics of the damped newton s method applied to polynomials of degrees 2 , 3 , 4 and 5 . in order to proof that the tool works properly we are going to apply the iterative method to find the real roots of a cubic polynomial
the scaling theorem @xcite allows up to suitable change of coordinates , to reduce the study of the dynamics of iterations of general maps , to the study of specific families of iterations of simpler maps .
specifically , the study of cubic polynomials reduce to the study of @xmath19 , @xmath20 @xmath3 and the uniparametric family @xmath21 . in @xcite
the author shows the real dynamics of the damped newton s method applied to @xmath3 are not easy and that there exists some cycles and chaotic behavior of the iterations of some points by means of using lyapunov exponents , feigenbaum diagrams and analytical techniques .
we are going to use the new tool _ the convergence plane _ to study the dynamics of the damped newton s method applied to @xmath22 .
we denote the three roots of @xmath22 as @xmath23 , @xmath24 and @xmath25 . before , applying the new tool we have to see what information give to us the lyapunov exponents and feigenbaum diagrams .
these two tools gives information only about the point which is being iterated . in concrete ,
the lyapunov exponent is defined , for an orbit @xmath26 as @xmath27 the applicability of this tool resides in the following result .
@xcite an orbit @xmath28 is chaotic if and only if the following conditions hold : * the orbit is not asymptotically periodic . *
@xmath29 . as a consequence ,
this tool shows which orbits are chaotical or not . on the other hand ,
the feigenbaum diagrams shows if the orbit of a point converges to a cycle , to a point , if it is chaotic or it diverges . in figure [ fff1 ]
it is shown the results obtained using the both two tools .
fftt.pngfeigenbaum diagram taking into account that a brief study has been made in @xcite and the idea of not making this paper very long we are going to focus on the interval @xmath30 which is sufficient to show the powerful of the new tool . in figure [ ff1 ]
we can see both tools centered on that interval .
and we distinguish three different zones .
fft.pngfeigenbaum diagram there exists 3 clear zones of different behavior . in the first one which corresponds with the interval @xmath31 ( see left side of the figure [ ff2 ] ) , we see the zone in which the iterations converge to the fixed points of the damped newton s method or equivalently to the roots of the polynomial @xmath22 .
the second zone , shown in the center of figure [ ff2 ] is where it appears cycles of different orders , in concrete cycles of order @xmath32 , @xmath33 and @xmath34 .
the third zone , shown in the right side of figure [ ff2 ] is where we found chaotical behavior or convergence to different cycles of periods different to @xmath35 .
ll2.png@xmath36$]ll3.png@xmath37$]ff2.png@xmath36$]ff3.png@xmath37 $ ] now if we use _ the convergence plane _ in the region @xmath12\times[0,2.6]$ ] ( see figure [ fig1 ] ) the goodness of this tool is going to be proof . in this case
, we use the program _
mathematica 5.0 _ ( the code appears in section 2.1 ) as in @xcite , with tolerance @xmath38 , a maximum of 4000 iterations and the following palette of colors : * cian , if the iterations converge to root @xmath39 , magenta , if the iterations converge to root @xmath40 and yellow , if the iterations converge to root @xmath41 . * red , if the iterations converge to a @xmath32-cycle , green , if converge to a @xmath42-cycle , orange , if converge to a @xmath33-cycle , dark blue , if converge to a @xmath43-cycle , dark green , if to a @xmath44-cycle , dark yellow , if converge to a @xmath45-cycle and dark red if the iterations converge to a @xmath34-cycle . * white , if the iterations diverge to @xmath2 . * black , in other case
. on the region @xmath46\times[0,2.6]$].,title="fig:",scaledwidth=70.0% ] + again we distinguish 3 clear zones with different dynamical behavior , which corresponds with the 3 zones using feiganbaum diagrams and lyapunov exponents .
the first one , shown in the left hand of figure [ fig2 ] corresponds to the interval @xmath47 $ ] and it is the zone in which every point ( except the poles and its preimages ) , converges to any of the roots of the polynomial or equivalently to any fixed point of the damped newton s method .
additionally , in this zone the tool gives the idea of how the basins of attraction changes with the value of @xmath48 .
for example , the basin of @xmath24 is getting lower when the value of @xmath48 is closer to two , an the basin of the other two roots are getting bigger as the parameter is closer to @xmath32 .
furthermore , it it shown that the julia set gets more intricate when the value of the parameter increases until @xmath32 . the second zone , see the center of figure [ fig2 ] , corresponds with the zone in which there exists convergence to cycles of orders @xmath32 , @xmath33 and @xmath34 . and
the third zone , shown in the right hand of the figure [ fig2 ] , the zone in which there exists chaotical behavior and convergence to cycles of order different to @xmath35 . in concrete in the interval @xmath49 appears cycles of order @xmath43 , @xmath44 and @xmath45 . summarizing
, the author has shown the helpfulness of this new tool using it to study the behavior of the damped newton s method applied to a cubic polynomial .
the conclusions drawn this study are : the damped newton s method is a good method as a root - finding algorithm if @xmath50 $ ] , if @xmath51 $ ] , the iteration can converge to cycles , diverge to infinity , etc . ; the basin of attraction of @xmath24 increases when the damping factor decreases , and the other basin increases with the parameter ; the julia set associated with the damped newton s method applied to @xmath3 is more intricate as the damping factor increases to @xmath32 .
all these conclusions coincide with the ones given in @xcite . p2.png@xmath36$]p3.png@xmath37 $ ]
in the present paper , the author has presented a new tool that allows to study the convergence of a family of iterative methods , taking into account every initial point . _ the convergence plane _ , includes the information given by tools such as lyapunov exponents and feigenbaum diagrams , indeed , the new tool provides more information because it considers each initial point of the real line and every value of the parameter .
moreover , this technique can be used to choose the best value of the parameter that ensure the convergence zone as large as possible and shows how the basins of attraction changes with the value of the parameter . on the other hand , this technique can be modified in order to : study methods applied to polynomials which depends on a parameter , to other kind of methods such as two - point methods ( for example secant - type methods ) , methods applied to non - differentiable functions , etc . therefore , this new tool can be used in a large amount of situations , making the study of real dynamics easier , deeper and in a more compact way . | in this paper , the author presents a new tool , called _ the convergence plane _ , that allows to study the real dynamics of iterative methods whose iterations depends on one parameter in an easy and compact way .
this tool can be used , inter alia , to find the elements of a family that have good convergence properties and discard the bad ones or to see how the basins of attraction changes along the elements of the family . to show the applicability of the tool an example of the dynamics of the damped newton s method applied to a cubic polynomial is presented . +
* keywords : * real dynamics , nonlinear equations , the convergence plane , iterative methods , basins of attraction . + +
ngel alberto magren + universidad de la rioja + departamento de matemticas y computacin + 26002 logroo , la rioja , spain + [email protected] + |
many liquids when cooled below the melting temperature do not crystallize . instead , as the temperature is lowered , the relaxation time ( or the viscosity ) increases dramatically , eventually becoming so large that the liquid appears frozen on experimental timescales . for all practical purposes
the system has solidified yet there is no long range order .
the system is said to have become a glass . despite over fifty years of work , a consensus is lacking on a theory of this phenomenon@xcite . even as basic a question as the existence of a genuine thermodynamic glass transition is difficult to decide experimentally . therefore , much work in this field has been numerical .
simulations in the continuum of glassy systems remain difficult , and , despite the substantial progress in this direction@xcite , there are still many unresolved issues . because of these problems it would be useful to have simplified lattice models that behave in some respects like glasses and which do not have quenched disorder . recently , lipowski@xcite and lipowski and johnston@xcite studied such a model by monte carlo simulations and provided evidence of glassy behavior .
the model is defined by the hamiltonian @xmath0 where the spins sit on a cubic lattice and the interactions are between the four spins at the corners of each plaquette .
the spin variables take values @xmath1 and @xmath2 is the strength of the ferromagnetic coupling .
( we will take @xmath3 in all that follows . )
their study was motivated by the fact that the spin - glass version of this model was found numerically @xcite to have several features in common with structural glasses .
in addition , there is a close connection@xcite between the plaquette model , eq .
[ eqn1 ] , and the ising model with competing interactions studied in the context of glasses by shore and sethna@xcite .
lipowski and johnston@xcite showed that , in equilibrium , the plaquette model has a first order transition as the temperature is lowered but that this transition is never observed in simulations within accessible timescales .
instead , one sees what appears to be a dynamical transition to a glassy state@xcite .
the work of these authors concentrated on single time quantities . to see clear signatures of glassy behaviour and to understand which particular aspects of glasses are present in this model , it is important to study two time quantities such as autocorrelation and response functions .
we do just that in this paper and our findings are organized as follows : in section ii , we summarize the results of lipowski and johnston on the statics of the plaquette model and the strong metastability that they observed . in section iii , we examine the metastable state in greater detail .
in particular , we calculate the timescale for nucleation and show that it is indeed extremely large . in section iv , we present results for the energy auto - correlation function in the supercooled phase which is unambiguously described by a stretched exponential .
we investigate the behavior of the associated timescale as a function of temperature , showing that it diverges with a power law at a temperature indistinguishable from the lower limit of metastability .
this temperature is that which the authors of @xcite called the glass transition temperature @xmath4 . in section v , we study the dynamical behavior below @xmath4 . the spin - spin autocorrelation function and overlap distribution function exhibit aging of a form that is characteristic of glasses , showing that the plaquette model has a complex free - energy landscape .
these findings provide further evidence to support the conjecture that the plaquette model exhibits an ideal glass transition .
in equilibrium , the model defined by eq . [ eqn1 ] has a first order transition at @xmath5 .
this system possesses a novel symmetry : flipping all the spins in any plane of the cubic lattice leaves the hamiltonian invariant , leading to a degeneracy of ground states @xmath6 for a cubic system of size @xmath7 in @xmath8 dimensions .
note , however , that the ground state entropy is finite . because of this symmetry , the magnetisation is not a good order parameter . instead , one can use the internal energy per spin , @xmath9 , to investigate this model further .
lipowski and johnston looked at the dynamics of @xmath9 following a quench from a high temperature state to a temperature @xmath10@xcite .
they found different behaviour depending on the value of @xmath11 .
in particular , for @xmath12 , @xmath9 appears not to relax to its value in the low temperature phase .
rather it settles into a plateau at an energy higher than that of the true low temperature phase at the same temperature .
furthermore , the time the system spends on the plateau ( in a `` supercooled '' state ) appears to increase at least exponentially with system size , suggesting the existence of a genuine metastable phase .
such behaviour , if present , would indeed be strange because one does not expect true metastability ( i.e. two local minima of the free energy separated by diverging free energy barriers ) in systems with finite range interactions .
moreover , nucleation theory tells us that the larger the system , the shorter the lifetime of a metastable state , provided that the system is large enough to support a nucleating droplet . to understand whether the strong metastability is indeed a finite size artifact , lipowski and johnston studied a quench starting from a configuration with a droplet of the low temperature phase in a high temperature background .
they reported that for sufficiently large seeds , the system always relaxed to the low temperature phase , concluding that ordinary nucleation was operative .
however , this does not explain why the supercooled phase appears so stable on observable timescales .
the analysis we report in the next section provides an explanation for this phenomenon . for quenches to below @xmath13 the behaviour
was found to be quite different .
the plateau disappeared and , after initial transients , the energy decreased extremely slowly .
moreover , it appeared to relax not to the value in the low temperature phase , but to a value extensively higher .
they identified this state as the glassy phase .
finally , they have recently shown that a characteristic zero - temperature length increases very slowly with the inverse cooling rate , as is seen in ordinary glasses@xcite . to summarize , this model has many properties that are similar to those seen in real glasses : a first order transition analogous to crystal melting , strong metastability effects in the supercooled phase , and a kinetic transition to a glassy state characterised by slow dynamics .
in the next section we will discuss the anomalous metastability described above before looking in some detail at the dynamical behavior of this model by studying the autocorrelation functions both in the supercooled phase and in the `` glassy '' state .
we begin by summarizing the basics of nucleation theory@xcite .
the idea is that if the system is in a metastable state , the process of finding the true free - energy minimum involves the formation of a droplet of the stable phase which can grow and take over the whole system .
phenomenologically , one can express the free energy barrier to forming a droplet of linear dimension @xmath14 as @xmath15 in @xmath8 dimensions . here
@xmath16 is the surface tension between the metastable and stable phases and @xmath17 is the difference in the bulk free energy densities of the two phases .
a and b are constants characterizing the geometrical shape of the droplet .
because the surface term is a cost and the bulk term is a gain , one can maximize @xmath18 to find a critical radius @xmath19 . droplets of a radius greater than @xmath20 can grow and take over the whole systems while those of radius less that @xmath20 will shrink . at low temperatures and in three dimensions , the timescale for nucleation of a droplet is then given by @xmath21 .
for the plaquette problem , the free energy difference between the two phases can be estimated from the data in @xcite .
we find @xmath22 .
the surface tension @xmath16 at @xmath23 can be calculated from the size dependence of the mean time it takes the system to flip between the high and low temperature phases .
this time @xmath24 is given by @xmath25 , where @xmath7 is the system size and the factor of @xmath26 comes from using period boundary conditions . we have performed standard single spin - flip monte carlo simulations to determine this transition time numerically for systems of size @xmath27 .
the transitions are identified from jumps in the energy between the values appropriate to the high and low temperature phases . for each system size
, an ` effective transition temperature ' , @xmath28 , is determined by the requirement that the system spend equal time in both phases , i.e. at each of the two possible values of the energy .
this gives @xmath29 , 4.19 , 4.03 , 3.92 and 3.86 for @xmath30 , 5 , 6 , 7 and 8 respectively .
the mean transition time , @xmath31 , is measured at @xmath28 , and the surface tension , @xmath16 , obtained by plotting @xmath32 against @xmath33 .
the data is presented in fig .
[ fig1 ] . for large @xmath7
, the points should approach a straight line with slope @xmath34 . since there is clear curvature in the data for the small values of @xmath7 available , we attempt an extrapolation to large @xmath7 by plotting @xmath35\ln \tau_{\rm flip}$ ] against @xmath36 , as shown in the inset to fig .
the intercept on the vertical axis gives @xmath34 .
while the extrapolation to @xmath37 is necessarily subjective , from the curvature of the data it is clear that any reasonable procedure will give a value of @xmath34 greater than the value @xmath38 obtained from a linear extrapolation of the last two points , and less than the value @xmath39 given by the @xmath40 point . to obtain lower bounds on critical drop sizes and nucleation times we use the estimate @xmath41 obtained from the linear extrapolation . assuming that @xmath16 depends only weakly on @xmath11 close to @xmath23 , and taking @xmath42
, we estimate the critical droplet radius at @xmath43 to be @xmath44 for a cubic droplet ( @xmath45 , @xmath46 ) and @xmath47 for a spherical droplet ( @xmath48 , @xmath49 ) . the corresponding time to nucleate
a cubic droplet is of the order of @xmath50 mcs ( and @xmath51 mcs for a spherical droplet ) .
these findings provide an explanation for the results of @xcite , in which it was noted that the relaxation time of the metastable phase grows rapidly with system size . in @xcite ,
the system sizes simulated were all less than , or of the same order as , @xmath20 . in this limit
, the free energy barrier to nucleation is set by the system size alone , as a droplet of the equilibrium phase can not fully form .
consequently , the lifetime of the metastable state will grow exponentially with the size of the system@xcite as reported .
we thus conclude that the supercooled state is indeed metastable but its lifetime is many orders of magnitude larger than the times accessible in monte carlo simulations .
the philosophy underlying the search for models without disorder that show glassy behavior is that in glasses , inhomogeneities are self - generated , even though the microscopic hamiltonian has no intrinsic disorder .
thus there is a spontaneous separation of the degrees of freedom into variables that exhibit either slow or fast relaxation .
one expects that it is the slow degrees of freedom that freeze as the glass transition is approached .
however , this separation is not necessarily a priori obvious , and one has to be careful about which quantities one studies .
it might happen , for example , that the spin - spin autocorrelations do not show any anomalous behavior , but that some complicated functions of the spins show a dramatic slowing down .
for this reason it is often useful to look at autocorrelation functions of global quantities because these presumably pick out the longest timescale or the slowest degree of freedom .
for the system defined by eq .
[ eqn1 ] , one such quantity is the energy per spin , @xmath9 , and we now discuss our findings for the energy autocorrelation function in the range of temperatures @xmath52 .
our results have been obtained from monte carlo simulations using single spin - flip glauber dynamics .
we have considered cubic systems of linear size @xmath7 , with @xmath7 ranging from @xmath53 to @xmath54 .
most of the data presented here is for a system size of @xmath55 . to measure the energy autocorrelation function at some temperature @xmath11 we start from a random configuration ( infinite temperature ) and quench to the temperature @xmath11 .
we then wait for a time @xmath56 and measure the autocorrelation function @xmath57 for subsequent times , @xmath58 , where @xmath59 .
the angular brackets denote averages over random initial conditions . in equilibrium one
expects @xmath57 to be independent of @xmath56 and it is only in equilibrium that one can define a meaningful timescale associated with relaxation . in the supercooled phase , we found that equilibrium is easily achieved in terms of this definition even though the system is in a metastable state .
in particular , for any temperature @xmath52 , the autocorrelation function becomes independent of @xmath56 once @xmath56 becomes of the order of @xmath60 mcs .
note that this time is at least @xmath61 orders of magnitude smaller than the lifetime of this metastable state . in fig .
[ fig2 ] we show @xmath57 for three different temperatures below @xmath23 and above @xmath4 , together with attempted fits to stretched exponentials of the form @xmath62 as can be seen from the figure , the fits are extremely good and we can identify a timescale for relaxation , @xmath63 , at each temperature .
the exponent @xmath64 is also temperature dependent and is found to lie in the range @xmath65 . in fig .
[ fig3 ] we plot the behaviour of @xmath63 as a function of @xmath11 .
it appears to increase sharply as @xmath11 is reduced and can be fit quite accurately by a power - law divergence , @xmath66 as shown in the figure .
this gives a `` critical '' temperature very close to that which was identified in @xcite as the lower limit of stability , @xmath4 , and interpreted as the glass transition temperature .
both stretched exponential relaxation functions and a diverging relaxation time are features of real glasses .
we also tried a vogel - fulcher ( stretched - exponential ) fit to @xmath63 but found that the corresponding `` critical temperature '' is significantly lower than @xmath4 . as we will show in the next section
, @xmath4 signifies the onset of aging phenomena and it is thus impossible to define a timescale below this temperature .
( we have also measured the spin - spin autocorrelation function but found that the behaviour was much harder to interpret .
in fact , the difficulty of interpreting relaxational behavior of local quantities has been noted before in molecular dynamics studies of lennard - jones systems in two dimensions @xcite . )
= 0.9 a possible explanation for this divergence is to note that in mean - field theory , the plaquette model has a first order transition at a temperature quite close to @xmath23@xcite . in any mean - field first order transition , one expects a region of metastability delimited by the spinodal temperature . above this temperature ,
there is genuine metastability and one expects timescales of autocorrelation functions to diverge as the spinodal temperature is approached when one moves away from mean - field , genuine metastability is no longer present and instead the spinodal gets rounded out .
our results for the autocorrelation functions , and those of the authors of @xcite for the energy relaxation , suggest that the spinodal is actually remarkably well - defined in this short - range system .
it is therefore plausible that what we are calling the glass transition temperature is really a remnant of the mean - field spinodal .
this too is something that is seen in lennard - jones systems in two dimensions @xcite .
however , the stretched exponential form of @xmath67 would not be expected in any simple mean - field treatment .
thus , the short - range plaquette model has a first order transition analogous to crystal melting , but because of the extraordinarily long - lived supercooled phase , it also exhibits timescales for relaxation that diverge at @xmath4 .
we would like to emphasize that in this model @xmath4 is not strictly well - defined . in fact , we expect that as one gets closer and closer to @xmath4 , or waits for longer in the metastable state , the divergence we are seeing will get rounded out . numerically , however , it is not possible to see this rounding in any reasonable amount of time .
this again is a feature in common with glass forming liquids .
for a quench to a temperature @xmath68 , as we have seen above , the system becomes trapped in a metastable state for times up to @xmath69 mcs . within this time window ,
two - time correlation functions become waiting - time independent and the system behaves as if in equilibrium .
this is no longer the case if @xmath70 .
the system now continues to evolve slowly for all accessible times and two - time quantities exhibit aging . to investigate this behaviour
we have measured the spin - spin autocorrelation function , @xmath71 for a range of waiting - times @xmath56 .
our results are shown in fig .
there is a strong dependence on @xmath56 for all times and temperatures we have considered . in the long time regime , @xmath72 decays as a power of @xmath58 with an exponent of around @xmath73 .
similar behaviour has been noted in md simulations of binary lennard - jones mixtures@xcite . we have also attempted to collapse the data at a fixed temperature with a scaling form @xmath74 where @xmath75 . while the fit is not perfect ,
our best estimate for @xmath76 is @xmath77 .
this type of behaviour has been termed super - aging@xcite as the characteristic timescale for relaxation , @xmath78 , grows faster than the weighting time itself .
note that the spin - glass version of this model exhibits sub - aging with @xmath79@xcite .
= 0.8 however , simply having a strong waiting - time dependence in a two - time correlation function does not imply glassy behaviour .
indeed , simple coarsening systems exhibit aging@xcite . to distinguish this type of dynamics , type i aging@xcite , from glassy dynamics , type ii aging ,
we have considered the overlap distribution function @xmath80 this is measured by first relaxing the system from a disordered state for a time @xmath56 . then two copies of the system are made , @xmath81 and @xmath82 , and each is evolved for a further time @xmath58 with independent thermal noise . in fig .
[ fig5 ] we show a plot of @xmath83 vs. @xmath84 for three different waiting times . in each case
@xmath85 as @xmath86 , indicating that the two copies of the system continue to move apart from each other irrespective of when the copies were made .
this behaviour is indicative of type ii , glassy aging@xcite .
note that this behaviour is not an automatic consequence of the existence of infinitely many ground states .
for example , we expect that for a triangular lattice anti - ferromagnet , @xmath83 would tend to a constant as @xmath84 tends to zero , despite the non - zero ground - state entropy per spin .
finally , we have observed that @xmath72 and @xmath87 are related by @xmath88 over a wide range of timescales .
this relation is known to hold in equilibrium , where both functions are independent of @xmath56@xcite .
it has also been verified in the aging regime of some trap models as well as in the spin - glass version of the plaquette model@xcite .
these results provide evidence that the free - energy landscape of the ferromagnetic plaquette model is indeed complex and has many mutually inaccessible minima .
this property is similar to what is seen in structural glasses and in mean field spin - glass models .
we have shown that the ferromagnetic plaquette model has many characteristics in common with glass forming liquids .
we provided an explanation for the long - lived nature of the supercooled phase in terms of standard nucleation theory . by measuring the two - time correlation functions , we have observed two distinct kinds of behavior : ( i ) in the supercooled phase
, the system appears to be stationary and the timescale of the energy autocorrelation functions diverges at a temperature identified in earlier work as a glass transition temperature @xmath4 , and ( ii ) below @xmath4 the spin - spin autocorrelation functions exhibit aging of a form characteristic of glasses .
our findings , along with the work reported in @xcite , indicate that a deeper understanding of this model would indeed be worthwhile . | recent work has suggested the existence of glassy behavior in a ferromagnetic model with a four - spin interaction . motivated by these findings , we have studied the dynamics of this model using monte carlo simulations with particular attention being paid to two - time quantities .
we find that the system shares many features in common with glass forming liquids .
in particular , the model exhibits : ( i ) a very long - lived metastable state , ( ii ) autocorrelation functions that show stretched exponential relaxation , ( iii ) a non - equilibrium timescale that appears to diverge at a well defined temperature , and ( iv ) low temperature aging behaviour characteristic of glasses . 2 |
* chemicals .
* 1,4-diphenyl-1,3-butadiyne-3,3``,5,5''-tetracarboxylic acid ( dpbdtc ) and _ p_-terphenyl-3,5,3``,5''-tetracarboxylic acid ( tptc ) were synthesized in - house , details of their synthesis can be found in the accompanying supplementary material and ref . respectively .
other chemicals used were coronene ( @xmath5695.0% , fluka ) , heptanoic acid ( @xmath5697% , sigma ) , octanoic acid ( @xmath5698% , aldrich ) , and nonanoic acid ( @xmath5695% , fluka ) .
* stm experiments .
* stm images were acquired with an agilent technologies 4500 picoplus stm using a picoscan controller and stm tips formed from mechanically cut ptir ( 80:20 ) wire .
saturated solutions of the desired adsorbate ( dpbdtc or tptc ) were prepared by placing an excess of solid in the desired solvent ( heptanoic , octanoic , or nonanoic acid ) . for experiment vi ,
the solution used combined equal volumes of a saturated solution of tptc in nonanoic acid and coronene in nonanoic acid ( conc .
@xmath57 mg / ml ) .
all solutions were ultrasonically agitated to ensure complete dissolution . to form the supramolecular networks , a 10 @xmath58l droplet of solution
was deposited onto a mechanically cleaved hopg substrate .
imaging at the alkanoic - acid - hopg interface commenced immediately after approach of the stm tip .
for experiment iv , the substrate was heated to 60 @xmath39c prior to solution deposition and held at this temperature for the duration of the experiment .
figure 1 stm imaging parameters ( @xmath59/@xmath60 ) : 1c , + 1.1 v/10 pa ; 1d , + 1.2 v/15 pa ; 1e , + 1.2 v/5 pa ; 1f , + 1.25 v/12 pa ; 1 g , + 1.0 v/13.5 pa ; 1h , + 1.0 v/30 pa . * numerical simulations . *
the rhombus tilings were simulated using the same model as in ref .
supplemented by interactions which varied depending on whether neighbouring tiles were parallel ( binding energy @xmath61 ) or non - parallel ( binding energy @xmath8 ) .
simulations were performed with continuous time monte carlo which allows to access very long times at low temperatures ( i.e. very low number of defects ) on systems sizes between @xmath62 to @xmath63 and averaged over @xmath64 samples for each state point .
the location of the continuous transition from the random tiled to the ordered non - parallel phase at @xmath65 ( the location at @xmath66 is known exactly@xcite ) was determined from finite size scaling analysis of susceptibilities and higher order cumulants of @xmath25 and of the ` columnar ' order parameter ( difference in density of non - parallel order in each sublattice ) as in ref . ;
the location of the first - order transition to the ordered parallel phase from the jump in @xmath25 and the ` tilt ' order parameter ( difference in density of each kind of tile ) , cf . ref . .
we are grateful to the uk engineering and physical sciences research council ( epsrc ) for financial support under grant ep / d048761/01 .
jpg was supported by epsrc grant no .
gr / s54074/01 . as was supported by the leverhulme trust [ ecf/2010/0380 ] and the epsrc [ ep / p502632/1 ] .
ms acknowledges receipt of a royal society wolfson merit award and an erc advanced grant .
nrc gratefully acknowledges receipt of a royal society leverhulme trust senior research fellowship .
30 elemans , j. a. a. w. , lei , s. & de feyter , s. molecular and supramolecular networks on surfaces : from two - dimensional crystal engineering to reactivity .
ed . _ * 48 , * 72987332 ( 2009 ) .
bartels , l. tailoring molecular layers at metal surfaces .
_ nature chem . _
* 2 , * 8795 ( 2010 ) .
blunt , m. o. _ et al .
_ random tiling and topological defects in a two - dimensional molecular network
. _ science _ * 322 , * 10771081 ( 2008 ) .
otero , r. _ et al .
_ elementary structural motifs in a random network of cytosine adsorbed on a gold(111 ) surface
. _ science _ * 319 , * 312315 ( 2008 ) .
marschall , m. _ et al .
_ random two - dimensional string networks based on divergent coordination assembly . _ nature chem .
_ * 2 , * 131137 ( 2010 ) .
zhou , h. _ et al . _ frustrated 2d molecular crystallization .
* 129 , * 1377413775 ( 2007 ) .
fisher , m. e. statistical mechanics of dimers on a plane lattice .
rev . _ * 124 , * 16641672 ( 1961 ) .
kasteleyn , p. w. dimer statistics and phase transitions . _ j. math .
phys . _ * 4 , * 287297 ( 1963 ) .
henley , c. l. _ quasicrystals , the state of the art _ ch .
di vincenzo d. p. & steinhardt p. j. ( world scientific , singapore , 1999 ) .
destainville , n. flip dynamics in octagonal rhombus tiling sets .
lett . _ * 88 , * 030601 ( 2002 ) .
wilson , d. b. mixing times of lozenge tiling and card shuffling markov chains .
probab . _ * 14 , * 274325 ( 2004 ) .
lu p. j. & steinhardt p. j. decagonal and quasi - crystalline tilings in medieval islamic architecture .
_ science _ * 315 , * 11061110 ( 2007 ) .
alet , f. , ikhlef , y. , jacobsen , j. l. , misguich , g. & pasquier v. classical dimers with aligning interactions on the square lattice .
e _ * 74 , * 041124 ( 2006 ) .
papanikolaou , s. , luijten , e. & fradkin , e. quantum criticality , lines of fixed points , and phase separation in doped two - dimensional quantum dimer models .
b _ * 76 , * 134514 ( 2007 ) .
castelnovo , c. , chamon , c. , mudry , c. & pujol , p. zero - temperature kosterlitz - thouless transition in a two - dimensional quantum system .
phys . _ * 322 , * 903934 ( 2007 ) .
jacobsen , j. l. & alet , f. semiflexible fully packed loop model and interacting rhombus tilings .
lett . _ * 102 , * 145702 ( 2009 ) .
lackinger , m. & heckl , w. m. carboxylic acids : versatile building blocks and mediators for two - dimensional supramolecular self - assembly .
_ langmuir _ * 25 , * 1130711321 ( 2009 ) .
zhao , j. f. _ et al . _ molecule length directed self - assembly behaviour of tetratopic oligomeric phenylene - ethynylenes end - capped with carboxylic groups by scanning tunneling microscopy . _
c _ * 114 , * 99319937 ( 2010 ) .
blunt , m. _ et al .
_ directing two - dimensional molecular crystallization using guest templates .
_ 23042306 ( 2008 ) .
this value is calculated in the limit of zero temperataure and the transition is of the kosterlitz - thouless kind@xcite . for non - zero temperature , of relevance to our experiments and simulations ,
the transition is second order and occurs at slightly larger values of @xmath43 .
kampschulte , l. _ et al . _ solvent induced polymorphism in supramolecular 1,3,5-benzenetribenzoic acid monolayers .
_ j. phys .
c _ * 110 , * 1082910836 ( 2006 ) .
tahara , k. _ et al .
_ two - dimensional porous molecular networks of dehydrobenzo[12]annulene derivatives via alkyl chain interdigitation .
soc . _ * 128 , * 1661316625 ( 2006 ) .
furukawa , s. _ et al .
_ structural transformation of a two - dimensional molecular network in response to selective guest inclusion .
_ * 46 , * 28312834 ( 2007 ) .
griessl , s. j. h. _ et al .
_ incorporation and manipulation of coronene in an organic template structure .
_ langmuir _ * 20 , * 94039407 ( 2004 ) .
wu , d. , deng , k. , he , m. , zeng , q. & wang , c. coadsorption - induced reconstruction of supramolecular assembly characteristics .
_ chemphyschem _ * 8 , * 15191523 ( 2007 ) .
garrahan , j. p. , stannard , a. , blunt , m. o. & beton , p. h. molecular random tilings as glasses .
* 106 , * 1520915213 ( 2009 ) .
stannard , a. , blunt , m. o. , beton , p. b. & garrahan , j. p. entropically stabilized growth of a two - dimensional random tiling .
e _ * 82 , * 041109 ( 2010 ) .
gutzler , r. , sirtl , t. , dienstmaier , j.f . ,
mahata , k. , heckl , w. , schmittel , m. and lackinger , m. reversible phase transitions in self - assembled monoayers at the liquid - solid interface : temperature - controlled opening and closing of nanopores .
soc . _ * 132 , * 50845090 ( 2010 ) .
blunt , m.o .
_ guest - induced growth of a surface - based supramolecular bilayer .
chem . _ * 3 * , 7478 ( 2011 ) .
cook , j.l . , hunter , c.a . ,
low , c.m.r .
, perez - velasco , a. , vinter , j.g .
solvent effects on hydrogen bonding .
chem . int .
ed . _ * 46 * , 37063709 ( 2007 ) . | the degree of randomness , or partial order , present in two - dimensional supramolecular arrays of isophthalate tetracarboxylic acids is shown to vary due to subtle chemical changes such as the choice of solvent or small differences in molecular dimensions .
this variation may be quantified using an order parameter and reveals a novel phase behaviour including random tiling with varying critical properties as well as ordered phases dominated by either parallel or non - parallel alignment of neighbouring molecules , consistent with long - standing theoretical studies . the balance between order and randomness
is driven by small differences in the intermolecular interaction energies , which we show , using numerical simulations , can be related to the measured order parameter .
significant variations occur even when the energy difference is much less than the thermal energy highlighting the delicate balance between entropic and energetic effects in complex self - assembly processes .
two - dimensional molecular networks provide an attractive and highly flexible route to the formation of surfaces with specific structural arrangements and functionalities @xcite . within the broad range of structures that have been explored ,
random molecular networks have gained considerable attention recently as exemplars of two - dimensional , nonperiodic glassy systems @xcite .
although these systems are of great interest , and represent examples of complex self - assembled sytems , a quantitative measure of the degree of randomness present in a molecular array , which is required for a theoretical understanding of their properties , is not generally available . in this paper
we report a set of supramolecular arrays in which the randomness varies due to small changes in chemical environment or molecular geometry .
we define an order parameter which quantifies these variations and provides a systematic comparison between supramolecular arrays prepared using similar methodologies .
furthermore , we show that the observed variations may be rationalised within a complex phase diagram which can be related to the adsorption of dimers on a periodic lattice@xcite , a paradigmatical problem in statistical mechanics which has been studied for several decades , and is a well - known example of a physical system which can be mapped onto a non - periodic tiling of the plane@xcite . in particular , using numerical simulations , we relate the order parameter to an orientational dependence of the intermolecular interactions which stabilise the network .
we focus on molecular rhombus tilings formed at the liquid - solid interface following the deposition of solvated molecules of interest on highly - oriented pyrolytic graphite ( hopg ) substrates . in recent work
we have shown@xcite that such arrays support the formation of a random tiling , the first experimental realisation of the dimer adsorption problem .
the molecular networks formed on graphite are imaged using scanning tunnelling microscopy ( stm ) and the observed structures can be directly mapped onto rhombus tilings . in - plane
stabilisation of the molecular networks arises from hydrogen bonds formed between carboxylic acid groups @xcite .
for the tetracarboxylic acids considered here this results in two orientations of intermolecular bonding with the molecular backbone axes either parallel or non - parallel at @xmath0 , as shown in fig .
1a .
to form a random tiling a close match between the two dimensions marked @xmath1 and @xmath2 in fig .
1a is required .
examples of suitable molecules are _ p_-terphenyl-3,5,3``,5''-tetracarboxylic acid ( tptc ) ( fig .
1j ) , with @xmath3 which we have studied previously@xcite , and 1,4-diphenyl-1,3-butadiyne-3,3``,5,5''-tetracarboxylic acid ( dpbdtc ) ( fig .
1i ) , with @xmath4 synthesised specifically for this study ;
@xmath5 for both molecules .
several closely related molecules have been studied previously @xcite , but do not display an extended random tiling , although frustrated crystallisation has been observed@xcite . to tile the plane perfectly a rhombus must have internal angles of @xmath0 and @xmath6 , a lozenge @xcite , for which @xmath7 .
a random tiling is considered _ ideal _ when the energies of the non - parallel and parallel alignments , given by @xmath8 and @xmath9 respectively , are degenerate , @xmath10 @xcite . in this case
the tiling configuration is determined solely by maximising configurational entropy .
such tilings possess no translational order and display correlations in tile orientation which have a logarithmic dependence on spatial separation , a signature of a ` coulomb ' phase @xcite .
however , if there is an energetic asymmetry , @xmath11 , the degeneracy between the exponentially large number of tilings is broken . in this case
an additional internal energy contribution to the free energy competes with the entropic term ( rhombus tiles with asymmetric interactions are described as _ interacting _ in many theoretical models@xcite ) . for small @xmath12 ,
random tiling phases ( with properties which vary continuously with @xmath13 ) are expected , but large @xmath12 values are predicted to lead to transitions to ordered phases @xcite .
the parameter @xmath12 may be considered to be analogous to the coupling constant @xmath14 which appears in ising models of magnetism and leads to ordered phases with broken symmetry for @xmath15 , which may be either ferromagnetic or antiferromagnetic depending on the sign of @xmath14 , while entropic terms in the free energy dominate for @xmath16 in which case a paramagnetic phase occurs .
our experiments and simulations show that this rich phase behaviour may be investigated by preparing molecular arrays with differing values of @xmath13 .
figures 1c to 1h show images of molecular networks prepared under several different conditions ( see methods ) , referred to as experiments i to vi , respectively , and the corresponding rhombus tiling representations , where each molecule is represented by a rhombus coloured according to its orientation . to characterise our experimental tilings we define an order parameter , @xmath17 , where @xmath18 and @xmath19 represent the fraction of rhombus tile junctions in non - parallel and parallel orientations , respectively , and @xmath20 and @xmath21 are the equivalent values for a defect - free , ideal , random tiling and are estimated numerically ( see methods ) .
as such , @xmath22 in a fully parallel phase , @xmath23 in a fully non - parallel phase , and @xmath24 for an ideal tiling .
@xmath25 is calculated for each experiment ( i to vi ) , and the images ( figs .
1c to 1h ) are placed in order of decreasing @xmath25 , spanning @xmath26 to @xmath27 .
fig .
1b is a schematic of the expected equilibrium phase diagram@xcite .
there are two relevant thermodynamic parameters : temperature , @xmath28 , and energetic bias , @xmath13 .
ideal random tilings are observed for @xmath29 . for @xmath30 non - parallel bonding
is favoured , resulting in increasingly negative values of @xmath25 . for sufficiently large @xmath13 the system orders into a crystalline phase dominated by non - parallel bonds .
this transition occurs at @xmath31@xcite . for @xmath32 parallel bonding
is favoured , resulting in increasingly positive values of @xmath25 .
the random tiled phase extends until @xmath33 where the system undergoes a first - order transition@xcite to a crystalline phase dominated by non - parallel bonds .
we re - iterate the analogy with magnetic systems where transitions from disordered to ordered phases occur as the absolute value of the coupling constant is increased beyond some critical value .
it is clear from fig . 1 that different experimental conditions give rise to tilings with varying degrees of order .
in experiment i we investigated dpbdtc in nonanoic acid ( fig .
1c ) .
for this arrangement there are more parallel bonds than expected for an ideal random tiling and @xmath34 .
a solution of tptc in heptanoic acid ( exp .
ii , fig .
1d ) results in a molecular tiling with @xmath35 .
this is the molecular tiling closest to ideal ( i.e. @xmath36 ) of all the systems we have studied . for tptc a change of solvent @xcite from heptanoic acid to octanoic ( exp .
iii , fig . 1e , @xmath37 ) or to nonanoic ( exp .
v , fig .
1 g , @xmath38 ) acid , solvent molecules with similar chemistry but slightly larger size , leads to a progressive increase in the fraction of non - parallel bonds .
experiment iv is performed with tptc and nonanoic acid at an elevated temperature of 60 @xmath39c ( fig .
1f , @xmath40 ) . comparing this molecular tiling with the equivalent tiling prepared at 19 @xmath39c ( room temperature )
we see that the increase in temperature leads , as expected , to a value of @xmath25 closer to zero . finally , the most negative order parameter , @xmath41 , is observed when coronene ( fig .
1k ) is added to a solution of tptc and nonanoic acid ( exp .
vi , fig .
1h ) .
coronene becomes incorporated in the molecular network in a site which enhances the stabilisation of the non - parallel arrangement ( fig .
1l ) , a capture process consistent with previous studies of coronene - carboxylic acid and other systems @xcite .
these results show that small modifications in experimental parameters lead to the exploration of the rhombus tiling model phase space . to determine the placement of the experimental structures on the phase diagram in fig . 1b
, we have undertaken extensive numerical simulations of rhombus tilings .
the details of the numerical scheme are provided in the methods section and represent a generalisation of previous work@xcite to the case @xmath42 .
fig .
2a shows the numerically computed order parameter @xmath25 as a function of scaled bias @xmath43 .
vertical dashed and solid lines indicate where transitions from random to ordered phases occur .
the six experiments are marked on the numerical curve according to their measured @xmath25 values .
experiments ii to v span the random tiled region with negative @xmath25 ; experiment vi lies beyond the phase boundary in the non - parallel phase ; experiment i is close to the phase boundary with the ordered parallel phase . as a function of scaled bias @xmath44 .
the full curve is the result of numerical simulations of the rhombus tiling model ( see methods ) for @xmath45 ( at this value the concentration of defects @xmath46 is comparable to that observed experimentally ) .
the vertical lines indicate the approximate location of the continuous transition to the non - parallel ordered phase , @xmath47 ( slightly larger than that calculated@xcite for @xmath48 - see methods and footnote@xcite ) , and the first - order transition to the parallel phase , @xmath49 .
the experimental points are placed on the curve according to the measured values of @xmath25 . *
b * , concentration of defects as a function of bias .
the full curve is the numerical result .
the abscissa values for experimentally measured defect concentrations are fixed by panel * a * : the experiments show the predicted non - monotonic dependence on bias .
* c * , estimate of bias @xmath13 for the six experiments .
notice that @xmath13 for experiments iv and v , which correspond to the same system ( tptc in nonanoic acid ) at different temperatures , is the same within experimental error , as expected.,width=291 ] the numerical results in fig .
2a provide a relationship between the order parameter @xmath25 and @xmath43 which allow an estimation of the latter for each of our experiments .
furthermore , using the known temperature at which the experiments were performed we can estimate values of @xmath13 , as tabulated in fig . 2c . a comparison of experiments iv and v is of particular interest : both were performed using tptc in nonanoic acid but with molecular networks prepared at different temperatures .
accordingly we would anticipate the same value of @xmath13 and we find excellent agreement between the values inferred from simulations .
this result provides strong support for our analysis and confirmation that the molecular arrays can be understood in terms of equilibrium structures . from our experiments
it is clear that measurable variations in the tiling statistics may be identified even for variations of @xmath13 of @xmath50 mev , much less than the thermal energy at room temperature .
a second experimental measure , which can also be compared with numerical results , is the concentration of defects .
these ` half - rhombus ' vacancies are topological defects and are present in several of the images shown in fig . 1 .
for an ideal tiling their concentration , @xmath46 , has been shown to be temperature dependent @xcite ( we choose an effective temperature in our simulations to match , approximately , the maximum defect density observed in experiments ) and we show in fig .
2b that @xmath46 is also controlled by @xmath13 .
we highlight the non - monotonic dependence on bias which results from the strong suppression of isolated triangular defects in the ordered phases .
the experimentally - determined defect data is plotted against the theoretical curve using the value of @xmath51 obtained from fig .
2a .
we note that the non - monotonic behaviour and suppression of defect formation in the ordered phases is reproduced in our data thus providing further support of the validity of our approach to determine the value of @xmath13 for our experiments .
rhombus tilings are commonly analysed as a projection of the interface of a simple cubic lattice onto a ( 111 ) plane @xcite . in this approach
each vertex is assigned an effective height @xmath52 ( @xmath53 and @xmath54 are the co - ordinates of the vertex ) using a simple procedure ( outlined in fig .
3j ) .
an analysis of our images using this approach shows explicitly the difference between random and ordered phases as shown in fig . 3 .
here we show larger area images for experiments i , iv and vi , in three different representations .
the first is the tiling representation , as in fig
. 1 ( figs . 3a , 3d , and 3 g ) .
the second representation is given by the height map @xmath52 ( figs .
3b , 3e , and 3h ) : experiment i shows a clear height gradient , fig . 3b , indicative of its closeness to the phase boundary to the parallel phase ; in contrast the height maps for experiments iv and vi , figs .
3e and 3h , show averaged height gradients which are close to zero .
this is expected for both the random and non - parallel ordered phase although these can be distinguished by the greater height fluctuations present for experiment iv consistent with a random tiling phase .
the third representation illustrates the spatial variation of non - parallel order . in a perfectly non - parallel configuration the unit cell ( e.g. see schematic inset of fig .
1b ) can be centred at any of the three sub - lattices of the triangular lattice . choosing a sublattice is the symmetry breaking associated with the transition to the non - parallel phase and in general we would expect to observe domains of non - parallel order corresponding to each of these three possible sub - lattices .
again a magnetic analogue is useful ; in a ferromagnetic system domains of spin - up and spin - down magnetisation are formed whose size diverges as the system undergoes a transition from a paramagnetic to a ferromagnetic phase . for the case considered here
there are three equivalent non - parallel ordered states into which the system can crystallise , rather than the two states available for a simple magnetic system . in figs .
3c , 3f , and 3i regions of non - parallel aligned tiles are coloured cyan / yellow / magenta depending on their assignment to one of the three possible non - parallel phases . for experiment i , fig .
3c , these domains are relatively small since the parallel tile alignment predominates .
the domains are larger but finite for experiment iv , figs .
3f , in the random tiling phase .
for experiment vi , fig .
3i , one domain has become dominant and is comparable with the size of the system , indicative of a transition to the non - parallel phase .
our results show conclusively that the phase space of two - dimensional rhombus tiles , which arises from asymmetric interactions , may be explored experimentally through the structural characterisation of molecular arrays .
the changes in energy which give rise to different ordering are small compared to the thermal energy , an essential requirement for systems where entropy is significant .
the dependence of the characteristic energetic bias , @xmath13 , on solvent indicates that the effective energetic difference between parallel and non - parallel molecular pairs adsorbed on graphite can not be simply described by the interactions between two tetracarboxylic acid molecules and must include other more complex contributions .
we highlight two possible influences of the solvent molecules on the effective interaction between adsorbed molecules .
firstly the co - adsorption of solvent molecules on the graphite surface is possible@xcite , and expected , since the the network of tetracarboxylic acid molecules results in an array of nanopores which can accommodate guest molecules . as shown in a recent paper@xcite these nanopores are hexagonal and may be classified into five inequivalent types which are formed from different combinations of parallel and non - parallel bonded molecular pairs .
there is a significant variation of the capture of guest molecules in pores of different type@xcite , and we therefore also expect variations in the stabilisation energies of the co - adsorption of solvent molecules in different pore types . since the energy of co - adsorbed molecules makes a contribution to the overall stability@xcite of a given pore , this effect provides a mechanism to break the degeneracy of the different pore types .
since different pores are formed by different combinations of parallel and non - paralell pairs , variations in guest inclusion stability will result in small effective energetic differences between the parallel and non - parallel molecular orientations .
a second consideration is that the solvents provide an effective medium with varying dielectric constant , as pointed out for these particular molecules by kampschulte _
et.al._@xcite , which would give rise to a solvent dependence of any energetic imbalance mediated , for example , by van der waals interactions , or hydrogen - bonding , between neighbouring molecules ( see cook et.al.@xcite for a discussion of solvent effects on intermolecular interactions in a solution environment ) .
a complete microscopic description of energy differences on this scale are beyond the confidence levels normally associated with modern numerical approaches to molecular modelling ( e.g. density functional theory , molecular dynamics ) .
furthermore the inclusion of solvent molecules makes a detailed microscopic description and prediction of energies particularly problematic .
however , our experiments give clear limits on the tolerance of the random tiling phase to symmetry breaking - a difference in binding energies of @xmath55mev will drive the structures into an ordered phase .
our study demonstrates that prototypical self - assembled systems such as supramolecular networks can reveal a richness of exotic phases which have been studied widely by theorists and are analogous to those found in complex condensed - matter systems .
in addition our work highlights the role of entropy in the balance between order and disorder in molecular templates . |
in nature , there exist various systems that involve chemical reactions .
some systems are on a geographical scale while others are on a nanoscale , in particular , the biochemical reactions in a cell . to study the dynamics of reaction systems , we often adopt rate equations in order to observe the change in the chemical concentrations . in rate equations , we consider the concentrations to be continuous variables and the rate of each reaction as a function of the concentrations .
in fact , in macroscopic systems , there are a large number of molecules ; therefore , continuous representations are usually applicable
. when the concentration of a certain chemical is low , fluctuations in the reactions or flows can not be negligible .
they are usually treated by employing stochastic differential equations , in which the noise is used as a continuum description of the fluctuations @xcite .
the employment of stochastic differential equations has led to some important discoveries such as noise - induced transitions @xcite , noise - induced order @xcite , and stochastic resonance @xcite . in stochastic differential equations , the quantities of chemicals
are still regarded as continuous variables . at a microscopic level , however , we need to seriously consider the fact that the number of molecules is an integer ( @xmath0 , @xmath1 , @xmath2 , @xmath3 ) that changes in a discrete manner .
fluctuations originate from the discrete stochastic changes ; thus , continuum descriptions of fluctuations are not always appropriate .
biological cells appear to provide a good example for such discreteness in molecule numbers .
the size of the cells is of the order of microns , in which nanoscale `` quantum '' effects can be ignored .
however , in these cells , some chemicals act at extremely low concentrations of the order of pm or nm . assuming that the typical volume of a cell ranges from @xmath1 to @xmath4 @xmath5 , the concentration of one molecule in the cell volume corresponds to @xmath6 pm to @xmath6 nm .
it is possible that the molecule numbers of some chemicals in a cell are of the order of @xmath1 or sometimes even @xmath0 .
if such chemicals play only a minor role in a cell , we can safely ignore these chemicals to study intracellular chemical processes .
however , this is not always the case . in biological systems , chemical species with a small number of molecules may critically affect the behavior of the entire system .
as an extreme example , there exist only one or a few copies of genetic molecules such as dna , which are important to characterize the behavior of each cell .
further , some experiments show that doses of particular chemicals at concentrations of the order of pm or fm may alter cell behavior ( e.g. , @xcite ) .
biological systems also include positive - feedback mechanisms such as autocatalytic reactions , which may amplify single molecular changes to a macroscopic level .
it is only recently that the stochastic effect due to small molecule numbers in cells has been noticed both theoretically @xcite and experimentally @xcite . in this paper , we focus on the possible effects of molecular discreteness . through stochastic simulations , we showed that the discreteness can induce transitions to novel states in autocatalytic systems @xcite , which may affect macroscopic chemical concentrations @xcite . in the first part of this paper , we briefly review these studies and explain other aspects of such effects .
see also @xcite for recent advances in the present topic by analytic methods and numerical simulations and @xcite for simulation methods concerned . in some cases , the discreteness in the molecule numbers may cause switches between two or more states with distinct concentrations and dynamical behaviors .
further , even though the concentration of chemicals is sufficiently high for one state , the concentration could be low in another state , in which a chemical with a very low concentration could work as a stochastic switch . in the second part of this paper ,
we discuss how molecular discreteness leads to switch among states with distinct dynamical behaviors in an autocatalytic chemical reaction network system .
this spontaneous switching is characterized as an alteration ( i.e. , disconnection and reconnection ) of the acting reaction paths .
for @xmath7 , @xmath8 , and @xmath9 . in this case
, @xmath10 can reach @xmath0 , and the switching states appear . in the 1 - 3 rich state
, the system successively switches between the @xmath11 and @xmath12 states .
the interval of switching is considerably longer than the period of continuous vibration ( @xmath13 ) . at around @xmath14 ,
a transition occurs from the 1 - 3 rich state to the 2 - 4 rich state.,width=302 ] in the autocatalytic loop system for @xmath15 and @xmath16 with nonequivalent reaction constants . for small @xmath17 ,
the flow of molecules dominates the system .
thus , @xmath18 , which simply reflects @xmath19 ; this does not depend on how the continuum limit is imbalanced by the reactions .
( a ) @xmath20 and @xmath21 .
( b ) @xmath22 and @xmath23.,title="fig:",width=241 ] in the autocatalytic loop system for @xmath15 and @xmath16 with nonequivalent reaction constants . for small @xmath17 ,
the flow of molecules dominates the system .
thus , @xmath18 , which simply reflects @xmath19 ; this does not depend on how the continuum limit is imbalanced by the reactions .
( a ) @xmath20 and @xmath21 .
( b ) @xmath22 and @xmath23.,title="fig:",width=241 ] we have previously reported that the discrete nature of molecules may induce transitions to novel states , which are not reproduced by the continuous descriptions of the dynamics ( stochastic differential equations ) @xcite .
here , we briefly review that result by including some novel results .
we consider a simple autocatalytic network ( loop ) with four chemicals @xmath24 ( @xmath25 ) .
we assume the reactions @xmath26 ( with @xmath27 ) between these chemicals .
all the reactions are irreversible .
we assume that the reactor is a well - stirred container with a volume @xmath17 .
the set of @xmath10 , the number of @xmath24 molecules , determines the state of the system .
the container is in contact with a chemical reservoir in which the concentration of @xmath24 is fixed at @xmath28 .
the flow rate of @xmath24 between the container and the reservoir is @xmath29 , which corresponds to the probability of the flow - out of a molecule per unit time .
is the diffusion rate across the surface of the container . here ,
we choose a flow proportional to @xmath17 in order to obtain a well - defined continuum limit . ]
we can consider the continuum limit as @xmath30 . in this limit
, the change in @xmath31 , the chemical concentration of @xmath24 in the container , obeys the following rate equation : @xmath32 where @xmath33 is the rate constant of the reaction @xmath26 , and @xmath34 . in @xcite
, we considered a case with four equivalent chemical species , given as @xmath35 , @xmath36 , and @xmath37 for all @xmath38 ( @xmath39 , @xmath40 , @xmath41 ) , @xmath42 . in the continuum limit , the dynamics is represented by the rate equation , which has only one attractor : a stable fixed point @xmath43 for all @xmath38 .
around the fixed point , @xmath31 vibrates with the frequency @xmath44 .
if the number of molecules is finite but fairly large , we can estimate the dynamical behavior of the system using the langevin equation , which is obtained by adding a noise term to the rate equation .
each concentration @xmath31 fluctuates and vibrates around the fixed point .
an increase in the noise ( corresponding to a decrease in the number of molecules ) merely amplifies the fluctuation .
however , as we have shown in @xcite , when the number of molecules is small , novel states that do not exist in the continuum limit are observed .
two chemicals are dominant and the other two are mostly extinct ( @xmath45 ) .
figure [ fig : a-032 ] shows the time series of @xmath10 in such a case . at @xmath46 , @xmath47 and
@xmath48 dominate the system and @xmath49 for the most part ( the 1 - 3 rich state ) .
once the system reaches @xmath49 , all the reactions stop .
the system remains at @xmath49 for a long time as compared with the ordinary time scale of the reactions ( @xmath50 ) .
conversely , at @xmath51 , @xmath52 and @xmath53 are large and usually @xmath54 ( the 2 - 4 rich state ) . in the 1 - 3 or 2 - 4 rich states , the system alternately switches between either @xmath11 and @xmath12 or @xmath55 and @xmath56 .
we name these states the `` switching states . ''
the appearance of discreteness - induced novel states is described as a phase transition with a decrease in the system size ( or flow rate ) , where the histogram of @xmath57 exhibits a change from single - peaked to double - peaked distribution . in this example , although the state at each instant exhibits a clear transition , the average concentrations of the chemicals are not altered since the system resides equally in the 1 - 3 rich and 2 - 4 rich states over a long time span . on the other hand , we found examples in @xcite , where the long - term average concentration of the molecule species is altered by the discreteness as well .
a simple example is provided by considering the same reaction model as that depicted by eq .
( [ eqn : tr1-rate ] ) , but by considering the case where the parameters @xmath29 , @xmath28 , or @xmath33 are species dependent .
note that the rate equation ( [ eqn : tr1-rate ] ) obtained in the continuum limit does not contain the volume @xmath17 ; hence , the average concentrations should be independent of @xmath17 .
here , we seek the possibility of the change in the average concentrations depending on the decrease in the system size @xmath17 by taking advantage of the switching states .
recall that for the transitions to the switching states to occur in @xcite , it was necessary for the interval of the inflow to be greater than the time scale of the reactions . in the present case , the inflow interval of @xmath24 is @xmath58 , and the time scale of the reaction @xmath26 in order to consume @xmath24 is @xmath59 .
if the conditions of all the chemicals are equivalent , the discreteness of all the chemicals takes effect equally and the 1 - 3 and 2 - 4 rich states coordinately appear at @xmath60 .
now , since the parameters are species dependent , the effect of the discreteness may be different for each species .
for example , by assuming that @xmath61 , the inflow interval of @xmath62 is greater than that of @xmath63 .
thus , the discreteness in the inflow of @xmath62 can be significant for larger @xmath17 . in the previous paper @xcite
, we studied the case in which only the external concentrations ( chemical inflows ) @xmath28 were dependent on the species .
based on the degree of @xmath64 , the discreteness - induced transition occurs successively with the decrease in @xmath17 , and the average concentrations of the chemicals take distinct values from those of the continuum limit case .
similarly , we can study the effect of the discreteness when each of the chemical reaction rates @xmath33 is species dependant .
in fact , the dependence of @xmath65 on @xmath17 in this case is different from the previous study in which only @xmath28 was species dependent .
for example , we consider the case that @xmath66 and @xmath67 . figure [ fig : tr2-average - r ] shows the dependence of @xmath65 on @xmath17 .
recall that the concentrations should not depend on @xmath17 as long as the continuum representations hold ( eq . ( [ eqn : tr1-rate ] ) does not contain @xmath17 ) . here , in the continuum limit or in the case of large @xmath17 , @xmath68 , as shown in fig .
[ fig : tr2-average - r ] ( a ) . in contrast ,
when @xmath17 is small , @xmath18 .
if @xmath17 is very small , so that the total number of molecules is mostly @xmath0 or @xmath1 , the reactions rarely occur and the flow of chemicals dominates the system .
thus , @xmath69 .
if both the reactions and the flows are species dependent , we can simply expect the behavior to be a combination of the above mentioned cases .
even this simple system can exhibit a multi - step change in the concentrations along with the change in @xmath17 .
furthermore , the present behavior is not limited to the simple autocatalytic reaction loop .
in fact , we observe this type of change in randomly connected reaction networks . for a large reaction network with multiple time scales of reactions and flows
, the discreteness effect may bring about behaviors that are more complicated , although our discussion is generally applicable to such cases if the time scales are appropriately defined . as seen above
, the discreteness of molecules can alter the average concentrations . when the rates of inflow and/or the reaction are species dependent , the transitions between the discreteness - induced states are imbalanced .
this may drastically alter the average concentrations from those of the continuum limit case .
note that although the concentrations are altered in both cases , their dependence on @xmath17 is different .
if the system is extremely small ( @xmath70 ) , the frequency of the reaction event , in comparison to the diffusion , is low .
the reaction is limited by the inflows , and therefore , the system is dominated by diffusion .
the average concentrations @xmath65 depend on @xmath17 , but the dependence is quite different from the case with uniform reaction rates ( @xmath33 ) and imbalanced inflows ( @xmath28 ) , which were previously reported by us .
( substrate ) .
( i ) if all the chemical species exist , all the reactions may occur .
the system exhibits damped oscillations .
( ii , iii ) if the system lacks one or more chemicals , some of the reactions can not proceed .
the portion of the reaction path beyond the stalled reaction is disconnected ; consequently , the actual topology of the network may change . ,
width=483 ] , the number of molecules .
( a ) @xmath71 , ( b ) @xmath72 , and ( c ) @xmath73.,title="fig:",width=241 ] , the number of molecules . ( a )
@xmath71 , ( b ) @xmath72 , and ( c ) @xmath73.,title="fig:",width=241 ] + , the number of molecules .
( a ) @xmath71 , ( b ) @xmath72 , and ( c ) @xmath73.,title="fig:",width=340 ] and @xmath74 with different @xmath17 .
transition is observed between the @xmath75 and @xmath76 cases.,title="fig:",width=241 ] and @xmath74 with different @xmath17 .
transition is observed between the @xmath75 and @xmath76 cases.,title="fig:",width=241 ] molecular discreteness may not only affect the chemical concentrations but also the network of reactions .
as seen above , if the number of molecules required for a certain reaction is @xmath0 , the reaction can not take place at all .
if such a situation continues for a long time , when compared to the time scales of other reactions , the system behaves as if the reaction never existed , i.e. , the reaction is virtually eliminated from the network .
furthermore , the existence of even one or a few molecules could cause the resumption of the reaction and the recovery of the network .
in contrast to the continuum limit , where decay or recovery of the chemical is gradual , such changes in the network structure are discrete and therefore quick . here , we show an example in which the discreteness alters the actual network and switches the dynamical behavior . we adopt a simple model with four chemicals and five reactions among them ( see fig .
[ fig : net2 ] ) such that @xmath77 again , we assume a well - mixed reactor of volume @xmath17 in contact with a chemical reservoir , where the concentration of @xmath24 is maintained at @xmath28 ( @xmath29 is the flow rate of @xmath24 ) . in the continuum limit , the system is governed by the following rate equations : @xmath78 where @xmath31 is the concentration of the chemical @xmath24 .
this reaction network mainly comprises constant flows of chemicals ( @xmath79 and @xmath80 ) and an autocatalytic loop ( @xmath81 , @xmath82 , and @xmath83 ) . here
, we set @xmath84 ( for all @xmath38 ) , @xmath85 , @xmath86 , and @xmath87 . with these settings ,
generally , @xmath62 molecules flow into the container and serve as substrates .
they are then converted into other chemicals , following which they flow out ; this maintains the nonequilibrium condition . in the continuum limit ,
the concentrations @xmath31 vibrate and converge to the fixed point . to elucidate the behavior at a condition distant from the continuum limit
, we have investigated the dynamical behavior in such a condition by stochastic simulation .
[ fig : ts - v ] shows the time series of @xmath10 , the number of @xmath24 molecules .
when @xmath17 is large , generally , @xmath10 remains large .
this behavior is similar to the rate equation with the addition of noise .
however , when @xmath17 is small , @xmath10 may reach @xmath0 . in our model , if the system lacks a substrate or a catalyst for a certain reaction , the reaction ceases completely .
consequently , the dynamics of the system with such a small @xmath17 are qualitatively different .
we define the state of the system based on the combination of the reactions that cease .
a system has the following three distinct states ( see fig . [
fig : net2 ] ) : state i. : : @xmath88 for all @xmath38 , and all the reactions occur .
+ this state is determined by the fixed point concentrations obtained by the continuum limit , and the system converges to the fixed point , while the vibration around it is sustained when the number of molecules is finite .
: : for the majority of the time , @xmath89 , and reactions @xmath80 , @xmath81 , and @xmath83 cease .
+ the reaction loop can not proceed , while reaction @xmath79 continuously converts @xmath62 into @xmath90 .
state iii .
: : for the majority of the time , @xmath91 , and reaction @xmath79 ceases .
+ in the absence of any reactions , the @xmath62 molecules accumulate .
an @xmath63 molecule flowing in may trigger reactions @xmath80 and @xmath81 and convert @xmath62 into @xmath92 . in the continuum limit , the concentrations can not reach @xmath0 due to the constant inflows , and the system remains at state i when @xmath17 is sufficiently large , as shown in fig .
[ fig : ts - v ] ( a ) , even though the concentrations fluctuate and vibrate around the fixed point . with a small @xmath17
, however , the other states appear .
for example , the time series of @xmath10 with @xmath72 is shown in fig .
[ fig : ts - v ] ( b ) . at around @xmath93 ,
the system is in state i , and it switches to state ii at around @xmath94 .
it then alternates between states ii and iii .
the system spontaneously switches among these states .
if @xmath17 is considerably smaller , state i is rarely observed , as shown in fig .
[ fig : ts - v ] ( c ) .
the distribution of @xmath31 is shown in fig .
[ fig : xdist ] .
a transition is observed with a decrease in @xmath17 . for a large @xmath17
, the distribution shows a peak at around @xmath95 and @xmath96 , corresponding to the fixed point of the rate equation . for a small @xmath17
, the distribution of @xmath74 shows peaks at around @xmath97 and @xmath98 , corresponding to state ii ( @xmath99 , @xmath100 ) and state iii ( mostly @xmath98 ) , respectively .
as mentioned above , these states are classified based on the reactions that cease ; in other words , the states are classified based on the part of the network that actually functions .
in state i , all the reactions in the network function ; in state ii , the autocatalytic loop does not function ; and in state iii , the conversion of @xmath62 into @xmath90 ceases .
the transitions to states ii or iii can be viewed as the disconnection of some parts in the reaction network .
such transitions are possible only if @xmath10 reaches @xmath0 , and therefore , molecular discreteness is essential .
the extinction of the @xmath63 and @xmath90 molecules makes the system switch to states ii and iii , respectively .
the question that arises here is as follows : in general , which chemicals can switch states in a network ? in our model , molecule @xmath62 can not serve as a switch , even though @xmath47 sometimes reaches @xmath0 in the case of @xmath72 .
first , for a molecule species to function as a key for switching , @xmath10 should be maintained at @xmath0 for a longer time than that for other reactions .
for @xmath62 , there is considerable inflow , and the inflow rate is not affected if @xmath47 reaches @xmath0 .
thus , @xmath47 can not remain at @xmath0 for a long time , and @xmath62 can not switch the dynamics .
second , a key chemical for a switch should be located within the reaction paths and the extinction of the molecule disconnects some reaction paths .
stochasticity in gene expression is widely studied with regard to the problem of a small number of molecules in a biological system .
it is often assumed that two states on and off are switched by a single regulatory site . the controlling chemicals and controlled
chemicals can be clearly separated .
in contrast , our result shows that chemical species , which are usually abundant , may sometimes work as a stochastic switch . in this sense , molecules that are common or ions such as ca@xmath101 ( see @xcite ) may cause stochastic effects .
the role of a chemical may change with time .
in this study , we have demonstrated that molecular discreteness may induce transitions to novel states in autocatalytic reaction systems , which may result in an alteration of macroscopic properties such as the average chemical concentrations . in biochemical pathways ,
it is not uncommon to find that the number of molecules of a chemical is of the order of @xmath102 or less in a cell .
there are thousands of protein species , and the total number of protein molecules in a cell is not very large .
for example , in signal transduction pathways , some chemicals work at concentrations of less than @xmath103 molecules per cell .
there exist only one or a few copies of genetic molecules such as dna ; furthermore , mrnas are not present in large numbers .
thus , regulation mechanisms involving genes are quite stochastic .
naturally , molecular discreteness involves such rare chemicals . in the second part of this paper
, we have shown that the molecular discreteness may change the dynamical behavior of reaction networks .
the reaction network is virtually disconnected by the extinction of certain chemicals , which is not possible in the continuum limit .
although the network studied here is a small model , similar phenomena can exist in a complex reaction network with a large number of chemicals and reactions .
we have also investigated random networks of catalytic reactions @xmath104 ( @xmath105 ) . in such systems ,
the dynamics of chemical concentrations also depend on the system size @xmath17 . in a small system , many of the chemical species become extinct ( @xmath45 ) , and the actual reaction network is disconnected into fragments , which may be occasionally reconnected by inflow or generation of a molecule .
the onset of change in the concentrations due to disconnection or reconnection is stochastic and sudden . in the continuum limit ,
in contrast , the concentrations gradually converge to a fixed point in most cases ( or to a limit cycle or other attractors ) .
the simple model in this paper can be viewed as a switching element of such a network ; however , the exact conditions that determine whether a chemical works as a stochastic switch or not should be addressed in future .
we observed the transitions in the distribution of the concentrations @xmath31 with respect to the change in the system size @xmath17 .
multiple transitions can also occur , especially if there are many chemical species for which the number of molecules is sometimes ( but not necessarily always ) small . in this paper
, we consider reactions in a well - stirred medium , where only the number of molecules is taken into account for determining the system behavior .
however , if the system is not mixed well , we need to take into account the diffusion of molecules in space . from a biological viewpoint ,
the diffusion in space is also important because the diffusion in cells is not always fast when compared with the time scales of the reactions . if the reactions are faster than the mixing , we should consider the system as a reaction - diffusion system , with discrete molecules diffusing in space .
the relation between these time scales will be important , as indicated by mikhailov and hess @xcite . with regards to these time scales
, we recently found that the spatial discreteness of molecules within the so - called kuramoto length @xcite , over which a molecule diffuses in its lifetime ( lapses before it undergoes reaction ) , can yield novel steady states that are not observed in the reaction - diffusion equations @xcite .
spatial domain structures due to molecular discreteness are also observed @xcite .
see also @xcite for relevance of the discreteness in a replicating molecule system .
the discreteness - induced effect present here does not depend on the characteristics of the reactions .
furthermore , it may be applicable to systems beyond reactions , such as ecosystems or economic systems . the inflow of chemicals in a reaction system can be seen as a model of intrusion or evolution in an ecosystem : for both systems , discrete agents ( molecules or individuals ) may become extinct . in this
regard , our result is relevant to the studies of ecosystems , e.g. , extinction dynamics with a replicator model by tokita and yasutomi @xcite ; strong dependence of the survival probability of new species in evolving networks on the population size was reported by ebeling _
the discreteness of agents or operations might also be relevant to some economic models , e.g. , artificial markets .
most mathematical methods that are applied to reaction systems can not appropriately describe the discreteness effect .
although the utility of simulations with the progress in computer technology has become convenient , it would also be important to develop a theoretical formulation applicable to discrete reaction systems .
on the other hand , in recent years , major advances have been made in the detection of a small number of molecules and the fabrication of small reactors , which raises the possibility to experimentally demonstrate the discreteness effect predicted here .
we believe that molecular discreteness has latent but actual importance with respect to biological mechanisms such as pattern formation , regulation of biochemical pathways , or evolution , which will be pursued in the future .
this research is supported by a grant - in - aid for scientific research from the ministry of education , culture , sports , science and technology of japan ( 11ce2006 , 15 - 11161 ) and research fellowships from the japan society for the promotion of science ( 15 - 11161 , abroad h17 ) .
x. wang , g. z. feuerstein , j. gu , p. g. lysko , and t. yue , `` interleukin-1@xmath106 induces expression of adhesion molecules in human vascular smooth muscle cells and enhances adhesion of leukocytes to smooth muscle cells '' , atherosclerosis * 115 * , 89 ( 1995 ) .
w. ebeling , r. feistel , i. hartmann - sonntag , l. schimansky - geier , and a. scharnhorst , `` new species in evolving networks
stochastic theory of sensitive networks and applications on the metaphorical level '' , biosystems * 85 * , 65 ( 2006 ) . | to study the fluctuations and dynamics in chemical reaction processes , stochastic differential equations based on the rate equation involving chemical concentrations are often adopted .
when the number of molecules is very small , however , the discreteness in the number of molecules can not be neglected since the number of molecules must be an integer .
this discreteness can be important in biochemical reactions , where the total number of molecules is not significantly larger than the number of chemical species . to elucidate the effects of such discreteness , we study autocatalytic reaction systems comprising several chemical species through stochastic particle simulations . the generation of novel states is observed ; it is caused by the extinction of some molecular species due to the discreteness in their number .
we demonstrate that the reaction dynamics are switched by a single molecule , which leads to the reconstruction of the acting network structure .
we also show the strong dependence of the chemical concentrations on the system size , which is caused by transitions to discreteness - induced novel states . |
theoretical work on gravitational lensing is traditionally done in a quasi - newtonian approximation formalism , see , e.g. , schneider , ehlers and falco @xcite or petters , levine and wambsganss @xcite , which is based , among other things , on the approximative assumptions that the gravitational field is weak and that the bending angles are small . under these assumptions ,
lensing is described in terms of a `` lens equation '' that determines a `` lens map '' from a `` deflector plane '' to a `` source plane '' , thereby relating image positions on the observer s sky to source positions .
although for all practical purposes up to now this formalism has proven to be very successful , there are two motivations for doing gravitational lens theory beyond the quasi - newtonian approximation .
first , from a methodological point of view it is desirable to investigate qualitative features of lensing , such as criteria for multiple imaging or for the formation of einstein rings , in a formalism without approximations , as far as possible , to be sure that these features are not just reflections of the approximations .
second , lensing phenomena where strong gravitational fields and large bending angles are involved are no longer as far away from observability as they have been a few years ago . in particular , the discovery that there is a black hole at the center of our galaxy @xcite , and probably at the center of most galaxies , has brought the matter of lensing in strong gravitational fields with large bending angles closer to practical astrophysical interest .
if a light ray comes sufficiently close to a black hole , the bending angle is not small ; in principle , it may even become arbitrarily large , corresponding to the light ray making arbitrarily many turns around the black hole . unboundedly large bending angles also occur e.g. with wormholes ; the latter are more exotic than black holes , in the sense that up to now there is no clear evidence for their existence , but nonetheless considered as hypothetical candidates for lensing by many authors .
if one wants to drop the assumptions of weak fields and small angles , gravitational lensing has to be based on the lightlike geodesic equation in a general - relativistic spacetime , without approximations . in this paper
we will discuss this issue for the special case of a spherically symmetric and static spacetime . in view of applications , this includes spherical non - rotating stars and black holes , and also more exotic objects such as wormholes and monopoles with the desired symmetries .
the main goal of this paper is to demonstrate that in this case lensing without approximations can be studied , quite conveniently , in terms of a lens equation that is not less explicit than the lens equation of the quasi - newtonian formalism .
lensing without weak - field or small - angle approximations was pioneered by darwin @xcite and by atkinson @xcite .
whereas darwin s work is restricted to the schwarzschild spacetime throughout , atkinson derives all relevant formulas for an unspecified spherically symmetric and static spacetime before specializing to the schwarzschild spacetime in schwarzschild and in isotropic coordinates .
all important features of schwarzschild lensing are clearly explained in both papers . in particular , they discuss the occurrence of infinitely many images , corresponding to light rays making arbitrarily many turns around the center and coming closer and closer to the light sphere at @xmath2 .
however , they do not derive anything like a lens equation .
the notion of a lens equation without weak - field or small - angle approximations was brought forward much later by frittelli and newman @xcite .
it is based on the idea of parametrizing the light cone of an arbitrary observation event in a particular way . for a general discussion of this idea and of the resulting `` exact gravitational lens map '' in arbitrary spacetimes the reader may consult ehlers , frittelli and newman @xcite or perlick @xcite . here
we are interested only in the special case of a spherically symmetric and static spacetime .
then the geodesic equation is completely integrable and the exact lens equation of frittelli and newman can be written quite explicitly .
one can evaluate this equation from the spacetime perspective , as has been demonstrated by frittelli , kling and newman @xcite for the case of the schwarzschild spacetime , thereby getting a good idea of the geometry of the light cone . here
we will use an alternative representation , using the symmetry for reducing the dimension of the problem . after fixing two radius values @xmath0 and @xmath1 , lensing for an observation event somewhere at @xmath0 and static light sources distributed at @xmath1
is coded in a lens equation , explicitly given in terms of integrals over the metric coefficients , that relates two angles to each other .
this representation results in a particularly convenient method of visualizing all relevant lensing properties , as will be demonstrated with two examples .
the lens equation discussed in this paper should be compared with the lens equation for spherically symmetric and static spacetimes that was introduced by virbhadra , narasimha and chitre @xcite and then , in a modified form , by virbhadra and ellis @xcite .
the virbhadra - ellis lens equation has found considerable interest .
it was applied to the schwarzschild spacetime @xcite and later also to other spherically symmetric and static spacetimes , e.g. to a boson star by dabrowski and schunck @xcite , to a fermion star by bili , nikoli and viollier @xcite , to spacetimes with naked singularities by virbhadra and ellis @xcite , to the reissner - nordstrm spacetime by eiroa , romero and torres @xcite and to a gibbons - maeda - garfinkle - horowitz - strominger black hole by bhadra @xcite . in the last two papers , the authors concentrate on light rays that make several turns around the center and they use analytical methods developed by bozza @xcite .
the virbhadra - ellis lens equation takes an intermediary position between the exact lens equation and the quasi - newtonian approximation .
it makes no assumptions as to the smallness of bending angles , but it does make approximative assumptions as to the position of light sources and observer .
for the virbhadra - ellis lens equation to be valid the spacetime must be asymptotically flat for @xmath3 and both observer and light sources must be at positions where @xmath4 is large ; moreover , one has to restrict to light sources close to the radial line opposite to the observer position , i.e. , to the case that there is only a small misalignment .
( the question of how one can free oneself from the latter assumption was addressed by dabrowski and schunck @xcite and by bozza @xcite . )
the lens equation to be discussed in the present paper is not restricted to the asymptotically flat case , and it makes no restriction on the position of light sources or observer .
we consider an arbitrary spherically symmetric and static spacetime . for our purpose
it will be advantageous to write the metric in the form @xmath5 here @xmath6 and @xmath7 are the standard coordinates on the sphere , @xmath8 ranges over @xmath9 and @xmath4 ranges over an open interval @xmath10\ , r_{\mathrm{min } } \ , , \ , r_{\mathrm{max } } \ , [ \;$ ] where @xmath11 .
we assume that the functions @xmath12 , @xmath13 , and @xmath14 are strictly positive and ( at least piecewise ) differentiable on the interval @xmath10\ , r_{\mathrm{min } } \ , , \ , r_{\mathrm{max } } \ , [ \;$ ] . as the lightlike geodesics are not affected by the conformal factor @xmath15 ( apart from their parametrizations )
, the lens equation will depend on the metric coefficients @xmath16 and @xmath17 only .
we will see below that many qualitative features of the lens equation are determined by the coefficient @xmath17 alone .
for introducing our lens equation we have to fix two radius values @xmath0 and @xmath1 between @xmath18 and @xmath19 .
the index @xmath20 stands for `` observer '' , the index @xmath13 stands for `` source '' .
we think of an observer at @xmath21 , @xmath22 , @xmath23 .
it is our goal to determine the appearance , on the observer s sky , of static light sources distributed on the sphere @xmath24 . by symmetry
, we may restrict to the plane @xmath23 .
we consider past - oriented lightlike geodesics that start at time @xmath25 at the observer and terminate , at some time @xmath26 which depends on the geodesic , somewhere on the sphere @xmath27 .
to each of those light rays we assign the angle @xmath28 , measured at the observer between the ray s tangent and the direction of @xmath29 , and the angle @xmath30 , swept out by the azimuth coordinate along the ray on its way from the observer to the source , see figure [ fig : lensmap ] .
the desired lens equation is an equation of the form @xmath31 which relates image positions on the observer s sky , given by @xmath28 , to source positions in the spacetime , given by @xmath30 modulo @xmath32 .
we restrict @xmath28 to values between @xmath33 and @xmath34 ; then @xmath35 can be viewed as a colatitude coordinate on the observer s celestial sphere . by symmetry
, @xmath36 must be equivalent to @xmath37 .
for a given angle @xmath28 , neither existence nor uniqueness of an angle @xmath30 with @xmath31 is guaranteed .
existence fails if the respective light ray never meets the sphere @xmath27 ; uniqueness fails if it meets this sphere several times . in the latter case
the observer sees two or more images of light sources at @xmath1 at the same point on the sky , one behind the other .
we will refer to images which are covered by other images as to `` hidden images '' .
the lens equation can be solved for @xmath30 , thereby giving a _ lens map _
@xmath38 , only if hidden images do not exist ( or are willfully ignored ) .
to work out the lens equation we have to calculate the lightlike geodesics in the plane @xmath23 of the metric ( [ eq : g ] ) , which is an elementary exercise . as a conformal factor
has no influence on the lightlike geodesics ( apart from their parametrization ) , they are solutions of the euler - lagrange equations of the lagrangian @xmath39 , i.e. @xmath40 @xmath41 @xmath42 where an overdot denotes differentiation with respect to the curve parameter @xmath43 . as an aside
we mention that , by ( [ eq : elr ] ) , a circular light ray exists at radius @xmath44 if and only if @xmath45 . comparing this condition with the equivalent but less convenient eq .
( 33 ) in atkinson s article @xcite shows that it is advantageous to write the metric in the form ( [ eq : g ] ) .
the relevance of circular light rays in view of lensing was discussed by hasse and perlick @xcite , also see claudel , virbhadra and ellis @xcite for related results . to get the past - oriented light ray that starts at time @xmath25 at the observer in the direction determined by the angle @xmath28 we have to impose the initial conditions @xmath46 @xmath47 @xmath48 for each @xmath28 , the initial value problem ( [ eq : elr ] ) , ( [ eq : elphi ] ) , ( [ eq : elt]),([eq : inr ] ) , ( [ eq : inphi ] ) , ( [ eq : int ] ) has a unique maximal solution @xmath49 where @xmath43 ranges from 0 up to some @xmath50 .
every image on the oberver s sky of a light source at @xmath1 corresponds to a pair @xmath51 such that @xmath52 with some parameter value @xmath53 \ , 0 , s_{\mathrm{max } } ( \theta ) \ , [ \:$ ] . in other words , we get the desired lens equation @xmath31 if we eliminate @xmath54 from the two equations ( [ eq : el ] ) . we get an explicit expression for the lens equation , and for the travel time @xmath54 , by writing the functions @xmath55 and @xmath56 in terms of integrals . from the constant of motion @xmath57
we find , with the help of ( [ eq : elphi ] ) , ( [ eq : elt]),([eq : inphi ] ) , ( [ eq : int ] ) , @xmath58 if @xmath59 does not change sign , integration of ( [ eq : rs ] ) yields @xmath60 with @xmath61 known , @xmath62 is determined by integrating ( [ eq : elphi ] ) with ( [ eq : inphi ] ) , @xmath63 ( [ eq : intphis ] ) can be rewritten as an integral over @xmath4 , with @xmath64 substituted from ( [ eq : rs ] ) .
this gives us the lens equation in the form @xmath65 if @xmath59 changes sign , ( [ eq : intrs ] ) has to be replaced by a piecewise integration .
similarly , the substitution from the @xmath43-integration in ( [ eq : intphis ] ) to an @xmath4-integration must be done piecewise . in this case , the lens equation is not of the form ( [ eq : phi ] ) ; in particular , it is not guaranteed that the lens equation can be solved for @xmath30 . in any case , we get exact integral expressions for the lens equation , and for the travel time @xmath54 , from which all relevant lensing features can be determined in a way that is not less explicit than the quasi - newtonian approximation formalism .
this will be demonstrated by two examples in section [ sec : examples ] . in subsection
[ subsec : mono ] we treat a particularly simple example where the metric coefficients @xmath17 and @xmath16 are analytic and the integral ( [ eq : intrs ] ) can be explicitly calculated in terms of elementary functions . in this case
it suffices to calculate ( [ eq : intrs ] ) for arbitrarily small @xmath43 to get the whole function @xmath66 by maximal analytic extension ; i.e. , in this case it is not necessary to determine the points where @xmath59 changes sign and to perform a piecewise integration .
in the first part of this section we want to discuss for which values of @xmath28 the lens equation @xmath67 admits a solution . in other words , we want to determine which part of the observer s sky is covered by the light sources distributed at @xmath24 .
we restrict to the case @xmath68 .
( the results for the case @xmath69 follow immediately from our discussion ; we just have to make a coordinate transformation @xmath70 and , correspondingly , to change @xmath28 into @xmath71 .
the case @xmath72 can be treated by a limit procedure . ) for a light ray with one end - point at @xmath0 and the other at @xmath1 the right - hand side of ( [ eq : rs ] ) must be non - negative for all @xmath4 between @xmath0 and @xmath1 . this condition restricts the possible values of @xmath28 by @xmath73 where @xmath74 note that our assumptions guarantee that this infimum is strictly positiv , @xmath75 .
furthermore , a light ray with @xmath76 can arrive at @xmath1 only if it passes through a minimal radius value @xmath77 .
as ( [ eq : rs ] ) requires @xmath78 , this can be true only if @xmath79 where @xmath80 @xmath81 .
so in general the light sources at @xmath1 cover on the observer s sky a disk of angular radius @xmath82 around the pole @xmath83 and , if @xmath84 , in addition a ring of angular width @xmath85 around the pole @xmath86 , see figure [ fig : sky ] .
the two domains join if @xmath87 .
we see that the allowed values of @xmath28 are determined by the metric coefficient @xmath14 alone .
we will now demonstrate that @xmath14 alone also determines the occurrence or non - occurrence of hidden images .
hidden images occur if a light ray from @xmath0 intersects the sphere @xmath24 at least two times ; between these two intersections it must pass through a maximal radius @xmath88 which , by ( [ eq : rs ] ) , has to satisfy @xmath89 .
such a radius @xmath90 exists for all @xmath28 with @xmath91 where @xmath92 as @xmath28 is restricted by @xmath73 , hidden images can not occur if @xmath93 .
the latter condition is satisfied in asymptotically flat spacetimes , where @xmath94 for @xmath95 , if we choose @xmath1 sufficiently large .
this is the reason why in the more special situation of the virbhadra - ellis lens equation @xcite hidden images can not occur . in the rest of this section
we discuss the question of multiple imaging and the occurrence of einstein rings . for a light source at @xmath24 , @xmath96 , @xmath23 with @xmath97 ,
images on the observer s sky are in one - to - one correspondence with solutions @xmath28 of the equation @xmath98 with @xmath99 .
we call the integer @xmath100 the `` winding number '' of the corresponding light ray .
an image with @xmath101 is called `` primary '' and an image with @xmath102 is called `` secondary ''
. images with other values of @xmath100 correspond to light rays that make at least one full turn and have been termed `` relativistic '' by virbhadra and ellis @xcite .
note that different images of a light source may have the same winding number .
if we send @xmath103 to 0 or to @xmath104 , solutions @xmath28 of equation ( [ eq : mult ] ) with @xmath105 come in pairs @xmath106 . by
spherical symmetry , every such pair gives rise to an einstein ring .
there are as many einstein rings as the equation @xmath107 admits solutions with positive integers @xmath108 .
even integers @xmath108 correspond to einstein rings of the source at @xmath109 , and odd integers @xmath108 correspond to einstein rings of the source at @xmath110 .
to each solution @xmath51 of the lens equation we can assign redshift , travel time , apparent brightness and image distortion .
the general redshift formula for static metrics ( see , e.g. , straumann @xcite , p. 97 ) specified to metrics of the form ( [ eq : g ] ) says that the redshift @xmath111 is given by @xmath112 if the observer s worldline is a @xmath8-line at @xmath21 and the source s worldline is a @xmath8-line at @xmath27 . in our situation @xmath0 and @xmath1
are fixed , so the redshift is a constant .
recall that @xmath51 is a solution of the lens equation if and only if there is a parameter @xmath54 such that the equations ( [ eq : el ] ) hold .
this assigns a travel time @xmath54 to each solution @xmath51 of the lens equation .
if there are no hidden images , the equation @xmath113 gives @xmath54 as a single - valued function of @xmath28 .
quite generally , determination of the angular diameter distance requires solving the sachs equations for the optical scalars along lightlike geodesics , see e.g. schneider , ehlers and falco @xcite . for the schwarzschild metric
, this has been explicitly worked out by dwivedi and kantowski @xcite .
their method easily carries over to arbitrary spherically symmetric and static spacetimes as was demonstrated by dyer @xcite . in what follows
we give a reformulation of these results in terms of our lens equation . to that end
we fix a solution @xmath114 of the lens equation and thereby a ( past - oriented ) light ray from the observer at @xmath0 to a light source at @xmath1 . around this ray ,
we consider an infinitesimally thin bundle of neighboring rays , with vertex at the observer .
the angular diameter distance is defined as the square - root of the ratio between the cross - sectional area of this bundle at the light source and the opening solid angle at the observer .
owing to the symmetry of our situation there are two preferred spatial directions perpendicular to the ray : a radial direction ( along a meridian on the observer s sky ) and a tangential direction ( along a circle of equal latitude on the observer s sky ) . therefore , the angular diameter distance naturally comes about as a product of a radial part and a tangential part . to calculate the radial part
, we consider the infinitesimally neighboring ray which corresponds to an infinitesimally neighboring solution @xmath115 of the lens equation , i.e. @xmath116 and @xmath117 satisfy @xmath118 we define the _ radial angular diameter distance _ as @xmath119 with @xmath120 given by figure [ fig : lensmap ] , i.e. , @xmath121 measures , in the direction perpendicular to the original ray , how far the neighboring ray is away . by ( [ eq : elphi ] ) ,
( [ eq : inphi ] ) and ( [ eq : rs ] ) , @xmath120 must satisfy @xmath122 with @xmath120 given by ( [ eq : psi ] ) and @xmath123 given by ( [ eq : dphidtheta ] ) , @xmath124 is determined by ( [ eq : dr ] ) for every solution @xmath114 of the lens equation .
note that @xmath124 is singular at those solutions of the lens equation where @xmath125 has a zero .
if the lens equation can be solved for @xmath30 , we may view @xmath124 as a ( single - valued ) function of @xmath28 . to calculate the tangential part we consider an infinitesimally neighboring light ray that results by applying a rotation around the axis @xmath22 , @xmath23 .
such rotations are generated by the killing vector field @xmath126 . at points with @xmath23
, this killing vector field takes the form @xmath127 .
hence , if we rotate by an infinitesimal angle @xmath128 , the neighboring ray intersects the sphere @xmath24 at a distance @xmath129 from the original ray . relating this distance to the angle @xmath130 between the two rays at the observer gives the _ tangential angular diameter distance
_ @xmath131 by this equation , @xmath132 is uniquely determined for each solution @xmath51 of the lens equation .
again , @xmath132 may be viewed as a function of @xmath28 if the lens equation can be solved for @xmath30 .
@xmath124 and @xmath132 together give the ( averaged ) _ angular diameter distance _ or _
area distance _
@xmath133 note that both @xmath124 and @xmath132 may be negative
. images with @xmath134 are said to have _ even parity _ and images with @xmath135 are said to have _
odd parity_. images with odd parity show the neighborhood of the light source side - inverted in comparison to images with even parity . a solution @xmath51 of the lens equation is called a _ radial critical point _ if @xmath136 and a _ tangential critical point _ if @xmath137 .
the latter condition is equivalent to @xmath138 and @xmath105 , i.e. , to the occurrence of an einstein ring .
note that ( radial and tangential ) critical points come in pairs , @xmath114 and @xmath139 .
every such pair corresponds to a circle of equal latitude on the observer s sky which may be called a ( radial or tangential ) _ critical circle _ , as in the quasi - newtonian approximation formalism , see schneider , ehlers and falco @xcite , p. 233
in the quasi - newtonian formalism one usually introduces the inverse magnification factors @xmath140 and @xmath141 as substitutes for @xmath124 and @xmath142 . in our situation , where there is no flat background metric , not even asymptotically , the magnification factors can not be defined in a reasonable way , but working with @xmath124 and @xmath142 is completely satisfactory . in arbitrary spacetimes , the angular diameter distance @xmath143 is related to the ( uncorrected ) _ luminosity distance _ @xmath144 by the universal formula @xmath145 , see , e.g. , schneider , ehlers and falco @xcite , eq .
( 3.80 ) . with the redshift @xmath111 given by ( [ eq : z ] ) and the angular diameter distance @xmath143 given by ( [ eq : d ] ) , we find @xmath146 for an isotropically radiating light source with bolometric luminosity @xmath147 , the total flux at the observer is @xmath148 , see again schneider , ehlers and falco @xcite , eq .
( 3.79 ) . hence , if we distribute standard candles at @xmath24 , their apparent brightness on the observer s sky is proportional to @xmath149 .
@xmath150 and @xmath132 immediately give the apparent distortion of images .
for the sake of illustration , we may think of small spheres , with infinitesimal diameter @xmath151 , distributed with their centers at @xmath24 . by definition of @xmath124 and @xmath132 ,
each solution @xmath152 of the lens equation corresponds to an elliptical image of such a sphere on the observer s sky , with the radial ( meridional ) diameter of the ellipse equal to @xmath153 and with the tangential ( latitudinal ) diameter of the ellipse equal to @xmath154 .
thus , we may use the _ ellipticity _ @xmath155 as a measure for image distortion .
we consider the metric @xmath156 where @xmath157 is a positive constant . for @xmath158 ,
this is just minkowski spacetime in spherical coordinates . for @xmath159 ,
there is a deficit solid angle and a singularity at @xmath160 ; the plane @xmath161 , @xmath23 has the geometry of a cone .
similarly , for @xmath162 there is a surplus solid angle and a singularity at @xmath163 . for @xmath164 ,
the metric is non - flat .
the einstein tensor has non - vanishing components @xmath165 , so the weak energy condition is satisfied ( without a cosmological constant ) if and only if @xmath166 . in that case
it was shown by barriola and vilenkin @xcite that the metric may be viewed as a model for the spacetime around a monopole resulting from breaking a global @xmath167 symmetry . to within the weak - field approximation ,
basic features of lensing by such a monopole were discussed in the original paper by barriola and vilenkin @xcite and also by durrer @xcite . in what follows
we give a detailed account in terms of our exact lens equation .
note that the virbhadra - ellis lens equation @xcite is not applicable to this case , at least not without modification , because for @xmath164 the spacetime is not asymptotically flat in the usual sense .
comparison of ( [ eq : monog ] ) with ( [ eq : g ] ) shows that the metric coefficients are given by @xmath168 on the interval @xmath169 . with these metric coefficients , the integrals ( [ eq : intrs ] ) and ( [ eq : intphis ] )
can be calculated in an elementary fashion , yielding the solution to the initial value problem in the form @xmath170 @xmath171 for @xmath172 , @xmath43 ranges from 0 to @xmath173 , so @xmath174 ranges from 0 to @xmath175 .
eliminating @xmath54 from the two equations ( [ eq : el ] ) gives the lens equation , @xmath176 which is to be considered on the domain @xmath177 we restrict to the case that the integer @xmath178 defined by @xmath179 is odd .
the lens equation is plotted for the case @xmath180 in figure [ fig : mono1 ] and for the case @xmath69 in figure [ fig : mono2 ] .
( for producing the pictures we have chosen @xmath157 such that @xmath181 . )
in either case we find that there are @xmath178 einstein rings .
for a light source at @xmath182 with @xmath183 there are @xmath178 images if @xmath184 ( non - shaded regions in figures [ fig : mono1 ] and [ fig : mono2 ] ) and @xmath185 images otherwise ( shaded regions in figures [ fig : mono1 ] and [ fig : mono2 ] ) . in the following we concentrate on the case @xmath180 .
then the lens equation can be solved for @xmath30 , giving a lens map @xmath186 on the domain @xmath187 , i.e. , in the notation of figure [ fig : sky ] we have @xmath87 and @xmath188 .
in the case @xmath158 ( flat spacetime ) , the lens map can be continuously extended into the point @xmath189 ( and by periodicity onto the full circle , @xmath190 mod @xmath32 ) . for @xmath164 the ray with @xmath189 can not pass through the singularity at @xmath163 ,
so the lens map is not to be extended beyond the open interval @xmath191-\pi,\pi \ , [ \ : \,$ ] . on this interval ,
@xmath30 increases monotonously from @xmath192 to @xmath193 , see figure [ fig : mono1 ] .
thus , the range of the lens map is independent of @xmath0 and @xmath1 .
the ratio @xmath194 influences the shape of the graph of the lens map in the following way . for @xmath195
it becomes a straight line , @xmath196 .
for @xmath197 it becomes a broken straight line , @xmath198 for @xmath199 and @xmath200 for @xmath201 .
note that linearity of the lens map implies that the angular distance on the observer s sky between consecutive images is the same for all light sources . as @xmath30 is a ( single - valued ) function of @xmath28
, so are all observables . by evaluating the formulas derived in section [ sec : obs ] for the case at hand
we find @xmath202 @xmath203 @xmath204 which gives @xmath205 and @xmath206 as functions of @xmath28 via ( [ eq : dlum ] ) and ( [ eq : epsilon ] ) .
the observables are plotted in figures [ fig : mono3 ] , [ fig : mono4 ] and [ fig : mono5 ] .
the metric @xmath207 where @xmath208 is a positive constant , is an example for a traversable wormhole of the morris - thorne class , see box 2 in morris and thorne @xcite .
it was investigated , already in the 1970s , by ellis @xcite who called it a `` drainhole '' .
lensing in the ellis spacetime was discussed , in a scattering formalism assuming that observer and light source are at infinity , by chetouani and clment @xcite . in the following
we give a detailed account of lensing in this spacetime with the help of our lens equation . by comparison of ( [ eq : wormg ] ) with ( [ eq : g ] )
we find @xmath209 where the radius coordinate @xmath4 ranges from @xmath210 to @xmath211 .
( we do _ not _ identify the region where @xmath4 is positive with the region where @xmath4 is negative . )
the function @xmath14 has a minimum at @xmath163 , thereby indicating the existence of circular light rays at the neck of the wormhole . in the following
we consider the case that observer and light sources are on different sides of the neck of the wormhole , @xmath212 .
as a first step , we determine for which angles
@xmath28 the lens equation admits a solution , recall figure [ fig : sky ] . in the case at hand , the angles @xmath82 and @xmath213 defined by ( [ eq : ci ] ) and ( [ eq : cii ] ) are given by @xmath214 and @xmath215 .
hence , the lens equation admits a solution for all angles @xmath28 with @xmath216 and @xmath217 light sources distributed at @xmath24 illuminate a disk of angular radius @xmath218 on the observer s sky .
the apparent rim of the disk corresponds to light rays that spiral asymptotically towards @xmath163 . as the constant @xmath219 defined by ( [ eq : c ] ) satisfies @xmath220 , there are no hidden images , i.e. , the lens equation can be solved for @xmath30 .
with @xmath28 restricted by ( [ eq : wormdelta ] ) we read from ( [ eq : rs ] ) that @xmath59 has no zeros along a ray that starts at @xmath221 and passes through @xmath24 .
hence , ( [ eq : intrs ] ) gives us @xmath55 and thereby the travel time in terms of an elliptic integral , @xmath222 similarly , ( [ eq : phi ] ) gives us @xmath30 as a ( single - valued ) function of @xmath28 in terms of an elliptic integral and thereby the lens equation , @xmath223 see figure [ fig : worm1 ] . as @xmath30 increases monotonously from @xmath224 to @xmath173 on the domain @xmath225- \delta _ i , \
, \delta _ i\ , [ \,$ ] , there are infinitely many einstein rings whose angular radii converge to @xmath82 .
if we fix a light source at @xmath96 with @xmath97 , figure [ fig : worm1 ] gives us infinitely many images which can be characterized in the following way .
for every @xmath99 there is a unique @xmath226 - \delta _ i , \
, \delta _ i \ , [ \:$ ] such that @xmath227 , and @xmath228 for @xmath229 . as @xmath30 is a ( single - valued ) function of @xmath28
, so are all observables . by evaluating the formulas derived in section [ sec : obs ] for the case at hand
we find @xmath230 @xmath231 @xmath232 @xmath233 which gives @xmath205 and @xmath206 as functions of @xmath28 via ( [ eq : dlum ] ) and ( [ eq : epsilon ] ) . the observables are plotted in figures [ fig : worm2 ] , [ fig : worm3 ] , [ fig : worm4 ] and [ fig : worm5 ]
. as there are infinitely many einstein rings whose angular radii converge to @xmath82 , the tangential angular diameter distance must have infinitely many zeros that converge to @xmath82 .
this is difficult to show in a picture unless one transforms @xmath28 into a new coordinate @xmath234 that goes to infinity for @xmath235 .
if one chooses a logarithmic transformation formula , as has been done in figure [ fig : worm3 ] , one sees that in terms of the new coordinate @xmath234 the einstein rings become equidistant .
this feature is not particular to the ellis wormhole ; @xmath236 as a function of @xmath28 always diverges logarithmically when a circular light ray at a radius @xmath44 with @xmath237 and @xmath238 is approached
. the proof can be taken from bozza @xcite .
one may also treat the case that observer and light sources are on the same side of the neck of the wormhole . if the observer is closer to the neck than the light sources , @xmath239 or @xmath240 , the results are quite similar to the case above .
the only difference is in the fact that the light sources appear as a disk of radius _ bigger _ than @xmath241 , i.e. , the disk covers more than one hemisphere . if the observer is farther from the neck than the light sources , @xmath242 or @xmath243 , there are hidden images .
i.e. , one does not get a single - valued lens map @xmath244 .
the qualitative features of lensing by an ellis wormhole are very similar to the qualitative features of lensing by a schwarzschild black hole .
the radii @xmath245 , @xmath246 , @xmath247 in the ellis case correspond respectively to the radii @xmath248 , @xmath249 , @xmath247 in the schwarzschild case . as a matter of fact
, we encounter these same features whenever the function @xmath14 has one minimum , @xmath45 and @xmath250 , and no other extrema on the considered interval .
the lens equation and the formulas for redshift , travel time and radial angular diameter distance used in this paper refer to lightlike geodesics of the @xmath251-dimensional metric @xmath252 , independently of whether this metric results from restricting a spherically symmetric and static spacetime to the equatorial plane .
therefore , these results apply equally well to the plane @xmath253 of a cylindrically symmetric and static spacetime and , of course , to genuinely @xmath251-dimensional spacetimes with the assumed symmetries such as the btz black hole .
( for lightlike and timelike geodesics in the metric of the btz black hole see cruz , martnez and pea @xcite . )
e.g. , the metric @xmath254 results not only by restricting the spacetime of a barriola - vilenkin monopole to the plane @xmath23 , as discussed in subsection [ subsec : mono ] , but also by restricting the cylindrically symmetric and static metric @xmath255 to the plane @xmath256 .
the latter metric is well - known to describe the spacetime around a static string , see vilenkin @xcite , gott @xcite and hiscock @xcite , and was investigated in detail already by marder @xcite .
hence , if re - interpreted appropriately , the results of subsection [ subsec : mono ] apply to light rays in the plane perpendicular to a static string . for treating _ all
_ light rays in a cylindrically symmetric and static spacetime one may introduce a modified lens equation , replacing the sphere @xmath24 with a cylinder . | lensing in a spherically symmetric and static spacetime is considered , based on the lightlike geodesic equation without approximations . after fixing two radius values @xmath0 and @xmath1 ,
lensing for an observation event somewhere at @xmath0 and static light sources distributed at @xmath1 is coded in a lens equation that is explicitly given in terms of integrals over the metric coefficients .
the lens equation relates two angle variables and can be easily plotted if the metric coefficients have been specified ; this allows to visualize in a convenient way all relevant lensing properties , giving image positions , apparent brightnesses , image distortions , etc .
two examples are treated : lensing by a barriola - vilenkin monopole and lensing by an ellis wormhole . |
the unitary extension of chiral perturbation theory , u@xmath0pt , has been successfully applied to study many meson - baryon and meson - meson interactions .
more recently , it has been used to study the lowest axial vector mesons @xmath3 , @xmath4 , @xmath5 , @xmath6 , @xmath7 , @xmath1 and @xmath8 @xcite .
both works generate most of the low - lying axial vector mesons dynamically .
however , there is a surprising discovery in ref .
@xcite , i.e. , two poles are found in the second riemann sheet in the @xmath9 and @xmath10 channel and both are attributed to the @xmath1 .
although the @xmath1 has been observed in various reactions , the most conclusive and high - statistics data of the @xmath1 come from the wa3 experiment at cern that accumulated data on the reaction @xmath11 at 63 gev .
these data were analyzed by the accmor collaboration @xcite . as will be shown in this paper , the two - peak structure , with a peak at lower energy depending drastically on the reaction channel investigated ,
can be easily explained in our model with two poles for the @xmath1 plus the @xmath8 . with only one pole , as has been noted long time ago @xcite , there is always a discrepancy for the peak positions observed in the @xmath12 and @xmath13 invariant mass distributions .
in the following , we briefly describe the chiral unitary approach , while detailed formalism can be found in refs . @xcite . in the bethe - salpeter formulation of the unitary chiral perturbation theory @xcite
, one has the following unitarized amplitude : @xmath14^{-1}(-v ) \,\vec{\epsilon}\cdot\vec{\epsilon}\ , ' , \label{bethes}\ ] ] where @xmath15 is a diagonal matrix with the @xmath16th element , @xmath17 , being the two meson loop function containing a vector and a pseudoscalar meson : @xmath18 with @xmath19 the total incident momentum , which in the center of mass frame is @xmath20 .
the loop function @xmath17 can be regularized either by a cutoff or by dimensional regularization . in the former case , one has cutoff values , whereas in the latter , one has subtraction constants as free parameters , which have to be fitted to the data . and @xmath21 channel.,title="fig : " ] and @xmath21 channel.,title="fig : " ]
the tree level amplitudes are calculated using the following interaction lagrangian @xcite : @xmath22 where @xmath23 means su(3 ) trace and @xmath24 is the covariant derivative defined as @xmath25,\ ] ] where @xmath26 $ ] stands for commutator and @xmath27 is the vector current @xmath28 with @xmath29 in the above equations @xmath30 is the pion decay constant in the chiral limit and @xmath19 and @xmath31 are the su(3 ) matrices containing the pseudoscalar octet of the pion and the vector nonet of the rho .
[ fig : gtvpvp ] shows the modulus squared of the @xmath9 , @xmath21 amplitudes multiplied by the corresponding loop functions obtained with @xmath32 mev , @xmath33 and @xmath34 mev .
the pole positions and corresponding widths obtained with this set of parameters are shown in table [ tab : poles2 ] . from fig .
[ fig : gtvpvp ] , the two poles are clearly seen : the higher pole manifests itself as one relatively narrower resonance around 1.28 gev and the lower pole as a broader resonance at @xmath35 gev .
the effective couplings for the coupled channels @xmath36 , @xmath37 , @xmath13 , @xmath38 and @xmath12 , calculated from the residues of the amplitudes at the complex pole positions , are tabulated in table [ tab : coup ] for both the lower pole and the higher pole , respectively .
.effective couplings of the two poles of the @xmath1 to the five channels : @xmath36 , @xmath37 , @xmath13 , @xmath38 and @xmath12 .
all the units are in mev.[tab : poles2 ] [ cols="^,^,^,^,^ " , ] [ tab : coup ] it is clearly seen that the lower pole couples dominantly to the @xmath12 channel while the higher pole couples more strongly to the @xmath13 channel .
if different reaction mechanisms favor one or the other channel , they will see different shapes for the resonance .
more importantly , it is to be noted that not only the two poles couple to different channels with different strengths , but also they manifest themselves in different final states . in other words , in the @xmath13 final states , one favors a narrower resonance around 1.28 gev , while in the @xmath12 final states , one would favor a broader resonance at a smaller invariant mass .
the reaction @xmath39 can be analyzed by the isobar model as @xmath40 .
therefore , one can construct the following amplitudes to simulate this process . assuming @xmath21 dominance for @xmath41 and @xmath42 as suggested by the experiment one has @xmath43 where @xmath44 are the coupled channel amplitudes obtained in section 2 and the clebsch - gordan coefficient @xmath45(@xmath46 ) accounts for projecting the @xmath21 @xmath12 ( @xmath13 ) state into @xmath41(@xmath42 ) .
the coefficients @xmath47 and @xmath48 are complex couplings . to contrast our model with data
, it is necessary to take into account the existence of the @xmath8 , which is not dynamically generated in our approach .
therefore , we add to the amplitudes in eq .
( [ eq : ts ] ) an explicit contribution of the @xmath8 @xmath49 where @xmath50 and @xmath51 are complex couplings , and @xmath52 and @xmath53 are the mass and width of the @xmath8 with the @xmath54-wave width given by @xmath55 @xmath56 and @xmath57 are calculated by @xmath58 in our model , eq .
( [ bwadd ] ) , we have the following adjustable parameters : @xmath47 , @xmath48 , @xmath59 , @xmath51 , @xmath52 and @xmath60 . in principle ,
@xmath30 and @xmath61 can also be taken as free parameters .
one can then fix these parameters by fitting the wa3 data ( see ref .
@xcite for details ) . and @xmath13 invariant mass distributions
. the data are from the wa3 reaction @xmath62 at 63 gev @xcite .
data in the upper panels are for @xmath63 gev@xmath64 , where @xmath65 is the four momentum transfer squared to the recoiling proton .
the data are further grouped by @xmath66 followed by the isobar and odd particle .
@xmath67 is the total angular momentum , @xmath19 the parity , @xmath68 the orbital angular momentum of the odd particle .
@xmath69 denotes the magnetic substate of the @xmath70 system and the naturality of the exchange.,title="fig : " ] and @xmath13 invariant mass distributions .
the data are from the wa3 reaction @xmath62 at 63 gev @xcite .
data in the upper panels are for @xmath63 gev@xmath64 , where @xmath65 is the four momentum transfer squared to the recoiling proton .
the data are further grouped by @xmath66 followed by the isobar and odd particle .
@xmath67 is the total angular momentum , @xmath19 the parity , @xmath68 the orbital angular momentum of the odd particle .
@xmath69 denotes the magnetic substate of the @xmath70 system and the naturality of the exchange.,title="fig : " ] and @xmath13 invariant mass distributions .
the data are from the wa3 reaction @xmath62 at 63 gev @xcite .
data in the upper panels are for @xmath63 gev@xmath64 , where @xmath65 is the four momentum transfer squared to the recoiling proton .
the data are further grouped by @xmath66 followed by the isobar and odd particle .
@xmath67 is the total angular momentum , @xmath19 the parity , @xmath68 the orbital angular momentum of the odd particle .
@xmath69 denotes the magnetic substate of the @xmath70 system and the naturality of the exchange.,title="fig : " ] and @xmath13 invariant mass distributions .
the data are from the wa3 reaction @xmath62 at 63 gev @xcite .
data in the upper panels are for @xmath63 gev@xmath64 , where @xmath65 is the four momentum transfer squared to the recoiling proton .
the data are further grouped by @xmath66 followed by the isobar and odd particle .
@xmath67 is the total angular momentum , @xmath19 the parity , @xmath68 the orbital angular momentum of the odd particle .
@xmath69 denotes the magnetic substate of the @xmath70 system and the naturality of the exchange.,title="fig : " ] and @xmath13 invariant mass distributions .
the data are from the wa3 reaction @xmath62 at 63 gev @xcite .
data in the upper panels are for @xmath63 gev@xmath64 , where @xmath65 is the four momentum transfer squared to the recoiling proton .
the data are further grouped by @xmath66 followed by the isobar and odd particle .
@xmath67 is the total angular momentum , @xmath19 the parity , @xmath68 the orbital angular momentum of the odd particle .
@xmath69 denotes the magnetic substate of the @xmath70 system and the naturality of the exchange.,title="fig : " ] and @xmath13 invariant mass distributions .
the data are from the wa3 reaction @xmath62 at 63 gev @xcite .
data in the upper panels are for @xmath63 gev@xmath64 , where @xmath65 is the four momentum transfer squared to the recoiling proton .
the data are further grouped by @xmath66 followed by the isobar and odd particle .
@xmath67 is the total angular momentum , @xmath19 the parity , @xmath68 the orbital angular momentum of the odd particle .
@xmath69 denotes the magnetic substate of the @xmath70 system and the naturality of the exchange.,title="fig : " ] according to ref .
@xcite , for an @xmath54-wave resonance , the theoretical differential cross section can be calculated by @xmath71 where @xmath52 is the invariant mass of the @xmath12 or @xmath13 systems , @xmath72 is a normalization constant , @xmath73 is the amplitude specified above for the @xmath12 or @xmath13 channels and @xmath74 is the center of mass three - momentum of @xmath12 or @xmath13 . we have taken @xmath72 to be 1 , or in other words , it has been absorbed into the coupling constants @xmath47 , @xmath48 , @xmath59 and @xmath51 . the theoretical invariant mass distributions calculated with eq .
( [ eq : inv ] ) are shown in fig .
[ fig : s1 ] in comparison with the wa3 data @xcite . from fig .
[ fig : s1 ] , it is clearly seen that our model can fit the data around the peaks very well . in fig .
[ fig : s1 ] , the dashed and dotted lines are the separate contributions of the @xmath1 and the @xmath8 .
one can easily see that the @xmath8 decays dominantly to @xmath12 , which is consistent with our present understanding of this resonance .
it should be mentioned that in our model the lower peak observed in the invariant mass distribution of the @xmath12 channel is due to the contribution of the two poles of the @xmath1 .
this is very different from the traditional interpretation .
for example , the lower peak observed in the @xmath12 invariant mass distributions of @xmath75 at 13 gev was interpreted as a pure gaussian background by carnegie et al .
@xcite , which has a shape similar to the contribution of the @xmath1 as shown in fig .
[ fig : s1 ] . on the other hand ,
the k - matrix approach was adopted to analyze the wa3 data @xcite and the slac data @xcite . in this latter approach ,
the lower peak mostly comes from the so - called deck background , which after unitarization , also has a shape of resonance .
as we mentioned in the introduction , even in the original wa3 paper @xcite , it was noted that their model failed to describe the @xmath76(@xmath12 ) data , in the notation @xmath77 with @xmath78 the naturality of the exchange @xcite .
the predicted peak is 20 mev higher than the data .
if the fit were done only to the @xmath12 data , the agreement was much better but then the predicted @xmath1 would be lower by 35 mev than that obtained when other channels were also considered in the fit .
it is worth stressing that the @xmath1 peak seen in the upper - left panel of fig .
[ fig : s1 ] is significantly broader than that in the upper - right panel .
furthermore the peak positions are also different in the two cases ( 1240 mev and 1280 mev respectively ) .
both features have a straightforward interpretation in our theoretical description since the first one is dominated by the low - energy ( broader ) @xmath1 state , while the second one is dominated by the higher - energy ( narrower ) @xmath1 state .
studies based on unitary chiral perturbation theory obtain two poles in the @xmath10 , @xmath9 , vector - pseudoscalar scattering amplitudes which can be assigned to two @xmath1 resonances .
one pole is at @xmath79 mev with a width of @xmath80 mev and the other is at @xmath81 mev with a width of @xmath82 mev .
the lower pole couples more to the @xmath12 channel whereas the higher pole couples dominantly to the @xmath13 channel .
different reaction mechanisms may prefer different channels and thus this explains the different invariant mass distributions seen in various experiments .
we have analyzed the wa3 data on the @xmath83 reaction since it is the most conclusive and high - statistics experiment quoted in the pdg on the @xmath1 resonance .
our model obtains a good description of the wa3 data both for the @xmath12 and @xmath13 final state channels . in our model ,
the peak in the @xmath70 mass distribution around the @xmath84 mev region is a superposition of the two poles , but in the @xmath12 channel the lower pole dominates and in the @xmath13 channel the higher pole gives the biggest contribution .
this work is partly supported by dgicyt contract no .
bfm2003 - 00856 , fpa2004 - 03470 , the generalitat valenciana , and the e.u . flavianet network contract no .
hprn - ct-2002 - 00311 .
this research is part of the eu integrated infrastructure initiative hadron physics project under contract no .
rii3-ct-2004 - 506078 .
m. c. birse , z. phys .
a * 355 * , 231 ( 1996 ) .
s. m. flatte , phys .
b * 63 * , 224 ( 1976 ) .
r. k. carnegie _
et al . _ ,
b * 127 * , 509 ( 1977 ) .
g. w. brandenburg _ et al .
_ , phys . rev
. lett . * 36 * , 703 ( 1976 ) . | recent studies based on unitary chiral perturbation theory ( u@xmath0pt ) found that the low - lying axial vector mesons can be dynamically generated due to the interaction of the pseudoscalar octet of the pion and the vector nonet of the rho .
in particular , two poles in the second riemann sheet have been associated to the nominal @xmath1 resonance . in this talk , we present a recent analysis of the wa3 data on @xmath2 at 63 gev using the u@xmath0pt amplitudes , and show that it is in favor of the existence of two @xmath1 s [ phys . rev .
d 75 , 014017 ( 2007 ) ] .
= 11.6pt |
since the early applications of the gw method to real materials ( see refs . 1 - 3 and references therein ) , the pseudopotential plane - wave approach has been the method of choice due to its accuracy and technical simplicity .
recent advances in lapw and lmto methodologies have allowed the implementation of `` all - electron '' applications of the gw method.@xcite one common feature of such calculations , using standard level of approximation for the self - energy , is an underestimation of the electronic energy gap compared to experimental measurements , whereas pseudopotential - based calculations show very good agreement with experiment . to explain this inconsistency , it was proposed that the pseudopotential approach does not correctly describe the effect of core orbitals in the self - energy corrections to the energy gaps , resulting in overestimated corrections.@xcite it is thus desirable to elucidate the effect of core orbitals in the quasiparticle band structure , and the preferred procedure is to perform a well converged all - electron calculation and compare its results with similarly converged pseudopotential - based calculations . obviously , numerical precision should not be neglected . in this work ,
we explicitly include semicore orbitals in the pseudopotential plane - wave approach and calculate the quasiparticle energy gap for three semiconductors of technical importance : si , ge and gaas .
the underlying description of the ground - state electronic structure is based on density functional theory in the local density approximation ( dft / lda).@xcite throughout this work , we are careful to converge all results systematically , and the final results are compared to previous pseudopotential results and to recent all - electron calculations .
the paper is organized as follows : we outline the theoretical method in section ii .
results are presented in section iii and discussed in section iv .
the pseudopotential formalism has two advantages that make it convenient for practical _ ab initio _ calculations .
first , degrees of freedom due to core electrons are removed from the system , resulting in a description that contains only valence electrons .
second , valence electronic pseudowavefunctions are smooth in the vicinity of atomic sites as well as in the interstitial region .
such valence wavefunctions can be expanded to convergence easily in a relatively small basis set such as a plane - wave basis . on the other hand ,
one possible problem with this formalism is that using pseudopotentials , instead of the true electron - ion potential , may not fully describe all effects produced by interactions between the valence and core electrons . in order to address this issue in gw calculations ,
we include in the present study electrons from the outermost valence shell as well as those from the second outermost ( `` semicore '' ) atomic shell as active `` valence '' electrons in the pseudopotential formalism .
only interactions between these `` valence '' electrons and electrons from the deeper core shells are described by pseudopotentials . taking si as an example ,
its `` core '' now contains only @xmath0 electrons .
electrons from the @xmath1,@xmath2,@xmath3,@xmath4 , and @xmath5 shells are all treated on equal footing in the subsequent calculations .
apart from the atomic configuration , we follow the standard prescription for generating _ ab initio _ pseudopotentials . @xcite
since the deeper core electrons have extremely large binding energy ( _ e.g. _ , 130 ry for the @xmath6 electrons in atomic si ) , their interaction with valence electrons is expected to be much weaker than the already small interaction between the outermost valence and the semicore electrons.@xcite the kohn - sham dft formalism within the lda is used to solve for the ground state electronic structure and to provide a starting point for the calculation of the electron self - energy .
we follow closely the gw method as developed by hybertsen and louie . in this method , the self - energy is given by the standard gw approximation , @xmath7 where @xmath8 is the one - electron green s function , calculated from lda energy eigenvalues and eigenstates ( @xmath9 for occupied states , @xmath10 for unoccupied states ) , @xmath11 the screened coulomb interaction @xmath12 is calculated within the random phase approximation ( rpa ) as @xmath13^{-1 } v$ ] , with @xmath14 being the bare coulomb interaction and the polarizability @xmath15.@xcite this is the commonly employed level of approximation for gw calculations ( _ i.e. _ neglecting self - consistency and vertex corrections ) @xcite , and we specifically compare results from both pseudopotential and all - electron calculations at this particular level of approximation to help untangle the effects of core states .
lccc + & reference configuration & @xmath16 & local channel + & & @xmath17 & + si & @xmath18 & 0.40 0.35 0.40 & @xmath19 + ge & @xmath20 & 0.50 0.50 0.50 & @xmath21 + ga & @xmath22 & 0.50 0.50 0.50 & @xmath21 + as & @xmath23 & 0.50 0.50 0.45 & @xmath24 + [ t.atomic ] the convolution integral in eq . ( [ sigma ] )
is performed using the generalized plasmon - pole ( gpp ) model , which enables one to distinguish two contributions to the self - energy : a screened exchange part ( @xmath25 ) arising from the poles of @xmath8 , and a dynamical coulomb interaction between an electron and the hole - like charge distribution around it ( @xmath26 ) stemming from the poles of @xmath12 .
the former tends to increase the quasiparticle energy , after the bare exchange is excluded .
the latter tends to decrease the quasiparticle energy .
in particular , the dynamical contribution @xmath26 is highly sensitive to the number of bands @xmath27 included in the calculation of the green s function in eq .
( [ green ] ) . the final quasiparticle energy of a state @xmath28 is given by ( @xmath29 is the lda exchange - correlation potential ) @xmath30
a semicore , non - relativistic pseudopotential was generated for si , using the troullier - martins scheme.@xcite for ga , ge and as , we constructed semi - relativistic pseudopotentials using the kerker scheme.@xcite these choices resulted in stable , transferable pseudopotentials in the kleinman - bylander form , without ghost states .
@xcite a summary of atomic parameters is presented in table [ t.atomic ] .
a good expansion of electronic wavefunctions in a plane - wave basis was obtained using a cutoff energy of 700 ry ( 600 ry for si ) , and the first brillouin zone was sampled using a 4x4x4 monkhorst - pack grid .
@xcite these numerical parameters ensure convergence in lda energy eigenvalues to 0.01 ev or better .
the ceperley - alder exchange - correlation potential is used.@xcite for the lattice parameter , we used the experimental values : 5.43 @xmath31 , 5.65 @xmath31 , and 5.66 @xmath31 for si , ge , and gaas , respectively .
the polarizability was expanded in a plane - wave basis with an energy cutoff of 45 ry ( 50 ry for si ) and numerically inverted for the calculation of the screened coulomb interaction @xmath12 .
numerical precision in the calculation of the self - energy is 0.05 ev or better .
lccc + & @xmath32 & @xmath33 & @xmath34 + + & 0.46 & 2.52 & 0.60 + + shirley _
et al._@xmath35 & 1.13 & 3.28 & 1.31 + + & 1.04 & 3.24 & 1.18 + + hamada _ et al._@xmath36 & 1.01 & 3.30 & 1.14 + kotani and van schilfgaarde@xmath37 & 0.89 & 3.12 & + ku and eguiluz@xmath38 & 0.85 & 3.12 & + experiment@xmath39 & 1.17 & 3.35 & 1.3 + @xmath35 reference 9 .
@xmath36 reference 5 .
@xmath37 reference 6 .
@xmath38 non - self - consistent results from reference 7 .
@xmath39 reference 23 .
[ t.si ] table [ t.si ] shows some of the energy gaps obtained in the present approach for si , compared with previous valence - only pseudopotential and all - electron calculations .
overall agreement between the present results ( which explicitly include the effect of the semicore states ) and experimental measurements is at the level of 0.1 ev , and discrepancies between our results and previous pseudopotential - based calculations of ref .
9 are equally small .
recent all - electron calculations carried out at the same level of the gw approximation , however , systematically underestimate the minimum gap and the direct @xmath40 gaps.@xcite we find that the convergence of the self - energy with respect to the number of unoccupied bands included in @xmath8 in eq .
( [ green ] ) is an important factor . in figure
[ f.si ] , we show the behavior of the calculated energy gap as function of the number of unoccupied bands , @xmath41 , included in @xmath8 .
convergence is typically very slow , and well - converged results require @xmath42 .
other energy transitions ( @xmath43 and @xmath40 ) show similar behavior and also approach the converged value from below .
in contrast , the results of ref . 7 were obtained with only @xmath44 and are closer to our results at approximately the same value of @xmath41 than to the converged results , as shown in figure [ f.si ] .
this fact points to a lack of numerical convergence in the evaluation of the self - energy corrections in ref .
lccc + & @xmath45 & @xmath33 & @xmath34 + + + & -0.04 & -0.26 & 0.56 + + shirley _ et al .
_ @xmath35 & 0.73 & 0.85 & 1.09 + + & 0.65 & 0.85 & 0.98 + gw all - electron & & & + kotani and van schilfgaarde@xmath36 & 0.47 & 0.79 + ku and eguiluz@xmath37 & 0.51 & 1.11 & 0.49 + experiment@xmath38 & 0.74 & 0.90 & 1.3 + @xmath35 reference 9 .
@xmath36 reference 6 after inclusion of spin - orbit effects .
@xmath37 non - self - consistent results from reference 7 .
@xmath38 reference 23 .
[ t.ge ] the energy gaps obtained for ge are presented in table [ t.ge ] .
spin - orbit interactions are included as first - order perturbations .
a common feature of lda - based calculations is the overlapping of the valence and conduction bands at the @xmath46 point .
the inclusion of @xmath47 electrons moves the conduction bands further down resulting in a sizable negative direct gap.@xcite this feature is verified in our lda calculation . as shown in table [ t.ge ] , self - energy corrections are responsible for an opening of the gap and the correct positioning of the minimum , indirect gap between points @xmath46 and @xmath48 in the brillouin zone .
our gw results compare well with experiment , although the @xmath49 gap still shows a large discrepancy . on the other hand , there are significant differences between the pseudopotential and all - electron gaps at this level of the gw approximation , in particular regarding the @xmath49 gap .
as found in previous studies , the self - energy corrections are needed to give the correct band topology .
figure [ f.ge ] shows direct and indirect energy gaps in ge as function of the number of unoccupied bands included in the calculation of @xmath8 . whereas the indirect gaps @xmath50 and @xmath51 approach the converged value from below , the direct gap @xmath52 approaches it from above .
this particular convergence behavior arises from the fact that we are plotting differences of quasi - particle energies : taken individually , all quasi - particle energies @xmath53 converge monotonically from above , reflecting the attractive nature of the coulomb - hole @xmath26 term .
additionaly , we note that in ref .
7 the direct gap at @xmath46 is strongly overestimated and the indirect gap @xmath49 is underestimated by @xmath54 ev . as expected , the same pattern of overestimation / underestimation is evident in figure [ f.ge ] when we reduce the number of unoccupied bands from 170 to 24 , which was the value used in ref .
gaas shows behavior similar to ge .
table [ t.gaas ] summarizes our results .
agreement with experimental data is now within 0.15 ev , and all - electron calculations again underestimate the energy gaps .
regarding the convergence with @xmath41 , figure [ f.gaas ] shows that the indirect gaps converges slowly from below . on the other hand
, the direct @xmath40 gap converges more quickly .
lccc + & @xmath33 & @xmath45 & @xmath34 + + & 0.13 & 0.70 & 1.21 + + shirley _
et al._@xmath35 & 1.42 & 1.75 & 1.95 + + & 1.38 & 1.65 & 1.83 + + kotani and van schilfgaarde@xmath36 & 1.20 & 1.40 & 1.46 + experiment@xmath37 & 1.52 & 1.815 & 1.98 + [ t.gaas ] @xmath35 reference 9 .
@xmath36 reference 6 after inclusion of spin - orbit effects .
@xmath37 references 23,24
. in all systems studied , we applied the gpp model in two ways : ( 1 ) using only the valence charge density in the f - sum rule , and ( 2 ) using a total charge density from both the valence and semicore bands .
method ( 1 ) is physically more realistic , and predicts an energy dependence of the inverse dielectric function that is consistent with the rpa .
method ( 2 ) implies that semicore electrons are able to screen electric fields as efficiently as valence electrons , which is not physical . nevertheless , we find agreement between the two schemes to better than 0.1 ev in the converged energy gap for all systems studied .
this is a consequence of cancellation of errors : in method ( 2 ) , the plasma frequency is overestimated , and therefore the contributions @xmath25 and @xmath26 are overestimated in absolute value .
since they have opposite sign , the final quasiparticle energies are weakly affected .
we also observe that method ( 2 ) shows slower convergence of energy gaps with respect to @xmath41 , which is expected since the @xmath26 term is enhanced . all results presented in this article
were obtained using the physically more appropriate method ( 1 ) .
as shown in figures [ f.si]-[f.gaas ] , energy gaps in the gw approximation have a significant dependence on the number of unoccupied bands that is slow to converge in many cases .
this fact was already reported in calculations using standard pseudopotential techniques.@xcite we can analyze the physics of this convergence by examining the gw approximation in the static limit , the so - called cohsex approximation .
bands gaps can be calculated more easily within the cohsex approximation and show convergence behavior similar to the full dynamic calculation . under this approximation , the coulomb - hole term has a simple form in terms of the polarization potential , @xmath55 , @xmath56 where the second equality follows from completeness of the basis of eigenvectors . in actual calculations ,
this sum over bands @xmath27 is always truncated , and the equality is violated .
the coulomb - hole energy evaluated at a given electronic state is calculated according to the expression @xmath57^\star { \cal m}^{nm}_{\bf g'}({\bf k , q})\ ; \ ; , \label{k_sigma_coh}\ ] ] where @xmath58 are the coefficients in the plane - wave expansion of the polarization potential , and we define @xmath59 physically , the summation in eq .
( [ sigma_coh ] ) describes virtual transitions produced when the quasiparticle induces a charge fluctuation around itself .
@xmath60 is the energy associated to the interaction between the quasiparticle and the induced charge fluctuation .
the matrix elements , eq .
( [ mat_el ] ) , decay slowly as the energy difference between bands @xmath61 and @xmath27 increases . monitoring the convergence of these matrix elements
provides a good estimate of the relative error in the @xmath60 , but gauging the absolute convergence of the self - energy requires knowing @xmath62 , which depends on the physical system .
although we have not directly investigated the importance of imposing self - consistency in the calculation of the self - energy , this is an unsetled issue and we address it briefly below .
von barth and holm have investigated the effect of self - consistency in the electron gas,@xcite and concluded that restricted self - consistency has small but significant effect on the full bandwidth and in the satellite structure of the electron gas . on the other hand ,
full self - consistency gives a poor description of the satellite structure and the bandwidth is drastically increased .
inclusion of vertex corrections are expected to recover the good , non - self - consistent results , but calculation of vertex corrections is not a simple task even for the electron gas system.@xcite self - consistency has been recently applied to real materials,@xcite and the valence bandwidth is also shown to increase when self - consistency is imposed .
it appears that one must therefore include self - consistency and vertex corrections together in order to obtain a meaningful picture .
inclusion of vertex corrections and self - consistency is tangential for the purpose of our work , which is to compare pseudopotential - based and all - electron gw calculations and understand the role of semicore electrons .
we conclude that for the systems considered , the valence - only pseudopotential method does not suffer from large errors from the neglect of core states , as claimed in some all - electron calculations . @xcite while semicore effects are negligible in si , they are important if one aims at good quantitative agreement with experiment in ge and gaas.@xcite however , discrepancies between pseudopotential and all - electron based gw calculations reported in recent works @xcite may be explained by a lack of numerical convergence in the latter . specifically , the self - energy calculated within the gw method has slow convergence with respect to the number of energy bands included in the calculation of the green s function , as is demonstrated in this work and has been pointed out in the past .
@xcite this convergence behavior is present in the static limit to gw , the cohsex approximation , and can be analyzed by comparing the cohsex coulomb - hole energy , eq .
( [ sigma_coh ] ) , obtained with and without explicit summation over energy bands .
we thank andrew canning and david clatterbuck for fruitful discussions regarding the flapw method .
this work was supported by national science foundation grant no .
dmr00 - 87088 and by the director , office of science , office of basic energy sciences , division of materials sciences and engineering , u.s .
department of energy under contract no .
de - ac03 - 76sf00098 .
computational resources have been provided by nsf at the national partnership for advanced computational infrastructure , and doe at the national energy research scientific computing center .
landolt - brnstein , _ numerical data and functional relationships in science and technology _ , eds .
hellwege , o. madelung , m. schulz , and h. weiss , new series , vol iii , pt .
22a , springer - verlag , new york ( 1987 ) . | we study the effect of semicore states on the self - energy corrections and electronic energy gaps of silicon , germanium and gaas .
self - energy effects are computed within the gw approach , and electronic states are expanded in a plane - wave basis . for these materials ,
we generate _ ab initio _ pseudopotentials treating as valence states the outermost two shells of atomic orbitals , rather than only the outermost valence shell as in traditional pseudopotential calculations .
the resulting direct and indirect energy gaps are compared with experimental measurements and with previous calculations based on pseudopotential and `` all - electron '' approaches .
our results show that , contrary to recent claims , self - energy effects due to semicore states on the band gaps can be well accounted for in the standard valence - only pseudopotential formalism . |
a complete characterization of the achievable rate distortion region for the classical lossy multiterminal source coding problem depicted in fig .
[ fig : mtsc ] has remained an open problem for over three decades .
several special cases have been solved : * the lossless case where @xmath0 .
slepian and wolf solved this case in their seminal work@xcite . * the case where one source is recovered losslessly : i.e. , @xmath1 .
this case corresponds to the source coding with side information problem of ahlswede - krner - wyner @xcite,@xcite . *
the wyner - ziv case @xcite where @xmath2 is available to the decoder as side information and @xmath3 should be recovered with distortion at most @xmath4 . * the berger - yeung case ( which subsumes the previous three cases )
@xcite where @xmath4 is arbitrary and @xmath5 . despite the apparent progress ,
other seemingly fundamental cases , such as when @xmath4 is arbitrary and @xmath6 , remain unsolved except perhaps in very special cases . in this paper , we give the achievable rate region for two cases subject to a particular choice of distortion measure @xmath7 , defined in section [ sec : probstatement ] .
specifically , for our particular choice of @xmath7 , we give the achievable rate distortion region for the following two cases : * the situation when @xmath8 and @xmath9 are subject to a joint distortion constraint given a reproduction @xmath10 : @xmath11\leq d.\end{aligned}\ ] ] * the case where @xmath8 is subject to a distortion constraint given a reproduction @xmath12 : @xmath13\leq d_x,\end{aligned}\ ] ] and there is no distortion constraint on the the reproduction of @xmath9 ( i.e. , @xmath14=@xmath15 ) .
the regions depend critically on our choice of @xmath7 , which can be interpreted as a natural measure of the soft information the reproduction @xmath10 symbol provides about the source symbols @xmath8 and @xmath9 ( resp .
the information @xmath12 provides about @xmath8 ) .
the remainder of this paper is organized as follows . in section
[ sec : probstatement ] we formally define the problem and provide our main results . in section [ sec : proofs ] , we discuss the properties of @xmath7 and provide the proofs of our main results .
section [ sec : conc ] delivers the conclusions and a brief discussion regarding further directions .
in this paper , we consider two cases of the lossy multiterminal source coding network presented in fig . [
fig : jdxd ] . in the first case ,
we study the achievable rates @xmath16 subject to the joint distortion constraint @xmath11\leq d,\end{aligned}\ ] ] where @xmath10 is the joint reproduction symbol computed at the decoder from the messages @xmath17 and @xmath18 received from the @xmath8- and @xmath9-encoders respectively . in the second case ,
we study the achievable rates @xmath16 subject to a distortion constraint on @xmath8 : @xmath13\leq d_x,\end{aligned}\ ] ] where @xmath12 is the reproduction symbol computed at the decoder from the messages @xmath17 and @xmath18 received from the @xmath8- and @xmath9-encoders respectively . in this second case
, there is no distortion constraint on @xmath9 . to simplify terminology
, we refer to the first and second cases described above as the joint distortion ( jd ) network and x - distortion ( xd ) network respectively .
formally , define the source alphabets as @xmath19 and @xmath20 .
we consider the discrete memoryless source sequences @xmath3 and @xmath2 drawn i.i.d . according to the joint distribution @xmath21 .
let @xmath3 be available at the @xmath8-encoder and @xmath2 be available at the @xmath9-encoder as depicted in fig .
[ fig : jdxd ] .
( we will informally refer to probability mass functions as distributions throughout this paper . ) for the case of joint distortion , we consider the reproduction alphabet @xmath22 , where @xmath23 denotes the set of probability distributions on @xmath24 points . in other words , for @xmath25 , @xmath26 where @xmath27 and @xmath28 . with @xmath29 defined in this way , it will be convenient to use the notation @xmath30 for @xmath31 .
note that the restriction of the reproduction alphabet to the probability simplex places constraints on the function @xmath32 .
for example , one can not choose @xmath33 .
define the joint distortion measure @xmath34 by @xmath35 and the corresponding distortion between the sequences @xmath36 and @xmath37 as @xmath38 as we will see in section [ sec : proofs ] , the distortion measure @xmath7 measures the amount of soft information that the reproduction symbols provide about the source symbols in such a way that the expected distortion can be described as an entropy .
for example , given the output from a discrete memoryless channel , the minimum distortion between the channel input and output is the conditional entropy .
for this reason , we refer to @xmath7 as an entropy - based distortion measure .
the function @xmath7 is a natural distortion measure for practical scenarios .
a similar distortion measure has appeared previously in the image processing literature @xcite and in the study of the information bottleneck problem @xcite .
however , it does not appear to have been studied in the context of multiterminal source coding .
a @xmath39-rate distortion code for the jd network consists of encoding functions , @xmath40 and a decoding function @xmath41 a vector @xmath42 with nonnegative components is achievable for the jd network if there exists a sequence of @xmath39-rate distortion codes satisfying @xmath43\leq d.\end{aligned}\ ] ] the achievable rate distortion region , @xmath44 , for the jd network is the closure of the set of all achievable vectors @xmath42 . in a similar manner , we can also consider the case when there is only a distortion constraint on @xmath8 rather than a joint distortion constraint on @xmath45 . for this , we consider the reproduction alphabet @xmath46
. with @xmath47 defined in this way , it will be convenient to use the notation @xmath48 for @xmath49 .
we define the distortion measure @xmath50 by @xmath51 and the corresponding distortion between the sequences @xmath52 and @xmath53 as @xmath54 identical to the case for the jd network , we can define a @xmath39-rate distortion code for the xd network , with the exception that the range of the decoding function @xmath55 is the reproduction alphabet @xmath56 .
a vector @xmath57 with nonnegative components is achievable for the xd network if there exists a sequence of @xmath39-rate distortion codes satisfying @xmath58\leq d_x.\end{aligned}\ ] ] the achievable rate distortion region , @xmath59 , for the xd network is the closure of the set of all achievable vectors @xmath57 .
our main results are stated in the following theorems : [ thm : achievableregion ] @xmath60 [ thm : achievableregionx ] @xmath61 since the distortion measure is reminiscent of discrete entropy , we can think of the units of distortion as bits " of distortion . thus , theorem [ thm : achievableregion ] states that for every bit of distortion we allow for @xmath45 jointly , we can remove exactly one bit of required rate from the constraints defining the slepian - wolf achievable rate region .
indeed , we prove the theorem by demonstrating a correspondence between a modified slepian - wolf network and the multiterminal source coding problem in question . similarly ,
when we only consider a distortion constraint on @xmath8 , theorem [ thm : achievableregionx ] states that for every bit of distortion we tolerate , we can remove one bit of rate required by the @xmath8-encoder in the ahlswede - krner - wyner region .
the proofs of theorems [ thm : achievableregion ] and [ thm : achievableregionx ] are given in the next section .
we choose to prove theorems [ thm : achievableregion ] and [ thm : achievableregionx ] by showing a correspondence between schemes that achieve a prescribed distortion constraint and the well - known lossless distributed source coding scheme of slepian and wolf , and the source coding with side - information scheme of ahlswede , krner , and wyner .
this provides a great deal of insight into how the various distortions are achieved . in each case
, the proof relies on a peculiar property of the distortion measure @xmath7 .
namely , the ability to convert expected distortions to entropies that are easily manipulated . in the following subsection
, we discuss the properties of the distortion measure @xmath7 .
[ lem : expecteddistortion ] given any @xmath62 arbitrarily correlated with @xmath63 , the estimator @xmath64 $ ] produces the expected distortion @xmath65 \geq \frac{1}{n}\sum_{i=1}^nh(x_i , y_i|u).\end{aligned}\ ] ] moreover , this lower bound can be achieved by setting @xmath66(x , y):=\pr\left(x_i = x , y_i = y|u = u\right)$ ] . given
any @xmath62 arbitrarily correlated with @xmath63 , denote the reproduction of @xmath63 from @xmath62 as @xmath64\in \hat{\mathcal{z}}^n$ ] . by definition of the reproduction alphabet
, we can consider the estimator @xmath64 $ ] to be some probability distribution on @xmath67 conditioned on @xmath62 .
then , we obtain the following lower bound on the expected distortion conditioned on @xmath68 : @xmath69 \\&= \frac{1}{n}\sum_{i=1}^n \sum_{x , y\in\mathcal{x}\times\mathcal{y } } p_i(x , y|u ) \log\left(\frac{1}{\hat{z}_i[u](x , y ) } \right)\\ & = \frac{1}{n}\sum_{i=1}^n d\left(p_i(x , y|u)||\hat{z}_i[u](x , y)\right)+h(x_i , y_i|u = u)\\ & \geq \frac{1}{n}\sum_{i=1}^n h(x_i , y_i|u = u),\end{aligned}\ ] ] where @xmath70 is the true conditional distribution . averaging both sides over all values of @xmath62 , we obtain the desired result .
note that the lower bound can always be achieved by setting @xmath66(x , y):=p_i(x , y|u)$ ] .
let @xmath71 be drawn i.i.d . and let @xmath72 be given .
the rate distortion function with side information is @xmath73 where the minimization is over all functions @xmath74 and conditional distributions @xmath75 such that @xmath76\leq d$ ] . in both examples
, we make the surprising observation that the distortion function @xmath7 yields a rate distortion function that is a multiple of the rate distortion function obtained using the erasure " distortion measure @xmath81 defined as follows : @xmath82 this is somewhat counter - intuitive given the fact that an estimator is able to pass much more soft " information to the distortion measure @xmath7 compared to @xmath81
. it would be interesting to understand whether or not this relationship holds for general multiterminal networks , however this issue remains open .
we have defined @xmath7 to be a joint distortion measure on @xmath67 , however it is possible to decompose it in a natural way .
we can define the marginal and conditional distortions for @xmath8 and @xmath83 respectively by decomposing @xmath66(x , y)=\hat{z}_i(x|u)\hat{z}_i(y|x , u)$ ] ( note the slight abuse of notation ) .
thus , if the total expected distortion is less than @xmath84 , we define the marginal and conditional distortions @xmath4 , and @xmath85 as follows : @xmath86 \\ & = \mathbb{e}\left[d_x(x^n,\hat{z}^n ) \right ] + \mathbb{e}\left[d_{y|x}(y^n,\hat{z}^n ) \right]\\ & : = d_x+d_{y|x}\\ & \geq \frac{1}{n } \sum_{i=1}^nh(x_i|u)+h(y_i|u , x_i).\end{aligned}\ ] ] in a complimentary manner , we can decompose the expected distortion into @xmath87 satisfying @xmath88 .
suppose a length @xmath91 code satisfies the expected distortion constraint @xmath92<d+\epsilon/2 $ ] . by repeating the code @xmath93 times
, we obtain @xmath93 i.i.d .
realizations of @xmath94 . by the weak law of large numbers : @xmath95 for @xmath93 sufficiently large . for each @xmath99 , we can rearrange ( [ eqn : dx_expression ] ) to obtain @xmath100 by the definition of @xmath37 , observe that @xmath101 is a valid probability measure on @xmath102 .
thus , for any subset @xmath103 , we have @xmath104 combining ( [ eqn : dx_expressionforcontradiction ] ) and ( [ eqn : subset ] ) gives the desired result : @xmath105 the basic idea is to let the @xmath8-encoder send @xmath166 ( requiring rate @xmath167 ) and have the @xmath9-encoder send @xmath168 ( requiring rate @xmath169 ) . by lemma
[ lem : conditionaldistortionsetsize ] , the number of @xmath3 sequences that lie in @xmath170 is less than @xmath171 .
therefore , if the @xmath8-encoder performs a random binning of the @xmath3 sequences into @xmath128 and sends the bin index corresponding to the observed sequence @xmath3 ( incurring an additional rate of @xmath172 ) , the decoder can recover @xmath3 losslessly with high probability .
since @xmath165 is an achievable rate vector , there exists some conditional distribution @xmath173 so that @xmath174 and @xmath175 .
wlog , reduce @xmath167 and @xmath169 if necessary so that @xmath176 and @xmath177 .
now , we construct a sequence of codes that achieve that point in the standard way .
in particular , generate @xmath178 different @xmath179 sequences independently i.i.d . according to @xmath180 . upon observing @xmath2 ,
the @xmath9-encoder finds a jointly typical @xmath179 and sends the corresponding index to the decoder . at the @xmath8-encoder ,
bin the @xmath3 sequences into @xmath181 bins and , upon observing the source sequence @xmath3 , send the corresponding bin index to the decoder . with high probability , the decoder can reconstruct @xmath3 losslessly . 1 .
use the lossless code described above with probability @xmath182 . in this case , the distortion on @xmath8 can be made arbitrarily small .
note that we can assume w.l.o.g . that @xmath183 since distortion @xmath184 can be achieved when the decoder only receives the sequence @xmath179 .
2 . with probability @xmath185 ,
the @xmath8-encoder sends nothing , while the @xmath9-encoder continues to send @xmath179 . in this case
, the distortion on @xmath8 is @xmath186 .
r. ahlswede and j. krner .
source coding with side information and a converse for the degraded broadcast channel .
ieee trans .
theory , it-21:629 - 637 , 1975 .
a. wyner . on source coding with side information at the decoder .
ieee trans .
theory , it-21:294 - 300 , 1975 .
i. csiszr and j. krner . towards a general theory of source networks .
ieee trans .
theory , it-26:155 - 165 , 1980 . i. csiszr and j. krner .
information theory : coding theorems for discrete memoryless systems . academic press , new york , 1981 . | in this paper , we consider a class of multiterminal source coding problems , each subject to distortion constraints computed using a specific , entropy - based , distortion measure .
we provide the achievable rate distortion region for two cases and , in so doing , we demonstrate a relationship between the lossy multiterminal source coding problems with our specific distortion measure and ( 1 ) the canonical slepian - wolf lossless distributed source coding network , and ( 2 ) the ahlswede - krner - wyner source coding with side information problem in which only one of the sources is recovered losslessly . |
the decays of heavy hadrons have recently received much attention in the literature @xcite . from the experimental standpoint ,
these decays allow access to some of the fundamental parameters of the standard model , such as the elements of the cabibbo - kobayashi - maskawa ( ckm ) matrix . questions of cp violation , heavy - flavor oscillations and many others have added to this interest . from the theoretical standpoint , these processes , and the heavy hadrons themselves , allow various quark models of qcd , as well effective theories , to be tested . in particular , the heavy quark effective theory ( hqet ) has both been tested by experimental observations , and has played a major role in the extraction of @xmath6 from experimental data @xcite .
much of the success of hqet has been in the treatment of decays from one heavy flavor to another , namely @xmath7 transitions .
the effective theory is more limited in scope when applied to heavy - to - light transitions , such as @xmath8 or @xmath9 .
neverthless , as we will outline in a later section , the scaling behavior of the form factors that describe various weak decays can be deduced @xcite .
this , in principle , allows the form factors for @xmath10 transitions to be inferred from those for @xmath8 . in this article
, we assume the validity of the heavy quark symmetry and examine the decays @xmath0 , @xmath1 , and the recently measured @xmath11 in a number of scenarios .
in particular , we seek a scenario in which all of these decays are adequately described by a single set of form factors .
a number of authors have performed similar analyses @xcite , using the decays @xmath12 and @xmath1 , with varying results .
once we find a scenario that is satisfactory for the decays mentioned above , we examine , briefly , the decays @xmath3 . in this way , we hope to make reliable estimates of the absolute decay rates for these processes . in a later article
, we will consider more details of these decays , such as forward - backward asymmetries .
these decays , as well as the decay @xmath2 , are particularly interesting as the short distance operators responsible arise first at the one - loop - level , and are therefore sensitive to new physics beyond the standard model @xcite . however , in order for any effects due to such new physics to be clearly identified , the long distance contributions that arise in the hadronic matrix elements must be well understood . as witness to this ,
we point out that the inclusive rate @xmath13 is reasonably well understood , while the exclusive rate @xmath2 has been predicted to be anywhere from 2% to 40% of this inclusive rate @xcite . in the case of the nonleptonic and the rare decays
, there arises the crucial issue of the form factors describing the @xmath14 transition .
such form factors may be estimated in various models , from qcd sum rules , or by applying the scaling relations predicted by hqet to the form factors for the corresponding @xmath15 transition .
the method suggested by hqet , while clearly model - independent , is itself somewhat problematic , as one must take into account the fact that predicting @xmath16 , for instance , requires that these relations be carefully extrapolated beyond the kinematic range accessible in the @xmath15 transitions .
this is because the maximum @xmath17 in the @xmath15 transitions is 1.95 ( 0.95 ) gev@xmath18 , while the @xmath17 appropriate to the @xmath19 transition is 9.6 gev@xmath18 .
the hqet symmetry predictions relate form factors at the same value of the kinematic variable @xmath20 ( or @xmath21 ) , not @xmath17 , where @xmath22 and @xmath23 are the four - velocities of the parent and daughter hadron , respectively . in this case , the extrapolation is from @xmath20 = 2.1 in @xmath24 to @xmath20 = 3.6 in @xmath25 .
the corresponding numbers for the @xmath26 decays are 1.3 and 2.0 .
the question of extrapolation also applies to the rare decays , and is particularly important for the decay @xmath2 , for which @xmath17 = 0 , but @xmath20 = 3.0 . for
the nonleptonic decays , a second issue is that of the factorization approximation , which is commonly used to calculate the hadronic matrix elements required .
this approximation is not very well founded in general theoretically , yet it appears to work well phenomenologically in the @xmath27 decays where it has been tested .
nevertheless , application of the form factors of some model or effective theory to the decay @xmath25 , in conjunction with the factorization approximation , serves to probe both issues , and may fail due either to inadequate choice of form factors , failure of the factorization approximation , or both .
the rest of this article is organized as follows . in the next section
we describe briefly the effective hamiltonians and the appropriate hadronic matrix elements for the processes of interest . in section
iii we use hqet to obtain relations among the form factors of interest . in section
iv we discuss our fitting procedure , and present the results for the decays that we are considering . in section v we discuss possible limitations of our results , and suggest questions that may be of interest to both theorists and experimentalists .
of the three processes we discuss , the semileptonic decays are perhaps the simplest to treat theoretically .
the effective hamiltonian for these decays is @xmath28 the hadronic matrix elements for the decays @xmath0 are @xmath29 these decays are thus described in terms of six independent , _ a priori _ unknown form factors .
the terms in @xmath30 and @xmath31 are unimportant when the lepton mass is ignored , since @xmath32 in experimental analyses , the form factors @xmath33 , @xmath34 , @xmath35 and @xmath36 are usually assumed to have the form @xmath37 where @xmath38 , and @xmath39 is a mass , usually taken to be that of the nearest resonance with the appropriate quantum numbers . to date , only the mass appropriate to @xmath33 has been measured , and its value measured by cleo is @xmath40 gev @xcite .
the masses appropriate to @xmath34 , @xmath35 and @xmath36 are assumed to be 2.5 gev , 2.5 gev and 2.1 gev respectively .
the @xmath41 have the values @xmath42 , @xmath43 , @xmath44 and @xmath45 , for @xmath33 , @xmath34 , @xmath36 and @xmath35 , respectively @xcite .
neglecting penguin contributions , the effective hamiltonian for the nonleptonic decays of interest here is @xmath46,\end{aligned}\ ] ] where @xmath47,\nonumber\\ c_2 ( m_b ) & = & \frac{1}{2 } \left [ \left ( \frac { \alpha_s ( m_b ) } { \alpha_s ( m_w ) } \right)^{-6/23 } - \left ( \frac { \alpha_s ( m_b ) } { \alpha_s ( m_w ) } \right)^{12/23 } \right].\end{aligned}\ ] ] here , @xmath48 is the mass of the @xmath49 boson , and @xmath50 is the running coupling of qcd .
this effective hamiltonian mediates two classes of @xmath27 nonleptonic decays .
the first class contains a @xmath51 in the final state : @xmath52 where @xmath53 may be @xmath54 .
the second class contains a light meson in the final state : @xmath55 ` @xmath56 ' @xmath57 where @xmath57 is now a charmonium state . to evaluate the matrix elements of the effective hamiltonian we employ the factorization assumption .
by fierz rearrangement we rewrite the effective hamiltonian in a form which is suitable for use with this assumption .
both terms of the effective hamiltonian contribute , in general , but for the decays in which we are interested , only the second term is of interest , and it may be written @xmath58 where @xmath59 is the number of colors . at this point , it has become customary to replace the coupling coefficient by a phenomenological constant @xmath60 , whose absolute value is measured to be approximately 0.24 .
the sign of @xmath60 is not important for our discussion at this point
. it will become important if long distance contributions to the dileptonic rare decays are included .
for the decay @xmath61 , we therefore write , after using factorization , @xmath62 the hadronic matrix elements @xmath63 are analogs of those of the previous subsection , so that the form factors required for these matrix elements may , in principle , be obtained from the ` semileptonic ' decays of @xmath27 mesons into kaons .
such decays do not take place in the standard model , but we can invoke heavy quark symmetries to relate the form factors needed to those for the semileptonic decays of @xmath51 mesons to kaons .
the remaining matrix element is @xmath64 the decay constant @xmath65 can be obtained from the leptonic width of the appropriate charmonium vector resonance as @xmath66 where we have ignored lepton masses . in this way
, we find @xmath67=0.382 gev , and @xmath68=0.302 gev . in the standard model ,
the effective hamiltonian for the decay @xmath69 is @xcite @xmath70b f^{\mu\nu},\ ] ] where @xmath71 is the electromagnetic field strength tensor , and the term in @xmath72 may be safely ignored . for the decay @xmath73 , the corresponding effective hamiltonian is @xmath74.\end{aligned}\ ] ]
the wilson coefficients @xmath75 are as in the article by buras _
_ @xcite . for the discussion at hand
, we have ignored contributions that arise from closed @xmath76 loops , although these may be easily included .
the only new hadronic matrix elements that arise in the rare decays of interest are @xmath77 , \nonumber\\ \left < k^*(p^\prime,\epsilon)\left|\bar s\sigma_{\mu\nu } b\right|b(p ) \right>&=&\epsilon_{\mu\nu\alpha\beta}\left[g_+\epsilon^{*\alpha } \left(p+p^\prime\right)^\beta + g_-\epsilon^{*\alpha}\left(p - p^\prime\right)^\beta\right.\nonumber\\ & & \left.+h\epsilon^*\cdot p\left(p+p^\prime\right)^\alpha\left(p - p^ \prime\right)^\beta\right],\end{aligned}\ ] ] introducing four new form factors .
due to the relation @xmath78 we can easily relate the matrix elements above to those in which the current is @xmath79 . experimentally , nothing is known about the form factors @xmath80 , @xmath81 nor @xmath82 .
using the dirac matrix representation of heavy mesons , one may treat heavy - to - light transitions using the same trace formalism that has been applied to heavy - to - heavy transitions @xcite . in the effective theory ,
a @xmath51 meson traveling with velocity @xmath22 is represented as @xcite @xmath83 with an identical representation for a @xmath27 meson .
the meson states of the effective theory are normalized so that @xmath84 the states of qcd and hqet are therefore related by @xmath85 in all that follows , we will represent the states of full qcd as @xmath86 , and the states of hqet as @xmath87 .
let us consider transitions between such a heavy meson ( @xmath51 meson ) and a light meson ( kaon ) of spin @xmath88 , through a generic flavor changing current . in the effective theory ,
the matrix element of interest is @xmath89 @xmath90 is an arbitrary combination of dirac @xmath91 matrices , and @xmath92 is the fully symmetric , traceless , transverse , @xmath88-index tensor that represents the polarization of the state with spin @xmath88
. the matrix @xmath93 must be the most general that can be constructed from the kinematic variables available , and dirac @xmath91 matrices .
the most general form for this is @xcite @xmath94\left(\matrix{1\cr\gamma_5}\right).\end{aligned}\ ] ] the @xmath95 s are uncalculable , nonperturbative functions of the kinematic variable @xmath96 , and the 1 ( @xmath97 ) is present if the resonance @xmath98 has natural ( unnatural ) parity . from this point on , let us limit the discussion to only two of the kaon resonances , namely the ground - state pseudoscalar kaon itself , and its vector counterpart , @xmath26 .
the above then leads to @xmath99\gamma m_d(v)\right\}.\end{aligned}\ ] ] at this point , let us emphasize that the form factors @xmath95 are independent of the form of @xmath90 , and so are valid for decays through the left handed current ( @xmath100 ) , as well as for rare decays ( @xmath101 ) .
these form factors are also independent of the mass of the heavy quark , and are therefore universal functions .
thus , they are valid for @xmath102 decays , as well as for @xmath103 decays .
this independence of the quark mass allows us to deduce , in a relatively straightforward manner , the scaling behavior of the usual form factors that describe these transitions @xcite .
we illustrate this by examining one matrix element in detail .
consider @xmath104 from these equations , one finds that @xmath105 since the @xmath95 do not scale with the mass of the heavy quark ( or meson ) , it is trivial to deduce that @xmath106 for the other transitions of interest , the form factors are as defined in the previous section , and the relationships between these and the @xmath95 are @xmath107 these expressions yield the additional scaling behavior @xmath108 eqns .
( [ formsgeneral ] ) and ( [ ff1 ] ) , and consequently eqns .
( [ formsscaling]-[xitof ] ) , contain all of the leading order @xmath109 dependence in the form factors , and are valid irrespective of the mass of the strange quark .
if the strange quark could be treated as heavy , then we could think of the @xmath95 as arising from an infinite sum of terms in the @xmath110 expansion .
the leading order forms ( in @xmath110 ) are also contained in these expressions .
these expressions are also valid in the limit of a very light strange quark .
isgur and wise @xcite have used the scaling of @xmath111 ( @xmath112 ) to say that this combination of form factors vanishes ( at leading order in @xmath113 ) , and suggest that one can write @xmath114 .
implicit in this argument is the assumption that the strange quark is very light . in our formalism
, this amounts to setting @xmath115 to zero , and we would automatically lose the full scope of our predictions . this is because we could then never recover the limit of a heavy strange quark , for in this limit , @xmath116 , where @xmath117 is the usual isgur - wise function .
the strange quark is such that it may be treated as either heavy or light .
we believe that neither the full heavy @xmath80 limit ( @xmath116 ) , nor the limit of a very light @xmath80 quark ( @xmath118 ) is completely satisfactory .
we assume neither limit in our analysis , and therefore make full use of the predictions of the heavy quark effective theory , which are valid independent of the mass of the strange quark .
this means that we retain the form factor @xmath115 in our discussion and treat it as a completely independent form factor , tying it to neither of the two limits discussed . at the risk of boring the overly patient reader to tears , and perhaps even to death , we list one more set of relationships among form factors , this time writing the usual form factors in terms of the @xmath95 .
the relations are @xmath119,\nonumber\\ a_-&=&-\frac{1}{m_d^{3/2}}\left[\xi_3-m_d\left(\xi_6-\xi_4\right ) \right],\nonumber\\ g&=&\frac{1}{\sqrt{m_d}}\xi_6.\end{aligned}\ ] ] for the corresponding transitions with a @xmath120 quark ( and a @xmath27 meson ) in the initial state , all factors of @xmath109 above must be replaced by @xmath121 . using this , rearrangement of eqns .
( [ ff2 ] ) yields @xmath122 , \nonumber\\ f_-^b(v\cdot p)&=&\frac{1}{2}\left(\frac{m_b}{m_d}\right)^{1/2 } \left[f_-^d(v\cdot p)\left(1+\frac{m_d}{m_b}\right ) + f_+^d(v\cdot p)\left(\frac{m_d}{m_b}-1\right)\right ] , \nonumber\\ f^b(v\cdot p)&=&\left(\frac{m_b}{m_d}\right)^{1/2}f^d(v\cdot p ) , \nonumber\\ g^b(v\cdot p)&=&\left(\frac{m_d}{m_b}\right)^{1/2}g^d(v\cdot p ) , \nonumber\\ a_+^b(v\cdot p)&=&\frac{1}{2}\left(\frac{m_d}{m_b}\right)^{1/2 } \left[a_+^d(v\cdot p)\left(1+\frac{m_d}{m_b}\right ) + a_-^d(v\cdot p)\left(\frac{m_d}{m_b}-1\right)\right ] , \nonumber\\ a_-^b(v\cdot p)&=&\frac{1}{2}\left(\frac{m_d}{m_b}\right)^{1/2 } \left[a_-^d(v\cdot p)\left(1+\frac{m_d}{m_b}\right ) + a_+^d(v\cdot p)\left(\frac{m_d}{m_b}-1\right)\right],\end{aligned}\ ] ] where @xmath123 is the form factor appropriate to the @xmath102 transition , while @xmath124 is the form factor appropriate to the @xmath125 transition , and quantities on the left - hand - sides of eqns .
( [ ffmix ] ) are evaluated at the same values of @xmath96 as those on the right - hand - sides .
omitted from each of eqn .
( [ ffmix ] ) is a qcd scaling factor , discussed below .
( [ ffmix ] ) illustrate two effects , namely the scaling of form factors in going from the @xmath51 system to the @xmath27 system , as well as the mixing of @xmath34 with @xmath31 , and @xmath33 with @xmath30 .
the effect of this mixing is very important in going from @xmath51 transitions to @xmath27 transitions , as it introduces form factors that have not yet been measured experimentally , or to which experiments are not yet sensitive , namely @xmath30 and @xmath31 . in the rates for the semileptonic
decays @xmath12 , terms dependent on @xmath30 and @xmath31 are proportional to the mass of the lepton , and thus play a miniscule role , except near @xmath126 .
such terms may also be significant if the polarization of the charged lepton is measured .
we close this section with a brief discussion of radiative corrections to the currents and matrix elements that we have discussed . in the limit of a heavy @xmath120 quark ,
the full current of qcd is replaced by @xcite @xmath127^{-\frac{6}{25}}.\ ] ] this arises from integrating out the @xmath120 quark , and matching the resulting effective theory onto full qcd at the scale @xmath128 , at one loop level . at the scale @xmath129 , we must also integrate out the @xmath130 quark , but there is also the effect due to running between @xmath128 and @xmath129 .
the net effect of this is that the form factors @xmath95 appropriate to the @xmath10 transitions are related to those for the @xmath8 transitions by @xmath131^{-\frac{6}{25}}.\ ] ] the forms of the matrix elements that we discuss above are valid in the limit of infinitely heavy @xmath120 and @xmath130 quarks . for quarks of finite mass ,
there are clearly going to be corrections to the relations we have obtained . in other words ,
new form factors that appear first at order @xmath132 and @xmath133 will begin to make contributions .
it is expected that such contributions will become more significant away from the ` non - recoil ' point , or for @xmath134 .
this is particularly important for the @xmath14 decays , as @xmath96 can become very large .
nevertheless , we will apply the relations we have found through all of the available phase space .
it is our hope that by doing this , we will at least be able to indicate the suitability of hqet for these transitions .
however , since we fit the form factors rather than attempt to calculate them in some model , some of these higher order effects may , in fact , have been included .
all of the results we describe are for decays with @xmath135 ( or @xmath136 ) in the final state .
the treatment of charged kaons would be identical , and we believe that our results in these channels would be similar in quality to those we obtain for neutral kaons . before describing the results of our fits , we must comment on how we treat the available data , particularly in the case of the semileptonic decays .
very few of the experimental collaborations have extracted acceptance - corrected distributions @xcite . instead , the form factors are all assumed to be of the monopole form @xcite , and the parameters are then extracted from the monte carlo simulations , with acceptances and efficiencies folded in .
because of this , we proceed in the inverse sense to generate some ` simulated data ' .
we use the published monopole parameters for the form factors to generate @xmath137 spectra for the semileptonic decays , using the published uncertainties in the monopole parameters to generate uncertainties in the simulated data . in general , these errors are correlated , but we ignore this correlation .
the simulated data generated in this way are completely smooth .
we introduce an ` anti - smoothing ' by smearing the simulated data with a pseudo - randomly generated gaussian distribution of mean zero and standard deviation determined by the errors in the unsmeared simulated data .
it is this smeared simulated data that we use for fitting . for the decays
@xmath138 , we also include the ratios @xmath139 and @xmath140 in the fit , where @xmath141 are as defined in pdg @xcite .
in addition , we must point out that the measured decay rates for @xmath142 are somewhat smaller than those for @xmath143 , while the published form factor parametrizations are for the average over charge states . to account for this ,
we rescale the values of the @xmath41 to correspond to the smaller rates for neutral kaons .
it is these rescaled values that are cited in section [ sldecays ] , and that we use to generate the simulated data .
for the nonleptonic decays , we fit to the pdg averaged decay widths for @xmath144 and @xmath145 . in the case of the latter
, we also include the ratio @xmath139 . for this ratio , we take the averaged value of @xmath146 as calculated by gourdin _
masses and lifetimes of mesons are all taken from pdg @xcite , and we use @xmath147 , @xmath148 , @xmath149 , @xmath150 , @xmath151 gev , @xmath129=1.5 gev , @xmath152=177 gev .
it is worth mentioning that the experimental choice of monopole form factors may be inappropriate , particularly for @xmath35 . in the limit of a heavy strange quark
, one finds that @xmath153 , where @xmath117 is the isgur - wise function .
if @xmath154 is assumed to be monopolar in @xmath17 , then simple pole dependence for @xmath35 is inappropriate . even in the case of a light strange quark
, we find that @xmath155 , again indicating that the dependence on the kinematic variable is not simply a monopole form .
however , for the range of @xmath17 accessible in @xmath156 decays , the decay rate is not sensitive to the kinematic dependence .
the effect on the decay rate for the nonleptonic and rare decays being considered here , however , are quite significant . in our fitting , we have separated the decays containing a @xmath56 meson in the final state from those containing a @xmath26 meson in the final state .
thus , for instance , we do not include data on ratios of rates like @xmath157 .
our reason is that such ratios introduce correlations between the parameters of the form factors for the two sets of decays .
one approximation used recently in the literature has been to treat the strange quark as heavy @xcite , so that the decays of interest can be treated in the heavy - to - heavy limit . in this limit , the form factors of eqn .
( [ ff1 ] ) may be written @xmath158 where @xmath117 is the isgur - wise function for heavy - to heavy transitions , and in this limit , @xmath159 .
in particular , this limit means that @xmath160 .
we have used this form in fitting the data , and have obtained reasonable fits to the differential decay rates in the semileptonic decays , as well as to the total decay rates in the nonleptonic decays .
polarization ratios , however , are poorly reproduced . in the case of the nonleptonic decay @xmath161 , the ratio @xmath139 depends only on kinematics , and has a value of 0.43 , independent of the form chosen for @xmath117 .
the experimental value is @xmath162 .
in addition , the ratio @xmath140 in @xmath163 always has a value of about 0.4 , significantly different from the experimental value of [email protected] , and largely independent of the form chosen for @xmath154 .
this indicates that the value of 0.43 obtained for the ratio @xmath139 in @xmath161 is not necessarily due to the breakdown of factorization in the heavy @xmath80 limit , as this limit does not even provide an adequate description of all measurements in the semileptonic decays .
relaxing the strict heavy-@xmath80 limit , but constraining the form factors to be near this limit , does not help much , as the polarization observables are still not well reproduced .
the conclusion that we draw from this is that the heavy-@xmath80 limit may give an acceptable description of unpolarized data , but may be dangerous when applied to polarization observables .
all of the results we present are obtained by performing four kinds of fits , namely ( 1 ) include the semileptonic decays @xmath165 only ; ( 1 ) include the semileptonic decays as well as the nonleptonic decays @xmath16 ; ( 3 ) include the semileptonic decays , the nonleptonic decays @xmath16 , and the nonleptonic decays @xmath166 ; ( 4 ) include the measured decay rate for @xmath11 in the fit , as well as the three other decays .
clearly , in the case of the decays to @xmath56 mesons , we need perform only three kinds of fits . in cases
where a measured quantity is not included in the fit , we calculate that quantity using the fit parameters .
we have explored two sets of parametrizations of the form factors . in the first scenario , which we refer to as the exponential scenario ,
each @xmath95 has one of the forms @xmath167}=a_i\exp{\left[-\frac{b_i}{2m_d}\left(q^2_{{\rm max } } - -q^2\right)\right]}\label{exp1},\\ \xi_i&=&a_i\exp{\left[-b_i\left(v\cdot p - m_{k^{(*)}}\right)^2 \right]}=a_i\exp{\left[-\frac{b_i}{4m_d^2}\left(q^2_{{\rm max}}-q^2 \right)^2\right ] } \label{exp2},\\ \xi_i&=&a_i\exp{\left[-b_i\left(v\cdot p\right)^2\right]}\label{exp3},\end{aligned}\ ] ] while in the second scenario , which we call the multipolar scenario , the forms chosen are @xmath168 with @xmath169 . in the exponential scenario , eqn .
( [ exp1 ] ) most closely corresponds to the forms that arise in some quark models , most notably that of isgur and collaborators @xcite .
however , in such models , the exponential of eqn .
( [ exp1 ] ) is usually multiplied by a polynomial in @xmath96 , or equivalently , in @xmath17 . in the multipolar scenario
, @xmath170 corresponds to a dipole form factor , while @xmath171 represents a monopole . in any one fit
, we do not choose all the form factors to have the same form .
this means that , for instance , in the case of the decays to @xmath56 mesons , for which there are two form factors , the second scenario corresponds to sixteen different possible combinations of forms for @xmath115 and @xmath172 . in each case , the @xmath173 and @xmath174 are the free parameters to be varied in the fit .
there are therefore twelve free parameters in each fit , to be compared with five extracted and three assumed parameters in the measured semileptonic decays .
since we have more free parameters than are extracted from the experimental data , one might expect that it should be very easy to account for all of the data .
in fact , the number of free parameters poses some problems , as it means that the problem is not very well constrained .
one consequence of this is that there appear to be several local minima for any particular choice of form factors . nevertheless , there are some combinations of form factors that simply do not provide adequate descriptions of the data , despite the large number of free parameters .
when only the semileptonic decays are included in the fit , we find that almost any combination of forms for the form factors leads to reasonable results .
the few combinations that do not provide good descriptions fail only in their description of the polarization observables . when we use the exponential forms , we find that we are not able to obtain an adequate description of all of the data simultaneously . in particular ,
when we include the nonleptonic decays in the fit , the prediction for the rare decay @xmath175 is significantly different from the measured rate . in some cases ,
however , we find that if we fit the semileptonic decays alone , omitting all of the nonleptonic decays , the prediction for the rare decays is of the right order of magnitude .
one possible conclusion here is that factorization is not applicable to these nonleptonic decays , or that there are significant non - factorizable contributions to the amplitude .
in contrast with the exponentials of the first scenario , many combinations of the ` multipolar ' forms of the second scenario lead to good descriptions of all the data simultaneously .
one outstanding feature of all of our results in this scenario ( which also exists in the exponential scenario , but to a lesser extent ) can be easily understood by examining eqn .
( [ ff2 ] ) , where it is seen that @xmath176 is present in all of the form factors that describe the semileptonic decays @xmath177 .
it is therefore not at all surprising that our results are most sensitive to this form factor .
invariably , we have found that the best fits occur for @xmath178 linear in @xmath96 ( _ i.e. _ , @xmath179 ) , independent of the forms chosen for the other form factors .
furthermore , the slope parameter @xmath180 is almost always negative , with values lying between -0.4 and -0.65 gev@xmath181 .
the only positive values for @xmath180 occur when only the semileptonic decays are included .
the fact that @xmath176 is so well constrained in our fits means that @xmath35 and @xmath36 are also quite well constrained .
since these are the only two form factors that are needed for the decay @xmath2 , it is no surprise that we find little variation in the predictions for this decay rate as we vary the parametrizations of @xmath182 , @xmath183 and @xmath184 , provided that the choice of parametrization of @xmath176 is unchanged . for other forms of @xmath176 , we find that at most two of the semileptonic , nonleptonic or rare decays are well accomodated .
for instance , if we choose a monopole form for @xmath176 , then in addition to the semileptonic decays , we find that we can accomodate either the nonleptonic decays , the rare decay , but not both .
in addition , if we choose all form factors to be monopolar , we fail to find adequate descriptions of the polarization observables , particularly for the ratio @xmath185 measured in the nonleptonic decay .
our results for the form factors are comparable with those of other authors .
@xcite have found that a number of scenarios , including monopolar form factors , are unable to describe the nonleptonic measurements .
_ @xcite have found that softening the scaling relations allows an adequate description , while in a second article , gourdin _ et al . _ @xcite have found that allowing the form factor @xmath35 to decrease linearly with @xmath17 allows an adequate description of the data .
we find that @xmath35 is quadratic in @xmath17 , but the absolute value at @xmath186 for the @xmath187 transitions is larger than the absolute value at @xmath188 , in keeping with the results of @xcite , and quite different from pole models . for each scenario ,
we have selected a set of fits that we consider to be representative .
the parameters for these fits are displayed in table [ fitparame ] for the exponential forms , and in table [ fitparam ] for the multipolar forms . in the first scenario ,
the results we have selected correspond to @xmath172 and @xmath182 as in eqn .
( [ exp1 ] ) , @xmath115 and @xmath183 as in eqn .
( [ exp2 ] ) , and @xmath184 and @xmath176 as in eqn .
( [ exp3 ] ) . in the second scenario ,
the @xmath189 have the values @xmath190 .
.values of the parameters that result from four different fits , for the exponential scenario . in this table ,
fit 1 means that only @xmath0 is included in the fit ; fit 2 means @xmath0 and @xmath191 are included ; fit 3 means @xmath0 , @xmath191 and @xmath192 are included ; fit 4 means @xmath163 , @xmath193 , @xmath194 and @xmath2 are all included , and applies only to decays with @xmath26 s in the final state .
[ fitparame ] [ cols="^,^,^,^,^",options="header " , ] given the failure of the exponential scenario to explain the @xmath195 data , one might be tempted to discard its predictions for the dileptonic decays .
however , we see that the integrated rates are , for the most part , quite similar to those predicted in the multipolar scenario .
the spectra that result from the two scenarios are very different , however , and the predictions for the relative amount of longitudinally polarized and transversely polarized @xmath26 produced are also somewhat different .
the exponential forms predict @xmath196 , while in the multipolar scenario , the ratio ranges between 0.15 and 0.24 .
the moral here may be that the exponential forms may be adequate for predicting total rates , but not decay spectra , nor polarization observables .
we limit our discussion to the multipolar scenario in what follows . 0.25 in 0.25 in the predictions for the process @xmath197 are two to three orders of magnitude smaller than present experimental upper limits , while those for @xmath198 are smaller than the experimental limits by factors of four to six .
we point out , however , that the calculated rates for these last two processes do not include possible contributions from charmonium resonances , which will certainly alter the shape of the lepton spectrum , and should also increase the total decay rate .
an investigation of this effect will be left for a future article . however , in at least one experimental analysis , kinematic cuts are imposed on the total mass of the lepton pair , so that events that may arise from either of the first two vector charmonium resonances are excluded @xcite . in any case , our results suggest that the exclusive dileptonic decay to the @xmath26 should be observed in the near future .
the results that we have obtained here again illustrate that the fit to the present @xmath15 semileptonic spectra alone is inadequate for providing information on @xmath187 processes .
even when we include the nonleptonic decays , the predictions for different decay modes ( particularly @xmath199 with longitudinally polarized @xmath26 s ) are sensitive to the nonleptonic modes we include in the fit . if any of these predictions are to be taken seriously , we would suggest that most attention be paid to the predictions of fit 4 in the multipolar scenario , as this is the only scenario that adequately describes all of the data available .
one of the features of the predicted spectra are the minima in the differential decay rates .
these minima are the result of zeroes in the respective helicity amplitudes , and the question of whether or not these zeroes do indeed exist , and of their exact locations , will have to await a @xmath27-factory .
however , long - distance effects , such as those that arise from charmonium resonances , or even from the charmonium continuum , will at least alter the positions of the zeroes , and may wash out the effect altogether .
we have used the scaling predictions of hqet , together with the most recent data on @xmath0 semileptonic decays , to extract the form factors that describe the @xmath15 and @xmath14 processes .
the latter we have applied to other processes , namely the nonleptonic decays @xmath200 and @xmath192 , as well as the rare decays @xmath11 and @xmath3 . in the case of
the nonleptonic decays , we have assumed factorization of the transition amplitude is valid .
we have also performed simultaneous fits of the semileptonic , nonleptonic and rare processes , and have found that hqet , together with factorization , provide an adequate framework for describing the observations .
our predictions for the modes @xmath3 suggest that these should be measurable in the next generation of experiments , and certainly at the proposed @xmath27 factory .
perhaps the greatest shortcoming of our fit procedure lies in how we handle the data , or simulated data , for the semileptonic decays .
ideally , we should have attempted to fit our choices of form factors to the experimentally measured differential decay rates . as a second choice
, our choices of form factors should have been input into the experimental monte - carlo programs to obtain the fit parameters . in any case , it is clear that the present data , particularly in the @xmath163 mode , are inadequate to sufficiently constrain the form factors .
in addition , the differences in form factors between fit 1 and fits 2 , 3 and 4 are quite striking .
the scenario that best describes all of the experimental data is the multipolar one , and in this scenario , we find that the universal form factor @xmath176 is linear in @xmath96 .
using this scenario , we predict @xmath201 and @xmath202 .
these numbers are consistent with other model calculations @xcite .
we also predict @xmath203 in @xmath204 to be @xmath205 .
to fully constrain the predictions of hqet , information on the form factors @xmath31 and @xmath30 is needed from the semileptonic decays .
such information can only be obtained from high - precision measurements of the decay spectra at low values of @xmath17 , particularly for semileptonic decays to muons , as well as by measuring the polarization of the charged lepton , again preferably the muon .
in addition , the precision and statistics in the @xmath17 spectrum must be improved so that the form factor parameters can be extracted from the data , particularly for the decays @xmath163 .
perhaps the ideal experiment would be the equivalent of present cleo experiments , in which the machine is tuned to be a source of @xmath206 pairs , produced from the strong decays of the @xmath207 . for @xmath51 decays
, the equivalent would be to produce a copious number of @xmath208 s , which can be realized at the proposed tau - charm factory .
we gratefully acknowledge helpful conversations with n. isgur , particularly for some very useful discussions on scaling relations .
we also thank c. carlson , a. freyberger , j. goity and k. protasov for discussions .
w. r. acknowledges the support of the national science foundation under grant phy 9457892 , and the u. s. department of energy under contracts de - ac05 - 84er40150 and de fg05 - 94er40832 .
w. r. also acknowledges the hospitality and support of institut des sciences nuclaires , grenoble , france , where much of this work was done , and of centre international des etudiants et stagiaires . | we examine the measured rates for the decays @xmath0 , @xmath1 and @xmath2 in a number of scenarios , in the framework of the heavy quark effective theory .
we attempt to find a scenario in which all of these decays are described by a single set of form factors .
once such a scenario is found , we make predictions for the rare decays @xmath3 .
while we find that many scenarios can provide adequate descriptions of all the data , somewhat surprisingly , we observe that two popular choices of form factors , namely monopolar forms and exponential forms , exhibit some shortcomings , especially when confronted with polarization observables .
we predict @xmath4 and @xmath5 .
we also make predictions for polarization observables in these decays .
= 10000 /#1#1 -0.5em / |
the conventional framework for particle physics beyond the standard model ( sm ) assumes that the fundamental mass scale of nature is the planck mass : @xmath15 gev .
it is then natural to ask : why are the masses of the elementary particles so small ? proposed solutions to this hierarchy problem have a common feature : new non - perturbative gauge interactions dynamically generate a much lower scale , @xmath16 , from which electroweak symmetry breaking is generated , and hence all the masses of the known elementary particles .
schematically , this mass hierarchy is @xmath17 in supersymmetric theories , @xmath16 is the scale at which supersymmetry is broken , and the triggering of electroweak symmetry breaking may be mediated , for example , by gravitational - scale physics , or by gauge interactions at much lower energy scales .
alternatively , @xmath16 may be the scale of a new gauge force , technicolor , which forms fermion condensates that directly break @xmath9 .
finally , new strong gauge forces could bind a composite higgs boson .
recently an alternative framework has been proposed @xcite in which spacetime is enlarged to contain large extra compact spatial dimensions . at distances smaller than the size of these extra dimensions
the gravitational force varies more rapidly than the inverse square law , so that the fundamental mass scale of gravity can be made much smaller than @xmath18 .
the conventional mass hierarchy of ( [ eq : masshier ] ) is completely avoided if this fundamental mass scale is of order the weak scale . in this case , the length scale of the extra dimensions is much larger than the scales probed experimentally at colliders , and hence this framework requires that the quarks , leptons and gauge quanta of the sm are spatially confined to a @xmath19 dimensional sub - space of the enlarged spacetime . the physics at the fundamental scale , @xmath0 , which may well be that of string theory , will be directly accessible to colliders of sufficiently high energy ; but even at lower energies this physics may be experimentally probed . at energies below the fundamental mass scale ,
physics is described by an effective lagrangian , which we take to be the most general set of @xmath20 invariant operators involving quark , lepton and higgs doublet fields of the sm : _ eff = _ sm + _
i c_i ^p _ i^4+p [ eq : leff ] where @xmath21 is the sm lagrangian , @xmath22 runs over all gauge invariant operators , @xmath23 , of dimension @xmath24 with @xmath25 , and @xmath26 are unknown dimensionless couplings .
in this letter we study consequences of several of the dimension-6 operators .
first we derive bounds on the @xmath27 from existing experimental results under very conservative assumptions about flavor - breaking in the ultraviolet theory .
we then re - examine the precision electroweak bounds on the higgs boson mass .
analyses within the standard model find a light higgs ; however , we will show that such results do not survive the addition of non - renormalizable operators , even if those operators are suppressed by scales as large as @xmath28 . in theories with large extra dimensions there is no good argument for a light higgs over a heavy higgs or a non - linearly realized @xmath29 symmetry , in which case ( [ eq : leff ] ) must be replaced by a chiral lagrangian .
finally we examine two operators in particular and their effects on the discovery of higgs bosons : _ g & = & ^ g^a_g^a [ eq : hop1 ] + _ & = & ^ f_f^ [ eq : hop2 ] where @xmath30 and @xmath31 are the qcd and electromagnetic field strengths , and @xmath7 is the higgs doublet with re@xmath32 .
the first operator contributes to higgs production at hadron colliders via gluon - gluon fusion , and the second to higgs decay to @xmath13 .
there are two reasons why these effects provide a significant discovery potential for extra dimensions : first , they are competing against a sm signal which is suppressed by loop factors , and second , the sm @xmath33 is further suppressed by @xmath34 , where @xmath35 is the electromagnetic coupling constant .
however we assume that the physics at scale @xmath0 which generates ( [ eq : hop1])([eq : hop2 ] ) , does so in a way that the coefficients are not suppressed by powers of the sm gauge coupling constants ( see also @xcite ) .
such a behavior is certainly _ not _ expected if the theory at @xmath0 is a 4-dimensional gauge field theory : in that case operators of the form ( [ eq : hop1])([eq : hop2 ] ) would arise by integrating out heavy fields , but these fields must couple to @xmath12 and @xmath11 with the usual sm gauge couplings , and further , as shown in @xmath36 , they must be also be loop - suppressed .
thus even if the gauge theory at @xmath0 were strongly - coupled , it seems unlikely that coefficients of @xmath37 could be generated .
this is very important the effect of the interaction @xmath38 on the @xmath39 branching ratio has been studied , and is small for @xmath40 @xcite .
thus observation of the physics we will describe in section [ sec : expt ] would provide support for an extra - dimensional theory .
are the coefficients @xmath41 expected to be large enough for an observable @xmath39 signal ? in general this can not be excluded , since physics induced by operators @xmath42 will place bounds on ( f_i=1 ) not on @xmath0 .
however , it would be unreasonable to expect @xmath43 to be orders of magnitude larger than all the other @xmath26 .
it is tempting to assume that although the dimensionless coefficients @xmath26 are unknown , they are all of order unity .
however , in this case operators which violate baryon number constrain @xmath44 , and @xmath45 violating operators contributing to @xmath46 constrain @xmath47 . thus the framework of large compact extra dimensions , allowing a fundamental scale close to the weak scale , is clearly excluded unless the low energy effective theory possesses an approximate flavor symmetry , in which case one expects c_i = _ fi c_i [ eq : cprime ] with @xmath48 of order unity .
the flavor symmetry breaking parameters , @xmath49 , depend on the flavor symmetry group and the pattern of flavor symmetry breaking . for operators which violate flavor and @xmath45 they must be small , while for operators which conserve flavor and @xmath45 they may be set to unity . to allow low values for @xmath0 ,
the flavor group should be large , and its breaking should be kept to a minimum , consistent with the observed quark and lepton masses and mixings .
the maximum flavor group of the sm is @xmath50 .
the three generations of quarks and leptons transform as @xmath51 ; @xmath52 ; @xmath53 ; @xmath54 ; @xmath55 .
if there are only three symmetry breaking parameters , one for each of the up , down and charged lepton mass matrices , @xmath56 ; @xmath57 ; @xmath58 , then baryon number and lepton number remain unbroken .
( the @xmath59 are equal to the yukawa couplings up to an @xmath37 factor , @xmath26 : @xmath60 .
) however , even after imposing such a flavor symmetry , there remain operators such as _ qq = ( |q_l ^_u _ u^q_l)^2 = c_u^4(|q_l^_u_u^q_l)^2 [ eq : oqq ] which contribute to @xmath46 and constrain @xmath61 .
there are two ways to avoid this bound .
first , since the bound depends quadratically on @xmath62 , values slightly larger than 1 will weaken the bound significantly ; this seems entirely natural to us .
second , one could postulate that @xmath63 are real and the observed @xmath46 has an exotic origin ; we view this as disfavored given that measurements of @xmath64 and @xmath65 mixing indicate values of the ckm matrix elements consistent with a standard model origin of @xmath46 to better than @xmath66 . for the @xmath39 signal ,
we are interested in the operators ( [ eq : hop1])([eq : hop2 ] ) , which conserve @xmath50 .
hence , even if the higher dimension flavor violating operators , such as ( [ eq : oqq ] ) , are completely absent , it is important to study constraints on @xmath0 expected from operators which conserve @xmath50 .
such operators include flavor - conserving four - fermion operators and operators involving the higgs doublet and the gauge fields .
there have been many analyses to date which obtain constraints from these operators , and here we will simply repeat the results of these analyses , in the notation we are using for @xmath0 .
( an analysis similar to ours was recently presented in @xcite . ) among the @xmath45-conserving four - fermion operators , the strongest constraints come from atomic parity violation .
the operator @xmath67 gives a constraint @xmath68 @xcite at 95% cl .
if the operator @xmath69 were generated with the same coefficient , @xmath70 would be preserved in atomic systems and the previous limit would vanish . although we do not expect @xmath70 to be a good symmetry of the underlying theory , a partial cancellation could easily weaken this bound . apart from @xmath70-violation , the best bounds on @xmath71 currently come from opal @xcite , using the @xmath72 component , and from cdf @xcite , using the @xmath73 component .
both find @xmath74 at 95% cl .
the bounds on the coefficients of the operators @xmath75 of ( [ eq : oqq])([eq : olq ] ) do not provide strict bounds on the scale @xmath0 , because @xmath76 , and the @xmath26 are unknown .
nevertheless , if the ( flavor - conserving ) @xmath77 are expected to be of order unity for these operators , then @xmath78 is clearly allowed , while a value of @xmath0 as low as @xmath79 seems disfavored .
a second class of constraints arise from precision measurements in the electroweak gauge sector , namely from the @xmath80 and @xmath81 parameters ( see , , @xcite ) .
the strongest of these constraints arise from the operators : _ bw&= & b^ ( ^^a w^a ) [ obw ] + _ & = & ( ^d^ ) ( d_^ ) [ op1 ] which contribute s_new & = & -f_bw + t_new & = & -f _ where @xmath82 are the sine and cosine of the weak angle and @xmath83 , @xmath84 are unknown signs . a global fit to electroweak observables @xcite yields @xmath85 and @xmath86 assuming @xmath87 . and defines @xmath88 in the sm .
we rescale to @xmath87 using the parameterization of ref .
@xcite ( see eqs .
( [ ds])([dt ] ) ) .
we then treat deviations from @xmath87 as `` new physics . ''
] since each operator contributes only to one of @xmath80 or @xmath81 , we can find independent bounds on each .
we find that at 95% cl : _ bw & > & 3.6[bwbound ] + _ & > & 3.0 .
we can also extract a bound if @xmath89 : @xmath90 , allowing the higgs mass to vary over the range @xmath91 .
we see that the constraints from precision electroweak physics are very similar in magnitude to those obtained in the previous section .
how important are these constraints for restricting @xmath92 ?
although the electromagnetic field strength , @xmath93 , is not @xmath94 invariant , the operator @xmath95 is generated , after electroweak symmetry - breaking , from the invariant operators @xmath96 , @xmath97 and @xmath98 of eq .
( [ obw ] ) : = c_w^2+s_w^2 + c_ws_w .
[ foper ] all @xmath99 and @xmath100 on the right side of eq .
( [ foper ] ) were equal , then the bound ( [ bwbound ] ) on @xmath101 implies @xmath102 .
however , changes in the relative signs or sizes of each contribution significantly reduces the bound ; thus we have no strong lower bound on the scale @xmath92 itself .
likewise we know of no strong constraint on the scale @xmath103 either . finally we wish to address the question of the higgs mass .
it is well - known that fits to the electroweak data indicate a light higgs .
a simple fit can be done using only @xmath80 and @xmath81 as given above and the following parameterization of the higgs contributions from ref .
@xcite : s_h&=&0.091 x_h - 0.010 x_h^2 [ ds ] + t_h&=&-0.079 x_h-0.028 x_h^2 + 0.0026 x_h^3 [ dt ] where @xmath104 . using these forms ,
one can do a fit demanding @xmath105 and likewise for @xmath81 . for the sm alone , a 95% cl
upper bound of @xmath106 has been obtained @xcite .
however it is clear that from the point of view of the oblique parameters , shifts in @xmath107 and @xmath108 can be compensated by similar shifts in @xmath109 and @xmath110 .
thus we can derive an effective `` 95% cl bound '' on the higgs mass as a function of @xmath0 under the requirement that the fit to the experimentally obtained @xmath111 and @xmath112 be no worse than that obtained for @xmath113 and @xmath114 .
( we do this by constructing a @xmath115 distribution from @xmath80 and @xmath81 alone . )
how large can the higgs mass become with the inclusion of @xmath98 and @xmath116 ?
the answer is : quite large .
fitting to @xmath117 as a function of @xmath0 and using @xmath80 and @xmath81 as `` experimental '' inputs , we find for particular choices of the signs of the operators ( , @xmath118 ) that _ the precision electroweak bound on the higgs mass disappears completely for @xmath119 ! _
( by `` disappear ''
we mean that the 95% upper bound on @xmath117 exceeds the unitarity bound of approximately @xmath120 and so is meaningless . ) thus , in the context of gravitational physics at or below @xmath121 , the usual claims that electroweak physics prefers a light higgs do not hold . and even for @xmath0 as high as @xmath122 , the upper limit on the higgs mass exceeds @xmath123 .
these results are summarized in figure [ massfig ] where we show the 95% cl allowed range for @xmath117 as a function of @xmath124 .
the hatched region at small @xmath0 is ruled out because of its large contribution to @xmath80 and @xmath81 , while the region at large @xmath0 and large @xmath117 is ruled out because the new operators contribute too little to @xmath80 and @xmath81 to significantly effect the sm fit to the higgs mass .
however for intermediate @xmath0 ( unhatched region ) it is clear that there is effectively no limit on the higgs mass thanks to the effects of the new operators .
= 4truein ( if the physics at @xmath0 were weakly - coupled then we would expect that @xmath125 ; then allowing @xmath126 would reproduce fig . [ massfig ] , only with the @xmath0 rescaled by @xmath127 .
thus the preference for a light higgs in the sm is even removed for a weakly - coupled gauge theory if @xmath128 . ) finally , we note that the one other argument for a light higgs , namely triviality , is no longer applicable in these models either . with such a low ultraviolet cutoff ( @xmath129 few tev ) , the higgs self - coupling can not run to its landau pole for @xmath130 .
the mechanism for electroweak symmetry breaking ( ewsb ) is unknown .
nevertheless , it is commonly believed that the higgs boson exists , and is light .
the two indirect indications for this are : * the successful prediction of the weak mixing angle from gauge coupling constant unification .
this prediction results in theories with weak scale supersymmetry which are perturbative to a high scale ; such theories have a light higgs boson , @xmath131 @xcite . *
the experimental values of the precision electroweak observables are consistent with the standard model , at 95% c.l .
, only if @xmath132 @xcite .
if there are large extra dimensions allowing the fundamental scale , @xmath0 , to be in the tev domain , neither of these points can be used to argue that the higgs boson is light . for the first
: it has not been demonstrated that it is possible to predict the weak mixing angle to the percent level of accuracy in these theories ; furthermore , there is no need for the field theory below @xmath0 to be supersymmetric since there is no large hierarchy between the weak scale and @xmath0 .
the argument from fits to the precision electroweak observables applies only if the standard model is the correct theory up to scales of at least 10 tev ; it is a very weak bound which is immediately evaded by large extra dimensions , allowing several scenarios for ewsb : * _ light higgs _ ( @xmath133 ) : for @xmath134 some protection mechanism for the higgs mass would be required ; if this is supersymmetry , the higgs will be light . for @xmath135 ,
if the tree level higgs mass happened to vanish , ewsb and a light higgs boson could result from 1 loop radiative corrections . *
_ heavy higgs _ ( @xmath136 ) : this could arise for @xmath137 tev , if the higgs mass parameter is somewhat less than @xmath0 , or alternatively for @xmath138 if the higgs mass parameter vanishes at tree level but arises at 1 loop . in both cases a large value for the higgs self coupling is needed , and the operators ( [ obw ] ) and ( [ op1 ] ) must mimic the effects of a light higgs in the @xmath80 and @xmath81 parameters . * _ no higgs _ : physics at the fundamental scale @xmath139 may itself cause ewsb .
an example of this has already been proposed @xcite . in this case
the theory below @xmath0 will have @xmath9 realized non - linearly , and the chiral lagrangian will have operators analagous to ( [ obw ] ) and ( [ op1 ] ) which mimic the effects of a light higgs in the @xmath80 and @xmath81 parameters .
a light higgs boson is just one possibility amongst several for ewsb , and is not preferred .
we have shown that , in theories with large extra dimensions having @xmath140 with @xmath141 of order unity , the precision electroweak data provide a lower bound on the fundamental scale , @xmath142 . for values of @xmath0 in the range ( 13)@xmath143 , the signs @xmath144 are critical . for two sign choices , no successful fit can be found for any higgs mass . for a third choice
, a good fit to the data is found for higgs masses all the way up to @xmath145 .
for the final choice , masses up to @xmath120 are also obtained , though the fits are less convincing . only in the case of very large @xmath0
does the data still prefer a light higgs , but then the quadratic finetuning of the light higgs mass to one part in @xmath146 is reintroduced . in view of the bounds on @xmath147 of @xmath148 from each of @xmath149 ( [ eq : oqq]),@xmath150 ( [ eq : olq ] ) , @xmath98 ( [ obw ] ) , and @xmath116 ( [ op1 ] ) , it may be felt that the exciting possibility of @xmath0 in the @xmath151 range is unlikely .
why would all the relevant @xmath26 coefficients be small ?
one possibility is that the dominant interactions of the new physics at @xmath0 preserve symmetries that are broken by the electroweak gauge interactions , including @xmath70 , @xmath45 and custodial @xmath5 .
if these symmetries are broken by sub - dominant interactions at @xmath0 , then the smallness of the relevant @xmath26 can be naturally explained .
for the case that there is a higgs boson , either light or heavy , we now study the effects of @xmath152 of ( [ eq : hop1])([eq : hop2 ] ) on the signal for @xmath39 at hadron colliders .
these operators have two immediate consequences .
first , when both higgs fields are set to their vacuum expectation values ( vev s ) , the gauge couplings of qed and qcd are shifted .
but these shifts can be reabsorbed into the definition of the gauge couplings and therefore have no observable implications .
( if one attempts to unify the sm gauge couplings at some ultraviolet scale , or otherwise define theoretical relations among them , then these shifts will enter into the relation between the theoretical couplings and those extracted from data . however , for all but the lightest @xmath0 , this shift is smaller than the experimental uncertainties . ) the second consequence is the possibility of unusual production and decay modes of the ( physical ) higgs bosons . taking one of the higgs fields to its vev , one obtains terms in the effective lagrangian : _ eff = + f_hf_f^ + f_g hg^a_g^a + [ eq : hleff ] where @xmath153 is the physical higgs boson , @xmath154 and @xmath155 are unknown signs .
first , @xmath156 can contribute to the gluon fusion process @xmath157 .
it is well - known that the dominant production mode for higgs bosons at the tevatron and the lhc is through gluon fusion , via a loop of @xmath158-quarks .
because the process occurs at one - loop , non - renormalizable operators are more likely to provide a significant correction to the cross - section . integrating out the @xmath158-quark , the relevant low - energy operator is then ( for a recent discussion of the relevant sm higgs physics , see @xcite ) : _ g , eff=(-i_g+f_g ) hg^a_g^a where @xmath159 is the su(2 ) coupling constant and @xmath160 for @xmath161 @xmath162 . for @xmath163 , the new physics will actually dominate the production of higgs bosons .
note that the cross - section is maximized for constructive interference , @xmath164 , and minimized for @xmath165 .
the operator @xmath95 does not contribute to higgs production could turn the nlc into an @xmath166-channel higgs factory when run in @xmath167 mode . ] .
however it can contribute to the decay of the higgs into photons : ( h)= for @xmath168 . in the sm , this process is dominated by loops of @xmath169-bosons and @xmath158-quarks . integrating them out yields an effective operator : _
, eff=(-i_+ f _ ) hf_f^ where @xmath170 varies from roughly @xmath171 to @xmath172 as @xmath117 is varied . once again
, the new physics will dominate the width for @xmath173 given @xmath174 . if @xmath175 , its decay width is dominated by final state @xmath176-quarks
; then @xmath173 becomes the dominant decay mode given @xmath177 .
however , even for larger @xmath92 , the branching ratio @xmath173 may be more than sufficient to provide a strong signal .
the signal is maximized for @xmath178 ( , constructive interference of the sm and new physics ) and minimized for @xmath179 .
( in the context of lep , ref .
@xcite recently examined the effect of @xmath95 and related operators on @xmath180 and found sensitivity there to new physics roughly below a scale @xmath181 . ) unfortunately , the operator @xmath156 can also contribute to the higgs decay width via @xmath182 which is unobservable among the qcd backgrounds .
in fact , to lowest order , ( hgg)=8()^4 ( h ) . in the limit in which the new physics is dominating the higgs decays and @xmath183 , the @xmath182 decays suppress the branching ratio into @xmath173 by about a factor of 10 .
however , once final state @xmath184 dominate the higgs width , the decays to gluons provide no real additional suppression of the @xmath173 branching fraction .
finally we note that the interference of @xmath156 with the sm gives simultaneously larger ( smaller ) higgs cross - sections and larger ( smaller ) @xmath185 .
the sensitivity of any experiment to new physics in the higgs channel is then a function of several variables : @xmath117 , @xmath186 , @xmath187 , @xmath92 and @xmath103 .
there are four sign choices for @xmath188 ; we choose to study the two cases which maximize / minimize the signal at current and future colliders .
the maximum signal case has @xmath178 and @xmath164 ; we checked that over the entire range of interest the increase in the cross - section implied by @xmath164 more than offset the corresponding increase in @xmath189 .
the minimum signal case has the opposite choice of both signs .
our analysis then has two parts .
first we ignore the @xmath156 operator ( , @xmath190 ) and determine the sensitivity of current and future experiments to new physics through @xmath95 alone . in this case , the production cross - section is simply that of the sm . then in a second analysis we include both @xmath95 and @xmath156 . as we already noted , the effect of @xmath156 is both to enhance the production but also to diminish the relative branching ratio of @xmath173 . for the purposes of doing the numerical calculations , we have used ( in a greatly modified form ) the programs of m. spira and collaborators @xcite . in all cases , we will work only to leading order . in the sm
it has been found that nlo qcd corrections can change the cross - sections and decay widths by @xmath191 @xcite .
naively such changes appear to correspond only to @xmath192 shifts in @xmath0 , which are too small for the physics we are interested in here .
however , it is possible that interference effects and enhanced backgrounds ( , @xmath173 in the sm ) could produce a larger effect we will not consider that possibility here . throughout our analysis
we also have to address issues of acceptances and backgrounds in an approximate manner . in run
i , cdf reported an efficiency times acceptance approaching 15% in inclusive @xmath193 higgs searches @xcite ; we will assume that this figure prevails at all future facilities .
there are also two major sources of backgrounds for our @xmath167 signal : sm processes which produce or fake @xmath167 , and the usual sm decay of @xmath173 itself .
the latter can be calculated explicitly .
for the former we estimate by fitting to the cdf background spectrum @xcite , appropriately scaled to the luminosity of future tevatron runs , or the atlas background spectrum @xcite appropriately scaled for lhc runs . in figures [ fig1](a)-(b )
we show the sensitivity to @xmath92 that can be obtained at various machines by plotting their @xmath194 discovery reaches ( with no @xmath156 contribution ) .
the colliders shown are : the tevatron with @xmath195 and @xmath196 of luminosity ( run i ) , with @xmath197 and @xmath198 of luminosity ( run ii ) , with @xmath197 and @xmath199 ( a proposed run iii ) , and the lhc with @xmath200 and @xmath201 ( initial luminosity ) and @xmath202 ( final luminosity ) respectively .
( note that the tev run i line falls below the region of parameter space plotted . ) as one expects , once the @xmath203 threshold opens up at @xmath204 , the large @xmath205 is sufficient to overwhelm the photonic width and our experimental sensitivity drops significantly .
nonetheless , given the possibility of a light higgs ( and the robust arguments for one in supersymmetric frameworks ) experimentalists should be encouraged to view @xmath173 as a viable and potentially large signal .
= 6truein in terms of extracting a conservative discovery reach for @xmath0 , figure [ fig1](b ) should be used since it chooses @xmath186 in order to minimize the signal .
we note , for example , that the data from run i can not presently probe ( or exclude ) @xmath0 above @xmath206 , but that run ii should have a reach of approximately 1
@xmath207 for a light higgs .
however it is important to realize that for generic @xmath186 , the various colliders may have reaches as high as those shown in figure [ fig1](a ) .
thus , for example , if the higgs mass is below the @xmath208 threshold , the lhc can possibly find a signal for @xmath0 up to @xmath209 for a light higgs !
( unfortunately , that scale could also be as low as @xmath210 . )
figures [ fig2](a)-(b ) repeat the same analysis , but now with @xmath156 included such that @xmath211 .
we view these results as more realistic compared to those above in which only the @xmath95 operator was kept .
we again show the same set of 5 collider options .
figure [ fig2](b ) is the conservative @xmath194 discovery reach , chosen to minimize the @xmath212 rate .
it is interesting that for a light higgs , the limits are slightly stronger than those obtained with @xmath190 ; now even the tevatron run i data has the ability to probe scales above @xmath206 .
however the more noticable difference is the ability to produce larger numbers of heavy higgs bosons and observe their @xmath167 decays .
for example , the lhc is capable of probing scales near @xmath213 even for @xmath214 .
= 6truein figure [ fig2](b ) shows the maximal reach of the various colliders , with the lhc now extending its sensitivity to @xmath0 as high as @xmath121 for a light higgs ! finally , we summarize a few of our results for @xmath215 , @xmath216 and @xmath123 for both exclusion and discovery in table [ table1 ] .
all bounds assume @xmath217 . for each choice of the higgs mass ,
we have shown a conservative limit on @xmath0 which can be excluded , and a maximum @xmath0 below which a signal may be discovered .
thus for the exclusion bounds ( @xmath218 ) we have taken the interference effects to minimize the signal ; for the maximum discovery reaches ( @xmath194 ) , we have chosen the interference effects to maximize the signal . .exclusion limits and maximum discovery reaches ( in tev ) for various collider runs for 3 representative higgs masses . the first column for each
@xmath117 is a conservative @xmath218 exclusion reach for each machine ; the second column is the optimistic @xmath194 discovery reach .
unfilled columns represent limits below @xmath206 .
we take @xmath219 for the table . [
cols="<,^,^,^,^,^,^ " , ] [ table1 ] we have attempted in this analysis to be rather conservative . for one thing , the @xmath218 exclusion limits of the various colliders are often several tev higher than the @xmath194 discovery limits .
secondly , we have treated the discovery of the @xmath173 signal as simply a counting experiment , throwing away useful experimental information , for example on the shape of the diphoton mass spectrum , which would be available experimentally to help extract the signal from the backgrounds .
lastly , we have not included qcd corrections to the amplitudes , which we believe could increase the signal ( though also increasing the `` background '' @xmath173 signal ) by @xmath220 .
therefore we believe that the reaches given here are to be taken as conservative values , insofar as one should take the scales deduced from naive power - counting seriously .
in this paper we have studied two consequences of large extra dimensions for electroweak symmetry breaking : a relaxation of the precision electroweak bound on the higgs boson mass , and an enhanced rate for @xmath13 events at hadron colliders from higgs decay .
the relaxation of the precision electroweak bound on the higgs mass applies when any new physics generates ( [ obw ] ) and ( [ op1 ] ) at a scale of several tev .
it is well known that @xmath80 and @xmath81 depend only logarithmically on the higgs boson mass , but it may not be appreciated that the mass bound can be evaded completely for a wide range of values of @xmath0 , extending as high as 10 tev . for example ,
even a weakened bound of @xmath221 , only applies if the standard model is the correct description of nature up to energies of @xmath122 .
we find this implausible , since it implies a fine tuning in the higgs mass squared parameter of 1 part in 2000 .
there is only one strong argument for a light higgs boson : the correct successful prediction of the weak mixing angle at the percent level of accuracy requires weak scale supersymmetry , and therefore a light higgs boson . in theories with large extra dimensions this argument is not applicable , since the percent level prediction for the weak mixing angle is lost .
hence , in these theories , there is no preference for a light higgs boson , and thus alternatives with a heavy higgs or no higgs should be considered seriously .
if there is a higgs boson , we have shown that a generic signal of large extra dimensions is an anomalously large @xmath167 signal at machines capable of producing higgs bosons .
expectations from the sm put such a signal out of reach of the tevatron . in figure [ fig2 ]
we showed the @xmath222 discovery reaches for @xmath223 at the tevatron and lhc . at run ii of the tevatron collider this signal would be discovered for a light higgs if @xmath0 is less than 2 ( 3 ) tev for destructive ( constructive ) interference .
lhc not only increases the discovery potential for a light higgs boson mass , up to @xmath121 for constructive interference , but also has significant discovery potential up to the largest higgs masses .
this signal compares favorably with that of graviton production at colliders @xcite , especially if the scale which sets the size of the @xmath224 dimensional gravitational coupling is somewhat larger than the scale @xmath0 .
we are grateful to nima arkani - hamed , michael chanowitz , savas dimopoulos and henry frisch for many useful conversations .
this work was supported in part by the u.s.department of energy under contract de ac0376sf00098 and by the national science foundation under grant phy9514797 . | if spacetime contains large compact extra dimensions , the fundamental mass scale of nature , @xmath0 , may be close to the weak scale , allowing gravitational physics to significantly modify electroweak symmetry breaking .
operators of the form @xmath1 and @xmath2 , where @xmath3 and @xmath4 are the @xmath5 and @xmath6 field strengths and @xmath7 is the higgs field , remove the precision electroweak bound on the higgs boson mass for values of @xmath0 in a wide range : @xmath8 . within this framework
, there is no preference between a light higgs boson , a heavy higgs boson , or a non - linearly realized @xmath9 symmetry beneath @xmath0 . if there is a higgs doublet , then operators of the form @xmath10 , where @xmath11 and @xmath12 are the qcd and electromagnetic field strengths , modify the production of the higgs boson by gluon - gluon fusion , and the decay of the higgs boson to @xmath13 , respectively . at run ii of the tevatron collider ,
a @xmath13 signal for extra dimensions will be discovered if @xmath0 is below 2.5 ( 1 ) tev for a higgs boson of mass 100 ( 300 ) gev .
furthermore , such a signal would point to gravitational physics , rather than to new conventional gauge theories at @xmath0 .
the discovery potential of the lhc depends sensitively on whether the gravitational amplitudes interfere constructively or destructively with the standard model amplitudes , and ranges from @xmath14 3 10 ( 2 4 ) tev for a light ( heavy ) higgs boson .
@=11 caption#1[#2]#3 @=12 .7ex .7ex gev mev ev kev tev m_z m_pl ^-1 ^-1 # 1#2#3nucl . phys . * b#1 * ( 19#2 ) # 3 # 1#2#3phys . lett . * b#1 * ( 19#2 ) # 3 # 1#2#3phys . lett .
* # 1b * ( 19#2 ) # 3 # 1#2#3phys . rev .
* d#1 * ( 19#2 ) # 3 # 1#2#3phys .
rev .
lett . *
# 1 * ( 19#2 ) # 3 # 1#2#3phys . rep . *
# 1 * ( 19#2 ) # 3 # 1#2#3ann .
rev .
. astrophys . *
# 1 * ( 19#2 ) # 3 # 1#2#3ann .
rev .
nucl . part .
sci . * # 1 * ( 19#2 ) # 3 # 1#2#3mod .
. lett . * # 1 * ( 19#2 ) # 3 # 1#2#3zeit . fr physik * c#1 * ( 19#2 ) # 3 # 1#2#3ap .
j. * # 1 * ( 19#2 ) # 3 # 1#2#3ann . phys . * # 1 * ( 19#2 ) # 3 # 1#2#3rev .
mod .
phys . *
# 1 * ( 19#2 ) # 3 # 1#2#3comm .
math .
phys . *
# 1 * ( 19#2 ) # 3 _ i.e. _ _ et al . _ _ e.g. _ _ etc
. _ _ c.f .
_ /#1/ |#1 # 1#1 ucb pth99/13 + lbnl43085 + hep - ph/9904236 + april 1999 + * electroweak symmetry breaking and large extra dimensions * + 0.5 cm _ department of physics + university of california + berkeley , ca 94530 , usa + and + theory group + lawrence berkeley national laboratory + berkeley , ca 94530 , usa + _ .5 cm |
the set of dyson
schwinger equations form a poincar covariant framework within which to study hadrons @xcite . in rainbow - ladder truncation , they have been successfully applied to calculate a range of properties of the light pseudoscalar and vector mesons , see ref .
@xcite and references therein .
the dse for the renormalized quark propagator @xmath1 in euclidean space is @xcite @xmath2 where @xmath3 and @xmath4 are the renormalized dressed gluon propagator and quark - gluon vertex , respectively .
the most general solution of eq .
( [ gendse ] ) has the form , renormalized at spacelike @xmath5 according to and with @xmath6 the current quark mass .
mesons are described by solutions of the homogeneous bethe
salpeter equation ( bse ) @xmath7 at discrete values of @xmath8 , where @xmath9 is the meson mass . in this equation ,
@xmath10 and @xmath11 are the outgoing and incoming quark momenta respectively , and similarly for @xmath12 .
the kernel @xmath13 is the renormalized , amputated @xmath14 scattering kernel that is irreducible with respect to a pair of @xmath14 lines . together with the canonical normalization condition for @xmath14 bound states , eq .
( [ hombse ] ) completely determines the bound state bethe
salpeter amplitude ( bsa ) @xmath15 .
different types of mesons , such as pseudoscalar or vector mesons , are characterized by different dirac structures . to solve the bse
, we use the ladder truncation , @xmath16 in conjunction with the rainbow truncation for the quark dse , eq .
( [ gendse ] ) : and , with @xmath17 . here
, @xmath18 is the free gluon propagator in landau gauge , and @xmath19 an effective quark - quark interaction , which reduces to the one - loop running coupling of perturbative qcd ( pqcd ) in the perturbative region .
this truncation preserves both the vector ward
takahashi identity ( wti ) for the @xmath20 vertex and the axial - vector wti , independent of the details of the effective interaction .
the latter ensures the existence of massless pseudoscalar mesons associated with dynamical chiral symmetry breaking ( d@xmath21sb ) in the chiral limit @xcite . in combination with an impulse approximation ,
the former ensures electromagnetic current conservation @xcite .
a momentum - dependent quark mass function @xmath22 is central to qcd . in the perturbative region
this mass function gives the one - loop perturbative running quark mass @xmath23\right)^{\gamma_m } } \ , , \label{eq : currentm}\end{aligned}\ ] ] with the anomalous mass dimension @xmath24 .
dynamical chiral symmetry breaking means that this mass function is nonzero even though the current - quark masses are zero . in the chiral limit
the mass function is @xcite @xmath25 \right)^{1-\gamma_m}}\ , , \label{eq : chiralm}\end{aligned}\ ] ] with @xmath26 the renormalization - point - independent vacuum quark condensate @xcite .
it is a longstanding prediction of dse studies in qcd that the dressed quark propagator receives strong momentum - dependent corrections at infrared momenta , see e.g. refs .
@xcite and references therein .
provided that the ( effective ) quark - quark interaction reduces to the perturbative running coupling in the ultraviolet region , it is also straightforward to reproduce the asymptotic behavior of eqs .
( [ eq : currentm ] ) and ( [ eq : chiralm ] ) @xcite .
both these phenomena are illustrated in the left panel of fig .
[ fig : quark ] . in this figure
one can also see that the dynamical mass function of the @xmath27 and @xmath28 quarks becomes very similar to that of quarks in the chiral limit in the infrared region .
this is a direct consequence of d@xmath21sb , and leads to a mass function of the order of several hundred mev for the light quarks in the infrared , providing one with a constituent mass for quarks inside hadrons , even though the corresponding current quark masses are only a few mev .
the dynamical quark mass function @xmath29 for different quark flavors ( left , adapted from ref .
@xcite ) and a comparison of the chiral - limit quark mass function with the lattice data @xcite ( right , adapted from ref .
@xcite).,title="fig : " ] the dynamical quark mass function @xmath29 for different quark flavors ( left , adapted from ref .
@xcite ) and a comparison of the chiral - limit quark mass function with the lattice data @xcite ( right , adapted from ref . @xcite).,title="fig : " ] these predictions were recently confirmed in lattice simulations of qcd @xcite .
quantitative agreement between the lattice simulations and the dse results for the quark propagator functions can be obtained within the rainbow truncation via a suitable choice for the effective quark - quark interaction @xcite .
pointwise agreement for a range of quark masses requires this interaction to be flavor - dependent @xcite , suggesting that dressing the quark - gluon vertex @xmath30 is important .
indeed , both lattice simulations @xcite and dse studies @xcite of this vertex indicate that @xmath30 deviates significantly from a bare vertex in the nonperturbative region .
a ( flavor - dependent ) nonperturbative vertex dressing could make a significant difference for the solution of the quark dse @xcite , as can be seen from the right panel of fig .
[ fig : quark ] .
the consequences of a dressed vertex for the meson bses are currently being explored @xcite and indications are that in the pseudoscalar and vector channels , the effects are small @xcite .
the meson spectrum contains three pseudoscalars with quantum numbers @xmath31 and masses below @xmath32gev : @xmath33 ; @xmath34 ; and @xmath35 . in a constituent - quark model
, these mesons are viewed as the first three members of a @xmath14 @xmath36 trajectory , where @xmath37 is the principal quantum number with the ground state @xmath38 ( the @xmath33 ) , and the others are its first two radial excitations , @xmath39 and @xmath40 .
the pseudoscalar trajectory is particularly interesting , because its lowest mass member is qcd s goldstone mode .
therefore an explanation should describe both chiral symmetry and its dynamical breaking as well as a correlation of the ground and excited states via an approximately linear radial regge trajectory .
the latter is easily realized in poincar invariant quantum mechanics @xcite but the former is not .
the chiral properties of qcd are manifest in the axial - vector wti , which reads @xmath41 with @xmath42 and @xmath43 the renormalized dressed axial - vector and pseudoscalar vertices , each satisfying an inhomogeneous extension of eq .
( [ hombse ] ) .
equation ( [ eq : avwti ] ) is an exact statement in qcd implying that the kernels of the quark dse , eq .
( [ gendse ] ) , and of the bses have to be intimately related . a weak coupling expansion of the dses yields perturbation theory and satisfies this constraint , but is not useful in the study of intrinsically nonperturbative phenomena .
however , a systematic and symmetry preserving nonperturbative truncation scheme exists @xcite , allowing for both elucidation and illustration of the consequences of the axial - vector wti .
pseudoscalar mesons appear as pole contributions to the axial - vector and pseudoscalar vertices at @xmath44 .
the residues of these poles are @xmath45 \ , , \\
i \rho_{\pi_n}\!(\zeta ) & = & z_4(\zeta)\,\ , \int\!\!\frac{d^4q}{(2\pi)^4 } { \rm tr}\left [ \gamma_5 \ ; s(q_+ ) \ , \gamma_{\pi_n}(q_+,q_-;p ) \ ,
s(q_-)\ , \right ] \,,\end{aligned}\ ] ] both of which are gauge invariant and cutoff independent .
it follows from eq .
( [ eq : avwti ] ) that these residues satisfy the following exact identity in qcd @xcite @xmath46 valid for every @xmath47 meson @xcite , irrespective of the magnitude of the current quark mass @xcite . for the @xmath38 , d@xmath21sb yields : @xmath48 and @xmath49 .
hence , the gell - mann oakes renner relation emerges as a corollary of eq .
( [ eq : gmorgen ] ) and the ground state pion is qcd s goldstone mode @xcite . for the @xmath50 pseudoscalar mesons one
has @xmath51 by assumption , and hence @xmath52 in the chiral limit . furthermore , the ultraviolet behavior of the quark - antiquark scattering kernel in qcd guarantees that @xmath53 is finite in the chiral limit .
hence , it is a necessary consequence of chiral symmetry and the axial - vector wti that @xmath54 vanishes in the chiral limit for all excited pseudoscalar mesons ( i.e. for all @xmath55 ) @xcite .
we now illustrate the exact results reviewed above in a model that both preserves qcd s ultraviolet properties and exhibits d@xmath21sb , namely the rainbow - ladder truncation of the set of dses . for the infrared behavior of the effective quark - quark interaction , @xmath19
, we employ an ansatz @xcite that is sufficiently enhanced in the infrared to produce a realistic value for the vacuum quark condensate of about @xmath56 .
the model parameters , along with the quark masses @xcite , are fitted to give a good description of the chiral condensate , @xmath57 and @xmath58 .
the obtained quark propagator functions agree qualitatively with lattice simulations .
l|ll|ll|ll|l|ll|ll @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & @xmath61 & @xmath65 & @xmath66 & @xmath61 & @xmath67 & @xmath63 + @xmath68 & 0.14 & 0.14 & 0.131 & 0.131 & 1.10 & 1.3 & -0.002 & 0.74 & 0.77 & 0.206 & 0.22 + @xmath69 & 0.70 & & 0.182 & & 1.41 & 1.47 & -0.033 & 1.08 & 1.02 & 0.257 & 0.23 + @xmath70 & 2.98 & 2.98 & 0.33 & 0.34 & 3.45 & 3.65 & -0.15 & 3.13 & 3.10 & 0.33 & 0.41 + the main results for the pseudoscalar and vector meson masses and leptonic decay constants are summarized in table [ tab : excpion ] and illustrated in fig .
[ fig : mesons ] . regarding the table
one should note that the rainbow - ladder truncation gives `` ideal '' flavor mixing and the study uses @xmath71 .
thus the first row simultaneously describes four degenerate mesons for each column , namely , a \{@xmath72 , @xmath73,@xmath74 } isotriplet and a @xmath75 isosinglet .
the second row describes @xmath76 mesons , and the third row @xmath77 mesons . for vector mesons ideal
mixing is a very good approximation , though this is not the case for the pseudoscalar ground states .
however , the experimental degeneracy of the @xmath34 and @xmath78 suggests @xcite that ideal mixing is almost realized for the excited pseudoscalar mesons .
this supports the interpretation of the @xmath78 and @xmath79 as the radial excitations of the @xmath80 and @xmath81 , with quark content almost entirely @xmath75 and @xmath76 respectively @xcite .
the masses @xmath82 of the pseudoscalar ground and first radially excited states ( left ) and the corresponding weak decay constants ( right ) as functions of the current - quark mass @xmath83 . for comparison
, we also include the mass of the ground - state vector mesons.,title="fig : " ] the masses @xmath82 of the pseudoscalar ground and first radially excited states ( left ) and the corresponding weak decay constants ( right ) as functions of the current - quark mass @xmath83 . for comparison , we also include the mass of the ground - state vector mesons.,title="fig : " ] figure [ fig : mesons ] shows bound - state masses and leptonic decay constants for the ground and excited states of pseudoscalar mesons as well as ground - state vector - meson masses as functions of the current - quark mass .
it illustrates that @xmath84 vanishes in the chiral limit , while @xmath85 and @xmath86 do not ; on the other hand , @xmath65 vanishes in the chiral limit , while @xmath62 does not . satisfaction of the axial - vector wti can be checked by virtue of eq .
( [ eq : gmorgen ] ) for any pseudoscalar meson in table [ tab : excpion ] . for the ground and excited states the inaccuracies are below 1% and 5% ,
the latter are larger because of : ( 1 ) the need to project out the ground state in order to calculate the excited state , ( 2 ) the smallness of @xmath65 close to the chiral limit , and ( 3 ) the larger bound - state mass , leading to a larger region of analytic continuation for the quark dse solution in the complex @xmath87 plane .
meson - meson - photon interactions can be described in generalized impulse approximation by @xmath88 \ , , \label{eq : generictri}\ ] ] where @xmath89 , @xmath90 , @xmath91 , and momentum conservation dictates @xmath92 . in eq .
( [ eq : generictri ] ) , @xmath93 is the dressed quark propagator with flavor index @xmath94 , and @xmath95 stands for a generic vertex function with incoming quark flavor @xmath96 and momentum @xmath97 , and outgoing quark flavor @xmath94 and momentum @xmath98 . depending on the specific process under consideration
, this vertex function could be a meson bsa or a quark - photon vertex .
the quark - photon vertex , , with @xmath99 the photon momentum and @xmath100 the quark momenta , is the solution of the inhomogeneous bse @xmath101 solutions of the homogeneous version of eq .
( [ verbse ] ) define vector meson bound states at timelike photon momenta .
it follows that @xmath102 has poles at these locations @xcite ( see also the vector - meson masses in table [ tab : excpion ] ) .
there are two diagrams that contribute to meson electromagnetic form factors : one with the photon coupled to the quark and one with the photon coupled to the antiquark respectively .
with photon momentum @xmath99 , and incoming and outgoing meson momenta @xmath103 , we can define a form factor for each of these diagrams @xcite @xmath104 we work in the isospin symmetric limit , and thus .
the @xmath105 and @xmath106 form factors are given by and , respectively .
left : the pion form factor @xmath107 , compared to the experimental results from refs .
right : the transition form factor @xmath108 together with experimental data @xcite.,title="fig : " ] left : the pion form factor @xmath107 , compared to the experimental results from refs .
right : the transition form factor @xmath108 together with experimental data @xcite.,title="fig : " ] our result for @xmath109 are shown in the left panel of fig .
[ fig : emff ] . in the timelike region , and in the spacelike region up to about @xmath110 , @xmath111 can be described very well by a monopole with our calculated @xmath112-mass , @xmath113 ( note that our calculated @xmath112-mass is slightly below the experimental value ) . above this value ,
our curve starts to deviate more and more from this naive vector meson dominance ( vmd ) monopole .
our result is in excellent agreement with the most recent jlab data @xcite ; it would be very interesting to compare with future jlab data in the 3 to 5 gev@xmath114 range , where we expect to see a significant deviation from the naive monopole behavior .
the pqcd prediction is significantly smaller than our results at 4 gev@xmath114 ; we anticipate that true perturbative behavior sets in somewhere between 10 and 20 gev@xmath114 @xcite .
recent lattice simulation also indicate that in the region between 0 and 2 gev@xmath114 the pion form factor can be represented by a vmd - like monopole @xcite .
this vmd - like behavior of the pion form factor in the spacelike region appears to be valid almost independent of the pion mass , both in our calculations and in the lattice simulations , though a monopole fit results in a vmd - mass which is slightly less than the actual vector meson mass .
current lattice simulations are not accurate enough , nor do they extend to large enough values of @xmath115 , to detect a significant deviation from vmd - like behavior above 2 gev@xmath114 .
l|lll|ll|ll|ll|ll & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 & @xmath121 & @xmath120 & @xmath122 & @xmath120 & @xmath123 & @xmath120 + calc . &
0.44 & 0.38 & -0.085 & 53 & 0.69 & 479 & 2.07 & 90 & 0.99 & 130 & 1.19 + expt . &
0.44 & 0.34(5 ) & -0.052(26 ) & 68 & 0.74 & 757 & 2.38 & 50 & 0.84 & 116 & 1.27 + also our results @xcite for @xmath124 agree quite well with the available experimental data , as do both the neutral and the charged kaon charge radius @xcite , see table [ tab : radiiradec ] .
it should be noted here that the @xmath106 form factor is obtained by taking the difference of two numbers , @xmath125 and @xmath126 , which are both close to one for @xmath115 near zero .
it is therefore much more sensitive to details of the model : both kaon loops and a flavor dependence of the ( effective ) quark - quark interaction will have a significantly bigger effect on the @xmath106 form factor than on the @xmath105 form factor .
this could explain why it does not agree as well with experiment as our other results .
in this respect one should note that the absolute deviation between our calculated charge radius and the experimental charge radius is similar for the charged and neutral kaon .
we can describe the radiative decay of the vector mesons using the same loop integral , eq .
( [ eq : generictri ] ) , this time with one vector meson bsa , one pseudoscalar bsa , and one @xmath127-vertex @xcite . the generic structure of a vector - pseudoscalar - photon vertex is @xmath128 where @xmath129 is the vector momentum and @xmath99 the photon momentum
. the on - shell value gives us the coupling constant , and can be used to calculate the partial decay width of the vector mesons . for virtual photons
, we can define a transition form factor @xmath130 , normalized to 1 at @xmath131 , which can be used in estimating meson - exchange contributions to hadronic processes @xcite . in the isospin limit ,
both the @xmath132 and @xmath133 vertices are given by @xmath134 the @xmath135 vertex is a factor of 3 larger , due to the difference in isospin factors ; however , the form factor @xmath130 is the same for @xmath136 and @xmath135 .
in contrast to the elastic form factors , this transition form factor falls off significantly faster than a vmd - like monopole , as can be seen in the right panel of fig .
[ fig : emff ] . only in the timelike region , near the vector meson pole , do we see a true vmd - like behavior . as eq .
( [ eq : rpgver ] ) shows , it is @xmath137 that is the natural outcome of our calculations . therefore , it is this combination that we give in table [ tab : radiiradec ] , together with the corresponding partial decay widths @xcite .
as anticipated , the partial decay width @xmath138 is indeed ( approximately ) nine times larger than the @xmath138 partial decay width .
note that part of the difference between the experimental and calculated decay width comes from the phase space factor because our calculated vector meson masses deviate up to 5% from the physical masses .
for the @xmath139 decays ( and corresponding form factors ) @xcite the situation is more complicated owing to the interference of the diagrams with the photon coupled to the @xmath69-quark and to the @xmath27- or @xmath28-quark . in the su(3 )
flavor limit , the charged @xmath140 vertex becomes equal to the @xmath136 vertex , whereas the neutral @xmath141 vertex is twice as large in magnitude @xmath142 in table [ tab : radiiradec ] we see indeed that the partial decay width of the neutral @xmath143 is larger than that of the charged @xmath143 , though not by a factor of four as it would in the su(3 ) flavor limit . furthermore , we see that the deviation between our calculation and experiment is largest for the charged @xmath143 decay . again , this can be understood since this decay is sensitive to the difference between the impulse diagrams , and therefore more sensitive to details of the model and its omissions .
we would like to thank craig roberts and peter tandy for useful discussions .
this work was supported by the austrian research foundation _
fwf , erwin - schrdinger - stipendium _ no.j2233-n08 , the department of energy , office of nuclear physics , contract no .
w-31 - 109-eng-38 , and benefited from the anl computing resource center s facilities ; part of the computations were performed on the national science foundation terascale computing system at the pittsburgh supercomputing center .
99 c. d. roberts and a. g. williams , _ prog .
nucl . phys . _ * 33 * , 477 ( 1994 ) ; c. d. roberts and s. m. schmidt , _ prog . part . nucl . phys . _
* 45s1 * , 1 ( 2000 ) ; r. alkofer and l. von smekal , _ phys .
_ * 353 * , 281 ( 2001 ) . p.
maris and c. d. roberts , _ int .
j. mod .
_ e * 12 * , 297 ( 2003 ) .
maris and c. d. roberts , _ phys .
_ c * 56 * , 3369 ( 1997 ) .
p. maris , c. d. roberts and p. c. tandy , _ phys .
_ b * 420 * , 267 ( 1998 ) . c. d. roberts , _ nucl .
_ a * 605 * , 475 ( 1996 ) .
p. maris and p. c. tandy , _ phys .
_ c * 62 * , 055204 ( 2000 ) ; p. maris and p. c. tandy , _ phys .
c * 61 * , 045202 ( 2000 ) . h. d. politzer , _ nucl .
b * 117 * , 397 ( 1976 ) .
k. higashijima , _ phys .
_ d * 29 * , 1228 ( 1984 ) .
i. fomin , v. p. gusynin , v. a. miransky and y. a. sitenko , _ riv .
nuovo cim . _
* 6n5 * , 1 ( 1983 ) . j. i. skullerud and a. g. williams , _ phys . rev .
_ d * 63 * , 054508 ( 2001 ) ; j. skullerud , d. b. leinweber and a. g. williams , _ phys .
_ d * 64 * , 074508 ( 2001 ) .
o. bowman , u. m. heller and a. g. williams , _ phys .
_ d * 66 * , 014505 ( 2002 ) .
j. b. zhang , p. o. bowman , d. b. leinweber , a. g. williams and f. d. r. bonnet _ phys .
_ d * 70 * , 034505 ( 2004 ) .
p. maris , a. raya , c. d. roberts and s. m. schmidt , _ eur .
j. _ a * 18 * , 231 ( 2003 ) .
m. s. bhagwat , m. a. pichowsky , c. d. roberts and p. c. tandy , _ phys .
_ c * 68 * , 015203 ( 2003 ) .
j. skullerud and a. kizilersu , _ jhep _ * 0209 * , 013 ( 2002 ) ; j. i. skullerud , p. o. bowman , a. kizilersu , d. b. leinweber and a. g. williams , _ jhep _ * 0304 * , 047 ( 2003 ) . m. s. bhagwat , a. holl , a. krassnigg , c. d. roberts and p. c. tandy , _ phys .
_ c * 70 * , 035205 ( 2004 ) .
m. s. bhagwat and p. c. tandy , arxiv : hep - ph/0407163 . c. s. fischer , f. j. llanes - estrada and r. alkofer , arxiv : hep - ph/0407294 ; f. j. llanes - estrada , c. s. fischer and r. alkofer , arxiv : hep - ph/0407332 . c. s. fischer and r. alkofer , _ phys .
_ d * 67 * , 094020 ( 2003 ) .
a. bender , c. d. roberts and l. von smekal , _ phys .
_ b * 380 * , 7 ( 1996 ) .
a. bender , w. detmold , c. d. roberts and a. w. thomas , _ phys .
_ c * 65 * , 065203 ( 2002 ) .
a. krassnigg , w. schweiger and w. h. klink , _ phys .
_ c * 67 * , 064003 ( 2003 ) .
a. krassnigg and c. d. roberts , arxiv : nucl - th/0308039 .
a. holl , a. krassnigg and c. d. roberts , _ phys .
c * 70 * , 042203 ( 2004 ) .
m. a. ivanov , y. l. kalinovsky and c. d. roberts , _ phys .
_ d * 60 * , 034018 ( 1999 ) .
p. maris and p. c. tandy , _ phys .
_ c * 60 * , 055214 ( 1999 ) .
a. hoell , a. krassnigg , c. d. roberts and s. v. wright , arxiv : nucl - th/0411065 . s. eidelman _ et al .
_ [ particle data group collaboration ] , _ phys . lett .
_ b * 592 * , 1 ( 2004 ) .
k. w. edwards _ et al . _
[ cleo collaboration ] , _ phys .
lett . _ * 86 * , 30 ( 2001 ) .
j. volmer _ et al .
_ [ the jefferson lab f(pi ) collaboration ] , _ phys .
lett . _ * 86 * , 1713 ( 2001 )
. s. r. amendolia _ et al . _
[ na7 collaboration ] , _ nucl .
_ b * 277 * , 168 ( 1986 ) .
maris and c. d. roberts , _ phys .
_ c * 58 * , 3659 ( 1998 ) .
j. van der heide , j. h. koch and e. laermann , _ phys .
_ d * 69 * , 094511 ( 2004 ) .
f. d. r. bonnet , r. g. edwards , g. t. fleming , r. lewis and d. g. richards arxiv : hep - lat/0411028 . s. r. amendolia _ et al .
_ , _ phys .
_ b * 178 * , 435 ( 1986 ) .
w. r. molzon _ et al .
_ , _ phys .
lett . _ * 41 * , 1213 ( 1978 ) [ erratum - ibid . *
41 * , 1523 ( 1978 errat,41,1835.1978 ) ] .
p. maris and p. c. tandy , _ phys .
_ c * 65 * , 045211 ( 2002 ) .
j. w. van orden , n. devine and f. gross , _ phys .
lett . _ * 75 * , 4369 ( 1995 ) .
p. c. tandy , _ prog .
_ * 39 * , 117 ( 1997 )
. r. i. dzhelyadin _ et al .
_ , _ phys .
b * 102 * , 296 ( 1981 ) [ _ jetp lett .
_ * 33 * , 228 ( 1981 ) ] .
p. maris , arxiv : nucl - th/0209048 . | two - body bound states such as mesons are described by solutions of the bethe salpeter equation .
we discuss recent results for the pseudoscalar and vector meson masses and leptonic decay constants , ranging from pions up to @xmath0 bound states .
our results are in good agreement with data .
essential in these calculation is a momentum - dependent quark mass function , which evolves from a constituent - quark mass in the infrared region to a current - quark mass in the perturbative region .
in addition to the mass spectrum , we review the electromagnetic form factors of the light mesons . electromagnetic current conservation is manifest and the influence of intermediate vector mesons is incorporated self - consistently .
the results for the pion form factor are in excellent agreement with experiment . |
the kondo effect in quantum dots has been studied in various kinds of systems in recent years , and the triple quantum dots with a triangular configuration is one of the interesting systems @xcite .
specifically , the closed path along the triangle causes a high - spin ground state with the nagaoka mechanism @xcite , and also an su(4 ) kondo effect , depending on the gate voltage . in this report
we study the triangular triple dot connected to two noninteracting leads , and present the results of the scattering phase shifts obtained with the numerical renormalization group ( nrg ) .
we also calculate the temperature dependence of the entropy and spin susceptibility @xcite , which show clearly the two - stage screening process of the high - spin nagaoka state and also the feature of the kondo screening in the su(4 ) case at half - filling .
we consider a three - site hubbard model on a triangle , which is connected to two non - interacting leads on the left ( @xmath1 ) and right ( @xmath2 ) , at @xmath3 and @xmath4 ( @xmath5 ) , respectively , as shown in fig .
[ fig : system ] ( a ) .
the total hamiltonian has the form @xmath6 , with @xmath7 here , @xmath8 @xmath9 is the hopping matrix element between the dots , @xmath10 the onsite energy in the dots , and @xmath11 the coulomb interaction .
the conduction electrons around the dots , @xmath12 , can tunnel into the triangle via @xmath13 , and it causes the level broadening @xmath14 with @xmath15 the density of states of the leads . in the limit of @xmath16 and @xmath17 ,
the circular motion along the triangle forms a single orbital at @xmath18 and two degenerate orbitals at @xmath19 .
this degeneracy brings interesting varieties to the kondo effect , occurring in the triangular quantum dot at different values of @xmath10 , which can be controlled by the gate voltage .
the low - energy states of the whole system including the leads show a local fermi - liquid behavior , which is characterized by the two phase shifts , @xmath20 and @xmath21 , for the quasi - particles with the even and odd parities .
specifically , at @xmath22 , the conductance @xmath23 and the local charge @xmath24 in the triangle can be expressed , respectively , in the landauer and friedel - sum - rule forms @xcite , @xmath25 owing to an inversion symmetry , the conductance @xmath26 for the current flowing along the horizontal direction in a four - terminal geometry , shown in fig . [
fig : system ] ( b ) , can also be obtained from these two phase shifts defined with respect to the two - channel model @xmath27 , as @xmath28 .
we have deduced the value of @xmath20 and @xmath21 from the fixed point of nrg @xcite .
furthermore , the quantum dots contribution of the free energy @xmath29 may be defined by @xcite , @xmath30 % , \quad % f_\mathrm{lead } = % -t\,\log \left [ \mbox{tr}\ , e^{-\mathcal{h}_\mathrm{lead}/t } % \right ] % % f_\mathrm{d } \
, \equiv \ , -t \left [ \ , % \log \left ( \mbox{tr}\ , e^{-\mathcal{h}/t } \right ) % \,-\ , % \log \left ( \mbox{tr}\ , e^{-\mathcal{h}_\mathrm{lead}/t } % \right ) \ , % \right ] % \end{aligned}\ ] ] one can also calculate the entropy @xmath31 and spin susceptibility @xmath32 . here , the magnetic field @xmath33 is introduced by replacing the energy levels @xmath10 and @xmath34 , respectively , by @xmath35 and @xmath36 .
( 110,20)(0,0 ) ( 8.0,7)(0,0)[bl](a ) ( 19,6)(1,0)20 ( 19,14)(1,0)20 ( 39,6)(0,1)8 ( 71,6)(1,0)20 ( 71,14)(1,0)20 ( 71,6)(0,1)8 ( 40.0,10)(2,0)4(1,0)1 ( 62.9,10)(2,0)4(1,0)1 ( 55,18.6 ) ( 50,10 ) ( 60,10 ) ( 51,10)(1,0)8 ( 49.5,10)(3,5)6 ( 60.5,10)(-3,5)6 ( 42.5,5)(0,0)[bl]@xmath13 ( 65,5)(0,0)[bl]@xmath13 ( 110,30)(-4,0 ) ( 4.5,13)(0,0)[bl](b ) ( 16.5,4.5)(1,0)20 ( 16.5,12.5)(1,0)20 ( 36.5,4.5)(0,1)8 ( 16.5,17.5)(1,0)20 ( 16.5,25.5)(1,0)20 ( 36.5,17.5)(0,1)8 ( 67.5,4.5)(1,0)20 ( 67.5,12.5)(1,0)20 ( 67.5,4.5)(0,1)8 ( 67.5,17.5)(1,0)20 ( 67.5,25.5)(1,0)20 ( 67.5,17.5)(0,1)8 ( 38.0,21)(2,0)5(1,0)1 ( 38.0,9.5)(2,0)5(1,0)1 ( 53.2,9.5)(2,0)7(1,0)1 ( 53.2,21)(2,0)7(1,0)1 ( 58.6,15 ) ( 50,10 ) ( 50,20 ) ( 50,10)(0,1)10 ( 51,10)(3,2)7 ( 51,20)(3,-2)7 ( 37.5,22.5)(0,0)[bl]@xmath37 ( 54,22.5)(0,0)[bl]@xmath37 ( 37.5,1)(0,0)[bl]@xmath37 ( 54,1)(0,0)[bl]@xmath37 for @xmath38 and @xmath39 .
left : the number of electrons @xmath40 in the triangle , the phase shifts @xmath41 and @xmath42 .
the vertical dashed lines correspond to the values of @xmath10 , at which @xmath40 changes discontinuously in the isolated limit @xmath43 .
right : series ( @xmath44 , @xmath23 ) and parallel ( , @xmath26 ) conductances . ] for @xmath38 and @xmath39 . left : the number of electrons @xmath40 in the triangle , the phase shifts @xmath41 and @xmath42 .
the vertical dashed lines correspond to the values of @xmath10 , at which @xmath40 changes discontinuously in the isolated limit @xmath43 .
right : series ( @xmath44 , @xmath23 ) and parallel ( , @xmath26 ) conductances . ]
in fig . [ fig : cond_u1 ] , the results of ( a ) the phase shifts @xmath20 , @xmath21 and local charge @xmath24 , and ( b ) the conductances @xmath23 and @xmath26 , are plotted as functions of @xmath10 .
the number of electrons in the triangle @xmath24 increases with decreasing @xmath10 , showing the steps for @xmath45 and @xmath46 .
it changes directly from @xmath47 to @xmath48 at @xmath49 without taking the step for @xmath50 .
this is caused by a property of the triangle : the ground state becomes degenerate for the three charge states @xmath51 , and @xmath46 at @xmath49 in the isolated limit @xmath52 .
the coulomb repulsion lifts partly the degeneracy due to the orbitals with the energy @xmath53 . specifically at the filling @xmath54 ,
a high - spin @xmath0 state becomes the ground state for @xmath55 by the nagaoka mechanism for the ferromagnetism . for finite @xmath56 , the conduction electrons tunneling from the two leads screen the @xmath0 moment at low temperatures to form a kondo singlet .
the series and parallel conductances through the singlet ground state show a clear difference in this region @xmath57 , as seen in fig .
[ fig : cond_u1 ] ( b ) .
we see that both of the partial waves contribute to the parallel conductance to give the plateau of @xmath58 , while the series conductance is suppressed @xmath59 by the interference .
correspondingly , the phase shifts take the values @xmath60 and @xmath61 . for small values of the electron occupancies @xmath62 , seen at @xmath63 , the phase shift @xmath20 for the even - parity partial wave increases for decreasing @xmath10 , while the odd one is almost zero @xmath64 .
it implies that in this region the electrons occupy mainly the orbital of @xmath65 that has an even parity , although generally each phase shift does not necessary correspond to the occupancy of the eigenstates with the same parity .
nevertheless we have confirmed , by calculating directly the occupation number of the each the partial waves @xmath66 and @xmath67 which satisfies @xmath68 , that an approximate relation @xmath69 holds well for each partial wave in the two regions , @xmath70 and @xmath71 .
however , such a separation of the friedel sum rule does not hold for the intermediate region @xmath72 : while each phase shift takes a constant value @xmath73 and @xmath74 , the occupation calculated directly has a different constant @xmath75 and @xmath76 .
these values suggest that , after two electrons occupy the orbital of the energy @xmath65 , the third electron occupies the degenerate orbitals of the energy @xmath53 in equal weights .
although the coupling with the two leads breaks the full triangular symmetry of the point group , it seems as if the charge distribution recovers effectively the degeneracy at @xmath77 . during the sudden change of the phase shifts at @xmath78 , the sum @xmath79 and
the parallel conductance @xmath80 change their values very little .
on the other hand , the phase difference @xmath81 , varies rapidly from @xmath82 to @xmath83 .
this rapid and continuous change of @xmath81 causes the narrow dip observed for the series conductance @xmath23 at @xmath78 in the middle of the conductance plateau . the dip is also related to the su(4 ) symmetry .
in contrast , the conductance at @xmath84 is caused by the usual su(2 ) kondo effect due to @xmath85 moment . in fig .
[ fig : s_tchi ] ( a ) the temperature dependence of the entropy @xmath86 and @xmath87 , which are obtained at @xmath88 in the middle of the charge step for @xmath54 , is plotted as a function of @xmath89 , where @xmath90 is the half - width of the conduction bands .
the results clearly show that the screening of the spin @xmath0 local moment due to the nagaoka mechanism occurs via two separate stages at @xmath91 and @xmath92 .
we see that the entropy takes the value @xmath93 at high temperatures for @xmath94 , and it means that the @xmath0 moment is free .
then , at intermediate temperature @xmath95 , the entropy becomes @xmath96 .
it shows that the half of the moment is screened by the conduction electrons from one of the channel degrees , and the local moment is still in an under - screened state .
the full screening is completed finally at very low temperatures @xmath97 , as we can seen also in the behavior of @xmath98 .
similarly , the entropy and the spin susceptibility for @xmath99 , just at the dip of @xmath23 for @xmath100 , are plotted in fig .
[ fig : s_tchi ] ( b ) . at the high temperatures @xmath101
, the entropy takes a constant value @xmath102 , which implies that there exist four degenerate local states at the triangle .
the screening completes at @xmath103 , and it may be regarded as the su(4 ) kondo effect .
we see that the spin susceptibility for @xmath104 shows the curie behavior @xmath105 with the coefficient corresponding to @xmath85 .
therefore the degenerate states have the spin @xmath85 degrees of freedom , and thus the remaining degrees of freedoms of the four fold degeneracy should have an orbital origin in the triangle . as a function of @xmath106 for @xmath38 and @xmath39 . here , @xmath98 is the spin susceptibility , and @xmath90 is the half - width of the conduction band .
left ( a ) : results at @xmath88 ( @xmath107 ) , where the triangle has the @xmath0 moment .
right ( b ) : results at @xmath99 ( @xmath108 ) , where @xmath109 shows the sharp dip . ] as a function of @xmath106 for @xmath38 and @xmath39 . here
, @xmath98 is the spin susceptibility , and @xmath90 is the half - width of the conduction band
. left ( a ) : results at @xmath88 ( @xmath107 ) , where the triangle has the @xmath0 moment .
right ( b ) : results at @xmath99 ( @xmath108 ) , where @xmath109 shows the sharp dip . ]
we have studied the ground state properties of a triangular triple quantum dot connected to two noninteracting leads , and have found that the system shows the su(4 ) and the @xmath0 two - stage kondo effects , as well as the usual su(2 ) one , depending on the gate voltage @xmath10 .
we would like to thank j. bauer and v. meden for valuable discussions .
this work was supported by jsps grant - in - aid for scientific research ( c ) .
numerical computation was partly carried out in yukawa institute computer facility .
9 vidan a , westervelt r m , stopa m , hanson m and gossard a c 2004 _ appl .
lett . _ * 85 * 3602 | we study the conductance through a triangular triple quantum dot , which are connected to two noninteracting leads , using the numerical renormalization group ( nrg ) .
it is found that the system shows a variety of kondo effects depending on the filling of the triangle .
the su(4 ) kondo effect occurs at half - filling , and a sharp conductance dip due to a phase lapse appears in the gate - voltage dependence .
furthermore , when four electrons occupy the three sites on average , a local @xmath0 moment , which is caused by the nagaoka mechanism , is induced along the triangle .
the temperature dependence of the entropy and spin susceptibility of the triangle shows that this moment is screened by the conduction electrons via two separate stages at different temperatures . the two - terminal and
four - terminal conductances show a clear difference at the gate voltages , where the su(4 ) or the @xmath0 kondo effects occurring . |
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.